diff options
Diffstat (limited to 'pixman')
| -rw-r--r-- | pixman/Makefile.am | 264 | ||||
| -rw-r--r-- | pixman/configure.ac | 1642 | ||||
| -rw-r--r-- | pixman/pixman/config.h | 227 | ||||
| -rw-r--r-- | pixman/pixman/makefile | 76 | ||||
| -rw-r--r-- | pixman/pixman/pixman-arm-common.h | 728 | ||||
| -rw-r--r-- | pixman/pixman/pixman-arm-neon-asm.S | 4786 | ||||
| -rw-r--r-- | pixman/pixman/pixman-arm-neon.c | 882 | ||||
| -rw-r--r-- | pixman/pixman/pixman-compiler.h | 18 | ||||
| -rw-r--r-- | pixman/pixman/pixman-fast-path.c | 4456 | ||||
| -rw-r--r-- | pixman/pixman/pixman-fast-path.h | 1180 | ||||
| -rw-r--r-- | pixman/pixman/pixman-mmx.c | 24 | ||||
| -rw-r--r-- | pixman/pixman/pixman-sse2.c | 12848 | ||||
| -rw-r--r-- | pixman/pixman/pixman-trap.c | 4 | ||||
| -rw-r--r-- | pixman/pixman/pixman-version.h | 50 | ||||
| -rw-r--r-- | pixman/test/Makefile.am | 88 | ||||
| -rw-r--r-- | pixman/test/scaling-test.c | 736 |
16 files changed, 14195 insertions, 13814 deletions
diff --git a/pixman/Makefile.am b/pixman/Makefile.am index 062c58a89..b12b212fd 100644 --- a/pixman/Makefile.am +++ b/pixman/Makefile.am @@ -1,132 +1,132 @@ -SUBDIRS = pixman test demos - -pkgconfigdir=$(libdir)/pkgconfig -pkgconfig_DATA=pixman-1.pc - -$(pkgconfig_DATA): pixman-1.pc.in - -snapshot: - distdir="$(distdir)-`date '+%Y%m%d'`"; \ - test -d "$(srcdir)/.git" && distdir=$$distdir-`cd "$(srcdir)" && git rev-parse HEAD | cut -c 1-6`; \ - $(MAKE) $(AM_MAKEFLAGS) distdir="$$distdir" dist - -GPGKEY=6FF7C1A8 -USERNAME=$$USER -RELEASE_OR_SNAPSHOT = $$(if test "x$(CAIRO_VERSION_MINOR)" = "x$$(echo "$(CAIRO_VERSION_MINOR)/2*2" | bc)" ; then echo release; else echo snapshot; fi) -RELEASE_CAIRO_HOST = $(USERNAME)@cairographics.org -RELEASE_CAIRO_DIR = /srv/cairo.freedesktop.org/www/releases -RELEASE_CAIRO_URL = http://cairographics.org/releases -RELEASE_XORG_URL = http://xorg.freedesktop.org/archive/individual/lib -RELEASE_XORG_HOST = $(USERNAME)@xorg.freedesktop.org -RELEASE_XORG_DIR = /srv/xorg.freedesktop.org/archive/individual/lib -RELEASE_ANNOUNCE_LIST = cairo-announce@cairographics.org, xorg-announce@lists.freedesktop.org - -tar_gz = $(PACKAGE)-$(VERSION).tar.gz -tar_bz2 = $(PACKAGE)-$(VERSION).tar.bz2 - -sha1_tgz = $(tar_gz).sha1 -md5_tgz = $(tar_gz).md5 - -sha1_tbz2 = $(tar_bz2).sha1 -md5_tbz2 = $(tar_bz2).md5 - -gpg_file = $(sha1_tgz).asc - -$(sha1_tgz): $(tar_gz) - sha1sum $^ > $@ - -$(md5_tgz): $(tar_gz) - md5sum $^ > $@ - -$(sha1_tbz2): $(tar_bz2) - sha1sum $^ > $@ - -$(md5_tbz2): $(tar_bz2) - md5sum $^ > $@ - -$(gpg_file): $(sha1_tgz) - @echo "Please enter your GPG password to sign the checksum." - gpg --armor --sign $^ - -HASHFILES = $(sha1_tgz) $(sha1_tbz2) $(md5_tgz) $(md5_tbz2) - -release-verify-newer: - @echo -n "Checking that no $(VERSION) release already exists at $(RELEASE_XORG_HOST)..." - @ssh $(RELEASE_XORG_HOST) test ! -e $(RELEASE_XORG_DIR)/$(tar_gz) \ - || (echo "Ouch." && echo "Found: $(RELEASE_XORG_HOST):$(RELEASE_XORG_DIR)/$(tar_gz)" \ - && echo "Refusing to try to generate a new release of the same name." \ - && false) - @ssh $(RELEASE_CAIRO_HOST) test ! -e $(RELEASE_CAIRO_DIR)/$(tar_gz) \ - || (echo "Ouch." && echo "Found: $(RELEASE_CAIRO_HOST):$(RELEASE_CAIRO_DIR)/$(tar_gz)" \ - && echo "Refusing to try to generate a new release of the same name." \ - && false) - @echo "Good." - -release-remove-old: - $(RM) $(tar_gz) $(tar_bz2) $(HASHFILES) $(gpg_file) - -ensure-prev: - @if [[ "$(PREV)" == "" ]]; then \ - echo "" && \ - echo "You must set the PREV variable on the make command line to" && \ - echo "the last version." && \ - echo "" && \ - echo "For example:" && \ - echo " make PREV=0.7.3" && \ - echo "" && \ - false; \ - fi - -release-check: ensure-prev release-verify-newer release-remove-old distcheck - -release-tag: - git tag -u $(GPGKEY) -m "$(PACKAGE) $(VERSION) release" $(PACKAGE)-$(VERSION) - -release-upload: release-check $(tar_gz) $(tar_bz2) $(sha1_tgz) $(sha1_tbz2) $(md5_tgz) $(gpg_file) - mkdir -p releases - scp $(tar_gz) $(sha1_tgz) $(gpg_file) $(RELEASE_CAIRO_HOST):$(RELEASE_CAIRO_DIR) - scp $(tar_gz) $(tar_bz2) $(RELEASE_XORG_HOST):$(RELEASE_XORG_DIR) - ssh $(RELEASE_CAIRO_HOST) "rm -f $(RELEASE_CAIRO_DIR)/LATEST-$(PACKAGE)-[0-9]* && ln -s $(tar_gz) $(RELEASE_CAIRO_DIR)/LATEST-$(PACKAGE)-$(VERSION)" - -release-publish-message: $(HASHFILES) ensure-prev - @echo "Please follow the instructions in RELEASING to push stuff out and" - @echo "send out the announcement mails. Here is the excerpt you need:" - @echo "" - @echo "Lists: $(RELEASE_ANNOUNCE_LIST)" - @echo "Subject: [ANNOUNCE] $(PACKAGE) release $(VERSION) now available" - @echo "============================== CUT HERE ==============================" - @echo "A new $(PACKAGE) release $(VERSION) is now available" - @echo "" - @echo "tar.gz:" - @echo " $(RELEASE_CAIRO_URL)/$(tar_gz)" - @echo " $(RELEASE_XORG_URL)/$(tar_gz)" - @echo "" - @echo "tar.bz2:" - @echo " $(RELEASE_XORG_URL)/$(tar_bz2)" - @echo "" - @echo "Hashes:" - @echo -n " MD5: " - @cat $(md5_tgz) - @echo -n " MD5: " - @cat $(md5_tbz2) - @echo -n " SHA1: " - @cat $(sha1_tgz) - @echo -n " SHA1: " - @cat $(sha1_tbz2) - @echo "" - @echo "GPG signature:" - @echo " $(RELEASE_CAIRO_URL)/$(gpg_file)" - @echo " (signed by `git config --get user.name` <`git config --get user.email`>)" - @echo "" - @echo "Git:" - @echo " git://git.freedesktop.org/git/pixman" - @echo " tag: $(PACKAGE)-$(VERSION)" - @echo "" - @echo "Log:" - @git log --no-merges "$(PACKAGE)-$(PREV)".."$(PACKAGE)-$(VERSION)" | git shortlog | awk '{ printf "\t"; print ; }' | cut -b1-80 - @echo "============================== CUT HERE ==============================" - @echo "" - -release-publish: release-upload release-tag release-publish-message - -.PHONY: release-upload release-publish release-publish-message release-tag +SUBDIRS = pixman test demos
+
+pkgconfigdir=$(libdir)/pkgconfig
+pkgconfig_DATA=pixman-1.pc
+
+$(pkgconfig_DATA): pixman-1.pc.in
+
+snapshot:
+ distdir="$(distdir)-`date '+%Y%m%d'`"; \
+ test -d "$(srcdir)/.git" && distdir=$$distdir-`cd "$(srcdir)" && git rev-parse HEAD | cut -c 1-6`; \
+ $(MAKE) $(AM_MAKEFLAGS) distdir="$$distdir" dist
+
+GPGKEY=6FF7C1A8
+USERNAME=$$USER
+RELEASE_OR_SNAPSHOT = $$(if test "x$(CAIRO_VERSION_MINOR)" = "x$$(echo "$(CAIRO_VERSION_MINOR)/2*2" | bc)" ; then echo release; else echo snapshot; fi)
+RELEASE_CAIRO_HOST = $(USERNAME)@cairographics.org
+RELEASE_CAIRO_DIR = /srv/cairo.freedesktop.org/www/releases
+RELEASE_CAIRO_URL = http://cairographics.org/releases
+RELEASE_XORG_URL = http://xorg.freedesktop.org/archive/individual/lib
+RELEASE_XORG_HOST = $(USERNAME)@xorg.freedesktop.org
+RELEASE_XORG_DIR = /srv/xorg.freedesktop.org/archive/individual/lib
+RELEASE_ANNOUNCE_LIST = cairo-announce@cairographics.org, xorg-announce@lists.freedesktop.org
+
+tar_gz = $(PACKAGE)-$(VERSION).tar.gz
+tar_bz2 = $(PACKAGE)-$(VERSION).tar.bz2
+
+sha1_tgz = $(tar_gz).sha1
+md5_tgz = $(tar_gz).md5
+
+sha1_tbz2 = $(tar_bz2).sha1
+md5_tbz2 = $(tar_bz2).md5
+
+gpg_file = $(sha1_tgz).asc
+
+$(sha1_tgz): $(tar_gz)
+ sha1sum $^ > $@
+
+$(md5_tgz): $(tar_gz)
+ md5sum $^ > $@
+
+$(sha1_tbz2): $(tar_bz2)
+ sha1sum $^ > $@
+
+$(md5_tbz2): $(tar_bz2)
+ md5sum $^ > $@
+
+$(gpg_file): $(sha1_tgz)
+ @echo "Please enter your GPG password to sign the checksum."
+ gpg --armor --sign $^
+
+HASHFILES = $(sha1_tgz) $(sha1_tbz2) $(md5_tgz) $(md5_tbz2)
+
+release-verify-newer:
+ @echo -n "Checking that no $(VERSION) release already exists at $(RELEASE_XORG_HOST)..."
+ @ssh $(RELEASE_XORG_HOST) test ! -e $(RELEASE_XORG_DIR)/$(tar_gz) \
+ || (echo "Ouch." && echo "Found: $(RELEASE_XORG_HOST):$(RELEASE_XORG_DIR)/$(tar_gz)" \
+ && echo "Refusing to try to generate a new release of the same name." \
+ && false)
+ @ssh $(RELEASE_CAIRO_HOST) test ! -e $(RELEASE_CAIRO_DIR)/$(tar_gz) \
+ || (echo "Ouch." && echo "Found: $(RELEASE_CAIRO_HOST):$(RELEASE_CAIRO_DIR)/$(tar_gz)" \
+ && echo "Refusing to try to generate a new release of the same name." \
+ && false)
+ @echo "Good."
+
+release-remove-old:
+ $(RM) $(tar_gz) $(tar_bz2) $(HASHFILES) $(gpg_file)
+
+ensure-prev:
+ @if [[ "$(PREV)" == "" ]]; then \
+ echo "" && \
+ echo "You must set the PREV variable on the make command line to" && \
+ echo "the last version." && \
+ echo "" && \
+ echo "For example:" && \
+ echo " make PREV=0.7.3" && \
+ echo "" && \
+ false; \
+ fi
+
+release-check: ensure-prev release-verify-newer release-remove-old distcheck
+
+release-tag:
+ git tag -u $(GPGKEY) -m "$(PACKAGE) $(VERSION) release" $(PACKAGE)-$(VERSION)
+
+release-upload: release-check $(tar_gz) $(tar_bz2) $(sha1_tgz) $(sha1_tbz2) $(md5_tgz) $(gpg_file)
+ mkdir -p releases
+ scp $(tar_gz) $(sha1_tgz) $(gpg_file) $(RELEASE_CAIRO_HOST):$(RELEASE_CAIRO_DIR)
+ scp $(tar_gz) $(tar_bz2) $(RELEASE_XORG_HOST):$(RELEASE_XORG_DIR)
+ ssh $(RELEASE_CAIRO_HOST) "rm -f $(RELEASE_CAIRO_DIR)/LATEST-$(PACKAGE)-[0-9]* && ln -s $(tar_gz) $(RELEASE_CAIRO_DIR)/LATEST-$(PACKAGE)-$(VERSION)"
+
+release-publish-message: $(HASHFILES) ensure-prev
+ @echo "Please follow the instructions in RELEASING to push stuff out and"
+ @echo "send out the announcement mails. Here is the excerpt you need:"
+ @echo ""
+ @echo "Lists: $(RELEASE_ANNOUNCE_LIST)"
+ @echo "Subject: [ANNOUNCE] $(PACKAGE) release $(VERSION) now available"
+ @echo "============================== CUT HERE =============================="
+ @echo "A new $(PACKAGE) release $(VERSION) is now available"
+ @echo ""
+ @echo "tar.gz:"
+ @echo " $(RELEASE_CAIRO_URL)/$(tar_gz)"
+ @echo " $(RELEASE_XORG_URL)/$(tar_gz)"
+ @echo ""
+ @echo "tar.bz2:"
+ @echo " $(RELEASE_XORG_URL)/$(tar_bz2)"
+ @echo ""
+ @echo "Hashes:"
+ @echo -n " MD5: "
+ @cat $(md5_tgz)
+ @echo -n " MD5: "
+ @cat $(md5_tbz2)
+ @echo -n " SHA1: "
+ @cat $(sha1_tgz)
+ @echo -n " SHA1: "
+ @cat $(sha1_tbz2)
+ @echo ""
+ @echo "GPG signature:"
+ @echo " $(RELEASE_CAIRO_URL)/$(gpg_file)"
+ @echo " (signed by `git config --get user.name` <`git config --get user.email`>)"
+ @echo ""
+ @echo "Git:"
+ @echo " git://git.freedesktop.org/git/pixman"
+ @echo " tag: $(PACKAGE)-$(VERSION)"
+ @echo ""
+ @echo "Log:"
+ @git log --no-merges "$(PACKAGE)-$(PREV)".."$(PACKAGE)-$(VERSION)" | git shortlog | awk '{ printf "\t"; print ; }' | cut -b1-80
+ @echo "============================== CUT HERE =============================="
+ @echo ""
+
+release-publish: release-upload release-tag release-publish-message
+
+.PHONY: release-upload release-publish release-publish-message release-tag
diff --git a/pixman/configure.ac b/pixman/configure.ac index 5242799bb..189cc0667 100644 --- a/pixman/configure.ac +++ b/pixman/configure.ac @@ -1,821 +1,821 @@ -dnl Copyright 2005 Red Hat, Inc. -dnl -dnl Permission to use, copy, modify, distribute, and sell this software and its -dnl documentation for any purpose is hereby granted without fee, provided that -dnl the above copyright notice appear in all copies and that both that -dnl copyright notice and this permission notice appear in supporting -dnl documentation, and that the name of Red Hat not be used in -dnl advertising or publicity pertaining to distribution of the software without -dnl specific, written prior permission. Red Hat makes no -dnl representations about the suitability of this software for any purpose. It -dnl is provided "as is" without express or implied warranty. -dnl -dnl RED HAT DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, -dnl INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO -dnl EVENT SHALL RED HAT BE LIABLE FOR ANY SPECIAL, INDIRECT OR -dnl CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, -dnl DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER -dnl TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR -dnl PERFORMANCE OF THIS SOFTWARE. -dnl -dnl Process this file with autoconf to create configure. - -AC_PREREQ([2.57]) - -# Pixman versioning scheme -# -# - The version in git has an odd MICRO version number -# -# - Released versions both development and stable have an even MICRO -# version number -# -# - Released development versions have an odd MINOR number -# -# - Released stable versions have an even MINOR number -# -# - Versions that break ABI must have a new MAJOR number -# -# - If you break the ABI, then at least this must be done: -# -# - increment MAJOR -# -# - In the first development release where you break ABI, find -# all instances of "pixman-n" and change them to pixman-(n+1) -# -# This needs to be done at least in -# configure.ac -# all Makefile.am's -# pixman-n.pc.in -# -# This ensures that binary incompatible versions can be installed -# in parallel. See http://www106.pair.com/rhp/parallel.html for -# more information -# - -m4_define([pixman_major], 0) -m4_define([pixman_minor], 21) -m4_define([pixman_micro], 5) - -m4_define([pixman_version],[pixman_major.pixman_minor.pixman_micro]) - -AC_INIT(pixman, pixman_version, [pixman@lists.freedesktop.org], pixman) -AM_INIT_AUTOMAKE([foreign dist-bzip2]) - -# Suppress verbose compile lines -m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])]) - -AM_CONFIG_HEADER(config.h) - -AC_CANONICAL_HOST - -test_CFLAGS=${CFLAGS+set} # We may override autoconf default CFLAGS. - -AC_PROG_CC -AM_PROG_AS -AC_PROG_LIBTOOL -AC_CHECK_FUNCS([getisax]) -AC_C_BIGENDIAN -AC_C_INLINE - -dnl PIXMAN_LINK_WITH_ENV(env-setup, program, true-action, false-action) -dnl -dnl Compiles and links the given program in the environment setup by env-setup -dnl and executes true-action on success and false-action on failure. -AC_DEFUN([PIXMAN_LINK_WITH_ENV],[dnl - save_CFLAGS="$CFLAGS" - save_LDFLAGS="$LDFLAGS" - save_LIBS="$LIBS" - CFLAGS="" - LDFLAGS="" - LIBS="" - $1 - AC_LINK_IFELSE( - [$2], - [pixman_cc_stderr=`test -f conftest.err && cat conftest.err` - pixman_cc_flag=yes], - [pixman_cc_stderr=`test -f conftest.err && cat conftest.err` - pixman_cc_flag=no]) - - if test "x$pixman_cc_stderr" != "x"; then - pixman_cc_flag=no - fi - - if test "x$pixman_cc_flag" = "xyes"; then - ifelse([$3], , :, [$3]) - else - ifelse([$4], , :, [$4]) - fi - CFLAGS="$save_CFLAGS" - LDFLAGS="$save_LDFLAGS" - LIBS="$save_LIBS" -]) - -dnl Find a -Werror for catching warnings. -WERROR= -for w in -Werror -errwarn; do - if test "z$WERROR" = "z"; then - AC_MSG_CHECKING([whether the compiler supports $w]) - PIXMAN_LINK_WITH_ENV( - [CFLAGS=$w], - [int main(int c, char **v) { (void)c; (void)v; return 0; }], - [WERROR=$w; yesno=yes], [yesno=no]) - AC_MSG_RESULT($_yesno) - fi -done - -dnl PIXMAN_CHECK_CFLAG(flag, [program]) -dnl Adds flag to CFLAGS if the given program links without warnings or errors. -AC_DEFUN([PIXMAN_CHECK_CFLAG], [dnl - AC_MSG_CHECKING([whether the compiler supports $1]) - PIXMAN_LINK_WITH_ENV( - [CFLAGS="$WERROR $1"], - [$2 - int main(int c, char **v) { (void)c; (void)v; return 0; } - ], - [_yesno=yes], - [_yesno=no]) - if test "x$_yesno" = xyes; then - CFLAGS="$CFLAGS $1" - fi - AC_MSG_RESULT($_yesno) -]) - -AC_CHECK_SIZEOF(long) - -# Checks for Sun Studio compilers -AC_CHECK_DECL([__SUNPRO_C], [SUNCC="yes"], [SUNCC="no"]) -AC_CHECK_DECL([__amd64], [AMD64_ABI="yes"], [AMD64_ABI="no"]) - -# Default CFLAGS to -O -g rather than just the -g from AC_PROG_CC -# if we're using Sun Studio and neither the user nor a config.site -# has set CFLAGS. -if test $SUNCC = yes && \ - test "$test_CFLAGS" == "" && \ - test "$CFLAGS" = "-g" -then - CFLAGS="-O -g" -fi - -# -# We ignore pixman_major in the version here because the major version should -# always be encoded in the actual library name. Ie., the soname is: -# -# pixman-$(pixman_major).0.minor.micro -# -m4_define([lt_current], [pixman_minor]) -m4_define([lt_revision], [pixman_micro]) -m4_define([lt_age], [pixman_minor]) - -LT_VERSION_INFO="lt_current:lt_revision:lt_age" - -PIXMAN_VERSION_MAJOR=pixman_major() -AC_SUBST(PIXMAN_VERSION_MAJOR) -PIXMAN_VERSION_MINOR=pixman_minor() -AC_SUBST(PIXMAN_VERSION_MINOR) -PIXMAN_VERSION_MICRO=pixman_micro() -AC_SUBST(PIXMAN_VERSION_MICRO) - -AC_SUBST(LT_VERSION_INFO) - -# Check for dependencies - -PIXMAN_CHECK_CFLAG([-Wall]) -PIXMAN_CHECK_CFLAG([-fno-strict-aliasing]) - -AC_PATH_PROG(PERL, perl, no) -if test "x$PERL" = xno; then - AC_MSG_ERROR([Perl is required to build pixman.]) -fi -AC_SUBST(PERL) - -dnl ========================================================================= -dnl OpenMP for the test suite? -dnl - -# Check for OpenMP support (only supported by autoconf >=2.62) -OPENMP_CFLAGS= -m4_ifdef([AC_OPENMP], [AC_OPENMP]) - -m4_define([openmp_test_program],[dnl -#include <stdio.h> - -extern unsigned int lcg_seed; -#pragma omp threadprivate(lcg_seed) -unsigned int lcg_seed; - -unsigned function(unsigned a, unsigned b) -{ - lcg_seed ^= b; - return ((a + b) ^ a ) + lcg_seed; -} - -int main(int argc, char **argv) -{ - int i; - int n1 = 0, n2 = argc; - unsigned checksum = 0; - int verbose = argv != NULL; - unsigned (*test_function)(unsigned, unsigned); - test_function = function; - #pragma omp parallel for reduction(+:checksum) default(none) \ - shared(n1, n2, test_function, verbose) - for (i = n1; i < n2; i++) - { - unsigned crc = test_function (i, 0); - if (verbose) - printf ("%d: %08X\n", i, crc); - checksum += crc; - } - printf("%u\n", checksum); - return 0; -} -]) - -PIXMAN_LINK_WITH_ENV( - [CFLAGS="$OPENMP_CFLAGS" LDFLAGS="$OPENMP_CFLAGS"], - [openmp_test_program], - [have_openmp=yes], - [have_openmp=no]) -if test "x$have_openmp" = "xyes"; then - AC_DEFINE(USE_OPENMP, 1, [use OpenMP in the test suite]) -else - OPENMP_CFLAGS="" -fi -AC_SUBST(OPENMP_CFLAGS) - -dnl ========================================================================= -dnl -fvisibility stuff - -PIXMAN_CHECK_CFLAG([-fvisibility=hidden], [dnl -#if defined(__GNUC__) && (__GNUC__ >= 4) -#ifdef _WIN32 -#error Have -fvisibility but it is ignored and generates a warning -#endif -#else -error Need GCC 4.0 for visibility -#endif -]) - -PIXMAN_CHECK_CFLAG([-xldscope=hidden], [dnl -#if defined(__SUNPRO_C) && (__SUNPRO_C >= 0x550) -#else -error Need Sun Studio 8 for visibility -#endif -]) - -dnl =========================================================================== -dnl Check for MMX - -if test "x$MMX_CFLAGS" = "x" ; then - if test "x$SUNCC" = "xyes"; then - # Sun Studio doesn't have an -xarch=mmx flag, so we have to use sse - # but if we're building 64-bit, mmx & sse support is on by default and - # -xarch=sse throws an error instead - if test "$AMD64_ABI" = "no" ; then - MMX_CFLAGS="-xarch=sse" - fi - else - MMX_CFLAGS="-mmmx -Winline" - fi -fi - -have_mmx_intrinsics=no -AC_MSG_CHECKING(whether to use MMX intrinsics) -xserver_save_CFLAGS=$CFLAGS -CFLAGS="$MMX_CFLAGS $CFLAGS" -AC_COMPILE_IFELSE([ -#if defined(__GNUC__) && (__GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 4)) -error "Need GCC >= 3.4 for MMX intrinsics" -#endif -#include <mmintrin.h> -int main () { - __m64 v = _mm_cvtsi32_si64 (1); - return _mm_cvtsi64_si32 (v); -}], have_mmx_intrinsics=yes) -CFLAGS=$xserver_save_CFLAGS - -AC_ARG_ENABLE(mmx, - [AC_HELP_STRING([--disable-mmx], - [disable MMX fast paths])], - [enable_mmx=$enableval], [enable_mmx=auto]) - -if test $enable_mmx = no ; then - have_mmx_intrinsics=disabled -fi - -if test $have_mmx_intrinsics = yes ; then - AC_DEFINE(USE_MMX, 1, [use MMX compiler intrinsics]) -else - MMX_CFLAGS= -fi - -AC_MSG_RESULT($have_mmx_intrinsics) -if test $enable_mmx = yes && test $have_mmx_intrinsics = no ; then - AC_MSG_ERROR([MMX intrinsics not detected]) -fi - -AM_CONDITIONAL(USE_MMX, test $have_mmx_intrinsics = yes) - -dnl =========================================================================== -dnl Check for SSE2 - -if test "x$SSE2_CFLAGS" = "x" ; then - if test "x$SUNCC" = "xyes"; then - # SSE2 is enabled by default in the Sun Studio 64-bit environment - if test "$AMD64_ABI" = "no" ; then - SSE2_CFLAGS="-xarch=sse2" - fi - else - SSE2_CFLAGS="-mmmx -msse2 -Winline" - fi -fi - -have_sse2_intrinsics=no -AC_MSG_CHECKING(whether to use SSE2 intrinsics) -xserver_save_CFLAGS=$CFLAGS -CFLAGS="$SSE2_CFLAGS $CFLAGS" - -AC_COMPILE_IFELSE([ -#if defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 2)) -# if !defined(__amd64__) && !defined(__x86_64__) -# error "Need GCC >= 4.2 for SSE2 intrinsics on x86" -# endif -#endif -#include <mmintrin.h> -#include <xmmintrin.h> -#include <emmintrin.h> -int main () { - __m128i a = _mm_set1_epi32 (0), b = _mm_set1_epi32 (0), c; - c = _mm_xor_si128 (a, b); - return 0; -}], have_sse2_intrinsics=yes) -CFLAGS=$xserver_save_CFLAGS - -AC_ARG_ENABLE(sse2, - [AC_HELP_STRING([--disable-sse2], - [disable SSE2 fast paths])], - [enable_sse2=$enableval], [enable_sse2=auto]) - -if test $enable_sse2 = no ; then - have_sse2_intrinsics=disabled -fi - -if test $have_sse2_intrinsics = yes ; then - AC_DEFINE(USE_SSE2, 1, [use SSE2 compiler intrinsics]) -fi - -AC_MSG_RESULT($have_sse2_intrinsics) -if test $enable_sse2 = yes && test $have_sse2_intrinsics = no ; then - AC_MSG_ERROR([SSE2 intrinsics not detected]) -fi - -AM_CONDITIONAL(USE_SSE2, test $have_sse2_intrinsics = yes) - -dnl =========================================================================== -dnl Other special flags needed when building code using MMX or SSE instructions -case $host_os in - solaris*) - # When building 32-bit binaries, apply a mapfile to ensure that the - # binaries aren't flagged as only able to run on MMX+SSE capable CPUs - # since they check at runtime before using those instructions. - # Not all linkers grok the mapfile format so we check for that first. - if test "$AMD64_ABI" = "no" ; then - use_hwcap_mapfile=no - AC_MSG_CHECKING(whether to use a hardware capability map file) - hwcap_save_LDFLAGS="$LDFLAGS" - HWCAP_LDFLAGS='-Wl,-M,$(srcdir)/solaris-hwcap.mapfile' - LDFLAGS="$LDFLAGS -Wl,-M,pixman/solaris-hwcap.mapfile" - AC_LINK_IFELSE([int main() { return 0; }], - use_hwcap_mapfile=yes, - HWCAP_LDFLAGS="") - LDFLAGS="$hwcap_save_LDFLAGS" - AC_MSG_RESULT($use_hwcap_mapfile) - fi - if test "x$MMX_LDFLAGS" = "x" ; then - MMX_LDFLAGS="$HWCAP_LDFLAGS" - fi - if test "x$SSE2_LDFLAGS" = "x" ; then - SSE2_LDFLAGS="$HWCAP_LDFLAGS" - fi - ;; -esac - -AC_SUBST(MMX_CFLAGS) -AC_SUBST(MMX_LDFLAGS) -AC_SUBST(SSE2_CFLAGS) -AC_SUBST(SSE2_LDFLAGS) - -dnl =========================================================================== -dnl Check for VMX/Altivec -if test -n "`$CC -v 2>&1 | grep version | grep Apple`"; then - VMX_CFLAGS="-faltivec" -else - VMX_CFLAGS="-maltivec -mabi=altivec" -fi - -have_vmx_intrinsics=no -AC_MSG_CHECKING(whether to use VMX/Altivec intrinsics) -xserver_save_CFLAGS=$CFLAGS -CFLAGS="$VMX_CFLAGS $CFLAGS" -AC_COMPILE_IFELSE([ -#if defined(__GNUC__) && (__GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 4)) -error "Need GCC >= 3.4 for sane altivec support" -#endif -#include <altivec.h> -int main () { - vector unsigned int v = vec_splat_u32 (1); - v = vec_sub (v, v); - return 0; -}], have_vmx_intrinsics=yes) -CFLAGS=$xserver_save_CFLAGS - -AC_ARG_ENABLE(vmx, - [AC_HELP_STRING([--disable-vmx], - [disable VMX fast paths])], - [enable_vmx=$enableval], [enable_vmx=auto]) - -if test $enable_vmx = no ; then - have_vmx_intrinsics=disabled -fi - -if test $have_vmx_intrinsics = yes ; then - AC_DEFINE(USE_VMX, 1, [use VMX compiler intrinsics]) -else - VMX_CFLAGS= -fi - -AC_MSG_RESULT($have_vmx_intrinsics) -if test $enable_vmx = yes && test $have_vmx_intrinsics = no ; then - AC_MSG_ERROR([VMX intrinsics not detected]) -fi - -AC_SUBST(VMX_CFLAGS) - -AM_CONDITIONAL(USE_VMX, test $have_vmx_intrinsics = yes) - -dnl ========================================================================== -dnl Check if assembler is gas compatible and supports ARM SIMD instructions -have_arm_simd=no -AC_MSG_CHECKING(whether to use ARM SIMD assembler) -xserver_save_CFLAGS=$CFLAGS -CFLAGS="-x assembler-with-cpp $CFLAGS" -AC_COMPILE_IFELSE([[ -.text -.arch armv6 -.object_arch armv4 -.arm -.altmacro -#ifndef __ARM_EABI__ -#error EABI is required (to be sure that calling conventions are compatible) -#endif -pld [r0] -uqadd8 r0, r0, r0]], have_arm_simd=yes) -CFLAGS=$xserver_save_CFLAGS - -AC_ARG_ENABLE(arm-simd, - [AC_HELP_STRING([--disable-arm-simd], - [disable ARM SIMD fast paths])], - [enable_arm_simd=$enableval], [enable_arm_simd=auto]) - -if test $enable_arm_simd = no ; then - have_arm_simd=disabled -fi - -if test $have_arm_simd = yes ; then - AC_DEFINE(USE_ARM_SIMD, 1, [use ARM SIMD assembly optimizations]) -fi - -AM_CONDITIONAL(USE_ARM_SIMD, test $have_arm_simd = yes) - -AC_MSG_RESULT($have_arm_simd) -if test $enable_arm_simd = yes && test $have_arm_simd = no ; then - AC_MSG_ERROR([ARM SIMD intrinsics not detected]) -fi - -dnl ========================================================================== -dnl Check if assembler is gas compatible and supports NEON instructions -have_arm_neon=no -AC_MSG_CHECKING(whether to use ARM NEON assembler) -xserver_save_CFLAGS=$CFLAGS -CFLAGS="-x assembler-with-cpp $CFLAGS" -AC_COMPILE_IFELSE([[ -.text -.fpu neon -.arch armv7a -.object_arch armv4 -.eabi_attribute 10, 0 -.arm -.altmacro -#ifndef __ARM_EABI__ -#error EABI is required (to be sure that calling conventions are compatible) -#endif -pld [r0] -vmovn.u16 d0, q0]], have_arm_neon=yes) -CFLAGS=$xserver_save_CFLAGS - -AC_ARG_ENABLE(arm-neon, - [AC_HELP_STRING([--disable-arm-neon], - [disable ARM NEON fast paths])], - [enable_arm_neon=$enableval], [enable_arm_neon=auto]) - -if test $enable_arm_neon = no ; then - have_arm_neon=disabled -fi - -if test $have_arm_neon = yes ; then - AC_DEFINE(USE_ARM_NEON, 1, [use ARM NEON assembly optimizations]) -fi - -AM_CONDITIONAL(USE_ARM_NEON, test $have_arm_neon = yes) - -AC_MSG_RESULT($have_arm_neon) -if test $enable_arm_neon = yes && test $have_arm_neon = no ; then - AC_MSG_ERROR([ARM NEON intrinsics not detected]) -fi - -dnl ========================================================================================= -dnl Check for GNU-style inline assembly support - -have_gcc_inline_asm=no -AC_MSG_CHECKING(whether to use GNU-style inline assembler) -AC_COMPILE_IFELSE([ -int main () { - /* Most modern architectures have a NOP instruction, so this is a fairly generic test. */ - asm volatile ( "\tnop\n" : : : "cc", "memory" ); - return 0; -}], have_gcc_inline_asm=yes) - -AC_ARG_ENABLE(gcc-inline-asm, - [AC_HELP_STRING([--disable-gcc-inline-asm], - [disable GNU-style inline assembler])], - [enable_gcc_inline_asm=$enableval], [enable_gcc_inline_asm=auto]) - -if test $enable_gcc_inline_asm = no ; then - have_gcc_inline_asm=disabled -fi - -if test $have_gcc_inline_asm = yes ; then - AC_DEFINE(USE_GCC_INLINE_ASM, 1, [use GNU-style inline assembler]) -fi - -AC_MSG_RESULT($have_gcc_inline_asm) -if test $enable_gcc_inline_asm = yes && test $have_gcc_inline_asm = no ; then - AC_MSG_ERROR([GNU-style inline assembler not detected]) -fi - -AM_CONDITIONAL(USE_GCC_INLINE_ASM, test $have_gcc_inline_asm = yes) - -dnl ============================================== -dnl Static test programs - -AC_ARG_ENABLE(static-testprogs, - [AC_HELP_STRING([--enable-static-testprogs], - [build test programs as static binaries [default=no]])], - [enable_static_testprogs=$enableval], [enable_static_testprogs=no]) - -TESTPROGS_EXTRA_LDFLAGS= -if test "x$enable_static_testprogs" = "xyes" ; then - TESTPROGS_EXTRA_LDFLAGS="-all-static" -fi -AC_SUBST(TESTPROGS_EXTRA_LDFLAGS) - -dnl ============================================== -dnl Timers - -AC_ARG_ENABLE(timers, - [AC_HELP_STRING([--enable-timers], - [enable TIMER_BEGIN and TIMER_END macros [default=no]])], - [enable_timers=$enableval], [enable_timers=no]) - -if test $enable_timers = yes ; then - AC_DEFINE(PIXMAN_TIMERS, 1, [enable TIMER_BEGIN/TIMER_END macros]) -fi -AC_SUBST(PIXMAN_TIMERS) - -dnl =================================== -dnl GTK+ - -AC_ARG_ENABLE(gtk, - [AC_HELP_STRING([--enable-gtk], - [enable tests using GTK+ [default=auto]])], - [enable_gtk=$enableval], [enable_gtk=auto]) - -PKG_PROG_PKG_CONFIG - -if test $enable_gtk = yes ; then - AC_CHECK_LIB([pixman-1], [pixman_version_string]) - PKG_CHECK_MODULES(GTK, [gtk+-2.0 pixman-1]) -fi - -if test $enable_gtk = auto ; then - AC_CHECK_LIB([pixman-1], [pixman_version_string], [enable_gtk=auto], [enable_gtk=no]) -fi - -if test $enable_gtk = auto ; then - PKG_CHECK_MODULES(GTK, [gtk+-2.0 pixman-1], [enable_gtk=yes], [enable_gtk=no]) -fi - -AM_CONDITIONAL(HAVE_GTK, [test "x$enable_gtk" = xyes]) - -AC_SUBST(GTK_CFLAGS) -AC_SUBST(GTK_LIBS) -AC_SUBST(DEP_CFLAGS) -AC_SUBST(DEP_LIBS) - -dnl ===================================== -dnl posix_memalign, sigaction, alarm, gettimeofday - -AC_CHECK_FUNC(posix_memalign, have_posix_memalign=yes, have_posix_memalign=no) -if test x$have_posix_memalign = xyes; then - AC_DEFINE(HAVE_POSIX_MEMALIGN, 1, [Whether we have posix_memalign()]) -fi - -AC_CHECK_FUNC(sigaction, have_sigaction=yes, have_sigaction=no) -if test x$have_sigaction = xyes; then - AC_DEFINE(HAVE_SIGACTION, 1, [Whether we have sigaction()]) -fi - -AC_CHECK_FUNC(alarm, have_alarm=yes, have_alarm=no) -if test x$have_alarm = xyes; then - AC_DEFINE(HAVE_ALARM, 1, [Whether we have alarm()]) -fi - -AC_CHECK_HEADER([sys/mman.h], - [AC_DEFINE(HAVE_SYS_MMAN_H, [1], [Define to 1 if we have <sys/mman.h>])]) - -AC_CHECK_FUNC(mprotect, have_mprotect=yes, have_mprotect=no) -if test x$have_mprotect = xyes; then - AC_DEFINE(HAVE_MPROTECT, 1, [Whether we have mprotect()]) -fi - -AC_CHECK_FUNC(getpagesize, have_getpagesize=yes, have_getpagesize=no) -if test x$have_getpagesize = xyes; then - AC_DEFINE(HAVE_GETPAGESIZE, 1, [Whether we have getpagesize()]) -fi - -AC_CHECK_HEADER([fenv.h], - [AC_DEFINE(HAVE_FENV_H, [1], [Define to 1 if we have <fenv.h>])]) - -AC_CHECK_LIB(m, feenableexcept, have_feenableexcept=yes, have_feenableexcept=no) -if test x$have_feenableexcept = xyes; then - AC_DEFINE(HAVE_FEENABLEEXCEPT, 1, [Whether we have feenableexcept()]) -fi - -AC_CHECK_FUNC(gettimeofday, have_gettimeofday=yes, have_gettimeofday=no) -AC_CHECK_HEADER(sys/time.h, have_sys_time_h=yes, have_sys_time_h=no) -if test x$have_gettimeofday = xyes && test x$have_sys_time_h = xyes; then - AC_DEFINE(HAVE_GETTIMEOFDAY, 1, [Whether we have gettimeofday()]) -fi - -dnl ===================================== -dnl Thread local storage - -support_for__thread=no - -AC_MSG_CHECKING(for __thread) -AC_LINK_IFELSE([ -#if defined(__MINGW32__) && !(__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 5)) -#error This MinGW version has broken __thread support -#endif -#ifdef __OpenBSD__ -#error OpenBSD has broken __thread support -#endif -static __thread int x ; -int main () { x = 123; return x; } -], support_for__thread=yes) - -if test $support_for__thread = yes; then - AC_DEFINE([TOOLCHAIN_SUPPORTS__THREAD],[],[Whether the tool chain supports __thread]) -fi - -AC_MSG_RESULT($support_for__thread) - -dnl -dnl posix tls -dnl - -m4_define([pthread_test_program],[dnl -#include <stdlib.h> -#include <pthread.h> - -static pthread_once_t once_control = PTHREAD_ONCE_INIT; -static pthread_key_t key; - -static void -make_key (void) -{ - pthread_key_create (&key, NULL); -} - -int -main () -{ - void *value = NULL; - - if (pthread_once (&once_control, make_key) != 0) - { - value = NULL; - } - else - { - value = pthread_getspecific (key); - if (!value) - { - value = malloc (100); - pthread_setspecific (key, value); - } - } - return 0; -} -]) - -AC_DEFUN([PIXMAN_CHECK_PTHREAD],[dnl - if test "z$support_for_pthread_setspecific" != "zyes"; then - PIXMAN_LINK_WITH_ENV( - [$1], [pthread_test_program], - [PTHREAD_CFLAGS="$CFLAGS" - PTHREAD_LIBS="$LIBS" - PTHREAD_LDFLAGS="$LDFLAGS" - support_for_pthread_setspecific=yes]) - fi -]) - -if test $support_for__thread = no; then - support_for_pthread_setspecific=no - - AC_MSG_CHECKING(for pthread_setspecific) - - PIXMAN_CHECK_PTHREAD([CFLAGS="-D_REENTRANT"; LIBS="-lpthread"]) - PIXMAN_CHECK_PTHREAD([CFLAGS="-pthread"; LDFLAGS="-pthread"]) - PIXMAN_CHECK_PTHREAD([CFLAGS="-D_REENTRANT"; LDFLAGS="-lroot"]) - - if test $support_for_pthread_setspecific = yes; then - CFLAGS="$CFLAGS $PTHREAD_CFLAGS" - AC_DEFINE([HAVE_PTHREAD_SETSPECIFIC], [], [Whether pthread_setspecific() is supported]) - fi - - AC_MSG_RESULT($support_for_pthread_setspecific); -fi - -AC_SUBST(TOOLCHAIN_SUPPORTS__THREAD) -AC_SUBST(HAVE_PTHREAD_SETSPECIFIC) -AC_SUBST(PTHREAD_LDFLAGS) -AC_SUBST(PTHREAD_LIBS) - -dnl ===================================== -dnl __attribute__((constructor)) - -support_for_attribute_constructor=no - -AC_MSG_CHECKING(for __attribute__((constructor))) -AC_LINK_IFELSE([ -#if defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 7)) -/* attribute 'constructor' is supported since gcc 2.7, but some compilers - * may only pretend to be gcc, so let's try to actually use it - */ -static int x = 1; -static void __attribute__((constructor)) constructor_function () { x = 0; } -int main (void) { return x; } -#else -#error not gcc or gcc version is older than 2.7 -#endif -], support_for_attribute_constructor=yes) - -if test x$support_for_attribute_constructor = xyes; then - AC_DEFINE([TOOLCHAIN_SUPPORTS_ATTRIBUTE_CONSTRUCTOR], - [],[Whether the tool chain supports __attribute__((constructor))]) -fi - -AC_MSG_RESULT($support_for_attribute_constructor) -AC_SUBST(TOOLCHAIN_SUPPORTS_ATTRIBUTE_CONSTRUCTOR) - -AC_OUTPUT([pixman-1.pc - pixman-1-uninstalled.pc - Makefile - pixman/Makefile - pixman/pixman-version.h - demos/Makefile - test/Makefile]) - -m4_if(m4_eval(pixman_minor % 2), [1], [ - echo - echo "@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@" - echo - echo " Thanks for testing this development snapshot of pixman. Please" - echo " report any problems you find, either by sending email to " - echo - echo " pixman@lists.freedesktop.org" - echo - echo " or by filing a bug at " - echo - echo " https://bugs.freedesktop.org/enter_bug.cgi?product=pixman " - echo - echo " If you are looking for a stable release of pixman, please note " - echo " that stable releases have _even_ minor version numbers. Ie., " - echo " pixman-0.]m4_eval(pixman_minor & ~1)[.x are stable releases, whereas pixman-$PIXMAN_VERSION_MAJOR.$PIXMAN_VERSION_MINOR.$PIXMAN_VERSION_MICRO is a " - echo " development snapshot that may contain bugs and experimental " - echo " features. " - echo - echo "@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@" - echo -]) +dnl Copyright 2005 Red Hat, Inc.
+dnl
+dnl Permission to use, copy, modify, distribute, and sell this software and its
+dnl documentation for any purpose is hereby granted without fee, provided that
+dnl the above copyright notice appear in all copies and that both that
+dnl copyright notice and this permission notice appear in supporting
+dnl documentation, and that the name of Red Hat not be used in
+dnl advertising or publicity pertaining to distribution of the software without
+dnl specific, written prior permission. Red Hat makes no
+dnl representations about the suitability of this software for any purpose. It
+dnl is provided "as is" without express or implied warranty.
+dnl
+dnl RED HAT DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE,
+dnl INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO
+dnl EVENT SHALL RED HAT BE LIABLE FOR ANY SPECIAL, INDIRECT OR
+dnl CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
+dnl DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
+dnl TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
+dnl PERFORMANCE OF THIS SOFTWARE.
+dnl
+dnl Process this file with autoconf to create configure.
+
+AC_PREREQ([2.57])
+
+# Pixman versioning scheme
+#
+# - The version in git has an odd MICRO version number
+#
+# - Released versions both development and stable have an even MICRO
+# version number
+#
+# - Released development versions have an odd MINOR number
+#
+# - Released stable versions have an even MINOR number
+#
+# - Versions that break ABI must have a new MAJOR number
+#
+# - If you break the ABI, then at least this must be done:
+#
+# - increment MAJOR
+#
+# - In the first development release where you break ABI, find
+# all instances of "pixman-n" and change them to pixman-(n+1)
+#
+# This needs to be done at least in
+# configure.ac
+# all Makefile.am's
+# pixman-n.pc.in
+#
+# This ensures that binary incompatible versions can be installed
+# in parallel. See http://www106.pair.com/rhp/parallel.html for
+# more information
+#
+
+m4_define([pixman_major], 0)
+m4_define([pixman_minor], 21)
+m4_define([pixman_micro], 5)
+
+m4_define([pixman_version],[pixman_major.pixman_minor.pixman_micro])
+
+AC_INIT(pixman, pixman_version, [pixman@lists.freedesktop.org], pixman)
+AM_INIT_AUTOMAKE([foreign dist-bzip2])
+
+# Suppress verbose compile lines
+m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
+
+AM_CONFIG_HEADER(config.h)
+
+AC_CANONICAL_HOST
+
+test_CFLAGS=${CFLAGS+set} # We may override autoconf default CFLAGS.
+
+AC_PROG_CC
+AM_PROG_AS
+AC_PROG_LIBTOOL
+AC_CHECK_FUNCS([getisax])
+AC_C_BIGENDIAN
+AC_C_INLINE
+
+dnl PIXMAN_LINK_WITH_ENV(env-setup, program, true-action, false-action)
+dnl
+dnl Compiles and links the given program in the environment setup by env-setup
+dnl and executes true-action on success and false-action on failure.
+AC_DEFUN([PIXMAN_LINK_WITH_ENV],[dnl
+ save_CFLAGS="$CFLAGS"
+ save_LDFLAGS="$LDFLAGS"
+ save_LIBS="$LIBS"
+ CFLAGS=""
+ LDFLAGS=""
+ LIBS=""
+ $1
+ AC_LINK_IFELSE(
+ [$2],
+ [pixman_cc_stderr=`test -f conftest.err && cat conftest.err`
+ pixman_cc_flag=yes],
+ [pixman_cc_stderr=`test -f conftest.err && cat conftest.err`
+ pixman_cc_flag=no])
+
+ if test "x$pixman_cc_stderr" != "x"; then
+ pixman_cc_flag=no
+ fi
+
+ if test "x$pixman_cc_flag" = "xyes"; then
+ ifelse([$3], , :, [$3])
+ else
+ ifelse([$4], , :, [$4])
+ fi
+ CFLAGS="$save_CFLAGS"
+ LDFLAGS="$save_LDFLAGS"
+ LIBS="$save_LIBS"
+])
+
+dnl Find a -Werror for catching warnings.
+WERROR=
+for w in -Werror -errwarn; do
+ if test "z$WERROR" = "z"; then
+ AC_MSG_CHECKING([whether the compiler supports $w])
+ PIXMAN_LINK_WITH_ENV(
+ [CFLAGS=$w],
+ [int main(int c, char **v) { (void)c; (void)v; return 0; }],
+ [WERROR=$w; yesno=yes], [yesno=no])
+ AC_MSG_RESULT($_yesno)
+ fi
+done
+
+dnl PIXMAN_CHECK_CFLAG(flag, [program])
+dnl Adds flag to CFLAGS if the given program links without warnings or errors.
+AC_DEFUN([PIXMAN_CHECK_CFLAG], [dnl
+ AC_MSG_CHECKING([whether the compiler supports $1])
+ PIXMAN_LINK_WITH_ENV(
+ [CFLAGS="$WERROR $1"],
+ [$2
+ int main(int c, char **v) { (void)c; (void)v; return 0; }
+ ],
+ [_yesno=yes],
+ [_yesno=no])
+ if test "x$_yesno" = xyes; then
+ CFLAGS="$CFLAGS $1"
+ fi
+ AC_MSG_RESULT($_yesno)
+])
+
+AC_CHECK_SIZEOF(long)
+
+# Checks for Sun Studio compilers
+AC_CHECK_DECL([__SUNPRO_C], [SUNCC="yes"], [SUNCC="no"])
+AC_CHECK_DECL([__amd64], [AMD64_ABI="yes"], [AMD64_ABI="no"])
+
+# Default CFLAGS to -O -g rather than just the -g from AC_PROG_CC
+# if we're using Sun Studio and neither the user nor a config.site
+# has set CFLAGS.
+if test $SUNCC = yes && \
+ test "$test_CFLAGS" == "" && \
+ test "$CFLAGS" = "-g"
+then
+ CFLAGS="-O -g"
+fi
+
+#
+# We ignore pixman_major in the version here because the major version should
+# always be encoded in the actual library name. Ie., the soname is:
+#
+# pixman-$(pixman_major).0.minor.micro
+#
+m4_define([lt_current], [pixman_minor])
+m4_define([lt_revision], [pixman_micro])
+m4_define([lt_age], [pixman_minor])
+
+LT_VERSION_INFO="lt_current:lt_revision:lt_age"
+
+PIXMAN_VERSION_MAJOR=pixman_major()
+AC_SUBST(PIXMAN_VERSION_MAJOR)
+PIXMAN_VERSION_MINOR=pixman_minor()
+AC_SUBST(PIXMAN_VERSION_MINOR)
+PIXMAN_VERSION_MICRO=pixman_micro()
+AC_SUBST(PIXMAN_VERSION_MICRO)
+
+AC_SUBST(LT_VERSION_INFO)
+
+# Check for dependencies
+
+PIXMAN_CHECK_CFLAG([-Wall])
+PIXMAN_CHECK_CFLAG([-fno-strict-aliasing])
+
+AC_PATH_PROG(PERL, perl, no)
+if test "x$PERL" = xno; then
+ AC_MSG_ERROR([Perl is required to build pixman.])
+fi
+AC_SUBST(PERL)
+
+dnl =========================================================================
+dnl OpenMP for the test suite?
+dnl
+
+# Check for OpenMP support (only supported by autoconf >=2.62)
+OPENMP_CFLAGS=
+m4_ifdef([AC_OPENMP], [AC_OPENMP])
+
+m4_define([openmp_test_program],[dnl
+#include <stdio.h>
+
+extern unsigned int lcg_seed;
+#pragma omp threadprivate(lcg_seed)
+unsigned int lcg_seed;
+
+unsigned function(unsigned a, unsigned b)
+{
+ lcg_seed ^= b;
+ return ((a + b) ^ a ) + lcg_seed;
+}
+
+int main(int argc, char **argv)
+{
+ int i;
+ int n1 = 0, n2 = argc;
+ unsigned checksum = 0;
+ int verbose = argv != NULL;
+ unsigned (*test_function)(unsigned, unsigned);
+ test_function = function;
+ #pragma omp parallel for reduction(+:checksum) default(none) \
+ shared(n1, n2, test_function, verbose)
+ for (i = n1; i < n2; i++)
+ {
+ unsigned crc = test_function (i, 0);
+ if (verbose)
+ printf ("%d: %08X\n", i, crc);
+ checksum += crc;
+ }
+ printf("%u\n", checksum);
+ return 0;
+}
+])
+
+PIXMAN_LINK_WITH_ENV(
+ [CFLAGS="$OPENMP_CFLAGS" LDFLAGS="$OPENMP_CFLAGS"],
+ [openmp_test_program],
+ [have_openmp=yes],
+ [have_openmp=no])
+if test "x$have_openmp" = "xyes"; then
+ AC_DEFINE(USE_OPENMP, 1, [use OpenMP in the test suite])
+else
+ OPENMP_CFLAGS=""
+fi
+AC_SUBST(OPENMP_CFLAGS)
+
+dnl =========================================================================
+dnl -fvisibility stuff
+
+PIXMAN_CHECK_CFLAG([-fvisibility=hidden], [dnl
+#if defined(__GNUC__) && (__GNUC__ >= 4)
+#ifdef _WIN32
+#error Have -fvisibility but it is ignored and generates a warning
+#endif
+#else
+error Need GCC 4.0 for visibility
+#endif
+])
+
+PIXMAN_CHECK_CFLAG([-xldscope=hidden], [dnl
+#if defined(__SUNPRO_C) && (__SUNPRO_C >= 0x550)
+#else
+error Need Sun Studio 8 for visibility
+#endif
+])
+
+dnl ===========================================================================
+dnl Check for MMX
+
+if test "x$MMX_CFLAGS" = "x" ; then
+ if test "x$SUNCC" = "xyes"; then
+ # Sun Studio doesn't have an -xarch=mmx flag, so we have to use sse
+ # but if we're building 64-bit, mmx & sse support is on by default and
+ # -xarch=sse throws an error instead
+ if test "$AMD64_ABI" = "no" ; then
+ MMX_CFLAGS="-xarch=sse"
+ fi
+ else
+ MMX_CFLAGS="-mmmx -Winline"
+ fi
+fi
+
+have_mmx_intrinsics=no
+AC_MSG_CHECKING(whether to use MMX intrinsics)
+xserver_save_CFLAGS=$CFLAGS
+CFLAGS="$MMX_CFLAGS $CFLAGS"
+AC_COMPILE_IFELSE([
+#if defined(__GNUC__) && (__GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 4))
+error "Need GCC >= 3.4 for MMX intrinsics"
+#endif
+#include <mmintrin.h>
+int main () {
+ __m64 v = _mm_cvtsi32_si64 (1);
+ return _mm_cvtsi64_si32 (v);
+}], have_mmx_intrinsics=yes)
+CFLAGS=$xserver_save_CFLAGS
+
+AC_ARG_ENABLE(mmx,
+ [AC_HELP_STRING([--disable-mmx],
+ [disable MMX fast paths])],
+ [enable_mmx=$enableval], [enable_mmx=auto])
+
+if test $enable_mmx = no ; then
+ have_mmx_intrinsics=disabled
+fi
+
+if test $have_mmx_intrinsics = yes ; then
+ AC_DEFINE(USE_MMX, 1, [use MMX compiler intrinsics])
+else
+ MMX_CFLAGS=
+fi
+
+AC_MSG_RESULT($have_mmx_intrinsics)
+if test $enable_mmx = yes && test $have_mmx_intrinsics = no ; then
+ AC_MSG_ERROR([MMX intrinsics not detected])
+fi
+
+AM_CONDITIONAL(USE_MMX, test $have_mmx_intrinsics = yes)
+
+dnl ===========================================================================
+dnl Check for SSE2
+
+if test "x$SSE2_CFLAGS" = "x" ; then
+ if test "x$SUNCC" = "xyes"; then
+ # SSE2 is enabled by default in the Sun Studio 64-bit environment
+ if test "$AMD64_ABI" = "no" ; then
+ SSE2_CFLAGS="-xarch=sse2"
+ fi
+ else
+ SSE2_CFLAGS="-mmmx -msse2 -Winline"
+ fi
+fi
+
+have_sse2_intrinsics=no
+AC_MSG_CHECKING(whether to use SSE2 intrinsics)
+xserver_save_CFLAGS=$CFLAGS
+CFLAGS="$SSE2_CFLAGS $CFLAGS"
+
+AC_COMPILE_IFELSE([
+#if defined(__GNUC__) && (__GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 2))
+# if !defined(__amd64__) && !defined(__x86_64__)
+# error "Need GCC >= 4.2 for SSE2 intrinsics on x86"
+# endif
+#endif
+#include <mmintrin.h>
+#include <xmmintrin.h>
+#include <emmintrin.h>
+int main () {
+ __m128i a = _mm_set1_epi32 (0), b = _mm_set1_epi32 (0), c;
+ c = _mm_xor_si128 (a, b);
+ return 0;
+}], have_sse2_intrinsics=yes)
+CFLAGS=$xserver_save_CFLAGS
+
+AC_ARG_ENABLE(sse2,
+ [AC_HELP_STRING([--disable-sse2],
+ [disable SSE2 fast paths])],
+ [enable_sse2=$enableval], [enable_sse2=auto])
+
+if test $enable_sse2 = no ; then
+ have_sse2_intrinsics=disabled
+fi
+
+if test $have_sse2_intrinsics = yes ; then
+ AC_DEFINE(USE_SSE2, 1, [use SSE2 compiler intrinsics])
+fi
+
+AC_MSG_RESULT($have_sse2_intrinsics)
+if test $enable_sse2 = yes && test $have_sse2_intrinsics = no ; then
+ AC_MSG_ERROR([SSE2 intrinsics not detected])
+fi
+
+AM_CONDITIONAL(USE_SSE2, test $have_sse2_intrinsics = yes)
+
+dnl ===========================================================================
+dnl Other special flags needed when building code using MMX or SSE instructions
+case $host_os in
+ solaris*)
+ # When building 32-bit binaries, apply a mapfile to ensure that the
+ # binaries aren't flagged as only able to run on MMX+SSE capable CPUs
+ # since they check at runtime before using those instructions.
+ # Not all linkers grok the mapfile format so we check for that first.
+ if test "$AMD64_ABI" = "no" ; then
+ use_hwcap_mapfile=no
+ AC_MSG_CHECKING(whether to use a hardware capability map file)
+ hwcap_save_LDFLAGS="$LDFLAGS"
+ HWCAP_LDFLAGS='-Wl,-M,$(srcdir)/solaris-hwcap.mapfile'
+ LDFLAGS="$LDFLAGS -Wl,-M,pixman/solaris-hwcap.mapfile"
+ AC_LINK_IFELSE([int main() { return 0; }],
+ use_hwcap_mapfile=yes,
+ HWCAP_LDFLAGS="")
+ LDFLAGS="$hwcap_save_LDFLAGS"
+ AC_MSG_RESULT($use_hwcap_mapfile)
+ fi
+ if test "x$MMX_LDFLAGS" = "x" ; then
+ MMX_LDFLAGS="$HWCAP_LDFLAGS"
+ fi
+ if test "x$SSE2_LDFLAGS" = "x" ; then
+ SSE2_LDFLAGS="$HWCAP_LDFLAGS"
+ fi
+ ;;
+esac
+
+AC_SUBST(MMX_CFLAGS)
+AC_SUBST(MMX_LDFLAGS)
+AC_SUBST(SSE2_CFLAGS)
+AC_SUBST(SSE2_LDFLAGS)
+
+dnl ===========================================================================
+dnl Check for VMX/Altivec
+if test -n "`$CC -v 2>&1 | grep version | grep Apple`"; then
+ VMX_CFLAGS="-faltivec"
+else
+ VMX_CFLAGS="-maltivec -mabi=altivec"
+fi
+
+have_vmx_intrinsics=no
+AC_MSG_CHECKING(whether to use VMX/Altivec intrinsics)
+xserver_save_CFLAGS=$CFLAGS
+CFLAGS="$VMX_CFLAGS $CFLAGS"
+AC_COMPILE_IFELSE([
+#if defined(__GNUC__) && (__GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 4))
+error "Need GCC >= 3.4 for sane altivec support"
+#endif
+#include <altivec.h>
+int main () {
+ vector unsigned int v = vec_splat_u32 (1);
+ v = vec_sub (v, v);
+ return 0;
+}], have_vmx_intrinsics=yes)
+CFLAGS=$xserver_save_CFLAGS
+
+AC_ARG_ENABLE(vmx,
+ [AC_HELP_STRING([--disable-vmx],
+ [disable VMX fast paths])],
+ [enable_vmx=$enableval], [enable_vmx=auto])
+
+if test $enable_vmx = no ; then
+ have_vmx_intrinsics=disabled
+fi
+
+if test $have_vmx_intrinsics = yes ; then
+ AC_DEFINE(USE_VMX, 1, [use VMX compiler intrinsics])
+else
+ VMX_CFLAGS=
+fi
+
+AC_MSG_RESULT($have_vmx_intrinsics)
+if test $enable_vmx = yes && test $have_vmx_intrinsics = no ; then
+ AC_MSG_ERROR([VMX intrinsics not detected])
+fi
+
+AC_SUBST(VMX_CFLAGS)
+
+AM_CONDITIONAL(USE_VMX, test $have_vmx_intrinsics = yes)
+
+dnl ==========================================================================
+dnl Check if assembler is gas compatible and supports ARM SIMD instructions
+have_arm_simd=no
+AC_MSG_CHECKING(whether to use ARM SIMD assembler)
+xserver_save_CFLAGS=$CFLAGS
+CFLAGS="-x assembler-with-cpp $CFLAGS"
+AC_COMPILE_IFELSE([[
+.text
+.arch armv6
+.object_arch armv4
+.arm
+.altmacro
+#ifndef __ARM_EABI__
+#error EABI is required (to be sure that calling conventions are compatible)
+#endif
+pld [r0]
+uqadd8 r0, r0, r0]], have_arm_simd=yes)
+CFLAGS=$xserver_save_CFLAGS
+
+AC_ARG_ENABLE(arm-simd,
+ [AC_HELP_STRING([--disable-arm-simd],
+ [disable ARM SIMD fast paths])],
+ [enable_arm_simd=$enableval], [enable_arm_simd=auto])
+
+if test $enable_arm_simd = no ; then
+ have_arm_simd=disabled
+fi
+
+if test $have_arm_simd = yes ; then
+ AC_DEFINE(USE_ARM_SIMD, 1, [use ARM SIMD assembly optimizations])
+fi
+
+AM_CONDITIONAL(USE_ARM_SIMD, test $have_arm_simd = yes)
+
+AC_MSG_RESULT($have_arm_simd)
+if test $enable_arm_simd = yes && test $have_arm_simd = no ; then
+ AC_MSG_ERROR([ARM SIMD intrinsics not detected])
+fi
+
+dnl ==========================================================================
+dnl Check if assembler is gas compatible and supports NEON instructions
+have_arm_neon=no
+AC_MSG_CHECKING(whether to use ARM NEON assembler)
+xserver_save_CFLAGS=$CFLAGS
+CFLAGS="-x assembler-with-cpp $CFLAGS"
+AC_COMPILE_IFELSE([[
+.text
+.fpu neon
+.arch armv7a
+.object_arch armv4
+.eabi_attribute 10, 0
+.arm
+.altmacro
+#ifndef __ARM_EABI__
+#error EABI is required (to be sure that calling conventions are compatible)
+#endif
+pld [r0]
+vmovn.u16 d0, q0]], have_arm_neon=yes)
+CFLAGS=$xserver_save_CFLAGS
+
+AC_ARG_ENABLE(arm-neon,
+ [AC_HELP_STRING([--disable-arm-neon],
+ [disable ARM NEON fast paths])],
+ [enable_arm_neon=$enableval], [enable_arm_neon=auto])
+
+if test $enable_arm_neon = no ; then
+ have_arm_neon=disabled
+fi
+
+if test $have_arm_neon = yes ; then
+ AC_DEFINE(USE_ARM_NEON, 1, [use ARM NEON assembly optimizations])
+fi
+
+AM_CONDITIONAL(USE_ARM_NEON, test $have_arm_neon = yes)
+
+AC_MSG_RESULT($have_arm_neon)
+if test $enable_arm_neon = yes && test $have_arm_neon = no ; then
+ AC_MSG_ERROR([ARM NEON intrinsics not detected])
+fi
+
+dnl =========================================================================================
+dnl Check for GNU-style inline assembly support
+
+have_gcc_inline_asm=no
+AC_MSG_CHECKING(whether to use GNU-style inline assembler)
+AC_COMPILE_IFELSE([
+int main () {
+ /* Most modern architectures have a NOP instruction, so this is a fairly generic test. */
+ asm volatile ( "\tnop\n" : : : "cc", "memory" );
+ return 0;
+}], have_gcc_inline_asm=yes)
+
+AC_ARG_ENABLE(gcc-inline-asm,
+ [AC_HELP_STRING([--disable-gcc-inline-asm],
+ [disable GNU-style inline assembler])],
+ [enable_gcc_inline_asm=$enableval], [enable_gcc_inline_asm=auto])
+
+if test $enable_gcc_inline_asm = no ; then
+ have_gcc_inline_asm=disabled
+fi
+
+if test $have_gcc_inline_asm = yes ; then
+ AC_DEFINE(USE_GCC_INLINE_ASM, 1, [use GNU-style inline assembler])
+fi
+
+AC_MSG_RESULT($have_gcc_inline_asm)
+if test $enable_gcc_inline_asm = yes && test $have_gcc_inline_asm = no ; then
+ AC_MSG_ERROR([GNU-style inline assembler not detected])
+fi
+
+AM_CONDITIONAL(USE_GCC_INLINE_ASM, test $have_gcc_inline_asm = yes)
+
+dnl ==============================================
+dnl Static test programs
+
+AC_ARG_ENABLE(static-testprogs,
+ [AC_HELP_STRING([--enable-static-testprogs],
+ [build test programs as static binaries [default=no]])],
+ [enable_static_testprogs=$enableval], [enable_static_testprogs=no])
+
+TESTPROGS_EXTRA_LDFLAGS=
+if test "x$enable_static_testprogs" = "xyes" ; then
+ TESTPROGS_EXTRA_LDFLAGS="-all-static"
+fi
+AC_SUBST(TESTPROGS_EXTRA_LDFLAGS)
+
+dnl ==============================================
+dnl Timers
+
+AC_ARG_ENABLE(timers,
+ [AC_HELP_STRING([--enable-timers],
+ [enable TIMER_BEGIN and TIMER_END macros [default=no]])],
+ [enable_timers=$enableval], [enable_timers=no])
+
+if test $enable_timers = yes ; then
+ AC_DEFINE(PIXMAN_TIMERS, 1, [enable TIMER_BEGIN/TIMER_END macros])
+fi
+AC_SUBST(PIXMAN_TIMERS)
+
+dnl ===================================
+dnl GTK+
+
+AC_ARG_ENABLE(gtk,
+ [AC_HELP_STRING([--enable-gtk],
+ [enable tests using GTK+ [default=auto]])],
+ [enable_gtk=$enableval], [enable_gtk=auto])
+
+PKG_PROG_PKG_CONFIG
+
+if test $enable_gtk = yes ; then
+ AC_CHECK_LIB([pixman-1], [pixman_version_string])
+ PKG_CHECK_MODULES(GTK, [gtk+-2.0 pixman-1])
+fi
+
+if test $enable_gtk = auto ; then
+ AC_CHECK_LIB([pixman-1], [pixman_version_string], [enable_gtk=auto], [enable_gtk=no])
+fi
+
+if test $enable_gtk = auto ; then
+ PKG_CHECK_MODULES(GTK, [gtk+-2.0 pixman-1], [enable_gtk=yes], [enable_gtk=no])
+fi
+
+AM_CONDITIONAL(HAVE_GTK, [test "x$enable_gtk" = xyes])
+
+AC_SUBST(GTK_CFLAGS)
+AC_SUBST(GTK_LIBS)
+AC_SUBST(DEP_CFLAGS)
+AC_SUBST(DEP_LIBS)
+
+dnl =====================================
+dnl posix_memalign, sigaction, alarm, gettimeofday
+
+AC_CHECK_FUNC(posix_memalign, have_posix_memalign=yes, have_posix_memalign=no)
+if test x$have_posix_memalign = xyes; then
+ AC_DEFINE(HAVE_POSIX_MEMALIGN, 1, [Whether we have posix_memalign()])
+fi
+
+AC_CHECK_FUNC(sigaction, have_sigaction=yes, have_sigaction=no)
+if test x$have_sigaction = xyes; then
+ AC_DEFINE(HAVE_SIGACTION, 1, [Whether we have sigaction()])
+fi
+
+AC_CHECK_FUNC(alarm, have_alarm=yes, have_alarm=no)
+if test x$have_alarm = xyes; then
+ AC_DEFINE(HAVE_ALARM, 1, [Whether we have alarm()])
+fi
+
+AC_CHECK_HEADER([sys/mman.h],
+ [AC_DEFINE(HAVE_SYS_MMAN_H, [1], [Define to 1 if we have <sys/mman.h>])])
+
+AC_CHECK_FUNC(mprotect, have_mprotect=yes, have_mprotect=no)
+if test x$have_mprotect = xyes; then
+ AC_DEFINE(HAVE_MPROTECT, 1, [Whether we have mprotect()])
+fi
+
+AC_CHECK_FUNC(getpagesize, have_getpagesize=yes, have_getpagesize=no)
+if test x$have_getpagesize = xyes; then
+ AC_DEFINE(HAVE_GETPAGESIZE, 1, [Whether we have getpagesize()])
+fi
+
+AC_CHECK_HEADER([fenv.h],
+ [AC_DEFINE(HAVE_FENV_H, [1], [Define to 1 if we have <fenv.h>])])
+
+AC_CHECK_LIB(m, feenableexcept, have_feenableexcept=yes, have_feenableexcept=no)
+if test x$have_feenableexcept = xyes; then
+ AC_DEFINE(HAVE_FEENABLEEXCEPT, 1, [Whether we have feenableexcept()])
+fi
+
+AC_CHECK_FUNC(gettimeofday, have_gettimeofday=yes, have_gettimeofday=no)
+AC_CHECK_HEADER(sys/time.h, have_sys_time_h=yes, have_sys_time_h=no)
+if test x$have_gettimeofday = xyes && test x$have_sys_time_h = xyes; then
+ AC_DEFINE(HAVE_GETTIMEOFDAY, 1, [Whether we have gettimeofday()])
+fi
+
+dnl =====================================
+dnl Thread local storage
+
+support_for__thread=no
+
+AC_MSG_CHECKING(for __thread)
+AC_LINK_IFELSE([
+#if defined(__MINGW32__) && !(__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 5))
+#error This MinGW version has broken __thread support
+#endif
+#ifdef __OpenBSD__
+#error OpenBSD has broken __thread support
+#endif
+static __thread int x ;
+int main () { x = 123; return x; }
+], support_for__thread=yes)
+
+if test $support_for__thread = yes; then
+ AC_DEFINE([TOOLCHAIN_SUPPORTS__THREAD],[],[Whether the tool chain supports __thread])
+fi
+
+AC_MSG_RESULT($support_for__thread)
+
+dnl
+dnl posix tls
+dnl
+
+m4_define([pthread_test_program],[dnl
+#include <stdlib.h>
+#include <pthread.h>
+
+static pthread_once_t once_control = PTHREAD_ONCE_INIT;
+static pthread_key_t key;
+
+static void
+make_key (void)
+{
+ pthread_key_create (&key, NULL);
+}
+
+int
+main ()
+{
+ void *value = NULL;
+
+ if (pthread_once (&once_control, make_key) != 0)
+ {
+ value = NULL;
+ }
+ else
+ {
+ value = pthread_getspecific (key);
+ if (!value)
+ {
+ value = malloc (100);
+ pthread_setspecific (key, value);
+ }
+ }
+ return 0;
+}
+])
+
+AC_DEFUN([PIXMAN_CHECK_PTHREAD],[dnl
+ if test "z$support_for_pthread_setspecific" != "zyes"; then
+ PIXMAN_LINK_WITH_ENV(
+ [$1], [pthread_test_program],
+ [PTHREAD_CFLAGS="$CFLAGS"
+ PTHREAD_LIBS="$LIBS"
+ PTHREAD_LDFLAGS="$LDFLAGS"
+ support_for_pthread_setspecific=yes])
+ fi
+])
+
+if test $support_for__thread = no; then
+ support_for_pthread_setspecific=no
+
+ AC_MSG_CHECKING(for pthread_setspecific)
+
+ PIXMAN_CHECK_PTHREAD([CFLAGS="-D_REENTRANT"; LIBS="-lpthread"])
+ PIXMAN_CHECK_PTHREAD([CFLAGS="-pthread"; LDFLAGS="-pthread"])
+ PIXMAN_CHECK_PTHREAD([CFLAGS="-D_REENTRANT"; LDFLAGS="-lroot"])
+
+ if test $support_for_pthread_setspecific = yes; then
+ CFLAGS="$CFLAGS $PTHREAD_CFLAGS"
+ AC_DEFINE([HAVE_PTHREAD_SETSPECIFIC], [], [Whether pthread_setspecific() is supported])
+ fi
+
+ AC_MSG_RESULT($support_for_pthread_setspecific);
+fi
+
+AC_SUBST(TOOLCHAIN_SUPPORTS__THREAD)
+AC_SUBST(HAVE_PTHREAD_SETSPECIFIC)
+AC_SUBST(PTHREAD_LDFLAGS)
+AC_SUBST(PTHREAD_LIBS)
+
+dnl =====================================
+dnl __attribute__((constructor))
+
+support_for_attribute_constructor=no
+
+AC_MSG_CHECKING(for __attribute__((constructor)))
+AC_LINK_IFELSE([
+#if defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 7))
+/* attribute 'constructor' is supported since gcc 2.7, but some compilers
+ * may only pretend to be gcc, so let's try to actually use it
+ */
+static int x = 1;
+static void __attribute__((constructor)) constructor_function () { x = 0; }
+int main (void) { return x; }
+#else
+#error not gcc or gcc version is older than 2.7
+#endif
+], support_for_attribute_constructor=yes)
+
+if test x$support_for_attribute_constructor = xyes; then
+ AC_DEFINE([TOOLCHAIN_SUPPORTS_ATTRIBUTE_CONSTRUCTOR],
+ [],[Whether the tool chain supports __attribute__((constructor))])
+fi
+
+AC_MSG_RESULT($support_for_attribute_constructor)
+AC_SUBST(TOOLCHAIN_SUPPORTS_ATTRIBUTE_CONSTRUCTOR)
+
+AC_OUTPUT([pixman-1.pc
+ pixman-1-uninstalled.pc
+ Makefile
+ pixman/Makefile
+ pixman/pixman-version.h
+ demos/Makefile
+ test/Makefile])
+
+m4_if(m4_eval(pixman_minor % 2), [1], [
+ echo
+ echo "@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@"
+ echo
+ echo " Thanks for testing this development snapshot of pixman. Please"
+ echo " report any problems you find, either by sending email to "
+ echo
+ echo " pixman@lists.freedesktop.org"
+ echo
+ echo " or by filing a bug at "
+ echo
+ echo " https://bugs.freedesktop.org/enter_bug.cgi?product=pixman "
+ echo
+ echo " If you are looking for a stable release of pixman, please note "
+ echo " that stable releases have _even_ minor version numbers. Ie., "
+ echo " pixman-0.]m4_eval(pixman_minor & ~1)[.x are stable releases, whereas pixman-$PIXMAN_VERSION_MAJOR.$PIXMAN_VERSION_MINOR.$PIXMAN_VERSION_MICRO is a "
+ echo " development snapshot that may contain bugs and experimental "
+ echo " features. "
+ echo
+ echo "@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@"
+ echo
+])
diff --git a/pixman/pixman/config.h b/pixman/pixman/config.h new file mode 100644 index 000000000..50adacc11 --- /dev/null +++ b/pixman/pixman/config.h @@ -0,0 +1,227 @@ +/* config.h. Generated from config.h.in by configure. */ +/* config.h.in. Generated from configure.ac by autoheader. */ + +/* Define if building universal (internal helper macro) */ +#undef AC_APPLE_UNIVERSAL_BUILD + +/* Include compose table cache support */ +#define COMPOSECACHE 1 + + +/* Define to 1 if you have the <dlfcn.h> header file. */ +#define HAVE_DLFCN_H 1 + +/* Define to 1 if you have the <endian.h> header file. */ +#define HAVE_ENDIAN_H 1 + +/* Use dlopen to load shared libraries */ +#define HAVE_DLOPEN 1 + + +/* Define to 1 if you have the `getpagesize' function. */ +#define HAVE_GETPAGESIZE 1 + +/* Define to 1 if you have the <inttypes.h> header file. */ +#define HAVE_INTTYPES_H 1 + +/* launchd support available */ +/* #undef HAVE_LAUNCHD */ + +/* Define to 1 if you have the <memory.h> header file. */ +#define HAVE_MEMORY_H 1 + +/* Define to 1 if you have the `poll' function. */ +#define HAVE_POLL 1 + +/* Define to 1 if you have a working `mmap' system call. */ +#define HAVE_MMAP 1 + +/* Use shl_load to load shared libraries */ +/* #undef HAVE_SHL_LOAD */ + +/* Define to 1 if the system has the type `socklen_t'. */ +#define HAVE_SOCKLEN_T 0 + +/* Define to 1 if you have the <stdint.h> header file. */ +#define HAVE_STDINT_H 1 + +/* Define to 1 if you have the <stdlib.h> header file. */ +#define HAVE_STDLIB_H 1 + +/* Define to 1 if you have the <strings.h> header file. */ +#define HAVE_STRINGS_H 1 + +/* Define to 1 if you have the <string.h> header file. */ +#define HAVE_STRING_H 1 + +/* Define to 1 if you have the <sys/poll.h> header file. */ +#define HAVE_SYS_POLL_H 1 + +/* Define to 1 if you have the `strtol' function. */ +#define HAVE_STRTOL 1 + +/* Define to 1 if you have the <sys/stat.h> header file. */ +#define HAVE_SYS_STAT_H 1 + +/* Define to 1 if you have the <sys/types.h> header file. */ +#define HAVE_SYS_TYPES_H 1 + +/* Define to 1 if you have the <unistd.h> header file. */ +#define HAVE_UNISTD_H 1 + +/* Support IPv6 for TCP connections */ +/* #undef IPv6 */ + +/* Support dynamically loaded font modules */ +#define LOADABLEFONTS 1 + +/* Support os-specific local connections */ +/* #undef LOCALCONN */ + +/* Define to the sub-directory in which libtool stores uninstalled libraries. + */ +#define LT_OBJDIR ".libs/" + +/* Disable XLOCALEDIR environment variable */ +#define NO_XLOCALEDIR 1 + +/* Name of package */ +#define PACKAGE "pixman" + +/* Define to the address where bug reports for this package should be sent. */ +#define PACKAGE_BUGREPORT ""sandmann@daimi.au.dk"" + +/* Define to the full name of this package. */ +#define PACKAGE_NAME "pixman" + +/* Define to the full name and version of this package. */ +#define PACKAGE_STRING "pixman 0.19.1" + +/* Define to the one symbol short name of this package. */ +#define PACKAGE_TARNAME "pixman" + +/* Define to the version of this package. */ +#define PACKAGE_VERSION "0.19.1" + +/* Major version of this package */ +#define PACKAGE_VERSION_MAJOR 0 + +/* Minor version of this package */ +#define PACKAGE_VERSION_MINOR 19 + +/* Patch version of this package */ +#define PACKAGE_VERSION_PATCHLEVEL 1 + +/* Define to 1 if you have the ANSI C header files. */ +#define STDC_HEADERS 1 + +/* Support TCP socket connections */ +#define TCPCONN 1 + +/* launchd support available */ +/* #undef TRANS_REOPEN */ + +/* use MMX compiler intrinsics */ +#define USE_MMX 1 + +/* Support UNIX socket connections */ +#define UNIXCONN 1 + +/* Split some i18n functions into loadable modules */ +/* #undef USE_DYNAMIC_LC */ + +/* Use the X cursor library to load cursors */ +#define USE_DYNAMIC_XCURSOR 1 + +/* use SSE2 compiler intrinsics */ +#undef USE_SSE2 + +/* use VMX compiler intrinsics */ +#undef USE_VMX + +/* poll() function is available */ +#define USE_POLL 1 + +/* Use XCB for low-level protocol implementation */ +#define USE_XCB 1 + +/* Version number of package */ +#define VERSION "0.15.18" + +/* Support bdf format bitmap font files */ +#define XFONT_BDFFORMAT 1 + +/* Location of libX11 data */ +#define X11_DATADIR "/usr/share/X11" + +/* Location of libX11 library data */ +#define X11_LIBDIR "/usr/lib/X11" + +/* Include support for XCMS */ +#define XCMS 1 + +/* Location of error message database */ +#define XERRORDB "XErrorDB" + +/* Enable XF86BIGFONT extension */ +/* #undef XF86BIGFONT */ + +/* Use XKB */ +#define XKB 1 + +/* Location of keysym database */ +#define XKEYSYMDB "XKeysymDB" + +/* support for X Locales */ +#define XLOCALE 1 + +/* Location of libX11 locale data */ +#define XLOCALEDATADIR "locale" + +/* Location of libX11 locale data */ +#define XLOCALEDIR "locale" + +/* Location of libX11 locale libraries */ +#define XLOCALELIBDIR "locale" + +/* Whether libX11 is compiled with thread support */ +#define XTHREADS /**/ + +/* Whether libX11 needs to use MT safe API's */ +#define XUSE_MTSAFE_API /**/ + +/* Enable GNU and other extensions to the C environment for glibc */ +/* #undef _GNU_SOURCE */ + +/* Support bitmap font files */ +#define XFONT_BITMAP 1 + +/* Support built-in fonts */ +#define XFONT_BUILTINS 1 + +/* Support the X Font Services Protocol */ +#define XFONT_FC 1 + +/* Support fonts in files */ +#define XFONT_FONTFILE 1 + +/* Support FreeType rasterizer for nearly all font file formats */ +#define XFONT_FREETYPE 1 + +/* Support pcf format bitmap font files */ +#define XFONT_PCFFORMAT 1 + +/* Support snf format bitmap font files */ +#define XFONT_SNFFORMAT 1 + +/* Support Speedo font files */ +#define XFONT_SPEEDO 1 + +/* Support IBM Type 1 rasterizer for Type1 font files */ +#define XFONT_TYPE1 1 + +/* Support bzip2 for bitmap fonts */ +/* #undef X_BZIP2_FONT_COMPRESSION */ + +/* Support gzip for bitmap fonts */ +#define X_GZIP_FONT_COMPRESSION 1 diff --git a/pixman/pixman/makefile b/pixman/pixman/makefile new file mode 100644 index 000000000..eef65121f --- /dev/null +++ b/pixman/pixman/makefile @@ -0,0 +1,76 @@ +LIBRARY = libpixman-1 + +INCLUDES += $(OBJDIR) + +CSRCS = \ + pixman-access.c \ + pixman-access-accessors.c \ + pixman-cpu.c \ + pixman-gradient-walker.c \ + pixman-region16.c \ + pixman-region32.c \ + pixman-image.c \ + pixman-implementation.c \ + pixman-combine32.c \ + pixman-combine64.c \ + pixman-general.c \ + pixman.c \ + pixman-fast-path.c \ + pixman-solid-fill.c \ + pixman-conical-gradient.c \ + pixman-linear-gradient.c \ + pixman-radial-gradient.c \ + pixman-bits-image.c \ + pixman-utils.c \ + pixman-edge.c \ + pixman-edge-accessors.c \ + pixman-trap.c \ + pixman-timer.c \ + pixman-matrix.c + +BUILT_SOURCES = pixman-combine32.h pixman-combine32.c pixman-combine64.h pixman-combine64.c + +$(OBJDIR)\pixman-combine32.c : pixman-combine.c.template $(OBJDIR)\pixman-combine32.h make-combine.pl + perl make-combine.pl 8 < pixman-combine.c.template > $@ +$(OBJDIR)\pixman-combine32.h : pixman-combine.h.template make-combine.pl + perl make-combine.pl 8 < pixman-combine.h.template > $@ + +$(OBJDIR)\pixman-combine64.c : pixman-combine.c.template $(OBJDIR)\pixman-combine64.h make-combine.pl + perl make-combine.pl 16 < pixman-combine.c.template > $@ +$(OBJDIR)\pixman-combine64.h : pixman-combine.h.template make-combine.pl + perl make-combine.pl 16 < pixman-combine.h.template > $@ + +## mmx code +#if USE_MMX +#noinst_LTLIBRARIES += libpixman-mmx.la +CSRCS += \ + pixman-mmx.c +# pixman-mmx.h +#libpixman_mmx_la_CFLAGS = $(DEP_CFLAGS) $(MMX_CFLAGS) +#libpixman_mmx_la_LIBADD = $(DEP_LIBS) +#libpixman_1_la_LIBADD += libpixman-mmx.la +#endif + +## vmx code +#if USE_VMX +#noinst_LTLIBRARIES += libpixman-vmx.la +#libpixman_vmx_la_SOURCES = \ +# pixman-vmx.c \ +# pixman-vmx.h \ +# pixman-combine32.h +#libpixman_vmx_la_CFLAGS = $(DEP_CFLAGS) $(VMX_CFLAGS) +#libpixman_vmx_la_LIBADD = $(DEP_LIBS) +#libpixman_1_la_LIBADD += libpixman-vmx.la +#endif + +# sse2 code +#if USE_SSE2 +#noinst_LTLIBRARIES += libpixman-sse2.la +#libpixman_sse2_la_SOURCES = \ +# pixman-sse2.c \ +# pixman-sse2.h +#libpixman_sse2_la_CFLAGS = $(DEP_CFLAGS) $(SSE2_CFLAGS) +#libpixman_sse2_la_LIBADD = $(DEP_LIBS) +#libpixman_1_la_LIBADD += libpixman-sse2.la +#endif + diff --git a/pixman/pixman/pixman-arm-common.h b/pixman/pixman/pixman-arm-common.h index 9b1322b6c..7420d9305 100644 --- a/pixman/pixman/pixman-arm-common.h +++ b/pixman/pixman/pixman-arm-common.h @@ -1,364 +1,364 @@ -/* - * Copyright © 2010 Nokia Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - * - * Author: Siarhei Siamashka (siarhei.siamashka@nokia.com) - */ - -#ifndef PIXMAN_ARM_COMMON_H -#define PIXMAN_ARM_COMMON_H - -#include "pixman-fast-path.h" - -/* Define some macros which can expand into proxy functions between - * ARM assembly optimized functions and the rest of pixman fast path API. - * - * All the low level ARM assembly functions have to use ARM EABI - * calling convention and take up to 8 arguments: - * width, height, dst, dst_stride, src, src_stride, mask, mask_stride - * - * The arguments are ordered with the most important coming first (the - * first 4 arguments are passed to function in registers, the rest are - * on stack). The last arguments are optional, for example if the - * function is not using mask, then 'mask' and 'mask_stride' can be - * omitted when doing a function call. - * - * Arguments 'src' and 'mask' contain either a pointer to the top left - * pixel of the composited rectangle or a pixel color value depending - * on the function type. In the case of just a color value (solid source - * or mask), the corresponding stride argument is unused. - */ - -#define SKIP_ZERO_SRC 1 -#define SKIP_ZERO_MASK 2 - -#define PIXMAN_ARM_BIND_FAST_PATH_SRC_DST(cputype, name, \ - src_type, src_cnt, \ - dst_type, dst_cnt) \ -void \ -pixman_composite_##name##_asm_##cputype (int32_t w, \ - int32_t h, \ - dst_type *dst, \ - int32_t dst_stride, \ - src_type *src, \ - int32_t src_stride); \ - \ -static void \ -cputype##_composite_##name (pixman_implementation_t *imp, \ - pixman_op_t op, \ - pixman_image_t * src_image, \ - pixman_image_t * mask_image, \ - pixman_image_t * dst_image, \ - int32_t src_x, \ - int32_t src_y, \ - int32_t mask_x, \ - int32_t mask_y, \ - int32_t dest_x, \ - int32_t dest_y, \ - int32_t width, \ - int32_t height) \ -{ \ - dst_type *dst_line; \ - src_type *src_line; \ - int32_t dst_stride, src_stride; \ - \ - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, src_type, \ - src_stride, src_line, src_cnt); \ - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \ - dst_stride, dst_line, dst_cnt); \ - \ - pixman_composite_##name##_asm_##cputype (width, height, \ - dst_line, dst_stride, \ - src_line, src_stride); \ -} - -#define PIXMAN_ARM_BIND_FAST_PATH_N_DST(flags, cputype, name, \ - dst_type, dst_cnt) \ -void \ -pixman_composite_##name##_asm_##cputype (int32_t w, \ - int32_t h, \ - dst_type *dst, \ - int32_t dst_stride, \ - uint32_t src); \ - \ -static void \ -cputype##_composite_##name (pixman_implementation_t *imp, \ - pixman_op_t op, \ - pixman_image_t * src_image, \ - pixman_image_t * mask_image, \ - pixman_image_t * dst_image, \ - int32_t src_x, \ - int32_t src_y, \ - int32_t mask_x, \ - int32_t mask_y, \ - int32_t dest_x, \ - int32_t dest_y, \ - int32_t width, \ - int32_t height) \ -{ \ - dst_type *dst_line; \ - int32_t dst_stride; \ - uint32_t src; \ - \ - src = _pixman_image_get_solid ( \ - imp, src_image, dst_image->bits.format); \ - \ - if ((flags & SKIP_ZERO_SRC) && src == 0) \ - return; \ - \ - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \ - dst_stride, dst_line, dst_cnt); \ - \ - pixman_composite_##name##_asm_##cputype (width, height, \ - dst_line, dst_stride, \ - src); \ -} - -#define PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST(flags, cputype, name, \ - mask_type, mask_cnt, \ - dst_type, dst_cnt) \ -void \ -pixman_composite_##name##_asm_##cputype (int32_t w, \ - int32_t h, \ - dst_type *dst, \ - int32_t dst_stride, \ - uint32_t src, \ - int32_t unused, \ - mask_type *mask, \ - int32_t mask_stride); \ - \ -static void \ -cputype##_composite_##name (pixman_implementation_t *imp, \ - pixman_op_t op, \ - pixman_image_t * src_image, \ - pixman_image_t * mask_image, \ - pixman_image_t * dst_image, \ - int32_t src_x, \ - int32_t src_y, \ - int32_t mask_x, \ - int32_t mask_y, \ - int32_t dest_x, \ - int32_t dest_y, \ - int32_t width, \ - int32_t height) \ -{ \ - dst_type *dst_line; \ - mask_type *mask_line; \ - int32_t dst_stride, mask_stride; \ - uint32_t src; \ - \ - src = _pixman_image_get_solid ( \ - imp, src_image, dst_image->bits.format); \ - \ - if ((flags & SKIP_ZERO_SRC) && src == 0) \ - return; \ - \ - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \ - dst_stride, dst_line, dst_cnt); \ - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, mask_type, \ - mask_stride, mask_line, mask_cnt); \ - \ - pixman_composite_##name##_asm_##cputype (width, height, \ - dst_line, dst_stride, \ - src, 0, \ - mask_line, mask_stride); \ -} - -#define PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST(flags, cputype, name, \ - src_type, src_cnt, \ - dst_type, dst_cnt) \ -void \ -pixman_composite_##name##_asm_##cputype (int32_t w, \ - int32_t h, \ - dst_type *dst, \ - int32_t dst_stride, \ - src_type *src, \ - int32_t src_stride, \ - uint32_t mask); \ - \ -static void \ -cputype##_composite_##name (pixman_implementation_t *imp, \ - pixman_op_t op, \ - pixman_image_t * src_image, \ - pixman_image_t * mask_image, \ - pixman_image_t * dst_image, \ - int32_t src_x, \ - int32_t src_y, \ - int32_t mask_x, \ - int32_t mask_y, \ - int32_t dest_x, \ - int32_t dest_y, \ - int32_t width, \ - int32_t height) \ -{ \ - dst_type *dst_line; \ - src_type *src_line; \ - int32_t dst_stride, src_stride; \ - uint32_t mask; \ - \ - mask = _pixman_image_get_solid ( \ - imp, mask_image, dst_image->bits.format); \ - \ - if ((flags & SKIP_ZERO_MASK) && mask == 0) \ - return; \ - \ - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \ - dst_stride, dst_line, dst_cnt); \ - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, src_type, \ - src_stride, src_line, src_cnt); \ - \ - pixman_composite_##name##_asm_##cputype (width, height, \ - dst_line, dst_stride, \ - src_line, src_stride, \ - mask); \ -} - -#define PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST(cputype, name, \ - src_type, src_cnt, \ - mask_type, mask_cnt, \ - dst_type, dst_cnt) \ -void \ -pixman_composite_##name##_asm_##cputype (int32_t w, \ - int32_t h, \ - dst_type *dst, \ - int32_t dst_stride, \ - src_type *src, \ - int32_t src_stride, \ - mask_type *mask, \ - int32_t mask_stride); \ - \ -static void \ -cputype##_composite_##name (pixman_implementation_t *imp, \ - pixman_op_t op, \ - pixman_image_t * src_image, \ - pixman_image_t * mask_image, \ - pixman_image_t * dst_image, \ - int32_t src_x, \ - int32_t src_y, \ - int32_t mask_x, \ - int32_t mask_y, \ - int32_t dest_x, \ - int32_t dest_y, \ - int32_t width, \ - int32_t height) \ -{ \ - dst_type *dst_line; \ - src_type *src_line; \ - mask_type *mask_line; \ - int32_t dst_stride, src_stride, mask_stride; \ - \ - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \ - dst_stride, dst_line, dst_cnt); \ - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, src_type, \ - src_stride, src_line, src_cnt); \ - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, mask_type, \ - mask_stride, mask_line, mask_cnt); \ - \ - pixman_composite_##name##_asm_##cputype (width, height, \ - dst_line, dst_stride, \ - src_line, src_stride, \ - mask_line, mask_stride); \ -} - -#define PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST(cputype, name, op, \ - src_type, dst_type) \ -void \ -pixman_scaled_nearest_scanline_##name##_##op##_asm_##cputype ( \ - int32_t w, \ - dst_type * dst, \ - const src_type * src, \ - pixman_fixed_t vx, \ - pixman_fixed_t unit_x); \ - \ -static force_inline void \ -scaled_nearest_scanline_##cputype##_##name##_##op (dst_type * pd, \ - const src_type * ps, \ - int32_t w, \ - pixman_fixed_t vx, \ - pixman_fixed_t unit_x, \ - pixman_fixed_t max_vx, \ - pixman_bool_t zero_src) \ -{ \ - pixman_scaled_nearest_scanline_##name##_##op##_asm_##cputype (w, pd, ps, \ - vx, unit_x);\ -} \ - \ -FAST_NEAREST_MAINLOOP (cputype##_##name##_cover_##op, \ - scaled_nearest_scanline_##cputype##_##name##_##op, \ - src_type, dst_type, COVER) \ -FAST_NEAREST_MAINLOOP (cputype##_##name##_none_##op, \ - scaled_nearest_scanline_##cputype##_##name##_##op, \ - src_type, dst_type, NONE) \ -FAST_NEAREST_MAINLOOP (cputype##_##name##_pad_##op, \ - scaled_nearest_scanline_##cputype##_##name##_##op, \ - src_type, dst_type, PAD) - -/* Provide entries for the fast path table */ -#define PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH(op,s,d,func) \ - SIMPLE_NEAREST_FAST_PATH_COVER (op,s,d,func), \ - SIMPLE_NEAREST_FAST_PATH_NONE (op,s,d,func), \ - SIMPLE_NEAREST_FAST_PATH_PAD (op,s,d,func) - -#define PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_A8_DST(flags, cputype, name, op, \ - src_type, dst_type) \ -void \ -pixman_scaled_nearest_scanline_##name##_##op##_asm_##cputype ( \ - int32_t w, \ - dst_type * dst, \ - const src_type * src, \ - pixman_fixed_t vx, \ - pixman_fixed_t unit_x, \ - const uint8_t * mask); \ - \ -static force_inline void \ -scaled_nearest_scanline_##cputype##_##name##_##op (const uint8_t * mask, \ - dst_type * pd, \ - const src_type * ps, \ - int32_t w, \ - pixman_fixed_t vx, \ - pixman_fixed_t unit_x, \ - pixman_fixed_t max_vx, \ - pixman_bool_t zero_src) \ -{ \ - if ((flags & SKIP_ZERO_SRC) && zero_src) \ - return; \ - pixman_scaled_nearest_scanline_##name##_##op##_asm_##cputype (w, pd, ps, \ - vx, unit_x, \ - mask); \ -} \ - \ -FAST_NEAREST_MAINLOOP_COMMON (cputype##_##name##_cover_##op, \ - scaled_nearest_scanline_##cputype##_##name##_##op,\ - src_type, uint8_t, dst_type, COVER, TRUE, FALSE)\ -FAST_NEAREST_MAINLOOP_COMMON (cputype##_##name##_none_##op, \ - scaled_nearest_scanline_##cputype##_##name##_##op,\ - src_type, uint8_t, dst_type, NONE, TRUE, FALSE) \ -FAST_NEAREST_MAINLOOP_COMMON (cputype##_##name##_pad_##op, \ - scaled_nearest_scanline_##cputype##_##name##_##op,\ - src_type, uint8_t, dst_type, PAD, TRUE, FALSE) - -/* Provide entries for the fast path table */ -#define PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH(op,s,d,func) \ - SIMPLE_NEAREST_A8_MASK_FAST_PATH_COVER (op,s,d,func), \ - SIMPLE_NEAREST_A8_MASK_FAST_PATH_NONE (op,s,d,func), \ - SIMPLE_NEAREST_A8_MASK_FAST_PATH_PAD (op,s,d,func) - -#endif +/*
+ * Copyright © 2010 Nokia Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Author: Siarhei Siamashka (siarhei.siamashka@nokia.com)
+ */
+
+#ifndef PIXMAN_ARM_COMMON_H
+#define PIXMAN_ARM_COMMON_H
+
+#include "pixman-fast-path.h"
+
+/* Define some macros which can expand into proxy functions between
+ * ARM assembly optimized functions and the rest of pixman fast path API.
+ *
+ * All the low level ARM assembly functions have to use ARM EABI
+ * calling convention and take up to 8 arguments:
+ * width, height, dst, dst_stride, src, src_stride, mask, mask_stride
+ *
+ * The arguments are ordered with the most important coming first (the
+ * first 4 arguments are passed to function in registers, the rest are
+ * on stack). The last arguments are optional, for example if the
+ * function is not using mask, then 'mask' and 'mask_stride' can be
+ * omitted when doing a function call.
+ *
+ * Arguments 'src' and 'mask' contain either a pointer to the top left
+ * pixel of the composited rectangle or a pixel color value depending
+ * on the function type. In the case of just a color value (solid source
+ * or mask), the corresponding stride argument is unused.
+ */
+
+#define SKIP_ZERO_SRC 1
+#define SKIP_ZERO_MASK 2
+
+#define PIXMAN_ARM_BIND_FAST_PATH_SRC_DST(cputype, name, \
+ src_type, src_cnt, \
+ dst_type, dst_cnt) \
+void \
+pixman_composite_##name##_asm_##cputype (int32_t w, \
+ int32_t h, \
+ dst_type *dst, \
+ int32_t dst_stride, \
+ src_type *src, \
+ int32_t src_stride); \
+ \
+static void \
+cputype##_composite_##name (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ pixman_image_t * src_image, \
+ pixman_image_t * mask_image, \
+ pixman_image_t * dst_image, \
+ int32_t src_x, \
+ int32_t src_y, \
+ int32_t mask_x, \
+ int32_t mask_y, \
+ int32_t dest_x, \
+ int32_t dest_y, \
+ int32_t width, \
+ int32_t height) \
+{ \
+ dst_type *dst_line; \
+ src_type *src_line; \
+ int32_t dst_stride, src_stride; \
+ \
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, src_type, \
+ src_stride, src_line, src_cnt); \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \
+ dst_stride, dst_line, dst_cnt); \
+ \
+ pixman_composite_##name##_asm_##cputype (width, height, \
+ dst_line, dst_stride, \
+ src_line, src_stride); \
+}
+
+#define PIXMAN_ARM_BIND_FAST_PATH_N_DST(flags, cputype, name, \
+ dst_type, dst_cnt) \
+void \
+pixman_composite_##name##_asm_##cputype (int32_t w, \
+ int32_t h, \
+ dst_type *dst, \
+ int32_t dst_stride, \
+ uint32_t src); \
+ \
+static void \
+cputype##_composite_##name (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ pixman_image_t * src_image, \
+ pixman_image_t * mask_image, \
+ pixman_image_t * dst_image, \
+ int32_t src_x, \
+ int32_t src_y, \
+ int32_t mask_x, \
+ int32_t mask_y, \
+ int32_t dest_x, \
+ int32_t dest_y, \
+ int32_t width, \
+ int32_t height) \
+{ \
+ dst_type *dst_line; \
+ int32_t dst_stride; \
+ uint32_t src; \
+ \
+ src = _pixman_image_get_solid ( \
+ imp, src_image, dst_image->bits.format); \
+ \
+ if ((flags & SKIP_ZERO_SRC) && src == 0) \
+ return; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \
+ dst_stride, dst_line, dst_cnt); \
+ \
+ pixman_composite_##name##_asm_##cputype (width, height, \
+ dst_line, dst_stride, \
+ src); \
+}
+
+#define PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST(flags, cputype, name, \
+ mask_type, mask_cnt, \
+ dst_type, dst_cnt) \
+void \
+pixman_composite_##name##_asm_##cputype (int32_t w, \
+ int32_t h, \
+ dst_type *dst, \
+ int32_t dst_stride, \
+ uint32_t src, \
+ int32_t unused, \
+ mask_type *mask, \
+ int32_t mask_stride); \
+ \
+static void \
+cputype##_composite_##name (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ pixman_image_t * src_image, \
+ pixman_image_t * mask_image, \
+ pixman_image_t * dst_image, \
+ int32_t src_x, \
+ int32_t src_y, \
+ int32_t mask_x, \
+ int32_t mask_y, \
+ int32_t dest_x, \
+ int32_t dest_y, \
+ int32_t width, \
+ int32_t height) \
+{ \
+ dst_type *dst_line; \
+ mask_type *mask_line; \
+ int32_t dst_stride, mask_stride; \
+ uint32_t src; \
+ \
+ src = _pixman_image_get_solid ( \
+ imp, src_image, dst_image->bits.format); \
+ \
+ if ((flags & SKIP_ZERO_SRC) && src == 0) \
+ return; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \
+ dst_stride, dst_line, dst_cnt); \
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, mask_type, \
+ mask_stride, mask_line, mask_cnt); \
+ \
+ pixman_composite_##name##_asm_##cputype (width, height, \
+ dst_line, dst_stride, \
+ src, 0, \
+ mask_line, mask_stride); \
+}
+
+#define PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST(flags, cputype, name, \
+ src_type, src_cnt, \
+ dst_type, dst_cnt) \
+void \
+pixman_composite_##name##_asm_##cputype (int32_t w, \
+ int32_t h, \
+ dst_type *dst, \
+ int32_t dst_stride, \
+ src_type *src, \
+ int32_t src_stride, \
+ uint32_t mask); \
+ \
+static void \
+cputype##_composite_##name (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ pixman_image_t * src_image, \
+ pixman_image_t * mask_image, \
+ pixman_image_t * dst_image, \
+ int32_t src_x, \
+ int32_t src_y, \
+ int32_t mask_x, \
+ int32_t mask_y, \
+ int32_t dest_x, \
+ int32_t dest_y, \
+ int32_t width, \
+ int32_t height) \
+{ \
+ dst_type *dst_line; \
+ src_type *src_line; \
+ int32_t dst_stride, src_stride; \
+ uint32_t mask; \
+ \
+ mask = _pixman_image_get_solid ( \
+ imp, mask_image, dst_image->bits.format); \
+ \
+ if ((flags & SKIP_ZERO_MASK) && mask == 0) \
+ return; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \
+ dst_stride, dst_line, dst_cnt); \
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, src_type, \
+ src_stride, src_line, src_cnt); \
+ \
+ pixman_composite_##name##_asm_##cputype (width, height, \
+ dst_line, dst_stride, \
+ src_line, src_stride, \
+ mask); \
+}
+
+#define PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST(cputype, name, \
+ src_type, src_cnt, \
+ mask_type, mask_cnt, \
+ dst_type, dst_cnt) \
+void \
+pixman_composite_##name##_asm_##cputype (int32_t w, \
+ int32_t h, \
+ dst_type *dst, \
+ int32_t dst_stride, \
+ src_type *src, \
+ int32_t src_stride, \
+ mask_type *mask, \
+ int32_t mask_stride); \
+ \
+static void \
+cputype##_composite_##name (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ pixman_image_t * src_image, \
+ pixman_image_t * mask_image, \
+ pixman_image_t * dst_image, \
+ int32_t src_x, \
+ int32_t src_y, \
+ int32_t mask_x, \
+ int32_t mask_y, \
+ int32_t dest_x, \
+ int32_t dest_y, \
+ int32_t width, \
+ int32_t height) \
+{ \
+ dst_type *dst_line; \
+ src_type *src_line; \
+ mask_type *mask_line; \
+ int32_t dst_stride, src_stride, mask_stride; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, dst_type, \
+ dst_stride, dst_line, dst_cnt); \
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, src_type, \
+ src_stride, src_line, src_cnt); \
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, mask_type, \
+ mask_stride, mask_line, mask_cnt); \
+ \
+ pixman_composite_##name##_asm_##cputype (width, height, \
+ dst_line, dst_stride, \
+ src_line, src_stride, \
+ mask_line, mask_stride); \
+}
+
+#define PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST(cputype, name, op, \
+ src_type, dst_type) \
+void \
+pixman_scaled_nearest_scanline_##name##_##op##_asm_##cputype ( \
+ int32_t w, \
+ dst_type * dst, \
+ const src_type * src, \
+ pixman_fixed_t vx, \
+ pixman_fixed_t unit_x); \
+ \
+static force_inline void \
+scaled_nearest_scanline_##cputype##_##name##_##op (dst_type * pd, \
+ const src_type * ps, \
+ int32_t w, \
+ pixman_fixed_t vx, \
+ pixman_fixed_t unit_x, \
+ pixman_fixed_t max_vx, \
+ pixman_bool_t zero_src) \
+{ \
+ pixman_scaled_nearest_scanline_##name##_##op##_asm_##cputype (w, pd, ps, \
+ vx, unit_x);\
+} \
+ \
+FAST_NEAREST_MAINLOOP (cputype##_##name##_cover_##op, \
+ scaled_nearest_scanline_##cputype##_##name##_##op, \
+ src_type, dst_type, COVER) \
+FAST_NEAREST_MAINLOOP (cputype##_##name##_none_##op, \
+ scaled_nearest_scanline_##cputype##_##name##_##op, \
+ src_type, dst_type, NONE) \
+FAST_NEAREST_MAINLOOP (cputype##_##name##_pad_##op, \
+ scaled_nearest_scanline_##cputype##_##name##_##op, \
+ src_type, dst_type, PAD)
+
+/* Provide entries for the fast path table */
+#define PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH(op,s,d,func) \
+ SIMPLE_NEAREST_FAST_PATH_COVER (op,s,d,func), \
+ SIMPLE_NEAREST_FAST_PATH_NONE (op,s,d,func), \
+ SIMPLE_NEAREST_FAST_PATH_PAD (op,s,d,func)
+
+#define PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_A8_DST(flags, cputype, name, op, \
+ src_type, dst_type) \
+void \
+pixman_scaled_nearest_scanline_##name##_##op##_asm_##cputype ( \
+ int32_t w, \
+ dst_type * dst, \
+ const src_type * src, \
+ pixman_fixed_t vx, \
+ pixman_fixed_t unit_x, \
+ const uint8_t * mask); \
+ \
+static force_inline void \
+scaled_nearest_scanline_##cputype##_##name##_##op (const uint8_t * mask, \
+ dst_type * pd, \
+ const src_type * ps, \
+ int32_t w, \
+ pixman_fixed_t vx, \
+ pixman_fixed_t unit_x, \
+ pixman_fixed_t max_vx, \
+ pixman_bool_t zero_src) \
+{ \
+ if ((flags & SKIP_ZERO_SRC) && zero_src) \
+ return; \
+ pixman_scaled_nearest_scanline_##name##_##op##_asm_##cputype (w, pd, ps, \
+ vx, unit_x, \
+ mask); \
+} \
+ \
+FAST_NEAREST_MAINLOOP_COMMON (cputype##_##name##_cover_##op, \
+ scaled_nearest_scanline_##cputype##_##name##_##op,\
+ src_type, uint8_t, dst_type, COVER, TRUE, FALSE)\
+FAST_NEAREST_MAINLOOP_COMMON (cputype##_##name##_none_##op, \
+ scaled_nearest_scanline_##cputype##_##name##_##op,\
+ src_type, uint8_t, dst_type, NONE, TRUE, FALSE) \
+FAST_NEAREST_MAINLOOP_COMMON (cputype##_##name##_pad_##op, \
+ scaled_nearest_scanline_##cputype##_##name##_##op,\
+ src_type, uint8_t, dst_type, PAD, TRUE, FALSE)
+
+/* Provide entries for the fast path table */
+#define PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH(op,s,d,func) \
+ SIMPLE_NEAREST_A8_MASK_FAST_PATH_COVER (op,s,d,func), \
+ SIMPLE_NEAREST_A8_MASK_FAST_PATH_NONE (op,s,d,func), \
+ SIMPLE_NEAREST_A8_MASK_FAST_PATH_PAD (op,s,d,func)
+
+#endif
diff --git a/pixman/pixman/pixman-arm-neon-asm.S b/pixman/pixman/pixman-arm-neon-asm.S index 47daf457c..617217beb 100644 --- a/pixman/pixman/pixman-arm-neon-asm.S +++ b/pixman/pixman/pixman-arm-neon-asm.S @@ -1,2393 +1,2393 @@ -/* - * Copyright © 2009 Nokia Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - * - * Author: Siarhei Siamashka (siarhei.siamashka@nokia.com) - */ - -/* - * This file contains implementations of NEON optimized pixel processing - * functions. There is no full and detailed tutorial, but some functions - * (those which are exposing some new or interesting features) are - * extensively commented and can be used as examples. - * - * You may want to have a look at the comments for following functions: - * - pixman_composite_over_8888_0565_asm_neon - * - pixman_composite_over_n_8_0565_asm_neon - */ - -/* Prevent the stack from becoming executable for no reason... */ -#if defined(__linux__) && defined(__ELF__) -.section .note.GNU-stack,"",%progbits -#endif - - .text - .fpu neon - .arch armv7a - .object_arch armv4 - .eabi_attribute 10, 0 /* suppress Tag_FP_arch */ - .eabi_attribute 12, 0 /* suppress Tag_Advanced_SIMD_arch */ - .arm - .altmacro - -#include "pixman-arm-neon-asm.h" - -/* Global configuration options and preferences */ - -/* - * The code can optionally make use of unaligned memory accesses to improve - * performance of handling leading/trailing pixels for each scanline. - * Configuration variable RESPECT_STRICT_ALIGNMENT can be set to 0 for - * example in linux if unaligned memory accesses are not configured to - * generate.exceptions. - */ -.set RESPECT_STRICT_ALIGNMENT, 1 - -/* - * Set default prefetch type. There is a choice between the following options: - * - * PREFETCH_TYPE_NONE (may be useful for the ARM cores where PLD is set to work - * as NOP to workaround some HW bugs or for whatever other reason) - * - * PREFETCH_TYPE_SIMPLE (may be useful for simple single-issue ARM cores where - * advanced prefetch intruduces heavy overhead) - * - * PREFETCH_TYPE_ADVANCED (useful for superscalar cores such as ARM Cortex-A8 - * which can run ARM and NEON instructions simultaneously so that extra ARM - * instructions do not add (many) extra cycles, but improve prefetch efficiency) - * - * Note: some types of function can't support advanced prefetch and fallback - * to simple one (those which handle 24bpp pixels) - */ -.set PREFETCH_TYPE_DEFAULT, PREFETCH_TYPE_ADVANCED - -/* Prefetch distance in pixels for simple prefetch */ -.set PREFETCH_DISTANCE_SIMPLE, 64 - -/* - * Implementation of pixman_composite_over_8888_0565_asm_neon - * - * This function takes a8r8g8b8 source buffer, r5g6b5 destination buffer and - * performs OVER compositing operation. Function fast_composite_over_8888_0565 - * from pixman-fast-path.c does the same in C and can be used as a reference. - * - * First we need to have some NEON assembly code which can do the actual - * operation on the pixels and provide it to the template macro. - * - * Template macro quite conveniently takes care of emitting all the necessary - * code for memory reading and writing (including quite tricky cases of - * handling unaligned leading/trailing pixels), so we only need to deal with - * the data in NEON registers. - * - * NEON registers allocation in general is recommented to be the following: - * d0, d1, d2, d3 - contain loaded source pixel data - * d4, d5, d6, d7 - contain loaded destination pixels (if they are needed) - * d24, d25, d26, d27 - contain loading mask pixel data (if mask is used) - * d28, d29, d30, d31 - place for storing the result (destination pixels) - * - * As can be seen above, four 64-bit NEON registers are used for keeping - * intermediate pixel data and up to 8 pixels can be processed in one step - * for 32bpp formats (16 pixels for 16bpp, 32 pixels for 8bpp). - * - * This particular function uses the following registers allocation: - * d0, d1, d2, d3 - contain loaded source pixel data - * d4, d5 - contain loaded destination pixels (they are needed) - * d28, d29 - place for storing the result (destination pixels) - */ - -/* - * Step one. We need to have some code to do some arithmetics on pixel data. - * This is implemented as a pair of macros: '*_head' and '*_tail'. When used - * back-to-back, they take pixel data from {d0, d1, d2, d3} and {d4, d5}, - * perform all the needed calculations and write the result to {d28, d29}. - * The rationale for having two macros and not just one will be explained - * later. In practice, any single monolitic function which does the work can - * be split into two parts in any arbitrary way without affecting correctness. - * - * There is one special trick here too. Common template macro can optionally - * make our life a bit easier by doing R, G, B, A color components - * deinterleaving for 32bpp pixel formats (and this feature is used in - * 'pixman_composite_over_8888_0565_asm_neon' function). So it means that - * instead of having 8 packed pixels in {d0, d1, d2, d3} registers, we - * actually use d0 register for blue channel (a vector of eight 8-bit - * values), d1 register for green, d2 for red and d3 for alpha. This - * simple conversion can be also done with a few NEON instructions: - * - * Packed to planar conversion: - * vuzp.8 d0, d1 - * vuzp.8 d2, d3 - * vuzp.8 d1, d3 - * vuzp.8 d0, d2 - * - * Planar to packed conversion: - * vzip.8 d0, d2 - * vzip.8 d1, d3 - * vzip.8 d2, d3 - * vzip.8 d0, d1 - * - * But pixel can be loaded directly in planar format using VLD4.8 NEON - * instruction. It is 1 cycle slower than VLD1.32, so this is not always - * desirable, that's why deinterleaving is optional. - * - * But anyway, here is the code: - */ -.macro pixman_composite_over_8888_0565_process_pixblock_head - /* convert 8 r5g6b5 pixel data from {d4, d5} to planar 8-bit format - and put data into d6 - red, d7 - green, d30 - blue */ - vshrn.u16 d6, q2, #8 - vshrn.u16 d7, q2, #3 - vsli.u16 q2, q2, #5 - vsri.u8 d6, d6, #5 - vmvn.8 d3, d3 /* invert source alpha */ - vsri.u8 d7, d7, #6 - vshrn.u16 d30, q2, #2 - /* now do alpha blending, storing results in 8-bit planar format - into d16 - red, d19 - green, d18 - blue */ - vmull.u8 q10, d3, d6 - vmull.u8 q11, d3, d7 - vmull.u8 q12, d3, d30 - vrshr.u16 q13, q10, #8 - vrshr.u16 q3, q11, #8 - vrshr.u16 q15, q12, #8 - vraddhn.u16 d20, q10, q13 - vraddhn.u16 d23, q11, q3 - vraddhn.u16 d22, q12, q15 -.endm - -.macro pixman_composite_over_8888_0565_process_pixblock_tail - /* ... continue alpha blending */ - vqadd.u8 d16, d2, d20 - vqadd.u8 q9, q0, q11 - /* convert the result to r5g6b5 and store it into {d28, d29} */ - vshll.u8 q14, d16, #8 - vshll.u8 q8, d19, #8 - vshll.u8 q9, d18, #8 - vsri.u16 q14, q8, #5 - vsri.u16 q14, q9, #11 -.endm - -/* - * OK, now we got almost everything that we need. Using the above two - * macros, the work can be done right. But now we want to optimize - * it a bit. ARM Cortex-A8 is an in-order core, and benefits really - * a lot from good code scheduling and software pipelining. - * - * Let's construct some code, which will run in the core main loop. - * Some pseudo-code of the main loop will look like this: - * head - * while (...) { - * tail - * head - * } - * tail - * - * It may look a bit weird, but this setup allows to hide instruction - * latencies better and also utilize dual-issue capability more - * efficiently (make pairs of load-store and ALU instructions). - * - * So what we need now is a '*_tail_head' macro, which will be used - * in the core main loop. A trivial straightforward implementation - * of this macro would look like this: - * - * pixman_composite_over_8888_0565_process_pixblock_tail - * vst1.16 {d28, d29}, [DST_W, :128]! - * vld1.16 {d4, d5}, [DST_R, :128]! - * vld4.32 {d0, d1, d2, d3}, [SRC]! - * pixman_composite_over_8888_0565_process_pixblock_head - * cache_preload 8, 8 - * - * Now it also got some VLD/VST instructions. We simply can't move from - * processing one block of pixels to the other one with just arithmetics. - * The previously processed data needs to be written to memory and new - * data needs to be fetched. Fortunately, this main loop does not deal - * with partial leading/trailing pixels and can load/store a full block - * of pixels in a bulk. Additionally, destination buffer is already - * 16 bytes aligned here (which is good for performance). - * - * New things here are DST_R, DST_W, SRC and MASK identifiers. These - * are the aliases for ARM registers which are used as pointers for - * accessing data. We maintain separate pointers for reading and writing - * destination buffer (DST_R and DST_W). - * - * Another new thing is 'cache_preload' macro. It is used for prefetching - * data into CPU L2 cache and improve performance when dealing with large - * images which are far larger than cache size. It uses one argument - * (actually two, but they need to be the same here) - number of pixels - * in a block. Looking into 'pixman-arm-neon-asm.h' can provide some - * details about this macro. Moreover, if good performance is needed - * the code from this macro needs to be copied into '*_tail_head' macro - * and mixed with the rest of code for optimal instructions scheduling. - * We are actually doing it below. - * - * Now after all the explanations, here is the optimized code. - * Different instruction streams (originaling from '*_head', '*_tail' - * and 'cache_preload' macro) use different indentation levels for - * better readability. Actually taking the code from one of these - * indentation levels and ignoring a few VLD/VST instructions would - * result in exactly the code from '*_head', '*_tail' or 'cache_preload' - * macro! - */ - -#if 1 - -.macro pixman_composite_over_8888_0565_process_pixblock_tail_head - vqadd.u8 d16, d2, d20 - vld1.16 {d4, d5}, [DST_R, :128]! - vqadd.u8 q9, q0, q11 - vshrn.u16 d6, q2, #8 - fetch_src_pixblock - vshrn.u16 d7, q2, #3 - vsli.u16 q2, q2, #5 - vshll.u8 q14, d16, #8 - PF add PF_X, PF_X, #8 - vshll.u8 q8, d19, #8 - PF tst PF_CTL, #0xF - vsri.u8 d6, d6, #5 - PF addne PF_X, PF_X, #8 - vmvn.8 d3, d3 - PF subne PF_CTL, PF_CTL, #1 - vsri.u8 d7, d7, #6 - vshrn.u16 d30, q2, #2 - vmull.u8 q10, d3, d6 - PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift] - vmull.u8 q11, d3, d7 - vmull.u8 q12, d3, d30 - PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift] - vsri.u16 q14, q8, #5 - PF cmp PF_X, ORIG_W - vshll.u8 q9, d18, #8 - vrshr.u16 q13, q10, #8 - PF subge PF_X, PF_X, ORIG_W - vrshr.u16 q3, q11, #8 - vrshr.u16 q15, q12, #8 - PF subges PF_CTL, PF_CTL, #0x10 - vsri.u16 q14, q9, #11 - PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]! - vraddhn.u16 d20, q10, q13 - vraddhn.u16 d23, q11, q3 - PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]! - vraddhn.u16 d22, q12, q15 - vst1.16 {d28, d29}, [DST_W, :128]! -.endm - -#else - -/* If we did not care much about the performance, we would just use this... */ -.macro pixman_composite_over_8888_0565_process_pixblock_tail_head - pixman_composite_over_8888_0565_process_pixblock_tail - vst1.16 {d28, d29}, [DST_W, :128]! - vld1.16 {d4, d5}, [DST_R, :128]! - fetch_src_pixblock - pixman_composite_over_8888_0565_process_pixblock_head - cache_preload 8, 8 -.endm - -#endif - -/* - * And now the final part. We are using 'generate_composite_function' macro - * to put all the stuff together. We are specifying the name of the function - * which we want to get, number of bits per pixel for the source, mask and - * destination (0 if unused, like mask in this case). Next come some bit - * flags: - * FLAG_DST_READWRITE - tells that the destination buffer is both read - * and written, for write-only buffer we would use - * FLAG_DST_WRITEONLY flag instead - * FLAG_DEINTERLEAVE_32BPP - tells that we prefer to work with planar data - * and separate color channels for 32bpp format. - * The next things are: - * - the number of pixels processed per iteration (8 in this case, because - * that's the maximum what can fit into four 64-bit NEON registers). - * - prefetch distance, measured in pixel blocks. In this case it is 5 times - * by 8 pixels. That would be 40 pixels, or up to 160 bytes. Optimal - * prefetch distance can be selected by running some benchmarks. - * - * After that we specify some macros, these are 'default_init', - * 'default_cleanup' here which are empty (but it is possible to have custom - * init/cleanup macros to be able to save/restore some extra NEON registers - * like d8-d15 or do anything else) followed by - * 'pixman_composite_over_8888_0565_process_pixblock_head', - * 'pixman_composite_over_8888_0565_process_pixblock_tail' and - * 'pixman_composite_over_8888_0565_process_pixblock_tail_head' - * which we got implemented above. - * - * The last part is the NEON registers allocation scheme. - */ -generate_composite_function \ - pixman_composite_over_8888_0565_asm_neon, 32, 0, 16, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_over_8888_0565_process_pixblock_head, \ - pixman_composite_over_8888_0565_process_pixblock_tail, \ - pixman_composite_over_8888_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 24 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_over_n_0565_process_pixblock_head - /* convert 8 r5g6b5 pixel data from {d4, d5} to planar 8-bit format - and put data into d6 - red, d7 - green, d30 - blue */ - vshrn.u16 d6, q2, #8 - vshrn.u16 d7, q2, #3 - vsli.u16 q2, q2, #5 - vsri.u8 d6, d6, #5 - vsri.u8 d7, d7, #6 - vshrn.u16 d30, q2, #2 - /* now do alpha blending, storing results in 8-bit planar format - into d16 - red, d19 - green, d18 - blue */ - vmull.u8 q10, d3, d6 - vmull.u8 q11, d3, d7 - vmull.u8 q12, d3, d30 - vrshr.u16 q13, q10, #8 - vrshr.u16 q3, q11, #8 - vrshr.u16 q15, q12, #8 - vraddhn.u16 d20, q10, q13 - vraddhn.u16 d23, q11, q3 - vraddhn.u16 d22, q12, q15 -.endm - -.macro pixman_composite_over_n_0565_process_pixblock_tail - /* ... continue alpha blending */ - vqadd.u8 d16, d2, d20 - vqadd.u8 q9, q0, q11 - /* convert the result to r5g6b5 and store it into {d28, d29} */ - vshll.u8 q14, d16, #8 - vshll.u8 q8, d19, #8 - vshll.u8 q9, d18, #8 - vsri.u16 q14, q8, #5 - vsri.u16 q14, q9, #11 -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_over_n_0565_process_pixblock_tail_head - pixman_composite_over_n_0565_process_pixblock_tail - vld1.16 {d4, d5}, [DST_R, :128]! - vst1.16 {d28, d29}, [DST_W, :128]! - pixman_composite_over_n_0565_process_pixblock_head - cache_preload 8, 8 -.endm - -.macro pixman_composite_over_n_0565_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vld1.32 {d3[0]}, [DUMMY] - vdup.8 d0, d3[0] - vdup.8 d1, d3[1] - vdup.8 d2, d3[2] - vdup.8 d3, d3[3] - vmvn.8 d3, d3 /* invert source alpha */ -.endm - -generate_composite_function \ - pixman_composite_over_n_0565_asm_neon, 0, 0, 16, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_n_0565_init, \ - default_cleanup, \ - pixman_composite_over_n_0565_process_pixblock_head, \ - pixman_composite_over_n_0565_process_pixblock_tail, \ - pixman_composite_over_n_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 24 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_8888_0565_process_pixblock_head - vshll.u8 q8, d1, #8 - vshll.u8 q14, d2, #8 - vshll.u8 q9, d0, #8 -.endm - -.macro pixman_composite_src_8888_0565_process_pixblock_tail - vsri.u16 q14, q8, #5 - vsri.u16 q14, q9, #11 -.endm - -.macro pixman_composite_src_8888_0565_process_pixblock_tail_head - vsri.u16 q14, q8, #5 - PF add PF_X, PF_X, #8 - PF tst PF_CTL, #0xF - fetch_src_pixblock - PF addne PF_X, PF_X, #8 - PF subne PF_CTL, PF_CTL, #1 - vsri.u16 q14, q9, #11 - PF cmp PF_X, ORIG_W - PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift] - vshll.u8 q8, d1, #8 - vst1.16 {d28, d29}, [DST_W, :128]! - PF subge PF_X, PF_X, ORIG_W - PF subges PF_CTL, PF_CTL, #0x10 - vshll.u8 q14, d2, #8 - PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]! - vshll.u8 q9, d0, #8 -.endm - -generate_composite_function \ - pixman_composite_src_8888_0565_asm_neon, 32, 0, 16, \ - FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_8888_0565_process_pixblock_head, \ - pixman_composite_src_8888_0565_process_pixblock_tail, \ - pixman_composite_src_8888_0565_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_src_0565_8888_process_pixblock_head - vshrn.u16 d30, q0, #8 - vshrn.u16 d29, q0, #3 - vsli.u16 q0, q0, #5 - vmov.u8 d31, #255 - vsri.u8 d30, d30, #5 - vsri.u8 d29, d29, #6 - vshrn.u16 d28, q0, #2 -.endm - -.macro pixman_composite_src_0565_8888_process_pixblock_tail -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_src_0565_8888_process_pixblock_tail_head - pixman_composite_src_0565_8888_process_pixblock_tail - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - fetch_src_pixblock - pixman_composite_src_0565_8888_process_pixblock_head - cache_preload 8, 8 -.endm - -generate_composite_function \ - pixman_composite_src_0565_8888_asm_neon, 16, 0, 32, \ - FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_0565_8888_process_pixblock_head, \ - pixman_composite_src_0565_8888_process_pixblock_tail, \ - pixman_composite_src_0565_8888_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_add_8_8_process_pixblock_head - vqadd.u8 q14, q0, q2 - vqadd.u8 q15, q1, q3 -.endm - -.macro pixman_composite_add_8_8_process_pixblock_tail -.endm - -.macro pixman_composite_add_8_8_process_pixblock_tail_head - fetch_src_pixblock - PF add PF_X, PF_X, #32 - PF tst PF_CTL, #0xF - vld1.8 {d4, d5, d6, d7}, [DST_R, :128]! - PF addne PF_X, PF_X, #32 - PF subne PF_CTL, PF_CTL, #1 - vst1.8 {d28, d29, d30, d31}, [DST_W, :128]! - PF cmp PF_X, ORIG_W - PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift] - PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift] - PF subge PF_X, PF_X, ORIG_W - PF subges PF_CTL, PF_CTL, #0x10 - vqadd.u8 q14, q0, q2 - PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]! - PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]! - vqadd.u8 q15, q1, q3 -.endm - -generate_composite_function \ - pixman_composite_add_8_8_asm_neon, 8, 0, 8, \ - FLAG_DST_READWRITE, \ - 32, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_add_8_8_process_pixblock_head, \ - pixman_composite_add_8_8_process_pixblock_tail, \ - pixman_composite_add_8_8_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_add_8888_8888_process_pixblock_tail_head - fetch_src_pixblock - PF add PF_X, PF_X, #8 - PF tst PF_CTL, #0xF - vld1.32 {d4, d5, d6, d7}, [DST_R, :128]! - PF addne PF_X, PF_X, #8 - PF subne PF_CTL, PF_CTL, #1 - vst1.32 {d28, d29, d30, d31}, [DST_W, :128]! - PF cmp PF_X, ORIG_W - PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift] - PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift] - PF subge PF_X, PF_X, ORIG_W - PF subges PF_CTL, PF_CTL, #0x10 - vqadd.u8 q14, q0, q2 - PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]! - PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]! - vqadd.u8 q15, q1, q3 -.endm - -generate_composite_function \ - pixman_composite_add_8888_8888_asm_neon, 32, 0, 32, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_add_8_8_process_pixblock_head, \ - pixman_composite_add_8_8_process_pixblock_tail, \ - pixman_composite_add_8888_8888_process_pixblock_tail_head - -generate_composite_function_single_scanline \ - pixman_composite_scanline_add_asm_neon, 32, 0, 32, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - default_init, \ - default_cleanup, \ - pixman_composite_add_8_8_process_pixblock_head, \ - pixman_composite_add_8_8_process_pixblock_tail, \ - pixman_composite_add_8888_8888_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_out_reverse_8888_8888_process_pixblock_head - vmvn.8 d24, d3 /* get inverted alpha */ - /* do alpha blending */ - vmull.u8 q8, d24, d4 - vmull.u8 q9, d24, d5 - vmull.u8 q10, d24, d6 - vmull.u8 q11, d24, d7 -.endm - -.macro pixman_composite_out_reverse_8888_8888_process_pixblock_tail - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - vraddhn.u16 d30, q12, q10 - vraddhn.u16 d31, q13, q11 -.endm - -.macro pixman_composite_out_reverse_8888_8888_process_pixblock_tail_head - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - vrshr.u16 q14, q8, #8 - PF add PF_X, PF_X, #8 - PF tst PF_CTL, #0xF - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - PF addne PF_X, PF_X, #8 - PF subne PF_CTL, PF_CTL, #1 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - PF cmp PF_X, ORIG_W - vraddhn.u16 d30, q12, q10 - vraddhn.u16 d31, q13, q11 - fetch_src_pixblock - PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift] - vmvn.8 d22, d3 - PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift] - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - PF subge PF_X, PF_X, ORIG_W - vmull.u8 q8, d22, d4 - PF subges PF_CTL, PF_CTL, #0x10 - vmull.u8 q9, d22, d5 - PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]! - vmull.u8 q10, d22, d6 - PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]! - vmull.u8 q11, d22, d7 -.endm - -generate_composite_function_single_scanline \ - pixman_composite_scanline_out_reverse_asm_neon, 32, 0, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init, \ - default_cleanup, \ - pixman_composite_out_reverse_8888_8888_process_pixblock_head, \ - pixman_composite_out_reverse_8888_8888_process_pixblock_tail, \ - pixman_composite_out_reverse_8888_8888_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_over_8888_8888_process_pixblock_head - pixman_composite_out_reverse_8888_8888_process_pixblock_head -.endm - -.macro pixman_composite_over_8888_8888_process_pixblock_tail - pixman_composite_out_reverse_8888_8888_process_pixblock_tail - vqadd.u8 q14, q0, q14 - vqadd.u8 q15, q1, q15 -.endm - -.macro pixman_composite_over_8888_8888_process_pixblock_tail_head - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - vrshr.u16 q14, q8, #8 - PF add PF_X, PF_X, #8 - PF tst PF_CTL, #0xF - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - PF addne PF_X, PF_X, #8 - PF subne PF_CTL, PF_CTL, #1 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - PF cmp PF_X, ORIG_W - vraddhn.u16 d30, q12, q10 - vraddhn.u16 d31, q13, q11 - vqadd.u8 q14, q0, q14 - vqadd.u8 q15, q1, q15 - fetch_src_pixblock - PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift] - vmvn.8 d22, d3 - PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift] - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - PF subge PF_X, PF_X, ORIG_W - vmull.u8 q8, d22, d4 - PF subges PF_CTL, PF_CTL, #0x10 - vmull.u8 q9, d22, d5 - PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]! - vmull.u8 q10, d22, d6 - PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]! - vmull.u8 q11, d22, d7 -.endm - -generate_composite_function \ - pixman_composite_over_8888_8888_asm_neon, 32, 0, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_over_8888_8888_process_pixblock_head, \ - pixman_composite_over_8888_8888_process_pixblock_tail, \ - pixman_composite_over_8888_8888_process_pixblock_tail_head - -generate_composite_function_single_scanline \ - pixman_composite_scanline_over_asm_neon, 32, 0, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init, \ - default_cleanup, \ - pixman_composite_over_8888_8888_process_pixblock_head, \ - pixman_composite_over_8888_8888_process_pixblock_tail, \ - pixman_composite_over_8888_8888_process_pixblock_tail_head - -/******************************************************************************/ - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_over_n_8888_process_pixblock_tail_head - pixman_composite_over_8888_8888_process_pixblock_tail - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - pixman_composite_over_8888_8888_process_pixblock_head - cache_preload 8, 8 -.endm - -.macro pixman_composite_over_n_8888_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vld1.32 {d3[0]}, [DUMMY] - vdup.8 d0, d3[0] - vdup.8 d1, d3[1] - vdup.8 d2, d3[2] - vdup.8 d3, d3[3] -.endm - -generate_composite_function \ - pixman_composite_over_n_8888_asm_neon, 0, 0, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_n_8888_init, \ - default_cleanup, \ - pixman_composite_over_8888_8888_process_pixblock_head, \ - pixman_composite_over_8888_8888_process_pixblock_tail, \ - pixman_composite_over_n_8888_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_over_reverse_n_8888_process_pixblock_tail_head - vrshr.u16 q14, q8, #8 - PF add PF_X, PF_X, #8 - PF tst PF_CTL, #0xF - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - PF addne PF_X, PF_X, #8 - PF subne PF_CTL, PF_CTL, #1 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - PF cmp PF_X, ORIG_W - vraddhn.u16 d30, q12, q10 - vraddhn.u16 d31, q13, q11 - vqadd.u8 q14, q0, q14 - vqadd.u8 q15, q1, q15 - vld4.8 {d0, d1, d2, d3}, [DST_R, :128]! - vmvn.8 d22, d3 - PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift] - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - PF subge PF_X, PF_X, ORIG_W - vmull.u8 q8, d22, d4 - PF subges PF_CTL, PF_CTL, #0x10 - vmull.u8 q9, d22, d5 - vmull.u8 q10, d22, d6 - PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]! - vmull.u8 q11, d22, d7 -.endm - -.macro pixman_composite_over_reverse_n_8888_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vld1.32 {d7[0]}, [DUMMY] - vdup.8 d4, d7[0] - vdup.8 d5, d7[1] - vdup.8 d6, d7[2] - vdup.8 d7, d7[3] -.endm - -generate_composite_function \ - pixman_composite_over_reverse_n_8888_asm_neon, 0, 0, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_reverse_n_8888_init, \ - default_cleanup, \ - pixman_composite_over_8888_8888_process_pixblock_head, \ - pixman_composite_over_8888_8888_process_pixblock_tail, \ - pixman_composite_over_reverse_n_8888_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 4, /* src_basereg */ \ - 24 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_over_8888_8_0565_process_pixblock_head - vmull.u8 q0, d24, d8 /* IN for SRC pixels (part1) */ - vmull.u8 q1, d24, d9 - vmull.u8 q6, d24, d10 - vmull.u8 q7, d24, d11 - vshrn.u16 d6, q2, #8 /* convert DST_R data to 32-bpp (part1) */ - vshrn.u16 d7, q2, #3 - vsli.u16 q2, q2, #5 - vrshr.u16 q8, q0, #8 /* IN for SRC pixels (part2) */ - vrshr.u16 q9, q1, #8 - vrshr.u16 q10, q6, #8 - vrshr.u16 q11, q7, #8 - vraddhn.u16 d0, q0, q8 - vraddhn.u16 d1, q1, q9 - vraddhn.u16 d2, q6, q10 - vraddhn.u16 d3, q7, q11 - vsri.u8 d6, d6, #5 /* convert DST_R data to 32-bpp (part2) */ - vsri.u8 d7, d7, #6 - vmvn.8 d3, d3 - vshrn.u16 d30, q2, #2 - vmull.u8 q8, d3, d6 /* now do alpha blending */ - vmull.u8 q9, d3, d7 - vmull.u8 q10, d3, d30 -.endm - -.macro pixman_composite_over_8888_8_0565_process_pixblock_tail - /* 3 cycle bubble (after vmull.u8) */ - vrshr.u16 q13, q8, #8 - vrshr.u16 q11, q9, #8 - vrshr.u16 q15, q10, #8 - vraddhn.u16 d16, q8, q13 - vraddhn.u16 d27, q9, q11 - vraddhn.u16 d26, q10, q15 - vqadd.u8 d16, d2, d16 - /* 1 cycle bubble */ - vqadd.u8 q9, q0, q13 - vshll.u8 q14, d16, #8 /* convert to 16bpp */ - vshll.u8 q8, d19, #8 - vshll.u8 q9, d18, #8 - vsri.u16 q14, q8, #5 - /* 1 cycle bubble */ - vsri.u16 q14, q9, #11 -.endm - -.macro pixman_composite_over_8888_8_0565_process_pixblock_tail_head - vld1.16 {d4, d5}, [DST_R, :128]! - vshrn.u16 d6, q2, #8 - fetch_mask_pixblock - vshrn.u16 d7, q2, #3 - fetch_src_pixblock - vmull.u8 q6, d24, d10 - vrshr.u16 q13, q8, #8 - vrshr.u16 q11, q9, #8 - vrshr.u16 q15, q10, #8 - vraddhn.u16 d16, q8, q13 - vraddhn.u16 d27, q9, q11 - vraddhn.u16 d26, q10, q15 - vqadd.u8 d16, d2, d16 - vmull.u8 q1, d24, d9 - vqadd.u8 q9, q0, q13 - vshll.u8 q14, d16, #8 - vmull.u8 q0, d24, d8 - vshll.u8 q8, d19, #8 - vshll.u8 q9, d18, #8 - vsri.u16 q14, q8, #5 - vmull.u8 q7, d24, d11 - vsri.u16 q14, q9, #11 - - cache_preload 8, 8 - - vsli.u16 q2, q2, #5 - vrshr.u16 q8, q0, #8 - vrshr.u16 q9, q1, #8 - vrshr.u16 q10, q6, #8 - vrshr.u16 q11, q7, #8 - vraddhn.u16 d0, q0, q8 - vraddhn.u16 d1, q1, q9 - vraddhn.u16 d2, q6, q10 - vraddhn.u16 d3, q7, q11 - vsri.u8 d6, d6, #5 - vsri.u8 d7, d7, #6 - vmvn.8 d3, d3 - vshrn.u16 d30, q2, #2 - vst1.16 {d28, d29}, [DST_W, :128]! - vmull.u8 q8, d3, d6 - vmull.u8 q9, d3, d7 - vmull.u8 q10, d3, d30 -.endm - -generate_composite_function \ - pixman_composite_over_8888_8_0565_asm_neon, 32, 8, 16, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_over_8888_8_0565_process_pixblock_head, \ - pixman_composite_over_8888_8_0565_process_pixblock_tail, \ - pixman_composite_over_8888_8_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 8, /* src_basereg */ \ - 24 /* mask_basereg */ - -/******************************************************************************/ - -/* - * This function needs a special initialization of solid mask. - * Solid source pixel data is fetched from stack at ARGS_STACK_OFFSET - * offset, split into color components and replicated in d8-d11 - * registers. Additionally, this function needs all the NEON registers, - * so it has to save d8-d15 registers which are callee saved according - * to ABI. These registers are restored from 'cleanup' macro. All the - * other NEON registers are caller saved, so can be clobbered freely - * without introducing any problems. - */ -.macro pixman_composite_over_n_8_0565_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vpush {d8-d15} - vld1.32 {d11[0]}, [DUMMY] - vdup.8 d8, d11[0] - vdup.8 d9, d11[1] - vdup.8 d10, d11[2] - vdup.8 d11, d11[3] -.endm - -.macro pixman_composite_over_n_8_0565_cleanup - vpop {d8-d15} -.endm - -generate_composite_function \ - pixman_composite_over_n_8_0565_asm_neon, 0, 8, 16, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_n_8_0565_init, \ - pixman_composite_over_n_8_0565_cleanup, \ - pixman_composite_over_8888_8_0565_process_pixblock_head, \ - pixman_composite_over_8888_8_0565_process_pixblock_tail, \ - pixman_composite_over_8888_8_0565_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_over_8888_n_0565_init - add DUMMY, sp, #(ARGS_STACK_OFFSET + 8) - vpush {d8-d15} - vld1.32 {d24[0]}, [DUMMY] - vdup.8 d24, d24[3] -.endm - -.macro pixman_composite_over_8888_n_0565_cleanup - vpop {d8-d15} -.endm - -generate_composite_function \ - pixman_composite_over_8888_n_0565_asm_neon, 32, 0, 16, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_8888_n_0565_init, \ - pixman_composite_over_8888_n_0565_cleanup, \ - pixman_composite_over_8888_8_0565_process_pixblock_head, \ - pixman_composite_over_8888_8_0565_process_pixblock_tail, \ - pixman_composite_over_8888_8_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 8, /* src_basereg */ \ - 24 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_0565_0565_process_pixblock_head -.endm - -.macro pixman_composite_src_0565_0565_process_pixblock_tail -.endm - -.macro pixman_composite_src_0565_0565_process_pixblock_tail_head - vst1.16 {d0, d1, d2, d3}, [DST_W, :128]! - fetch_src_pixblock - cache_preload 16, 16 -.endm - -generate_composite_function \ - pixman_composite_src_0565_0565_asm_neon, 16, 0, 16, \ - FLAG_DST_WRITEONLY, \ - 16, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_0565_0565_process_pixblock_head, \ - pixman_composite_src_0565_0565_process_pixblock_tail, \ - pixman_composite_src_0565_0565_process_pixblock_tail_head, \ - 0, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_n_8_process_pixblock_head -.endm - -.macro pixman_composite_src_n_8_process_pixblock_tail -.endm - -.macro pixman_composite_src_n_8_process_pixblock_tail_head - vst1.8 {d0, d1, d2, d3}, [DST_W, :128]! -.endm - -.macro pixman_composite_src_n_8_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vld1.32 {d0[0]}, [DUMMY] - vsli.u64 d0, d0, #8 - vsli.u64 d0, d0, #16 - vsli.u64 d0, d0, #32 - vorr d1, d0, d0 - vorr q1, q0, q0 -.endm - -.macro pixman_composite_src_n_8_cleanup -.endm - -generate_composite_function \ - pixman_composite_src_n_8_asm_neon, 0, 0, 8, \ - FLAG_DST_WRITEONLY, \ - 32, /* number of pixels, processed in a single block */ \ - 0, /* prefetch distance */ \ - pixman_composite_src_n_8_init, \ - pixman_composite_src_n_8_cleanup, \ - pixman_composite_src_n_8_process_pixblock_head, \ - pixman_composite_src_n_8_process_pixblock_tail, \ - pixman_composite_src_n_8_process_pixblock_tail_head, \ - 0, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_n_0565_process_pixblock_head -.endm - -.macro pixman_composite_src_n_0565_process_pixblock_tail -.endm - -.macro pixman_composite_src_n_0565_process_pixblock_tail_head - vst1.16 {d0, d1, d2, d3}, [DST_W, :128]! -.endm - -.macro pixman_composite_src_n_0565_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vld1.32 {d0[0]}, [DUMMY] - vsli.u64 d0, d0, #16 - vsli.u64 d0, d0, #32 - vorr d1, d0, d0 - vorr q1, q0, q0 -.endm - -.macro pixman_composite_src_n_0565_cleanup -.endm - -generate_composite_function \ - pixman_composite_src_n_0565_asm_neon, 0, 0, 16, \ - FLAG_DST_WRITEONLY, \ - 16, /* number of pixels, processed in a single block */ \ - 0, /* prefetch distance */ \ - pixman_composite_src_n_0565_init, \ - pixman_composite_src_n_0565_cleanup, \ - pixman_composite_src_n_0565_process_pixblock_head, \ - pixman_composite_src_n_0565_process_pixblock_tail, \ - pixman_composite_src_n_0565_process_pixblock_tail_head, \ - 0, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_n_8888_process_pixblock_head -.endm - -.macro pixman_composite_src_n_8888_process_pixblock_tail -.endm - -.macro pixman_composite_src_n_8888_process_pixblock_tail_head - vst1.32 {d0, d1, d2, d3}, [DST_W, :128]! -.endm - -.macro pixman_composite_src_n_8888_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vld1.32 {d0[0]}, [DUMMY] - vsli.u64 d0, d0, #32 - vorr d1, d0, d0 - vorr q1, q0, q0 -.endm - -.macro pixman_composite_src_n_8888_cleanup -.endm - -generate_composite_function \ - pixman_composite_src_n_8888_asm_neon, 0, 0, 32, \ - FLAG_DST_WRITEONLY, \ - 8, /* number of pixels, processed in a single block */ \ - 0, /* prefetch distance */ \ - pixman_composite_src_n_8888_init, \ - pixman_composite_src_n_8888_cleanup, \ - pixman_composite_src_n_8888_process_pixblock_head, \ - pixman_composite_src_n_8888_process_pixblock_tail, \ - pixman_composite_src_n_8888_process_pixblock_tail_head, \ - 0, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_8888_8888_process_pixblock_head -.endm - -.macro pixman_composite_src_8888_8888_process_pixblock_tail -.endm - -.macro pixman_composite_src_8888_8888_process_pixblock_tail_head - vst1.32 {d0, d1, d2, d3}, [DST_W, :128]! - fetch_src_pixblock - cache_preload 8, 8 -.endm - -generate_composite_function \ - pixman_composite_src_8888_8888_asm_neon, 32, 0, 32, \ - FLAG_DST_WRITEONLY, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_8888_8888_process_pixblock_head, \ - pixman_composite_src_8888_8888_process_pixblock_tail, \ - pixman_composite_src_8888_8888_process_pixblock_tail_head, \ - 0, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_x888_8888_process_pixblock_head - vorr q0, q0, q2 - vorr q1, q1, q2 -.endm - -.macro pixman_composite_src_x888_8888_process_pixblock_tail -.endm - -.macro pixman_composite_src_x888_8888_process_pixblock_tail_head - vst1.32 {d0, d1, d2, d3}, [DST_W, :128]! - fetch_src_pixblock - vorr q0, q0, q2 - vorr q1, q1, q2 - cache_preload 8, 8 -.endm - -.macro pixman_composite_src_x888_8888_init - vmov.u8 q2, #0xFF - vshl.u32 q2, q2, #24 -.endm - -generate_composite_function \ - pixman_composite_src_x888_8888_asm_neon, 32, 0, 32, \ - FLAG_DST_WRITEONLY, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - pixman_composite_src_x888_8888_init, \ - default_cleanup, \ - pixman_composite_src_x888_8888_process_pixblock_head, \ - pixman_composite_src_x888_8888_process_pixblock_tail, \ - pixman_composite_src_x888_8888_process_pixblock_tail_head, \ - 0, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_over_n_8_8888_process_pixblock_head - /* expecting deinterleaved source data in {d8, d9, d10, d11} */ - /* d8 - blue, d9 - green, d10 - red, d11 - alpha */ - /* and destination data in {d4, d5, d6, d7} */ - /* mask is in d24 (d25, d26, d27 are unused) */ - - /* in */ - vmull.u8 q0, d24, d8 - vmull.u8 q1, d24, d9 - vmull.u8 q6, d24, d10 - vmull.u8 q7, d24, d11 - vrshr.u16 q10, q0, #8 - vrshr.u16 q11, q1, #8 - vrshr.u16 q12, q6, #8 - vrshr.u16 q13, q7, #8 - vraddhn.u16 d0, q0, q10 - vraddhn.u16 d1, q1, q11 - vraddhn.u16 d2, q6, q12 - vraddhn.u16 d3, q7, q13 - vmvn.8 d24, d3 /* get inverted alpha */ - /* source: d0 - blue, d1 - green, d2 - red, d3 - alpha */ - /* destination: d4 - blue, d5 - green, d6 - red, d7 - alpha */ - /* now do alpha blending */ - vmull.u8 q8, d24, d4 - vmull.u8 q9, d24, d5 - vmull.u8 q10, d24, d6 - vmull.u8 q11, d24, d7 -.endm - -.macro pixman_composite_over_n_8_8888_process_pixblock_tail - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - vraddhn.u16 d30, q12, q10 - vraddhn.u16 d31, q13, q11 - vqadd.u8 q14, q0, q14 - vqadd.u8 q15, q1, q15 -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_over_n_8_8888_process_pixblock_tail_head - pixman_composite_over_n_8_8888_process_pixblock_tail - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - fetch_mask_pixblock - cache_preload 8, 8 - pixman_composite_over_n_8_8888_process_pixblock_head -.endm - -.macro pixman_composite_over_n_8_8888_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vpush {d8-d15} - vld1.32 {d11[0]}, [DUMMY] - vdup.8 d8, d11[0] - vdup.8 d9, d11[1] - vdup.8 d10, d11[2] - vdup.8 d11, d11[3] -.endm - -.macro pixman_composite_over_n_8_8888_cleanup - vpop {d8-d15} -.endm - -generate_composite_function \ - pixman_composite_over_n_8_8888_asm_neon, 0, 8, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_n_8_8888_init, \ - pixman_composite_over_n_8_8888_cleanup, \ - pixman_composite_over_n_8_8888_process_pixblock_head, \ - pixman_composite_over_n_8_8888_process_pixblock_tail, \ - pixman_composite_over_n_8_8888_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_over_n_8_8_process_pixblock_head - vmull.u8 q0, d24, d8 - vmull.u8 q1, d25, d8 - vmull.u8 q6, d26, d8 - vmull.u8 q7, d27, d8 - vrshr.u16 q10, q0, #8 - vrshr.u16 q11, q1, #8 - vrshr.u16 q12, q6, #8 - vrshr.u16 q13, q7, #8 - vraddhn.u16 d0, q0, q10 - vraddhn.u16 d1, q1, q11 - vraddhn.u16 d2, q6, q12 - vraddhn.u16 d3, q7, q13 - vmvn.8 q12, q0 - vmvn.8 q13, q1 - vmull.u8 q8, d24, d4 - vmull.u8 q9, d25, d5 - vmull.u8 q10, d26, d6 - vmull.u8 q11, d27, d7 -.endm - -.macro pixman_composite_over_n_8_8_process_pixblock_tail - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - vraddhn.u16 d30, q12, q10 - vraddhn.u16 d31, q13, q11 - vqadd.u8 q14, q0, q14 - vqadd.u8 q15, q1, q15 -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_over_n_8_8_process_pixblock_tail_head - vld1.8 {d4, d5, d6, d7}, [DST_R, :128]! - pixman_composite_over_n_8_8_process_pixblock_tail - fetch_mask_pixblock - cache_preload 32, 32 - vst1.8 {d28, d29, d30, d31}, [DST_W, :128]! - pixman_composite_over_n_8_8_process_pixblock_head -.endm - -.macro pixman_composite_over_n_8_8_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vpush {d8-d15} - vld1.32 {d8[0]}, [DUMMY] - vdup.8 d8, d8[3] -.endm - -.macro pixman_composite_over_n_8_8_cleanup - vpop {d8-d15} -.endm - -generate_composite_function \ - pixman_composite_over_n_8_8_asm_neon, 0, 8, 8, \ - FLAG_DST_READWRITE, \ - 32, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_n_8_8_init, \ - pixman_composite_over_n_8_8_cleanup, \ - pixman_composite_over_n_8_8_process_pixblock_head, \ - pixman_composite_over_n_8_8_process_pixblock_tail, \ - pixman_composite_over_n_8_8_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_over_n_8888_8888_ca_process_pixblock_head - /* - * 'combine_mask_ca' replacement - * - * input: solid src (n) in {d8, d9, d10, d11} - * dest in {d4, d5, d6, d7 } - * mask in {d24, d25, d26, d27} - * output: updated src in {d0, d1, d2, d3 } - * updated mask in {d24, d25, d26, d3 } - */ - vmull.u8 q0, d24, d8 - vmull.u8 q1, d25, d9 - vmull.u8 q6, d26, d10 - vmull.u8 q7, d27, d11 - vmull.u8 q9, d11, d25 - vmull.u8 q12, d11, d24 - vmull.u8 q13, d11, d26 - vrshr.u16 q8, q0, #8 - vrshr.u16 q10, q1, #8 - vrshr.u16 q11, q6, #8 - vraddhn.u16 d0, q0, q8 - vraddhn.u16 d1, q1, q10 - vraddhn.u16 d2, q6, q11 - vrshr.u16 q11, q12, #8 - vrshr.u16 q8, q9, #8 - vrshr.u16 q6, q13, #8 - vrshr.u16 q10, q7, #8 - vraddhn.u16 d24, q12, q11 - vraddhn.u16 d25, q9, q8 - vraddhn.u16 d26, q13, q6 - vraddhn.u16 d3, q7, q10 - /* - * 'combine_over_ca' replacement - * - * output: updated dest in {d28, d29, d30, d31} - */ - vmvn.8 d24, d24 - vmvn.8 d25, d25 - vmull.u8 q8, d24, d4 - vmull.u8 q9, d25, d5 - vmvn.8 d26, d26 - vmvn.8 d27, d3 - vmull.u8 q10, d26, d6 - vmull.u8 q11, d27, d7 -.endm - -.macro pixman_composite_over_n_8888_8888_ca_process_pixblock_tail - /* ... continue 'combine_over_ca' replacement */ - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vrshr.u16 q6, q10, #8 - vrshr.u16 q7, q11, #8 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - vraddhn.u16 d30, q6, q10 - vraddhn.u16 d31, q7, q11 - vqadd.u8 q14, q0, q14 - vqadd.u8 q15, q1, q15 -.endm - -.macro pixman_composite_over_n_8888_8888_ca_process_pixblock_tail_head - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - vrshr.u16 q6, q10, #8 - vrshr.u16 q7, q11, #8 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - vraddhn.u16 d30, q6, q10 - vraddhn.u16 d31, q7, q11 - fetch_mask_pixblock - vqadd.u8 q14, q0, q14 - vqadd.u8 q15, q1, q15 - cache_preload 8, 8 - pixman_composite_over_n_8888_8888_ca_process_pixblock_head - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! -.endm - -.macro pixman_composite_over_n_8888_8888_ca_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vpush {d8-d15} - vld1.32 {d11[0]}, [DUMMY] - vdup.8 d8, d11[0] - vdup.8 d9, d11[1] - vdup.8 d10, d11[2] - vdup.8 d11, d11[3] -.endm - -.macro pixman_composite_over_n_8888_8888_ca_cleanup - vpop {d8-d15} -.endm - -generate_composite_function \ - pixman_composite_over_n_8888_8888_ca_asm_neon, 0, 32, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_n_8888_8888_ca_init, \ - pixman_composite_over_n_8888_8888_ca_cleanup, \ - pixman_composite_over_n_8888_8888_ca_process_pixblock_head, \ - pixman_composite_over_n_8888_8888_ca_process_pixblock_tail, \ - pixman_composite_over_n_8888_8888_ca_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_in_n_8_process_pixblock_head - /* expecting source data in {d0, d1, d2, d3} */ - /* and destination data in {d4, d5, d6, d7} */ - vmull.u8 q8, d4, d3 - vmull.u8 q9, d5, d3 - vmull.u8 q10, d6, d3 - vmull.u8 q11, d7, d3 -.endm - -.macro pixman_composite_in_n_8_process_pixblock_tail - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - vraddhn.u16 d28, q8, q14 - vraddhn.u16 d29, q9, q15 - vraddhn.u16 d30, q10, q12 - vraddhn.u16 d31, q11, q13 -.endm - -.macro pixman_composite_in_n_8_process_pixblock_tail_head - pixman_composite_in_n_8_process_pixblock_tail - vld1.8 {d4, d5, d6, d7}, [DST_R, :128]! - cache_preload 32, 32 - pixman_composite_in_n_8_process_pixblock_head - vst1.8 {d28, d29, d30, d31}, [DST_W, :128]! -.endm - -.macro pixman_composite_in_n_8_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vld1.32 {d3[0]}, [DUMMY] - vdup.8 d3, d3[3] -.endm - -.macro pixman_composite_in_n_8_cleanup -.endm - -generate_composite_function \ - pixman_composite_in_n_8_asm_neon, 0, 0, 8, \ - FLAG_DST_READWRITE, \ - 32, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_in_n_8_init, \ - pixman_composite_in_n_8_cleanup, \ - pixman_composite_in_n_8_process_pixblock_head, \ - pixman_composite_in_n_8_process_pixblock_tail, \ - pixman_composite_in_n_8_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 24 /* mask_basereg */ - -.macro pixman_composite_add_n_8_8_process_pixblock_head - /* expecting source data in {d8, d9, d10, d11} */ - /* d8 - blue, d9 - green, d10 - red, d11 - alpha */ - /* and destination data in {d4, d5, d6, d7} */ - /* mask is in d24, d25, d26, d27 */ - vmull.u8 q0, d24, d11 - vmull.u8 q1, d25, d11 - vmull.u8 q6, d26, d11 - vmull.u8 q7, d27, d11 - vrshr.u16 q10, q0, #8 - vrshr.u16 q11, q1, #8 - vrshr.u16 q12, q6, #8 - vrshr.u16 q13, q7, #8 - vraddhn.u16 d0, q0, q10 - vraddhn.u16 d1, q1, q11 - vraddhn.u16 d2, q6, q12 - vraddhn.u16 d3, q7, q13 - vqadd.u8 q14, q0, q2 - vqadd.u8 q15, q1, q3 -.endm - -.macro pixman_composite_add_n_8_8_process_pixblock_tail -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_add_n_8_8_process_pixblock_tail_head - pixman_composite_add_n_8_8_process_pixblock_tail - vst1.8 {d28, d29, d30, d31}, [DST_W, :128]! - vld1.8 {d4, d5, d6, d7}, [DST_R, :128]! - fetch_mask_pixblock - cache_preload 32, 32 - pixman_composite_add_n_8_8_process_pixblock_head -.endm - -.macro pixman_composite_add_n_8_8_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vpush {d8-d15} - vld1.32 {d11[0]}, [DUMMY] - vdup.8 d11, d11[3] -.endm - -.macro pixman_composite_add_n_8_8_cleanup - vpop {d8-d15} -.endm - -generate_composite_function \ - pixman_composite_add_n_8_8_asm_neon, 0, 8, 8, \ - FLAG_DST_READWRITE, \ - 32, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_add_n_8_8_init, \ - pixman_composite_add_n_8_8_cleanup, \ - pixman_composite_add_n_8_8_process_pixblock_head, \ - pixman_composite_add_n_8_8_process_pixblock_tail, \ - pixman_composite_add_n_8_8_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_add_8_8_8_process_pixblock_head - /* expecting source data in {d0, d1, d2, d3} */ - /* destination data in {d4, d5, d6, d7} */ - /* mask in {d24, d25, d26, d27} */ - vmull.u8 q8, d24, d0 - vmull.u8 q9, d25, d1 - vmull.u8 q10, d26, d2 - vmull.u8 q11, d27, d3 - vrshr.u16 q0, q8, #8 - vrshr.u16 q1, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - vraddhn.u16 d0, q0, q8 - vraddhn.u16 d1, q1, q9 - vraddhn.u16 d2, q12, q10 - vraddhn.u16 d3, q13, q11 - vqadd.u8 q14, q0, q2 - vqadd.u8 q15, q1, q3 -.endm - -.macro pixman_composite_add_8_8_8_process_pixblock_tail -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_add_8_8_8_process_pixblock_tail_head - pixman_composite_add_8_8_8_process_pixblock_tail - vst1.8 {d28, d29, d30, d31}, [DST_W, :128]! - vld1.8 {d4, d5, d6, d7}, [DST_R, :128]! - fetch_mask_pixblock - fetch_src_pixblock - cache_preload 32, 32 - pixman_composite_add_8_8_8_process_pixblock_head -.endm - -.macro pixman_composite_add_8_8_8_init -.endm - -.macro pixman_composite_add_8_8_8_cleanup -.endm - -generate_composite_function \ - pixman_composite_add_8_8_8_asm_neon, 8, 8, 8, \ - FLAG_DST_READWRITE, \ - 32, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_add_8_8_8_init, \ - pixman_composite_add_8_8_8_cleanup, \ - pixman_composite_add_8_8_8_process_pixblock_head, \ - pixman_composite_add_8_8_8_process_pixblock_tail, \ - pixman_composite_add_8_8_8_process_pixblock_tail_head - -/******************************************************************************/ - -.macro pixman_composite_add_8888_8888_8888_process_pixblock_head - /* expecting source data in {d0, d1, d2, d3} */ - /* destination data in {d4, d5, d6, d7} */ - /* mask in {d24, d25, d26, d27} */ - vmull.u8 q8, d27, d0 - vmull.u8 q9, d27, d1 - vmull.u8 q10, d27, d2 - vmull.u8 q11, d27, d3 - /* 1 cycle bubble */ - vrsra.u16 q8, q8, #8 - vrsra.u16 q9, q9, #8 - vrsra.u16 q10, q10, #8 - vrsra.u16 q11, q11, #8 -.endm - -.macro pixman_composite_add_8888_8888_8888_process_pixblock_tail - /* 2 cycle bubble */ - vrshrn.u16 d28, q8, #8 - vrshrn.u16 d29, q9, #8 - vrshrn.u16 d30, q10, #8 - vrshrn.u16 d31, q11, #8 - vqadd.u8 q14, q2, q14 - /* 1 cycle bubble */ - vqadd.u8 q15, q3, q15 -.endm - -.macro pixman_composite_add_8888_8888_8888_process_pixblock_tail_head - fetch_src_pixblock - vrshrn.u16 d28, q8, #8 - fetch_mask_pixblock - vrshrn.u16 d29, q9, #8 - vmull.u8 q8, d27, d0 - vrshrn.u16 d30, q10, #8 - vmull.u8 q9, d27, d1 - vrshrn.u16 d31, q11, #8 - vmull.u8 q10, d27, d2 - vqadd.u8 q14, q2, q14 - vmull.u8 q11, d27, d3 - vqadd.u8 q15, q3, q15 - vrsra.u16 q8, q8, #8 - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - vrsra.u16 q9, q9, #8 - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - vrsra.u16 q10, q10, #8 - - cache_preload 8, 8 - - vrsra.u16 q11, q11, #8 -.endm - -generate_composite_function \ - pixman_composite_add_8888_8888_8888_asm_neon, 32, 32, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_add_8888_8888_8888_process_pixblock_head, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail_head - -generate_composite_function_single_scanline \ - pixman_composite_scanline_add_mask_asm_neon, 32, 32, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init, \ - default_cleanup, \ - pixman_composite_add_8888_8888_8888_process_pixblock_head, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail_head - -/******************************************************************************/ - -generate_composite_function \ - pixman_composite_add_8888_8_8888_asm_neon, 32, 8, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_add_8888_8888_8888_process_pixblock_head, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 27 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_add_n_8_8888_init - add DUMMY, sp, #ARGS_STACK_OFFSET - vld1.32 {d3[0]}, [DUMMY] - vdup.8 d0, d3[0] - vdup.8 d1, d3[1] - vdup.8 d2, d3[2] - vdup.8 d3, d3[3] -.endm - -.macro pixman_composite_add_n_8_8888_cleanup -.endm - -generate_composite_function \ - pixman_composite_add_n_8_8888_asm_neon, 0, 8, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_add_n_8_8888_init, \ - pixman_composite_add_n_8_8888_cleanup, \ - pixman_composite_add_8888_8888_8888_process_pixblock_head, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 27 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_add_8888_n_8888_init - add DUMMY, sp, #(ARGS_STACK_OFFSET + 8) - vld1.32 {d27[0]}, [DUMMY] - vdup.8 d27, d27[3] -.endm - -.macro pixman_composite_add_8888_n_8888_cleanup -.endm - -generate_composite_function \ - pixman_composite_add_8888_n_8888_asm_neon, 32, 0, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_add_8888_n_8888_init, \ - pixman_composite_add_8888_n_8888_cleanup, \ - pixman_composite_add_8888_8888_8888_process_pixblock_head, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail, \ - pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 27 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_out_reverse_8888_n_8888_process_pixblock_head - /* expecting source data in {d0, d1, d2, d3} */ - /* destination data in {d4, d5, d6, d7} */ - /* solid mask is in d15 */ - - /* 'in' */ - vmull.u8 q8, d15, d3 - vmull.u8 q6, d15, d2 - vmull.u8 q5, d15, d1 - vmull.u8 q4, d15, d0 - vrshr.u16 q13, q8, #8 - vrshr.u16 q12, q6, #8 - vrshr.u16 q11, q5, #8 - vrshr.u16 q10, q4, #8 - vraddhn.u16 d3, q8, q13 - vraddhn.u16 d2, q6, q12 - vraddhn.u16 d1, q5, q11 - vraddhn.u16 d0, q4, q10 - vmvn.8 d24, d3 /* get inverted alpha */ - /* now do alpha blending */ - vmull.u8 q8, d24, d4 - vmull.u8 q9, d24, d5 - vmull.u8 q10, d24, d6 - vmull.u8 q11, d24, d7 -.endm - -.macro pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vrshr.u16 q13, q11, #8 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - vraddhn.u16 d30, q12, q10 - vraddhn.u16 d31, q13, q11 -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_out_reverse_8888_8888_8888_process_pixblock_tail_head - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail - fetch_src_pixblock - cache_preload 8, 8 - fetch_mask_pixblock - pixman_composite_out_reverse_8888_n_8888_process_pixblock_head - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! -.endm - -generate_composite_function_single_scanline \ - pixman_composite_scanline_out_reverse_mask_asm_neon, 32, 32, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_out_reverse_8888_n_8888_process_pixblock_head, \ - pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail, \ - pixman_composite_out_reverse_8888_8888_8888_process_pixblock_tail_head \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 12 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_over_8888_n_8888_process_pixblock_head - pixman_composite_out_reverse_8888_n_8888_process_pixblock_head -.endm - -.macro pixman_composite_over_8888_n_8888_process_pixblock_tail - pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail - vqadd.u8 q14, q0, q14 - vqadd.u8 q15, q1, q15 -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_over_8888_n_8888_process_pixblock_tail_head - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - pixman_composite_over_8888_n_8888_process_pixblock_tail - fetch_src_pixblock - cache_preload 8, 8 - pixman_composite_over_8888_n_8888_process_pixblock_head - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! -.endm - -.macro pixman_composite_over_8888_n_8888_init - add DUMMY, sp, #48 - vpush {d8-d15} - vld1.32 {d15[0]}, [DUMMY] - vdup.8 d15, d15[3] -.endm - -.macro pixman_composite_over_8888_n_8888_cleanup - vpop {d8-d15} -.endm - -generate_composite_function \ - pixman_composite_over_8888_n_8888_asm_neon, 32, 0, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_8888_n_8888_init, \ - pixman_composite_over_8888_n_8888_cleanup, \ - pixman_composite_over_8888_n_8888_process_pixblock_head, \ - pixman_composite_over_8888_n_8888_process_pixblock_tail, \ - pixman_composite_over_8888_n_8888_process_pixblock_tail_head - -/******************************************************************************/ - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_over_8888_8888_8888_process_pixblock_tail_head - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - pixman_composite_over_8888_n_8888_process_pixblock_tail - fetch_src_pixblock - cache_preload 8, 8 - fetch_mask_pixblock - pixman_composite_over_8888_n_8888_process_pixblock_head - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! -.endm - -generate_composite_function \ - pixman_composite_over_8888_8888_8888_asm_neon, 32, 32, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_over_8888_n_8888_process_pixblock_head, \ - pixman_composite_over_8888_n_8888_process_pixblock_tail, \ - pixman_composite_over_8888_8888_8888_process_pixblock_tail_head \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 12 /* mask_basereg */ - -generate_composite_function_single_scanline \ - pixman_composite_scanline_over_mask_asm_neon, 32, 32, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_over_8888_n_8888_process_pixblock_head, \ - pixman_composite_over_8888_n_8888_process_pixblock_tail, \ - pixman_composite_over_8888_8888_8888_process_pixblock_tail_head \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 12 /* mask_basereg */ - -/******************************************************************************/ - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_over_8888_8_8888_process_pixblock_tail_head - vld4.8 {d4, d5, d6, d7}, [DST_R, :128]! - pixman_composite_over_8888_n_8888_process_pixblock_tail - fetch_src_pixblock - cache_preload 8, 8 - fetch_mask_pixblock - pixman_composite_over_8888_n_8888_process_pixblock_head - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! -.endm - -generate_composite_function \ - pixman_composite_over_8888_8_8888_asm_neon, 32, 8, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_over_8888_n_8888_process_pixblock_head, \ - pixman_composite_over_8888_n_8888_process_pixblock_tail, \ - pixman_composite_over_8888_8_8888_process_pixblock_tail_head \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 15 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_0888_0888_process_pixblock_head -.endm - -.macro pixman_composite_src_0888_0888_process_pixblock_tail -.endm - -.macro pixman_composite_src_0888_0888_process_pixblock_tail_head - vst3.8 {d0, d1, d2}, [DST_W]! - fetch_src_pixblock - cache_preload 8, 8 -.endm - -generate_composite_function \ - pixman_composite_src_0888_0888_asm_neon, 24, 0, 24, \ - FLAG_DST_WRITEONLY, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_0888_0888_process_pixblock_head, \ - pixman_composite_src_0888_0888_process_pixblock_tail, \ - pixman_composite_src_0888_0888_process_pixblock_tail_head, \ - 0, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_0888_8888_rev_process_pixblock_head - vswp d0, d2 -.endm - -.macro pixman_composite_src_0888_8888_rev_process_pixblock_tail -.endm - -.macro pixman_composite_src_0888_8888_rev_process_pixblock_tail_head - vst4.8 {d0, d1, d2, d3}, [DST_W]! - fetch_src_pixblock - vswp d0, d2 - cache_preload 8, 8 -.endm - -.macro pixman_composite_src_0888_8888_rev_init - veor d3, d3, d3 -.endm - -generate_composite_function \ - pixman_composite_src_0888_8888_rev_asm_neon, 24, 0, 32, \ - FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - pixman_composite_src_0888_8888_rev_init, \ - default_cleanup, \ - pixman_composite_src_0888_8888_rev_process_pixblock_head, \ - pixman_composite_src_0888_8888_rev_process_pixblock_tail, \ - pixman_composite_src_0888_8888_rev_process_pixblock_tail_head, \ - 0, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_0888_0565_rev_process_pixblock_head - vshll.u8 q8, d1, #8 - vshll.u8 q9, d2, #8 -.endm - -.macro pixman_composite_src_0888_0565_rev_process_pixblock_tail - vshll.u8 q14, d0, #8 - vsri.u16 q14, q8, #5 - vsri.u16 q14, q9, #11 -.endm - -.macro pixman_composite_src_0888_0565_rev_process_pixblock_tail_head - vshll.u8 q14, d0, #8 - fetch_src_pixblock - vsri.u16 q14, q8, #5 - vsri.u16 q14, q9, #11 - vshll.u8 q8, d1, #8 - vst1.16 {d28, d29}, [DST_W, :128]! - vshll.u8 q9, d2, #8 -.endm - -generate_composite_function \ - pixman_composite_src_0888_0565_rev_asm_neon, 24, 0, 16, \ - FLAG_DST_WRITEONLY, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_0888_0565_rev_process_pixblock_head, \ - pixman_composite_src_0888_0565_rev_process_pixblock_tail, \ - pixman_composite_src_0888_0565_rev_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_pixbuf_8888_process_pixblock_head - vmull.u8 q8, d3, d0 - vmull.u8 q9, d3, d1 - vmull.u8 q10, d3, d2 -.endm - -.macro pixman_composite_src_pixbuf_8888_process_pixblock_tail - vrshr.u16 q11, q8, #8 - vswp d3, d31 - vrshr.u16 q12, q9, #8 - vrshr.u16 q13, q10, #8 - vraddhn.u16 d30, q11, q8 - vraddhn.u16 d29, q12, q9 - vraddhn.u16 d28, q13, q10 -.endm - -.macro pixman_composite_src_pixbuf_8888_process_pixblock_tail_head - vrshr.u16 q11, q8, #8 - vswp d3, d31 - vrshr.u16 q12, q9, #8 - vrshr.u16 q13, q10, #8 - fetch_src_pixblock - vraddhn.u16 d30, q11, q8 - PF add PF_X, PF_X, #8 - PF tst PF_CTL, #0xF - PF addne PF_X, PF_X, #8 - PF subne PF_CTL, PF_CTL, #1 - vraddhn.u16 d29, q12, q9 - vraddhn.u16 d28, q13, q10 - vmull.u8 q8, d3, d0 - vmull.u8 q9, d3, d1 - vmull.u8 q10, d3, d2 - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - PF cmp PF_X, ORIG_W - PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift] - PF subge PF_X, PF_X, ORIG_W - PF subges PF_CTL, PF_CTL, #0x10 - PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]! -.endm - -generate_composite_function \ - pixman_composite_src_pixbuf_8888_asm_neon, 32, 0, 32, \ - FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_pixbuf_8888_process_pixblock_head, \ - pixman_composite_src_pixbuf_8888_process_pixblock_tail, \ - pixman_composite_src_pixbuf_8888_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_src_rpixbuf_8888_process_pixblock_head - vmull.u8 q8, d3, d0 - vmull.u8 q9, d3, d1 - vmull.u8 q10, d3, d2 -.endm - -.macro pixman_composite_src_rpixbuf_8888_process_pixblock_tail - vrshr.u16 q11, q8, #8 - vswp d3, d31 - vrshr.u16 q12, q9, #8 - vrshr.u16 q13, q10, #8 - vraddhn.u16 d28, q11, q8 - vraddhn.u16 d29, q12, q9 - vraddhn.u16 d30, q13, q10 -.endm - -.macro pixman_composite_src_rpixbuf_8888_process_pixblock_tail_head - vrshr.u16 q11, q8, #8 - vswp d3, d31 - vrshr.u16 q12, q9, #8 - vrshr.u16 q13, q10, #8 - fetch_src_pixblock - vraddhn.u16 d28, q11, q8 - PF add PF_X, PF_X, #8 - PF tst PF_CTL, #0xF - PF addne PF_X, PF_X, #8 - PF subne PF_CTL, PF_CTL, #1 - vraddhn.u16 d29, q12, q9 - vraddhn.u16 d30, q13, q10 - vmull.u8 q8, d3, d0 - vmull.u8 q9, d3, d1 - vmull.u8 q10, d3, d2 - vst4.8 {d28, d29, d30, d31}, [DST_W, :128]! - PF cmp PF_X, ORIG_W - PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift] - PF subge PF_X, PF_X, ORIG_W - PF subges PF_CTL, PF_CTL, #0x10 - PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]! -.endm - -generate_composite_function \ - pixman_composite_src_rpixbuf_8888_asm_neon, 32, 0, 32, \ - FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - 10, /* prefetch distance */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_rpixbuf_8888_process_pixblock_head, \ - pixman_composite_src_rpixbuf_8888_process_pixblock_tail, \ - pixman_composite_src_rpixbuf_8888_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 0, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_over_0565_8_0565_process_pixblock_head - /* mask is in d15 */ - convert_0565_to_x888 q4, d2, d1, d0 - convert_0565_to_x888 q5, d6, d5, d4 - /* source pixel data is in {d0, d1, d2, XX} */ - /* destination pixel data is in {d4, d5, d6, XX} */ - vmvn.8 d7, d15 - vmull.u8 q6, d15, d2 - vmull.u8 q5, d15, d1 - vmull.u8 q4, d15, d0 - vmull.u8 q8, d7, d4 - vmull.u8 q9, d7, d5 - vmull.u8 q13, d7, d6 - vrshr.u16 q12, q6, #8 - vrshr.u16 q11, q5, #8 - vrshr.u16 q10, q4, #8 - vraddhn.u16 d2, q6, q12 - vraddhn.u16 d1, q5, q11 - vraddhn.u16 d0, q4, q10 -.endm - -.macro pixman_composite_over_0565_8_0565_process_pixblock_tail - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q13, #8 - vraddhn.u16 d28, q14, q8 - vraddhn.u16 d29, q15, q9 - vraddhn.u16 d30, q12, q13 - vqadd.u8 q0, q0, q14 - vqadd.u8 q1, q1, q15 - /* 32bpp result is in {d0, d1, d2, XX} */ - convert_8888_to_0565 d2, d1, d0, q14, q15, q3 -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_over_0565_8_0565_process_pixblock_tail_head - fetch_mask_pixblock - pixman_composite_over_0565_8_0565_process_pixblock_tail - fetch_src_pixblock - vld1.16 {d10, d11}, [DST_R, :128]! - cache_preload 8, 8 - pixman_composite_over_0565_8_0565_process_pixblock_head - vst1.16 {d28, d29}, [DST_W, :128]! -.endm - -generate_composite_function \ - pixman_composite_over_0565_8_0565_asm_neon, 16, 8, 16, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_over_0565_8_0565_process_pixblock_head, \ - pixman_composite_over_0565_8_0565_process_pixblock_tail, \ - pixman_composite_over_0565_8_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 10, /* dst_r_basereg */ \ - 8, /* src_basereg */ \ - 15 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_over_0565_n_0565_init - add DUMMY, sp, #(ARGS_STACK_OFFSET + 8) - vpush {d8-d15} - vld1.32 {d15[0]}, [DUMMY] - vdup.8 d15, d15[3] -.endm - -.macro pixman_composite_over_0565_n_0565_cleanup - vpop {d8-d15} -.endm - -generate_composite_function \ - pixman_composite_over_0565_n_0565_asm_neon, 16, 0, 16, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - pixman_composite_over_0565_n_0565_init, \ - pixman_composite_over_0565_n_0565_cleanup, \ - pixman_composite_over_0565_8_0565_process_pixblock_head, \ - pixman_composite_over_0565_8_0565_process_pixblock_tail, \ - pixman_composite_over_0565_8_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 10, /* dst_r_basereg */ \ - 8, /* src_basereg */ \ - 15 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_add_0565_8_0565_process_pixblock_head - /* mask is in d15 */ - convert_0565_to_x888 q4, d2, d1, d0 - convert_0565_to_x888 q5, d6, d5, d4 - /* source pixel data is in {d0, d1, d2, XX} */ - /* destination pixel data is in {d4, d5, d6, XX} */ - vmull.u8 q6, d15, d2 - vmull.u8 q5, d15, d1 - vmull.u8 q4, d15, d0 - vrshr.u16 q12, q6, #8 - vrshr.u16 q11, q5, #8 - vrshr.u16 q10, q4, #8 - vraddhn.u16 d2, q6, q12 - vraddhn.u16 d1, q5, q11 - vraddhn.u16 d0, q4, q10 -.endm - -.macro pixman_composite_add_0565_8_0565_process_pixblock_tail - vqadd.u8 q0, q0, q2 - vqadd.u8 q1, q1, q3 - /* 32bpp result is in {d0, d1, d2, XX} */ - convert_8888_to_0565 d2, d1, d0, q14, q15, q3 -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_add_0565_8_0565_process_pixblock_tail_head - fetch_mask_pixblock - pixman_composite_add_0565_8_0565_process_pixblock_tail - fetch_src_pixblock - vld1.16 {d10, d11}, [DST_R, :128]! - cache_preload 8, 8 - pixman_composite_add_0565_8_0565_process_pixblock_head - vst1.16 {d28, d29}, [DST_W, :128]! -.endm - -generate_composite_function \ - pixman_composite_add_0565_8_0565_asm_neon, 16, 8, 16, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_add_0565_8_0565_process_pixblock_head, \ - pixman_composite_add_0565_8_0565_process_pixblock_tail, \ - pixman_composite_add_0565_8_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 10, /* dst_r_basereg */ \ - 8, /* src_basereg */ \ - 15 /* mask_basereg */ - -/******************************************************************************/ - -.macro pixman_composite_out_reverse_8_0565_process_pixblock_head - /* mask is in d15 */ - convert_0565_to_x888 q5, d6, d5, d4 - /* destination pixel data is in {d4, d5, d6, xx} */ - vmvn.8 d24, d15 /* get inverted alpha */ - /* now do alpha blending */ - vmull.u8 q8, d24, d4 - vmull.u8 q9, d24, d5 - vmull.u8 q10, d24, d6 -.endm - -.macro pixman_composite_out_reverse_8_0565_process_pixblock_tail - vrshr.u16 q14, q8, #8 - vrshr.u16 q15, q9, #8 - vrshr.u16 q12, q10, #8 - vraddhn.u16 d0, q14, q8 - vraddhn.u16 d1, q15, q9 - vraddhn.u16 d2, q12, q10 - /* 32bpp result is in {d0, d1, d2, XX} */ - convert_8888_to_0565 d2, d1, d0, q14, q15, q3 -.endm - -/* TODO: expand macros and do better instructions scheduling */ -.macro pixman_composite_out_reverse_8_0565_process_pixblock_tail_head - fetch_src_pixblock - pixman_composite_out_reverse_8_0565_process_pixblock_tail - vld1.16 {d10, d11}, [DST_R, :128]! - cache_preload 8, 8 - pixman_composite_out_reverse_8_0565_process_pixblock_head - vst1.16 {d28, d29}, [DST_W, :128]! -.endm - -generate_composite_function \ - pixman_composite_out_reverse_8_0565_asm_neon, 8, 0, 16, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - 5, /* prefetch distance */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_out_reverse_8_0565_process_pixblock_head, \ - pixman_composite_out_reverse_8_0565_process_pixblock_tail, \ - pixman_composite_out_reverse_8_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 10, /* dst_r_basereg */ \ - 15, /* src_basereg */ \ - 0 /* mask_basereg */ - -/******************************************************************************/ - -generate_composite_function_nearest_scanline \ - pixman_scaled_nearest_scanline_8888_8888_OVER_asm_neon, 32, 0, 32, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init, \ - default_cleanup, \ - pixman_composite_over_8888_8888_process_pixblock_head, \ - pixman_composite_over_8888_8888_process_pixblock_tail, \ - pixman_composite_over_8888_8888_process_pixblock_tail_head - -generate_composite_function_nearest_scanline \ - pixman_scaled_nearest_scanline_8888_0565_OVER_asm_neon, 32, 0, 16, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init, \ - default_cleanup, \ - pixman_composite_over_8888_0565_process_pixblock_head, \ - pixman_composite_over_8888_0565_process_pixblock_tail, \ - pixman_composite_over_8888_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 0, /* src_basereg */ \ - 24 /* mask_basereg */ - -generate_composite_function_nearest_scanline \ - pixman_scaled_nearest_scanline_8888_0565_SRC_asm_neon, 32, 0, 16, \ - FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_8888_0565_process_pixblock_head, \ - pixman_composite_src_8888_0565_process_pixblock_tail, \ - pixman_composite_src_8888_0565_process_pixblock_tail_head - -generate_composite_function_nearest_scanline \ - pixman_scaled_nearest_scanline_0565_8888_SRC_asm_neon, 16, 0, 32, \ - FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init, \ - default_cleanup, \ - pixman_composite_src_0565_8888_process_pixblock_head, \ - pixman_composite_src_0565_8888_process_pixblock_tail, \ - pixman_composite_src_0565_8888_process_pixblock_tail_head - -generate_composite_function_nearest_scanline \ - pixman_scaled_nearest_scanline_8888_8_0565_OVER_asm_neon, 32, 8, 16, \ - FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \ - 8, /* number of pixels, processed in a single block */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_over_8888_8_0565_process_pixblock_head, \ - pixman_composite_over_8888_8_0565_process_pixblock_tail, \ - pixman_composite_over_8888_8_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 4, /* dst_r_basereg */ \ - 8, /* src_basereg */ \ - 24 /* mask_basereg */ - -generate_composite_function_nearest_scanline \ - pixman_scaled_nearest_scanline_0565_8_0565_OVER_asm_neon, 16, 8, 16, \ - FLAG_DST_READWRITE, \ - 8, /* number of pixels, processed in a single block */ \ - default_init_need_all_regs, \ - default_cleanup_need_all_regs, \ - pixman_composite_over_0565_8_0565_process_pixblock_head, \ - pixman_composite_over_0565_8_0565_process_pixblock_tail, \ - pixman_composite_over_0565_8_0565_process_pixblock_tail_head, \ - 28, /* dst_w_basereg */ \ - 10, /* dst_r_basereg */ \ - 8, /* src_basereg */ \ - 15 /* mask_basereg */ +/*
+ * Copyright © 2009 Nokia Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Author: Siarhei Siamashka (siarhei.siamashka@nokia.com)
+ */
+
+/*
+ * This file contains implementations of NEON optimized pixel processing
+ * functions. There is no full and detailed tutorial, but some functions
+ * (those which are exposing some new or interesting features) are
+ * extensively commented and can be used as examples.
+ *
+ * You may want to have a look at the comments for following functions:
+ * - pixman_composite_over_8888_0565_asm_neon
+ * - pixman_composite_over_n_8_0565_asm_neon
+ */
+
+/* Prevent the stack from becoming executable for no reason... */
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif
+
+ .text
+ .fpu neon
+ .arch armv7a
+ .object_arch armv4
+ .eabi_attribute 10, 0 /* suppress Tag_FP_arch */
+ .eabi_attribute 12, 0 /* suppress Tag_Advanced_SIMD_arch */
+ .arm
+ .altmacro
+
+#include "pixman-arm-neon-asm.h"
+
+/* Global configuration options and preferences */
+
+/*
+ * The code can optionally make use of unaligned memory accesses to improve
+ * performance of handling leading/trailing pixels for each scanline.
+ * Configuration variable RESPECT_STRICT_ALIGNMENT can be set to 0 for
+ * example in linux if unaligned memory accesses are not configured to
+ * generate.exceptions.
+ */
+.set RESPECT_STRICT_ALIGNMENT, 1
+
+/*
+ * Set default prefetch type. There is a choice between the following options:
+ *
+ * PREFETCH_TYPE_NONE (may be useful for the ARM cores where PLD is set to work
+ * as NOP to workaround some HW bugs or for whatever other reason)
+ *
+ * PREFETCH_TYPE_SIMPLE (may be useful for simple single-issue ARM cores where
+ * advanced prefetch intruduces heavy overhead)
+ *
+ * PREFETCH_TYPE_ADVANCED (useful for superscalar cores such as ARM Cortex-A8
+ * which can run ARM and NEON instructions simultaneously so that extra ARM
+ * instructions do not add (many) extra cycles, but improve prefetch efficiency)
+ *
+ * Note: some types of function can't support advanced prefetch and fallback
+ * to simple one (those which handle 24bpp pixels)
+ */
+.set PREFETCH_TYPE_DEFAULT, PREFETCH_TYPE_ADVANCED
+
+/* Prefetch distance in pixels for simple prefetch */
+.set PREFETCH_DISTANCE_SIMPLE, 64
+
+/*
+ * Implementation of pixman_composite_over_8888_0565_asm_neon
+ *
+ * This function takes a8r8g8b8 source buffer, r5g6b5 destination buffer and
+ * performs OVER compositing operation. Function fast_composite_over_8888_0565
+ * from pixman-fast-path.c does the same in C and can be used as a reference.
+ *
+ * First we need to have some NEON assembly code which can do the actual
+ * operation on the pixels and provide it to the template macro.
+ *
+ * Template macro quite conveniently takes care of emitting all the necessary
+ * code for memory reading and writing (including quite tricky cases of
+ * handling unaligned leading/trailing pixels), so we only need to deal with
+ * the data in NEON registers.
+ *
+ * NEON registers allocation in general is recommented to be the following:
+ * d0, d1, d2, d3 - contain loaded source pixel data
+ * d4, d5, d6, d7 - contain loaded destination pixels (if they are needed)
+ * d24, d25, d26, d27 - contain loading mask pixel data (if mask is used)
+ * d28, d29, d30, d31 - place for storing the result (destination pixels)
+ *
+ * As can be seen above, four 64-bit NEON registers are used for keeping
+ * intermediate pixel data and up to 8 pixels can be processed in one step
+ * for 32bpp formats (16 pixels for 16bpp, 32 pixels for 8bpp).
+ *
+ * This particular function uses the following registers allocation:
+ * d0, d1, d2, d3 - contain loaded source pixel data
+ * d4, d5 - contain loaded destination pixels (they are needed)
+ * d28, d29 - place for storing the result (destination pixels)
+ */
+
+/*
+ * Step one. We need to have some code to do some arithmetics on pixel data.
+ * This is implemented as a pair of macros: '*_head' and '*_tail'. When used
+ * back-to-back, they take pixel data from {d0, d1, d2, d3} and {d4, d5},
+ * perform all the needed calculations and write the result to {d28, d29}.
+ * The rationale for having two macros and not just one will be explained
+ * later. In practice, any single monolitic function which does the work can
+ * be split into two parts in any arbitrary way without affecting correctness.
+ *
+ * There is one special trick here too. Common template macro can optionally
+ * make our life a bit easier by doing R, G, B, A color components
+ * deinterleaving for 32bpp pixel formats (and this feature is used in
+ * 'pixman_composite_over_8888_0565_asm_neon' function). So it means that
+ * instead of having 8 packed pixels in {d0, d1, d2, d3} registers, we
+ * actually use d0 register for blue channel (a vector of eight 8-bit
+ * values), d1 register for green, d2 for red and d3 for alpha. This
+ * simple conversion can be also done with a few NEON instructions:
+ *
+ * Packed to planar conversion:
+ * vuzp.8 d0, d1
+ * vuzp.8 d2, d3
+ * vuzp.8 d1, d3
+ * vuzp.8 d0, d2
+ *
+ * Planar to packed conversion:
+ * vzip.8 d0, d2
+ * vzip.8 d1, d3
+ * vzip.8 d2, d3
+ * vzip.8 d0, d1
+ *
+ * But pixel can be loaded directly in planar format using VLD4.8 NEON
+ * instruction. It is 1 cycle slower than VLD1.32, so this is not always
+ * desirable, that's why deinterleaving is optional.
+ *
+ * But anyway, here is the code:
+ */
+.macro pixman_composite_over_8888_0565_process_pixblock_head
+ /* convert 8 r5g6b5 pixel data from {d4, d5} to planar 8-bit format
+ and put data into d6 - red, d7 - green, d30 - blue */
+ vshrn.u16 d6, q2, #8
+ vshrn.u16 d7, q2, #3
+ vsli.u16 q2, q2, #5
+ vsri.u8 d6, d6, #5
+ vmvn.8 d3, d3 /* invert source alpha */
+ vsri.u8 d7, d7, #6
+ vshrn.u16 d30, q2, #2
+ /* now do alpha blending, storing results in 8-bit planar format
+ into d16 - red, d19 - green, d18 - blue */
+ vmull.u8 q10, d3, d6
+ vmull.u8 q11, d3, d7
+ vmull.u8 q12, d3, d30
+ vrshr.u16 q13, q10, #8
+ vrshr.u16 q3, q11, #8
+ vrshr.u16 q15, q12, #8
+ vraddhn.u16 d20, q10, q13
+ vraddhn.u16 d23, q11, q3
+ vraddhn.u16 d22, q12, q15
+.endm
+
+.macro pixman_composite_over_8888_0565_process_pixblock_tail
+ /* ... continue alpha blending */
+ vqadd.u8 d16, d2, d20
+ vqadd.u8 q9, q0, q11
+ /* convert the result to r5g6b5 and store it into {d28, d29} */
+ vshll.u8 q14, d16, #8
+ vshll.u8 q8, d19, #8
+ vshll.u8 q9, d18, #8
+ vsri.u16 q14, q8, #5
+ vsri.u16 q14, q9, #11
+.endm
+
+/*
+ * OK, now we got almost everything that we need. Using the above two
+ * macros, the work can be done right. But now we want to optimize
+ * it a bit. ARM Cortex-A8 is an in-order core, and benefits really
+ * a lot from good code scheduling and software pipelining.
+ *
+ * Let's construct some code, which will run in the core main loop.
+ * Some pseudo-code of the main loop will look like this:
+ * head
+ * while (...) {
+ * tail
+ * head
+ * }
+ * tail
+ *
+ * It may look a bit weird, but this setup allows to hide instruction
+ * latencies better and also utilize dual-issue capability more
+ * efficiently (make pairs of load-store and ALU instructions).
+ *
+ * So what we need now is a '*_tail_head' macro, which will be used
+ * in the core main loop. A trivial straightforward implementation
+ * of this macro would look like this:
+ *
+ * pixman_composite_over_8888_0565_process_pixblock_tail
+ * vst1.16 {d28, d29}, [DST_W, :128]!
+ * vld1.16 {d4, d5}, [DST_R, :128]!
+ * vld4.32 {d0, d1, d2, d3}, [SRC]!
+ * pixman_composite_over_8888_0565_process_pixblock_head
+ * cache_preload 8, 8
+ *
+ * Now it also got some VLD/VST instructions. We simply can't move from
+ * processing one block of pixels to the other one with just arithmetics.
+ * The previously processed data needs to be written to memory and new
+ * data needs to be fetched. Fortunately, this main loop does not deal
+ * with partial leading/trailing pixels and can load/store a full block
+ * of pixels in a bulk. Additionally, destination buffer is already
+ * 16 bytes aligned here (which is good for performance).
+ *
+ * New things here are DST_R, DST_W, SRC and MASK identifiers. These
+ * are the aliases for ARM registers which are used as pointers for
+ * accessing data. We maintain separate pointers for reading and writing
+ * destination buffer (DST_R and DST_W).
+ *
+ * Another new thing is 'cache_preload' macro. It is used for prefetching
+ * data into CPU L2 cache and improve performance when dealing with large
+ * images which are far larger than cache size. It uses one argument
+ * (actually two, but they need to be the same here) - number of pixels
+ * in a block. Looking into 'pixman-arm-neon-asm.h' can provide some
+ * details about this macro. Moreover, if good performance is needed
+ * the code from this macro needs to be copied into '*_tail_head' macro
+ * and mixed with the rest of code for optimal instructions scheduling.
+ * We are actually doing it below.
+ *
+ * Now after all the explanations, here is the optimized code.
+ * Different instruction streams (originaling from '*_head', '*_tail'
+ * and 'cache_preload' macro) use different indentation levels for
+ * better readability. Actually taking the code from one of these
+ * indentation levels and ignoring a few VLD/VST instructions would
+ * result in exactly the code from '*_head', '*_tail' or 'cache_preload'
+ * macro!
+ */
+
+#if 1
+
+.macro pixman_composite_over_8888_0565_process_pixblock_tail_head
+ vqadd.u8 d16, d2, d20
+ vld1.16 {d4, d5}, [DST_R, :128]!
+ vqadd.u8 q9, q0, q11
+ vshrn.u16 d6, q2, #8
+ fetch_src_pixblock
+ vshrn.u16 d7, q2, #3
+ vsli.u16 q2, q2, #5
+ vshll.u8 q14, d16, #8
+ PF add PF_X, PF_X, #8
+ vshll.u8 q8, d19, #8
+ PF tst PF_CTL, #0xF
+ vsri.u8 d6, d6, #5
+ PF addne PF_X, PF_X, #8
+ vmvn.8 d3, d3
+ PF subne PF_CTL, PF_CTL, #1
+ vsri.u8 d7, d7, #6
+ vshrn.u16 d30, q2, #2
+ vmull.u8 q10, d3, d6
+ PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+ vmull.u8 q11, d3, d7
+ vmull.u8 q12, d3, d30
+ PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
+ vsri.u16 q14, q8, #5
+ PF cmp PF_X, ORIG_W
+ vshll.u8 q9, d18, #8
+ vrshr.u16 q13, q10, #8
+ PF subge PF_X, PF_X, ORIG_W
+ vrshr.u16 q3, q11, #8
+ vrshr.u16 q15, q12, #8
+ PF subges PF_CTL, PF_CTL, #0x10
+ vsri.u16 q14, q9, #11
+ PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+ vraddhn.u16 d20, q10, q13
+ vraddhn.u16 d23, q11, q3
+ PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
+ vraddhn.u16 d22, q12, q15
+ vst1.16 {d28, d29}, [DST_W, :128]!
+.endm
+
+#else
+
+/* If we did not care much about the performance, we would just use this... */
+.macro pixman_composite_over_8888_0565_process_pixblock_tail_head
+ pixman_composite_over_8888_0565_process_pixblock_tail
+ vst1.16 {d28, d29}, [DST_W, :128]!
+ vld1.16 {d4, d5}, [DST_R, :128]!
+ fetch_src_pixblock
+ pixman_composite_over_8888_0565_process_pixblock_head
+ cache_preload 8, 8
+.endm
+
+#endif
+
+/*
+ * And now the final part. We are using 'generate_composite_function' macro
+ * to put all the stuff together. We are specifying the name of the function
+ * which we want to get, number of bits per pixel for the source, mask and
+ * destination (0 if unused, like mask in this case). Next come some bit
+ * flags:
+ * FLAG_DST_READWRITE - tells that the destination buffer is both read
+ * and written, for write-only buffer we would use
+ * FLAG_DST_WRITEONLY flag instead
+ * FLAG_DEINTERLEAVE_32BPP - tells that we prefer to work with planar data
+ * and separate color channels for 32bpp format.
+ * The next things are:
+ * - the number of pixels processed per iteration (8 in this case, because
+ * that's the maximum what can fit into four 64-bit NEON registers).
+ * - prefetch distance, measured in pixel blocks. In this case it is 5 times
+ * by 8 pixels. That would be 40 pixels, or up to 160 bytes. Optimal
+ * prefetch distance can be selected by running some benchmarks.
+ *
+ * After that we specify some macros, these are 'default_init',
+ * 'default_cleanup' here which are empty (but it is possible to have custom
+ * init/cleanup macros to be able to save/restore some extra NEON registers
+ * like d8-d15 or do anything else) followed by
+ * 'pixman_composite_over_8888_0565_process_pixblock_head',
+ * 'pixman_composite_over_8888_0565_process_pixblock_tail' and
+ * 'pixman_composite_over_8888_0565_process_pixblock_tail_head'
+ * which we got implemented above.
+ *
+ * The last part is the NEON registers allocation scheme.
+ */
+generate_composite_function \
+ pixman_composite_over_8888_0565_asm_neon, 32, 0, 16, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_over_8888_0565_process_pixblock_head, \
+ pixman_composite_over_8888_0565_process_pixblock_tail, \
+ pixman_composite_over_8888_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 24 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_over_n_0565_process_pixblock_head
+ /* convert 8 r5g6b5 pixel data from {d4, d5} to planar 8-bit format
+ and put data into d6 - red, d7 - green, d30 - blue */
+ vshrn.u16 d6, q2, #8
+ vshrn.u16 d7, q2, #3
+ vsli.u16 q2, q2, #5
+ vsri.u8 d6, d6, #5
+ vsri.u8 d7, d7, #6
+ vshrn.u16 d30, q2, #2
+ /* now do alpha blending, storing results in 8-bit planar format
+ into d16 - red, d19 - green, d18 - blue */
+ vmull.u8 q10, d3, d6
+ vmull.u8 q11, d3, d7
+ vmull.u8 q12, d3, d30
+ vrshr.u16 q13, q10, #8
+ vrshr.u16 q3, q11, #8
+ vrshr.u16 q15, q12, #8
+ vraddhn.u16 d20, q10, q13
+ vraddhn.u16 d23, q11, q3
+ vraddhn.u16 d22, q12, q15
+.endm
+
+.macro pixman_composite_over_n_0565_process_pixblock_tail
+ /* ... continue alpha blending */
+ vqadd.u8 d16, d2, d20
+ vqadd.u8 q9, q0, q11
+ /* convert the result to r5g6b5 and store it into {d28, d29} */
+ vshll.u8 q14, d16, #8
+ vshll.u8 q8, d19, #8
+ vshll.u8 q9, d18, #8
+ vsri.u16 q14, q8, #5
+ vsri.u16 q14, q9, #11
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_n_0565_process_pixblock_tail_head
+ pixman_composite_over_n_0565_process_pixblock_tail
+ vld1.16 {d4, d5}, [DST_R, :128]!
+ vst1.16 {d28, d29}, [DST_W, :128]!
+ pixman_composite_over_n_0565_process_pixblock_head
+ cache_preload 8, 8
+.endm
+
+.macro pixman_composite_over_n_0565_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vld1.32 {d3[0]}, [DUMMY]
+ vdup.8 d0, d3[0]
+ vdup.8 d1, d3[1]
+ vdup.8 d2, d3[2]
+ vdup.8 d3, d3[3]
+ vmvn.8 d3, d3 /* invert source alpha */
+.endm
+
+generate_composite_function \
+ pixman_composite_over_n_0565_asm_neon, 0, 0, 16, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_n_0565_init, \
+ default_cleanup, \
+ pixman_composite_over_n_0565_process_pixblock_head, \
+ pixman_composite_over_n_0565_process_pixblock_tail, \
+ pixman_composite_over_n_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 24 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_8888_0565_process_pixblock_head
+ vshll.u8 q8, d1, #8
+ vshll.u8 q14, d2, #8
+ vshll.u8 q9, d0, #8
+.endm
+
+.macro pixman_composite_src_8888_0565_process_pixblock_tail
+ vsri.u16 q14, q8, #5
+ vsri.u16 q14, q9, #11
+.endm
+
+.macro pixman_composite_src_8888_0565_process_pixblock_tail_head
+ vsri.u16 q14, q8, #5
+ PF add PF_X, PF_X, #8
+ PF tst PF_CTL, #0xF
+ fetch_src_pixblock
+ PF addne PF_X, PF_X, #8
+ PF subne PF_CTL, PF_CTL, #1
+ vsri.u16 q14, q9, #11
+ PF cmp PF_X, ORIG_W
+ PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+ vshll.u8 q8, d1, #8
+ vst1.16 {d28, d29}, [DST_W, :128]!
+ PF subge PF_X, PF_X, ORIG_W
+ PF subges PF_CTL, PF_CTL, #0x10
+ vshll.u8 q14, d2, #8
+ PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+ vshll.u8 q9, d0, #8
+.endm
+
+generate_composite_function \
+ pixman_composite_src_8888_0565_asm_neon, 32, 0, 16, \
+ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_8888_0565_process_pixblock_head, \
+ pixman_composite_src_8888_0565_process_pixblock_tail, \
+ pixman_composite_src_8888_0565_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_src_0565_8888_process_pixblock_head
+ vshrn.u16 d30, q0, #8
+ vshrn.u16 d29, q0, #3
+ vsli.u16 q0, q0, #5
+ vmov.u8 d31, #255
+ vsri.u8 d30, d30, #5
+ vsri.u8 d29, d29, #6
+ vshrn.u16 d28, q0, #2
+.endm
+
+.macro pixman_composite_src_0565_8888_process_pixblock_tail
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_src_0565_8888_process_pixblock_tail_head
+ pixman_composite_src_0565_8888_process_pixblock_tail
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ fetch_src_pixblock
+ pixman_composite_src_0565_8888_process_pixblock_head
+ cache_preload 8, 8
+.endm
+
+generate_composite_function \
+ pixman_composite_src_0565_8888_asm_neon, 16, 0, 32, \
+ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_0565_8888_process_pixblock_head, \
+ pixman_composite_src_0565_8888_process_pixblock_tail, \
+ pixman_composite_src_0565_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_add_8_8_process_pixblock_head
+ vqadd.u8 q14, q0, q2
+ vqadd.u8 q15, q1, q3
+.endm
+
+.macro pixman_composite_add_8_8_process_pixblock_tail
+.endm
+
+.macro pixman_composite_add_8_8_process_pixblock_tail_head
+ fetch_src_pixblock
+ PF add PF_X, PF_X, #32
+ PF tst PF_CTL, #0xF
+ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ PF addne PF_X, PF_X, #32
+ PF subne PF_CTL, PF_CTL, #1
+ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ PF cmp PF_X, ORIG_W
+ PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+ PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
+ PF subge PF_X, PF_X, ORIG_W
+ PF subges PF_CTL, PF_CTL, #0x10
+ vqadd.u8 q14, q0, q2
+ PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+ PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
+ vqadd.u8 q15, q1, q3
+.endm
+
+generate_composite_function \
+ pixman_composite_add_8_8_asm_neon, 8, 0, 8, \
+ FLAG_DST_READWRITE, \
+ 32, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_add_8_8_process_pixblock_head, \
+ pixman_composite_add_8_8_process_pixblock_tail, \
+ pixman_composite_add_8_8_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_add_8888_8888_process_pixblock_tail_head
+ fetch_src_pixblock
+ PF add PF_X, PF_X, #8
+ PF tst PF_CTL, #0xF
+ vld1.32 {d4, d5, d6, d7}, [DST_R, :128]!
+ PF addne PF_X, PF_X, #8
+ PF subne PF_CTL, PF_CTL, #1
+ vst1.32 {d28, d29, d30, d31}, [DST_W, :128]!
+ PF cmp PF_X, ORIG_W
+ PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+ PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
+ PF subge PF_X, PF_X, ORIG_W
+ PF subges PF_CTL, PF_CTL, #0x10
+ vqadd.u8 q14, q0, q2
+ PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+ PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
+ vqadd.u8 q15, q1, q3
+.endm
+
+generate_composite_function \
+ pixman_composite_add_8888_8888_asm_neon, 32, 0, 32, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_add_8_8_process_pixblock_head, \
+ pixman_composite_add_8_8_process_pixblock_tail, \
+ pixman_composite_add_8888_8888_process_pixblock_tail_head
+
+generate_composite_function_single_scanline \
+ pixman_composite_scanline_add_asm_neon, 32, 0, 32, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_add_8_8_process_pixblock_head, \
+ pixman_composite_add_8_8_process_pixblock_tail, \
+ pixman_composite_add_8888_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_out_reverse_8888_8888_process_pixblock_head
+ vmvn.8 d24, d3 /* get inverted alpha */
+ /* do alpha blending */
+ vmull.u8 q8, d24, d4
+ vmull.u8 q9, d24, d5
+ vmull.u8 q10, d24, d6
+ vmull.u8 q11, d24, d7
+.endm
+
+.macro pixman_composite_out_reverse_8888_8888_process_pixblock_tail
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ vraddhn.u16 d30, q12, q10
+ vraddhn.u16 d31, q13, q11
+.endm
+
+.macro pixman_composite_out_reverse_8888_8888_process_pixblock_tail_head
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ vrshr.u16 q14, q8, #8
+ PF add PF_X, PF_X, #8
+ PF tst PF_CTL, #0xF
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ PF addne PF_X, PF_X, #8
+ PF subne PF_CTL, PF_CTL, #1
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ PF cmp PF_X, ORIG_W
+ vraddhn.u16 d30, q12, q10
+ vraddhn.u16 d31, q13, q11
+ fetch_src_pixblock
+ PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+ vmvn.8 d22, d3
+ PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ PF subge PF_X, PF_X, ORIG_W
+ vmull.u8 q8, d22, d4
+ PF subges PF_CTL, PF_CTL, #0x10
+ vmull.u8 q9, d22, d5
+ PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+ vmull.u8 q10, d22, d6
+ PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
+ vmull.u8 q11, d22, d7
+.endm
+
+generate_composite_function_single_scanline \
+ pixman_composite_scanline_out_reverse_asm_neon, 32, 0, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_out_reverse_8888_8888_process_pixblock_head, \
+ pixman_composite_out_reverse_8888_8888_process_pixblock_tail, \
+ pixman_composite_out_reverse_8888_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_over_8888_8888_process_pixblock_head
+ pixman_composite_out_reverse_8888_8888_process_pixblock_head
+.endm
+
+.macro pixman_composite_over_8888_8888_process_pixblock_tail
+ pixman_composite_out_reverse_8888_8888_process_pixblock_tail
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+.endm
+
+.macro pixman_composite_over_8888_8888_process_pixblock_tail_head
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ vrshr.u16 q14, q8, #8
+ PF add PF_X, PF_X, #8
+ PF tst PF_CTL, #0xF
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ PF addne PF_X, PF_X, #8
+ PF subne PF_CTL, PF_CTL, #1
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ PF cmp PF_X, ORIG_W
+ vraddhn.u16 d30, q12, q10
+ vraddhn.u16 d31, q13, q11
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+ fetch_src_pixblock
+ PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+ vmvn.8 d22, d3
+ PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ PF subge PF_X, PF_X, ORIG_W
+ vmull.u8 q8, d22, d4
+ PF subges PF_CTL, PF_CTL, #0x10
+ vmull.u8 q9, d22, d5
+ PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+ vmull.u8 q10, d22, d6
+ PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
+ vmull.u8 q11, d22, d7
+.endm
+
+generate_composite_function \
+ pixman_composite_over_8888_8888_asm_neon, 32, 0, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_over_8888_8888_process_pixblock_head, \
+ pixman_composite_over_8888_8888_process_pixblock_tail, \
+ pixman_composite_over_8888_8888_process_pixblock_tail_head
+
+generate_composite_function_single_scanline \
+ pixman_composite_scanline_over_asm_neon, 32, 0, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_over_8888_8888_process_pixblock_head, \
+ pixman_composite_over_8888_8888_process_pixblock_tail, \
+ pixman_composite_over_8888_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_n_8888_process_pixblock_tail_head
+ pixman_composite_over_8888_8888_process_pixblock_tail
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ pixman_composite_over_8888_8888_process_pixblock_head
+ cache_preload 8, 8
+.endm
+
+.macro pixman_composite_over_n_8888_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vld1.32 {d3[0]}, [DUMMY]
+ vdup.8 d0, d3[0]
+ vdup.8 d1, d3[1]
+ vdup.8 d2, d3[2]
+ vdup.8 d3, d3[3]
+.endm
+
+generate_composite_function \
+ pixman_composite_over_n_8888_asm_neon, 0, 0, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_n_8888_init, \
+ default_cleanup, \
+ pixman_composite_over_8888_8888_process_pixblock_head, \
+ pixman_composite_over_8888_8888_process_pixblock_tail, \
+ pixman_composite_over_n_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_over_reverse_n_8888_process_pixblock_tail_head
+ vrshr.u16 q14, q8, #8
+ PF add PF_X, PF_X, #8
+ PF tst PF_CTL, #0xF
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ PF addne PF_X, PF_X, #8
+ PF subne PF_CTL, PF_CTL, #1
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ PF cmp PF_X, ORIG_W
+ vraddhn.u16 d30, q12, q10
+ vraddhn.u16 d31, q13, q11
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+ vld4.8 {d0, d1, d2, d3}, [DST_R, :128]!
+ vmvn.8 d22, d3
+ PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ PF subge PF_X, PF_X, ORIG_W
+ vmull.u8 q8, d22, d4
+ PF subges PF_CTL, PF_CTL, #0x10
+ vmull.u8 q9, d22, d5
+ vmull.u8 q10, d22, d6
+ PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
+ vmull.u8 q11, d22, d7
+.endm
+
+.macro pixman_composite_over_reverse_n_8888_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vld1.32 {d7[0]}, [DUMMY]
+ vdup.8 d4, d7[0]
+ vdup.8 d5, d7[1]
+ vdup.8 d6, d7[2]
+ vdup.8 d7, d7[3]
+.endm
+
+generate_composite_function \
+ pixman_composite_over_reverse_n_8888_asm_neon, 0, 0, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_reverse_n_8888_init, \
+ default_cleanup, \
+ pixman_composite_over_8888_8888_process_pixblock_head, \
+ pixman_composite_over_8888_8888_process_pixblock_tail, \
+ pixman_composite_over_reverse_n_8888_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 4, /* src_basereg */ \
+ 24 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_over_8888_8_0565_process_pixblock_head
+ vmull.u8 q0, d24, d8 /* IN for SRC pixels (part1) */
+ vmull.u8 q1, d24, d9
+ vmull.u8 q6, d24, d10
+ vmull.u8 q7, d24, d11
+ vshrn.u16 d6, q2, #8 /* convert DST_R data to 32-bpp (part1) */
+ vshrn.u16 d7, q2, #3
+ vsli.u16 q2, q2, #5
+ vrshr.u16 q8, q0, #8 /* IN for SRC pixels (part2) */
+ vrshr.u16 q9, q1, #8
+ vrshr.u16 q10, q6, #8
+ vrshr.u16 q11, q7, #8
+ vraddhn.u16 d0, q0, q8
+ vraddhn.u16 d1, q1, q9
+ vraddhn.u16 d2, q6, q10
+ vraddhn.u16 d3, q7, q11
+ vsri.u8 d6, d6, #5 /* convert DST_R data to 32-bpp (part2) */
+ vsri.u8 d7, d7, #6
+ vmvn.8 d3, d3
+ vshrn.u16 d30, q2, #2
+ vmull.u8 q8, d3, d6 /* now do alpha blending */
+ vmull.u8 q9, d3, d7
+ vmull.u8 q10, d3, d30
+.endm
+
+.macro pixman_composite_over_8888_8_0565_process_pixblock_tail
+ /* 3 cycle bubble (after vmull.u8) */
+ vrshr.u16 q13, q8, #8
+ vrshr.u16 q11, q9, #8
+ vrshr.u16 q15, q10, #8
+ vraddhn.u16 d16, q8, q13
+ vraddhn.u16 d27, q9, q11
+ vraddhn.u16 d26, q10, q15
+ vqadd.u8 d16, d2, d16
+ /* 1 cycle bubble */
+ vqadd.u8 q9, q0, q13
+ vshll.u8 q14, d16, #8 /* convert to 16bpp */
+ vshll.u8 q8, d19, #8
+ vshll.u8 q9, d18, #8
+ vsri.u16 q14, q8, #5
+ /* 1 cycle bubble */
+ vsri.u16 q14, q9, #11
+.endm
+
+.macro pixman_composite_over_8888_8_0565_process_pixblock_tail_head
+ vld1.16 {d4, d5}, [DST_R, :128]!
+ vshrn.u16 d6, q2, #8
+ fetch_mask_pixblock
+ vshrn.u16 d7, q2, #3
+ fetch_src_pixblock
+ vmull.u8 q6, d24, d10
+ vrshr.u16 q13, q8, #8
+ vrshr.u16 q11, q9, #8
+ vrshr.u16 q15, q10, #8
+ vraddhn.u16 d16, q8, q13
+ vraddhn.u16 d27, q9, q11
+ vraddhn.u16 d26, q10, q15
+ vqadd.u8 d16, d2, d16
+ vmull.u8 q1, d24, d9
+ vqadd.u8 q9, q0, q13
+ vshll.u8 q14, d16, #8
+ vmull.u8 q0, d24, d8
+ vshll.u8 q8, d19, #8
+ vshll.u8 q9, d18, #8
+ vsri.u16 q14, q8, #5
+ vmull.u8 q7, d24, d11
+ vsri.u16 q14, q9, #11
+
+ cache_preload 8, 8
+
+ vsli.u16 q2, q2, #5
+ vrshr.u16 q8, q0, #8
+ vrshr.u16 q9, q1, #8
+ vrshr.u16 q10, q6, #8
+ vrshr.u16 q11, q7, #8
+ vraddhn.u16 d0, q0, q8
+ vraddhn.u16 d1, q1, q9
+ vraddhn.u16 d2, q6, q10
+ vraddhn.u16 d3, q7, q11
+ vsri.u8 d6, d6, #5
+ vsri.u8 d7, d7, #6
+ vmvn.8 d3, d3
+ vshrn.u16 d30, q2, #2
+ vst1.16 {d28, d29}, [DST_W, :128]!
+ vmull.u8 q8, d3, d6
+ vmull.u8 q9, d3, d7
+ vmull.u8 q10, d3, d30
+.endm
+
+generate_composite_function \
+ pixman_composite_over_8888_8_0565_asm_neon, 32, 8, 16, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_over_8888_8_0565_process_pixblock_head, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 8, /* src_basereg */ \
+ 24 /* mask_basereg */
+
+/******************************************************************************/
+
+/*
+ * This function needs a special initialization of solid mask.
+ * Solid source pixel data is fetched from stack at ARGS_STACK_OFFSET
+ * offset, split into color components and replicated in d8-d11
+ * registers. Additionally, this function needs all the NEON registers,
+ * so it has to save d8-d15 registers which are callee saved according
+ * to ABI. These registers are restored from 'cleanup' macro. All the
+ * other NEON registers are caller saved, so can be clobbered freely
+ * without introducing any problems.
+ */
+.macro pixman_composite_over_n_8_0565_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vpush {d8-d15}
+ vld1.32 {d11[0]}, [DUMMY]
+ vdup.8 d8, d11[0]
+ vdup.8 d9, d11[1]
+ vdup.8 d10, d11[2]
+ vdup.8 d11, d11[3]
+.endm
+
+.macro pixman_composite_over_n_8_0565_cleanup
+ vpop {d8-d15}
+.endm
+
+generate_composite_function \
+ pixman_composite_over_n_8_0565_asm_neon, 0, 8, 16, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_n_8_0565_init, \
+ pixman_composite_over_n_8_0565_cleanup, \
+ pixman_composite_over_8888_8_0565_process_pixblock_head, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_over_8888_n_0565_init
+ add DUMMY, sp, #(ARGS_STACK_OFFSET + 8)
+ vpush {d8-d15}
+ vld1.32 {d24[0]}, [DUMMY]
+ vdup.8 d24, d24[3]
+.endm
+
+.macro pixman_composite_over_8888_n_0565_cleanup
+ vpop {d8-d15}
+.endm
+
+generate_composite_function \
+ pixman_composite_over_8888_n_0565_asm_neon, 32, 0, 16, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_8888_n_0565_init, \
+ pixman_composite_over_8888_n_0565_cleanup, \
+ pixman_composite_over_8888_8_0565_process_pixblock_head, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 8, /* src_basereg */ \
+ 24 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_0565_0565_process_pixblock_head
+.endm
+
+.macro pixman_composite_src_0565_0565_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_0565_0565_process_pixblock_tail_head
+ vst1.16 {d0, d1, d2, d3}, [DST_W, :128]!
+ fetch_src_pixblock
+ cache_preload 16, 16
+.endm
+
+generate_composite_function \
+ pixman_composite_src_0565_0565_asm_neon, 16, 0, 16, \
+ FLAG_DST_WRITEONLY, \
+ 16, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_0565_0565_process_pixblock_head, \
+ pixman_composite_src_0565_0565_process_pixblock_tail, \
+ pixman_composite_src_0565_0565_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_n_8_process_pixblock_head
+.endm
+
+.macro pixman_composite_src_n_8_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_n_8_process_pixblock_tail_head
+ vst1.8 {d0, d1, d2, d3}, [DST_W, :128]!
+.endm
+
+.macro pixman_composite_src_n_8_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vld1.32 {d0[0]}, [DUMMY]
+ vsli.u64 d0, d0, #8
+ vsli.u64 d0, d0, #16
+ vsli.u64 d0, d0, #32
+ vorr d1, d0, d0
+ vorr q1, q0, q0
+.endm
+
+.macro pixman_composite_src_n_8_cleanup
+.endm
+
+generate_composite_function \
+ pixman_composite_src_n_8_asm_neon, 0, 0, 8, \
+ FLAG_DST_WRITEONLY, \
+ 32, /* number of pixels, processed in a single block */ \
+ 0, /* prefetch distance */ \
+ pixman_composite_src_n_8_init, \
+ pixman_composite_src_n_8_cleanup, \
+ pixman_composite_src_n_8_process_pixblock_head, \
+ pixman_composite_src_n_8_process_pixblock_tail, \
+ pixman_composite_src_n_8_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_n_0565_process_pixblock_head
+.endm
+
+.macro pixman_composite_src_n_0565_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_n_0565_process_pixblock_tail_head
+ vst1.16 {d0, d1, d2, d3}, [DST_W, :128]!
+.endm
+
+.macro pixman_composite_src_n_0565_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vld1.32 {d0[0]}, [DUMMY]
+ vsli.u64 d0, d0, #16
+ vsli.u64 d0, d0, #32
+ vorr d1, d0, d0
+ vorr q1, q0, q0
+.endm
+
+.macro pixman_composite_src_n_0565_cleanup
+.endm
+
+generate_composite_function \
+ pixman_composite_src_n_0565_asm_neon, 0, 0, 16, \
+ FLAG_DST_WRITEONLY, \
+ 16, /* number of pixels, processed in a single block */ \
+ 0, /* prefetch distance */ \
+ pixman_composite_src_n_0565_init, \
+ pixman_composite_src_n_0565_cleanup, \
+ pixman_composite_src_n_0565_process_pixblock_head, \
+ pixman_composite_src_n_0565_process_pixblock_tail, \
+ pixman_composite_src_n_0565_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_n_8888_process_pixblock_head
+.endm
+
+.macro pixman_composite_src_n_8888_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_n_8888_process_pixblock_tail_head
+ vst1.32 {d0, d1, d2, d3}, [DST_W, :128]!
+.endm
+
+.macro pixman_composite_src_n_8888_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vld1.32 {d0[0]}, [DUMMY]
+ vsli.u64 d0, d0, #32
+ vorr d1, d0, d0
+ vorr q1, q0, q0
+.endm
+
+.macro pixman_composite_src_n_8888_cleanup
+.endm
+
+generate_composite_function \
+ pixman_composite_src_n_8888_asm_neon, 0, 0, 32, \
+ FLAG_DST_WRITEONLY, \
+ 8, /* number of pixels, processed in a single block */ \
+ 0, /* prefetch distance */ \
+ pixman_composite_src_n_8888_init, \
+ pixman_composite_src_n_8888_cleanup, \
+ pixman_composite_src_n_8888_process_pixblock_head, \
+ pixman_composite_src_n_8888_process_pixblock_tail, \
+ pixman_composite_src_n_8888_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_8888_8888_process_pixblock_head
+.endm
+
+.macro pixman_composite_src_8888_8888_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_8888_8888_process_pixblock_tail_head
+ vst1.32 {d0, d1, d2, d3}, [DST_W, :128]!
+ fetch_src_pixblock
+ cache_preload 8, 8
+.endm
+
+generate_composite_function \
+ pixman_composite_src_8888_8888_asm_neon, 32, 0, 32, \
+ FLAG_DST_WRITEONLY, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_8888_8888_process_pixblock_head, \
+ pixman_composite_src_8888_8888_process_pixblock_tail, \
+ pixman_composite_src_8888_8888_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_x888_8888_process_pixblock_head
+ vorr q0, q0, q2
+ vorr q1, q1, q2
+.endm
+
+.macro pixman_composite_src_x888_8888_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_x888_8888_process_pixblock_tail_head
+ vst1.32 {d0, d1, d2, d3}, [DST_W, :128]!
+ fetch_src_pixblock
+ vorr q0, q0, q2
+ vorr q1, q1, q2
+ cache_preload 8, 8
+.endm
+
+.macro pixman_composite_src_x888_8888_init
+ vmov.u8 q2, #0xFF
+ vshl.u32 q2, q2, #24
+.endm
+
+generate_composite_function \
+ pixman_composite_src_x888_8888_asm_neon, 32, 0, 32, \
+ FLAG_DST_WRITEONLY, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ pixman_composite_src_x888_8888_init, \
+ default_cleanup, \
+ pixman_composite_src_x888_8888_process_pixblock_head, \
+ pixman_composite_src_x888_8888_process_pixblock_tail, \
+ pixman_composite_src_x888_8888_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_over_n_8_8888_process_pixblock_head
+ /* expecting deinterleaved source data in {d8, d9, d10, d11} */
+ /* d8 - blue, d9 - green, d10 - red, d11 - alpha */
+ /* and destination data in {d4, d5, d6, d7} */
+ /* mask is in d24 (d25, d26, d27 are unused) */
+
+ /* in */
+ vmull.u8 q0, d24, d8
+ vmull.u8 q1, d24, d9
+ vmull.u8 q6, d24, d10
+ vmull.u8 q7, d24, d11
+ vrshr.u16 q10, q0, #8
+ vrshr.u16 q11, q1, #8
+ vrshr.u16 q12, q6, #8
+ vrshr.u16 q13, q7, #8
+ vraddhn.u16 d0, q0, q10
+ vraddhn.u16 d1, q1, q11
+ vraddhn.u16 d2, q6, q12
+ vraddhn.u16 d3, q7, q13
+ vmvn.8 d24, d3 /* get inverted alpha */
+ /* source: d0 - blue, d1 - green, d2 - red, d3 - alpha */
+ /* destination: d4 - blue, d5 - green, d6 - red, d7 - alpha */
+ /* now do alpha blending */
+ vmull.u8 q8, d24, d4
+ vmull.u8 q9, d24, d5
+ vmull.u8 q10, d24, d6
+ vmull.u8 q11, d24, d7
+.endm
+
+.macro pixman_composite_over_n_8_8888_process_pixblock_tail
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ vraddhn.u16 d30, q12, q10
+ vraddhn.u16 d31, q13, q11
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_n_8_8888_process_pixblock_tail_head
+ pixman_composite_over_n_8_8888_process_pixblock_tail
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ fetch_mask_pixblock
+ cache_preload 8, 8
+ pixman_composite_over_n_8_8888_process_pixblock_head
+.endm
+
+.macro pixman_composite_over_n_8_8888_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vpush {d8-d15}
+ vld1.32 {d11[0]}, [DUMMY]
+ vdup.8 d8, d11[0]
+ vdup.8 d9, d11[1]
+ vdup.8 d10, d11[2]
+ vdup.8 d11, d11[3]
+.endm
+
+.macro pixman_composite_over_n_8_8888_cleanup
+ vpop {d8-d15}
+.endm
+
+generate_composite_function \
+ pixman_composite_over_n_8_8888_asm_neon, 0, 8, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_n_8_8888_init, \
+ pixman_composite_over_n_8_8888_cleanup, \
+ pixman_composite_over_n_8_8888_process_pixblock_head, \
+ pixman_composite_over_n_8_8888_process_pixblock_tail, \
+ pixman_composite_over_n_8_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_over_n_8_8_process_pixblock_head
+ vmull.u8 q0, d24, d8
+ vmull.u8 q1, d25, d8
+ vmull.u8 q6, d26, d8
+ vmull.u8 q7, d27, d8
+ vrshr.u16 q10, q0, #8
+ vrshr.u16 q11, q1, #8
+ vrshr.u16 q12, q6, #8
+ vrshr.u16 q13, q7, #8
+ vraddhn.u16 d0, q0, q10
+ vraddhn.u16 d1, q1, q11
+ vraddhn.u16 d2, q6, q12
+ vraddhn.u16 d3, q7, q13
+ vmvn.8 q12, q0
+ vmvn.8 q13, q1
+ vmull.u8 q8, d24, d4
+ vmull.u8 q9, d25, d5
+ vmull.u8 q10, d26, d6
+ vmull.u8 q11, d27, d7
+.endm
+
+.macro pixman_composite_over_n_8_8_process_pixblock_tail
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ vraddhn.u16 d30, q12, q10
+ vraddhn.u16 d31, q13, q11
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_n_8_8_process_pixblock_tail_head
+ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ pixman_composite_over_n_8_8_process_pixblock_tail
+ fetch_mask_pixblock
+ cache_preload 32, 32
+ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ pixman_composite_over_n_8_8_process_pixblock_head
+.endm
+
+.macro pixman_composite_over_n_8_8_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vpush {d8-d15}
+ vld1.32 {d8[0]}, [DUMMY]
+ vdup.8 d8, d8[3]
+.endm
+
+.macro pixman_composite_over_n_8_8_cleanup
+ vpop {d8-d15}
+.endm
+
+generate_composite_function \
+ pixman_composite_over_n_8_8_asm_neon, 0, 8, 8, \
+ FLAG_DST_READWRITE, \
+ 32, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_n_8_8_init, \
+ pixman_composite_over_n_8_8_cleanup, \
+ pixman_composite_over_n_8_8_process_pixblock_head, \
+ pixman_composite_over_n_8_8_process_pixblock_tail, \
+ pixman_composite_over_n_8_8_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_over_n_8888_8888_ca_process_pixblock_head
+ /*
+ * 'combine_mask_ca' replacement
+ *
+ * input: solid src (n) in {d8, d9, d10, d11}
+ * dest in {d4, d5, d6, d7 }
+ * mask in {d24, d25, d26, d27}
+ * output: updated src in {d0, d1, d2, d3 }
+ * updated mask in {d24, d25, d26, d3 }
+ */
+ vmull.u8 q0, d24, d8
+ vmull.u8 q1, d25, d9
+ vmull.u8 q6, d26, d10
+ vmull.u8 q7, d27, d11
+ vmull.u8 q9, d11, d25
+ vmull.u8 q12, d11, d24
+ vmull.u8 q13, d11, d26
+ vrshr.u16 q8, q0, #8
+ vrshr.u16 q10, q1, #8
+ vrshr.u16 q11, q6, #8
+ vraddhn.u16 d0, q0, q8
+ vraddhn.u16 d1, q1, q10
+ vraddhn.u16 d2, q6, q11
+ vrshr.u16 q11, q12, #8
+ vrshr.u16 q8, q9, #8
+ vrshr.u16 q6, q13, #8
+ vrshr.u16 q10, q7, #8
+ vraddhn.u16 d24, q12, q11
+ vraddhn.u16 d25, q9, q8
+ vraddhn.u16 d26, q13, q6
+ vraddhn.u16 d3, q7, q10
+ /*
+ * 'combine_over_ca' replacement
+ *
+ * output: updated dest in {d28, d29, d30, d31}
+ */
+ vmvn.8 d24, d24
+ vmvn.8 d25, d25
+ vmull.u8 q8, d24, d4
+ vmull.u8 q9, d25, d5
+ vmvn.8 d26, d26
+ vmvn.8 d27, d3
+ vmull.u8 q10, d26, d6
+ vmull.u8 q11, d27, d7
+.endm
+
+.macro pixman_composite_over_n_8888_8888_ca_process_pixblock_tail
+ /* ... continue 'combine_over_ca' replacement */
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q6, q10, #8
+ vrshr.u16 q7, q11, #8
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ vraddhn.u16 d30, q6, q10
+ vraddhn.u16 d31, q7, q11
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+.endm
+
+.macro pixman_composite_over_n_8888_8888_ca_process_pixblock_tail_head
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ vrshr.u16 q6, q10, #8
+ vrshr.u16 q7, q11, #8
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ vraddhn.u16 d30, q6, q10
+ vraddhn.u16 d31, q7, q11
+ fetch_mask_pixblock
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+ cache_preload 8, 8
+ pixman_composite_over_n_8888_8888_ca_process_pixblock_head
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+.endm
+
+.macro pixman_composite_over_n_8888_8888_ca_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vpush {d8-d15}
+ vld1.32 {d11[0]}, [DUMMY]
+ vdup.8 d8, d11[0]
+ vdup.8 d9, d11[1]
+ vdup.8 d10, d11[2]
+ vdup.8 d11, d11[3]
+.endm
+
+.macro pixman_composite_over_n_8888_8888_ca_cleanup
+ vpop {d8-d15}
+.endm
+
+generate_composite_function \
+ pixman_composite_over_n_8888_8888_ca_asm_neon, 0, 32, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_n_8888_8888_ca_init, \
+ pixman_composite_over_n_8888_8888_ca_cleanup, \
+ pixman_composite_over_n_8888_8888_ca_process_pixblock_head, \
+ pixman_composite_over_n_8888_8888_ca_process_pixblock_tail, \
+ pixman_composite_over_n_8888_8888_ca_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_in_n_8_process_pixblock_head
+ /* expecting source data in {d0, d1, d2, d3} */
+ /* and destination data in {d4, d5, d6, d7} */
+ vmull.u8 q8, d4, d3
+ vmull.u8 q9, d5, d3
+ vmull.u8 q10, d6, d3
+ vmull.u8 q11, d7, d3
+.endm
+
+.macro pixman_composite_in_n_8_process_pixblock_tail
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ vraddhn.u16 d28, q8, q14
+ vraddhn.u16 d29, q9, q15
+ vraddhn.u16 d30, q10, q12
+ vraddhn.u16 d31, q11, q13
+.endm
+
+.macro pixman_composite_in_n_8_process_pixblock_tail_head
+ pixman_composite_in_n_8_process_pixblock_tail
+ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ cache_preload 32, 32
+ pixman_composite_in_n_8_process_pixblock_head
+ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
+.endm
+
+.macro pixman_composite_in_n_8_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vld1.32 {d3[0]}, [DUMMY]
+ vdup.8 d3, d3[3]
+.endm
+
+.macro pixman_composite_in_n_8_cleanup
+.endm
+
+generate_composite_function \
+ pixman_composite_in_n_8_asm_neon, 0, 0, 8, \
+ FLAG_DST_READWRITE, \
+ 32, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_in_n_8_init, \
+ pixman_composite_in_n_8_cleanup, \
+ pixman_composite_in_n_8_process_pixblock_head, \
+ pixman_composite_in_n_8_process_pixblock_tail, \
+ pixman_composite_in_n_8_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 24 /* mask_basereg */
+
+.macro pixman_composite_add_n_8_8_process_pixblock_head
+ /* expecting source data in {d8, d9, d10, d11} */
+ /* d8 - blue, d9 - green, d10 - red, d11 - alpha */
+ /* and destination data in {d4, d5, d6, d7} */
+ /* mask is in d24, d25, d26, d27 */
+ vmull.u8 q0, d24, d11
+ vmull.u8 q1, d25, d11
+ vmull.u8 q6, d26, d11
+ vmull.u8 q7, d27, d11
+ vrshr.u16 q10, q0, #8
+ vrshr.u16 q11, q1, #8
+ vrshr.u16 q12, q6, #8
+ vrshr.u16 q13, q7, #8
+ vraddhn.u16 d0, q0, q10
+ vraddhn.u16 d1, q1, q11
+ vraddhn.u16 d2, q6, q12
+ vraddhn.u16 d3, q7, q13
+ vqadd.u8 q14, q0, q2
+ vqadd.u8 q15, q1, q3
+.endm
+
+.macro pixman_composite_add_n_8_8_process_pixblock_tail
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_add_n_8_8_process_pixblock_tail_head
+ pixman_composite_add_n_8_8_process_pixblock_tail
+ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ fetch_mask_pixblock
+ cache_preload 32, 32
+ pixman_composite_add_n_8_8_process_pixblock_head
+.endm
+
+.macro pixman_composite_add_n_8_8_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vpush {d8-d15}
+ vld1.32 {d11[0]}, [DUMMY]
+ vdup.8 d11, d11[3]
+.endm
+
+.macro pixman_composite_add_n_8_8_cleanup
+ vpop {d8-d15}
+.endm
+
+generate_composite_function \
+ pixman_composite_add_n_8_8_asm_neon, 0, 8, 8, \
+ FLAG_DST_READWRITE, \
+ 32, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_add_n_8_8_init, \
+ pixman_composite_add_n_8_8_cleanup, \
+ pixman_composite_add_n_8_8_process_pixblock_head, \
+ pixman_composite_add_n_8_8_process_pixblock_tail, \
+ pixman_composite_add_n_8_8_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_add_8_8_8_process_pixblock_head
+ /* expecting source data in {d0, d1, d2, d3} */
+ /* destination data in {d4, d5, d6, d7} */
+ /* mask in {d24, d25, d26, d27} */
+ vmull.u8 q8, d24, d0
+ vmull.u8 q9, d25, d1
+ vmull.u8 q10, d26, d2
+ vmull.u8 q11, d27, d3
+ vrshr.u16 q0, q8, #8
+ vrshr.u16 q1, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ vraddhn.u16 d0, q0, q8
+ vraddhn.u16 d1, q1, q9
+ vraddhn.u16 d2, q12, q10
+ vraddhn.u16 d3, q13, q11
+ vqadd.u8 q14, q0, q2
+ vqadd.u8 q15, q1, q3
+.endm
+
+.macro pixman_composite_add_8_8_8_process_pixblock_tail
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_add_8_8_8_process_pixblock_tail_head
+ pixman_composite_add_8_8_8_process_pixblock_tail
+ vst1.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ vld1.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ fetch_mask_pixblock
+ fetch_src_pixblock
+ cache_preload 32, 32
+ pixman_composite_add_8_8_8_process_pixblock_head
+.endm
+
+.macro pixman_composite_add_8_8_8_init
+.endm
+
+.macro pixman_composite_add_8_8_8_cleanup
+.endm
+
+generate_composite_function \
+ pixman_composite_add_8_8_8_asm_neon, 8, 8, 8, \
+ FLAG_DST_READWRITE, \
+ 32, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_add_8_8_8_init, \
+ pixman_composite_add_8_8_8_cleanup, \
+ pixman_composite_add_8_8_8_process_pixblock_head, \
+ pixman_composite_add_8_8_8_process_pixblock_tail, \
+ pixman_composite_add_8_8_8_process_pixblock_tail_head
+
+/******************************************************************************/
+
+.macro pixman_composite_add_8888_8888_8888_process_pixblock_head
+ /* expecting source data in {d0, d1, d2, d3} */
+ /* destination data in {d4, d5, d6, d7} */
+ /* mask in {d24, d25, d26, d27} */
+ vmull.u8 q8, d27, d0
+ vmull.u8 q9, d27, d1
+ vmull.u8 q10, d27, d2
+ vmull.u8 q11, d27, d3
+ /* 1 cycle bubble */
+ vrsra.u16 q8, q8, #8
+ vrsra.u16 q9, q9, #8
+ vrsra.u16 q10, q10, #8
+ vrsra.u16 q11, q11, #8
+.endm
+
+.macro pixman_composite_add_8888_8888_8888_process_pixblock_tail
+ /* 2 cycle bubble */
+ vrshrn.u16 d28, q8, #8
+ vrshrn.u16 d29, q9, #8
+ vrshrn.u16 d30, q10, #8
+ vrshrn.u16 d31, q11, #8
+ vqadd.u8 q14, q2, q14
+ /* 1 cycle bubble */
+ vqadd.u8 q15, q3, q15
+.endm
+
+.macro pixman_composite_add_8888_8888_8888_process_pixblock_tail_head
+ fetch_src_pixblock
+ vrshrn.u16 d28, q8, #8
+ fetch_mask_pixblock
+ vrshrn.u16 d29, q9, #8
+ vmull.u8 q8, d27, d0
+ vrshrn.u16 d30, q10, #8
+ vmull.u8 q9, d27, d1
+ vrshrn.u16 d31, q11, #8
+ vmull.u8 q10, d27, d2
+ vqadd.u8 q14, q2, q14
+ vmull.u8 q11, d27, d3
+ vqadd.u8 q15, q3, q15
+ vrsra.u16 q8, q8, #8
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ vrsra.u16 q9, q9, #8
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ vrsra.u16 q10, q10, #8
+
+ cache_preload 8, 8
+
+ vrsra.u16 q11, q11, #8
+.endm
+
+generate_composite_function \
+ pixman_composite_add_8888_8888_8888_asm_neon, 32, 32, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_head, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail_head
+
+generate_composite_function_single_scanline \
+ pixman_composite_scanline_add_mask_asm_neon, 32, 32, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_head, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+generate_composite_function \
+ pixman_composite_add_8888_8_8888_asm_neon, 32, 8, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_head, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 27 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_add_n_8_8888_init
+ add DUMMY, sp, #ARGS_STACK_OFFSET
+ vld1.32 {d3[0]}, [DUMMY]
+ vdup.8 d0, d3[0]
+ vdup.8 d1, d3[1]
+ vdup.8 d2, d3[2]
+ vdup.8 d3, d3[3]
+.endm
+
+.macro pixman_composite_add_n_8_8888_cleanup
+.endm
+
+generate_composite_function \
+ pixman_composite_add_n_8_8888_asm_neon, 0, 8, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_add_n_8_8888_init, \
+ pixman_composite_add_n_8_8888_cleanup, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_head, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 27 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_add_8888_n_8888_init
+ add DUMMY, sp, #(ARGS_STACK_OFFSET + 8)
+ vld1.32 {d27[0]}, [DUMMY]
+ vdup.8 d27, d27[3]
+.endm
+
+.macro pixman_composite_add_8888_n_8888_cleanup
+.endm
+
+generate_composite_function \
+ pixman_composite_add_8888_n_8888_asm_neon, 32, 0, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_add_8888_n_8888_init, \
+ pixman_composite_add_8888_n_8888_cleanup, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_head, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail, \
+ pixman_composite_add_8888_8888_8888_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 27 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_out_reverse_8888_n_8888_process_pixblock_head
+ /* expecting source data in {d0, d1, d2, d3} */
+ /* destination data in {d4, d5, d6, d7} */
+ /* solid mask is in d15 */
+
+ /* 'in' */
+ vmull.u8 q8, d15, d3
+ vmull.u8 q6, d15, d2
+ vmull.u8 q5, d15, d1
+ vmull.u8 q4, d15, d0
+ vrshr.u16 q13, q8, #8
+ vrshr.u16 q12, q6, #8
+ vrshr.u16 q11, q5, #8
+ vrshr.u16 q10, q4, #8
+ vraddhn.u16 d3, q8, q13
+ vraddhn.u16 d2, q6, q12
+ vraddhn.u16 d1, q5, q11
+ vraddhn.u16 d0, q4, q10
+ vmvn.8 d24, d3 /* get inverted alpha */
+ /* now do alpha blending */
+ vmull.u8 q8, d24, d4
+ vmull.u8 q9, d24, d5
+ vmull.u8 q10, d24, d6
+ vmull.u8 q11, d24, d7
+.endm
+
+.macro pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vrshr.u16 q13, q11, #8
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ vraddhn.u16 d30, q12, q10
+ vraddhn.u16 d31, q13, q11
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_out_reverse_8888_8888_8888_process_pixblock_tail_head
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail
+ fetch_src_pixblock
+ cache_preload 8, 8
+ fetch_mask_pixblock
+ pixman_composite_out_reverse_8888_n_8888_process_pixblock_head
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+.endm
+
+generate_composite_function_single_scanline \
+ pixman_composite_scanline_out_reverse_mask_asm_neon, 32, 32, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_out_reverse_8888_n_8888_process_pixblock_head, \
+ pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail, \
+ pixman_composite_out_reverse_8888_8888_8888_process_pixblock_tail_head \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 12 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_over_8888_n_8888_process_pixblock_head
+ pixman_composite_out_reverse_8888_n_8888_process_pixblock_head
+.endm
+
+.macro pixman_composite_over_8888_n_8888_process_pixblock_tail
+ pixman_composite_out_reverse_8888_n_8888_process_pixblock_tail
+ vqadd.u8 q14, q0, q14
+ vqadd.u8 q15, q1, q15
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_8888_n_8888_process_pixblock_tail_head
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ pixman_composite_over_8888_n_8888_process_pixblock_tail
+ fetch_src_pixblock
+ cache_preload 8, 8
+ pixman_composite_over_8888_n_8888_process_pixblock_head
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+.endm
+
+.macro pixman_composite_over_8888_n_8888_init
+ add DUMMY, sp, #48
+ vpush {d8-d15}
+ vld1.32 {d15[0]}, [DUMMY]
+ vdup.8 d15, d15[3]
+.endm
+
+.macro pixman_composite_over_8888_n_8888_cleanup
+ vpop {d8-d15}
+.endm
+
+generate_composite_function \
+ pixman_composite_over_8888_n_8888_asm_neon, 32, 0, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_8888_n_8888_init, \
+ pixman_composite_over_8888_n_8888_cleanup, \
+ pixman_composite_over_8888_n_8888_process_pixblock_head, \
+ pixman_composite_over_8888_n_8888_process_pixblock_tail, \
+ pixman_composite_over_8888_n_8888_process_pixblock_tail_head
+
+/******************************************************************************/
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_8888_8888_8888_process_pixblock_tail_head
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ pixman_composite_over_8888_n_8888_process_pixblock_tail
+ fetch_src_pixblock
+ cache_preload 8, 8
+ fetch_mask_pixblock
+ pixman_composite_over_8888_n_8888_process_pixblock_head
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+.endm
+
+generate_composite_function \
+ pixman_composite_over_8888_8888_8888_asm_neon, 32, 32, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_over_8888_n_8888_process_pixblock_head, \
+ pixman_composite_over_8888_n_8888_process_pixblock_tail, \
+ pixman_composite_over_8888_8888_8888_process_pixblock_tail_head \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 12 /* mask_basereg */
+
+generate_composite_function_single_scanline \
+ pixman_composite_scanline_over_mask_asm_neon, 32, 32, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_over_8888_n_8888_process_pixblock_head, \
+ pixman_composite_over_8888_n_8888_process_pixblock_tail, \
+ pixman_composite_over_8888_8888_8888_process_pixblock_tail_head \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 12 /* mask_basereg */
+
+/******************************************************************************/
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_8888_8_8888_process_pixblock_tail_head
+ vld4.8 {d4, d5, d6, d7}, [DST_R, :128]!
+ pixman_composite_over_8888_n_8888_process_pixblock_tail
+ fetch_src_pixblock
+ cache_preload 8, 8
+ fetch_mask_pixblock
+ pixman_composite_over_8888_n_8888_process_pixblock_head
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+.endm
+
+generate_composite_function \
+ pixman_composite_over_8888_8_8888_asm_neon, 32, 8, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_over_8888_n_8888_process_pixblock_head, \
+ pixman_composite_over_8888_n_8888_process_pixblock_tail, \
+ pixman_composite_over_8888_8_8888_process_pixblock_tail_head \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 15 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_0888_0888_process_pixblock_head
+.endm
+
+.macro pixman_composite_src_0888_0888_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_0888_0888_process_pixblock_tail_head
+ vst3.8 {d0, d1, d2}, [DST_W]!
+ fetch_src_pixblock
+ cache_preload 8, 8
+.endm
+
+generate_composite_function \
+ pixman_composite_src_0888_0888_asm_neon, 24, 0, 24, \
+ FLAG_DST_WRITEONLY, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_0888_0888_process_pixblock_head, \
+ pixman_composite_src_0888_0888_process_pixblock_tail, \
+ pixman_composite_src_0888_0888_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_0888_8888_rev_process_pixblock_head
+ vswp d0, d2
+.endm
+
+.macro pixman_composite_src_0888_8888_rev_process_pixblock_tail
+.endm
+
+.macro pixman_composite_src_0888_8888_rev_process_pixblock_tail_head
+ vst4.8 {d0, d1, d2, d3}, [DST_W]!
+ fetch_src_pixblock
+ vswp d0, d2
+ cache_preload 8, 8
+.endm
+
+.macro pixman_composite_src_0888_8888_rev_init
+ veor d3, d3, d3
+.endm
+
+generate_composite_function \
+ pixman_composite_src_0888_8888_rev_asm_neon, 24, 0, 32, \
+ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ pixman_composite_src_0888_8888_rev_init, \
+ default_cleanup, \
+ pixman_composite_src_0888_8888_rev_process_pixblock_head, \
+ pixman_composite_src_0888_8888_rev_process_pixblock_tail, \
+ pixman_composite_src_0888_8888_rev_process_pixblock_tail_head, \
+ 0, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_0888_0565_rev_process_pixblock_head
+ vshll.u8 q8, d1, #8
+ vshll.u8 q9, d2, #8
+.endm
+
+.macro pixman_composite_src_0888_0565_rev_process_pixblock_tail
+ vshll.u8 q14, d0, #8
+ vsri.u16 q14, q8, #5
+ vsri.u16 q14, q9, #11
+.endm
+
+.macro pixman_composite_src_0888_0565_rev_process_pixblock_tail_head
+ vshll.u8 q14, d0, #8
+ fetch_src_pixblock
+ vsri.u16 q14, q8, #5
+ vsri.u16 q14, q9, #11
+ vshll.u8 q8, d1, #8
+ vst1.16 {d28, d29}, [DST_W, :128]!
+ vshll.u8 q9, d2, #8
+.endm
+
+generate_composite_function \
+ pixman_composite_src_0888_0565_rev_asm_neon, 24, 0, 16, \
+ FLAG_DST_WRITEONLY, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_0888_0565_rev_process_pixblock_head, \
+ pixman_composite_src_0888_0565_rev_process_pixblock_tail, \
+ pixman_composite_src_0888_0565_rev_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_pixbuf_8888_process_pixblock_head
+ vmull.u8 q8, d3, d0
+ vmull.u8 q9, d3, d1
+ vmull.u8 q10, d3, d2
+.endm
+
+.macro pixman_composite_src_pixbuf_8888_process_pixblock_tail
+ vrshr.u16 q11, q8, #8
+ vswp d3, d31
+ vrshr.u16 q12, q9, #8
+ vrshr.u16 q13, q10, #8
+ vraddhn.u16 d30, q11, q8
+ vraddhn.u16 d29, q12, q9
+ vraddhn.u16 d28, q13, q10
+.endm
+
+.macro pixman_composite_src_pixbuf_8888_process_pixblock_tail_head
+ vrshr.u16 q11, q8, #8
+ vswp d3, d31
+ vrshr.u16 q12, q9, #8
+ vrshr.u16 q13, q10, #8
+ fetch_src_pixblock
+ vraddhn.u16 d30, q11, q8
+ PF add PF_X, PF_X, #8
+ PF tst PF_CTL, #0xF
+ PF addne PF_X, PF_X, #8
+ PF subne PF_CTL, PF_CTL, #1
+ vraddhn.u16 d29, q12, q9
+ vraddhn.u16 d28, q13, q10
+ vmull.u8 q8, d3, d0
+ vmull.u8 q9, d3, d1
+ vmull.u8 q10, d3, d2
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ PF cmp PF_X, ORIG_W
+ PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+ PF subge PF_X, PF_X, ORIG_W
+ PF subges PF_CTL, PF_CTL, #0x10
+ PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+.endm
+
+generate_composite_function \
+ pixman_composite_src_pixbuf_8888_asm_neon, 32, 0, 32, \
+ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_pixbuf_8888_process_pixblock_head, \
+ pixman_composite_src_pixbuf_8888_process_pixblock_tail, \
+ pixman_composite_src_pixbuf_8888_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_src_rpixbuf_8888_process_pixblock_head
+ vmull.u8 q8, d3, d0
+ vmull.u8 q9, d3, d1
+ vmull.u8 q10, d3, d2
+.endm
+
+.macro pixman_composite_src_rpixbuf_8888_process_pixblock_tail
+ vrshr.u16 q11, q8, #8
+ vswp d3, d31
+ vrshr.u16 q12, q9, #8
+ vrshr.u16 q13, q10, #8
+ vraddhn.u16 d28, q11, q8
+ vraddhn.u16 d29, q12, q9
+ vraddhn.u16 d30, q13, q10
+.endm
+
+.macro pixman_composite_src_rpixbuf_8888_process_pixblock_tail_head
+ vrshr.u16 q11, q8, #8
+ vswp d3, d31
+ vrshr.u16 q12, q9, #8
+ vrshr.u16 q13, q10, #8
+ fetch_src_pixblock
+ vraddhn.u16 d28, q11, q8
+ PF add PF_X, PF_X, #8
+ PF tst PF_CTL, #0xF
+ PF addne PF_X, PF_X, #8
+ PF subne PF_CTL, PF_CTL, #1
+ vraddhn.u16 d29, q12, q9
+ vraddhn.u16 d30, q13, q10
+ vmull.u8 q8, d3, d0
+ vmull.u8 q9, d3, d1
+ vmull.u8 q10, d3, d2
+ vst4.8 {d28, d29, d30, d31}, [DST_W, :128]!
+ PF cmp PF_X, ORIG_W
+ PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+ PF subge PF_X, PF_X, ORIG_W
+ PF subges PF_CTL, PF_CTL, #0x10
+ PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+.endm
+
+generate_composite_function \
+ pixman_composite_src_rpixbuf_8888_asm_neon, 32, 0, 32, \
+ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ 10, /* prefetch distance */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_rpixbuf_8888_process_pixblock_head, \
+ pixman_composite_src_rpixbuf_8888_process_pixblock_tail, \
+ pixman_composite_src_rpixbuf_8888_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 0, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_over_0565_8_0565_process_pixblock_head
+ /* mask is in d15 */
+ convert_0565_to_x888 q4, d2, d1, d0
+ convert_0565_to_x888 q5, d6, d5, d4
+ /* source pixel data is in {d0, d1, d2, XX} */
+ /* destination pixel data is in {d4, d5, d6, XX} */
+ vmvn.8 d7, d15
+ vmull.u8 q6, d15, d2
+ vmull.u8 q5, d15, d1
+ vmull.u8 q4, d15, d0
+ vmull.u8 q8, d7, d4
+ vmull.u8 q9, d7, d5
+ vmull.u8 q13, d7, d6
+ vrshr.u16 q12, q6, #8
+ vrshr.u16 q11, q5, #8
+ vrshr.u16 q10, q4, #8
+ vraddhn.u16 d2, q6, q12
+ vraddhn.u16 d1, q5, q11
+ vraddhn.u16 d0, q4, q10
+.endm
+
+.macro pixman_composite_over_0565_8_0565_process_pixblock_tail
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q13, #8
+ vraddhn.u16 d28, q14, q8
+ vraddhn.u16 d29, q15, q9
+ vraddhn.u16 d30, q12, q13
+ vqadd.u8 q0, q0, q14
+ vqadd.u8 q1, q1, q15
+ /* 32bpp result is in {d0, d1, d2, XX} */
+ convert_8888_to_0565 d2, d1, d0, q14, q15, q3
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_over_0565_8_0565_process_pixblock_tail_head
+ fetch_mask_pixblock
+ pixman_composite_over_0565_8_0565_process_pixblock_tail
+ fetch_src_pixblock
+ vld1.16 {d10, d11}, [DST_R, :128]!
+ cache_preload 8, 8
+ pixman_composite_over_0565_8_0565_process_pixblock_head
+ vst1.16 {d28, d29}, [DST_W, :128]!
+.endm
+
+generate_composite_function \
+ pixman_composite_over_0565_8_0565_asm_neon, 16, 8, 16, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_over_0565_8_0565_process_pixblock_head, \
+ pixman_composite_over_0565_8_0565_process_pixblock_tail, \
+ pixman_composite_over_0565_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 10, /* dst_r_basereg */ \
+ 8, /* src_basereg */ \
+ 15 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_over_0565_n_0565_init
+ add DUMMY, sp, #(ARGS_STACK_OFFSET + 8)
+ vpush {d8-d15}
+ vld1.32 {d15[0]}, [DUMMY]
+ vdup.8 d15, d15[3]
+.endm
+
+.macro pixman_composite_over_0565_n_0565_cleanup
+ vpop {d8-d15}
+.endm
+
+generate_composite_function \
+ pixman_composite_over_0565_n_0565_asm_neon, 16, 0, 16, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ pixman_composite_over_0565_n_0565_init, \
+ pixman_composite_over_0565_n_0565_cleanup, \
+ pixman_composite_over_0565_8_0565_process_pixblock_head, \
+ pixman_composite_over_0565_8_0565_process_pixblock_tail, \
+ pixman_composite_over_0565_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 10, /* dst_r_basereg */ \
+ 8, /* src_basereg */ \
+ 15 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_add_0565_8_0565_process_pixblock_head
+ /* mask is in d15 */
+ convert_0565_to_x888 q4, d2, d1, d0
+ convert_0565_to_x888 q5, d6, d5, d4
+ /* source pixel data is in {d0, d1, d2, XX} */
+ /* destination pixel data is in {d4, d5, d6, XX} */
+ vmull.u8 q6, d15, d2
+ vmull.u8 q5, d15, d1
+ vmull.u8 q4, d15, d0
+ vrshr.u16 q12, q6, #8
+ vrshr.u16 q11, q5, #8
+ vrshr.u16 q10, q4, #8
+ vraddhn.u16 d2, q6, q12
+ vraddhn.u16 d1, q5, q11
+ vraddhn.u16 d0, q4, q10
+.endm
+
+.macro pixman_composite_add_0565_8_0565_process_pixblock_tail
+ vqadd.u8 q0, q0, q2
+ vqadd.u8 q1, q1, q3
+ /* 32bpp result is in {d0, d1, d2, XX} */
+ convert_8888_to_0565 d2, d1, d0, q14, q15, q3
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_add_0565_8_0565_process_pixblock_tail_head
+ fetch_mask_pixblock
+ pixman_composite_add_0565_8_0565_process_pixblock_tail
+ fetch_src_pixblock
+ vld1.16 {d10, d11}, [DST_R, :128]!
+ cache_preload 8, 8
+ pixman_composite_add_0565_8_0565_process_pixblock_head
+ vst1.16 {d28, d29}, [DST_W, :128]!
+.endm
+
+generate_composite_function \
+ pixman_composite_add_0565_8_0565_asm_neon, 16, 8, 16, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_add_0565_8_0565_process_pixblock_head, \
+ pixman_composite_add_0565_8_0565_process_pixblock_tail, \
+ pixman_composite_add_0565_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 10, /* dst_r_basereg */ \
+ 8, /* src_basereg */ \
+ 15 /* mask_basereg */
+
+/******************************************************************************/
+
+.macro pixman_composite_out_reverse_8_0565_process_pixblock_head
+ /* mask is in d15 */
+ convert_0565_to_x888 q5, d6, d5, d4
+ /* destination pixel data is in {d4, d5, d6, xx} */
+ vmvn.8 d24, d15 /* get inverted alpha */
+ /* now do alpha blending */
+ vmull.u8 q8, d24, d4
+ vmull.u8 q9, d24, d5
+ vmull.u8 q10, d24, d6
+.endm
+
+.macro pixman_composite_out_reverse_8_0565_process_pixblock_tail
+ vrshr.u16 q14, q8, #8
+ vrshr.u16 q15, q9, #8
+ vrshr.u16 q12, q10, #8
+ vraddhn.u16 d0, q14, q8
+ vraddhn.u16 d1, q15, q9
+ vraddhn.u16 d2, q12, q10
+ /* 32bpp result is in {d0, d1, d2, XX} */
+ convert_8888_to_0565 d2, d1, d0, q14, q15, q3
+.endm
+
+/* TODO: expand macros and do better instructions scheduling */
+.macro pixman_composite_out_reverse_8_0565_process_pixblock_tail_head
+ fetch_src_pixblock
+ pixman_composite_out_reverse_8_0565_process_pixblock_tail
+ vld1.16 {d10, d11}, [DST_R, :128]!
+ cache_preload 8, 8
+ pixman_composite_out_reverse_8_0565_process_pixblock_head
+ vst1.16 {d28, d29}, [DST_W, :128]!
+.endm
+
+generate_composite_function \
+ pixman_composite_out_reverse_8_0565_asm_neon, 8, 0, 16, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ 5, /* prefetch distance */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_out_reverse_8_0565_process_pixblock_head, \
+ pixman_composite_out_reverse_8_0565_process_pixblock_tail, \
+ pixman_composite_out_reverse_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 10, /* dst_r_basereg */ \
+ 15, /* src_basereg */ \
+ 0 /* mask_basereg */
+
+/******************************************************************************/
+
+generate_composite_function_nearest_scanline \
+ pixman_scaled_nearest_scanline_8888_8888_OVER_asm_neon, 32, 0, 32, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_over_8888_8888_process_pixblock_head, \
+ pixman_composite_over_8888_8888_process_pixblock_tail, \
+ pixman_composite_over_8888_8888_process_pixblock_tail_head
+
+generate_composite_function_nearest_scanline \
+ pixman_scaled_nearest_scanline_8888_0565_OVER_asm_neon, 32, 0, 16, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_over_8888_0565_process_pixblock_head, \
+ pixman_composite_over_8888_0565_process_pixblock_tail, \
+ pixman_composite_over_8888_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 0, /* src_basereg */ \
+ 24 /* mask_basereg */
+
+generate_composite_function_nearest_scanline \
+ pixman_scaled_nearest_scanline_8888_0565_SRC_asm_neon, 32, 0, 16, \
+ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_8888_0565_process_pixblock_head, \
+ pixman_composite_src_8888_0565_process_pixblock_tail, \
+ pixman_composite_src_8888_0565_process_pixblock_tail_head
+
+generate_composite_function_nearest_scanline \
+ pixman_scaled_nearest_scanline_0565_8888_SRC_asm_neon, 16, 0, 32, \
+ FLAG_DST_WRITEONLY | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init, \
+ default_cleanup, \
+ pixman_composite_src_0565_8888_process_pixblock_head, \
+ pixman_composite_src_0565_8888_process_pixblock_tail, \
+ pixman_composite_src_0565_8888_process_pixblock_tail_head
+
+generate_composite_function_nearest_scanline \
+ pixman_scaled_nearest_scanline_8888_8_0565_OVER_asm_neon, 32, 8, 16, \
+ FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_over_8888_8_0565_process_pixblock_head, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail, \
+ pixman_composite_over_8888_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 4, /* dst_r_basereg */ \
+ 8, /* src_basereg */ \
+ 24 /* mask_basereg */
+
+generate_composite_function_nearest_scanline \
+ pixman_scaled_nearest_scanline_0565_8_0565_OVER_asm_neon, 16, 8, 16, \
+ FLAG_DST_READWRITE, \
+ 8, /* number of pixels, processed in a single block */ \
+ default_init_need_all_regs, \
+ default_cleanup_need_all_regs, \
+ pixman_composite_over_0565_8_0565_process_pixblock_head, \
+ pixman_composite_over_0565_8_0565_process_pixblock_tail, \
+ pixman_composite_over_0565_8_0565_process_pixblock_tail_head, \
+ 28, /* dst_w_basereg */ \
+ 10, /* dst_r_basereg */ \
+ 8, /* src_basereg */ \
+ 15 /* mask_basereg */
diff --git a/pixman/pixman/pixman-arm-neon.c b/pixman/pixman/pixman-arm-neon.c index 3e0c0d1c2..24a8b0886 100644 --- a/pixman/pixman/pixman-arm-neon.c +++ b/pixman/pixman/pixman-arm-neon.c @@ -1,441 +1,441 @@ -/* - * Copyright © 2009 ARM Ltd, Movial Creative Technologies Oy - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of ARM Ltd not be used in - * advertising or publicity pertaining to distribution of the software without - * specific, written prior permission. ARM Ltd makes no - * representations about the suitability of this software for any purpose. It - * is provided "as is" without express or implied warranty. - * - * THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS - * SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND - * FITNESS, IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY - * SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES - * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN - * AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING - * OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS - * SOFTWARE. - * - * Author: Ian Rickards (ian.rickards@arm.com) - * Author: Jonathan Morton (jonathan.morton@movial.com) - * Author: Markku Vire (markku.vire@movial.com) - * - */ - -#ifdef HAVE_CONFIG_H -#include <config.h> -#endif - -#include <string.h> -#include "pixman-private.h" -#include "pixman-arm-common.h" - -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_8888_8888, - uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_x888_8888, - uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0565_0565, - uint16_t, 1, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0888_0888, - uint8_t, 3, uint8_t, 3) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_8888_0565, - uint32_t, 1, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0565_8888, - uint16_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0888_8888_rev, - uint8_t, 3, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0888_0565_rev, - uint8_t, 3, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_pixbuf_8888, - uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_rpixbuf_8888, - uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, add_8_8, - uint8_t, 1, uint8_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, add_8888_8888, - uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, over_8888_0565, - uint32_t, 1, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, over_8888_8888, - uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, out_reverse_8_0565, - uint8_t, 1, uint16_t, 1) - -PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_n_0565, - uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_n_8888, - uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_reverse_n_8888, - uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_N_DST (0, neon, in_n_8, - uint8_t, 1) - -PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_0565, - uint8_t, 1, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_8888, - uint8_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8888_8888_ca, - uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_8, - uint8_t, 1, uint8_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, add_n_8_8, - uint8_t, 1, uint8_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, add_n_8_8888, - uint8_t, 1, uint32_t, 1) - -PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_8888_n_8888, - uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_8888_n_0565, - uint32_t, 1, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_0565_n_0565, - uint16_t, 1, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, add_8888_n_8888, - uint32_t, 1, uint32_t, 1) - -PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8_8_8, - uint8_t, 1, uint8_t, 1, uint8_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_0565_8_0565, - uint16_t, 1, uint8_t, 1, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8888_8_8888, - uint32_t, 1, uint8_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8888_8888_8888, - uint32_t, 1, uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_8888_8_8888, - uint32_t, 1, uint8_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_8888_8888_8888, - uint32_t, 1, uint32_t, 1, uint32_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_8888_8_0565, - uint32_t, 1, uint8_t, 1, uint16_t, 1) -PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_0565_8_0565, - uint16_t, 1, uint8_t, 1, uint16_t, 1) - -PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (neon, 8888_8888, OVER, - uint32_t, uint32_t) -PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (neon, 8888_0565, OVER, - uint32_t, uint16_t) -PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (neon, 8888_0565, SRC, - uint32_t, uint16_t) -PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (neon, 0565_8888, SRC, - uint16_t, uint32_t) - -PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_A8_DST (SKIP_ZERO_SRC, neon, 8888_8_0565, - OVER, uint32_t, uint16_t) -PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_A8_DST (SKIP_ZERO_SRC, neon, 0565_8_0565, - OVER, uint16_t, uint16_t) - -void -pixman_composite_src_n_8_asm_neon (int32_t w, - int32_t h, - uint8_t *dst, - int32_t dst_stride, - uint8_t src); - -void -pixman_composite_src_n_0565_asm_neon (int32_t w, - int32_t h, - uint16_t *dst, - int32_t dst_stride, - uint16_t src); - -void -pixman_composite_src_n_8888_asm_neon (int32_t w, - int32_t h, - uint32_t *dst, - int32_t dst_stride, - uint32_t src); - -static pixman_bool_t -pixman_fill_neon (uint32_t *bits, - int stride, - int bpp, - int x, - int y, - int width, - int height, - uint32_t _xor) -{ - /* stride is always multiple of 32bit units in pixman */ - uint32_t byte_stride = stride * sizeof(uint32_t); - - switch (bpp) - { - case 8: - pixman_composite_src_n_8_asm_neon ( - width, - height, - (uint8_t *)(((char *) bits) + y * byte_stride + x), - byte_stride, - _xor & 0xff); - return TRUE; - case 16: - pixman_composite_src_n_0565_asm_neon ( - width, - height, - (uint16_t *)(((char *) bits) + y * byte_stride + x * 2), - byte_stride / 2, - _xor & 0xffff); - return TRUE; - case 32: - pixman_composite_src_n_8888_asm_neon ( - width, - height, - (uint32_t *)(((char *) bits) + y * byte_stride + x * 4), - byte_stride / 4, - _xor); - return TRUE; - default: - return FALSE; - } -} - -static pixman_bool_t -pixman_blt_neon (uint32_t *src_bits, - uint32_t *dst_bits, - int src_stride, - int dst_stride, - int src_bpp, - int dst_bpp, - int src_x, - int src_y, - int dst_x, - int dst_y, - int width, - int height) -{ - if (src_bpp != dst_bpp) - return FALSE; - - switch (src_bpp) - { - case 16: - pixman_composite_src_0565_0565_asm_neon ( - width, height, - (uint16_t *)(((char *) dst_bits) + - dst_y * dst_stride * 4 + dst_x * 2), dst_stride * 2, - (uint16_t *)(((char *) src_bits) + - src_y * src_stride * 4 + src_x * 2), src_stride * 2); - return TRUE; - case 32: - pixman_composite_src_8888_8888_asm_neon ( - width, height, - (uint32_t *)(((char *) dst_bits) + - dst_y * dst_stride * 4 + dst_x * 4), dst_stride, - (uint32_t *)(((char *) src_bits) + - src_y * src_stride * 4 + src_x * 4), src_stride); - return TRUE; - default: - return FALSE; - } -} - -static const pixman_fast_path_t arm_neon_fast_paths[] = -{ - PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, r5g6b5, neon_composite_src_0565_0565), - PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, b5g6r5, neon_composite_src_0565_0565), - PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, r5g6b5, neon_composite_src_8888_0565), - PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, r5g6b5, neon_composite_src_8888_0565), - PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, b5g6r5, neon_composite_src_8888_0565), - PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, b5g6r5, neon_composite_src_8888_0565), - PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, a8r8g8b8, neon_composite_src_0565_8888), - PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, x8r8g8b8, neon_composite_src_0565_8888), - PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, a8b8g8r8, neon_composite_src_0565_8888), - PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, x8b8g8r8, neon_composite_src_0565_8888), - PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, x8r8g8b8, neon_composite_src_8888_8888), - PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, x8r8g8b8, neon_composite_src_8888_8888), - PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, x8b8g8r8, neon_composite_src_8888_8888), - PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, x8b8g8r8, neon_composite_src_8888_8888), - PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, a8r8g8b8, neon_composite_src_8888_8888), - PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, a8b8g8r8, neon_composite_src_8888_8888), - PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, a8r8g8b8, neon_composite_src_x888_8888), - PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, a8b8g8r8, neon_composite_src_x888_8888), - PIXMAN_STD_FAST_PATH (SRC, r8g8b8, null, r8g8b8, neon_composite_src_0888_0888), - PIXMAN_STD_FAST_PATH (SRC, b8g8r8, null, x8r8g8b8, neon_composite_src_0888_8888_rev), - PIXMAN_STD_FAST_PATH (SRC, b8g8r8, null, r5g6b5, neon_composite_src_0888_0565_rev), - PIXMAN_STD_FAST_PATH (SRC, pixbuf, pixbuf, a8r8g8b8, neon_composite_src_pixbuf_8888), - PIXMAN_STD_FAST_PATH (SRC, pixbuf, pixbuf, a8b8g8r8, neon_composite_src_rpixbuf_8888), - PIXMAN_STD_FAST_PATH (SRC, rpixbuf, rpixbuf, a8r8g8b8, neon_composite_src_rpixbuf_8888), - PIXMAN_STD_FAST_PATH (SRC, rpixbuf, rpixbuf, a8b8g8r8, neon_composite_src_pixbuf_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8, neon_composite_over_n_8_8), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, r5g6b5, neon_composite_over_n_8_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, b5g6r5, neon_composite_over_n_8_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8r8g8b8, neon_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8r8g8b8, neon_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8b8g8r8, neon_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8b8g8r8, neon_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, null, r5g6b5, neon_composite_over_n_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, null, a8r8g8b8, neon_composite_over_n_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, null, x8r8g8b8, neon_composite_over_n_8888), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, a8r8g8b8, neon_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, x8r8g8b8, neon_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, a8b8g8r8, neon_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, x8b8g8r8, neon_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, a8r8g8b8, neon_composite_over_8888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, x8r8g8b8, neon_composite_over_8888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, r5g6b5, neon_composite_over_8888_n_0565), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, solid, b5g6r5, neon_composite_over_8888_n_0565), - PIXMAN_STD_FAST_PATH (OVER, r5g6b5, solid, r5g6b5, neon_composite_over_0565_n_0565), - PIXMAN_STD_FAST_PATH (OVER, b5g6r5, solid, b5g6r5, neon_composite_over_0565_n_0565), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, a8r8g8b8, neon_composite_over_8888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, x8r8g8b8, neon_composite_over_8888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, a8b8g8r8, neon_composite_over_8888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, x8b8g8r8, neon_composite_over_8888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, r5g6b5, neon_composite_over_8888_8_0565), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, b5g6r5, neon_composite_over_8888_8_0565), - PIXMAN_STD_FAST_PATH (OVER, r5g6b5, a8, r5g6b5, neon_composite_over_0565_8_0565), - PIXMAN_STD_FAST_PATH (OVER, b5g6r5, a8, b5g6r5, neon_composite_over_0565_8_0565), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, a8r8g8b8, neon_composite_over_8888_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, r5g6b5, neon_composite_over_8888_0565), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, b5g6r5, neon_composite_over_8888_0565), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, a8r8g8b8, neon_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, x8r8g8b8, neon_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, a8b8g8r8, neon_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, x8b8g8r8, neon_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, null, a8r8g8b8, neon_composite_src_x888_8888), - PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, null, a8b8g8r8, neon_composite_src_x888_8888), - PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8, neon_composite_add_n_8_8), - PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8r8g8b8, neon_composite_add_n_8_8888), - PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8b8g8r8, neon_composite_add_n_8_8888), - PIXMAN_STD_FAST_PATH (ADD, a8, a8, a8, neon_composite_add_8_8_8), - PIXMAN_STD_FAST_PATH (ADD, r5g6b5, a8, r5g6b5, neon_composite_add_0565_8_0565), - PIXMAN_STD_FAST_PATH (ADD, b5g6r5, a8, b5g6r5, neon_composite_add_0565_8_0565), - PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, a8, a8r8g8b8, neon_composite_add_8888_8_8888), - PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, a8, a8b8g8r8, neon_composite_add_8888_8_8888), - PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, a8r8g8b8, a8r8g8b8, neon_composite_add_8888_8888_8888), - PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, solid, a8r8g8b8, neon_composite_add_8888_n_8888), - PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, solid, a8b8g8r8, neon_composite_add_8888_n_8888), - PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, neon_composite_add_8_8), - PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, neon_composite_add_8888_8888), - PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, null, a8b8g8r8, neon_composite_add_8888_8888), - PIXMAN_STD_FAST_PATH (IN, solid, null, a8, neon_composite_in_n_8), - PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8r8g8b8, neon_composite_over_reverse_n_8888), - PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8b8g8r8, neon_composite_over_reverse_n_8888), - PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8, null, r5g6b5, neon_composite_out_reverse_8_0565), - PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8, null, b5g6r5, neon_composite_out_reverse_8_0565), - - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, neon_8888_8888), - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8, neon_8888_8888), - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, neon_8888_8888), - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8, neon_8888_8888), - - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, r5g6b5, neon_8888_0565), - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, b5g6r5, neon_8888_0565), - - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, a8r8g8b8, r5g6b5, neon_8888_0565), - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, x8r8g8b8, r5g6b5, neon_8888_0565), - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, a8b8g8r8, b5g6r5, neon_8888_0565), - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, x8b8g8r8, b5g6r5, neon_8888_0565), - - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, b5g6r5, x8b8g8r8, neon_0565_8888), - PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, r5g6b5, x8r8g8b8, neon_0565_8888), - /* Note: NONE repeat is not supported yet */ - SIMPLE_NEAREST_FAST_PATH_COVER (SRC, r5g6b5, a8r8g8b8, neon_0565_8888), - SIMPLE_NEAREST_FAST_PATH_COVER (SRC, b5g6r5, a8b8g8r8, neon_0565_8888), - SIMPLE_NEAREST_FAST_PATH_PAD (SRC, r5g6b5, a8r8g8b8, neon_0565_8888), - SIMPLE_NEAREST_FAST_PATH_PAD (SRC, b5g6r5, a8b8g8r8, neon_0565_8888), - - PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, a8r8g8b8, r5g6b5, neon_8888_8_0565), - PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, a8b8g8r8, b5g6r5, neon_8888_8_0565), - - PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, r5g6b5, r5g6b5, neon_0565_8_0565), - PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, b5g6r5, b5g6r5, neon_0565_8_0565), - - { PIXMAN_OP_NONE }, -}; - -static pixman_bool_t -arm_neon_blt (pixman_implementation_t *imp, - uint32_t * src_bits, - uint32_t * dst_bits, - int src_stride, - int dst_stride, - int src_bpp, - int dst_bpp, - int src_x, - int src_y, - int dst_x, - int dst_y, - int width, - int height) -{ - if (!pixman_blt_neon ( - src_bits, dst_bits, src_stride, dst_stride, src_bpp, dst_bpp, - src_x, src_y, dst_x, dst_y, width, height)) - - { - return _pixman_implementation_blt ( - imp->delegate, - src_bits, dst_bits, src_stride, dst_stride, src_bpp, dst_bpp, - src_x, src_y, dst_x, dst_y, width, height); - } - - return TRUE; -} - -static pixman_bool_t -arm_neon_fill (pixman_implementation_t *imp, - uint32_t * bits, - int stride, - int bpp, - int x, - int y, - int width, - int height, - uint32_t xor) -{ - if (pixman_fill_neon (bits, stride, bpp, x, y, width, height, xor)) - return TRUE; - - return _pixman_implementation_fill ( - imp->delegate, bits, stride, bpp, x, y, width, height, xor); -} - -#define BIND_COMBINE_U(name) \ -void \ -pixman_composite_scanline_##name##_mask_asm_neon (int32_t w, \ - const uint32_t *dst, \ - const uint32_t *src, \ - const uint32_t *mask); \ - \ -void \ -pixman_composite_scanline_##name##_asm_neon (int32_t w, \ - const uint32_t *dst, \ - const uint32_t *src); \ - \ -static void \ -neon_combine_##name##_u (pixman_implementation_t *imp, \ - pixman_op_t op, \ - uint32_t * dest, \ - const uint32_t * src, \ - const uint32_t * mask, \ - int width) \ -{ \ - if (mask) \ - pixman_composite_scanline_##name##_mask_asm_neon (width, dest, \ - src, mask); \ - else \ - pixman_composite_scanline_##name##_asm_neon (width, dest, src); \ -} - -BIND_COMBINE_U (over) -BIND_COMBINE_U (add) -BIND_COMBINE_U (out_reverse) - -pixman_implementation_t * -_pixman_implementation_create_arm_neon (pixman_implementation_t *fallback) -{ - pixman_implementation_t *imp = - _pixman_implementation_create (fallback, arm_neon_fast_paths); - - imp->combine_32[PIXMAN_OP_OVER] = neon_combine_over_u; - imp->combine_32[PIXMAN_OP_ADD] = neon_combine_add_u; - imp->combine_32[PIXMAN_OP_OUT_REVERSE] = neon_combine_out_reverse_u; - - imp->blt = arm_neon_blt; - imp->fill = arm_neon_fill; - - return imp; -} +/*
+ * Copyright © 2009 ARM Ltd, Movial Creative Technologies Oy
+ *
+ * Permission to use, copy, modify, distribute, and sell this software and its
+ * documentation for any purpose is hereby granted without fee, provided that
+ * the above copyright notice appear in all copies and that both that
+ * copyright notice and this permission notice appear in supporting
+ * documentation, and that the name of ARM Ltd not be used in
+ * advertising or publicity pertaining to distribution of the software without
+ * specific, written prior permission. ARM Ltd makes no
+ * representations about the suitability of this software for any purpose. It
+ * is provided "as is" without express or implied warranty.
+ *
+ * THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS
+ * SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND
+ * FITNESS, IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN
+ * AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING
+ * OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
+ * SOFTWARE.
+ *
+ * Author: Ian Rickards (ian.rickards@arm.com)
+ * Author: Jonathan Morton (jonathan.morton@movial.com)
+ * Author: Markku Vire (markku.vire@movial.com)
+ *
+ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <string.h>
+#include "pixman-private.h"
+#include "pixman-arm-common.h"
+
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_8888_8888,
+ uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_x888_8888,
+ uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0565_0565,
+ uint16_t, 1, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0888_0888,
+ uint8_t, 3, uint8_t, 3)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_8888_0565,
+ uint32_t, 1, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0565_8888,
+ uint16_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0888_8888_rev,
+ uint8_t, 3, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_0888_0565_rev,
+ uint8_t, 3, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_pixbuf_8888,
+ uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, src_rpixbuf_8888,
+ uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, add_8_8,
+ uint8_t, 1, uint8_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, add_8888_8888,
+ uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, over_8888_0565,
+ uint32_t, 1, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, over_8888_8888,
+ uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_DST (neon, out_reverse_8_0565,
+ uint8_t, 1, uint16_t, 1)
+
+PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_n_0565,
+ uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_n_8888,
+ uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_DST (SKIP_ZERO_SRC, neon, over_reverse_n_8888,
+ uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_DST (0, neon, in_n_8,
+ uint8_t, 1)
+
+PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_0565,
+ uint8_t, 1, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_8888,
+ uint8_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8888_8888_ca,
+ uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, over_n_8_8,
+ uint8_t, 1, uint8_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, add_n_8_8,
+ uint8_t, 1, uint8_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_N_MASK_DST (SKIP_ZERO_SRC, neon, add_n_8_8888,
+ uint8_t, 1, uint32_t, 1)
+
+PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_8888_n_8888,
+ uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_8888_n_0565,
+ uint32_t, 1, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, over_0565_n_0565,
+ uint16_t, 1, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_N_DST (SKIP_ZERO_MASK, neon, add_8888_n_8888,
+ uint32_t, 1, uint32_t, 1)
+
+PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8_8_8,
+ uint8_t, 1, uint8_t, 1, uint8_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_0565_8_0565,
+ uint16_t, 1, uint8_t, 1, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8888_8_8888,
+ uint32_t, 1, uint8_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, add_8888_8888_8888,
+ uint32_t, 1, uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_8888_8_8888,
+ uint32_t, 1, uint8_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_8888_8888_8888,
+ uint32_t, 1, uint32_t, 1, uint32_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_8888_8_0565,
+ uint32_t, 1, uint8_t, 1, uint16_t, 1)
+PIXMAN_ARM_BIND_FAST_PATH_SRC_MASK_DST (neon, over_0565_8_0565,
+ uint16_t, 1, uint8_t, 1, uint16_t, 1)
+
+PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (neon, 8888_8888, OVER,
+ uint32_t, uint32_t)
+PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (neon, 8888_0565, OVER,
+ uint32_t, uint16_t)
+PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (neon, 8888_0565, SRC,
+ uint32_t, uint16_t)
+PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_DST (neon, 0565_8888, SRC,
+ uint16_t, uint32_t)
+
+PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_A8_DST (SKIP_ZERO_SRC, neon, 8888_8_0565,
+ OVER, uint32_t, uint16_t)
+PIXMAN_ARM_BIND_SCALED_NEAREST_SRC_A8_DST (SKIP_ZERO_SRC, neon, 0565_8_0565,
+ OVER, uint16_t, uint16_t)
+
+void
+pixman_composite_src_n_8_asm_neon (int32_t w,
+ int32_t h,
+ uint8_t *dst,
+ int32_t dst_stride,
+ uint8_t src);
+
+void
+pixman_composite_src_n_0565_asm_neon (int32_t w,
+ int32_t h,
+ uint16_t *dst,
+ int32_t dst_stride,
+ uint16_t src);
+
+void
+pixman_composite_src_n_8888_asm_neon (int32_t w,
+ int32_t h,
+ uint32_t *dst,
+ int32_t dst_stride,
+ uint32_t src);
+
+static pixman_bool_t
+pixman_fill_neon (uint32_t *bits,
+ int stride,
+ int bpp,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t _xor)
+{
+ /* stride is always multiple of 32bit units in pixman */
+ uint32_t byte_stride = stride * sizeof(uint32_t);
+
+ switch (bpp)
+ {
+ case 8:
+ pixman_composite_src_n_8_asm_neon (
+ width,
+ height,
+ (uint8_t *)(((char *) bits) + y * byte_stride + x),
+ byte_stride,
+ _xor & 0xff);
+ return TRUE;
+ case 16:
+ pixman_composite_src_n_0565_asm_neon (
+ width,
+ height,
+ (uint16_t *)(((char *) bits) + y * byte_stride + x * 2),
+ byte_stride / 2,
+ _xor & 0xffff);
+ return TRUE;
+ case 32:
+ pixman_composite_src_n_8888_asm_neon (
+ width,
+ height,
+ (uint32_t *)(((char *) bits) + y * byte_stride + x * 4),
+ byte_stride / 4,
+ _xor);
+ return TRUE;
+ default:
+ return FALSE;
+ }
+}
+
+static pixman_bool_t
+pixman_blt_neon (uint32_t *src_bits,
+ uint32_t *dst_bits,
+ int src_stride,
+ int dst_stride,
+ int src_bpp,
+ int dst_bpp,
+ int src_x,
+ int src_y,
+ int dst_x,
+ int dst_y,
+ int width,
+ int height)
+{
+ if (src_bpp != dst_bpp)
+ return FALSE;
+
+ switch (src_bpp)
+ {
+ case 16:
+ pixman_composite_src_0565_0565_asm_neon (
+ width, height,
+ (uint16_t *)(((char *) dst_bits) +
+ dst_y * dst_stride * 4 + dst_x * 2), dst_stride * 2,
+ (uint16_t *)(((char *) src_bits) +
+ src_y * src_stride * 4 + src_x * 2), src_stride * 2);
+ return TRUE;
+ case 32:
+ pixman_composite_src_8888_8888_asm_neon (
+ width, height,
+ (uint32_t *)(((char *) dst_bits) +
+ dst_y * dst_stride * 4 + dst_x * 4), dst_stride,
+ (uint32_t *)(((char *) src_bits) +
+ src_y * src_stride * 4 + src_x * 4), src_stride);
+ return TRUE;
+ default:
+ return FALSE;
+ }
+}
+
+static const pixman_fast_path_t arm_neon_fast_paths[] =
+{
+ PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, r5g6b5, neon_composite_src_0565_0565),
+ PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, b5g6r5, neon_composite_src_0565_0565),
+ PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, r5g6b5, neon_composite_src_8888_0565),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, r5g6b5, neon_composite_src_8888_0565),
+ PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, b5g6r5, neon_composite_src_8888_0565),
+ PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, b5g6r5, neon_composite_src_8888_0565),
+ PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, a8r8g8b8, neon_composite_src_0565_8888),
+ PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, x8r8g8b8, neon_composite_src_0565_8888),
+ PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, a8b8g8r8, neon_composite_src_0565_8888),
+ PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, x8b8g8r8, neon_composite_src_0565_8888),
+ PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, x8r8g8b8, neon_composite_src_8888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, x8r8g8b8, neon_composite_src_8888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, x8b8g8r8, neon_composite_src_8888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, x8b8g8r8, neon_composite_src_8888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, a8r8g8b8, neon_composite_src_8888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, a8b8g8r8, neon_composite_src_8888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, a8r8g8b8, neon_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, a8b8g8r8, neon_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, r8g8b8, null, r8g8b8, neon_composite_src_0888_0888),
+ PIXMAN_STD_FAST_PATH (SRC, b8g8r8, null, x8r8g8b8, neon_composite_src_0888_8888_rev),
+ PIXMAN_STD_FAST_PATH (SRC, b8g8r8, null, r5g6b5, neon_composite_src_0888_0565_rev),
+ PIXMAN_STD_FAST_PATH (SRC, pixbuf, pixbuf, a8r8g8b8, neon_composite_src_pixbuf_8888),
+ PIXMAN_STD_FAST_PATH (SRC, pixbuf, pixbuf, a8b8g8r8, neon_composite_src_rpixbuf_8888),
+ PIXMAN_STD_FAST_PATH (SRC, rpixbuf, rpixbuf, a8r8g8b8, neon_composite_src_rpixbuf_8888),
+ PIXMAN_STD_FAST_PATH (SRC, rpixbuf, rpixbuf, a8b8g8r8, neon_composite_src_pixbuf_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8, neon_composite_over_n_8_8),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, r5g6b5, neon_composite_over_n_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, b5g6r5, neon_composite_over_n_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8r8g8b8, neon_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8r8g8b8, neon_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8b8g8r8, neon_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8b8g8r8, neon_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, null, r5g6b5, neon_composite_over_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, null, a8r8g8b8, neon_composite_over_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, null, x8r8g8b8, neon_composite_over_n_8888),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, a8r8g8b8, neon_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, x8r8g8b8, neon_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, a8b8g8r8, neon_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, x8b8g8r8, neon_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, a8r8g8b8, neon_composite_over_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, x8r8g8b8, neon_composite_over_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, r5g6b5, neon_composite_over_8888_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, solid, b5g6r5, neon_composite_over_8888_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, r5g6b5, solid, r5g6b5, neon_composite_over_0565_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, b5g6r5, solid, b5g6r5, neon_composite_over_0565_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, a8r8g8b8, neon_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, x8r8g8b8, neon_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, a8b8g8r8, neon_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, x8b8g8r8, neon_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, r5g6b5, neon_composite_over_8888_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, b5g6r5, neon_composite_over_8888_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, r5g6b5, a8, r5g6b5, neon_composite_over_0565_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, b5g6r5, a8, b5g6r5, neon_composite_over_0565_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, a8r8g8b8, neon_composite_over_8888_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, r5g6b5, neon_composite_over_8888_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, b5g6r5, neon_composite_over_8888_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, a8r8g8b8, neon_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, x8r8g8b8, neon_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, a8b8g8r8, neon_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, x8b8g8r8, neon_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, null, a8r8g8b8, neon_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, null, a8b8g8r8, neon_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8, neon_composite_add_n_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8r8g8b8, neon_composite_add_n_8_8888),
+ PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8b8g8r8, neon_composite_add_n_8_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8, a8, a8, neon_composite_add_8_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, r5g6b5, a8, r5g6b5, neon_composite_add_0565_8_0565),
+ PIXMAN_STD_FAST_PATH (ADD, b5g6r5, a8, b5g6r5, neon_composite_add_0565_8_0565),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, a8, a8r8g8b8, neon_composite_add_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, a8, a8b8g8r8, neon_composite_add_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, a8r8g8b8, a8r8g8b8, neon_composite_add_8888_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, solid, a8r8g8b8, neon_composite_add_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, solid, a8b8g8r8, neon_composite_add_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, neon_composite_add_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, neon_composite_add_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, null, a8b8g8r8, neon_composite_add_8888_8888),
+ PIXMAN_STD_FAST_PATH (IN, solid, null, a8, neon_composite_in_n_8),
+ PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8r8g8b8, neon_composite_over_reverse_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8b8g8r8, neon_composite_over_reverse_n_8888),
+ PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8, null, r5g6b5, neon_composite_out_reverse_8_0565),
+ PIXMAN_STD_FAST_PATH (OUT_REVERSE, a8, null, b5g6r5, neon_composite_out_reverse_8_0565),
+
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, neon_8888_8888),
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8, neon_8888_8888),
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, neon_8888_8888),
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8, neon_8888_8888),
+
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, r5g6b5, neon_8888_0565),
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, b5g6r5, neon_8888_0565),
+
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, a8r8g8b8, r5g6b5, neon_8888_0565),
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, x8r8g8b8, r5g6b5, neon_8888_0565),
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, a8b8g8r8, b5g6r5, neon_8888_0565),
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, x8b8g8r8, b5g6r5, neon_8888_0565),
+
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, b5g6r5, x8b8g8r8, neon_0565_8888),
+ PIXMAN_ARM_SIMPLE_NEAREST_FAST_PATH (SRC, r5g6b5, x8r8g8b8, neon_0565_8888),
+ /* Note: NONE repeat is not supported yet */
+ SIMPLE_NEAREST_FAST_PATH_COVER (SRC, r5g6b5, a8r8g8b8, neon_0565_8888),
+ SIMPLE_NEAREST_FAST_PATH_COVER (SRC, b5g6r5, a8b8g8r8, neon_0565_8888),
+ SIMPLE_NEAREST_FAST_PATH_PAD (SRC, r5g6b5, a8r8g8b8, neon_0565_8888),
+ SIMPLE_NEAREST_FAST_PATH_PAD (SRC, b5g6r5, a8b8g8r8, neon_0565_8888),
+
+ PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, a8r8g8b8, r5g6b5, neon_8888_8_0565),
+ PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, a8b8g8r8, b5g6r5, neon_8888_8_0565),
+
+ PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, r5g6b5, r5g6b5, neon_0565_8_0565),
+ PIXMAN_ARM_SIMPLE_NEAREST_A8_MASK_FAST_PATH (OVER, b5g6r5, b5g6r5, neon_0565_8_0565),
+
+ { PIXMAN_OP_NONE },
+};
+
+static pixman_bool_t
+arm_neon_blt (pixman_implementation_t *imp,
+ uint32_t * src_bits,
+ uint32_t * dst_bits,
+ int src_stride,
+ int dst_stride,
+ int src_bpp,
+ int dst_bpp,
+ int src_x,
+ int src_y,
+ int dst_x,
+ int dst_y,
+ int width,
+ int height)
+{
+ if (!pixman_blt_neon (
+ src_bits, dst_bits, src_stride, dst_stride, src_bpp, dst_bpp,
+ src_x, src_y, dst_x, dst_y, width, height))
+
+ {
+ return _pixman_implementation_blt (
+ imp->delegate,
+ src_bits, dst_bits, src_stride, dst_stride, src_bpp, dst_bpp,
+ src_x, src_y, dst_x, dst_y, width, height);
+ }
+
+ return TRUE;
+}
+
+static pixman_bool_t
+arm_neon_fill (pixman_implementation_t *imp,
+ uint32_t * bits,
+ int stride,
+ int bpp,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t xor)
+{
+ if (pixman_fill_neon (bits, stride, bpp, x, y, width, height, xor))
+ return TRUE;
+
+ return _pixman_implementation_fill (
+ imp->delegate, bits, stride, bpp, x, y, width, height, xor);
+}
+
+#define BIND_COMBINE_U(name) \
+void \
+pixman_composite_scanline_##name##_mask_asm_neon (int32_t w, \
+ const uint32_t *dst, \
+ const uint32_t *src, \
+ const uint32_t *mask); \
+ \
+void \
+pixman_composite_scanline_##name##_asm_neon (int32_t w, \
+ const uint32_t *dst, \
+ const uint32_t *src); \
+ \
+static void \
+neon_combine_##name##_u (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ uint32_t * dest, \
+ const uint32_t * src, \
+ const uint32_t * mask, \
+ int width) \
+{ \
+ if (mask) \
+ pixman_composite_scanline_##name##_mask_asm_neon (width, dest, \
+ src, mask); \
+ else \
+ pixman_composite_scanline_##name##_asm_neon (width, dest, src); \
+}
+
+BIND_COMBINE_U (over)
+BIND_COMBINE_U (add)
+BIND_COMBINE_U (out_reverse)
+
+pixman_implementation_t *
+_pixman_implementation_create_arm_neon (pixman_implementation_t *fallback)
+{
+ pixman_implementation_t *imp =
+ _pixman_implementation_create (fallback, arm_neon_fast_paths);
+
+ imp->combine_32[PIXMAN_OP_OVER] = neon_combine_over_u;
+ imp->combine_32[PIXMAN_OP_ADD] = neon_combine_add_u;
+ imp->combine_32[PIXMAN_OP_OUT_REVERSE] = neon_combine_out_reverse_u;
+
+ imp->blt = arm_neon_blt;
+ imp->fill = arm_neon_fill;
+
+ return imp;
+}
diff --git a/pixman/pixman/pixman-compiler.h b/pixman/pixman/pixman-compiler.h index 8f6c787f6..ebbffc3f3 100644 --- a/pixman/pixman/pixman-compiler.h +++ b/pixman/pixman/pixman-compiler.h @@ -89,10 +89,22 @@ # define PIXMAN_GET_THREAD_LOCAL(name) \ (&name) -#elif defined(__MINGW32__) +#elif defined(__MINGW32__) && !defined(__WIN64) -# define _NO_W32_PSEUDO_MODIFIERS -# include <windows.h> +/* We can't include <windows.h> as it causes carious clashes with + * identifiers in pixman, sigh. So just declare the functions we need + * here. + */ +extern long __stdcall InterlockedCompareExchange(long volatile *, long, long); +#define InterlockedCompareExchangePointer(d,e,c) \ + (void *)InterlockedCompareExchange((long volatile *)(d),(long)(e),(long)(c)) +extern int __stdcall TlsAlloc (void); +extern void * __stdcall TlsGetValue (unsigned); +extern int __stdcall TlsSetValue (unsigned, void *); +extern void * __stdcall CreateMutexA(void *, int, char *); +extern int __stdcall CloseHandle(void *); +extern unsigned __stdcall WaitForSingleObject (void *, unsigned); +extern int __stdcall ReleaseMutex (void *); # define PIXMAN_DEFINE_THREAD_LOCAL(type, name) \ static volatile int tls_ ## name ## _initialized = 0; \ diff --git a/pixman/pixman/pixman-fast-path.c b/pixman/pixman/pixman-fast-path.c index 92f030871..5c0965d83 100644 --- a/pixman/pixman/pixman-fast-path.c +++ b/pixman/pixman/pixman-fast-path.c @@ -1,2228 +1,2228 @@ -/* -*- Mode: c; c-basic-offset: 4; tab-width: 8; indent-tabs-mode: t; -*- */ -/* - * Copyright © 2000 SuSE, Inc. - * Copyright © 2007 Red Hat, Inc. - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of SuSE not be used in advertising or - * publicity pertaining to distribution of the software without specific, - * written prior permission. SuSE makes no representations about the - * suitability of this software for any purpose. It is provided "as is" - * without express or implied warranty. - * - * SuSE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL - * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL SuSE - * BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES - * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION - * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN - * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. - * - * Author: Keith Packard, SuSE, Inc. - */ - -#ifdef HAVE_CONFIG_H -#include <config.h> -#endif -#include <string.h> -#include <stdlib.h> -#include "pixman-private.h" -#include "pixman-combine32.h" -#include "pixman-fast-path.h" - -static force_inline uint32_t -fetch_24 (uint8_t *a) -{ - if (((unsigned long)a) & 1) - { -#ifdef WORDS_BIGENDIAN - return (*a << 16) | (*(uint16_t *)(a + 1)); -#else - return *a | (*(uint16_t *)(a + 1) << 8); -#endif - } - else - { -#ifdef WORDS_BIGENDIAN - return (*(uint16_t *)a << 8) | *(a + 2); -#else - return *(uint16_t *)a | (*(a + 2) << 16); -#endif - } -} - -static force_inline void -store_24 (uint8_t *a, - uint32_t v) -{ - if (((unsigned long)a) & 1) - { -#ifdef WORDS_BIGENDIAN - *a = (uint8_t) (v >> 16); - *(uint16_t *)(a + 1) = (uint16_t) (v); -#else - *a = (uint8_t) (v); - *(uint16_t *)(a + 1) = (uint16_t) (v >> 8); -#endif - } - else - { -#ifdef WORDS_BIGENDIAN - *(uint16_t *)a = (uint16_t)(v >> 8); - *(a + 2) = (uint8_t)v; -#else - *(uint16_t *)a = (uint16_t)v; - *(a + 2) = (uint8_t)(v >> 16); -#endif - } -} - -static force_inline uint32_t -over (uint32_t src, - uint32_t dest) -{ - uint32_t a = ~src >> 24; - - UN8x4_MUL_UN8_ADD_UN8x4 (dest, a, src); - - return dest; -} - -static uint32_t -in (uint32_t x, - uint8_t y) -{ - uint16_t a = y; - - UN8x4_MUL_UN8 (x, a); - - return x; -} - -/* - * Naming convention: - * - * op_src_mask_dest - */ -static void -fast_composite_over_x888_8_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *src, *src_line; - uint32_t *dst, *dst_line; - uint8_t *mask, *mask_line; - int src_stride, mask_stride, dst_stride; - uint8_t m; - uint32_t s, d; - int32_t w; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - while (height--) - { - src = src_line; - src_line += src_stride; - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - - w = width; - while (w--) - { - m = *mask++; - if (m) - { - s = *src | 0xff000000; - - if (m == 0xff) - { - *dst = s; - } - else - { - d = in (s, m); - *dst = over (d, *dst); - } - } - src++; - dst++; - } - } -} - -static void -fast_composite_in_n_8_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dest_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint8_t *dst_line, *dst; - uint8_t *mask_line, *mask, m; - int dst_stride, mask_stride; - int32_t w; - uint16_t t; - - src = _pixman_image_get_solid (imp, src_image, dest_image->bits.format); - - srca = src >> 24; - - PIXMAN_IMAGE_GET_LINE (dest_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - if (srca == 0xff) - { - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - m = *mask++; - - if (m == 0) - *dst = 0; - else if (m != 0xff) - *dst = MUL_UN8 (m, *dst, t); - - dst++; - } - } - } - else - { - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - m = *mask++; - m = MUL_UN8 (m, srca, t); - - if (m == 0) - *dst = 0; - else if (m != 0xff) - *dst = MUL_UN8 (m, *dst, t); - - dst++; - } - } - } -} - -static void -fast_composite_in_8_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dest_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - uint8_t *src_line, *src; - int dst_stride, src_stride; - int32_t w; - uint8_t s; - uint16_t t; - - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint8_t, src_stride, src_line, 1); - PIXMAN_IMAGE_GET_LINE (dest_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - { - s = *src++; - - if (s == 0) - *dst = 0; - else if (s != 0xff) - *dst = MUL_UN8 (s, *dst, t); - - dst++; - } - } -} - -static void -fast_composite_over_n_8_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint32_t *dst_line, *dst, d; - uint8_t *mask_line, *mask, m; - int dst_stride, mask_stride; - int32_t w; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - srca = src >> 24; - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - m = *mask++; - if (m == 0xff) - { - if (srca == 0xff) - *dst = src; - else - *dst = over (src, *dst); - } - else if (m) - { - d = in (src, m); - *dst = over (d, *dst); - } - dst++; - } - } -} - -static void -fast_composite_add_n_8888_8888_ca (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, s; - uint32_t *dst_line, *dst, d; - uint32_t *mask_line, *mask, ma; - int dst_stride, mask_stride; - int32_t w; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - ma = *mask++; - - if (ma) - { - d = *dst; - s = src; - - UN8x4_MUL_UN8x4_ADD_UN8x4 (s, ma, d); - - *dst = s; - } - - dst++; - } - } -} - -static void -fast_composite_over_n_8888_8888_ca (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca, s; - uint32_t *dst_line, *dst, d; - uint32_t *mask_line, *mask, ma; - int dst_stride, mask_stride; - int32_t w; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - srca = src >> 24; - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - ma = *mask++; - if (ma == 0xffffffff) - { - if (srca == 0xff) - *dst = src; - else - *dst = over (src, *dst); - } - else if (ma) - { - d = *dst; - s = src; - - UN8x4_MUL_UN8x4 (s, ma); - UN8x4_MUL_UN8 (ma, srca); - ma = ~ma; - UN8x4_MUL_UN8x4_ADD_UN8x4 (d, ma, s); - - *dst = d; - } - - dst++; - } - } -} - -static void -fast_composite_over_n_8_0888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint8_t *dst_line, *dst; - uint32_t d; - uint8_t *mask_line, *mask, m; - int dst_stride, mask_stride; - int32_t w; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - srca = src >> 24; - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 3); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - m = *mask++; - if (m == 0xff) - { - if (srca == 0xff) - { - d = src; - } - else - { - d = fetch_24 (dst); - d = over (src, d); - } - store_24 (dst, d); - } - else if (m) - { - d = over (in (src, m), fetch_24 (dst)); - store_24 (dst, d); - } - dst += 3; - } - } -} - -static void -fast_composite_over_n_8_0565 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint16_t *dst_line, *dst; - uint32_t d; - uint8_t *mask_line, *mask, m; - int dst_stride, mask_stride; - int32_t w; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - srca = src >> 24; - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - m = *mask++; - if (m == 0xff) - { - if (srca == 0xff) - { - d = src; - } - else - { - d = *dst; - d = over (src, CONVERT_0565_TO_0888 (d)); - } - *dst = CONVERT_8888_TO_0565 (d); - } - else if (m) - { - d = *dst; - d = over (in (src, m), CONVERT_0565_TO_0888 (d)); - *dst = CONVERT_8888_TO_0565 (d); - } - dst++; - } - } -} - -static void -fast_composite_over_n_8888_0565_ca (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca, s; - uint16_t src16; - uint16_t *dst_line, *dst; - uint32_t d; - uint32_t *mask_line, *mask, ma; - int dst_stride, mask_stride; - int32_t w; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - srca = src >> 24; - if (src == 0) - return; - - src16 = CONVERT_8888_TO_0565 (src); - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - ma = *mask++; - if (ma == 0xffffffff) - { - if (srca == 0xff) - { - *dst = src16; - } - else - { - d = *dst; - d = over (src, CONVERT_0565_TO_0888 (d)); - *dst = CONVERT_8888_TO_0565 (d); - } - } - else if (ma) - { - d = *dst; - d = CONVERT_0565_TO_0888 (d); - - s = src; - - UN8x4_MUL_UN8x4 (s, ma); - UN8x4_MUL_UN8 (ma, srca); - ma = ~ma; - UN8x4_MUL_UN8x4_ADD_UN8x4 (d, ma, s); - - *dst = CONVERT_8888_TO_0565 (d); - } - dst++; - } - } -} - -static void -fast_composite_over_8888_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst; - uint32_t *src_line, *src, s; - int dst_stride, src_stride; - uint8_t a; - int32_t w; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - { - s = *src++; - a = s >> 24; - if (a == 0xff) - *dst = s; - else if (s) - *dst = over (s, *dst); - dst++; - } - } -} - -static void -fast_composite_src_x888_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst; - uint32_t *src_line, *src; - int dst_stride, src_stride; - int32_t w; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - *dst++ = (*src++) | 0xff000000; - } -} - -#if 0 -static void -fast_composite_over_8888_0888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - uint32_t d; - uint32_t *src_line, *src, s; - uint8_t a; - int dst_stride, src_stride; - int32_t w; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 3); - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - { - s = *src++; - a = s >> 24; - if (a) - { - if (a == 0xff) - d = s; - else - d = over (s, fetch_24 (dst)); - - store_24 (dst, d); - } - dst += 3; - } - } -} -#endif - -static void -fast_composite_over_8888_0565 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint16_t *dst_line, *dst; - uint32_t d; - uint32_t *src_line, *src, s; - uint8_t a; - int dst_stride, src_stride; - int32_t w; - - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - { - s = *src++; - a = s >> 24; - if (s) - { - if (a == 0xff) - { - d = s; - } - else - { - d = *dst; - d = over (s, CONVERT_0565_TO_0888 (d)); - } - *dst = CONVERT_8888_TO_0565 (d); - } - dst++; - } - } -} - -static void -fast_composite_src_x888_0565 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint16_t *dst_line, *dst; - uint32_t *src_line, *src, s; - int dst_stride, src_stride; - int32_t w; - - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - { - s = *src++; - *dst = CONVERT_8888_TO_0565 (s); - dst++; - } - } -} - -static void -fast_composite_add_8_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - uint8_t *src_line, *src; - int dst_stride, src_stride; - int32_t w; - uint8_t s, d; - uint16_t t; - - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint8_t, src_stride, src_line, 1); - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - { - s = *src++; - if (s) - { - if (s != 0xff) - { - d = *dst; - t = d + s; - s = t | (0 - (t >> 8)); - } - *dst = s; - } - dst++; - } - } -} - -static void -fast_composite_add_8888_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst; - uint32_t *src_line, *src; - int dst_stride, src_stride; - int32_t w; - uint32_t s, d; - - PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - { - s = *src++; - if (s) - { - if (s != 0xffffffff) - { - d = *dst; - if (d) - UN8x4_ADD_UN8x4 (s, d); - } - *dst = s; - } - dst++; - } - } -} - -static void -fast_composite_add_n_8_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - uint8_t *mask_line, *mask; - int dst_stride, mask_stride; - int32_t w; - uint32_t src; - uint8_t sa; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - sa = (src >> 24); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w--) - { - uint16_t tmp; - uint16_t a; - uint32_t m, d; - uint32_t r; - - a = *mask++; - d = *dst; - - m = MUL_UN8 (sa, a, tmp); - r = ADD_UN8 (m, d, tmp); - - *dst++ = r; - } - } -} - -#ifdef WORDS_BIGENDIAN -#define CREATE_BITMASK(n) (0x80000000 >> (n)) -#define UPDATE_BITMASK(n) ((n) >> 1) -#else -#define CREATE_BITMASK(n) (1 << (n)) -#define UPDATE_BITMASK(n) ((n) << 1) -#endif - -#define TEST_BIT(p, n) \ - (*((p) + ((n) >> 5)) & CREATE_BITMASK ((n) & 31)) -#define SET_BIT(p, n) \ - do { *((p) + ((n) >> 5)) |= CREATE_BITMASK ((n) & 31); } while (0); - -static void -fast_composite_add_1000_1000 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst; - uint32_t *src_line, *src; - int dst_stride, src_stride; - int32_t w; - - PIXMAN_IMAGE_GET_LINE (src_image, 0, src_y, uint32_t, - src_stride, src_line, 1); - PIXMAN_IMAGE_GET_LINE (dst_image, 0, dest_y, uint32_t, - dst_stride, dst_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w--) - { - /* - * TODO: improve performance by processing uint32_t data instead - * of individual bits - */ - if (TEST_BIT (src, src_x + w)) - SET_BIT (dst, dest_x + w); - } - } -} - -static void -fast_composite_over_n_1_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint32_t *dst, *dst_line; - uint32_t *mask, *mask_line; - int mask_stride, dst_stride; - uint32_t bitcache, bitmask; - int32_t w; - - if (width <= 0) - return; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - srca = src >> 24; - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, - dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, 0, mask_y, uint32_t, - mask_stride, mask_line, 1); - mask_line += mask_x >> 5; - - if (srca == 0xff) - { - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - bitcache = *mask++; - bitmask = CREATE_BITMASK (mask_x & 31); - - while (w--) - { - if (bitmask == 0) - { - bitcache = *mask++; - bitmask = CREATE_BITMASK (0); - } - if (bitcache & bitmask) - *dst = src; - bitmask = UPDATE_BITMASK (bitmask); - dst++; - } - } - } - else - { - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - bitcache = *mask++; - bitmask = CREATE_BITMASK (mask_x & 31); - - while (w--) - { - if (bitmask == 0) - { - bitcache = *mask++; - bitmask = CREATE_BITMASK (0); - } - if (bitcache & bitmask) - *dst = over (src, *dst); - bitmask = UPDATE_BITMASK (bitmask); - dst++; - } - } - } -} - -static void -fast_composite_over_n_1_0565 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint16_t *dst, *dst_line; - uint32_t *mask, *mask_line; - int mask_stride, dst_stride; - uint32_t bitcache, bitmask; - int32_t w; - uint32_t d; - uint16_t src565; - - if (width <= 0) - return; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - srca = src >> 24; - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, - dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE (mask_image, 0, mask_y, uint32_t, - mask_stride, mask_line, 1); - mask_line += mask_x >> 5; - - if (srca == 0xff) - { - src565 = CONVERT_8888_TO_0565 (src); - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - bitcache = *mask++; - bitmask = CREATE_BITMASK (mask_x & 31); - - while (w--) - { - if (bitmask == 0) - { - bitcache = *mask++; - bitmask = CREATE_BITMASK (0); - } - if (bitcache & bitmask) - *dst = src565; - bitmask = UPDATE_BITMASK (bitmask); - dst++; - } - } - } - else - { - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - bitcache = *mask++; - bitmask = CREATE_BITMASK (mask_x & 31); - - while (w--) - { - if (bitmask == 0) - { - bitcache = *mask++; - bitmask = CREATE_BITMASK (0); - } - if (bitcache & bitmask) - { - d = over (src, CONVERT_0565_TO_0888 (*dst)); - *dst = CONVERT_8888_TO_0565 (d); - } - bitmask = UPDATE_BITMASK (bitmask); - dst++; - } - } - } -} - -/* - * Simple bitblt - */ - -static void -fast_composite_solid_fill (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - if (dst_image->bits.format == PIXMAN_a1) - { - src = src >> 31; - } - else if (dst_image->bits.format == PIXMAN_a8) - { - src = src >> 24; - } - else if (dst_image->bits.format == PIXMAN_r5g6b5 || - dst_image->bits.format == PIXMAN_b5g6r5) - { - src = CONVERT_8888_TO_0565 (src); - } - - pixman_fill (dst_image->bits.bits, dst_image->bits.rowstride, - PIXMAN_FORMAT_BPP (dst_image->bits.format), - dest_x, dest_y, - width, height, - src); -} - -static void -fast_composite_src_memcpy (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - int bpp = PIXMAN_FORMAT_BPP (dst_image->bits.format) / 8; - uint32_t n_bytes = width * bpp; - int dst_stride, src_stride; - uint8_t *dst; - uint8_t *src; - - src_stride = src_image->bits.rowstride * 4; - dst_stride = dst_image->bits.rowstride * 4; - - src = (uint8_t *)src_image->bits.bits + src_y * src_stride + src_x * bpp; - dst = (uint8_t *)dst_image->bits.bits + dest_y * dst_stride + dest_x * bpp; - - while (height--) - { - memcpy (dst, src, n_bytes); - - dst += dst_stride; - src += src_stride; - } -} - -FAST_NEAREST (8888_8888_cover, 8888, 8888, uint32_t, uint32_t, SRC, COVER) -FAST_NEAREST (8888_8888_none, 8888, 8888, uint32_t, uint32_t, SRC, NONE) -FAST_NEAREST (8888_8888_pad, 8888, 8888, uint32_t, uint32_t, SRC, PAD) -FAST_NEAREST (8888_8888_normal, 8888, 8888, uint32_t, uint32_t, SRC, NORMAL) -FAST_NEAREST (8888_8888_cover, 8888, 8888, uint32_t, uint32_t, OVER, COVER) -FAST_NEAREST (8888_8888_none, 8888, 8888, uint32_t, uint32_t, OVER, NONE) -FAST_NEAREST (8888_8888_pad, 8888, 8888, uint32_t, uint32_t, OVER, PAD) -FAST_NEAREST (8888_8888_normal, 8888, 8888, uint32_t, uint32_t, OVER, NORMAL) -FAST_NEAREST (8888_565_cover, 8888, 0565, uint32_t, uint16_t, SRC, COVER) -FAST_NEAREST (8888_565_none, 8888, 0565, uint32_t, uint16_t, SRC, NONE) -FAST_NEAREST (8888_565_pad, 8888, 0565, uint32_t, uint16_t, SRC, PAD) -FAST_NEAREST (8888_565_normal, 8888, 0565, uint32_t, uint16_t, SRC, NORMAL) -FAST_NEAREST (565_565_normal, 0565, 0565, uint16_t, uint16_t, SRC, NORMAL) -FAST_NEAREST (8888_565_cover, 8888, 0565, uint32_t, uint16_t, OVER, COVER) -FAST_NEAREST (8888_565_none, 8888, 0565, uint32_t, uint16_t, OVER, NONE) -FAST_NEAREST (8888_565_pad, 8888, 0565, uint32_t, uint16_t, OVER, PAD) -FAST_NEAREST (8888_565_normal, 8888, 0565, uint32_t, uint16_t, OVER, NORMAL) - -/* Use more unrolling for src_0565_0565 because it is typically CPU bound */ -static force_inline void -scaled_nearest_scanline_565_565_SRC (uint16_t * dst, - const uint16_t * src, - int32_t w, - pixman_fixed_t vx, - pixman_fixed_t unit_x, - pixman_fixed_t max_vx, - pixman_bool_t fully_transparent_src) -{ - uint16_t tmp1, tmp2, tmp3, tmp4; - while ((w -= 4) >= 0) - { - tmp1 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - tmp2 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - tmp3 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - tmp4 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - *dst++ = tmp1; - *dst++ = tmp2; - *dst++ = tmp3; - *dst++ = tmp4; - } - if (w & 2) - { - tmp1 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - tmp2 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - *dst++ = tmp1; - *dst++ = tmp2; - } - if (w & 1) - *dst++ = src[pixman_fixed_to_int (vx)]; -} - -FAST_NEAREST_MAINLOOP (565_565_cover_SRC, - scaled_nearest_scanline_565_565_SRC, - uint16_t, uint16_t, COVER) -FAST_NEAREST_MAINLOOP (565_565_none_SRC, - scaled_nearest_scanline_565_565_SRC, - uint16_t, uint16_t, NONE) -FAST_NEAREST_MAINLOOP (565_565_pad_SRC, - scaled_nearest_scanline_565_565_SRC, - uint16_t, uint16_t, PAD) - -static force_inline uint32_t -fetch_nearest (pixman_repeat_t src_repeat, - pixman_format_code_t format, - uint32_t *src, int x, int src_width) -{ - if (repeat (src_repeat, &x, src_width)) - { - if (format == PIXMAN_x8r8g8b8) - return *(src + x) | 0xff000000; - else - return *(src + x); - } - else - { - return 0; - } -} - -static force_inline void -combine_over (uint32_t s, uint32_t *dst) -{ - if (s) - { - uint8_t ia = 0xff - (s >> 24); - - if (ia) - UN8x4_MUL_UN8_ADD_UN8x4 (*dst, ia, s); - else - *dst = s; - } -} - -static force_inline void -combine_src (uint32_t s, uint32_t *dst) -{ - *dst = s; -} - -static void -fast_composite_scaled_nearest (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line; - uint32_t *src_line; - int dst_stride, src_stride; - int src_width, src_height; - pixman_repeat_t src_repeat; - pixman_fixed_t unit_x, unit_y; - pixman_format_code_t src_format; - pixman_vector_t v; - pixman_fixed_t vy; - - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - /* pass in 0 instead of src_x and src_y because src_x and src_y need to be - * transformed from destination space to source space - */ - PIXMAN_IMAGE_GET_LINE (src_image, 0, 0, uint32_t, src_stride, src_line, 1); - - /* reference point is the center of the pixel */ - v.vector[0] = pixman_int_to_fixed (src_x) + pixman_fixed_1 / 2; - v.vector[1] = pixman_int_to_fixed (src_y) + pixman_fixed_1 / 2; - v.vector[2] = pixman_fixed_1; - - if (!pixman_transform_point_3d (src_image->common.transform, &v)) - return; - - unit_x = src_image->common.transform->matrix[0][0]; - unit_y = src_image->common.transform->matrix[1][1]; - - /* Round down to closest integer, ensuring that 0.5 rounds to 0, not 1 */ - v.vector[0] -= pixman_fixed_e; - v.vector[1] -= pixman_fixed_e; - - src_height = src_image->bits.height; - src_width = src_image->bits.width; - src_repeat = src_image->common.repeat; - src_format = src_image->bits.format; - - vy = v.vector[1]; - while (height--) - { - pixman_fixed_t vx = v.vector[0]; - int y = pixman_fixed_to_int (vy); - uint32_t *dst = dst_line; - - dst_line += dst_stride; - - /* adjust the y location by a unit vector in the y direction - * this is equivalent to transforming y+1 of the destination point to source space */ - vy += unit_y; - - if (!repeat (src_repeat, &y, src_height)) - { - if (op == PIXMAN_OP_SRC) - memset (dst, 0, sizeof (*dst) * width); - } - else - { - int w = width; - - uint32_t *src = src_line + y * src_stride; - - while (w >= 2) - { - uint32_t s1, s2; - int x1, x2; - - x1 = pixman_fixed_to_int (vx); - vx += unit_x; - - x2 = pixman_fixed_to_int (vx); - vx += unit_x; - - w -= 2; - - s1 = fetch_nearest (src_repeat, src_format, src, x1, src_width); - s2 = fetch_nearest (src_repeat, src_format, src, x2, src_width); - - if (op == PIXMAN_OP_OVER) - { - combine_over (s1, dst++); - combine_over (s2, dst++); - } - else - { - combine_src (s1, dst++); - combine_src (s2, dst++); - } - } - - while (w--) - { - uint32_t s; - int x; - - x = pixman_fixed_to_int (vx); - vx += unit_x; - - s = fetch_nearest (src_repeat, src_format, src, x, src_width); - - if (op == PIXMAN_OP_OVER) - combine_over (s, dst++); - else - combine_src (s, dst++); - } - } - } -} - -#define CACHE_LINE_SIZE 64 - -#define FAST_SIMPLE_ROTATE(suffix, pix_type) \ - \ -static void \ -blt_rotated_90_trivial_##suffix (pix_type *dst, \ - int dst_stride, \ - const pix_type *src, \ - int src_stride, \ - int w, \ - int h) \ -{ \ - int x, y; \ - for (y = 0; y < h; y++) \ - { \ - const pix_type *s = src + (h - y - 1); \ - pix_type *d = dst + dst_stride * y; \ - for (x = 0; x < w; x++) \ - { \ - *d++ = *s; \ - s += src_stride; \ - } \ - } \ -} \ - \ -static void \ -blt_rotated_270_trivial_##suffix (pix_type *dst, \ - int dst_stride, \ - const pix_type *src, \ - int src_stride, \ - int w, \ - int h) \ -{ \ - int x, y; \ - for (y = 0; y < h; y++) \ - { \ - const pix_type *s = src + src_stride * (w - 1) + y; \ - pix_type *d = dst + dst_stride * y; \ - for (x = 0; x < w; x++) \ - { \ - *d++ = *s; \ - s -= src_stride; \ - } \ - } \ -} \ - \ -static void \ -blt_rotated_90_##suffix (pix_type *dst, \ - int dst_stride, \ - const pix_type *src, \ - int src_stride, \ - int W, \ - int H) \ -{ \ - int x; \ - int leading_pixels = 0, trailing_pixels = 0; \ - const int TILE_SIZE = CACHE_LINE_SIZE / sizeof(pix_type); \ - \ - /* \ - * split processing into handling destination as TILE_SIZExH cache line \ - * aligned vertical stripes (optimistically assuming that destination \ - * stride is a multiple of cache line, if not - it will be just a bit \ - * slower) \ - */ \ - \ - if ((uintptr_t)dst & (CACHE_LINE_SIZE - 1)) \ - { \ - leading_pixels = TILE_SIZE - (((uintptr_t)dst & \ - (CACHE_LINE_SIZE - 1)) / sizeof(pix_type)); \ - if (leading_pixels > W) \ - leading_pixels = W; \ - \ - /* unaligned leading part NxH (where N < TILE_SIZE) */ \ - blt_rotated_90_trivial_##suffix ( \ - dst, \ - dst_stride, \ - src, \ - src_stride, \ - leading_pixels, \ - H); \ - \ - dst += leading_pixels; \ - src += leading_pixels * src_stride; \ - W -= leading_pixels; \ - } \ - \ - if ((uintptr_t)(dst + W) & (CACHE_LINE_SIZE - 1)) \ - { \ - trailing_pixels = (((uintptr_t)(dst + W) & \ - (CACHE_LINE_SIZE - 1)) / sizeof(pix_type)); \ - if (trailing_pixels > W) \ - trailing_pixels = W; \ - W -= trailing_pixels; \ - } \ - \ - for (x = 0; x < W; x += TILE_SIZE) \ - { \ - /* aligned middle part TILE_SIZExH */ \ - blt_rotated_90_trivial_##suffix ( \ - dst + x, \ - dst_stride, \ - src + src_stride * x, \ - src_stride, \ - TILE_SIZE, \ - H); \ - } \ - \ - if (trailing_pixels) \ - { \ - /* unaligned trailing part NxH (where N < TILE_SIZE) */ \ - blt_rotated_90_trivial_##suffix ( \ - dst + W, \ - dst_stride, \ - src + W * src_stride, \ - src_stride, \ - trailing_pixels, \ - H); \ - } \ -} \ - \ -static void \ -blt_rotated_270_##suffix (pix_type *dst, \ - int dst_stride, \ - const pix_type *src, \ - int src_stride, \ - int W, \ - int H) \ -{ \ - int x; \ - int leading_pixels = 0, trailing_pixels = 0; \ - const int TILE_SIZE = CACHE_LINE_SIZE / sizeof(pix_type); \ - \ - /* \ - * split processing into handling destination as TILE_SIZExH cache line \ - * aligned vertical stripes (optimistically assuming that destination \ - * stride is a multiple of cache line, if not - it will be just a bit \ - * slower) \ - */ \ - \ - if ((uintptr_t)dst & (CACHE_LINE_SIZE - 1)) \ - { \ - leading_pixels = TILE_SIZE - (((uintptr_t)dst & \ - (CACHE_LINE_SIZE - 1)) / sizeof(pix_type)); \ - if (leading_pixels > W) \ - leading_pixels = W; \ - \ - /* unaligned leading part NxH (where N < TILE_SIZE) */ \ - blt_rotated_270_trivial_##suffix ( \ - dst, \ - dst_stride, \ - src + src_stride * (W - leading_pixels), \ - src_stride, \ - leading_pixels, \ - H); \ - \ - dst += leading_pixels; \ - W -= leading_pixels; \ - } \ - \ - if ((uintptr_t)(dst + W) & (CACHE_LINE_SIZE - 1)) \ - { \ - trailing_pixels = (((uintptr_t)(dst + W) & \ - (CACHE_LINE_SIZE - 1)) / sizeof(pix_type)); \ - if (trailing_pixels > W) \ - trailing_pixels = W; \ - W -= trailing_pixels; \ - src += trailing_pixels * src_stride; \ - } \ - \ - for (x = 0; x < W; x += TILE_SIZE) \ - { \ - /* aligned middle part TILE_SIZExH */ \ - blt_rotated_270_trivial_##suffix ( \ - dst + x, \ - dst_stride, \ - src + src_stride * (W - x - TILE_SIZE), \ - src_stride, \ - TILE_SIZE, \ - H); \ - } \ - \ - if (trailing_pixels) \ - { \ - /* unaligned trailing part NxH (where N < TILE_SIZE) */ \ - blt_rotated_270_trivial_##suffix ( \ - dst + W, \ - dst_stride, \ - src - trailing_pixels * src_stride, \ - src_stride, \ - trailing_pixels, \ - H); \ - } \ -} \ - \ -static void \ -fast_composite_rotate_90_##suffix (pixman_implementation_t *imp, \ - pixman_op_t op, \ - pixman_image_t * src_image, \ - pixman_image_t * mask_image, \ - pixman_image_t * dst_image, \ - int32_t src_x, \ - int32_t src_y, \ - int32_t mask_x, \ - int32_t mask_y, \ - int32_t dest_x, \ - int32_t dest_y, \ - int32_t width, \ - int32_t height) \ -{ \ - pix_type *dst_line; \ - pix_type *src_line; \ - int dst_stride, src_stride; \ - int src_x_t, src_y_t; \ - \ - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, pix_type, \ - dst_stride, dst_line, 1); \ - src_x_t = -src_y + pixman_fixed_to_int ( \ - src_image->common.transform->matrix[0][2] + \ - pixman_fixed_1 / 2 - pixman_fixed_e) - height;\ - src_y_t = src_x + pixman_fixed_to_int ( \ - src_image->common.transform->matrix[1][2] + \ - pixman_fixed_1 / 2 - pixman_fixed_e); \ - PIXMAN_IMAGE_GET_LINE (src_image, src_x_t, src_y_t, pix_type, \ - src_stride, src_line, 1); \ - blt_rotated_90_##suffix (dst_line, dst_stride, src_line, src_stride, \ - width, height); \ -} \ - \ -static void \ -fast_composite_rotate_270_##suffix (pixman_implementation_t *imp, \ - pixman_op_t op, \ - pixman_image_t * src_image, \ - pixman_image_t * mask_image, \ - pixman_image_t * dst_image, \ - int32_t src_x, \ - int32_t src_y, \ - int32_t mask_x, \ - int32_t mask_y, \ - int32_t dest_x, \ - int32_t dest_y, \ - int32_t width, \ - int32_t height) \ -{ \ - pix_type *dst_line; \ - pix_type *src_line; \ - int dst_stride, src_stride; \ - int src_x_t, src_y_t; \ - \ - PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, pix_type, \ - dst_stride, dst_line, 1); \ - src_x_t = src_y + pixman_fixed_to_int ( \ - src_image->common.transform->matrix[0][2] + \ - pixman_fixed_1 / 2 - pixman_fixed_e); \ - src_y_t = -src_x + pixman_fixed_to_int ( \ - src_image->common.transform->matrix[1][2] + \ - pixman_fixed_1 / 2 - pixman_fixed_e) - width; \ - PIXMAN_IMAGE_GET_LINE (src_image, src_x_t, src_y_t, pix_type, \ - src_stride, src_line, 1); \ - blt_rotated_270_##suffix (dst_line, dst_stride, src_line, src_stride, \ - width, height); \ -} - -FAST_SIMPLE_ROTATE (8, uint8_t) -FAST_SIMPLE_ROTATE (565, uint16_t) -FAST_SIMPLE_ROTATE (8888, uint32_t) - -static const pixman_fast_path_t c_fast_paths[] = -{ - PIXMAN_STD_FAST_PATH (OVER, solid, a8, r5g6b5, fast_composite_over_n_8_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, b5g6r5, fast_composite_over_n_8_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, r8g8b8, fast_composite_over_n_8_0888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, b8g8r8, fast_composite_over_n_8_0888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8r8g8b8, fast_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8r8g8b8, fast_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8b8g8r8, fast_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8b8g8r8, fast_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a1, a8r8g8b8, fast_composite_over_n_1_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a1, x8r8g8b8, fast_composite_over_n_1_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a1, a8b8g8r8, fast_composite_over_n_1_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a1, x8b8g8r8, fast_composite_over_n_1_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a1, r5g6b5, fast_composite_over_n_1_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, a1, b5g6r5, fast_composite_over_n_1_0565), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, a8r8g8b8, fast_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, x8r8g8b8, fast_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, r5g6b5, fast_composite_over_n_8888_0565_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, a8b8g8r8, fast_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, x8b8g8r8, fast_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, b5g6r5, fast_composite_over_n_8888_0565_ca), - PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, a8, x8r8g8b8, fast_composite_over_x888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, a8, a8r8g8b8, fast_composite_over_x888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, a8, x8b8g8r8, fast_composite_over_x888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, a8, a8b8g8r8, fast_composite_over_x888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, a8r8g8b8, fast_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, x8r8g8b8, fast_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, r5g6b5, fast_composite_over_8888_0565), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, a8b8g8r8, fast_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, x8b8g8r8, fast_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, b5g6r5, fast_composite_over_8888_0565), - PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, fast_composite_add_8888_8888), - PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, null, a8b8g8r8, fast_composite_add_8888_8888), - PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, fast_composite_add_8_8), - PIXMAN_STD_FAST_PATH (ADD, a1, null, a1, fast_composite_add_1000_1000), - PIXMAN_STD_FAST_PATH_CA (ADD, solid, a8r8g8b8, a8r8g8b8, fast_composite_add_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8, fast_composite_add_n_8_8), - PIXMAN_STD_FAST_PATH (SRC, solid, null, a8r8g8b8, fast_composite_solid_fill), - PIXMAN_STD_FAST_PATH (SRC, solid, null, x8r8g8b8, fast_composite_solid_fill), - PIXMAN_STD_FAST_PATH (SRC, solid, null, a8b8g8r8, fast_composite_solid_fill), - PIXMAN_STD_FAST_PATH (SRC, solid, null, x8b8g8r8, fast_composite_solid_fill), - PIXMAN_STD_FAST_PATH (SRC, solid, null, a1, fast_composite_solid_fill), - PIXMAN_STD_FAST_PATH (SRC, solid, null, a8, fast_composite_solid_fill), - PIXMAN_STD_FAST_PATH (SRC, solid, null, r5g6b5, fast_composite_solid_fill), - PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, a8r8g8b8, fast_composite_src_x888_8888), - PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, a8b8g8r8, fast_composite_src_x888_8888), - PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, x8r8g8b8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, a8r8g8b8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, x8r8g8b8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, x8b8g8r8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, a8b8g8r8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, x8b8g8r8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, b8g8r8a8, null, b8g8r8x8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, b8g8r8a8, null, b8g8r8a8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, b8g8r8x8, null, b8g8r8x8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, r5g6b5, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, b5g6r5, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, r8g8b8, null, r8g8b8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, b8g8r8, null, b8g8r8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, x1r5g5b5, null, x1r5g5b5, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, a1r5g5b5, null, x1r5g5b5, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, a8, null, a8, fast_composite_src_memcpy), - PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, r5g6b5, fast_composite_src_x888_0565), - PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, r5g6b5, fast_composite_src_x888_0565), - PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, b5g6r5, fast_composite_src_x888_0565), - PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, b5g6r5, fast_composite_src_x888_0565), - PIXMAN_STD_FAST_PATH (IN, a8, null, a8, fast_composite_in_8_8), - PIXMAN_STD_FAST_PATH (IN, solid, a8, a8, fast_composite_in_n_8_8), - - SIMPLE_NEAREST_FAST_PATH (SRC, x8r8g8b8, x8r8g8b8, 8888_8888), - SIMPLE_NEAREST_FAST_PATH (SRC, a8r8g8b8, x8r8g8b8, 8888_8888), - SIMPLE_NEAREST_FAST_PATH (SRC, x8b8g8r8, x8b8g8r8, 8888_8888), - SIMPLE_NEAREST_FAST_PATH (SRC, a8b8g8r8, x8b8g8r8, 8888_8888), - - SIMPLE_NEAREST_FAST_PATH (SRC, a8r8g8b8, a8r8g8b8, 8888_8888), - SIMPLE_NEAREST_FAST_PATH (SRC, a8b8g8r8, a8b8g8r8, 8888_8888), - - SIMPLE_NEAREST_FAST_PATH (SRC, x8r8g8b8, r5g6b5, 8888_565), - SIMPLE_NEAREST_FAST_PATH (SRC, a8r8g8b8, r5g6b5, 8888_565), - - SIMPLE_NEAREST_FAST_PATH (SRC, r5g6b5, r5g6b5, 565_565), - - SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, 8888_8888), - SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8, 8888_8888), - SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, 8888_8888), - SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8, 8888_8888), - - SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, r5g6b5, 8888_565), - -#define NEAREST_FAST_PATH(op,s,d) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, SCALED_NEAREST_FLAGS, \ - PIXMAN_null, 0, \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest, \ - } - - NEAREST_FAST_PATH (SRC, x8r8g8b8, x8r8g8b8), - NEAREST_FAST_PATH (SRC, a8r8g8b8, x8r8g8b8), - NEAREST_FAST_PATH (SRC, x8b8g8r8, x8b8g8r8), - NEAREST_FAST_PATH (SRC, a8b8g8r8, x8b8g8r8), - - NEAREST_FAST_PATH (SRC, x8r8g8b8, a8r8g8b8), - NEAREST_FAST_PATH (SRC, a8r8g8b8, a8r8g8b8), - NEAREST_FAST_PATH (SRC, x8b8g8r8, a8b8g8r8), - NEAREST_FAST_PATH (SRC, a8b8g8r8, a8b8g8r8), - - NEAREST_FAST_PATH (OVER, x8r8g8b8, x8r8g8b8), - NEAREST_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8), - NEAREST_FAST_PATH (OVER, x8b8g8r8, x8b8g8r8), - NEAREST_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8), - - NEAREST_FAST_PATH (OVER, x8r8g8b8, a8r8g8b8), - NEAREST_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8), - NEAREST_FAST_PATH (OVER, x8b8g8r8, a8b8g8r8), - NEAREST_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8), - -#define SIMPLE_ROTATE_FLAGS(angle) \ - (FAST_PATH_ROTATE_ ## angle ## _TRANSFORM | \ - FAST_PATH_NEAREST_FILTER | \ - FAST_PATH_SAMPLES_COVER_CLIP | \ - FAST_PATH_STANDARD_FLAGS) - -#define SIMPLE_ROTATE_FAST_PATH(op,s,d,suffix) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, SIMPLE_ROTATE_FLAGS (90), \ - PIXMAN_null, 0, \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_rotate_90_##suffix, \ - }, \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, SIMPLE_ROTATE_FLAGS (270), \ - PIXMAN_null, 0, \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_rotate_270_##suffix, \ - } - - SIMPLE_ROTATE_FAST_PATH (SRC, a8r8g8b8, a8r8g8b8, 8888), - SIMPLE_ROTATE_FAST_PATH (SRC, a8r8g8b8, x8r8g8b8, 8888), - SIMPLE_ROTATE_FAST_PATH (SRC, x8r8g8b8, x8r8g8b8, 8888), - SIMPLE_ROTATE_FAST_PATH (SRC, r5g6b5, r5g6b5, 565), - SIMPLE_ROTATE_FAST_PATH (SRC, a8, a8, 8), - - { PIXMAN_OP_NONE }, -}; - -#ifdef WORDS_BIGENDIAN -#define A1_FILL_MASK(n, offs) (((1 << (n)) - 1) << (32 - (offs) - (n))) -#else -#define A1_FILL_MASK(n, offs) (((1 << (n)) - 1) << (offs)) -#endif - -static force_inline void -pixman_fill1_line (uint32_t *dst, int offs, int width, int v) -{ - if (offs) - { - int leading_pixels = 32 - offs; - if (leading_pixels >= width) - { - if (v) - *dst |= A1_FILL_MASK (width, offs); - else - *dst &= ~A1_FILL_MASK (width, offs); - return; - } - else - { - if (v) - *dst++ |= A1_FILL_MASK (leading_pixels, offs); - else - *dst++ &= ~A1_FILL_MASK (leading_pixels, offs); - width -= leading_pixels; - } - } - while (width >= 32) - { - if (v) - *dst++ = 0xFFFFFFFF; - else - *dst++ = 0; - width -= 32; - } - if (width > 0) - { - if (v) - *dst |= A1_FILL_MASK (width, 0); - else - *dst &= ~A1_FILL_MASK (width, 0); - } -} - -static void -pixman_fill1 (uint32_t *bits, - int stride, - int x, - int y, - int width, - int height, - uint32_t xor) -{ - uint32_t *dst = bits + y * stride + (x >> 5); - int offs = x & 31; - - if (xor & 1) - { - while (height--) - { - pixman_fill1_line (dst, offs, width, 1); - dst += stride; - } - } - else - { - while (height--) - { - pixman_fill1_line (dst, offs, width, 0); - dst += stride; - } - } -} - -static void -pixman_fill8 (uint32_t *bits, - int stride, - int x, - int y, - int width, - int height, - uint32_t xor) -{ - int byte_stride = stride * (int) sizeof (uint32_t); - uint8_t *dst = (uint8_t *) bits; - uint8_t v = xor & 0xff; - int i; - - dst = dst + y * byte_stride + x; - - while (height--) - { - for (i = 0; i < width; ++i) - dst[i] = v; - - dst += byte_stride; - } -} - -static void -pixman_fill16 (uint32_t *bits, - int stride, - int x, - int y, - int width, - int height, - uint32_t xor) -{ - int short_stride = - (stride * (int)sizeof (uint32_t)) / (int)sizeof (uint16_t); - uint16_t *dst = (uint16_t *)bits; - uint16_t v = xor & 0xffff; - int i; - - dst = dst + y * short_stride + x; - - while (height--) - { - for (i = 0; i < width; ++i) - dst[i] = v; - - dst += short_stride; - } -} - -static void -pixman_fill32 (uint32_t *bits, - int stride, - int x, - int y, - int width, - int height, - uint32_t xor) -{ - int i; - - bits = bits + y * stride + x; - - while (height--) - { - for (i = 0; i < width; ++i) - bits[i] = xor; - - bits += stride; - } -} - -static pixman_bool_t -fast_path_fill (pixman_implementation_t *imp, - uint32_t * bits, - int stride, - int bpp, - int x, - int y, - int width, - int height, - uint32_t xor) -{ - switch (bpp) - { - case 1: - pixman_fill1 (bits, stride, x, y, width, height, xor); - break; - - case 8: - pixman_fill8 (bits, stride, x, y, width, height, xor); - break; - - case 16: - pixman_fill16 (bits, stride, x, y, width, height, xor); - break; - - case 32: - pixman_fill32 (bits, stride, x, y, width, height, xor); - break; - - default: - return _pixman_implementation_fill ( - imp->delegate, bits, stride, bpp, x, y, width, height, xor); - break; - } - - return TRUE; -} - -pixman_implementation_t * -_pixman_implementation_create_fast_path (pixman_implementation_t *fallback) -{ - pixman_implementation_t *imp = _pixman_implementation_create (fallback, c_fast_paths); - - imp->fill = fast_path_fill; - - return imp; -} +/* -*- Mode: c; c-basic-offset: 4; tab-width: 8; indent-tabs-mode: t; -*- */
+/*
+ * Copyright © 2000 SuSE, Inc.
+ * Copyright © 2007 Red Hat, Inc.
+ *
+ * Permission to use, copy, modify, distribute, and sell this software and its
+ * documentation for any purpose is hereby granted without fee, provided that
+ * the above copyright notice appear in all copies and that both that
+ * copyright notice and this permission notice appear in supporting
+ * documentation, and that the name of SuSE not be used in advertising or
+ * publicity pertaining to distribution of the software without specific,
+ * written prior permission. SuSE makes no representations about the
+ * suitability of this software for any purpose. It is provided "as is"
+ * without express or implied warranty.
+ *
+ * SuSE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL SuSE
+ * BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * Author: Keith Packard, SuSE, Inc.
+ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+#include <string.h>
+#include <stdlib.h>
+#include "pixman-private.h"
+#include "pixman-combine32.h"
+#include "pixman-fast-path.h"
+
+static force_inline uint32_t
+fetch_24 (uint8_t *a)
+{
+ if (((unsigned long)a) & 1)
+ {
+#ifdef WORDS_BIGENDIAN
+ return (*a << 16) | (*(uint16_t *)(a + 1));
+#else
+ return *a | (*(uint16_t *)(a + 1) << 8);
+#endif
+ }
+ else
+ {
+#ifdef WORDS_BIGENDIAN
+ return (*(uint16_t *)a << 8) | *(a + 2);
+#else
+ return *(uint16_t *)a | (*(a + 2) << 16);
+#endif
+ }
+}
+
+static force_inline void
+store_24 (uint8_t *a,
+ uint32_t v)
+{
+ if (((unsigned long)a) & 1)
+ {
+#ifdef WORDS_BIGENDIAN
+ *a = (uint8_t) (v >> 16);
+ *(uint16_t *)(a + 1) = (uint16_t) (v);
+#else
+ *a = (uint8_t) (v);
+ *(uint16_t *)(a + 1) = (uint16_t) (v >> 8);
+#endif
+ }
+ else
+ {
+#ifdef WORDS_BIGENDIAN
+ *(uint16_t *)a = (uint16_t)(v >> 8);
+ *(a + 2) = (uint8_t)v;
+#else
+ *(uint16_t *)a = (uint16_t)v;
+ *(a + 2) = (uint8_t)(v >> 16);
+#endif
+ }
+}
+
+static force_inline uint32_t
+over (uint32_t src,
+ uint32_t dest)
+{
+ uint32_t a = ~src >> 24;
+
+ UN8x4_MUL_UN8_ADD_UN8x4 (dest, a, src);
+
+ return dest;
+}
+
+static uint32_t
+in (uint32_t x,
+ uint8_t y)
+{
+ uint16_t a = y;
+
+ UN8x4_MUL_UN8 (x, a);
+
+ return x;
+}
+
+/*
+ * Naming convention:
+ *
+ * op_src_mask_dest
+ */
+static void
+fast_composite_over_x888_8_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *src, *src_line;
+ uint32_t *dst, *dst_line;
+ uint8_t *mask, *mask_line;
+ int src_stride, mask_stride, dst_stride;
+ uint8_t m;
+ uint32_t s, d;
+ int32_t w;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ src = src_line;
+ src_line += src_stride;
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+
+ w = width;
+ while (w--)
+ {
+ m = *mask++;
+ if (m)
+ {
+ s = *src | 0xff000000;
+
+ if (m == 0xff)
+ {
+ *dst = s;
+ }
+ else
+ {
+ d = in (s, m);
+ *dst = over (d, *dst);
+ }
+ }
+ src++;
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_in_n_8_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dest_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint8_t *dst_line, *dst;
+ uint8_t *mask_line, *mask, m;
+ int dst_stride, mask_stride;
+ int32_t w;
+ uint16_t t;
+
+ src = _pixman_image_get_solid (imp, src_image, dest_image->bits.format);
+
+ srca = src >> 24;
+
+ PIXMAN_IMAGE_GET_LINE (dest_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ if (srca == 0xff)
+ {
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ m = *mask++;
+
+ if (m == 0)
+ *dst = 0;
+ else if (m != 0xff)
+ *dst = MUL_UN8 (m, *dst, t);
+
+ dst++;
+ }
+ }
+ }
+ else
+ {
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ m = *mask++;
+ m = MUL_UN8 (m, srca, t);
+
+ if (m == 0)
+ *dst = 0;
+ else if (m != 0xff)
+ *dst = MUL_UN8 (m, *dst, t);
+
+ dst++;
+ }
+ }
+ }
+}
+
+static void
+fast_composite_in_8_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dest_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ uint8_t *src_line, *src;
+ int dst_stride, src_stride;
+ int32_t w;
+ uint8_t s;
+ uint16_t t;
+
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint8_t, src_stride, src_line, 1);
+ PIXMAN_IMAGE_GET_LINE (dest_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ {
+ s = *src++;
+
+ if (s == 0)
+ *dst = 0;
+ else if (s != 0xff)
+ *dst = MUL_UN8 (s, *dst, t);
+
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_over_n_8_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint32_t *dst_line, *dst, d;
+ uint8_t *mask_line, *mask, m;
+ int dst_stride, mask_stride;
+ int32_t w;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ m = *mask++;
+ if (m == 0xff)
+ {
+ if (srca == 0xff)
+ *dst = src;
+ else
+ *dst = over (src, *dst);
+ }
+ else if (m)
+ {
+ d = in (src, m);
+ *dst = over (d, *dst);
+ }
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_add_n_8888_8888_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, s;
+ uint32_t *dst_line, *dst, d;
+ uint32_t *mask_line, *mask, ma;
+ int dst_stride, mask_stride;
+ int32_t w;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ ma = *mask++;
+
+ if (ma)
+ {
+ d = *dst;
+ s = src;
+
+ UN8x4_MUL_UN8x4_ADD_UN8x4 (s, ma, d);
+
+ *dst = s;
+ }
+
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_over_n_8888_8888_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca, s;
+ uint32_t *dst_line, *dst, d;
+ uint32_t *mask_line, *mask, ma;
+ int dst_stride, mask_stride;
+ int32_t w;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ ma = *mask++;
+ if (ma == 0xffffffff)
+ {
+ if (srca == 0xff)
+ *dst = src;
+ else
+ *dst = over (src, *dst);
+ }
+ else if (ma)
+ {
+ d = *dst;
+ s = src;
+
+ UN8x4_MUL_UN8x4 (s, ma);
+ UN8x4_MUL_UN8 (ma, srca);
+ ma = ~ma;
+ UN8x4_MUL_UN8x4_ADD_UN8x4 (d, ma, s);
+
+ *dst = d;
+ }
+
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_over_n_8_0888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint8_t *dst_line, *dst;
+ uint32_t d;
+ uint8_t *mask_line, *mask, m;
+ int dst_stride, mask_stride;
+ int32_t w;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 3);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ m = *mask++;
+ if (m == 0xff)
+ {
+ if (srca == 0xff)
+ {
+ d = src;
+ }
+ else
+ {
+ d = fetch_24 (dst);
+ d = over (src, d);
+ }
+ store_24 (dst, d);
+ }
+ else if (m)
+ {
+ d = over (in (src, m), fetch_24 (dst));
+ store_24 (dst, d);
+ }
+ dst += 3;
+ }
+ }
+}
+
+static void
+fast_composite_over_n_8_0565 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint16_t *dst_line, *dst;
+ uint32_t d;
+ uint8_t *mask_line, *mask, m;
+ int dst_stride, mask_stride;
+ int32_t w;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ m = *mask++;
+ if (m == 0xff)
+ {
+ if (srca == 0xff)
+ {
+ d = src;
+ }
+ else
+ {
+ d = *dst;
+ d = over (src, CONVERT_0565_TO_0888 (d));
+ }
+ *dst = CONVERT_8888_TO_0565 (d);
+ }
+ else if (m)
+ {
+ d = *dst;
+ d = over (in (src, m), CONVERT_0565_TO_0888 (d));
+ *dst = CONVERT_8888_TO_0565 (d);
+ }
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_over_n_8888_0565_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca, s;
+ uint16_t src16;
+ uint16_t *dst_line, *dst;
+ uint32_t d;
+ uint32_t *mask_line, *mask, ma;
+ int dst_stride, mask_stride;
+ int32_t w;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ src16 = CONVERT_8888_TO_0565 (src);
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ ma = *mask++;
+ if (ma == 0xffffffff)
+ {
+ if (srca == 0xff)
+ {
+ *dst = src16;
+ }
+ else
+ {
+ d = *dst;
+ d = over (src, CONVERT_0565_TO_0888 (d));
+ *dst = CONVERT_8888_TO_0565 (d);
+ }
+ }
+ else if (ma)
+ {
+ d = *dst;
+ d = CONVERT_0565_TO_0888 (d);
+
+ s = src;
+
+ UN8x4_MUL_UN8x4 (s, ma);
+ UN8x4_MUL_UN8 (ma, srca);
+ ma = ~ma;
+ UN8x4_MUL_UN8x4_ADD_UN8x4 (d, ma, s);
+
+ *dst = CONVERT_8888_TO_0565 (d);
+ }
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_over_8888_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src, s;
+ int dst_stride, src_stride;
+ uint8_t a;
+ int32_t w;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ {
+ s = *src++;
+ a = s >> 24;
+ if (a == 0xff)
+ *dst = s;
+ else if (s)
+ *dst = over (s, *dst);
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_src_x888_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src;
+ int dst_stride, src_stride;
+ int32_t w;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ *dst++ = (*src++) | 0xff000000;
+ }
+}
+
+#if 0
+static void
+fast_composite_over_8888_0888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ uint32_t d;
+ uint32_t *src_line, *src, s;
+ uint8_t a;
+ int dst_stride, src_stride;
+ int32_t w;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 3);
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ {
+ s = *src++;
+ a = s >> 24;
+ if (a)
+ {
+ if (a == 0xff)
+ d = s;
+ else
+ d = over (s, fetch_24 (dst));
+
+ store_24 (dst, d);
+ }
+ dst += 3;
+ }
+ }
+}
+#endif
+
+static void
+fast_composite_over_8888_0565 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint16_t *dst_line, *dst;
+ uint32_t d;
+ uint32_t *src_line, *src, s;
+ uint8_t a;
+ int dst_stride, src_stride;
+ int32_t w;
+
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ {
+ s = *src++;
+ a = s >> 24;
+ if (s)
+ {
+ if (a == 0xff)
+ {
+ d = s;
+ }
+ else
+ {
+ d = *dst;
+ d = over (s, CONVERT_0565_TO_0888 (d));
+ }
+ *dst = CONVERT_8888_TO_0565 (d);
+ }
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_src_x888_0565 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint16_t *dst_line, *dst;
+ uint32_t *src_line, *src, s;
+ int dst_stride, src_stride;
+ int32_t w;
+
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ {
+ s = *src++;
+ *dst = CONVERT_8888_TO_0565 (s);
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_add_8_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ uint8_t *src_line, *src;
+ int dst_stride, src_stride;
+ int32_t w;
+ uint8_t s, d;
+ uint16_t t;
+
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint8_t, src_stride, src_line, 1);
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ {
+ s = *src++;
+ if (s)
+ {
+ if (s != 0xff)
+ {
+ d = *dst;
+ t = d + s;
+ s = t | (0 - (t >> 8));
+ }
+ *dst = s;
+ }
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_add_8888_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src;
+ int dst_stride, src_stride;
+ int32_t w;
+ uint32_t s, d;
+
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ {
+ s = *src++;
+ if (s)
+ {
+ if (s != 0xffffffff)
+ {
+ d = *dst;
+ if (d)
+ UN8x4_ADD_UN8x4 (s, d);
+ }
+ *dst = s;
+ }
+ dst++;
+ }
+ }
+}
+
+static void
+fast_composite_add_n_8_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ uint8_t *mask_line, *mask;
+ int dst_stride, mask_stride;
+ int32_t w;
+ uint32_t src;
+ uint8_t sa;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+ sa = (src >> 24);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w--)
+ {
+ uint16_t tmp;
+ uint16_t a;
+ uint32_t m, d;
+ uint32_t r;
+
+ a = *mask++;
+ d = *dst;
+
+ m = MUL_UN8 (sa, a, tmp);
+ r = ADD_UN8 (m, d, tmp);
+
+ *dst++ = r;
+ }
+ }
+}
+
+#ifdef WORDS_BIGENDIAN
+#define CREATE_BITMASK(n) (0x80000000 >> (n))
+#define UPDATE_BITMASK(n) ((n) >> 1)
+#else
+#define CREATE_BITMASK(n) (1 << (n))
+#define UPDATE_BITMASK(n) ((n) << 1)
+#endif
+
+#define TEST_BIT(p, n) \
+ (*((p) + ((n) >> 5)) & CREATE_BITMASK ((n) & 31))
+#define SET_BIT(p, n) \
+ do { *((p) + ((n) >> 5)) |= CREATE_BITMASK ((n) & 31); } while (0);
+
+static void
+fast_composite_add_1000_1000 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src;
+ int dst_stride, src_stride;
+ int32_t w;
+
+ PIXMAN_IMAGE_GET_LINE (src_image, 0, src_y, uint32_t,
+ src_stride, src_line, 1);
+ PIXMAN_IMAGE_GET_LINE (dst_image, 0, dest_y, uint32_t,
+ dst_stride, dst_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w--)
+ {
+ /*
+ * TODO: improve performance by processing uint32_t data instead
+ * of individual bits
+ */
+ if (TEST_BIT (src, src_x + w))
+ SET_BIT (dst, dest_x + w);
+ }
+ }
+}
+
+static void
+fast_composite_over_n_1_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint32_t *dst, *dst_line;
+ uint32_t *mask, *mask_line;
+ int mask_stride, dst_stride;
+ uint32_t bitcache, bitmask;
+ int32_t w;
+
+ if (width <= 0)
+ return;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t,
+ dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, 0, mask_y, uint32_t,
+ mask_stride, mask_line, 1);
+ mask_line += mask_x >> 5;
+
+ if (srca == 0xff)
+ {
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ bitcache = *mask++;
+ bitmask = CREATE_BITMASK (mask_x & 31);
+
+ while (w--)
+ {
+ if (bitmask == 0)
+ {
+ bitcache = *mask++;
+ bitmask = CREATE_BITMASK (0);
+ }
+ if (bitcache & bitmask)
+ *dst = src;
+ bitmask = UPDATE_BITMASK (bitmask);
+ dst++;
+ }
+ }
+ }
+ else
+ {
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ bitcache = *mask++;
+ bitmask = CREATE_BITMASK (mask_x & 31);
+
+ while (w--)
+ {
+ if (bitmask == 0)
+ {
+ bitcache = *mask++;
+ bitmask = CREATE_BITMASK (0);
+ }
+ if (bitcache & bitmask)
+ *dst = over (src, *dst);
+ bitmask = UPDATE_BITMASK (bitmask);
+ dst++;
+ }
+ }
+ }
+}
+
+static void
+fast_composite_over_n_1_0565 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint16_t *dst, *dst_line;
+ uint32_t *mask, *mask_line;
+ int mask_stride, dst_stride;
+ uint32_t bitcache, bitmask;
+ int32_t w;
+ uint32_t d;
+ uint16_t src565;
+
+ if (width <= 0)
+ return;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint16_t,
+ dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (mask_image, 0, mask_y, uint32_t,
+ mask_stride, mask_line, 1);
+ mask_line += mask_x >> 5;
+
+ if (srca == 0xff)
+ {
+ src565 = CONVERT_8888_TO_0565 (src);
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ bitcache = *mask++;
+ bitmask = CREATE_BITMASK (mask_x & 31);
+
+ while (w--)
+ {
+ if (bitmask == 0)
+ {
+ bitcache = *mask++;
+ bitmask = CREATE_BITMASK (0);
+ }
+ if (bitcache & bitmask)
+ *dst = src565;
+ bitmask = UPDATE_BITMASK (bitmask);
+ dst++;
+ }
+ }
+ }
+ else
+ {
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ bitcache = *mask++;
+ bitmask = CREATE_BITMASK (mask_x & 31);
+
+ while (w--)
+ {
+ if (bitmask == 0)
+ {
+ bitcache = *mask++;
+ bitmask = CREATE_BITMASK (0);
+ }
+ if (bitcache & bitmask)
+ {
+ d = over (src, CONVERT_0565_TO_0888 (*dst));
+ *dst = CONVERT_8888_TO_0565 (d);
+ }
+ bitmask = UPDATE_BITMASK (bitmask);
+ dst++;
+ }
+ }
+ }
+}
+
+/*
+ * Simple bitblt
+ */
+
+static void
+fast_composite_solid_fill (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ if (dst_image->bits.format == PIXMAN_a1)
+ {
+ src = src >> 31;
+ }
+ else if (dst_image->bits.format == PIXMAN_a8)
+ {
+ src = src >> 24;
+ }
+ else if (dst_image->bits.format == PIXMAN_r5g6b5 ||
+ dst_image->bits.format == PIXMAN_b5g6r5)
+ {
+ src = CONVERT_8888_TO_0565 (src);
+ }
+
+ pixman_fill (dst_image->bits.bits, dst_image->bits.rowstride,
+ PIXMAN_FORMAT_BPP (dst_image->bits.format),
+ dest_x, dest_y,
+ width, height,
+ src);
+}
+
+static void
+fast_composite_src_memcpy (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ int bpp = PIXMAN_FORMAT_BPP (dst_image->bits.format) / 8;
+ uint32_t n_bytes = width * bpp;
+ int dst_stride, src_stride;
+ uint8_t *dst;
+ uint8_t *src;
+
+ src_stride = src_image->bits.rowstride * 4;
+ dst_stride = dst_image->bits.rowstride * 4;
+
+ src = (uint8_t *)src_image->bits.bits + src_y * src_stride + src_x * bpp;
+ dst = (uint8_t *)dst_image->bits.bits + dest_y * dst_stride + dest_x * bpp;
+
+ while (height--)
+ {
+ memcpy (dst, src, n_bytes);
+
+ dst += dst_stride;
+ src += src_stride;
+ }
+}
+
+FAST_NEAREST (8888_8888_cover, 8888, 8888, uint32_t, uint32_t, SRC, COVER)
+FAST_NEAREST (8888_8888_none, 8888, 8888, uint32_t, uint32_t, SRC, NONE)
+FAST_NEAREST (8888_8888_pad, 8888, 8888, uint32_t, uint32_t, SRC, PAD)
+FAST_NEAREST (8888_8888_normal, 8888, 8888, uint32_t, uint32_t, SRC, NORMAL)
+FAST_NEAREST (8888_8888_cover, 8888, 8888, uint32_t, uint32_t, OVER, COVER)
+FAST_NEAREST (8888_8888_none, 8888, 8888, uint32_t, uint32_t, OVER, NONE)
+FAST_NEAREST (8888_8888_pad, 8888, 8888, uint32_t, uint32_t, OVER, PAD)
+FAST_NEAREST (8888_8888_normal, 8888, 8888, uint32_t, uint32_t, OVER, NORMAL)
+FAST_NEAREST (8888_565_cover, 8888, 0565, uint32_t, uint16_t, SRC, COVER)
+FAST_NEAREST (8888_565_none, 8888, 0565, uint32_t, uint16_t, SRC, NONE)
+FAST_NEAREST (8888_565_pad, 8888, 0565, uint32_t, uint16_t, SRC, PAD)
+FAST_NEAREST (8888_565_normal, 8888, 0565, uint32_t, uint16_t, SRC, NORMAL)
+FAST_NEAREST (565_565_normal, 0565, 0565, uint16_t, uint16_t, SRC, NORMAL)
+FAST_NEAREST (8888_565_cover, 8888, 0565, uint32_t, uint16_t, OVER, COVER)
+FAST_NEAREST (8888_565_none, 8888, 0565, uint32_t, uint16_t, OVER, NONE)
+FAST_NEAREST (8888_565_pad, 8888, 0565, uint32_t, uint16_t, OVER, PAD)
+FAST_NEAREST (8888_565_normal, 8888, 0565, uint32_t, uint16_t, OVER, NORMAL)
+
+/* Use more unrolling for src_0565_0565 because it is typically CPU bound */
+static force_inline void
+scaled_nearest_scanline_565_565_SRC (uint16_t * dst,
+ const uint16_t * src,
+ int32_t w,
+ pixman_fixed_t vx,
+ pixman_fixed_t unit_x,
+ pixman_fixed_t max_vx,
+ pixman_bool_t fully_transparent_src)
+{
+ uint16_t tmp1, tmp2, tmp3, tmp4;
+ while ((w -= 4) >= 0)
+ {
+ tmp1 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp2 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp3 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp4 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ *dst++ = tmp1;
+ *dst++ = tmp2;
+ *dst++ = tmp3;
+ *dst++ = tmp4;
+ }
+ if (w & 2)
+ {
+ tmp1 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp2 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ *dst++ = tmp1;
+ *dst++ = tmp2;
+ }
+ if (w & 1)
+ *dst++ = src[pixman_fixed_to_int (vx)];
+}
+
+FAST_NEAREST_MAINLOOP (565_565_cover_SRC,
+ scaled_nearest_scanline_565_565_SRC,
+ uint16_t, uint16_t, COVER)
+FAST_NEAREST_MAINLOOP (565_565_none_SRC,
+ scaled_nearest_scanline_565_565_SRC,
+ uint16_t, uint16_t, NONE)
+FAST_NEAREST_MAINLOOP (565_565_pad_SRC,
+ scaled_nearest_scanline_565_565_SRC,
+ uint16_t, uint16_t, PAD)
+
+static force_inline uint32_t
+fetch_nearest (pixman_repeat_t src_repeat,
+ pixman_format_code_t format,
+ uint32_t *src, int x, int src_width)
+{
+ if (repeat (src_repeat, &x, src_width))
+ {
+ if (format == PIXMAN_x8r8g8b8)
+ return *(src + x) | 0xff000000;
+ else
+ return *(src + x);
+ }
+ else
+ {
+ return 0;
+ }
+}
+
+static force_inline void
+combine_over (uint32_t s, uint32_t *dst)
+{
+ if (s)
+ {
+ uint8_t ia = 0xff - (s >> 24);
+
+ if (ia)
+ UN8x4_MUL_UN8_ADD_UN8x4 (*dst, ia, s);
+ else
+ *dst = s;
+ }
+}
+
+static force_inline void
+combine_src (uint32_t s, uint32_t *dst)
+{
+ *dst = s;
+}
+
+static void
+fast_composite_scaled_nearest (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line;
+ uint32_t *src_line;
+ int dst_stride, src_stride;
+ int src_width, src_height;
+ pixman_repeat_t src_repeat;
+ pixman_fixed_t unit_x, unit_y;
+ pixman_format_code_t src_format;
+ pixman_vector_t v;
+ pixman_fixed_t vy;
+
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ /* pass in 0 instead of src_x and src_y because src_x and src_y need to be
+ * transformed from destination space to source space
+ */
+ PIXMAN_IMAGE_GET_LINE (src_image, 0, 0, uint32_t, src_stride, src_line, 1);
+
+ /* reference point is the center of the pixel */
+ v.vector[0] = pixman_int_to_fixed (src_x) + pixman_fixed_1 / 2;
+ v.vector[1] = pixman_int_to_fixed (src_y) + pixman_fixed_1 / 2;
+ v.vector[2] = pixman_fixed_1;
+
+ if (!pixman_transform_point_3d (src_image->common.transform, &v))
+ return;
+
+ unit_x = src_image->common.transform->matrix[0][0];
+ unit_y = src_image->common.transform->matrix[1][1];
+
+ /* Round down to closest integer, ensuring that 0.5 rounds to 0, not 1 */
+ v.vector[0] -= pixman_fixed_e;
+ v.vector[1] -= pixman_fixed_e;
+
+ src_height = src_image->bits.height;
+ src_width = src_image->bits.width;
+ src_repeat = src_image->common.repeat;
+ src_format = src_image->bits.format;
+
+ vy = v.vector[1];
+ while (height--)
+ {
+ pixman_fixed_t vx = v.vector[0];
+ int y = pixman_fixed_to_int (vy);
+ uint32_t *dst = dst_line;
+
+ dst_line += dst_stride;
+
+ /* adjust the y location by a unit vector in the y direction
+ * this is equivalent to transforming y+1 of the destination point to source space */
+ vy += unit_y;
+
+ if (!repeat (src_repeat, &y, src_height))
+ {
+ if (op == PIXMAN_OP_SRC)
+ memset (dst, 0, sizeof (*dst) * width);
+ }
+ else
+ {
+ int w = width;
+
+ uint32_t *src = src_line + y * src_stride;
+
+ while (w >= 2)
+ {
+ uint32_t s1, s2;
+ int x1, x2;
+
+ x1 = pixman_fixed_to_int (vx);
+ vx += unit_x;
+
+ x2 = pixman_fixed_to_int (vx);
+ vx += unit_x;
+
+ w -= 2;
+
+ s1 = fetch_nearest (src_repeat, src_format, src, x1, src_width);
+ s2 = fetch_nearest (src_repeat, src_format, src, x2, src_width);
+
+ if (op == PIXMAN_OP_OVER)
+ {
+ combine_over (s1, dst++);
+ combine_over (s2, dst++);
+ }
+ else
+ {
+ combine_src (s1, dst++);
+ combine_src (s2, dst++);
+ }
+ }
+
+ while (w--)
+ {
+ uint32_t s;
+ int x;
+
+ x = pixman_fixed_to_int (vx);
+ vx += unit_x;
+
+ s = fetch_nearest (src_repeat, src_format, src, x, src_width);
+
+ if (op == PIXMAN_OP_OVER)
+ combine_over (s, dst++);
+ else
+ combine_src (s, dst++);
+ }
+ }
+ }
+}
+
+#define CACHE_LINE_SIZE 64
+
+#define FAST_SIMPLE_ROTATE(suffix, pix_type) \
+ \
+static void \
+blt_rotated_90_trivial_##suffix (pix_type *dst, \
+ int dst_stride, \
+ const pix_type *src, \
+ int src_stride, \
+ int w, \
+ int h) \
+{ \
+ int x, y; \
+ for (y = 0; y < h; y++) \
+ { \
+ const pix_type *s = src + (h - y - 1); \
+ pix_type *d = dst + dst_stride * y; \
+ for (x = 0; x < w; x++) \
+ { \
+ *d++ = *s; \
+ s += src_stride; \
+ } \
+ } \
+} \
+ \
+static void \
+blt_rotated_270_trivial_##suffix (pix_type *dst, \
+ int dst_stride, \
+ const pix_type *src, \
+ int src_stride, \
+ int w, \
+ int h) \
+{ \
+ int x, y; \
+ for (y = 0; y < h; y++) \
+ { \
+ const pix_type *s = src + src_stride * (w - 1) + y; \
+ pix_type *d = dst + dst_stride * y; \
+ for (x = 0; x < w; x++) \
+ { \
+ *d++ = *s; \
+ s -= src_stride; \
+ } \
+ } \
+} \
+ \
+static void \
+blt_rotated_90_##suffix (pix_type *dst, \
+ int dst_stride, \
+ const pix_type *src, \
+ int src_stride, \
+ int W, \
+ int H) \
+{ \
+ int x; \
+ int leading_pixels = 0, trailing_pixels = 0; \
+ const int TILE_SIZE = CACHE_LINE_SIZE / sizeof(pix_type); \
+ \
+ /* \
+ * split processing into handling destination as TILE_SIZExH cache line \
+ * aligned vertical stripes (optimistically assuming that destination \
+ * stride is a multiple of cache line, if not - it will be just a bit \
+ * slower) \
+ */ \
+ \
+ if ((uintptr_t)dst & (CACHE_LINE_SIZE - 1)) \
+ { \
+ leading_pixels = TILE_SIZE - (((uintptr_t)dst & \
+ (CACHE_LINE_SIZE - 1)) / sizeof(pix_type)); \
+ if (leading_pixels > W) \
+ leading_pixels = W; \
+ \
+ /* unaligned leading part NxH (where N < TILE_SIZE) */ \
+ blt_rotated_90_trivial_##suffix ( \
+ dst, \
+ dst_stride, \
+ src, \
+ src_stride, \
+ leading_pixels, \
+ H); \
+ \
+ dst += leading_pixels; \
+ src += leading_pixels * src_stride; \
+ W -= leading_pixels; \
+ } \
+ \
+ if ((uintptr_t)(dst + W) & (CACHE_LINE_SIZE - 1)) \
+ { \
+ trailing_pixels = (((uintptr_t)(dst + W) & \
+ (CACHE_LINE_SIZE - 1)) / sizeof(pix_type)); \
+ if (trailing_pixels > W) \
+ trailing_pixels = W; \
+ W -= trailing_pixels; \
+ } \
+ \
+ for (x = 0; x < W; x += TILE_SIZE) \
+ { \
+ /* aligned middle part TILE_SIZExH */ \
+ blt_rotated_90_trivial_##suffix ( \
+ dst + x, \
+ dst_stride, \
+ src + src_stride * x, \
+ src_stride, \
+ TILE_SIZE, \
+ H); \
+ } \
+ \
+ if (trailing_pixels) \
+ { \
+ /* unaligned trailing part NxH (where N < TILE_SIZE) */ \
+ blt_rotated_90_trivial_##suffix ( \
+ dst + W, \
+ dst_stride, \
+ src + W * src_stride, \
+ src_stride, \
+ trailing_pixels, \
+ H); \
+ } \
+} \
+ \
+static void \
+blt_rotated_270_##suffix (pix_type *dst, \
+ int dst_stride, \
+ const pix_type *src, \
+ int src_stride, \
+ int W, \
+ int H) \
+{ \
+ int x; \
+ int leading_pixels = 0, trailing_pixels = 0; \
+ const int TILE_SIZE = CACHE_LINE_SIZE / sizeof(pix_type); \
+ \
+ /* \
+ * split processing into handling destination as TILE_SIZExH cache line \
+ * aligned vertical stripes (optimistically assuming that destination \
+ * stride is a multiple of cache line, if not - it will be just a bit \
+ * slower) \
+ */ \
+ \
+ if ((uintptr_t)dst & (CACHE_LINE_SIZE - 1)) \
+ { \
+ leading_pixels = TILE_SIZE - (((uintptr_t)dst & \
+ (CACHE_LINE_SIZE - 1)) / sizeof(pix_type)); \
+ if (leading_pixels > W) \
+ leading_pixels = W; \
+ \
+ /* unaligned leading part NxH (where N < TILE_SIZE) */ \
+ blt_rotated_270_trivial_##suffix ( \
+ dst, \
+ dst_stride, \
+ src + src_stride * (W - leading_pixels), \
+ src_stride, \
+ leading_pixels, \
+ H); \
+ \
+ dst += leading_pixels; \
+ W -= leading_pixels; \
+ } \
+ \
+ if ((uintptr_t)(dst + W) & (CACHE_LINE_SIZE - 1)) \
+ { \
+ trailing_pixels = (((uintptr_t)(dst + W) & \
+ (CACHE_LINE_SIZE - 1)) / sizeof(pix_type)); \
+ if (trailing_pixels > W) \
+ trailing_pixels = W; \
+ W -= trailing_pixels; \
+ src += trailing_pixels * src_stride; \
+ } \
+ \
+ for (x = 0; x < W; x += TILE_SIZE) \
+ { \
+ /* aligned middle part TILE_SIZExH */ \
+ blt_rotated_270_trivial_##suffix ( \
+ dst + x, \
+ dst_stride, \
+ src + src_stride * (W - x - TILE_SIZE), \
+ src_stride, \
+ TILE_SIZE, \
+ H); \
+ } \
+ \
+ if (trailing_pixels) \
+ { \
+ /* unaligned trailing part NxH (where N < TILE_SIZE) */ \
+ blt_rotated_270_trivial_##suffix ( \
+ dst + W, \
+ dst_stride, \
+ src - trailing_pixels * src_stride, \
+ src_stride, \
+ trailing_pixels, \
+ H); \
+ } \
+} \
+ \
+static void \
+fast_composite_rotate_90_##suffix (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ pixman_image_t * src_image, \
+ pixman_image_t * mask_image, \
+ pixman_image_t * dst_image, \
+ int32_t src_x, \
+ int32_t src_y, \
+ int32_t mask_x, \
+ int32_t mask_y, \
+ int32_t dest_x, \
+ int32_t dest_y, \
+ int32_t width, \
+ int32_t height) \
+{ \
+ pix_type *dst_line; \
+ pix_type *src_line; \
+ int dst_stride, src_stride; \
+ int src_x_t, src_y_t; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, pix_type, \
+ dst_stride, dst_line, 1); \
+ src_x_t = -src_y + pixman_fixed_to_int ( \
+ src_image->common.transform->matrix[0][2] + \
+ pixman_fixed_1 / 2 - pixman_fixed_e) - height;\
+ src_y_t = src_x + pixman_fixed_to_int ( \
+ src_image->common.transform->matrix[1][2] + \
+ pixman_fixed_1 / 2 - pixman_fixed_e); \
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x_t, src_y_t, pix_type, \
+ src_stride, src_line, 1); \
+ blt_rotated_90_##suffix (dst_line, dst_stride, src_line, src_stride, \
+ width, height); \
+} \
+ \
+static void \
+fast_composite_rotate_270_##suffix (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ pixman_image_t * src_image, \
+ pixman_image_t * mask_image, \
+ pixman_image_t * dst_image, \
+ int32_t src_x, \
+ int32_t src_y, \
+ int32_t mask_x, \
+ int32_t mask_y, \
+ int32_t dest_x, \
+ int32_t dest_y, \
+ int32_t width, \
+ int32_t height) \
+{ \
+ pix_type *dst_line; \
+ pix_type *src_line; \
+ int dst_stride, src_stride; \
+ int src_x_t, src_y_t; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dest_x, dest_y, pix_type, \
+ dst_stride, dst_line, 1); \
+ src_x_t = src_y + pixman_fixed_to_int ( \
+ src_image->common.transform->matrix[0][2] + \
+ pixman_fixed_1 / 2 - pixman_fixed_e); \
+ src_y_t = -src_x + pixman_fixed_to_int ( \
+ src_image->common.transform->matrix[1][2] + \
+ pixman_fixed_1 / 2 - pixman_fixed_e) - width; \
+ PIXMAN_IMAGE_GET_LINE (src_image, src_x_t, src_y_t, pix_type, \
+ src_stride, src_line, 1); \
+ blt_rotated_270_##suffix (dst_line, dst_stride, src_line, src_stride, \
+ width, height); \
+}
+
+FAST_SIMPLE_ROTATE (8, uint8_t)
+FAST_SIMPLE_ROTATE (565, uint16_t)
+FAST_SIMPLE_ROTATE (8888, uint32_t)
+
+static const pixman_fast_path_t c_fast_paths[] =
+{
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, r5g6b5, fast_composite_over_n_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, b5g6r5, fast_composite_over_n_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, r8g8b8, fast_composite_over_n_8_0888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, b8g8r8, fast_composite_over_n_8_0888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8r8g8b8, fast_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8r8g8b8, fast_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8b8g8r8, fast_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8b8g8r8, fast_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a1, a8r8g8b8, fast_composite_over_n_1_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a1, x8r8g8b8, fast_composite_over_n_1_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a1, a8b8g8r8, fast_composite_over_n_1_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a1, x8b8g8r8, fast_composite_over_n_1_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a1, r5g6b5, fast_composite_over_n_1_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a1, b5g6r5, fast_composite_over_n_1_0565),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, a8r8g8b8, fast_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, x8r8g8b8, fast_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, r5g6b5, fast_composite_over_n_8888_0565_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, a8b8g8r8, fast_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, x8b8g8r8, fast_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, b5g6r5, fast_composite_over_n_8888_0565_ca),
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, a8, x8r8g8b8, fast_composite_over_x888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, a8, a8r8g8b8, fast_composite_over_x888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, a8, x8b8g8r8, fast_composite_over_x888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, a8, a8b8g8r8, fast_composite_over_x888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, a8r8g8b8, fast_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, x8r8g8b8, fast_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, r5g6b5, fast_composite_over_8888_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, a8b8g8r8, fast_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, x8b8g8r8, fast_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, b5g6r5, fast_composite_over_8888_0565),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, fast_composite_add_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, null, a8b8g8r8, fast_composite_add_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, fast_composite_add_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, a1, null, a1, fast_composite_add_1000_1000),
+ PIXMAN_STD_FAST_PATH_CA (ADD, solid, a8r8g8b8, a8r8g8b8, fast_composite_add_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8, fast_composite_add_n_8_8),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, a8r8g8b8, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, x8r8g8b8, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, a8b8g8r8, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, x8b8g8r8, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, a1, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, a8, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, solid, null, r5g6b5, fast_composite_solid_fill),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, a8r8g8b8, fast_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, a8b8g8r8, fast_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, x8r8g8b8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, a8r8g8b8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, x8r8g8b8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, x8b8g8r8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, a8b8g8r8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, x8b8g8r8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, b8g8r8a8, null, b8g8r8x8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, b8g8r8a8, null, b8g8r8a8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, b8g8r8x8, null, b8g8r8x8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, r5g6b5, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, b5g6r5, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, r8g8b8, null, r8g8b8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, b8g8r8, null, b8g8r8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, x1r5g5b5, null, x1r5g5b5, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, a1r5g5b5, null, x1r5g5b5, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, a8, null, a8, fast_composite_src_memcpy),
+ PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, r5g6b5, fast_composite_src_x888_0565),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, r5g6b5, fast_composite_src_x888_0565),
+ PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, b5g6r5, fast_composite_src_x888_0565),
+ PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, b5g6r5, fast_composite_src_x888_0565),
+ PIXMAN_STD_FAST_PATH (IN, a8, null, a8, fast_composite_in_8_8),
+ PIXMAN_STD_FAST_PATH (IN, solid, a8, a8, fast_composite_in_n_8_8),
+
+ SIMPLE_NEAREST_FAST_PATH (SRC, x8r8g8b8, x8r8g8b8, 8888_8888),
+ SIMPLE_NEAREST_FAST_PATH (SRC, a8r8g8b8, x8r8g8b8, 8888_8888),
+ SIMPLE_NEAREST_FAST_PATH (SRC, x8b8g8r8, x8b8g8r8, 8888_8888),
+ SIMPLE_NEAREST_FAST_PATH (SRC, a8b8g8r8, x8b8g8r8, 8888_8888),
+
+ SIMPLE_NEAREST_FAST_PATH (SRC, a8r8g8b8, a8r8g8b8, 8888_8888),
+ SIMPLE_NEAREST_FAST_PATH (SRC, a8b8g8r8, a8b8g8r8, 8888_8888),
+
+ SIMPLE_NEAREST_FAST_PATH (SRC, x8r8g8b8, r5g6b5, 8888_565),
+ SIMPLE_NEAREST_FAST_PATH (SRC, a8r8g8b8, r5g6b5, 8888_565),
+
+ SIMPLE_NEAREST_FAST_PATH (SRC, r5g6b5, r5g6b5, 565_565),
+
+ SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, 8888_8888),
+ SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8, 8888_8888),
+ SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, 8888_8888),
+ SIMPLE_NEAREST_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8, 8888_8888),
+
+ SIMPLE_NEAREST_FAST_PATH (OVER, a8r8g8b8, r5g6b5, 8888_565),
+
+#define NEAREST_FAST_PATH(op,s,d) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, SCALED_NEAREST_FLAGS, \
+ PIXMAN_null, 0, \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest, \
+ }
+
+ NEAREST_FAST_PATH (SRC, x8r8g8b8, x8r8g8b8),
+ NEAREST_FAST_PATH (SRC, a8r8g8b8, x8r8g8b8),
+ NEAREST_FAST_PATH (SRC, x8b8g8r8, x8b8g8r8),
+ NEAREST_FAST_PATH (SRC, a8b8g8r8, x8b8g8r8),
+
+ NEAREST_FAST_PATH (SRC, x8r8g8b8, a8r8g8b8),
+ NEAREST_FAST_PATH (SRC, a8r8g8b8, a8r8g8b8),
+ NEAREST_FAST_PATH (SRC, x8b8g8r8, a8b8g8r8),
+ NEAREST_FAST_PATH (SRC, a8b8g8r8, a8b8g8r8),
+
+ NEAREST_FAST_PATH (OVER, x8r8g8b8, x8r8g8b8),
+ NEAREST_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8),
+ NEAREST_FAST_PATH (OVER, x8b8g8r8, x8b8g8r8),
+ NEAREST_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8),
+
+ NEAREST_FAST_PATH (OVER, x8r8g8b8, a8r8g8b8),
+ NEAREST_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8),
+ NEAREST_FAST_PATH (OVER, x8b8g8r8, a8b8g8r8),
+ NEAREST_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8),
+
+#define SIMPLE_ROTATE_FLAGS(angle) \
+ (FAST_PATH_ROTATE_ ## angle ## _TRANSFORM | \
+ FAST_PATH_NEAREST_FILTER | \
+ FAST_PATH_SAMPLES_COVER_CLIP | \
+ FAST_PATH_STANDARD_FLAGS)
+
+#define SIMPLE_ROTATE_FAST_PATH(op,s,d,suffix) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, SIMPLE_ROTATE_FLAGS (90), \
+ PIXMAN_null, 0, \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_rotate_90_##suffix, \
+ }, \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, SIMPLE_ROTATE_FLAGS (270), \
+ PIXMAN_null, 0, \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_rotate_270_##suffix, \
+ }
+
+ SIMPLE_ROTATE_FAST_PATH (SRC, a8r8g8b8, a8r8g8b8, 8888),
+ SIMPLE_ROTATE_FAST_PATH (SRC, a8r8g8b8, x8r8g8b8, 8888),
+ SIMPLE_ROTATE_FAST_PATH (SRC, x8r8g8b8, x8r8g8b8, 8888),
+ SIMPLE_ROTATE_FAST_PATH (SRC, r5g6b5, r5g6b5, 565),
+ SIMPLE_ROTATE_FAST_PATH (SRC, a8, a8, 8),
+
+ { PIXMAN_OP_NONE },
+};
+
+#ifdef WORDS_BIGENDIAN
+#define A1_FILL_MASK(n, offs) (((1 << (n)) - 1) << (32 - (offs) - (n)))
+#else
+#define A1_FILL_MASK(n, offs) (((1 << (n)) - 1) << (offs))
+#endif
+
+static force_inline void
+pixman_fill1_line (uint32_t *dst, int offs, int width, int v)
+{
+ if (offs)
+ {
+ int leading_pixels = 32 - offs;
+ if (leading_pixels >= width)
+ {
+ if (v)
+ *dst |= A1_FILL_MASK (width, offs);
+ else
+ *dst &= ~A1_FILL_MASK (width, offs);
+ return;
+ }
+ else
+ {
+ if (v)
+ *dst++ |= A1_FILL_MASK (leading_pixels, offs);
+ else
+ *dst++ &= ~A1_FILL_MASK (leading_pixels, offs);
+ width -= leading_pixels;
+ }
+ }
+ while (width >= 32)
+ {
+ if (v)
+ *dst++ = 0xFFFFFFFF;
+ else
+ *dst++ = 0;
+ width -= 32;
+ }
+ if (width > 0)
+ {
+ if (v)
+ *dst |= A1_FILL_MASK (width, 0);
+ else
+ *dst &= ~A1_FILL_MASK (width, 0);
+ }
+}
+
+static void
+pixman_fill1 (uint32_t *bits,
+ int stride,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t xor)
+{
+ uint32_t *dst = bits + y * stride + (x >> 5);
+ int offs = x & 31;
+
+ if (xor & 1)
+ {
+ while (height--)
+ {
+ pixman_fill1_line (dst, offs, width, 1);
+ dst += stride;
+ }
+ }
+ else
+ {
+ while (height--)
+ {
+ pixman_fill1_line (dst, offs, width, 0);
+ dst += stride;
+ }
+ }
+}
+
+static void
+pixman_fill8 (uint32_t *bits,
+ int stride,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t xor)
+{
+ int byte_stride = stride * (int) sizeof (uint32_t);
+ uint8_t *dst = (uint8_t *) bits;
+ uint8_t v = xor & 0xff;
+ int i;
+
+ dst = dst + y * byte_stride + x;
+
+ while (height--)
+ {
+ for (i = 0; i < width; ++i)
+ dst[i] = v;
+
+ dst += byte_stride;
+ }
+}
+
+static void
+pixman_fill16 (uint32_t *bits,
+ int stride,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t xor)
+{
+ int short_stride =
+ (stride * (int)sizeof (uint32_t)) / (int)sizeof (uint16_t);
+ uint16_t *dst = (uint16_t *)bits;
+ uint16_t v = xor & 0xffff;
+ int i;
+
+ dst = dst + y * short_stride + x;
+
+ while (height--)
+ {
+ for (i = 0; i < width; ++i)
+ dst[i] = v;
+
+ dst += short_stride;
+ }
+}
+
+static void
+pixman_fill32 (uint32_t *bits,
+ int stride,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t xor)
+{
+ int i;
+
+ bits = bits + y * stride + x;
+
+ while (height--)
+ {
+ for (i = 0; i < width; ++i)
+ bits[i] = xor;
+
+ bits += stride;
+ }
+}
+
+static pixman_bool_t
+fast_path_fill (pixman_implementation_t *imp,
+ uint32_t * bits,
+ int stride,
+ int bpp,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t xor)
+{
+ switch (bpp)
+ {
+ case 1:
+ pixman_fill1 (bits, stride, x, y, width, height, xor);
+ break;
+
+ case 8:
+ pixman_fill8 (bits, stride, x, y, width, height, xor);
+ break;
+
+ case 16:
+ pixman_fill16 (bits, stride, x, y, width, height, xor);
+ break;
+
+ case 32:
+ pixman_fill32 (bits, stride, x, y, width, height, xor);
+ break;
+
+ default:
+ return _pixman_implementation_fill (
+ imp->delegate, bits, stride, bpp, x, y, width, height, xor);
+ break;
+ }
+
+ return TRUE;
+}
+
+pixman_implementation_t *
+_pixman_implementation_create_fast_path (pixman_implementation_t *fallback)
+{
+ pixman_implementation_t *imp = _pixman_implementation_create (fallback, c_fast_paths);
+
+ imp->fill = fast_path_fill;
+
+ return imp;
+}
diff --git a/pixman/pixman/pixman-fast-path.h b/pixman/pixman/pixman-fast-path.h index d08122293..b5f524353 100644 --- a/pixman/pixman/pixman-fast-path.h +++ b/pixman/pixman/pixman-fast-path.h @@ -1,590 +1,590 @@ -/* -*- Mode: c; c-basic-offset: 4; tab-width: 8; indent-tabs-mode: t; -*- */ -/* - * Copyright © 2000 SuSE, Inc. - * Copyright © 2007 Red Hat, Inc. - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of SuSE not be used in advertising or - * publicity pertaining to distribution of the software without specific, - * written prior permission. SuSE makes no representations about the - * suitability of this software for any purpose. It is provided "as is" - * without express or implied warranty. - * - * SuSE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL - * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL SuSE - * BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES - * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION - * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN - * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. - * - * Author: Keith Packard, SuSE, Inc. - */ - -#ifndef PIXMAN_FAST_PATH_H__ -#define PIXMAN_FAST_PATH_H__ - -#include "pixman-private.h" - -#define PIXMAN_REPEAT_COVER -1 - -static force_inline pixman_bool_t -repeat (pixman_repeat_t repeat, int *c, int size) -{ - if (repeat == PIXMAN_REPEAT_NONE) - { - if (*c < 0 || *c >= size) - return FALSE; - } - else if (repeat == PIXMAN_REPEAT_NORMAL) - { - while (*c >= size) - *c -= size; - while (*c < 0) - *c += size; - } - else if (repeat == PIXMAN_REPEAT_PAD) - { - *c = CLIP (*c, 0, size - 1); - } - else /* REFLECT */ - { - *c = MOD (*c, size * 2); - if (*c >= size) - *c = size * 2 - *c - 1; - } - return TRUE; -} - -/* - * For each scanline fetched from source image with PAD repeat: - * - calculate how many pixels need to be padded on the left side - * - calculate how many pixels need to be padded on the right side - * - update width to only count pixels which are fetched from the image - * All this information is returned via 'width', 'left_pad', 'right_pad' - * arguments. The code is assuming that 'unit_x' is positive. - * - * Note: 64-bit math is used in order to avoid potential overflows, which - * is probably excessive in many cases. This particular function - * may need its own correctness test and performance tuning. - */ -static force_inline void -pad_repeat_get_scanline_bounds (int32_t source_image_width, - pixman_fixed_t vx, - pixman_fixed_t unit_x, - int32_t * width, - int32_t * left_pad, - int32_t * right_pad) -{ - int64_t max_vx = (int64_t) source_image_width << 16; - int64_t tmp; - if (vx < 0) - { - tmp = ((int64_t) unit_x - 1 - vx) / unit_x; - if (tmp > *width) - { - *left_pad = *width; - *width = 0; - } - else - { - *left_pad = (int32_t) tmp; - *width -= (int32_t) tmp; - } - } - else - { - *left_pad = 0; - } - tmp = ((int64_t) unit_x - 1 - vx + max_vx) / unit_x - *left_pad; - if (tmp < 0) - { - *right_pad = *width; - *width = 0; - } - else if (tmp >= *width) - { - *right_pad = 0; - } - else - { - *right_pad = *width - (int32_t) tmp; - *width = (int32_t) tmp; - } -} - -/* A macroified version of specialized nearest scalers for some - * common 8888 and 565 formats. It supports SRC and OVER ops. - * - * There are two repeat versions, one that handles repeat normal, - * and one without repeat handling that only works if the src region - * used is completely covered by the pre-repeated source samples. - * - * The loops are unrolled to process two pixels per iteration for better - * performance on most CPU architectures (superscalar processors - * can issue several operations simultaneously, other processors can hide - * instructions latencies by pipelining operations). Unrolling more - * does not make much sense because the compiler will start running out - * of spare registers soon. - */ - -#define GET_8888_ALPHA(s) ((s) >> 24) - /* This is not actually used since we don't have an OVER with - 565 source, but it is needed to build. */ -#define GET_0565_ALPHA(s) 0xff - -#define FAST_NEAREST_SCANLINE(scanline_func_name, SRC_FORMAT, DST_FORMAT, \ - src_type_t, dst_type_t, OP, repeat_mode) \ -static force_inline void \ -scanline_func_name (dst_type_t *dst, \ - const src_type_t *src, \ - int32_t w, \ - pixman_fixed_t vx, \ - pixman_fixed_t unit_x, \ - pixman_fixed_t max_vx, \ - pixman_bool_t fully_transparent_src) \ -{ \ - uint32_t d; \ - src_type_t s1, s2; \ - uint8_t a1, a2; \ - int x1, x2; \ - \ - if (PIXMAN_OP_ ## OP == PIXMAN_OP_OVER && fully_transparent_src) \ - return; \ - \ - if (PIXMAN_OP_ ## OP != PIXMAN_OP_SRC && PIXMAN_OP_ ## OP != PIXMAN_OP_OVER) \ - abort(); \ - \ - while ((w -= 2) >= 0) \ - { \ - x1 = vx >> 16; \ - vx += unit_x; \ - if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NORMAL) \ - { \ - /* This works because we know that unit_x is positive */ \ - while (vx >= max_vx) \ - vx -= max_vx; \ - } \ - s1 = src[x1]; \ - \ - x2 = vx >> 16; \ - vx += unit_x; \ - if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NORMAL) \ - { \ - /* This works because we know that unit_x is positive */ \ - while (vx >= max_vx) \ - vx -= max_vx; \ - } \ - s2 = src[x2]; \ - \ - if (PIXMAN_OP_ ## OP == PIXMAN_OP_OVER) \ - { \ - a1 = GET_ ## SRC_FORMAT ## _ALPHA(s1); \ - a2 = GET_ ## SRC_FORMAT ## _ALPHA(s2); \ - \ - if (a1 == 0xff) \ - { \ - *dst = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s1); \ - } \ - else if (s1) \ - { \ - d = CONVERT_ ## DST_FORMAT ## _TO_8888 (*dst); \ - s1 = CONVERT_ ## SRC_FORMAT ## _TO_8888 (s1); \ - a1 ^= 0xff; \ - UN8x4_MUL_UN8_ADD_UN8x4 (d, a1, s1); \ - *dst = CONVERT_8888_TO_ ## DST_FORMAT (d); \ - } \ - dst++; \ - \ - if (a2 == 0xff) \ - { \ - *dst = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s2); \ - } \ - else if (s2) \ - { \ - d = CONVERT_## DST_FORMAT ## _TO_8888 (*dst); \ - s2 = CONVERT_## SRC_FORMAT ## _TO_8888 (s2); \ - a2 ^= 0xff; \ - UN8x4_MUL_UN8_ADD_UN8x4 (d, a2, s2); \ - *dst = CONVERT_8888_TO_ ## DST_FORMAT (d); \ - } \ - dst++; \ - } \ - else /* PIXMAN_OP_SRC */ \ - { \ - *dst++ = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s1); \ - *dst++ = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s2); \ - } \ - } \ - \ - if (w & 1) \ - { \ - x1 = vx >> 16; \ - s1 = src[x1]; \ - \ - if (PIXMAN_OP_ ## OP == PIXMAN_OP_OVER) \ - { \ - a1 = GET_ ## SRC_FORMAT ## _ALPHA(s1); \ - \ - if (a1 == 0xff) \ - { \ - *dst = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s1); \ - } \ - else if (s1) \ - { \ - d = CONVERT_## DST_FORMAT ## _TO_8888 (*dst); \ - s1 = CONVERT_ ## SRC_FORMAT ## _TO_8888 (s1); \ - a1 ^= 0xff; \ - UN8x4_MUL_UN8_ADD_UN8x4 (d, a1, s1); \ - *dst = CONVERT_8888_TO_ ## DST_FORMAT (d); \ - } \ - dst++; \ - } \ - else /* PIXMAN_OP_SRC */ \ - { \ - *dst++ = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s1); \ - } \ - } \ -} - -#define FAST_NEAREST_MAINLOOP_INT(scale_func_name, scanline_func, src_type_t, mask_type_t, \ - dst_type_t, repeat_mode, have_mask, mask_is_solid) \ -static void \ -fast_composite_scaled_nearest ## scale_func_name (pixman_implementation_t *imp, \ - pixman_op_t op, \ - pixman_image_t * src_image, \ - pixman_image_t * mask_image, \ - pixman_image_t * dst_image, \ - int32_t src_x, \ - int32_t src_y, \ - int32_t mask_x, \ - int32_t mask_y, \ - int32_t dst_x, \ - int32_t dst_y, \ - int32_t width, \ - int32_t height) \ -{ \ - dst_type_t *dst_line; \ - mask_type_t *mask_line; \ - src_type_t *src_first_line; \ - int y; \ - pixman_fixed_t max_vx = INT32_MAX; /* suppress uninitialized variable warning */ \ - pixman_fixed_t max_vy; \ - pixman_vector_t v; \ - pixman_fixed_t vx, vy; \ - pixman_fixed_t unit_x, unit_y; \ - int32_t left_pad, right_pad; \ - \ - src_type_t *src; \ - dst_type_t *dst; \ - mask_type_t solid_mask; \ - const mask_type_t *mask = &solid_mask; \ - int src_stride, mask_stride, dst_stride; \ - \ - PIXMAN_IMAGE_GET_LINE (dst_image, dst_x, dst_y, dst_type_t, dst_stride, dst_line, 1); \ - if (have_mask) \ - { \ - if (mask_is_solid) \ - solid_mask = _pixman_image_get_solid (imp, mask_image, dst_image->bits.format); \ - else \ - PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, mask_type_t, \ - mask_stride, mask_line, 1); \ - } \ - /* pass in 0 instead of src_x and src_y because src_x and src_y need to be \ - * transformed from destination space to source space */ \ - PIXMAN_IMAGE_GET_LINE (src_image, 0, 0, src_type_t, src_stride, src_first_line, 1); \ - \ - /* reference point is the center of the pixel */ \ - v.vector[0] = pixman_int_to_fixed (src_x) + pixman_fixed_1 / 2; \ - v.vector[1] = pixman_int_to_fixed (src_y) + pixman_fixed_1 / 2; \ - v.vector[2] = pixman_fixed_1; \ - \ - if (!pixman_transform_point_3d (src_image->common.transform, &v)) \ - return; \ - \ - unit_x = src_image->common.transform->matrix[0][0]; \ - unit_y = src_image->common.transform->matrix[1][1]; \ - \ - /* Round down to closest integer, ensuring that 0.5 rounds to 0, not 1 */ \ - v.vector[0] -= pixman_fixed_e; \ - v.vector[1] -= pixman_fixed_e; \ - \ - vx = v.vector[0]; \ - vy = v.vector[1]; \ - \ - if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NORMAL) \ - { \ - /* Clamp repeating positions inside the actual samples */ \ - max_vx = src_image->bits.width << 16; \ - max_vy = src_image->bits.height << 16; \ - \ - repeat (PIXMAN_REPEAT_NORMAL, &vx, max_vx); \ - repeat (PIXMAN_REPEAT_NORMAL, &vy, max_vy); \ - } \ - \ - if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_PAD || \ - PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NONE) \ - { \ - pad_repeat_get_scanline_bounds (src_image->bits.width, vx, unit_x, \ - &width, &left_pad, &right_pad); \ - vx += left_pad * unit_x; \ - } \ - \ - while (--height >= 0) \ - { \ - dst = dst_line; \ - dst_line += dst_stride; \ - if (have_mask && !mask_is_solid) \ - { \ - mask = mask_line; \ - mask_line += mask_stride; \ - } \ - \ - y = vy >> 16; \ - vy += unit_y; \ - if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NORMAL) \ - repeat (PIXMAN_REPEAT_NORMAL, &vy, max_vy); \ - if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_PAD) \ - { \ - repeat (PIXMAN_REPEAT_PAD, &y, src_image->bits.height); \ - src = src_first_line + src_stride * y; \ - if (left_pad > 0) \ - { \ - scanline_func (mask, dst, src, left_pad, 0, 0, 0, FALSE); \ - } \ - if (width > 0) \ - { \ - scanline_func (mask + (mask_is_solid ? 0 : left_pad), \ - dst + left_pad, src, width, vx, unit_x, 0, FALSE); \ - } \ - if (right_pad > 0) \ - { \ - scanline_func (mask + (mask_is_solid ? 0 : left_pad + width), \ - dst + left_pad + width, src + src_image->bits.width - 1, \ - right_pad, 0, 0, 0, FALSE); \ - } \ - } \ - else if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NONE) \ - { \ - static const src_type_t zero[1] = { 0 }; \ - if (y < 0 || y >= src_image->bits.height) \ - { \ - scanline_func (mask, dst, zero, left_pad + width + right_pad, 0, 0, 0, TRUE); \ - continue; \ - } \ - src = src_first_line + src_stride * y; \ - if (left_pad > 0) \ - { \ - scanline_func (mask, dst, zero, left_pad, 0, 0, 0, TRUE); \ - } \ - if (width > 0) \ - { \ - scanline_func (mask + (mask_is_solid ? 0 : left_pad), \ - dst + left_pad, src, width, vx, unit_x, 0, FALSE); \ - } \ - if (right_pad > 0) \ - { \ - scanline_func (mask + (mask_is_solid ? 0 : left_pad + width), \ - dst + left_pad + width, zero, right_pad, 0, 0, 0, TRUE); \ - } \ - } \ - else \ - { \ - src = src_first_line + src_stride * y; \ - scanline_func (mask, dst, src, width, vx, unit_x, max_vx, FALSE); \ - } \ - } \ -} - -/* A workaround for old sun studio, see: https://bugs.freedesktop.org/show_bug.cgi?id=32764 */ -#define FAST_NEAREST_MAINLOOP_COMMON(scale_func_name, scanline_func, src_type_t, mask_type_t, \ - dst_type_t, repeat_mode, have_mask, mask_is_solid) \ - FAST_NEAREST_MAINLOOP_INT(_ ## scale_func_name, scanline_func, src_type_t, mask_type_t, \ - dst_type_t, repeat_mode, have_mask, mask_is_solid) - -#define FAST_NEAREST_MAINLOOP_NOMASK(scale_func_name, scanline_func, src_type_t, dst_type_t, \ - repeat_mode) \ - static force_inline void \ - scanline_func##scale_func_name##_wrapper ( \ - const uint8_t *mask, \ - dst_type_t *dst, \ - const src_type_t *src, \ - int32_t w, \ - pixman_fixed_t vx, \ - pixman_fixed_t unit_x, \ - pixman_fixed_t max_vx, \ - pixman_bool_t fully_transparent_src) \ - { \ - scanline_func (dst, src, w, vx, unit_x, max_vx, fully_transparent_src); \ - } \ - FAST_NEAREST_MAINLOOP_INT (scale_func_name, scanline_func##scale_func_name##_wrapper, \ - src_type_t, uint8_t, dst_type_t, repeat_mode, FALSE, FALSE) - -#define FAST_NEAREST_MAINLOOP(scale_func_name, scanline_func, src_type_t, dst_type_t, \ - repeat_mode) \ - FAST_NEAREST_MAINLOOP_NOMASK(_ ## scale_func_name, scanline_func, src_type_t, \ - dst_type_t, repeat_mode) - -#define FAST_NEAREST(scale_func_name, SRC_FORMAT, DST_FORMAT, \ - src_type_t, dst_type_t, OP, repeat_mode) \ - FAST_NEAREST_SCANLINE(scaled_nearest_scanline_ ## scale_func_name ## _ ## OP, \ - SRC_FORMAT, DST_FORMAT, src_type_t, dst_type_t, \ - OP, repeat_mode) \ - FAST_NEAREST_MAINLOOP_NOMASK(_ ## scale_func_name ## _ ## OP, \ - scaled_nearest_scanline_ ## scale_func_name ## _ ## OP, \ - src_type_t, dst_type_t, repeat_mode) - - -#define SCALED_NEAREST_FLAGS \ - (FAST_PATH_SCALE_TRANSFORM | \ - FAST_PATH_NO_ALPHA_MAP | \ - FAST_PATH_NEAREST_FILTER | \ - FAST_PATH_NO_ACCESSORS | \ - FAST_PATH_NARROW_FORMAT) - -#define SIMPLE_NEAREST_FAST_PATH_NORMAL(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_NORMAL_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_null, 0, \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _normal ## _ ## op, \ - } - -#define SIMPLE_NEAREST_FAST_PATH_PAD(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_PAD_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_null, 0, \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _pad ## _ ## op, \ - } - -#define SIMPLE_NEAREST_FAST_PATH_NONE(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_NONE_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_null, 0, \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _none ## _ ## op, \ - } - -#define SIMPLE_NEAREST_FAST_PATH_COVER(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - SCALED_NEAREST_FLAGS | FAST_PATH_SAMPLES_COVER_CLIP, \ - PIXMAN_null, 0, \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _cover ## _ ## op, \ - } - -#define SIMPLE_NEAREST_A8_MASK_FAST_PATH_NORMAL(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_NORMAL_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_a8, MASK_FLAGS (a8, FAST_PATH_UNIFIED_ALPHA), \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _normal ## _ ## op, \ - } - -#define SIMPLE_NEAREST_A8_MASK_FAST_PATH_PAD(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_PAD_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_a8, MASK_FLAGS (a8, FAST_PATH_UNIFIED_ALPHA), \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _pad ## _ ## op, \ - } - -#define SIMPLE_NEAREST_A8_MASK_FAST_PATH_NONE(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_NONE_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_a8, MASK_FLAGS (a8, FAST_PATH_UNIFIED_ALPHA), \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _none ## _ ## op, \ - } - -#define SIMPLE_NEAREST_A8_MASK_FAST_PATH_COVER(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - SCALED_NEAREST_FLAGS | FAST_PATH_SAMPLES_COVER_CLIP, \ - PIXMAN_a8, MASK_FLAGS (a8, FAST_PATH_UNIFIED_ALPHA), \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _cover ## _ ## op, \ - } - -#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_NORMAL(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_NORMAL_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_solid, MASK_FLAGS (solid, FAST_PATH_UNIFIED_ALPHA), \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _normal ## _ ## op, \ - } - -#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_PAD(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_PAD_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_solid, MASK_FLAGS (solid, FAST_PATH_UNIFIED_ALPHA), \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _pad ## _ ## op, \ - } - -#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_NONE(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - (SCALED_NEAREST_FLAGS | \ - FAST_PATH_NONE_REPEAT | \ - FAST_PATH_X_UNIT_POSITIVE), \ - PIXMAN_solid, MASK_FLAGS (solid, FAST_PATH_UNIFIED_ALPHA), \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _none ## _ ## op, \ - } - -#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_COVER(op,s,d,func) \ - { PIXMAN_OP_ ## op, \ - PIXMAN_ ## s, \ - SCALED_NEAREST_FLAGS | FAST_PATH_SAMPLES_COVER_CLIP, \ - PIXMAN_solid, MASK_FLAGS (solid, FAST_PATH_UNIFIED_ALPHA), \ - PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \ - fast_composite_scaled_nearest_ ## func ## _cover ## _ ## op, \ - } - -/* Prefer the use of 'cover' variant, because it is faster */ -#define SIMPLE_NEAREST_FAST_PATH(op,s,d,func) \ - SIMPLE_NEAREST_FAST_PATH_COVER (op,s,d,func), \ - SIMPLE_NEAREST_FAST_PATH_NONE (op,s,d,func), \ - SIMPLE_NEAREST_FAST_PATH_PAD (op,s,d,func), \ - SIMPLE_NEAREST_FAST_PATH_NORMAL (op,s,d,func) - -#define SIMPLE_NEAREST_A8_MASK_FAST_PATH(op,s,d,func) \ - SIMPLE_NEAREST_A8_MASK_FAST_PATH_COVER (op,s,d,func), \ - SIMPLE_NEAREST_A8_MASK_FAST_PATH_NONE (op,s,d,func), \ - SIMPLE_NEAREST_A8_MASK_FAST_PATH_PAD (op,s,d,func) - -#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH(op,s,d,func) \ - SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_COVER (op,s,d,func), \ - SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_NONE (op,s,d,func), \ - SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_PAD (op,s,d,func) - -#endif +/* -*- Mode: c; c-basic-offset: 4; tab-width: 8; indent-tabs-mode: t; -*- */
+/*
+ * Copyright © 2000 SuSE, Inc.
+ * Copyright © 2007 Red Hat, Inc.
+ *
+ * Permission to use, copy, modify, distribute, and sell this software and its
+ * documentation for any purpose is hereby granted without fee, provided that
+ * the above copyright notice appear in all copies and that both that
+ * copyright notice and this permission notice appear in supporting
+ * documentation, and that the name of SuSE not be used in advertising or
+ * publicity pertaining to distribution of the software without specific,
+ * written prior permission. SuSE makes no representations about the
+ * suitability of this software for any purpose. It is provided "as is"
+ * without express or implied warranty.
+ *
+ * SuSE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL SuSE
+ * BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+ * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * Author: Keith Packard, SuSE, Inc.
+ */
+
+#ifndef PIXMAN_FAST_PATH_H__
+#define PIXMAN_FAST_PATH_H__
+
+#include "pixman-private.h"
+
+#define PIXMAN_REPEAT_COVER -1
+
+static force_inline pixman_bool_t
+repeat (pixman_repeat_t repeat, int *c, int size)
+{
+ if (repeat == PIXMAN_REPEAT_NONE)
+ {
+ if (*c < 0 || *c >= size)
+ return FALSE;
+ }
+ else if (repeat == PIXMAN_REPEAT_NORMAL)
+ {
+ while (*c >= size)
+ *c -= size;
+ while (*c < 0)
+ *c += size;
+ }
+ else if (repeat == PIXMAN_REPEAT_PAD)
+ {
+ *c = CLIP (*c, 0, size - 1);
+ }
+ else /* REFLECT */
+ {
+ *c = MOD (*c, size * 2);
+ if (*c >= size)
+ *c = size * 2 - *c - 1;
+ }
+ return TRUE;
+}
+
+/*
+ * For each scanline fetched from source image with PAD repeat:
+ * - calculate how many pixels need to be padded on the left side
+ * - calculate how many pixels need to be padded on the right side
+ * - update width to only count pixels which are fetched from the image
+ * All this information is returned via 'width', 'left_pad', 'right_pad'
+ * arguments. The code is assuming that 'unit_x' is positive.
+ *
+ * Note: 64-bit math is used in order to avoid potential overflows, which
+ * is probably excessive in many cases. This particular function
+ * may need its own correctness test and performance tuning.
+ */
+static force_inline void
+pad_repeat_get_scanline_bounds (int32_t source_image_width,
+ pixman_fixed_t vx,
+ pixman_fixed_t unit_x,
+ int32_t * width,
+ int32_t * left_pad,
+ int32_t * right_pad)
+{
+ int64_t max_vx = (int64_t) source_image_width << 16;
+ int64_t tmp;
+ if (vx < 0)
+ {
+ tmp = ((int64_t) unit_x - 1 - vx) / unit_x;
+ if (tmp > *width)
+ {
+ *left_pad = *width;
+ *width = 0;
+ }
+ else
+ {
+ *left_pad = (int32_t) tmp;
+ *width -= (int32_t) tmp;
+ }
+ }
+ else
+ {
+ *left_pad = 0;
+ }
+ tmp = ((int64_t) unit_x - 1 - vx + max_vx) / unit_x - *left_pad;
+ if (tmp < 0)
+ {
+ *right_pad = *width;
+ *width = 0;
+ }
+ else if (tmp >= *width)
+ {
+ *right_pad = 0;
+ }
+ else
+ {
+ *right_pad = *width - (int32_t) tmp;
+ *width = (int32_t) tmp;
+ }
+}
+
+/* A macroified version of specialized nearest scalers for some
+ * common 8888 and 565 formats. It supports SRC and OVER ops.
+ *
+ * There are two repeat versions, one that handles repeat normal,
+ * and one without repeat handling that only works if the src region
+ * used is completely covered by the pre-repeated source samples.
+ *
+ * The loops are unrolled to process two pixels per iteration for better
+ * performance on most CPU architectures (superscalar processors
+ * can issue several operations simultaneously, other processors can hide
+ * instructions latencies by pipelining operations). Unrolling more
+ * does not make much sense because the compiler will start running out
+ * of spare registers soon.
+ */
+
+#define GET_8888_ALPHA(s) ((s) >> 24)
+ /* This is not actually used since we don't have an OVER with
+ 565 source, but it is needed to build. */
+#define GET_0565_ALPHA(s) 0xff
+
+#define FAST_NEAREST_SCANLINE(scanline_func_name, SRC_FORMAT, DST_FORMAT, \
+ src_type_t, dst_type_t, OP, repeat_mode) \
+static force_inline void \
+scanline_func_name (dst_type_t *dst, \
+ const src_type_t *src, \
+ int32_t w, \
+ pixman_fixed_t vx, \
+ pixman_fixed_t unit_x, \
+ pixman_fixed_t max_vx, \
+ pixman_bool_t fully_transparent_src) \
+{ \
+ uint32_t d; \
+ src_type_t s1, s2; \
+ uint8_t a1, a2; \
+ int x1, x2; \
+ \
+ if (PIXMAN_OP_ ## OP == PIXMAN_OP_OVER && fully_transparent_src) \
+ return; \
+ \
+ if (PIXMAN_OP_ ## OP != PIXMAN_OP_SRC && PIXMAN_OP_ ## OP != PIXMAN_OP_OVER) \
+ abort(); \
+ \
+ while ((w -= 2) >= 0) \
+ { \
+ x1 = vx >> 16; \
+ vx += unit_x; \
+ if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NORMAL) \
+ { \
+ /* This works because we know that unit_x is positive */ \
+ while (vx >= max_vx) \
+ vx -= max_vx; \
+ } \
+ s1 = src[x1]; \
+ \
+ x2 = vx >> 16; \
+ vx += unit_x; \
+ if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NORMAL) \
+ { \
+ /* This works because we know that unit_x is positive */ \
+ while (vx >= max_vx) \
+ vx -= max_vx; \
+ } \
+ s2 = src[x2]; \
+ \
+ if (PIXMAN_OP_ ## OP == PIXMAN_OP_OVER) \
+ { \
+ a1 = GET_ ## SRC_FORMAT ## _ALPHA(s1); \
+ a2 = GET_ ## SRC_FORMAT ## _ALPHA(s2); \
+ \
+ if (a1 == 0xff) \
+ { \
+ *dst = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s1); \
+ } \
+ else if (s1) \
+ { \
+ d = CONVERT_ ## DST_FORMAT ## _TO_8888 (*dst); \
+ s1 = CONVERT_ ## SRC_FORMAT ## _TO_8888 (s1); \
+ a1 ^= 0xff; \
+ UN8x4_MUL_UN8_ADD_UN8x4 (d, a1, s1); \
+ *dst = CONVERT_8888_TO_ ## DST_FORMAT (d); \
+ } \
+ dst++; \
+ \
+ if (a2 == 0xff) \
+ { \
+ *dst = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s2); \
+ } \
+ else if (s2) \
+ { \
+ d = CONVERT_## DST_FORMAT ## _TO_8888 (*dst); \
+ s2 = CONVERT_## SRC_FORMAT ## _TO_8888 (s2); \
+ a2 ^= 0xff; \
+ UN8x4_MUL_UN8_ADD_UN8x4 (d, a2, s2); \
+ *dst = CONVERT_8888_TO_ ## DST_FORMAT (d); \
+ } \
+ dst++; \
+ } \
+ else /* PIXMAN_OP_SRC */ \
+ { \
+ *dst++ = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s1); \
+ *dst++ = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s2); \
+ } \
+ } \
+ \
+ if (w & 1) \
+ { \
+ x1 = vx >> 16; \
+ s1 = src[x1]; \
+ \
+ if (PIXMAN_OP_ ## OP == PIXMAN_OP_OVER) \
+ { \
+ a1 = GET_ ## SRC_FORMAT ## _ALPHA(s1); \
+ \
+ if (a1 == 0xff) \
+ { \
+ *dst = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s1); \
+ } \
+ else if (s1) \
+ { \
+ d = CONVERT_## DST_FORMAT ## _TO_8888 (*dst); \
+ s1 = CONVERT_ ## SRC_FORMAT ## _TO_8888 (s1); \
+ a1 ^= 0xff; \
+ UN8x4_MUL_UN8_ADD_UN8x4 (d, a1, s1); \
+ *dst = CONVERT_8888_TO_ ## DST_FORMAT (d); \
+ } \
+ dst++; \
+ } \
+ else /* PIXMAN_OP_SRC */ \
+ { \
+ *dst++ = CONVERT_ ## SRC_FORMAT ## _TO_ ## DST_FORMAT (s1); \
+ } \
+ } \
+}
+
+#define FAST_NEAREST_MAINLOOP_INT(scale_func_name, scanline_func, src_type_t, mask_type_t, \
+ dst_type_t, repeat_mode, have_mask, mask_is_solid) \
+static void \
+fast_composite_scaled_nearest ## scale_func_name (pixman_implementation_t *imp, \
+ pixman_op_t op, \
+ pixman_image_t * src_image, \
+ pixman_image_t * mask_image, \
+ pixman_image_t * dst_image, \
+ int32_t src_x, \
+ int32_t src_y, \
+ int32_t mask_x, \
+ int32_t mask_y, \
+ int32_t dst_x, \
+ int32_t dst_y, \
+ int32_t width, \
+ int32_t height) \
+{ \
+ dst_type_t *dst_line; \
+ mask_type_t *mask_line; \
+ src_type_t *src_first_line; \
+ int y; \
+ pixman_fixed_t max_vx = INT32_MAX; /* suppress uninitialized variable warning */ \
+ pixman_fixed_t max_vy; \
+ pixman_vector_t v; \
+ pixman_fixed_t vx, vy; \
+ pixman_fixed_t unit_x, unit_y; \
+ int32_t left_pad, right_pad; \
+ \
+ src_type_t *src; \
+ dst_type_t *dst; \
+ mask_type_t solid_mask; \
+ const mask_type_t *mask = &solid_mask; \
+ int src_stride, mask_stride, dst_stride; \
+ \
+ PIXMAN_IMAGE_GET_LINE (dst_image, dst_x, dst_y, dst_type_t, dst_stride, dst_line, 1); \
+ if (have_mask) \
+ { \
+ if (mask_is_solid) \
+ solid_mask = _pixman_image_get_solid (imp, mask_image, dst_image->bits.format); \
+ else \
+ PIXMAN_IMAGE_GET_LINE (mask_image, mask_x, mask_y, mask_type_t, \
+ mask_stride, mask_line, 1); \
+ } \
+ /* pass in 0 instead of src_x and src_y because src_x and src_y need to be \
+ * transformed from destination space to source space */ \
+ PIXMAN_IMAGE_GET_LINE (src_image, 0, 0, src_type_t, src_stride, src_first_line, 1); \
+ \
+ /* reference point is the center of the pixel */ \
+ v.vector[0] = pixman_int_to_fixed (src_x) + pixman_fixed_1 / 2; \
+ v.vector[1] = pixman_int_to_fixed (src_y) + pixman_fixed_1 / 2; \
+ v.vector[2] = pixman_fixed_1; \
+ \
+ if (!pixman_transform_point_3d (src_image->common.transform, &v)) \
+ return; \
+ \
+ unit_x = src_image->common.transform->matrix[0][0]; \
+ unit_y = src_image->common.transform->matrix[1][1]; \
+ \
+ /* Round down to closest integer, ensuring that 0.5 rounds to 0, not 1 */ \
+ v.vector[0] -= pixman_fixed_e; \
+ v.vector[1] -= pixman_fixed_e; \
+ \
+ vx = v.vector[0]; \
+ vy = v.vector[1]; \
+ \
+ if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NORMAL) \
+ { \
+ /* Clamp repeating positions inside the actual samples */ \
+ max_vx = src_image->bits.width << 16; \
+ max_vy = src_image->bits.height << 16; \
+ \
+ repeat (PIXMAN_REPEAT_NORMAL, &vx, max_vx); \
+ repeat (PIXMAN_REPEAT_NORMAL, &vy, max_vy); \
+ } \
+ \
+ if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_PAD || \
+ PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NONE) \
+ { \
+ pad_repeat_get_scanline_bounds (src_image->bits.width, vx, unit_x, \
+ &width, &left_pad, &right_pad); \
+ vx += left_pad * unit_x; \
+ } \
+ \
+ while (--height >= 0) \
+ { \
+ dst = dst_line; \
+ dst_line += dst_stride; \
+ if (have_mask && !mask_is_solid) \
+ { \
+ mask = mask_line; \
+ mask_line += mask_stride; \
+ } \
+ \
+ y = vy >> 16; \
+ vy += unit_y; \
+ if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NORMAL) \
+ repeat (PIXMAN_REPEAT_NORMAL, &vy, max_vy); \
+ if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_PAD) \
+ { \
+ repeat (PIXMAN_REPEAT_PAD, &y, src_image->bits.height); \
+ src = src_first_line + src_stride * y; \
+ if (left_pad > 0) \
+ { \
+ scanline_func (mask, dst, src, left_pad, 0, 0, 0, FALSE); \
+ } \
+ if (width > 0) \
+ { \
+ scanline_func (mask + (mask_is_solid ? 0 : left_pad), \
+ dst + left_pad, src, width, vx, unit_x, 0, FALSE); \
+ } \
+ if (right_pad > 0) \
+ { \
+ scanline_func (mask + (mask_is_solid ? 0 : left_pad + width), \
+ dst + left_pad + width, src + src_image->bits.width - 1, \
+ right_pad, 0, 0, 0, FALSE); \
+ } \
+ } \
+ else if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NONE) \
+ { \
+ static const src_type_t zero[1] = { 0 }; \
+ if (y < 0 || y >= src_image->bits.height) \
+ { \
+ scanline_func (mask, dst, zero, left_pad + width + right_pad, 0, 0, 0, TRUE); \
+ continue; \
+ } \
+ src = src_first_line + src_stride * y; \
+ if (left_pad > 0) \
+ { \
+ scanline_func (mask, dst, zero, left_pad, 0, 0, 0, TRUE); \
+ } \
+ if (width > 0) \
+ { \
+ scanline_func (mask + (mask_is_solid ? 0 : left_pad), \
+ dst + left_pad, src, width, vx, unit_x, 0, FALSE); \
+ } \
+ if (right_pad > 0) \
+ { \
+ scanline_func (mask + (mask_is_solid ? 0 : left_pad + width), \
+ dst + left_pad + width, zero, right_pad, 0, 0, 0, TRUE); \
+ } \
+ } \
+ else \
+ { \
+ src = src_first_line + src_stride * y; \
+ scanline_func (mask, dst, src, width, vx, unit_x, max_vx, FALSE); \
+ } \
+ } \
+}
+
+/* A workaround for old sun studio, see: https://bugs.freedesktop.org/show_bug.cgi?id=32764 */
+#define FAST_NEAREST_MAINLOOP_COMMON(scale_func_name, scanline_func, src_type_t, mask_type_t, \
+ dst_type_t, repeat_mode, have_mask, mask_is_solid) \
+ FAST_NEAREST_MAINLOOP_INT(_ ## scale_func_name, scanline_func, src_type_t, mask_type_t, \
+ dst_type_t, repeat_mode, have_mask, mask_is_solid)
+
+#define FAST_NEAREST_MAINLOOP_NOMASK(scale_func_name, scanline_func, src_type_t, dst_type_t, \
+ repeat_mode) \
+ static force_inline void \
+ scanline_func##scale_func_name##_wrapper ( \
+ const uint8_t *mask, \
+ dst_type_t *dst, \
+ const src_type_t *src, \
+ int32_t w, \
+ pixman_fixed_t vx, \
+ pixman_fixed_t unit_x, \
+ pixman_fixed_t max_vx, \
+ pixman_bool_t fully_transparent_src) \
+ { \
+ scanline_func (dst, src, w, vx, unit_x, max_vx, fully_transparent_src); \
+ } \
+ FAST_NEAREST_MAINLOOP_INT (scale_func_name, scanline_func##scale_func_name##_wrapper, \
+ src_type_t, uint8_t, dst_type_t, repeat_mode, FALSE, FALSE)
+
+#define FAST_NEAREST_MAINLOOP(scale_func_name, scanline_func, src_type_t, dst_type_t, \
+ repeat_mode) \
+ FAST_NEAREST_MAINLOOP_NOMASK(_ ## scale_func_name, scanline_func, src_type_t, \
+ dst_type_t, repeat_mode)
+
+#define FAST_NEAREST(scale_func_name, SRC_FORMAT, DST_FORMAT, \
+ src_type_t, dst_type_t, OP, repeat_mode) \
+ FAST_NEAREST_SCANLINE(scaled_nearest_scanline_ ## scale_func_name ## _ ## OP, \
+ SRC_FORMAT, DST_FORMAT, src_type_t, dst_type_t, \
+ OP, repeat_mode) \
+ FAST_NEAREST_MAINLOOP_NOMASK(_ ## scale_func_name ## _ ## OP, \
+ scaled_nearest_scanline_ ## scale_func_name ## _ ## OP, \
+ src_type_t, dst_type_t, repeat_mode)
+
+
+#define SCALED_NEAREST_FLAGS \
+ (FAST_PATH_SCALE_TRANSFORM | \
+ FAST_PATH_NO_ALPHA_MAP | \
+ FAST_PATH_NEAREST_FILTER | \
+ FAST_PATH_NO_ACCESSORS | \
+ FAST_PATH_NARROW_FORMAT)
+
+#define SIMPLE_NEAREST_FAST_PATH_NORMAL(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_NORMAL_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_null, 0, \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _normal ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_FAST_PATH_PAD(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_PAD_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_null, 0, \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _pad ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_FAST_PATH_NONE(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_NONE_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_null, 0, \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _none ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_FAST_PATH_COVER(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ SCALED_NEAREST_FLAGS | FAST_PATH_SAMPLES_COVER_CLIP, \
+ PIXMAN_null, 0, \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _cover ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_A8_MASK_FAST_PATH_NORMAL(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_NORMAL_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_a8, MASK_FLAGS (a8, FAST_PATH_UNIFIED_ALPHA), \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _normal ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_A8_MASK_FAST_PATH_PAD(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_PAD_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_a8, MASK_FLAGS (a8, FAST_PATH_UNIFIED_ALPHA), \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _pad ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_A8_MASK_FAST_PATH_NONE(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_NONE_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_a8, MASK_FLAGS (a8, FAST_PATH_UNIFIED_ALPHA), \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _none ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_A8_MASK_FAST_PATH_COVER(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ SCALED_NEAREST_FLAGS | FAST_PATH_SAMPLES_COVER_CLIP, \
+ PIXMAN_a8, MASK_FLAGS (a8, FAST_PATH_UNIFIED_ALPHA), \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _cover ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_NORMAL(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_NORMAL_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_solid, MASK_FLAGS (solid, FAST_PATH_UNIFIED_ALPHA), \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _normal ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_PAD(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_PAD_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_solid, MASK_FLAGS (solid, FAST_PATH_UNIFIED_ALPHA), \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _pad ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_NONE(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ (SCALED_NEAREST_FLAGS | \
+ FAST_PATH_NONE_REPEAT | \
+ FAST_PATH_X_UNIT_POSITIVE), \
+ PIXMAN_solid, MASK_FLAGS (solid, FAST_PATH_UNIFIED_ALPHA), \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _none ## _ ## op, \
+ }
+
+#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_COVER(op,s,d,func) \
+ { PIXMAN_OP_ ## op, \
+ PIXMAN_ ## s, \
+ SCALED_NEAREST_FLAGS | FAST_PATH_SAMPLES_COVER_CLIP, \
+ PIXMAN_solid, MASK_FLAGS (solid, FAST_PATH_UNIFIED_ALPHA), \
+ PIXMAN_ ## d, FAST_PATH_STD_DEST_FLAGS, \
+ fast_composite_scaled_nearest_ ## func ## _cover ## _ ## op, \
+ }
+
+/* Prefer the use of 'cover' variant, because it is faster */
+#define SIMPLE_NEAREST_FAST_PATH(op,s,d,func) \
+ SIMPLE_NEAREST_FAST_PATH_COVER (op,s,d,func), \
+ SIMPLE_NEAREST_FAST_PATH_NONE (op,s,d,func), \
+ SIMPLE_NEAREST_FAST_PATH_PAD (op,s,d,func), \
+ SIMPLE_NEAREST_FAST_PATH_NORMAL (op,s,d,func)
+
+#define SIMPLE_NEAREST_A8_MASK_FAST_PATH(op,s,d,func) \
+ SIMPLE_NEAREST_A8_MASK_FAST_PATH_COVER (op,s,d,func), \
+ SIMPLE_NEAREST_A8_MASK_FAST_PATH_NONE (op,s,d,func), \
+ SIMPLE_NEAREST_A8_MASK_FAST_PATH_PAD (op,s,d,func)
+
+#define SIMPLE_NEAREST_SOLID_MASK_FAST_PATH(op,s,d,func) \
+ SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_COVER (op,s,d,func), \
+ SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_NONE (op,s,d,func), \
+ SIMPLE_NEAREST_SOLID_MASK_FAST_PATH_PAD (op,s,d,func)
+
+#endif
diff --git a/pixman/pixman/pixman-mmx.c b/pixman/pixman/pixman-mmx.c index 027234773..88442f70e 100644 --- a/pixman/pixman/pixman-mmx.c +++ b/pixman/pixman/pixman-mmx.c @@ -152,6 +152,9 @@ to_m64 (uint64_t x) #endif } +#ifdef _MSC_VER +#define to_uint64(arg) arg.M64_MEMBER +#else static force_inline uint64_t to_uint64 (__m64 x) { @@ -164,6 +167,7 @@ to_uint64 (__m64 x) return (uint64_t)x; #endif } +#endif static force_inline __m64 shift (__m64 v, @@ -310,11 +314,15 @@ pack8888 (__m64 lo, __m64 hi) return _mm_packs_pu16 (lo, hi); } +#ifdef _MSC_VER +#define store8888(v) _mm_cvtsi64_si32 (pack8888 (v, _mm_setzero_si64 ())) +#else static force_inline uint32_t store8888 (__m64 v) { return _mm_cvtsi64_si32 (pack8888 (v, _mm_setzero_si64 ())); } +#endif /* Expand 16 bits positioned at @pos (0-3) of a mmx register into * @@ -417,6 +425,13 @@ pix_add_mul (__m64 x, __m64 a, __m64 y, __m64 b) /* --------------- MMX code patch for fbcompose.c --------------------- */ +#ifdef _MSC_VER +#define combine(src, mask) \ + ((mask) ? \ + store8888 (pix_multiply (load8888 (*src), expand_alpha (load8888 (*mask)))) \ + : \ + *src) +#else static force_inline uint32_t combine (const uint32_t *src, const uint32_t *mask) { @@ -435,6 +450,7 @@ combine (const uint32_t *src, const uint32_t *mask) return ssrc; } +#endif static void mmx_combine_over_u (pixman_implementation_t *imp, @@ -1685,7 +1701,7 @@ mmx_composite_over_8888_0565 (pixman_implementation_t *imp, vdest = pack_565 ( over (vsrc, expand_alpha (vsrc), vdest), vdest, 0); - *dst = to_uint64 (vdest); + *dst = to_uint64 (vdest)&0xffff; w--; dst++; @@ -1728,7 +1744,7 @@ mmx_composite_over_8888_0565 (pixman_implementation_t *imp, vdest = pack_565 (over (vsrc, expand_alpha (vsrc), vdest), vdest, 0); - *dst = to_uint64 (vdest); + *dst = to_uint64 (vdest) & 0xffff; w--; dst++; @@ -1943,7 +1959,7 @@ pixman_fill_mmx (uint32_t *bits, while (w >= 2 && ((unsigned long)d & 3)) { - *(uint16_t *)d = xor; + *(uint16_t *)d = (xor & 0xffff); w -= 2; d += 2; } @@ -1996,7 +2012,7 @@ pixman_fill_mmx (uint32_t *bits, } while (w >= 2) { - *(uint16_t *)d = xor; + *(uint16_t *)d = (xor & 0xffff); w -= 2; d += 2; } diff --git a/pixman/pixman/pixman-sse2.c b/pixman/pixman/pixman-sse2.c index 2e135e2fe..b62a54c3f 100644 --- a/pixman/pixman/pixman-sse2.c +++ b/pixman/pixman/pixman-sse2.c @@ -1,6424 +1,6424 @@ -/* - * Copyright © 2008 Rodrigo Kumpera - * Copyright © 2008 André Tupinambá - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of Red Hat not be used in advertising or - * publicity pertaining to distribution of the software without specific, - * written prior permission. Red Hat makes no representations about the - * suitability of this software for any purpose. It is provided "as is" - * without express or implied warranty. - * - * THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS - * SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND - * FITNESS, IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY - * SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES - * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN - * AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING - * OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS - * SOFTWARE. - * - * Author: Rodrigo Kumpera (kumpera@gmail.com) - * André Tupinambá (andrelrt@gmail.com) - * - * Based on work by Owen Taylor and Søren Sandmann - */ -#ifdef HAVE_CONFIG_H -#include <config.h> -#endif - -#include <mmintrin.h> -#include <xmmintrin.h> /* for _mm_shuffle_pi16 and _MM_SHUFFLE */ -#include <emmintrin.h> /* for SSE2 intrinsics */ -#include "pixman-private.h" -#include "pixman-combine32.h" -#include "pixman-fast-path.h" - -#if defined(_MSC_VER) && defined(_M_AMD64) -/* Windows 64 doesn't allow MMX to be used, so - * the pixman-x64-mmx-emulation.h file contains - * implementations of those MMX intrinsics that - * are used in the SSE2 implementation. - */ -# include "pixman-x64-mmx-emulation.h" -#endif - -#ifdef USE_SSE2 - -/* -------------------------------------------------------------------- - * Locals - */ - -static __m64 mask_x0080; -static __m64 mask_x00ff; -static __m64 mask_x0101; -static __m64 mask_x_alpha; - -static __m64 mask_x565_rgb; -static __m64 mask_x565_unpack; - -static __m128i mask_0080; -static __m128i mask_00ff; -static __m128i mask_0101; -static __m128i mask_ffff; -static __m128i mask_ff000000; -static __m128i mask_alpha; - -static __m128i mask_565_r; -static __m128i mask_565_g1, mask_565_g2; -static __m128i mask_565_b; -static __m128i mask_red; -static __m128i mask_green; -static __m128i mask_blue; - -static __m128i mask_565_fix_rb; -static __m128i mask_565_fix_g; - -/* ---------------------------------------------------------------------- - * SSE2 Inlines - */ -static force_inline __m128i -unpack_32_1x128 (uint32_t data) -{ - return _mm_unpacklo_epi8 (_mm_cvtsi32_si128 (data), _mm_setzero_si128 ()); -} - -static force_inline void -unpack_128_2x128 (__m128i data, __m128i* data_lo, __m128i* data_hi) -{ - *data_lo = _mm_unpacklo_epi8 (data, _mm_setzero_si128 ()); - *data_hi = _mm_unpackhi_epi8 (data, _mm_setzero_si128 ()); -} - -static force_inline __m128i -unpack_565_to_8888 (__m128i lo) -{ - __m128i r, g, b, rb, t; - - r = _mm_and_si128 (_mm_slli_epi32 (lo, 8), mask_red); - g = _mm_and_si128 (_mm_slli_epi32 (lo, 5), mask_green); - b = _mm_and_si128 (_mm_slli_epi32 (lo, 3), mask_blue); - - rb = _mm_or_si128 (r, b); - t = _mm_and_si128 (rb, mask_565_fix_rb); - t = _mm_srli_epi32 (t, 5); - rb = _mm_or_si128 (rb, t); - - t = _mm_and_si128 (g, mask_565_fix_g); - t = _mm_srli_epi32 (t, 6); - g = _mm_or_si128 (g, t); - - return _mm_or_si128 (rb, g); -} - -static force_inline void -unpack_565_128_4x128 (__m128i data, - __m128i* data0, - __m128i* data1, - __m128i* data2, - __m128i* data3) -{ - __m128i lo, hi; - - lo = _mm_unpacklo_epi16 (data, _mm_setzero_si128 ()); - hi = _mm_unpackhi_epi16 (data, _mm_setzero_si128 ()); - - lo = unpack_565_to_8888 (lo); - hi = unpack_565_to_8888 (hi); - - unpack_128_2x128 (lo, data0, data1); - unpack_128_2x128 (hi, data2, data3); -} - -static force_inline uint16_t -pack_565_32_16 (uint32_t pixel) -{ - return (uint16_t) (((pixel >> 8) & 0xf800) | - ((pixel >> 5) & 0x07e0) | - ((pixel >> 3) & 0x001f)); -} - -static force_inline __m128i -pack_2x128_128 (__m128i lo, __m128i hi) -{ - return _mm_packus_epi16 (lo, hi); -} - -static force_inline __m128i -pack_565_2x128_128 (__m128i lo, __m128i hi) -{ - __m128i data; - __m128i r, g1, g2, b; - - data = pack_2x128_128 (lo, hi); - - r = _mm_and_si128 (data, mask_565_r); - g1 = _mm_and_si128 (_mm_slli_epi32 (data, 3), mask_565_g1); - g2 = _mm_and_si128 (_mm_srli_epi32 (data, 5), mask_565_g2); - b = _mm_and_si128 (_mm_srli_epi32 (data, 3), mask_565_b); - - return _mm_or_si128 (_mm_or_si128 (_mm_or_si128 (r, g1), g2), b); -} - -static force_inline __m128i -pack_565_4x128_128 (__m128i* xmm0, __m128i* xmm1, __m128i* xmm2, __m128i* xmm3) -{ - return _mm_packus_epi16 (pack_565_2x128_128 (*xmm0, *xmm1), - pack_565_2x128_128 (*xmm2, *xmm3)); -} - -static force_inline int -is_opaque (__m128i x) -{ - __m128i ffs = _mm_cmpeq_epi8 (x, x); - - return (_mm_movemask_epi8 (_mm_cmpeq_epi8 (x, ffs)) & 0x8888) == 0x8888; -} - -static force_inline int -is_zero (__m128i x) -{ - return _mm_movemask_epi8 ( - _mm_cmpeq_epi8 (x, _mm_setzero_si128 ())) == 0xffff; -} - -static force_inline int -is_transparent (__m128i x) -{ - return (_mm_movemask_epi8 ( - _mm_cmpeq_epi8 (x, _mm_setzero_si128 ())) & 0x8888) == 0x8888; -} - -static force_inline __m128i -expand_pixel_32_1x128 (uint32_t data) -{ - return _mm_shuffle_epi32 (unpack_32_1x128 (data), _MM_SHUFFLE (1, 0, 1, 0)); -} - -static force_inline __m128i -expand_alpha_1x128 (__m128i data) -{ - return _mm_shufflehi_epi16 (_mm_shufflelo_epi16 (data, - _MM_SHUFFLE (3, 3, 3, 3)), - _MM_SHUFFLE (3, 3, 3, 3)); -} - -static force_inline void -expand_alpha_2x128 (__m128i data_lo, - __m128i data_hi, - __m128i* alpha_lo, - __m128i* alpha_hi) -{ - __m128i lo, hi; - - lo = _mm_shufflelo_epi16 (data_lo, _MM_SHUFFLE (3, 3, 3, 3)); - hi = _mm_shufflelo_epi16 (data_hi, _MM_SHUFFLE (3, 3, 3, 3)); - - *alpha_lo = _mm_shufflehi_epi16 (lo, _MM_SHUFFLE (3, 3, 3, 3)); - *alpha_hi = _mm_shufflehi_epi16 (hi, _MM_SHUFFLE (3, 3, 3, 3)); -} - -static force_inline void -expand_alpha_rev_2x128 (__m128i data_lo, - __m128i data_hi, - __m128i* alpha_lo, - __m128i* alpha_hi) -{ - __m128i lo, hi; - - lo = _mm_shufflelo_epi16 (data_lo, _MM_SHUFFLE (0, 0, 0, 0)); - hi = _mm_shufflelo_epi16 (data_hi, _MM_SHUFFLE (0, 0, 0, 0)); - *alpha_lo = _mm_shufflehi_epi16 (lo, _MM_SHUFFLE (0, 0, 0, 0)); - *alpha_hi = _mm_shufflehi_epi16 (hi, _MM_SHUFFLE (0, 0, 0, 0)); -} - -static force_inline void -pix_multiply_2x128 (__m128i* data_lo, - __m128i* data_hi, - __m128i* alpha_lo, - __m128i* alpha_hi, - __m128i* ret_lo, - __m128i* ret_hi) -{ - __m128i lo, hi; - - lo = _mm_mullo_epi16 (*data_lo, *alpha_lo); - hi = _mm_mullo_epi16 (*data_hi, *alpha_hi); - lo = _mm_adds_epu16 (lo, mask_0080); - hi = _mm_adds_epu16 (hi, mask_0080); - *ret_lo = _mm_mulhi_epu16 (lo, mask_0101); - *ret_hi = _mm_mulhi_epu16 (hi, mask_0101); -} - -static force_inline void -pix_add_multiply_2x128 (__m128i* src_lo, - __m128i* src_hi, - __m128i* alpha_dst_lo, - __m128i* alpha_dst_hi, - __m128i* dst_lo, - __m128i* dst_hi, - __m128i* alpha_src_lo, - __m128i* alpha_src_hi, - __m128i* ret_lo, - __m128i* ret_hi) -{ - __m128i t1_lo, t1_hi; - __m128i t2_lo, t2_hi; - - pix_multiply_2x128 (src_lo, src_hi, alpha_dst_lo, alpha_dst_hi, &t1_lo, &t1_hi); - pix_multiply_2x128 (dst_lo, dst_hi, alpha_src_lo, alpha_src_hi, &t2_lo, &t2_hi); - - *ret_lo = _mm_adds_epu8 (t1_lo, t2_lo); - *ret_hi = _mm_adds_epu8 (t1_hi, t2_hi); -} - -static force_inline void -negate_2x128 (__m128i data_lo, - __m128i data_hi, - __m128i* neg_lo, - __m128i* neg_hi) -{ - *neg_lo = _mm_xor_si128 (data_lo, mask_00ff); - *neg_hi = _mm_xor_si128 (data_hi, mask_00ff); -} - -static force_inline void -invert_colors_2x128 (__m128i data_lo, - __m128i data_hi, - __m128i* inv_lo, - __m128i* inv_hi) -{ - __m128i lo, hi; - - lo = _mm_shufflelo_epi16 (data_lo, _MM_SHUFFLE (3, 0, 1, 2)); - hi = _mm_shufflelo_epi16 (data_hi, _MM_SHUFFLE (3, 0, 1, 2)); - *inv_lo = _mm_shufflehi_epi16 (lo, _MM_SHUFFLE (3, 0, 1, 2)); - *inv_hi = _mm_shufflehi_epi16 (hi, _MM_SHUFFLE (3, 0, 1, 2)); -} - -static force_inline void -over_2x128 (__m128i* src_lo, - __m128i* src_hi, - __m128i* alpha_lo, - __m128i* alpha_hi, - __m128i* dst_lo, - __m128i* dst_hi) -{ - __m128i t1, t2; - - negate_2x128 (*alpha_lo, *alpha_hi, &t1, &t2); - - pix_multiply_2x128 (dst_lo, dst_hi, &t1, &t2, dst_lo, dst_hi); - - *dst_lo = _mm_adds_epu8 (*src_lo, *dst_lo); - *dst_hi = _mm_adds_epu8 (*src_hi, *dst_hi); -} - -static force_inline void -over_rev_non_pre_2x128 (__m128i src_lo, - __m128i src_hi, - __m128i* dst_lo, - __m128i* dst_hi) -{ - __m128i lo, hi; - __m128i alpha_lo, alpha_hi; - - expand_alpha_2x128 (src_lo, src_hi, &alpha_lo, &alpha_hi); - - lo = _mm_or_si128 (alpha_lo, mask_alpha); - hi = _mm_or_si128 (alpha_hi, mask_alpha); - - invert_colors_2x128 (src_lo, src_hi, &src_lo, &src_hi); - - pix_multiply_2x128 (&src_lo, &src_hi, &lo, &hi, &lo, &hi); - - over_2x128 (&lo, &hi, &alpha_lo, &alpha_hi, dst_lo, dst_hi); -} - -static force_inline void -in_over_2x128 (__m128i* src_lo, - __m128i* src_hi, - __m128i* alpha_lo, - __m128i* alpha_hi, - __m128i* mask_lo, - __m128i* mask_hi, - __m128i* dst_lo, - __m128i* dst_hi) -{ - __m128i s_lo, s_hi; - __m128i a_lo, a_hi; - - pix_multiply_2x128 (src_lo, src_hi, mask_lo, mask_hi, &s_lo, &s_hi); - pix_multiply_2x128 (alpha_lo, alpha_hi, mask_lo, mask_hi, &a_lo, &a_hi); - - over_2x128 (&s_lo, &s_hi, &a_lo, &a_hi, dst_lo, dst_hi); -} - -/* load 4 pixels from a 16-byte boundary aligned address */ -static force_inline __m128i -load_128_aligned (__m128i* src) -{ - return _mm_load_si128 (src); -} - -/* load 4 pixels from a unaligned address */ -static force_inline __m128i -load_128_unaligned (const __m128i* src) -{ - return _mm_loadu_si128 (src); -} - -/* save 4 pixels using Write Combining memory on a 16-byte - * boundary aligned address - */ -static force_inline void -save_128_write_combining (__m128i* dst, - __m128i data) -{ - _mm_stream_si128 (dst, data); -} - -/* save 4 pixels on a 16-byte boundary aligned address */ -static force_inline void -save_128_aligned (__m128i* dst, - __m128i data) -{ - _mm_store_si128 (dst, data); -} - -/* save 4 pixels on a unaligned address */ -static force_inline void -save_128_unaligned (__m128i* dst, - __m128i data) -{ - _mm_storeu_si128 (dst, data); -} - -/* ------------------------------------------------------------------ - * MMX inlines - */ - -static force_inline __m64 -load_32_1x64 (uint32_t data) -{ - return _mm_cvtsi32_si64 (data); -} - -static force_inline __m64 -unpack_32_1x64 (uint32_t data) -{ - return _mm_unpacklo_pi8 (load_32_1x64 (data), _mm_setzero_si64 ()); -} - -static force_inline __m64 -expand_alpha_1x64 (__m64 data) -{ - return _mm_shuffle_pi16 (data, _MM_SHUFFLE (3, 3, 3, 3)); -} - -static force_inline __m64 -expand_alpha_rev_1x64 (__m64 data) -{ - return _mm_shuffle_pi16 (data, _MM_SHUFFLE (0, 0, 0, 0)); -} - -static force_inline __m64 -expand_pixel_8_1x64 (uint8_t data) -{ - return _mm_shuffle_pi16 ( - unpack_32_1x64 ((uint32_t)data), _MM_SHUFFLE (0, 0, 0, 0)); -} - -static force_inline __m64 -pix_multiply_1x64 (__m64 data, - __m64 alpha) -{ - return _mm_mulhi_pu16 (_mm_adds_pu16 (_mm_mullo_pi16 (data, alpha), - mask_x0080), - mask_x0101); -} - -static force_inline __m64 -pix_add_multiply_1x64 (__m64* src, - __m64* alpha_dst, - __m64* dst, - __m64* alpha_src) -{ - __m64 t1 = pix_multiply_1x64 (*src, *alpha_dst); - __m64 t2 = pix_multiply_1x64 (*dst, *alpha_src); - - return _mm_adds_pu8 (t1, t2); -} - -static force_inline __m64 -negate_1x64 (__m64 data) -{ - return _mm_xor_si64 (data, mask_x00ff); -} - -static force_inline __m64 -invert_colors_1x64 (__m64 data) -{ - return _mm_shuffle_pi16 (data, _MM_SHUFFLE (3, 0, 1, 2)); -} - -static force_inline __m64 -over_1x64 (__m64 src, __m64 alpha, __m64 dst) -{ - return _mm_adds_pu8 (src, pix_multiply_1x64 (dst, negate_1x64 (alpha))); -} - -static force_inline __m64 -in_over_1x64 (__m64* src, __m64* alpha, __m64* mask, __m64* dst) -{ - return over_1x64 (pix_multiply_1x64 (*src, *mask), - pix_multiply_1x64 (*alpha, *mask), - *dst); -} - -static force_inline __m64 -over_rev_non_pre_1x64 (__m64 src, __m64 dst) -{ - __m64 alpha = expand_alpha_1x64 (src); - - return over_1x64 (pix_multiply_1x64 (invert_colors_1x64 (src), - _mm_or_si64 (alpha, mask_x_alpha)), - alpha, - dst); -} - -static force_inline uint32_t -pack_1x64_32 (__m64 data) -{ - return _mm_cvtsi64_si32 (_mm_packs_pu16 (data, _mm_setzero_si64 ())); -} - -/* Expand 16 bits positioned at @pos (0-3) of a mmx register into - * - * 00RR00GG00BB - * - * --- Expanding 565 in the low word --- - * - * m = (m << (32 - 3)) | (m << (16 - 5)) | m; - * m = m & (01f0003f001f); - * m = m * (008404100840); - * m = m >> 8; - * - * Note the trick here - the top word is shifted by another nibble to - * avoid it bumping into the middle word - */ -static force_inline __m64 -expand565_16_1x64 (uint16_t pixel) -{ - __m64 p; - __m64 t1, t2; - - p = _mm_cvtsi32_si64 ((uint32_t) pixel); - - t1 = _mm_slli_si64 (p, 36 - 11); - t2 = _mm_slli_si64 (p, 16 - 5); - - p = _mm_or_si64 (t1, p); - p = _mm_or_si64 (t2, p); - p = _mm_and_si64 (p, mask_x565_rgb); - p = _mm_mullo_pi16 (p, mask_x565_unpack); - - return _mm_srli_pi16 (p, 8); -} - -/* ---------------------------------------------------------------------------- - * Compose Core transformations - */ -static force_inline uint32_t -core_combine_over_u_pixel_sse2 (uint32_t src, uint32_t dst) -{ - uint8_t a; - __m64 ms; - - a = src >> 24; - - if (a == 0xff) - { - return src; - } - else if (src) - { - ms = unpack_32_1x64 (src); - return pack_1x64_32 ( - over_1x64 (ms, expand_alpha_1x64 (ms), unpack_32_1x64 (dst))); - } - - return dst; -} - -static force_inline uint32_t -combine1 (const uint32_t *ps, const uint32_t *pm) -{ - uint32_t s = *ps; - - if (pm) - { - __m64 ms, mm; - - mm = unpack_32_1x64 (*pm); - mm = expand_alpha_1x64 (mm); - - ms = unpack_32_1x64 (s); - ms = pix_multiply_1x64 (ms, mm); - - s = pack_1x64_32 (ms); - } - - return s; -} - -static force_inline __m128i -combine4 (const __m128i *ps, const __m128i *pm) -{ - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_msk_lo, xmm_msk_hi; - __m128i s; - - if (pm) - { - xmm_msk_lo = load_128_unaligned (pm); - - if (is_transparent (xmm_msk_lo)) - return _mm_setzero_si128 (); - } - - s = load_128_unaligned (ps); - - if (pm) - { - unpack_128_2x128 (s, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_msk_lo, &xmm_msk_lo, &xmm_msk_hi); - - expand_alpha_2x128 (xmm_msk_lo, xmm_msk_hi, &xmm_msk_lo, &xmm_msk_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_msk_lo, &xmm_msk_hi, - &xmm_src_lo, &xmm_src_hi); - - s = pack_2x128_128 (xmm_src_lo, xmm_src_hi); - } - - return s; -} - -static force_inline void -core_combine_over_u_sse2_mask (uint32_t * pd, - const uint32_t* ps, - const uint32_t* pm, - int w) -{ - uint32_t s, d; - - /* Align dst on a 16-byte boundary */ - while (w && ((unsigned long)pd & 15)) - { - d = *pd; - s = combine1 (ps, pm); - - if (s) - *pd = core_combine_over_u_pixel_sse2 (s, d); - pd++; - ps++; - pm++; - w--; - } - - while (w >= 4) - { - __m128i mask = load_128_unaligned ((__m128i *)pm); - - if (!is_zero (mask)) - { - __m128i src; - __m128i src_hi, src_lo; - __m128i mask_hi, mask_lo; - __m128i alpha_hi, alpha_lo; - - src = load_128_unaligned ((__m128i *)ps); - - if (is_opaque (_mm_and_si128 (src, mask))) - { - save_128_aligned ((__m128i *)pd, src); - } - else - { - __m128i dst = load_128_aligned ((__m128i *)pd); - __m128i dst_hi, dst_lo; - - unpack_128_2x128 (mask, &mask_lo, &mask_hi); - unpack_128_2x128 (src, &src_lo, &src_hi); - - expand_alpha_2x128 (mask_lo, mask_hi, &mask_lo, &mask_hi); - pix_multiply_2x128 (&src_lo, &src_hi, - &mask_lo, &mask_hi, - &src_lo, &src_hi); - - unpack_128_2x128 (dst, &dst_lo, &dst_hi); - - expand_alpha_2x128 (src_lo, src_hi, - &alpha_lo, &alpha_hi); - - over_2x128 (&src_lo, &src_hi, &alpha_lo, &alpha_hi, - &dst_lo, &dst_hi); - - save_128_aligned ( - (__m128i *)pd, - pack_2x128_128 (dst_lo, dst_hi)); - } - } - - pm += 4; - ps += 4; - pd += 4; - w -= 4; - } - while (w) - { - d = *pd; - s = combine1 (ps, pm); - - if (s) - *pd = core_combine_over_u_pixel_sse2 (s, d); - pd++; - ps++; - pm++; - - w--; - } -} - -static force_inline void -core_combine_over_u_sse2_no_mask (uint32_t * pd, - const uint32_t* ps, - int w) -{ - uint32_t s, d; - - /* Align dst on a 16-byte boundary */ - while (w && ((unsigned long)pd & 15)) - { - d = *pd; - s = *ps; - - if (s) - *pd = core_combine_over_u_pixel_sse2 (s, d); - pd++; - ps++; - w--; - } - - while (w >= 4) - { - __m128i src; - __m128i src_hi, src_lo, dst_hi, dst_lo; - __m128i alpha_hi, alpha_lo; - - src = load_128_unaligned ((__m128i *)ps); - - if (!is_zero (src)) - { - if (is_opaque (src)) - { - save_128_aligned ((__m128i *)pd, src); - } - else - { - __m128i dst = load_128_aligned ((__m128i *)pd); - - unpack_128_2x128 (src, &src_lo, &src_hi); - unpack_128_2x128 (dst, &dst_lo, &dst_hi); - - expand_alpha_2x128 (src_lo, src_hi, - &alpha_lo, &alpha_hi); - over_2x128 (&src_lo, &src_hi, &alpha_lo, &alpha_hi, - &dst_lo, &dst_hi); - - save_128_aligned ( - (__m128i *)pd, - pack_2x128_128 (dst_lo, dst_hi)); - } - } - - ps += 4; - pd += 4; - w -= 4; - } - while (w) - { - d = *pd; - s = *ps; - - if (s) - *pd = core_combine_over_u_pixel_sse2 (s, d); - pd++; - ps++; - - w--; - } -} - -static force_inline void -core_combine_over_u_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t* pm, - int w) -{ - if (pm) - core_combine_over_u_sse2_mask (pd, ps, pm, w); - else - core_combine_over_u_sse2_no_mask (pd, ps, w); -} - -static force_inline void -core_combine_over_reverse_u_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t* pm, - int w) -{ - uint32_t s, d; - - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_alpha_lo, xmm_alpha_hi; - - /* Align dst on a 16-byte boundary */ - while (w && - ((unsigned long)pd & 15)) - { - d = *pd; - s = combine1 (ps, pm); - - *pd++ = core_combine_over_u_pixel_sse2 (d, s); - w--; - ps++; - if (pm) - pm++; - } - - while (w >= 4) - { - /* I'm loading unaligned because I'm not sure - * about the address alignment. - */ - xmm_src_hi = combine4 ((__m128i*)ps, (__m128i*)pm); - xmm_dst_hi = load_128_aligned ((__m128i*) pd); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - over_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_src_lo, &xmm_src_hi); - - /* rebuid the 4 pixel data and save*/ - save_128_aligned ((__m128i*)pd, - pack_2x128_128 (xmm_src_lo, xmm_src_hi)); - - w -= 4; - ps += 4; - pd += 4; - - if (pm) - pm += 4; - } - - while (w) - { - d = *pd; - s = combine1 (ps, pm); - - *pd++ = core_combine_over_u_pixel_sse2 (d, s); - ps++; - w--; - if (pm) - pm++; - } -} - -static force_inline uint32_t -core_combine_in_u_pixelsse2 (uint32_t src, uint32_t dst) -{ - uint32_t maska = src >> 24; - - if (maska == 0) - { - return 0; - } - else if (maska != 0xff) - { - return pack_1x64_32 ( - pix_multiply_1x64 (unpack_32_1x64 (dst), - expand_alpha_1x64 (unpack_32_1x64 (src)))); - } - - return dst; -} - -static force_inline void -core_combine_in_u_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t* pm, - int w) -{ - uint32_t s, d; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - - while (w && ((unsigned long) pd & 15)) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_in_u_pixelsse2 (d, s); - w--; - ps++; - if (pm) - pm++; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*) pd); - xmm_src_hi = combine4 ((__m128i*) ps, (__m128i*) pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_dst_lo, &xmm_dst_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ((__m128i*)pd, - pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - w -= 4; - if (pm) - pm += 4; - } - - while (w) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_in_u_pixelsse2 (d, s); - w--; - ps++; - if (pm) - pm++; - } -} - -static force_inline void -core_combine_reverse_in_u_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t *pm, - int w) -{ - uint32_t s, d; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - - while (w && ((unsigned long) pd & 15)) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_in_u_pixelsse2 (s, d); - ps++; - w--; - if (pm) - pm++; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*) pd); - xmm_src_hi = combine4 ((__m128i*) ps, (__m128i*)pm); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_src_lo, &xmm_src_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - w -= 4; - if (pm) - pm += 4; - } - - while (w) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_in_u_pixelsse2 (s, d); - w--; - ps++; - if (pm) - pm++; - } -} - -static force_inline void -core_combine_reverse_out_u_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t* pm, - int w) -{ - while (w && ((unsigned long) pd & 15)) - { - uint32_t s = combine1 (ps, pm); - uint32_t d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (d), negate_1x64 ( - expand_alpha_1x64 (unpack_32_1x64 (s))))); - - if (pm) - pm++; - ps++; - w--; - } - - while (w >= 4) - { - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - - xmm_src_hi = combine4 ((__m128i*)ps, (__m128i*)pm); - xmm_dst_hi = load_128_aligned ((__m128i*) pd); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - negate_2x128 (xmm_src_lo, xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - - pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_src_lo, &xmm_src_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - if (pm) - pm += 4; - - w -= 4; - } - - while (w) - { - uint32_t s = combine1 (ps, pm); - uint32_t d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (d), negate_1x64 ( - expand_alpha_1x64 (unpack_32_1x64 (s))))); - ps++; - if (pm) - pm++; - w--; - } -} - -static force_inline void -core_combine_out_u_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t* pm, - int w) -{ - while (w && ((unsigned long) pd & 15)) - { - uint32_t s = combine1 (ps, pm); - uint32_t d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (s), negate_1x64 ( - expand_alpha_1x64 (unpack_32_1x64 (d))))); - w--; - ps++; - if (pm) - pm++; - } - - while (w >= 4) - { - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - - xmm_src_hi = combine4 ((__m128i*) ps, (__m128i*)pm); - xmm_dst_hi = load_128_aligned ((__m128i*) pd); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - negate_2x128 (xmm_dst_lo, xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_dst_lo, &xmm_dst_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - w -= 4; - if (pm) - pm += 4; - } - - while (w) - { - uint32_t s = combine1 (ps, pm); - uint32_t d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (s), negate_1x64 ( - expand_alpha_1x64 (unpack_32_1x64 (d))))); - w--; - ps++; - if (pm) - pm++; - } -} - -static force_inline uint32_t -core_combine_atop_u_pixel_sse2 (uint32_t src, - uint32_t dst) -{ - __m64 s = unpack_32_1x64 (src); - __m64 d = unpack_32_1x64 (dst); - - __m64 sa = negate_1x64 (expand_alpha_1x64 (s)); - __m64 da = expand_alpha_1x64 (d); - - return pack_1x64_32 (pix_add_multiply_1x64 (&s, &da, &d, &sa)); -} - -static force_inline void -core_combine_atop_u_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t* pm, - int w) -{ - uint32_t s, d; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_alpha_src_lo, xmm_alpha_src_hi; - __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi; - - while (w && ((unsigned long) pd & 15)) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_atop_u_pixel_sse2 (s, d); - w--; - ps++; - if (pm) - pm++; - } - - while (w >= 4) - { - xmm_src_hi = combine4 ((__m128i*)ps, (__m128i*)pm); - xmm_dst_hi = load_128_aligned ((__m128i*) pd); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi); - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - negate_2x128 (xmm_alpha_src_lo, xmm_alpha_src_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi); - - pix_add_multiply_2x128 ( - &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi, - &xmm_dst_lo, &xmm_dst_hi, &xmm_alpha_src_lo, &xmm_alpha_src_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - w -= 4; - if (pm) - pm += 4; - } - - while (w) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_atop_u_pixel_sse2 (s, d); - w--; - ps++; - if (pm) - pm++; - } -} - -static force_inline uint32_t -core_combine_reverse_atop_u_pixel_sse2 (uint32_t src, - uint32_t dst) -{ - __m64 s = unpack_32_1x64 (src); - __m64 d = unpack_32_1x64 (dst); - - __m64 sa = expand_alpha_1x64 (s); - __m64 da = negate_1x64 (expand_alpha_1x64 (d)); - - return pack_1x64_32 (pix_add_multiply_1x64 (&s, &da, &d, &sa)); -} - -static force_inline void -core_combine_reverse_atop_u_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t* pm, - int w) -{ - uint32_t s, d; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_alpha_src_lo, xmm_alpha_src_hi; - __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi; - - while (w && ((unsigned long) pd & 15)) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_reverse_atop_u_pixel_sse2 (s, d); - ps++; - w--; - if (pm) - pm++; - } - - while (w >= 4) - { - xmm_src_hi = combine4 ((__m128i*)ps, (__m128i*)pm); - xmm_dst_hi = load_128_aligned ((__m128i*) pd); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi); - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - negate_2x128 (xmm_alpha_dst_lo, xmm_alpha_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - pix_add_multiply_2x128 ( - &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi, - &xmm_dst_lo, &xmm_dst_hi, &xmm_alpha_src_lo, &xmm_alpha_src_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - w -= 4; - if (pm) - pm += 4; - } - - while (w) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_reverse_atop_u_pixel_sse2 (s, d); - ps++; - w--; - if (pm) - pm++; - } -} - -static force_inline uint32_t -core_combine_xor_u_pixel_sse2 (uint32_t src, - uint32_t dst) -{ - __m64 s = unpack_32_1x64 (src); - __m64 d = unpack_32_1x64 (dst); - - __m64 neg_d = negate_1x64 (expand_alpha_1x64 (d)); - __m64 neg_s = negate_1x64 (expand_alpha_1x64 (s)); - - return pack_1x64_32 (pix_add_multiply_1x64 (&s, &neg_d, &d, &neg_s)); -} - -static force_inline void -core_combine_xor_u_sse2 (uint32_t* dst, - const uint32_t* src, - const uint32_t *mask, - int width) -{ - int w = width; - uint32_t s, d; - uint32_t* pd = dst; - const uint32_t* ps = src; - const uint32_t* pm = mask; - - __m128i xmm_src, xmm_src_lo, xmm_src_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_alpha_src_lo, xmm_alpha_src_hi; - __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi; - - while (w && ((unsigned long) pd & 15)) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_xor_u_pixel_sse2 (s, d); - w--; - ps++; - if (pm) - pm++; - } - - while (w >= 4) - { - xmm_src = combine4 ((__m128i*) ps, (__m128i*) pm); - xmm_dst = load_128_aligned ((__m128i*) pd); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi); - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - negate_2x128 (xmm_alpha_src_lo, xmm_alpha_src_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi); - negate_2x128 (xmm_alpha_dst_lo, xmm_alpha_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - pix_add_multiply_2x128 ( - &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi, - &xmm_dst_lo, &xmm_dst_hi, &xmm_alpha_src_lo, &xmm_alpha_src_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - w -= 4; - if (pm) - pm += 4; - } - - while (w) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_xor_u_pixel_sse2 (s, d); - w--; - ps++; - if (pm) - pm++; - } -} - -static force_inline void -core_combine_add_u_sse2 (uint32_t* dst, - const uint32_t* src, - const uint32_t* mask, - int width) -{ - int w = width; - uint32_t s, d; - uint32_t* pd = dst; - const uint32_t* ps = src; - const uint32_t* pm = mask; - - while (w && (unsigned long)pd & 15) - { - s = combine1 (ps, pm); - d = *pd; - - ps++; - if (pm) - pm++; - *pd++ = _mm_cvtsi64_si32 ( - _mm_adds_pu8 (_mm_cvtsi32_si64 (s), _mm_cvtsi32_si64 (d))); - w--; - } - - while (w >= 4) - { - __m128i s; - - s = combine4 ((__m128i*)ps, (__m128i*)pm); - - save_128_aligned ( - (__m128i*)pd, _mm_adds_epu8 (s, load_128_aligned ((__m128i*)pd))); - - pd += 4; - ps += 4; - if (pm) - pm += 4; - w -= 4; - } - - while (w--) - { - s = combine1 (ps, pm); - d = *pd; - - ps++; - *pd++ = _mm_cvtsi64_si32 ( - _mm_adds_pu8 (_mm_cvtsi32_si64 (s), _mm_cvtsi32_si64 (d))); - if (pm) - pm++; - } -} - -static force_inline uint32_t -core_combine_saturate_u_pixel_sse2 (uint32_t src, - uint32_t dst) -{ - __m64 ms = unpack_32_1x64 (src); - __m64 md = unpack_32_1x64 (dst); - uint32_t sa = src >> 24; - uint32_t da = ~dst >> 24; - - if (sa > da) - { - ms = pix_multiply_1x64 ( - ms, expand_alpha_1x64 (unpack_32_1x64 (DIV_UN8 (da, sa) << 24))); - } - - return pack_1x64_32 (_mm_adds_pu16 (md, ms)); -} - -static force_inline void -core_combine_saturate_u_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, d; - - uint32_t pack_cmp; - __m128i xmm_src, xmm_dst; - - while (w && (unsigned long)pd & 15) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_saturate_u_pixel_sse2 (s, d); - w--; - ps++; - if (pm) - pm++; - } - - while (w >= 4) - { - xmm_dst = load_128_aligned ((__m128i*)pd); - xmm_src = combine4 ((__m128i*)ps, (__m128i*)pm); - - pack_cmp = _mm_movemask_epi8 ( - _mm_cmpgt_epi32 ( - _mm_srli_epi32 (xmm_src, 24), - _mm_srli_epi32 (_mm_xor_si128 (xmm_dst, mask_ff000000), 24))); - - /* if some alpha src is grater than respective ~alpha dst */ - if (pack_cmp) - { - s = combine1 (ps++, pm); - d = *pd; - *pd++ = core_combine_saturate_u_pixel_sse2 (s, d); - if (pm) - pm++; - - s = combine1 (ps++, pm); - d = *pd; - *pd++ = core_combine_saturate_u_pixel_sse2 (s, d); - if (pm) - pm++; - - s = combine1 (ps++, pm); - d = *pd; - *pd++ = core_combine_saturate_u_pixel_sse2 (s, d); - if (pm) - pm++; - - s = combine1 (ps++, pm); - d = *pd; - *pd++ = core_combine_saturate_u_pixel_sse2 (s, d); - if (pm) - pm++; - } - else - { - save_128_aligned ((__m128i*)pd, _mm_adds_epu8 (xmm_dst, xmm_src)); - - pd += 4; - ps += 4; - if (pm) - pm += 4; - } - - w -= 4; - } - - while (w--) - { - s = combine1 (ps, pm); - d = *pd; - - *pd++ = core_combine_saturate_u_pixel_sse2 (s, d); - ps++; - if (pm) - pm++; - } -} - -static force_inline void -core_combine_src_ca_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m))); - w--; - } - - while (w >= 4) - { - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m))); - w--; - } -} - -static force_inline uint32_t -core_combine_over_ca_pixel_sse2 (uint32_t src, - uint32_t mask, - uint32_t dst) -{ - __m64 s = unpack_32_1x64 (src); - __m64 expAlpha = expand_alpha_1x64 (s); - __m64 unpk_mask = unpack_32_1x64 (mask); - __m64 unpk_dst = unpack_32_1x64 (dst); - - return pack_1x64_32 (in_over_1x64 (&s, &expAlpha, &unpk_mask, &unpk_dst)); -} - -static force_inline void -core_combine_over_ca_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_alpha_lo, xmm_alpha_hi; - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_over_ca_pixel_sse2 (s, m, d); - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - in_over_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_over_ca_pixel_sse2 (s, m, d); - w--; - } -} - -static force_inline uint32_t -core_combine_over_reverse_ca_pixel_sse2 (uint32_t src, - uint32_t mask, - uint32_t dst) -{ - __m64 d = unpack_32_1x64 (dst); - - return pack_1x64_32 ( - over_1x64 (d, expand_alpha_1x64 (d), - pix_multiply_1x64 (unpack_32_1x64 (src), - unpack_32_1x64 (mask)))); -} - -static force_inline void -core_combine_over_reverse_ca_sse2 (uint32_t* pd, - const uint32_t* ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_alpha_lo, xmm_alpha_hi; - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_over_reverse_ca_pixel_sse2 (s, m, d); - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - over_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_mask_lo, &xmm_mask_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_mask_lo, xmm_mask_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_over_reverse_ca_pixel_sse2 (s, m, d); - w--; - } -} - -static force_inline void -core_combine_in_ca_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_alpha_lo, xmm_alpha_hi; - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m)), - expand_alpha_1x64 (unpack_32_1x64 (d)))); - - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst_lo, &xmm_dst_hi); - - pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - pix_multiply_1x64 ( - unpack_32_1x64 (s), unpack_32_1x64 (m)), - expand_alpha_1x64 (unpack_32_1x64 (d)))); - - w--; - } -} - -static force_inline void -core_combine_in_reverse_ca_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_alpha_lo, xmm_alpha_hi; - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (d), - pix_multiply_1x64 (unpack_32_1x64 (m), - expand_alpha_1x64 (unpack_32_1x64 (s))))); - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (d), - pix_multiply_1x64 (unpack_32_1x64 (m), - expand_alpha_1x64 (unpack_32_1x64 (s))))); - w--; - } -} - -static force_inline void -core_combine_out_ca_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_alpha_lo, xmm_alpha_hi; - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - pix_multiply_1x64 ( - unpack_32_1x64 (s), unpack_32_1x64 (m)), - negate_1x64 (expand_alpha_1x64 (unpack_32_1x64 (d))))); - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - negate_2x128 (xmm_alpha_lo, xmm_alpha_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst_lo, &xmm_dst_hi); - pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - pix_multiply_1x64 ( - unpack_32_1x64 (s), unpack_32_1x64 (m)), - negate_1x64 (expand_alpha_1x64 (unpack_32_1x64 (d))))); - - w--; - } -} - -static force_inline void -core_combine_out_reverse_ca_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_alpha_lo, xmm_alpha_hi; - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (d), - negate_1x64 (pix_multiply_1x64 ( - unpack_32_1x64 (m), - expand_alpha_1x64 (unpack_32_1x64 (s)))))); - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_mask_lo, &xmm_mask_hi); - - negate_2x128 (xmm_mask_lo, xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (d), - negate_1x64 (pix_multiply_1x64 ( - unpack_32_1x64 (m), - expand_alpha_1x64 (unpack_32_1x64 (s)))))); - w--; - } -} - -static force_inline uint32_t -core_combine_atop_ca_pixel_sse2 (uint32_t src, - uint32_t mask, - uint32_t dst) -{ - __m64 m = unpack_32_1x64 (mask); - __m64 s = unpack_32_1x64 (src); - __m64 d = unpack_32_1x64 (dst); - __m64 sa = expand_alpha_1x64 (s); - __m64 da = expand_alpha_1x64 (d); - - s = pix_multiply_1x64 (s, m); - m = negate_1x64 (pix_multiply_1x64 (m, sa)); - - return pack_1x64_32 (pix_add_multiply_1x64 (&d, &m, &s, &da)); -} - -static force_inline void -core_combine_atop_ca_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_alpha_src_lo, xmm_alpha_src_hi; - __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_atop_ca_pixel_sse2 (s, m, d); - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi); - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_src_lo, &xmm_src_hi); - pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi, - &xmm_mask_lo, &xmm_mask_hi); - - negate_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - pix_add_multiply_2x128 ( - &xmm_dst_lo, &xmm_dst_hi, &xmm_mask_lo, &xmm_mask_hi, - &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_atop_ca_pixel_sse2 (s, m, d); - w--; - } -} - -static force_inline uint32_t -core_combine_reverse_atop_ca_pixel_sse2 (uint32_t src, - uint32_t mask, - uint32_t dst) -{ - __m64 m = unpack_32_1x64 (mask); - __m64 s = unpack_32_1x64 (src); - __m64 d = unpack_32_1x64 (dst); - - __m64 da = negate_1x64 (expand_alpha_1x64 (d)); - __m64 sa = expand_alpha_1x64 (s); - - s = pix_multiply_1x64 (s, m); - m = pix_multiply_1x64 (m, sa); - - return pack_1x64_32 (pix_add_multiply_1x64 (&d, &m, &s, &da)); -} - -static force_inline void -core_combine_reverse_atop_ca_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_alpha_src_lo, xmm_alpha_src_hi; - __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_reverse_atop_ca_pixel_sse2 (s, m, d); - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi); - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_src_lo, &xmm_src_hi); - pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi, - &xmm_mask_lo, &xmm_mask_hi); - - negate_2x128 (xmm_alpha_dst_lo, xmm_alpha_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - pix_add_multiply_2x128 ( - &xmm_dst_lo, &xmm_dst_hi, &xmm_mask_lo, &xmm_mask_hi, - &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_reverse_atop_ca_pixel_sse2 (s, m, d); - w--; - } -} - -static force_inline uint32_t -core_combine_xor_ca_pixel_sse2 (uint32_t src, - uint32_t mask, - uint32_t dst) -{ - __m64 a = unpack_32_1x64 (mask); - __m64 s = unpack_32_1x64 (src); - __m64 d = unpack_32_1x64 (dst); - - __m64 alpha_dst = negate_1x64 (pix_multiply_1x64 ( - a, expand_alpha_1x64 (s))); - __m64 dest = pix_multiply_1x64 (s, a); - __m64 alpha_src = negate_1x64 (expand_alpha_1x64 (d)); - - return pack_1x64_32 (pix_add_multiply_1x64 (&d, - &alpha_dst, - &dest, - &alpha_src)); -} - -static force_inline void -core_combine_xor_ca_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_alpha_src_lo, xmm_alpha_src_hi; - __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_xor_ca_pixel_sse2 (s, m, d); - w--; - } - - while (w >= 4) - { - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi); - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_src_lo, &xmm_src_hi); - pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi, - &xmm_alpha_src_lo, &xmm_alpha_src_hi, - &xmm_mask_lo, &xmm_mask_hi); - - negate_2x128 (xmm_alpha_dst_lo, xmm_alpha_dst_hi, - &xmm_alpha_dst_lo, &xmm_alpha_dst_hi); - negate_2x128 (xmm_mask_lo, xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - pix_add_multiply_2x128 ( - &xmm_dst_lo, &xmm_dst_hi, &xmm_mask_lo, &xmm_mask_hi, - &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = core_combine_xor_ca_pixel_sse2 (s, m, d); - w--; - } -} - -static force_inline void -core_combine_add_ca_sse2 (uint32_t * pd, - const uint32_t *ps, - const uint32_t *pm, - int w) -{ - uint32_t s, m, d; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask_lo, xmm_mask_hi; - - while (w && (unsigned long)pd & 15) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - _mm_adds_pu8 (pix_multiply_1x64 (unpack_32_1x64 (s), - unpack_32_1x64 (m)), - unpack_32_1x64 (d))); - w--; - } - - while (w >= 4) - { - xmm_src_hi = load_128_unaligned ((__m128i*)ps); - xmm_mask_hi = load_128_unaligned ((__m128i*)pm); - xmm_dst_hi = load_128_aligned ((__m128i*)pd); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_src_lo, &xmm_src_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 ( - _mm_adds_epu8 (xmm_src_lo, xmm_dst_lo), - _mm_adds_epu8 (xmm_src_hi, xmm_dst_hi))); - - ps += 4; - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - s = *ps++; - m = *pm++; - d = *pd; - - *pd++ = pack_1x64_32 ( - _mm_adds_pu8 (pix_multiply_1x64 (unpack_32_1x64 (s), - unpack_32_1x64 (m)), - unpack_32_1x64 (d))); - w--; - } -} - -/* --------------------------------------------------- - * fb_compose_setup_sSE2 - */ -static force_inline __m64 -create_mask_16_64 (uint16_t mask) -{ - return _mm_set1_pi16 (mask); -} - -static force_inline __m128i -create_mask_16_128 (uint16_t mask) -{ - return _mm_set1_epi16 (mask); -} - -static force_inline __m64 -create_mask_2x32_64 (uint32_t mask0, - uint32_t mask1) -{ - return _mm_set_pi32 (mask0, mask1); -} - -/* Work around a code generation bug in Sun Studio 12. */ -#if defined(__SUNPRO_C) && (__SUNPRO_C >= 0x590) -# define create_mask_2x32_128(mask0, mask1) \ - (_mm_set_epi32 ((mask0), (mask1), (mask0), (mask1))) -#else -static force_inline __m128i -create_mask_2x32_128 (uint32_t mask0, - uint32_t mask1) -{ - return _mm_set_epi32 (mask0, mask1, mask0, mask1); -} -#endif - -/* SSE2 code patch for fbcompose.c */ - -static void -sse2_combine_over_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_over_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_over_reverse_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_over_reverse_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_in_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_in_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_in_reverse_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_reverse_in_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_out_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_out_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_out_reverse_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_reverse_out_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_atop_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_atop_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_atop_reverse_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_reverse_atop_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_xor_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_xor_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_add_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_add_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_saturate_u (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_saturate_u_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_src_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_src_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_over_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_over_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_over_reverse_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_over_reverse_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_in_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_in_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_in_reverse_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_in_reverse_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_out_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_out_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_out_reverse_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_out_reverse_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_atop_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_atop_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_atop_reverse_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_reverse_atop_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_xor_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_xor_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -static void -sse2_combine_add_ca (pixman_implementation_t *imp, - pixman_op_t op, - uint32_t * dst, - const uint32_t * src, - const uint32_t * mask, - int width) -{ - core_combine_add_ca_sse2 (dst, src, mask, width); - _mm_empty (); -} - -/* ------------------------------------------------------------------- - * composite_over_n_8888 - */ - -static void -sse2_composite_over_n_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src; - uint32_t *dst_line, *dst, d; - int32_t w; - int dst_stride; - __m128i xmm_src, xmm_alpha; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - - xmm_src = expand_pixel_32_1x128 (src); - xmm_alpha = expand_alpha_1x128 (xmm_src); - - while (height--) - { - dst = dst_line; - - dst_line += dst_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - d = *dst; - *dst++ = pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src), - _mm_movepi64_pi64 (xmm_alpha), - unpack_32_1x64 (d))); - w--; - } - - while (w >= 4) - { - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_dst_lo, &xmm_dst_hi); - - /* rebuid the 4 pixel data and save*/ - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - w -= 4; - dst += 4; - } - - while (w) - { - d = *dst; - *dst++ = pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src), - _mm_movepi64_pi64 (xmm_alpha), - unpack_32_1x64 (d))); - w--; - } - - } - _mm_empty (); -} - -/* --------------------------------------------------------------------- - * composite_over_n_0565 - */ -static void -sse2_composite_over_n_0565 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src; - uint16_t *dst_line, *dst, d; - int32_t w; - int dst_stride; - __m128i xmm_src, xmm_alpha; - __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - - xmm_src = expand_pixel_32_1x128 (src); - xmm_alpha = expand_alpha_1x128 (xmm_src); - - while (height--) - { - dst = dst_line; - - dst_line += dst_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - d = *dst; - - *dst++ = pack_565_32_16 ( - pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src), - _mm_movepi64_pi64 (xmm_alpha), - expand565_16_1x64 (d)))); - w--; - } - - while (w >= 8) - { - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_565_128_4x128 (xmm_dst, - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3); - - over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_dst0, &xmm_dst1); - over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_dst2, &xmm_dst3); - - xmm_dst = pack_565_4x128_128 ( - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3); - - save_128_aligned ((__m128i*)dst, xmm_dst); - - dst += 8; - w -= 8; - } - - while (w--) - { - d = *dst; - *dst++ = pack_565_32_16 ( - pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src), - _mm_movepi64_pi64 (xmm_alpha), - expand565_16_1x64 (d)))); - } - } - - _mm_empty (); -} - -/* ------------------------------ - * composite_add_n_8888_8888_ca - */ -static void -sse2_composite_add_n_8888_8888_ca (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint32_t *dst_line, d; - uint32_t *mask_line, m; - uint32_t pack_cmp; - int dst_stride, mask_stride; - - __m128i xmm_src, xmm_alpha; - __m128i xmm_dst; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - - __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - srca = src >> 24; - - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1); - - xmm_src = _mm_unpacklo_epi8 ( - create_mask_2x32_128 (src, src), _mm_setzero_si128 ()); - xmm_alpha = expand_alpha_1x128 (xmm_src); - mmx_src = _mm_movepi64_pi64 (xmm_src); - mmx_alpha = _mm_movepi64_pi64 (xmm_alpha); - - while (height--) - { - int w = width; - const uint32_t *pm = (uint32_t *)mask_line; - uint32_t *pd = (uint32_t *)dst_line; - - dst_line += dst_stride; - mask_line += mask_stride; - - while (w && (unsigned long)pd & 15) - { - m = *pm++; - - if (m) - { - d = *pd; - - mmx_mask = unpack_32_1x64 (m); - mmx_dest = unpack_32_1x64 (d); - - *pd = pack_1x64_32 ( - _mm_adds_pu8 (pix_multiply_1x64 (mmx_mask, mmx_src), mmx_dest)); - } - - pd++; - w--; - } - - while (w >= 4) - { - xmm_mask = load_128_unaligned ((__m128i*)pm); - - pack_cmp = - _mm_movemask_epi8 ( - _mm_cmpeq_epi32 (xmm_mask, _mm_setzero_si128 ())); - - /* if all bits in mask are zero, pack_cmp are equal to 0xffff */ - if (pack_cmp != 0xffff) - { - xmm_dst = load_128_aligned ((__m128i*)pd); - - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - - pix_multiply_2x128 (&xmm_src, &xmm_src, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - xmm_mask_hi = pack_2x128_128 (xmm_mask_lo, xmm_mask_hi); - - save_128_aligned ( - (__m128i*)pd, _mm_adds_epu8 (xmm_mask_hi, xmm_dst)); - } - - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - m = *pm++; - - if (m) - { - d = *pd; - - mmx_mask = unpack_32_1x64 (m); - mmx_dest = unpack_32_1x64 (d); - - *pd = pack_1x64_32 ( - _mm_adds_pu8 (pix_multiply_1x64 (mmx_mask, mmx_src), mmx_dest)); - } - - pd++; - w--; - } - } - - _mm_empty (); -} - -/* --------------------------------------------------------------------------- - * composite_over_n_8888_8888_ca - */ - -static void -sse2_composite_over_n_8888_8888_ca (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src; - uint32_t *dst_line, d; - uint32_t *mask_line, m; - uint32_t pack_cmp; - int dst_stride, mask_stride; - - __m128i xmm_src, xmm_alpha; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - - __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1); - - xmm_src = _mm_unpacklo_epi8 ( - create_mask_2x32_128 (src, src), _mm_setzero_si128 ()); - xmm_alpha = expand_alpha_1x128 (xmm_src); - mmx_src = _mm_movepi64_pi64 (xmm_src); - mmx_alpha = _mm_movepi64_pi64 (xmm_alpha); - - while (height--) - { - int w = width; - const uint32_t *pm = (uint32_t *)mask_line; - uint32_t *pd = (uint32_t *)dst_line; - - dst_line += dst_stride; - mask_line += mask_stride; - - while (w && (unsigned long)pd & 15) - { - m = *pm++; - - if (m) - { - d = *pd; - mmx_mask = unpack_32_1x64 (m); - mmx_dest = unpack_32_1x64 (d); - - *pd = pack_1x64_32 (in_over_1x64 (&mmx_src, - &mmx_alpha, - &mmx_mask, - &mmx_dest)); - } - - pd++; - w--; - } - - while (w >= 4) - { - xmm_mask = load_128_unaligned ((__m128i*)pm); - - pack_cmp = - _mm_movemask_epi8 ( - _mm_cmpeq_epi32 (xmm_mask, _mm_setzero_si128 ())); - - /* if all bits in mask are zero, pack_cmp are equal to 0xffff */ - if (pack_cmp != 0xffff) - { - xmm_dst = load_128_aligned ((__m128i*)pd); - - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - in_over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - - pd += 4; - pm += 4; - w -= 4; - } - - while (w) - { - m = *pm++; - - if (m) - { - d = *pd; - mmx_mask = unpack_32_1x64 (m); - mmx_dest = unpack_32_1x64 (d); - - *pd = pack_1x64_32 ( - in_over_1x64 (&mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)); - } - - pd++; - w--; - } - } - - _mm_empty (); -} - -/*--------------------------------------------------------------------- - * composite_over_8888_n_8888 - */ - -static void -sse2_composite_over_8888_n_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst; - uint32_t *src_line, *src; - uint32_t mask; - int32_t w; - int dst_stride, src_stride; - - __m128i xmm_mask; - __m128i xmm_src, xmm_src_lo, xmm_src_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_alpha_lo, xmm_alpha_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - mask = _pixman_image_get_solid (imp, mask_image, PIXMAN_a8r8g8b8); - - xmm_mask = create_mask_16_128 (mask >> 24); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - uint32_t s = *src++; - - if (s) - { - uint32_t d = *dst; - - __m64 ms = unpack_32_1x64 (s); - __m64 alpha = expand_alpha_1x64 (ms); - __m64 dest = _mm_movepi64_pi64 (xmm_mask); - __m64 alpha_dst = unpack_32_1x64 (d); - - *dst = pack_1x64_32 ( - in_over_1x64 (&ms, &alpha, &dest, &alpha_dst)); - } - dst++; - w--; - } - - while (w >= 4) - { - xmm_src = load_128_unaligned ((__m128i*)src); - - if (!is_zero (xmm_src)) - { - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - in_over_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_mask, &xmm_mask, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - - dst += 4; - src += 4; - w -= 4; - } - - while (w) - { - uint32_t s = *src++; - - if (s) - { - uint32_t d = *dst; - - __m64 ms = unpack_32_1x64 (s); - __m64 alpha = expand_alpha_1x64 (ms); - __m64 mask = _mm_movepi64_pi64 (xmm_mask); - __m64 dest = unpack_32_1x64 (d); - - *dst = pack_1x64_32 ( - in_over_1x64 (&ms, &alpha, &mask, &dest)); - } - - dst++; - w--; - } - } - - _mm_empty (); -} - -/*--------------------------------------------------------------------- - * composite_over_8888_n_8888 - */ - -static void -sse2_composite_src_x888_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst; - uint32_t *src_line, *src; - int32_t w; - int dst_stride, src_stride; - - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - *dst++ = *src++ | 0xff000000; - w--; - } - - while (w >= 16) - { - __m128i xmm_src1, xmm_src2, xmm_src3, xmm_src4; - - xmm_src1 = load_128_unaligned ((__m128i*)src + 0); - xmm_src2 = load_128_unaligned ((__m128i*)src + 1); - xmm_src3 = load_128_unaligned ((__m128i*)src + 2); - xmm_src4 = load_128_unaligned ((__m128i*)src + 3); - - save_128_aligned ((__m128i*)dst + 0, _mm_or_si128 (xmm_src1, mask_ff000000)); - save_128_aligned ((__m128i*)dst + 1, _mm_or_si128 (xmm_src2, mask_ff000000)); - save_128_aligned ((__m128i*)dst + 2, _mm_or_si128 (xmm_src3, mask_ff000000)); - save_128_aligned ((__m128i*)dst + 3, _mm_or_si128 (xmm_src4, mask_ff000000)); - - dst += 16; - src += 16; - w -= 16; - } - - while (w) - { - *dst++ = *src++ | 0xff000000; - w--; - } - } - - _mm_empty (); -} - -/* --------------------------------------------------------------------- - * composite_over_x888_n_8888 - */ -static void -sse2_composite_over_x888_n_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst; - uint32_t *src_line, *src; - uint32_t mask; - int dst_stride, src_stride; - int32_t w; - - __m128i xmm_mask, xmm_alpha; - __m128i xmm_src, xmm_src_lo, xmm_src_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - mask = _pixman_image_get_solid (imp, mask_image, PIXMAN_a8r8g8b8); - - xmm_mask = create_mask_16_128 (mask >> 24); - xmm_alpha = mask_00ff; - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - uint32_t s = (*src++) | 0xff000000; - uint32_t d = *dst; - - __m64 src = unpack_32_1x64 (s); - __m64 alpha = _mm_movepi64_pi64 (xmm_alpha); - __m64 mask = _mm_movepi64_pi64 (xmm_mask); - __m64 dest = unpack_32_1x64 (d); - - *dst++ = pack_1x64_32 ( - in_over_1x64 (&src, &alpha, &mask, &dest)); - - w--; - } - - while (w >= 4) - { - xmm_src = _mm_or_si128 ( - load_128_unaligned ((__m128i*)src), mask_ff000000); - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - in_over_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_alpha, &xmm_alpha, - &xmm_mask, &xmm_mask, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - dst += 4; - src += 4; - w -= 4; - - } - - while (w) - { - uint32_t s = (*src++) | 0xff000000; - uint32_t d = *dst; - - __m64 src = unpack_32_1x64 (s); - __m64 alpha = _mm_movepi64_pi64 (xmm_alpha); - __m64 mask = _mm_movepi64_pi64 (xmm_mask); - __m64 dest = unpack_32_1x64 (d); - - *dst++ = pack_1x64_32 ( - in_over_1x64 (&src, &alpha, &mask, &dest)); - - w--; - } - } - - _mm_empty (); -} - -/* -------------------------------------------------------------------- - * composite_over_8888_8888 - */ -static void -sse2_composite_over_8888_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - int dst_stride, src_stride; - uint32_t *dst_line, *dst; - uint32_t *src_line, *src; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - dst = dst_line; - src = src_line; - - while (height--) - { - core_combine_over_u_sse2 (dst, src, NULL, width); - - dst += dst_stride; - src += src_stride; - } - _mm_empty (); -} - -/* ------------------------------------------------------------------ - * composite_over_8888_0565 - */ -static force_inline uint16_t -composite_over_8888_0565pixel (uint32_t src, uint16_t dst) -{ - __m64 ms; - - ms = unpack_32_1x64 (src); - return pack_565_32_16 ( - pack_1x64_32 ( - over_1x64 ( - ms, expand_alpha_1x64 (ms), expand565_16_1x64 (dst)))); -} - -static void -sse2_composite_over_8888_0565 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint16_t *dst_line, *dst, d; - uint32_t *src_line, *src, s; - int dst_stride, src_stride; - int32_t w; - - __m128i xmm_alpha_lo, xmm_alpha_hi; - __m128i xmm_src, xmm_src_lo, xmm_src_hi; - __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - -#if 0 - /* FIXME - * - * I copy the code from MMX one and keep the fixme. - * If it's a problem there, probably is a problem here. - */ - assert (src_image->drawable == mask_image->drawable); -#endif - - while (height--) - { - dst = dst_line; - src = src_line; - - dst_line += dst_stride; - src_line += src_stride; - w = width; - - /* Align dst on a 16-byte boundary */ - while (w && - ((unsigned long)dst & 15)) - { - s = *src++; - d = *dst; - - *dst++ = composite_over_8888_0565pixel (s, d); - w--; - } - - /* It's a 8 pixel loop */ - while (w >= 8) - { - /* I'm loading unaligned because I'm not sure - * about the address alignment. - */ - xmm_src = load_128_unaligned ((__m128i*) src); - xmm_dst = load_128_aligned ((__m128i*) dst); - - /* Unpacking */ - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_565_128_4x128 (xmm_dst, - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3); - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - /* I'm loading next 4 pixels from memory - * before to optimze the memory read. - */ - xmm_src = load_128_unaligned ((__m128i*) (src + 4)); - - over_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_dst0, &xmm_dst1); - - /* Unpacking */ - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - over_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_dst2, &xmm_dst3); - - save_128_aligned ( - (__m128i*)dst, pack_565_4x128_128 ( - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3)); - - w -= 8; - dst += 8; - src += 8; - } - - while (w--) - { - s = *src++; - d = *dst; - - *dst++ = composite_over_8888_0565pixel (s, d); - } - } - - _mm_empty (); -} - -/* ----------------------------------------------------------------- - * composite_over_n_8_8888 - */ - -static void -sse2_composite_over_n_8_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint32_t *dst_line, *dst; - uint8_t *mask_line, *mask; - int dst_stride, mask_stride; - int32_t w; - uint32_t m, d; - - __m128i xmm_src, xmm_alpha, xmm_def; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - - __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - srca = src >> 24; - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - xmm_def = create_mask_2x32_128 (src, src); - xmm_src = expand_pixel_32_1x128 (src); - xmm_alpha = expand_alpha_1x128 (xmm_src); - mmx_src = _mm_movepi64_pi64 (xmm_src); - mmx_alpha = _mm_movepi64_pi64 (xmm_alpha); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - uint8_t m = *mask++; - - if (m) - { - d = *dst; - mmx_mask = expand_pixel_8_1x64 (m); - mmx_dest = unpack_32_1x64 (d); - - *dst = pack_1x64_32 (in_over_1x64 (&mmx_src, - &mmx_alpha, - &mmx_mask, - &mmx_dest)); - } - - w--; - dst++; - } - - while (w >= 4) - { - m = *((uint32_t*)mask); - - if (srca == 0xff && m == 0xffffffff) - { - save_128_aligned ((__m128i*)dst, xmm_def); - } - else if (m) - { - xmm_dst = load_128_aligned ((__m128i*) dst); - xmm_mask = unpack_32_1x128 (m); - xmm_mask = _mm_unpacklo_epi8 (xmm_mask, _mm_setzero_si128 ()); - - /* Unpacking */ - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - in_over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - - w -= 4; - dst += 4; - mask += 4; - } - - while (w) - { - uint8_t m = *mask++; - - if (m) - { - d = *dst; - mmx_mask = expand_pixel_8_1x64 (m); - mmx_dest = unpack_32_1x64 (d); - - *dst = pack_1x64_32 (in_over_1x64 (&mmx_src, - &mmx_alpha, - &mmx_mask, - &mmx_dest)); - } - - w--; - dst++; - } - } - - _mm_empty (); -} - -/* ---------------------------------------------------------------- - * composite_over_n_8_8888 - */ - -pixman_bool_t -pixman_fill_sse2 (uint32_t *bits, - int stride, - int bpp, - int x, - int y, - int width, - int height, - uint32_t data) -{ - uint32_t byte_width; - uint8_t *byte_line; - - __m128i xmm_def; - - if (bpp == 8) - { - uint8_t b; - uint16_t w; - - stride = stride * (int) sizeof (uint32_t) / 1; - byte_line = (uint8_t *)(((uint8_t *)bits) + stride * y + x); - byte_width = width; - stride *= 1; - - b = data & 0xff; - w = (b << 8) | b; - data = (w << 16) | w; - } - else if (bpp == 16) - { - stride = stride * (int) sizeof (uint32_t) / 2; - byte_line = (uint8_t *)(((uint16_t *)bits) + stride * y + x); - byte_width = 2 * width; - stride *= 2; - - data = (data & 0xffff) * 0x00010001; - } - else if (bpp == 32) - { - stride = stride * (int) sizeof (uint32_t) / 4; - byte_line = (uint8_t *)(((uint32_t *)bits) + stride * y + x); - byte_width = 4 * width; - stride *= 4; - } - else - { - return FALSE; - } - - xmm_def = create_mask_2x32_128 (data, data); - - while (height--) - { - int w; - uint8_t *d = byte_line; - byte_line += stride; - w = byte_width; - - while (w >= 1 && ((unsigned long)d & 1)) - { - *(uint8_t *)d = data; - w -= 1; - d += 1; - } - - while (w >= 2 && ((unsigned long)d & 3)) - { - *(uint16_t *)d = data; - w -= 2; - d += 2; - } - - while (w >= 4 && ((unsigned long)d & 15)) - { - *(uint32_t *)d = data; - - w -= 4; - d += 4; - } - - while (w >= 128) - { - save_128_aligned ((__m128i*)(d), xmm_def); - save_128_aligned ((__m128i*)(d + 16), xmm_def); - save_128_aligned ((__m128i*)(d + 32), xmm_def); - save_128_aligned ((__m128i*)(d + 48), xmm_def); - save_128_aligned ((__m128i*)(d + 64), xmm_def); - save_128_aligned ((__m128i*)(d + 80), xmm_def); - save_128_aligned ((__m128i*)(d + 96), xmm_def); - save_128_aligned ((__m128i*)(d + 112), xmm_def); - - d += 128; - w -= 128; - } - - if (w >= 64) - { - save_128_aligned ((__m128i*)(d), xmm_def); - save_128_aligned ((__m128i*)(d + 16), xmm_def); - save_128_aligned ((__m128i*)(d + 32), xmm_def); - save_128_aligned ((__m128i*)(d + 48), xmm_def); - - d += 64; - w -= 64; - } - - if (w >= 32) - { - save_128_aligned ((__m128i*)(d), xmm_def); - save_128_aligned ((__m128i*)(d + 16), xmm_def); - - d += 32; - w -= 32; - } - - if (w >= 16) - { - save_128_aligned ((__m128i*)(d), xmm_def); - - d += 16; - w -= 16; - } - - while (w >= 4) - { - *(uint32_t *)d = data; - - w -= 4; - d += 4; - } - - if (w >= 2) - { - *(uint16_t *)d = data; - w -= 2; - d += 2; - } - - if (w >= 1) - { - *(uint8_t *)d = data; - w -= 1; - d += 1; - } - } - - _mm_empty (); - return TRUE; -} - -static void -sse2_composite_src_n_8_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint32_t *dst_line, *dst; - uint8_t *mask_line, *mask; - int dst_stride, mask_stride; - int32_t w; - uint32_t m; - - __m128i xmm_src, xmm_def; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - srca = src >> 24; - if (src == 0) - { - pixman_fill_sse2 (dst_image->bits.bits, dst_image->bits.rowstride, - PIXMAN_FORMAT_BPP (dst_image->bits.format), - dest_x, dest_y, width, height, 0); - return; - } - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - xmm_def = create_mask_2x32_128 (src, src); - xmm_src = expand_pixel_32_1x128 (src); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - uint8_t m = *mask++; - - if (m) - { - *dst = pack_1x64_32 ( - pix_multiply_1x64 ( - _mm_movepi64_pi64 (xmm_src), expand_pixel_8_1x64 (m))); - } - else - { - *dst = 0; - } - - w--; - dst++; - } - - while (w >= 4) - { - m = *((uint32_t*)mask); - - if (srca == 0xff && m == 0xffffffff) - { - save_128_aligned ((__m128i*)dst, xmm_def); - } - else if (m) - { - xmm_mask = unpack_32_1x128 (m); - xmm_mask = _mm_unpacklo_epi8 (xmm_mask, _mm_setzero_si128 ()); - - /* Unpacking */ - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - pix_multiply_2x128 (&xmm_src, &xmm_src, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_mask_lo, xmm_mask_hi)); - } - else - { - save_128_aligned ((__m128i*)dst, _mm_setzero_si128 ()); - } - - w -= 4; - dst += 4; - mask += 4; - } - - while (w) - { - uint8_t m = *mask++; - - if (m) - { - *dst = pack_1x64_32 ( - pix_multiply_1x64 ( - _mm_movepi64_pi64 (xmm_src), expand_pixel_8_1x64 (m))); - } - else - { - *dst = 0; - } - - w--; - dst++; - } - } - - _mm_empty (); -} - -/*----------------------------------------------------------------------- - * composite_over_n_8_0565 - */ - -static void -sse2_composite_over_n_8_0565 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src, srca; - uint16_t *dst_line, *dst, d; - uint8_t *mask_line, *mask; - int dst_stride, mask_stride; - int32_t w; - uint32_t m; - __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest; - - __m128i xmm_src, xmm_alpha; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - srca = src >> 24; - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - xmm_src = expand_pixel_32_1x128 (src); - xmm_alpha = expand_alpha_1x128 (xmm_src); - mmx_src = _mm_movepi64_pi64 (xmm_src); - mmx_alpha = _mm_movepi64_pi64 (xmm_alpha); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - m = *mask++; - - if (m) - { - d = *dst; - mmx_mask = expand_alpha_rev_1x64 (unpack_32_1x64 (m)); - mmx_dest = expand565_16_1x64 (d); - - *dst = pack_565_32_16 ( - pack_1x64_32 ( - in_over_1x64 ( - &mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest))); - } - - w--; - dst++; - } - - while (w >= 8) - { - xmm_dst = load_128_aligned ((__m128i*) dst); - unpack_565_128_4x128 (xmm_dst, - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3); - - m = *((uint32_t*)mask); - mask += 4; - - if (m) - { - xmm_mask = unpack_32_1x128 (m); - xmm_mask = _mm_unpacklo_epi8 (xmm_mask, _mm_setzero_si128 ()); - - /* Unpacking */ - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - in_over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst0, &xmm_dst1); - } - - m = *((uint32_t*)mask); - mask += 4; - - if (m) - { - xmm_mask = unpack_32_1x128 (m); - xmm_mask = _mm_unpacklo_epi8 (xmm_mask, _mm_setzero_si128 ()); - - /* Unpacking */ - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - - expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - in_over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst2, &xmm_dst3); - } - - save_128_aligned ( - (__m128i*)dst, pack_565_4x128_128 ( - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3)); - - w -= 8; - dst += 8; - } - - while (w) - { - m = *mask++; - - if (m) - { - d = *dst; - mmx_mask = expand_alpha_rev_1x64 (unpack_32_1x64 (m)); - mmx_dest = expand565_16_1x64 (d); - - *dst = pack_565_32_16 ( - pack_1x64_32 ( - in_over_1x64 ( - &mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest))); - } - - w--; - dst++; - } - } - - _mm_empty (); -} - -/* ----------------------------------------------------------------------- - * composite_over_pixbuf_0565 - */ - -static void -sse2_composite_over_pixbuf_0565 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint16_t *dst_line, *dst, d; - uint32_t *src_line, *src, s; - int dst_stride, src_stride; - int32_t w; - uint32_t opaque, zero; - - __m64 ms; - __m128i xmm_src, xmm_src_lo, xmm_src_hi; - __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - -#if 0 - /* FIXME - * - * I copy the code from MMX one and keep the fixme. - * If it's a problem there, probably is a problem here. - */ - assert (src_image->drawable == mask_image->drawable); -#endif - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - s = *src++; - d = *dst; - - ms = unpack_32_1x64 (s); - - *dst++ = pack_565_32_16 ( - pack_1x64_32 ( - over_rev_non_pre_1x64 (ms, expand565_16_1x64 (d)))); - w--; - } - - while (w >= 8) - { - /* First round */ - xmm_src = load_128_unaligned ((__m128i*)src); - xmm_dst = load_128_aligned ((__m128i*)dst); - - opaque = is_opaque (xmm_src); - zero = is_zero (xmm_src); - - unpack_565_128_4x128 (xmm_dst, - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3); - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - - /* preload next round*/ - xmm_src = load_128_unaligned ((__m128i*)(src + 4)); - - if (opaque) - { - invert_colors_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_dst0, &xmm_dst1); - } - else if (!zero) - { - over_rev_non_pre_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_dst0, &xmm_dst1); - } - - /* Second round */ - opaque = is_opaque (xmm_src); - zero = is_zero (xmm_src); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - - if (opaque) - { - invert_colors_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_dst2, &xmm_dst3); - } - else if (!zero) - { - over_rev_non_pre_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_dst2, &xmm_dst3); - } - - save_128_aligned ( - (__m128i*)dst, pack_565_4x128_128 ( - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3)); - - w -= 8; - src += 8; - dst += 8; - } - - while (w) - { - s = *src++; - d = *dst; - - ms = unpack_32_1x64 (s); - - *dst++ = pack_565_32_16 ( - pack_1x64_32 ( - over_rev_non_pre_1x64 (ms, expand565_16_1x64 (d)))); - w--; - } - } - - _mm_empty (); -} - -/* ------------------------------------------------------------------------- - * composite_over_pixbuf_8888 - */ - -static void -sse2_composite_over_pixbuf_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst, d; - uint32_t *src_line, *src, s; - int dst_stride, src_stride; - int32_t w; - uint32_t opaque, zero; - - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_dst_lo, xmm_dst_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - -#if 0 - /* FIXME - * - * I copy the code from MMX one and keep the fixme. - * If it's a problem there, probably is a problem here. - */ - assert (src_image->drawable == mask_image->drawable); -#endif - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - s = *src++; - d = *dst; - - *dst++ = pack_1x64_32 ( - over_rev_non_pre_1x64 ( - unpack_32_1x64 (s), unpack_32_1x64 (d))); - - w--; - } - - while (w >= 4) - { - xmm_src_hi = load_128_unaligned ((__m128i*)src); - - opaque = is_opaque (xmm_src_hi); - zero = is_zero (xmm_src_hi); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - - if (opaque) - { - invert_colors_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - else if (!zero) - { - xmm_dst_hi = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - over_rev_non_pre_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - - w -= 4; - dst += 4; - src += 4; - } - - while (w) - { - s = *src++; - d = *dst; - - *dst++ = pack_1x64_32 ( - over_rev_non_pre_1x64 ( - unpack_32_1x64 (s), unpack_32_1x64 (d))); - - w--; - } - } - - _mm_empty (); -} - -/* ------------------------------------------------------------------------------------------------- - * composite_over_n_8888_0565_ca - */ - -static void -sse2_composite_over_n_8888_0565_ca (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src; - uint16_t *dst_line, *dst, d; - uint32_t *mask_line, *mask, m; - int dst_stride, mask_stride; - int w; - uint32_t pack_cmp; - - __m128i xmm_src, xmm_alpha; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3; - - __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1); - - xmm_src = expand_pixel_32_1x128 (src); - xmm_alpha = expand_alpha_1x128 (xmm_src); - mmx_src = _mm_movepi64_pi64 (xmm_src); - mmx_alpha = _mm_movepi64_pi64 (xmm_alpha); - - while (height--) - { - w = width; - mask = mask_line; - dst = dst_line; - mask_line += mask_stride; - dst_line += dst_stride; - - while (w && ((unsigned long)dst & 15)) - { - m = *(uint32_t *) mask; - - if (m) - { - d = *dst; - mmx_mask = unpack_32_1x64 (m); - mmx_dest = expand565_16_1x64 (d); - - *dst = pack_565_32_16 ( - pack_1x64_32 ( - in_over_1x64 ( - &mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest))); - } - - w--; - dst++; - mask++; - } - - while (w >= 8) - { - /* First round */ - xmm_mask = load_128_unaligned ((__m128i*)mask); - xmm_dst = load_128_aligned ((__m128i*)dst); - - pack_cmp = _mm_movemask_epi8 ( - _mm_cmpeq_epi32 (xmm_mask, _mm_setzero_si128 ())); - - unpack_565_128_4x128 (xmm_dst, - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3); - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - - /* preload next round */ - xmm_mask = load_128_unaligned ((__m128i*)(mask + 4)); - - /* preload next round */ - if (pack_cmp != 0xffff) - { - in_over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst0, &xmm_dst1); - } - - /* Second round */ - pack_cmp = _mm_movemask_epi8 ( - _mm_cmpeq_epi32 (xmm_mask, _mm_setzero_si128 ())); - - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - - if (pack_cmp != 0xffff) - { - in_over_2x128 (&xmm_src, &xmm_src, - &xmm_alpha, &xmm_alpha, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_dst2, &xmm_dst3); - } - - save_128_aligned ( - (__m128i*)dst, pack_565_4x128_128 ( - &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3)); - - w -= 8; - dst += 8; - mask += 8; - } - - while (w) - { - m = *(uint32_t *) mask; - - if (m) - { - d = *dst; - mmx_mask = unpack_32_1x64 (m); - mmx_dest = expand565_16_1x64 (d); - - *dst = pack_565_32_16 ( - pack_1x64_32 ( - in_over_1x64 ( - &mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest))); - } - - w--; - dst++; - mask++; - } - } - - _mm_empty (); -} - -/* ----------------------------------------------------------------------- - * composite_in_n_8_8 - */ - -static void -sse2_composite_in_n_8_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - uint8_t *mask_line, *mask; - int dst_stride, mask_stride; - uint32_t d, m; - uint32_t src; - uint8_t sa; - int32_t w; - - __m128i xmm_alpha; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - sa = src >> 24; - - xmm_alpha = expand_alpha_1x128 (expand_pixel_32_1x128 (src)); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w && ((unsigned long)dst & 15)) - { - m = (uint32_t) *mask++; - d = (uint32_t) *dst; - - *dst++ = (uint8_t) pack_1x64_32 ( - pix_multiply_1x64 ( - pix_multiply_1x64 (_mm_movepi64_pi64 (xmm_alpha), - unpack_32_1x64 (m)), - unpack_32_1x64 (d))); - w--; - } - - while (w >= 16) - { - xmm_mask = load_128_unaligned ((__m128i*)mask); - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - pix_multiply_2x128 (&xmm_alpha, &xmm_alpha, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi, - &xmm_dst_lo, &xmm_dst_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - mask += 16; - dst += 16; - w -= 16; - } - - while (w) - { - m = (uint32_t) *mask++; - d = (uint32_t) *dst; - - *dst++ = (uint8_t) pack_1x64_32 ( - pix_multiply_1x64 ( - pix_multiply_1x64 ( - _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)), - unpack_32_1x64 (d))); - w--; - } - } - - _mm_empty (); -} - -/* ----------------------------------------------------------------------- - * composite_in_n_8 - */ - -static void -sse2_composite_in_n_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - int dst_stride; - uint32_t d; - uint32_t src; - int32_t w; - - __m128i xmm_alpha; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - xmm_alpha = expand_alpha_1x128 (expand_pixel_32_1x128 (src)); - - src = src >> 24; - - if (src == 0xff) - return; - - if (src == 0x00) - { - pixman_fill (dst_image->bits.bits, dst_image->bits.rowstride, - 8, dest_x, dest_y, width, height, src); - - return; - } - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - w = width; - - while (w && ((unsigned long)dst & 15)) - { - d = (uint32_t) *dst; - - *dst++ = (uint8_t) pack_1x64_32 ( - pix_multiply_1x64 ( - _mm_movepi64_pi64 (xmm_alpha), - unpack_32_1x64 (d))); - w--; - } - - while (w >= 16) - { - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - pix_multiply_2x128 (&xmm_alpha, &xmm_alpha, - &xmm_dst_lo, &xmm_dst_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - dst += 16; - w -= 16; - } - - while (w) - { - d = (uint32_t) *dst; - - *dst++ = (uint8_t) pack_1x64_32 ( - pix_multiply_1x64 ( - _mm_movepi64_pi64 (xmm_alpha), - unpack_32_1x64 (d))); - w--; - } - } - - _mm_empty (); -} - -/* --------------------------------------------------------------------------- - * composite_in_8_8 - */ - -static void -sse2_composite_in_8_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - uint8_t *src_line, *src; - int src_stride, dst_stride; - int32_t w; - uint32_t s, d; - - __m128i xmm_src, xmm_src_lo, xmm_src_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint8_t, src_stride, src_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - w = width; - - while (w && ((unsigned long)dst & 15)) - { - s = (uint32_t) *src++; - d = (uint32_t) *dst; - - *dst++ = (uint8_t) pack_1x64_32 ( - pix_multiply_1x64 ( - unpack_32_1x64 (s), unpack_32_1x64 (d))); - w--; - } - - while (w >= 16) - { - xmm_src = load_128_unaligned ((__m128i*)src); - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_dst_lo, &xmm_dst_hi, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - src += 16; - dst += 16; - w -= 16; - } - - while (w) - { - s = (uint32_t) *src++; - d = (uint32_t) *dst; - - *dst++ = (uint8_t) pack_1x64_32 ( - pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (d))); - w--; - } - } - - _mm_empty (); -} - -/* ------------------------------------------------------------------------- - * composite_add_n_8_8 - */ - -static void -sse2_composite_add_n_8_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - uint8_t *mask_line, *mask; - int dst_stride, mask_stride; - int32_t w; - uint32_t src; - uint8_t sa; - uint32_t m, d; - - __m128i xmm_alpha; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - sa = src >> 24; - - xmm_alpha = expand_alpha_1x128 (expand_pixel_32_1x128 (src)); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - w = width; - - while (w && ((unsigned long)dst & 15)) - { - m = (uint32_t) *mask++; - d = (uint32_t) *dst; - - *dst++ = (uint8_t) pack_1x64_32 ( - _mm_adds_pu16 ( - pix_multiply_1x64 ( - _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)), - unpack_32_1x64 (d))); - w--; - } - - while (w >= 16) - { - xmm_mask = load_128_unaligned ((__m128i*)mask); - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - pix_multiply_2x128 (&xmm_alpha, &xmm_alpha, - &xmm_mask_lo, &xmm_mask_hi, - &xmm_mask_lo, &xmm_mask_hi); - - xmm_dst_lo = _mm_adds_epu16 (xmm_mask_lo, xmm_dst_lo); - xmm_dst_hi = _mm_adds_epu16 (xmm_mask_hi, xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - - mask += 16; - dst += 16; - w -= 16; - } - - while (w) - { - m = (uint32_t) *mask++; - d = (uint32_t) *dst; - - *dst++ = (uint8_t) pack_1x64_32 ( - _mm_adds_pu16 ( - pix_multiply_1x64 ( - _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)), - unpack_32_1x64 (d))); - - w--; - } - } - - _mm_empty (); -} - -/* ------------------------------------------------------------------------- - * composite_add_n_8_8 - */ - -static void -sse2_composite_add_n_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - int dst_stride; - int32_t w; - uint32_t src; - - __m128i xmm_src; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - src >>= 24; - - if (src == 0x00) - return; - - if (src == 0xff) - { - pixman_fill (dst_image->bits.bits, dst_image->bits.rowstride, - 8, dest_x, dest_y, width, height, 0xff); - - return; - } - - src = (src << 24) | (src << 16) | (src << 8) | src; - xmm_src = _mm_set_epi32 (src, src, src, src); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - w = width; - - while (w && ((unsigned long)dst & 15)) - { - *dst = (uint8_t)_mm_cvtsi64_si32 ( - _mm_adds_pu8 ( - _mm_movepi64_pi64 (xmm_src), - _mm_cvtsi32_si64 (*dst))); - - w--; - dst++; - } - - while (w >= 16) - { - save_128_aligned ( - (__m128i*)dst, _mm_adds_epu8 (xmm_src, load_128_aligned ((__m128i*)dst))); - - dst += 16; - w -= 16; - } - - while (w) - { - *dst = (uint8_t)_mm_cvtsi64_si32 ( - _mm_adds_pu8 ( - _mm_movepi64_pi64 (xmm_src), - _mm_cvtsi32_si64 (*dst))); - - w--; - dst++; - } - } - - _mm_empty (); -} - -/* ---------------------------------------------------------------------- - * composite_add_8_8 - */ - -static void -sse2_composite_add_8_8 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint8_t *dst_line, *dst; - uint8_t *src_line, *src; - int dst_stride, src_stride; - int32_t w; - uint16_t t; - - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint8_t, src_stride, src_line, 1); - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1); - - while (height--) - { - dst = dst_line; - src = src_line; - - dst_line += dst_stride; - src_line += src_stride; - w = width; - - /* Small head */ - while (w && (unsigned long)dst & 3) - { - t = (*dst) + (*src++); - *dst++ = t | (0 - (t >> 8)); - w--; - } - - core_combine_add_u_sse2 ((uint32_t*)dst, (uint32_t*)src, NULL, w >> 2); - - /* Small tail */ - dst += w & 0xfffc; - src += w & 0xfffc; - - w &= 3; - - while (w) - { - t = (*dst) + (*src++); - *dst++ = t | (0 - (t >> 8)); - w--; - } - } - - _mm_empty (); -} - -/* --------------------------------------------------------------------- - * composite_add_8888_8888 - */ -static void -sse2_composite_add_8888_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *dst_line, *dst; - uint32_t *src_line, *src; - int dst_stride, src_stride; - - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - - while (height--) - { - dst = dst_line; - dst_line += dst_stride; - src = src_line; - src_line += src_stride; - - core_combine_add_u_sse2 (dst, src, NULL, width); - } - - _mm_empty (); -} - -/* ------------------------------------------------------------------------------------------------- - * sse2_composite_copy_area - */ - -static pixman_bool_t -pixman_blt_sse2 (uint32_t *src_bits, - uint32_t *dst_bits, - int src_stride, - int dst_stride, - int src_bpp, - int dst_bpp, - int src_x, - int src_y, - int dst_x, - int dst_y, - int width, - int height) -{ - uint8_t * src_bytes; - uint8_t * dst_bytes; - int byte_width; - - if (src_bpp != dst_bpp) - return FALSE; - - if (src_bpp == 16) - { - src_stride = src_stride * (int) sizeof (uint32_t) / 2; - dst_stride = dst_stride * (int) sizeof (uint32_t) / 2; - src_bytes =(uint8_t *)(((uint16_t *)src_bits) + src_stride * (src_y) + (src_x)); - dst_bytes = (uint8_t *)(((uint16_t *)dst_bits) + dst_stride * (dst_y) + (dst_x)); - byte_width = 2 * width; - src_stride *= 2; - dst_stride *= 2; - } - else if (src_bpp == 32) - { - src_stride = src_stride * (int) sizeof (uint32_t) / 4; - dst_stride = dst_stride * (int) sizeof (uint32_t) / 4; - src_bytes = (uint8_t *)(((uint32_t *)src_bits) + src_stride * (src_y) + (src_x)); - dst_bytes = (uint8_t *)(((uint32_t *)dst_bits) + dst_stride * (dst_y) + (dst_x)); - byte_width = 4 * width; - src_stride *= 4; - dst_stride *= 4; - } - else - { - return FALSE; - } - - while (height--) - { - int w; - uint8_t *s = src_bytes; - uint8_t *d = dst_bytes; - src_bytes += src_stride; - dst_bytes += dst_stride; - w = byte_width; - - while (w >= 2 && ((unsigned long)d & 3)) - { - *(uint16_t *)d = *(uint16_t *)s; - w -= 2; - s += 2; - d += 2; - } - - while (w >= 4 && ((unsigned long)d & 15)) - { - *(uint32_t *)d = *(uint32_t *)s; - - w -= 4; - s += 4; - d += 4; - } - - while (w >= 64) - { - __m128i xmm0, xmm1, xmm2, xmm3; - - xmm0 = load_128_unaligned ((__m128i*)(s)); - xmm1 = load_128_unaligned ((__m128i*)(s + 16)); - xmm2 = load_128_unaligned ((__m128i*)(s + 32)); - xmm3 = load_128_unaligned ((__m128i*)(s + 48)); - - save_128_aligned ((__m128i*)(d), xmm0); - save_128_aligned ((__m128i*)(d + 16), xmm1); - save_128_aligned ((__m128i*)(d + 32), xmm2); - save_128_aligned ((__m128i*)(d + 48), xmm3); - - s += 64; - d += 64; - w -= 64; - } - - while (w >= 16) - { - save_128_aligned ((__m128i*)d, load_128_unaligned ((__m128i*)s) ); - - w -= 16; - d += 16; - s += 16; - } - - while (w >= 4) - { - *(uint32_t *)d = *(uint32_t *)s; - - w -= 4; - s += 4; - d += 4; - } - - if (w >= 2) - { - *(uint16_t *)d = *(uint16_t *)s; - w -= 2; - s += 2; - d += 2; - } - } - - _mm_empty (); - - return TRUE; -} - -static void -sse2_composite_copy_area (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - pixman_blt_sse2 (src_image->bits.bits, - dst_image->bits.bits, - src_image->bits.rowstride, - dst_image->bits.rowstride, - PIXMAN_FORMAT_BPP (src_image->bits.format), - PIXMAN_FORMAT_BPP (dst_image->bits.format), - src_x, src_y, dest_x, dest_y, width, height); -} - -static void -sse2_composite_over_x888_8_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *src, *src_line, s; - uint32_t *dst, *dst_line, d; - uint8_t *mask, *mask_line; - uint32_t m; - int src_stride, mask_stride, dst_stride; - int32_t w; - __m64 ms; - - __m128i xmm_src, xmm_src_lo, xmm_src_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - while (height--) - { - src = src_line; - src_line += src_stride; - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - - w = width; - - while (w && (unsigned long)dst & 15) - { - s = 0xff000000 | *src++; - m = (uint32_t) *mask++; - d = *dst; - ms = unpack_32_1x64 (s); - - if (m != 0xff) - { - __m64 ma = expand_alpha_rev_1x64 (unpack_32_1x64 (m)); - __m64 md = unpack_32_1x64 (d); - - ms = in_over_1x64 (&ms, &mask_x00ff, &ma, &md); - } - - *dst++ = pack_1x64_32 (ms); - w--; - } - - while (w >= 4) - { - m = *(uint32_t*) mask; - xmm_src = _mm_or_si128 (load_128_unaligned ((__m128i*)src), mask_ff000000); - - if (m == 0xffffffff) - { - save_128_aligned ((__m128i*)dst, xmm_src); - } - else - { - xmm_dst = load_128_aligned ((__m128i*)dst); - - xmm_mask = _mm_unpacklo_epi16 (unpack_32_1x128 (m), _mm_setzero_si128()); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - in_over_2x128 (&xmm_src_lo, &xmm_src_hi, &mask_00ff, &mask_00ff, &xmm_mask_lo, &xmm_mask_hi, &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ((__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - - src += 4; - dst += 4; - mask += 4; - w -= 4; - } - - while (w) - { - m = (uint32_t) *mask++; - - if (m) - { - s = 0xff000000 | *src; - - if (m == 0xff) - { - *dst = s; - } - else - { - __m64 ma, md, ms; - - d = *dst; - - ma = expand_alpha_rev_1x64 (unpack_32_1x64 (m)); - md = unpack_32_1x64 (d); - ms = unpack_32_1x64 (s); - - *dst = pack_1x64_32 (in_over_1x64 (&ms, &mask_x00ff, &ma, &md)); - } - - } - - src++; - dst++; - w--; - } - } - - _mm_empty (); -} - -static void -sse2_composite_over_8888_8_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *src, *src_line, s; - uint32_t *dst, *dst_line, d; - uint8_t *mask, *mask_line; - uint32_t m; - int src_stride, mask_stride, dst_stride; - int32_t w; - - __m128i xmm_src, xmm_src_lo, xmm_src_hi, xmm_srca_lo, xmm_srca_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - while (height--) - { - src = src_line; - src_line += src_stride; - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - - w = width; - - while (w && (unsigned long)dst & 15) - { - uint32_t sa; - - s = *src++; - m = (uint32_t) *mask++; - d = *dst; - - sa = s >> 24; - - if (m) - { - if (sa == 0xff && m == 0xff) - { - *dst = s; - } - else - { - __m64 ms, md, ma, msa; - - ma = expand_alpha_rev_1x64 (load_32_1x64 (m)); - ms = unpack_32_1x64 (s); - md = unpack_32_1x64 (d); - - msa = expand_alpha_rev_1x64 (load_32_1x64 (sa)); - - *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md)); - } - } - - dst++; - w--; - } - - while (w >= 4) - { - m = *(uint32_t *) mask; - - if (m) - { - xmm_src = load_128_unaligned ((__m128i*)src); - - if (m == 0xffffffff && is_opaque (xmm_src)) - { - save_128_aligned ((__m128i *)dst, xmm_src); - } - else - { - xmm_dst = load_128_aligned ((__m128i *)dst); - - xmm_mask = _mm_unpacklo_epi16 (unpack_32_1x128 (m), _mm_setzero_si128()); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, &xmm_srca_lo, &xmm_srca_hi); - expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - in_over_2x128 (&xmm_src_lo, &xmm_src_hi, &xmm_srca_lo, &xmm_srca_hi, - &xmm_mask_lo, &xmm_mask_hi, &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ((__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - } - - src += 4; - dst += 4; - mask += 4; - w -= 4; - } - - while (w) - { - uint32_t sa; - - s = *src++; - m = (uint32_t) *mask++; - d = *dst; - - sa = s >> 24; - - if (m) - { - if (sa == 0xff && m == 0xff) - { - *dst = s; - } - else - { - __m64 ms, md, ma, msa; - - ma = expand_alpha_rev_1x64 (load_32_1x64 (m)); - ms = unpack_32_1x64 (s); - md = unpack_32_1x64 (d); - - msa = expand_alpha_rev_1x64 (load_32_1x64 (sa)); - - *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md)); - } - } - - dst++; - w--; - } - } - - _mm_empty (); -} - -static void -sse2_composite_over_reverse_n_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t src; - uint32_t *dst_line, *dst; - __m128i xmm_src; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_dsta_hi, xmm_dsta_lo; - int dst_stride; - int32_t w; - - src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format); - - if (src == 0) - return; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - - xmm_src = expand_pixel_32_1x128 (src); - - while (height--) - { - dst = dst_line; - - dst_line += dst_stride; - w = width; - - while (w && (unsigned long)dst & 15) - { - __m64 vd; - - vd = unpack_32_1x64 (*dst); - - *dst = pack_1x64_32 (over_1x64 (vd, expand_alpha_1x64 (vd), - _mm_movepi64_pi64 (xmm_src))); - w--; - dst++; - } - - while (w >= 4) - { - __m128i tmp_lo, tmp_hi; - - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, &xmm_dsta_lo, &xmm_dsta_hi); - - tmp_lo = xmm_src; - tmp_hi = xmm_src; - - over_2x128 (&xmm_dst_lo, &xmm_dst_hi, - &xmm_dsta_lo, &xmm_dsta_hi, - &tmp_lo, &tmp_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (tmp_lo, tmp_hi)); - - w -= 4; - dst += 4; - } - - while (w) - { - __m64 vd; - - vd = unpack_32_1x64 (*dst); - - *dst = pack_1x64_32 (over_1x64 (vd, expand_alpha_1x64 (vd), - _mm_movepi64_pi64 (xmm_src))); - w--; - dst++; - } - - } - - _mm_empty (); -} - -static void -sse2_composite_over_8888_8888_8888 (pixman_implementation_t *imp, - pixman_op_t op, - pixman_image_t * src_image, - pixman_image_t * mask_image, - pixman_image_t * dst_image, - int32_t src_x, - int32_t src_y, - int32_t mask_x, - int32_t mask_y, - int32_t dest_x, - int32_t dest_y, - int32_t width, - int32_t height) -{ - uint32_t *src, *src_line, s; - uint32_t *dst, *dst_line, d; - uint32_t *mask, *mask_line; - uint32_t m; - int src_stride, mask_stride, dst_stride; - int32_t w; - - __m128i xmm_src, xmm_src_lo, xmm_src_hi, xmm_srca_lo, xmm_srca_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi; - - PIXMAN_IMAGE_GET_LINE ( - dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1); - PIXMAN_IMAGE_GET_LINE ( - mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1); - PIXMAN_IMAGE_GET_LINE ( - src_image, src_x, src_y, uint32_t, src_stride, src_line, 1); - - while (height--) - { - src = src_line; - src_line += src_stride; - dst = dst_line; - dst_line += dst_stride; - mask = mask_line; - mask_line += mask_stride; - - w = width; - - while (w && (unsigned long)dst & 15) - { - uint32_t sa; - - s = *src++; - m = (*mask++) >> 24; - d = *dst; - - sa = s >> 24; - - if (m) - { - if (sa == 0xff && m == 0xff) - { - *dst = s; - } - else - { - __m64 ms, md, ma, msa; - - ma = expand_alpha_rev_1x64 (load_32_1x64 (m)); - ms = unpack_32_1x64 (s); - md = unpack_32_1x64 (d); - - msa = expand_alpha_rev_1x64 (load_32_1x64 (sa)); - - *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md)); - } - } - - dst++; - w--; - } - - while (w >= 4) - { - xmm_mask = load_128_unaligned ((__m128i*)mask); - - if (!is_transparent (xmm_mask)) - { - xmm_src = load_128_unaligned ((__m128i*)src); - - if (is_opaque (xmm_mask) && is_opaque (xmm_src)) - { - save_128_aligned ((__m128i *)dst, xmm_src); - } - else - { - xmm_dst = load_128_aligned ((__m128i *)dst); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, &xmm_srca_lo, &xmm_srca_hi); - expand_alpha_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi); - - in_over_2x128 (&xmm_src_lo, &xmm_src_hi, &xmm_srca_lo, &xmm_srca_hi, - &xmm_mask_lo, &xmm_mask_hi, &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ((__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - } - - src += 4; - dst += 4; - mask += 4; - w -= 4; - } - - while (w) - { - uint32_t sa; - - s = *src++; - m = (*mask++) >> 24; - d = *dst; - - sa = s >> 24; - - if (m) - { - if (sa == 0xff && m == 0xff) - { - *dst = s; - } - else - { - __m64 ms, md, ma, msa; - - ma = expand_alpha_rev_1x64 (load_32_1x64 (m)); - ms = unpack_32_1x64 (s); - md = unpack_32_1x64 (d); - - msa = expand_alpha_rev_1x64 (load_32_1x64 (sa)); - - *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md)); - } - } - - dst++; - w--; - } - } - - _mm_empty (); -} - -/* A variant of 'core_combine_over_u_sse2' with minor tweaks */ -static force_inline void -scaled_nearest_scanline_sse2_8888_8888_OVER (uint32_t* pd, - const uint32_t* ps, - int32_t w, - pixman_fixed_t vx, - pixman_fixed_t unit_x, - pixman_fixed_t max_vx, - pixman_bool_t fully_transparent_src) -{ - uint32_t s, d; - const uint32_t* pm = NULL; - - __m128i xmm_dst_lo, xmm_dst_hi; - __m128i xmm_src_lo, xmm_src_hi; - __m128i xmm_alpha_lo, xmm_alpha_hi; - - if (fully_transparent_src) - return; - - /* Align dst on a 16-byte boundary */ - while (w && ((unsigned long)pd & 15)) - { - d = *pd; - s = combine1 (ps + (vx >> 16), pm); - vx += unit_x; - - *pd++ = core_combine_over_u_pixel_sse2 (s, d); - if (pm) - pm++; - w--; - } - - while (w >= 4) - { - __m128i tmp; - uint32_t tmp1, tmp2, tmp3, tmp4; - - tmp1 = ps[vx >> 16]; - vx += unit_x; - tmp2 = ps[vx >> 16]; - vx += unit_x; - tmp3 = ps[vx >> 16]; - vx += unit_x; - tmp4 = ps[vx >> 16]; - vx += unit_x; - - tmp = _mm_set_epi32 (tmp4, tmp3, tmp2, tmp1); - - xmm_src_hi = combine4 ((__m128i*)&tmp, (__m128i*)pm); - - if (is_opaque (xmm_src_hi)) - { - save_128_aligned ((__m128i*)pd, xmm_src_hi); - } - else if (!is_zero (xmm_src_hi)) - { - xmm_dst_hi = load_128_aligned ((__m128i*) pd); - - unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi); - - expand_alpha_2x128 ( - xmm_src_lo, xmm_src_hi, &xmm_alpha_lo, &xmm_alpha_hi); - - over_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_dst_lo, &xmm_dst_hi); - - /* rebuid the 4 pixel data and save*/ - save_128_aligned ((__m128i*)pd, - pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - - w -= 4; - pd += 4; - if (pm) - pm += 4; - } - - while (w) - { - d = *pd; - s = combine1 (ps + (vx >> 16), pm); - vx += unit_x; - - *pd++ = core_combine_over_u_pixel_sse2 (s, d); - if (pm) - pm++; - - w--; - } - _mm_empty (); -} - -FAST_NEAREST_MAINLOOP (sse2_8888_8888_cover_OVER, - scaled_nearest_scanline_sse2_8888_8888_OVER, - uint32_t, uint32_t, COVER) -FAST_NEAREST_MAINLOOP (sse2_8888_8888_none_OVER, - scaled_nearest_scanline_sse2_8888_8888_OVER, - uint32_t, uint32_t, NONE) -FAST_NEAREST_MAINLOOP (sse2_8888_8888_pad_OVER, - scaled_nearest_scanline_sse2_8888_8888_OVER, - uint32_t, uint32_t, PAD) - -static force_inline void -scaled_nearest_scanline_sse2_8888_n_8888_OVER (const uint32_t * mask, - uint32_t * dst, - const uint32_t * src, - int32_t w, - pixman_fixed_t vx, - pixman_fixed_t unit_x, - pixman_fixed_t max_vx, - pixman_bool_t zero_src) -{ - __m128i xmm_mask; - __m128i xmm_src, xmm_src_lo, xmm_src_hi; - __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi; - __m128i xmm_alpha_lo, xmm_alpha_hi; - - if (zero_src || (*mask >> 24) == 0) - return; - - xmm_mask = create_mask_16_128 (*mask >> 24); - - while (w && (unsigned long)dst & 15) - { - uint32_t s = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - - if (s) - { - uint32_t d = *dst; - - __m64 ms = unpack_32_1x64 (s); - __m64 alpha = expand_alpha_1x64 (ms); - __m64 dest = _mm_movepi64_pi64 (xmm_mask); - __m64 alpha_dst = unpack_32_1x64 (d); - - *dst = pack_1x64_32 ( - in_over_1x64 (&ms, &alpha, &dest, &alpha_dst)); - } - dst++; - w--; - } - - while (w >= 4) - { - uint32_t tmp1, tmp2, tmp3, tmp4; - - tmp1 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - tmp2 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - tmp3 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - tmp4 = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - - xmm_src = _mm_set_epi32 (tmp4, tmp3, tmp2, tmp1); - - if (!is_zero (xmm_src)) - { - xmm_dst = load_128_aligned ((__m128i*)dst); - - unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi); - unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi); - expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi); - - in_over_2x128 (&xmm_src_lo, &xmm_src_hi, - &xmm_alpha_lo, &xmm_alpha_hi, - &xmm_mask, &xmm_mask, - &xmm_dst_lo, &xmm_dst_hi); - - save_128_aligned ( - (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi)); - } - - dst += 4; - w -= 4; - } - - while (w) - { - uint32_t s = src[pixman_fixed_to_int (vx)]; - vx += unit_x; - - if (s) - { - uint32_t d = *dst; - - __m64 ms = unpack_32_1x64 (s); - __m64 alpha = expand_alpha_1x64 (ms); - __m64 mask = _mm_movepi64_pi64 (xmm_mask); - __m64 dest = unpack_32_1x64 (d); - - *dst = pack_1x64_32 ( - in_over_1x64 (&ms, &alpha, &mask, &dest)); - } - - dst++; - w--; - } - - _mm_empty (); -} - -FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_cover_OVER, - scaled_nearest_scanline_sse2_8888_n_8888_OVER, - uint32_t, uint32_t, uint32_t, COVER, TRUE, TRUE) -FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_pad_OVER, - scaled_nearest_scanline_sse2_8888_n_8888_OVER, - uint32_t, uint32_t, uint32_t, PAD, TRUE, TRUE) -FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_none_OVER, - scaled_nearest_scanline_sse2_8888_n_8888_OVER, - uint32_t, uint32_t, uint32_t, NONE, TRUE, TRUE) - -static const pixman_fast_path_t sse2_fast_paths[] = -{ - /* PIXMAN_OP_OVER */ - PIXMAN_STD_FAST_PATH (OVER, solid, a8, r5g6b5, sse2_composite_over_n_8_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, b5g6r5, sse2_composite_over_n_8_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, null, a8r8g8b8, sse2_composite_over_n_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, null, x8r8g8b8, sse2_composite_over_n_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, null, r5g6b5, sse2_composite_over_n_0565), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, a8r8g8b8, sse2_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, x8r8g8b8, sse2_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, a8b8g8r8, sse2_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, x8b8g8r8, sse2_composite_over_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, r5g6b5, sse2_composite_over_8888_0565), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, b5g6r5, sse2_composite_over_8888_0565), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8r8g8b8, sse2_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8r8g8b8, sse2_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8b8g8r8, sse2_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8b8g8r8, sse2_composite_over_n_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, a8r8g8b8, sse2_composite_over_8888_8888_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, x8r8g8b8, sse2_composite_over_8888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, a8r8g8b8, sse2_composite_over_8888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, x8b8g8r8, sse2_composite_over_8888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, a8b8g8r8, sse2_composite_over_8888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, a8, x8r8g8b8, sse2_composite_over_x888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, a8, a8r8g8b8, sse2_composite_over_x888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, a8, x8b8g8r8, sse2_composite_over_x888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, a8, a8b8g8r8, sse2_composite_over_x888_8_8888), - PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, solid, a8r8g8b8, sse2_composite_over_x888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, solid, x8r8g8b8, sse2_composite_over_x888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, solid, a8b8g8r8, sse2_composite_over_x888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, solid, x8b8g8r8, sse2_composite_over_x888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, a8r8g8b8, sse2_composite_over_8888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, x8r8g8b8, sse2_composite_over_8888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, solid, a8b8g8r8, sse2_composite_over_8888_n_8888), - PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, solid, x8b8g8r8, sse2_composite_over_8888_n_8888), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, a8r8g8b8, sse2_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, x8r8g8b8, sse2_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, a8b8g8r8, sse2_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, x8b8g8r8, sse2_composite_over_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, r5g6b5, sse2_composite_over_n_8888_0565_ca), - PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, b5g6r5, sse2_composite_over_n_8888_0565_ca), - PIXMAN_STD_FAST_PATH (OVER, pixbuf, pixbuf, a8r8g8b8, sse2_composite_over_pixbuf_8888), - PIXMAN_STD_FAST_PATH (OVER, pixbuf, pixbuf, x8r8g8b8, sse2_composite_over_pixbuf_8888), - PIXMAN_STD_FAST_PATH (OVER, rpixbuf, rpixbuf, a8b8g8r8, sse2_composite_over_pixbuf_8888), - PIXMAN_STD_FAST_PATH (OVER, rpixbuf, rpixbuf, x8b8g8r8, sse2_composite_over_pixbuf_8888), - PIXMAN_STD_FAST_PATH (OVER, pixbuf, pixbuf, r5g6b5, sse2_composite_over_pixbuf_0565), - PIXMAN_STD_FAST_PATH (OVER, rpixbuf, rpixbuf, b5g6r5, sse2_composite_over_pixbuf_0565), - PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, null, x8r8g8b8, sse2_composite_copy_area), - PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, null, x8b8g8r8, sse2_composite_copy_area), - - /* PIXMAN_OP_OVER_REVERSE */ - PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8r8g8b8, sse2_composite_over_reverse_n_8888), - PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8b8g8r8, sse2_composite_over_reverse_n_8888), - - /* PIXMAN_OP_ADD */ - PIXMAN_STD_FAST_PATH_CA (ADD, solid, a8r8g8b8, a8r8g8b8, sse2_composite_add_n_8888_8888_ca), - PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, sse2_composite_add_8_8), - PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, sse2_composite_add_8888_8888), - PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, null, a8b8g8r8, sse2_composite_add_8888_8888), - PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8, sse2_composite_add_n_8_8), - PIXMAN_STD_FAST_PATH (ADD, solid, null, a8, sse2_composite_add_n_8), - - /* PIXMAN_OP_SRC */ - PIXMAN_STD_FAST_PATH (SRC, solid, a8, a8r8g8b8, sse2_composite_src_n_8_8888), - PIXMAN_STD_FAST_PATH (SRC, solid, a8, x8r8g8b8, sse2_composite_src_n_8_8888), - PIXMAN_STD_FAST_PATH (SRC, solid, a8, a8b8g8r8, sse2_composite_src_n_8_8888), - PIXMAN_STD_FAST_PATH (SRC, solid, a8, x8b8g8r8, sse2_composite_src_n_8_8888), - PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, a8r8g8b8, sse2_composite_src_x888_8888), - PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, a8b8g8r8, sse2_composite_src_x888_8888), - PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, a8r8g8b8, sse2_composite_copy_area), - PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, a8b8g8r8, sse2_composite_copy_area), - PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, x8r8g8b8, sse2_composite_copy_area), - PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, x8b8g8r8, sse2_composite_copy_area), - PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, x8r8g8b8, sse2_composite_copy_area), - PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, x8b8g8r8, sse2_composite_copy_area), - PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, r5g6b5, sse2_composite_copy_area), - PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, b5g6r5, sse2_composite_copy_area), - - /* PIXMAN_OP_IN */ - PIXMAN_STD_FAST_PATH (IN, a8, null, a8, sse2_composite_in_8_8), - PIXMAN_STD_FAST_PATH (IN, solid, a8, a8, sse2_composite_in_n_8_8), - PIXMAN_STD_FAST_PATH (IN, solid, null, a8, sse2_composite_in_n_8), - - SIMPLE_NEAREST_FAST_PATH_COVER (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_COVER (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_COVER (OVER, a8r8g8b8, a8r8g8b8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_COVER (OVER, a8b8g8r8, a8b8g8r8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_NONE (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_NONE (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_NONE (OVER, a8r8g8b8, a8r8g8b8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_NONE (OVER, a8b8g8r8, a8b8g8r8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_PAD (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_PAD (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_PAD (OVER, a8r8g8b8, a8r8g8b8, sse2_8888_8888), - SIMPLE_NEAREST_FAST_PATH_PAD (OVER, a8b8g8r8, a8b8g8r8, sse2_8888_8888), - - SIMPLE_NEAREST_SOLID_MASK_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, sse2_8888_n_8888), - SIMPLE_NEAREST_SOLID_MASK_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8, sse2_8888_n_8888), - SIMPLE_NEAREST_SOLID_MASK_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_n_8888), - SIMPLE_NEAREST_SOLID_MASK_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_n_8888), - - { PIXMAN_OP_NONE }, -}; - -static pixman_bool_t -sse2_blt (pixman_implementation_t *imp, - uint32_t * src_bits, - uint32_t * dst_bits, - int src_stride, - int dst_stride, - int src_bpp, - int dst_bpp, - int src_x, - int src_y, - int dst_x, - int dst_y, - int width, - int height) -{ - if (!pixman_blt_sse2 ( - src_bits, dst_bits, src_stride, dst_stride, src_bpp, dst_bpp, - src_x, src_y, dst_x, dst_y, width, height)) - - { - return _pixman_implementation_blt ( - imp->delegate, - src_bits, dst_bits, src_stride, dst_stride, src_bpp, dst_bpp, - src_x, src_y, dst_x, dst_y, width, height); - } - - return TRUE; -} - -#if defined(__GNUC__) && !defined(__x86_64__) && !defined(__amd64__) -__attribute__((__force_align_arg_pointer__)) -#endif -static pixman_bool_t -sse2_fill (pixman_implementation_t *imp, - uint32_t * bits, - int stride, - int bpp, - int x, - int y, - int width, - int height, - uint32_t xor) -{ - if (!pixman_fill_sse2 (bits, stride, bpp, x, y, width, height, xor)) - { - return _pixman_implementation_fill ( - imp->delegate, bits, stride, bpp, x, y, width, height, xor); - } - - return TRUE; -} - -static uint32_t * -sse2_fetch_x8r8g8b8 (pixman_iter_t *iter, const uint32_t *mask) -{ - int w = iter->width; - __m128i ff000000 = mask_ff000000; - uint32_t *dst = iter->buffer; - uint32_t *src = (uint32_t *)iter->bits; - - iter->bits += iter->stride; - - while (w && ((unsigned long)dst) & 0x0f) - { - *dst++ = (*src++) | 0xff000000; - w--; - } - - while (w >= 4) - { - save_128_aligned ( - (__m128i *)dst, _mm_or_si128 ( - load_128_unaligned ((__m128i *)src), ff000000)); - - dst += 4; - src += 4; - w -= 4; - } - - while (w) - { - *dst++ = (*src++) | 0xff000000; - w--; - } - - return iter->buffer; -} - -static uint32_t * -sse2_fetch_r5g6b5 (pixman_iter_t *iter, const uint32_t *mask) -{ - int w = iter->width; - uint32_t *dst = iter->buffer; - uint16_t *src = (uint16_t *)iter->bits; - __m128i ff000000 = mask_ff000000; - - iter->bits += iter->stride; - - while (w && ((unsigned long)dst) & 0x0f) - { - uint16_t s = *src++; - - *dst++ = CONVERT_0565_TO_8888 (s); - w--; - } - - while (w >= 8) - { - __m128i lo, hi, s; - - s = _mm_loadu_si128 ((__m128i *)src); - - lo = unpack_565_to_8888 (_mm_unpacklo_epi16 (s, _mm_setzero_si128 ())); - hi = unpack_565_to_8888 (_mm_unpackhi_epi16 (s, _mm_setzero_si128 ())); - - save_128_aligned ((__m128i *)(dst + 0), _mm_or_si128 (lo, ff000000)); - save_128_aligned ((__m128i *)(dst + 4), _mm_or_si128 (hi, ff000000)); - - dst += 8; - src += 8; - w -= 8; - } - - while (w) - { - uint16_t s = *src++; - - *dst++ = CONVERT_0565_TO_8888 (s); - w--; - } - - return iter->buffer; -} - -static uint32_t * -sse2_fetch_a8 (pixman_iter_t *iter, const uint32_t *mask) -{ - int w = iter->width; - uint32_t *dst = iter->buffer; - uint8_t *src = iter->bits; - __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6; - - iter->bits += iter->stride; - - while (w && (((unsigned long)dst) & 15)) - { - *dst++ = *(src++) << 24; - w--; - } - - while (w >= 16) - { - xmm0 = _mm_loadu_si128((__m128i *)src); - - xmm1 = _mm_unpacklo_epi8 (_mm_setzero_si128(), xmm0); - xmm2 = _mm_unpackhi_epi8 (_mm_setzero_si128(), xmm0); - xmm3 = _mm_unpacklo_epi16 (_mm_setzero_si128(), xmm1); - xmm4 = _mm_unpackhi_epi16 (_mm_setzero_si128(), xmm1); - xmm5 = _mm_unpacklo_epi16 (_mm_setzero_si128(), xmm2); - xmm6 = _mm_unpackhi_epi16 (_mm_setzero_si128(), xmm2); - - _mm_store_si128(((__m128i *)(dst + 0)), xmm3); - _mm_store_si128(((__m128i *)(dst + 4)), xmm4); - _mm_store_si128(((__m128i *)(dst + 8)), xmm5); - _mm_store_si128(((__m128i *)(dst + 12)), xmm6); - - dst += 16; - src += 16; - w -= 16; - } - - while (w) - { - *dst++ = *(src++) << 24; - w--; - } - - return iter->buffer; -} - -typedef struct -{ - pixman_format_code_t format; - pixman_iter_get_scanline_t get_scanline; -} fetcher_info_t; - -static const fetcher_info_t fetchers[] = -{ - { PIXMAN_x8r8g8b8, sse2_fetch_x8r8g8b8 }, - { PIXMAN_r5g6b5, sse2_fetch_r5g6b5 }, - { PIXMAN_a8, sse2_fetch_a8 }, - { PIXMAN_null } -}; - -static void -sse2_src_iter_init (pixman_implementation_t *imp, - pixman_iter_t *iter, - pixman_image_t *image, - int x, int y, int width, int height, - uint8_t *buffer, iter_flags_t flags) -{ -#define FLAGS \ - (FAST_PATH_STANDARD_FLAGS | FAST_PATH_ID_TRANSFORM) - - if ((flags & ITER_NARROW) && - (image->common.flags & FLAGS) == FLAGS && - x >= 0 && y >= 0 && - x + width <= image->bits.width && - y + height <= image->bits.height) - { - const fetcher_info_t *f; - - for (f = &fetchers[0]; f->format != PIXMAN_null; f++) - { - if (image->common.extended_format_code == f->format) - { - uint8_t *b = (uint8_t *)image->bits.bits; - int s = image->bits.rowstride * 4; - - iter->bits = b + s * y + x * PIXMAN_FORMAT_BPP (f->format) / 8; - iter->stride = s; - iter->width = width; - iter->buffer = (uint32_t *)buffer; - - iter->get_scanline = f->get_scanline; - return; - } - } - } - - _pixman_implementation_src_iter_init ( - imp->delegate, iter, image, x, y, width, height, buffer, flags); -} - -#if defined(__GNUC__) && !defined(__x86_64__) && !defined(__amd64__) -__attribute__((__force_align_arg_pointer__)) -#endif -pixman_implementation_t * -_pixman_implementation_create_sse2 (pixman_implementation_t *fallback) -{ - pixman_implementation_t *imp = _pixman_implementation_create (fallback, sse2_fast_paths); - - /* SSE2 constants */ - mask_565_r = create_mask_2x32_128 (0x00f80000, 0x00f80000); - mask_565_g1 = create_mask_2x32_128 (0x00070000, 0x00070000); - mask_565_g2 = create_mask_2x32_128 (0x000000e0, 0x000000e0); - mask_565_b = create_mask_2x32_128 (0x0000001f, 0x0000001f); - mask_red = create_mask_2x32_128 (0x00f80000, 0x00f80000); - mask_green = create_mask_2x32_128 (0x0000fc00, 0x0000fc00); - mask_blue = create_mask_2x32_128 (0x000000f8, 0x000000f8); - mask_565_fix_rb = create_mask_2x32_128 (0x00e000e0, 0x00e000e0); - mask_565_fix_g = create_mask_2x32_128 (0x0000c000, 0x0000c000); - mask_0080 = create_mask_16_128 (0x0080); - mask_00ff = create_mask_16_128 (0x00ff); - mask_0101 = create_mask_16_128 (0x0101); - mask_ffff = create_mask_16_128 (0xffff); - mask_ff000000 = create_mask_2x32_128 (0xff000000, 0xff000000); - mask_alpha = create_mask_2x32_128 (0x00ff0000, 0x00000000); - - /* MMX constants */ - mask_x565_rgb = create_mask_2x32_64 (0x000001f0, 0x003f001f); - mask_x565_unpack = create_mask_2x32_64 (0x00000084, 0x04100840); - - mask_x0080 = create_mask_16_64 (0x0080); - mask_x00ff = create_mask_16_64 (0x00ff); - mask_x0101 = create_mask_16_64 (0x0101); - mask_x_alpha = create_mask_2x32_64 (0x00ff0000, 0x00000000); - - _mm_empty (); - - /* Set up function pointers */ - - /* SSE code patch for fbcompose.c */ - imp->combine_32[PIXMAN_OP_OVER] = sse2_combine_over_u; - imp->combine_32[PIXMAN_OP_OVER_REVERSE] = sse2_combine_over_reverse_u; - imp->combine_32[PIXMAN_OP_IN] = sse2_combine_in_u; - imp->combine_32[PIXMAN_OP_IN_REVERSE] = sse2_combine_in_reverse_u; - imp->combine_32[PIXMAN_OP_OUT] = sse2_combine_out_u; - imp->combine_32[PIXMAN_OP_OUT_REVERSE] = sse2_combine_out_reverse_u; - imp->combine_32[PIXMAN_OP_ATOP] = sse2_combine_atop_u; - imp->combine_32[PIXMAN_OP_ATOP_REVERSE] = sse2_combine_atop_reverse_u; - imp->combine_32[PIXMAN_OP_XOR] = sse2_combine_xor_u; - imp->combine_32[PIXMAN_OP_ADD] = sse2_combine_add_u; - - imp->combine_32[PIXMAN_OP_SATURATE] = sse2_combine_saturate_u; - - imp->combine_32_ca[PIXMAN_OP_SRC] = sse2_combine_src_ca; - imp->combine_32_ca[PIXMAN_OP_OVER] = sse2_combine_over_ca; - imp->combine_32_ca[PIXMAN_OP_OVER_REVERSE] = sse2_combine_over_reverse_ca; - imp->combine_32_ca[PIXMAN_OP_IN] = sse2_combine_in_ca; - imp->combine_32_ca[PIXMAN_OP_IN_REVERSE] = sse2_combine_in_reverse_ca; - imp->combine_32_ca[PIXMAN_OP_OUT] = sse2_combine_out_ca; - imp->combine_32_ca[PIXMAN_OP_OUT_REVERSE] = sse2_combine_out_reverse_ca; - imp->combine_32_ca[PIXMAN_OP_ATOP] = sse2_combine_atop_ca; - imp->combine_32_ca[PIXMAN_OP_ATOP_REVERSE] = sse2_combine_atop_reverse_ca; - imp->combine_32_ca[PIXMAN_OP_XOR] = sse2_combine_xor_ca; - imp->combine_32_ca[PIXMAN_OP_ADD] = sse2_combine_add_ca; - - imp->blt = sse2_blt; - imp->fill = sse2_fill; - - imp->src_iter_init = sse2_src_iter_init; - - return imp; -} - -#endif /* USE_SSE2 */ +/*
+ * Copyright © 2008 Rodrigo Kumpera
+ * Copyright © 2008 André Tupinambá
+ *
+ * Permission to use, copy, modify, distribute, and sell this software and its
+ * documentation for any purpose is hereby granted without fee, provided that
+ * the above copyright notice appear in all copies and that both that
+ * copyright notice and this permission notice appear in supporting
+ * documentation, and that the name of Red Hat not be used in advertising or
+ * publicity pertaining to distribution of the software without specific,
+ * written prior permission. Red Hat makes no representations about the
+ * suitability of this software for any purpose. It is provided "as is"
+ * without express or implied warranty.
+ *
+ * THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS
+ * SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND
+ * FITNESS, IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY
+ * SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN
+ * AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING
+ * OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
+ * SOFTWARE.
+ *
+ * Author: Rodrigo Kumpera (kumpera@gmail.com)
+ * André Tupinambá (andrelrt@gmail.com)
+ *
+ * Based on work by Owen Taylor and Søren Sandmann
+ */
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <mmintrin.h>
+#include <xmmintrin.h> /* for _mm_shuffle_pi16 and _MM_SHUFFLE */
+#include <emmintrin.h> /* for SSE2 intrinsics */
+#include "pixman-private.h"
+#include "pixman-combine32.h"
+#include "pixman-fast-path.h"
+
+#if defined(_MSC_VER) && defined(_M_AMD64)
+/* Windows 64 doesn't allow MMX to be used, so
+ * the pixman-x64-mmx-emulation.h file contains
+ * implementations of those MMX intrinsics that
+ * are used in the SSE2 implementation.
+ */
+# include "pixman-x64-mmx-emulation.h"
+#endif
+
+#ifdef USE_SSE2
+
+/* --------------------------------------------------------------------
+ * Locals
+ */
+
+static __m64 mask_x0080;
+static __m64 mask_x00ff;
+static __m64 mask_x0101;
+static __m64 mask_x_alpha;
+
+static __m64 mask_x565_rgb;
+static __m64 mask_x565_unpack;
+
+static __m128i mask_0080;
+static __m128i mask_00ff;
+static __m128i mask_0101;
+static __m128i mask_ffff;
+static __m128i mask_ff000000;
+static __m128i mask_alpha;
+
+static __m128i mask_565_r;
+static __m128i mask_565_g1, mask_565_g2;
+static __m128i mask_565_b;
+static __m128i mask_red;
+static __m128i mask_green;
+static __m128i mask_blue;
+
+static __m128i mask_565_fix_rb;
+static __m128i mask_565_fix_g;
+
+/* ----------------------------------------------------------------------
+ * SSE2 Inlines
+ */
+static force_inline __m128i
+unpack_32_1x128 (uint32_t data)
+{
+ return _mm_unpacklo_epi8 (_mm_cvtsi32_si128 (data), _mm_setzero_si128 ());
+}
+
+static force_inline void
+unpack_128_2x128 (__m128i data, __m128i* data_lo, __m128i* data_hi)
+{
+ *data_lo = _mm_unpacklo_epi8 (data, _mm_setzero_si128 ());
+ *data_hi = _mm_unpackhi_epi8 (data, _mm_setzero_si128 ());
+}
+
+static force_inline __m128i
+unpack_565_to_8888 (__m128i lo)
+{
+ __m128i r, g, b, rb, t;
+
+ r = _mm_and_si128 (_mm_slli_epi32 (lo, 8), mask_red);
+ g = _mm_and_si128 (_mm_slli_epi32 (lo, 5), mask_green);
+ b = _mm_and_si128 (_mm_slli_epi32 (lo, 3), mask_blue);
+
+ rb = _mm_or_si128 (r, b);
+ t = _mm_and_si128 (rb, mask_565_fix_rb);
+ t = _mm_srli_epi32 (t, 5);
+ rb = _mm_or_si128 (rb, t);
+
+ t = _mm_and_si128 (g, mask_565_fix_g);
+ t = _mm_srli_epi32 (t, 6);
+ g = _mm_or_si128 (g, t);
+
+ return _mm_or_si128 (rb, g);
+}
+
+static force_inline void
+unpack_565_128_4x128 (__m128i data,
+ __m128i* data0,
+ __m128i* data1,
+ __m128i* data2,
+ __m128i* data3)
+{
+ __m128i lo, hi;
+
+ lo = _mm_unpacklo_epi16 (data, _mm_setzero_si128 ());
+ hi = _mm_unpackhi_epi16 (data, _mm_setzero_si128 ());
+
+ lo = unpack_565_to_8888 (lo);
+ hi = unpack_565_to_8888 (hi);
+
+ unpack_128_2x128 (lo, data0, data1);
+ unpack_128_2x128 (hi, data2, data3);
+}
+
+static force_inline uint16_t
+pack_565_32_16 (uint32_t pixel)
+{
+ return (uint16_t) (((pixel >> 8) & 0xf800) |
+ ((pixel >> 5) & 0x07e0) |
+ ((pixel >> 3) & 0x001f));
+}
+
+static force_inline __m128i
+pack_2x128_128 (__m128i lo, __m128i hi)
+{
+ return _mm_packus_epi16 (lo, hi);
+}
+
+static force_inline __m128i
+pack_565_2x128_128 (__m128i lo, __m128i hi)
+{
+ __m128i data;
+ __m128i r, g1, g2, b;
+
+ data = pack_2x128_128 (lo, hi);
+
+ r = _mm_and_si128 (data, mask_565_r);
+ g1 = _mm_and_si128 (_mm_slli_epi32 (data, 3), mask_565_g1);
+ g2 = _mm_and_si128 (_mm_srli_epi32 (data, 5), mask_565_g2);
+ b = _mm_and_si128 (_mm_srli_epi32 (data, 3), mask_565_b);
+
+ return _mm_or_si128 (_mm_or_si128 (_mm_or_si128 (r, g1), g2), b);
+}
+
+static force_inline __m128i
+pack_565_4x128_128 (__m128i* xmm0, __m128i* xmm1, __m128i* xmm2, __m128i* xmm3)
+{
+ return _mm_packus_epi16 (pack_565_2x128_128 (*xmm0, *xmm1),
+ pack_565_2x128_128 (*xmm2, *xmm3));
+}
+
+static force_inline int
+is_opaque (__m128i x)
+{
+ __m128i ffs = _mm_cmpeq_epi8 (x, x);
+
+ return (_mm_movemask_epi8 (_mm_cmpeq_epi8 (x, ffs)) & 0x8888) == 0x8888;
+}
+
+static force_inline int
+is_zero (__m128i x)
+{
+ return _mm_movemask_epi8 (
+ _mm_cmpeq_epi8 (x, _mm_setzero_si128 ())) == 0xffff;
+}
+
+static force_inline int
+is_transparent (__m128i x)
+{
+ return (_mm_movemask_epi8 (
+ _mm_cmpeq_epi8 (x, _mm_setzero_si128 ())) & 0x8888) == 0x8888;
+}
+
+static force_inline __m128i
+expand_pixel_32_1x128 (uint32_t data)
+{
+ return _mm_shuffle_epi32 (unpack_32_1x128 (data), _MM_SHUFFLE (1, 0, 1, 0));
+}
+
+static force_inline __m128i
+expand_alpha_1x128 (__m128i data)
+{
+ return _mm_shufflehi_epi16 (_mm_shufflelo_epi16 (data,
+ _MM_SHUFFLE (3, 3, 3, 3)),
+ _MM_SHUFFLE (3, 3, 3, 3));
+}
+
+static force_inline void
+expand_alpha_2x128 (__m128i data_lo,
+ __m128i data_hi,
+ __m128i* alpha_lo,
+ __m128i* alpha_hi)
+{
+ __m128i lo, hi;
+
+ lo = _mm_shufflelo_epi16 (data_lo, _MM_SHUFFLE (3, 3, 3, 3));
+ hi = _mm_shufflelo_epi16 (data_hi, _MM_SHUFFLE (3, 3, 3, 3));
+
+ *alpha_lo = _mm_shufflehi_epi16 (lo, _MM_SHUFFLE (3, 3, 3, 3));
+ *alpha_hi = _mm_shufflehi_epi16 (hi, _MM_SHUFFLE (3, 3, 3, 3));
+}
+
+static force_inline void
+expand_alpha_rev_2x128 (__m128i data_lo,
+ __m128i data_hi,
+ __m128i* alpha_lo,
+ __m128i* alpha_hi)
+{
+ __m128i lo, hi;
+
+ lo = _mm_shufflelo_epi16 (data_lo, _MM_SHUFFLE (0, 0, 0, 0));
+ hi = _mm_shufflelo_epi16 (data_hi, _MM_SHUFFLE (0, 0, 0, 0));
+ *alpha_lo = _mm_shufflehi_epi16 (lo, _MM_SHUFFLE (0, 0, 0, 0));
+ *alpha_hi = _mm_shufflehi_epi16 (hi, _MM_SHUFFLE (0, 0, 0, 0));
+}
+
+static force_inline void
+pix_multiply_2x128 (__m128i* data_lo,
+ __m128i* data_hi,
+ __m128i* alpha_lo,
+ __m128i* alpha_hi,
+ __m128i* ret_lo,
+ __m128i* ret_hi)
+{
+ __m128i lo, hi;
+
+ lo = _mm_mullo_epi16 (*data_lo, *alpha_lo);
+ hi = _mm_mullo_epi16 (*data_hi, *alpha_hi);
+ lo = _mm_adds_epu16 (lo, mask_0080);
+ hi = _mm_adds_epu16 (hi, mask_0080);
+ *ret_lo = _mm_mulhi_epu16 (lo, mask_0101);
+ *ret_hi = _mm_mulhi_epu16 (hi, mask_0101);
+}
+
+static force_inline void
+pix_add_multiply_2x128 (__m128i* src_lo,
+ __m128i* src_hi,
+ __m128i* alpha_dst_lo,
+ __m128i* alpha_dst_hi,
+ __m128i* dst_lo,
+ __m128i* dst_hi,
+ __m128i* alpha_src_lo,
+ __m128i* alpha_src_hi,
+ __m128i* ret_lo,
+ __m128i* ret_hi)
+{
+ __m128i t1_lo, t1_hi;
+ __m128i t2_lo, t2_hi;
+
+ pix_multiply_2x128 (src_lo, src_hi, alpha_dst_lo, alpha_dst_hi, &t1_lo, &t1_hi);
+ pix_multiply_2x128 (dst_lo, dst_hi, alpha_src_lo, alpha_src_hi, &t2_lo, &t2_hi);
+
+ *ret_lo = _mm_adds_epu8 (t1_lo, t2_lo);
+ *ret_hi = _mm_adds_epu8 (t1_hi, t2_hi);
+}
+
+static force_inline void
+negate_2x128 (__m128i data_lo,
+ __m128i data_hi,
+ __m128i* neg_lo,
+ __m128i* neg_hi)
+{
+ *neg_lo = _mm_xor_si128 (data_lo, mask_00ff);
+ *neg_hi = _mm_xor_si128 (data_hi, mask_00ff);
+}
+
+static force_inline void
+invert_colors_2x128 (__m128i data_lo,
+ __m128i data_hi,
+ __m128i* inv_lo,
+ __m128i* inv_hi)
+{
+ __m128i lo, hi;
+
+ lo = _mm_shufflelo_epi16 (data_lo, _MM_SHUFFLE (3, 0, 1, 2));
+ hi = _mm_shufflelo_epi16 (data_hi, _MM_SHUFFLE (3, 0, 1, 2));
+ *inv_lo = _mm_shufflehi_epi16 (lo, _MM_SHUFFLE (3, 0, 1, 2));
+ *inv_hi = _mm_shufflehi_epi16 (hi, _MM_SHUFFLE (3, 0, 1, 2));
+}
+
+static force_inline void
+over_2x128 (__m128i* src_lo,
+ __m128i* src_hi,
+ __m128i* alpha_lo,
+ __m128i* alpha_hi,
+ __m128i* dst_lo,
+ __m128i* dst_hi)
+{
+ __m128i t1, t2;
+
+ negate_2x128 (*alpha_lo, *alpha_hi, &t1, &t2);
+
+ pix_multiply_2x128 (dst_lo, dst_hi, &t1, &t2, dst_lo, dst_hi);
+
+ *dst_lo = _mm_adds_epu8 (*src_lo, *dst_lo);
+ *dst_hi = _mm_adds_epu8 (*src_hi, *dst_hi);
+}
+
+static force_inline void
+over_rev_non_pre_2x128 (__m128i src_lo,
+ __m128i src_hi,
+ __m128i* dst_lo,
+ __m128i* dst_hi)
+{
+ __m128i lo, hi;
+ __m128i alpha_lo, alpha_hi;
+
+ expand_alpha_2x128 (src_lo, src_hi, &alpha_lo, &alpha_hi);
+
+ lo = _mm_or_si128 (alpha_lo, mask_alpha);
+ hi = _mm_or_si128 (alpha_hi, mask_alpha);
+
+ invert_colors_2x128 (src_lo, src_hi, &src_lo, &src_hi);
+
+ pix_multiply_2x128 (&src_lo, &src_hi, &lo, &hi, &lo, &hi);
+
+ over_2x128 (&lo, &hi, &alpha_lo, &alpha_hi, dst_lo, dst_hi);
+}
+
+static force_inline void
+in_over_2x128 (__m128i* src_lo,
+ __m128i* src_hi,
+ __m128i* alpha_lo,
+ __m128i* alpha_hi,
+ __m128i* mask_lo,
+ __m128i* mask_hi,
+ __m128i* dst_lo,
+ __m128i* dst_hi)
+{
+ __m128i s_lo, s_hi;
+ __m128i a_lo, a_hi;
+
+ pix_multiply_2x128 (src_lo, src_hi, mask_lo, mask_hi, &s_lo, &s_hi);
+ pix_multiply_2x128 (alpha_lo, alpha_hi, mask_lo, mask_hi, &a_lo, &a_hi);
+
+ over_2x128 (&s_lo, &s_hi, &a_lo, &a_hi, dst_lo, dst_hi);
+}
+
+/* load 4 pixels from a 16-byte boundary aligned address */
+static force_inline __m128i
+load_128_aligned (__m128i* src)
+{
+ return _mm_load_si128 (src);
+}
+
+/* load 4 pixels from a unaligned address */
+static force_inline __m128i
+load_128_unaligned (const __m128i* src)
+{
+ return _mm_loadu_si128 (src);
+}
+
+/* save 4 pixels using Write Combining memory on a 16-byte
+ * boundary aligned address
+ */
+static force_inline void
+save_128_write_combining (__m128i* dst,
+ __m128i data)
+{
+ _mm_stream_si128 (dst, data);
+}
+
+/* save 4 pixels on a 16-byte boundary aligned address */
+static force_inline void
+save_128_aligned (__m128i* dst,
+ __m128i data)
+{
+ _mm_store_si128 (dst, data);
+}
+
+/* save 4 pixels on a unaligned address */
+static force_inline void
+save_128_unaligned (__m128i* dst,
+ __m128i data)
+{
+ _mm_storeu_si128 (dst, data);
+}
+
+/* ------------------------------------------------------------------
+ * MMX inlines
+ */
+
+static force_inline __m64
+load_32_1x64 (uint32_t data)
+{
+ return _mm_cvtsi32_si64 (data);
+}
+
+static force_inline __m64
+unpack_32_1x64 (uint32_t data)
+{
+ return _mm_unpacklo_pi8 (load_32_1x64 (data), _mm_setzero_si64 ());
+}
+
+static force_inline __m64
+expand_alpha_1x64 (__m64 data)
+{
+ return _mm_shuffle_pi16 (data, _MM_SHUFFLE (3, 3, 3, 3));
+}
+
+static force_inline __m64
+expand_alpha_rev_1x64 (__m64 data)
+{
+ return _mm_shuffle_pi16 (data, _MM_SHUFFLE (0, 0, 0, 0));
+}
+
+static force_inline __m64
+expand_pixel_8_1x64 (uint8_t data)
+{
+ return _mm_shuffle_pi16 (
+ unpack_32_1x64 ((uint32_t)data), _MM_SHUFFLE (0, 0, 0, 0));
+}
+
+static force_inline __m64
+pix_multiply_1x64 (__m64 data,
+ __m64 alpha)
+{
+ return _mm_mulhi_pu16 (_mm_adds_pu16 (_mm_mullo_pi16 (data, alpha),
+ mask_x0080),
+ mask_x0101);
+}
+
+static force_inline __m64
+pix_add_multiply_1x64 (__m64* src,
+ __m64* alpha_dst,
+ __m64* dst,
+ __m64* alpha_src)
+{
+ __m64 t1 = pix_multiply_1x64 (*src, *alpha_dst);
+ __m64 t2 = pix_multiply_1x64 (*dst, *alpha_src);
+
+ return _mm_adds_pu8 (t1, t2);
+}
+
+static force_inline __m64
+negate_1x64 (__m64 data)
+{
+ return _mm_xor_si64 (data, mask_x00ff);
+}
+
+static force_inline __m64
+invert_colors_1x64 (__m64 data)
+{
+ return _mm_shuffle_pi16 (data, _MM_SHUFFLE (3, 0, 1, 2));
+}
+
+static force_inline __m64
+over_1x64 (__m64 src, __m64 alpha, __m64 dst)
+{
+ return _mm_adds_pu8 (src, pix_multiply_1x64 (dst, negate_1x64 (alpha)));
+}
+
+static force_inline __m64
+in_over_1x64 (__m64* src, __m64* alpha, __m64* mask, __m64* dst)
+{
+ return over_1x64 (pix_multiply_1x64 (*src, *mask),
+ pix_multiply_1x64 (*alpha, *mask),
+ *dst);
+}
+
+static force_inline __m64
+over_rev_non_pre_1x64 (__m64 src, __m64 dst)
+{
+ __m64 alpha = expand_alpha_1x64 (src);
+
+ return over_1x64 (pix_multiply_1x64 (invert_colors_1x64 (src),
+ _mm_or_si64 (alpha, mask_x_alpha)),
+ alpha,
+ dst);
+}
+
+static force_inline uint32_t
+pack_1x64_32 (__m64 data)
+{
+ return _mm_cvtsi64_si32 (_mm_packs_pu16 (data, _mm_setzero_si64 ()));
+}
+
+/* Expand 16 bits positioned at @pos (0-3) of a mmx register into
+ *
+ * 00RR00GG00BB
+ *
+ * --- Expanding 565 in the low word ---
+ *
+ * m = (m << (32 - 3)) | (m << (16 - 5)) | m;
+ * m = m & (01f0003f001f);
+ * m = m * (008404100840);
+ * m = m >> 8;
+ *
+ * Note the trick here - the top word is shifted by another nibble to
+ * avoid it bumping into the middle word
+ */
+static force_inline __m64
+expand565_16_1x64 (uint16_t pixel)
+{
+ __m64 p;
+ __m64 t1, t2;
+
+ p = _mm_cvtsi32_si64 ((uint32_t) pixel);
+
+ t1 = _mm_slli_si64 (p, 36 - 11);
+ t2 = _mm_slli_si64 (p, 16 - 5);
+
+ p = _mm_or_si64 (t1, p);
+ p = _mm_or_si64 (t2, p);
+ p = _mm_and_si64 (p, mask_x565_rgb);
+ p = _mm_mullo_pi16 (p, mask_x565_unpack);
+
+ return _mm_srli_pi16 (p, 8);
+}
+
+/* ----------------------------------------------------------------------------
+ * Compose Core transformations
+ */
+static force_inline uint32_t
+core_combine_over_u_pixel_sse2 (uint32_t src, uint32_t dst)
+{
+ uint8_t a;
+ __m64 ms;
+
+ a = src >> 24;
+
+ if (a == 0xff)
+ {
+ return src;
+ }
+ else if (src)
+ {
+ ms = unpack_32_1x64 (src);
+ return pack_1x64_32 (
+ over_1x64 (ms, expand_alpha_1x64 (ms), unpack_32_1x64 (dst)));
+ }
+
+ return dst;
+}
+
+static force_inline uint32_t
+combine1 (const uint32_t *ps, const uint32_t *pm)
+{
+ uint32_t s = *ps;
+
+ if (pm)
+ {
+ __m64 ms, mm;
+
+ mm = unpack_32_1x64 (*pm);
+ mm = expand_alpha_1x64 (mm);
+
+ ms = unpack_32_1x64 (s);
+ ms = pix_multiply_1x64 (ms, mm);
+
+ s = pack_1x64_32 (ms);
+ }
+
+ return s;
+}
+
+static force_inline __m128i
+combine4 (const __m128i *ps, const __m128i *pm)
+{
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_msk_lo, xmm_msk_hi;
+ __m128i s;
+
+ if (pm)
+ {
+ xmm_msk_lo = load_128_unaligned (pm);
+
+ if (is_transparent (xmm_msk_lo))
+ return _mm_setzero_si128 ();
+ }
+
+ s = load_128_unaligned (ps);
+
+ if (pm)
+ {
+ unpack_128_2x128 (s, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_msk_lo, &xmm_msk_lo, &xmm_msk_hi);
+
+ expand_alpha_2x128 (xmm_msk_lo, xmm_msk_hi, &xmm_msk_lo, &xmm_msk_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_msk_lo, &xmm_msk_hi,
+ &xmm_src_lo, &xmm_src_hi);
+
+ s = pack_2x128_128 (xmm_src_lo, xmm_src_hi);
+ }
+
+ return s;
+}
+
+static force_inline void
+core_combine_over_u_sse2_mask (uint32_t * pd,
+ const uint32_t* ps,
+ const uint32_t* pm,
+ int w)
+{
+ uint32_t s, d;
+
+ /* Align dst on a 16-byte boundary */
+ while (w && ((unsigned long)pd & 15))
+ {
+ d = *pd;
+ s = combine1 (ps, pm);
+
+ if (s)
+ *pd = core_combine_over_u_pixel_sse2 (s, d);
+ pd++;
+ ps++;
+ pm++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ __m128i mask = load_128_unaligned ((__m128i *)pm);
+
+ if (!is_zero (mask))
+ {
+ __m128i src;
+ __m128i src_hi, src_lo;
+ __m128i mask_hi, mask_lo;
+ __m128i alpha_hi, alpha_lo;
+
+ src = load_128_unaligned ((__m128i *)ps);
+
+ if (is_opaque (_mm_and_si128 (src, mask)))
+ {
+ save_128_aligned ((__m128i *)pd, src);
+ }
+ else
+ {
+ __m128i dst = load_128_aligned ((__m128i *)pd);
+ __m128i dst_hi, dst_lo;
+
+ unpack_128_2x128 (mask, &mask_lo, &mask_hi);
+ unpack_128_2x128 (src, &src_lo, &src_hi);
+
+ expand_alpha_2x128 (mask_lo, mask_hi, &mask_lo, &mask_hi);
+ pix_multiply_2x128 (&src_lo, &src_hi,
+ &mask_lo, &mask_hi,
+ &src_lo, &src_hi);
+
+ unpack_128_2x128 (dst, &dst_lo, &dst_hi);
+
+ expand_alpha_2x128 (src_lo, src_hi,
+ &alpha_lo, &alpha_hi);
+
+ over_2x128 (&src_lo, &src_hi, &alpha_lo, &alpha_hi,
+ &dst_lo, &dst_hi);
+
+ save_128_aligned (
+ (__m128i *)pd,
+ pack_2x128_128 (dst_lo, dst_hi));
+ }
+ }
+
+ pm += 4;
+ ps += 4;
+ pd += 4;
+ w -= 4;
+ }
+ while (w)
+ {
+ d = *pd;
+ s = combine1 (ps, pm);
+
+ if (s)
+ *pd = core_combine_over_u_pixel_sse2 (s, d);
+ pd++;
+ ps++;
+ pm++;
+
+ w--;
+ }
+}
+
+static force_inline void
+core_combine_over_u_sse2_no_mask (uint32_t * pd,
+ const uint32_t* ps,
+ int w)
+{
+ uint32_t s, d;
+
+ /* Align dst on a 16-byte boundary */
+ while (w && ((unsigned long)pd & 15))
+ {
+ d = *pd;
+ s = *ps;
+
+ if (s)
+ *pd = core_combine_over_u_pixel_sse2 (s, d);
+ pd++;
+ ps++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ __m128i src;
+ __m128i src_hi, src_lo, dst_hi, dst_lo;
+ __m128i alpha_hi, alpha_lo;
+
+ src = load_128_unaligned ((__m128i *)ps);
+
+ if (!is_zero (src))
+ {
+ if (is_opaque (src))
+ {
+ save_128_aligned ((__m128i *)pd, src);
+ }
+ else
+ {
+ __m128i dst = load_128_aligned ((__m128i *)pd);
+
+ unpack_128_2x128 (src, &src_lo, &src_hi);
+ unpack_128_2x128 (dst, &dst_lo, &dst_hi);
+
+ expand_alpha_2x128 (src_lo, src_hi,
+ &alpha_lo, &alpha_hi);
+ over_2x128 (&src_lo, &src_hi, &alpha_lo, &alpha_hi,
+ &dst_lo, &dst_hi);
+
+ save_128_aligned (
+ (__m128i *)pd,
+ pack_2x128_128 (dst_lo, dst_hi));
+ }
+ }
+
+ ps += 4;
+ pd += 4;
+ w -= 4;
+ }
+ while (w)
+ {
+ d = *pd;
+ s = *ps;
+
+ if (s)
+ *pd = core_combine_over_u_pixel_sse2 (s, d);
+ pd++;
+ ps++;
+
+ w--;
+ }
+}
+
+static force_inline void
+core_combine_over_u_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t* pm,
+ int w)
+{
+ if (pm)
+ core_combine_over_u_sse2_mask (pd, ps, pm, w);
+ else
+ core_combine_over_u_sse2_no_mask (pd, ps, w);
+}
+
+static force_inline void
+core_combine_over_reverse_u_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t* pm,
+ int w)
+{
+ uint32_t s, d;
+
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+
+ /* Align dst on a 16-byte boundary */
+ while (w &&
+ ((unsigned long)pd & 15))
+ {
+ d = *pd;
+ s = combine1 (ps, pm);
+
+ *pd++ = core_combine_over_u_pixel_sse2 (d, s);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+
+ while (w >= 4)
+ {
+ /* I'm loading unaligned because I'm not sure
+ * about the address alignment.
+ */
+ xmm_src_hi = combine4 ((__m128i*)ps, (__m128i*)pm);
+ xmm_dst_hi = load_128_aligned ((__m128i*) pd);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ over_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_src_lo, &xmm_src_hi);
+
+ /* rebuid the 4 pixel data and save*/
+ save_128_aligned ((__m128i*)pd,
+ pack_2x128_128 (xmm_src_lo, xmm_src_hi));
+
+ w -= 4;
+ ps += 4;
+ pd += 4;
+
+ if (pm)
+ pm += 4;
+ }
+
+ while (w)
+ {
+ d = *pd;
+ s = combine1 (ps, pm);
+
+ *pd++ = core_combine_over_u_pixel_sse2 (d, s);
+ ps++;
+ w--;
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline uint32_t
+core_combine_in_u_pixelsse2 (uint32_t src, uint32_t dst)
+{
+ uint32_t maska = src >> 24;
+
+ if (maska == 0)
+ {
+ return 0;
+ }
+ else if (maska != 0xff)
+ {
+ return pack_1x64_32 (
+ pix_multiply_1x64 (unpack_32_1x64 (dst),
+ expand_alpha_1x64 (unpack_32_1x64 (src))));
+ }
+
+ return dst;
+}
+
+static force_inline void
+core_combine_in_u_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t* pm,
+ int w)
+{
+ uint32_t s, d;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+
+ while (w && ((unsigned long) pd & 15))
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_in_u_pixelsse2 (d, s);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*) pd);
+ xmm_src_hi = combine4 ((__m128i*) ps, (__m128i*) pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned ((__m128i*)pd,
+ pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ w -= 4;
+ if (pm)
+ pm += 4;
+ }
+
+ while (w)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_in_u_pixelsse2 (d, s);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline void
+core_combine_reverse_in_u_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, d;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+
+ while (w && ((unsigned long) pd & 15))
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_in_u_pixelsse2 (s, d);
+ ps++;
+ w--;
+ if (pm)
+ pm++;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*) pd);
+ xmm_src_hi = combine4 ((__m128i*) ps, (__m128i*)pm);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_src_lo, &xmm_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ w -= 4;
+ if (pm)
+ pm += 4;
+ }
+
+ while (w)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_in_u_pixelsse2 (s, d);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline void
+core_combine_reverse_out_u_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t* pm,
+ int w)
+{
+ while (w && ((unsigned long) pd & 15))
+ {
+ uint32_t s = combine1 (ps, pm);
+ uint32_t d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (d), negate_1x64 (
+ expand_alpha_1x64 (unpack_32_1x64 (s)))));
+
+ if (pm)
+ pm++;
+ ps++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+
+ xmm_src_hi = combine4 ((__m128i*)ps, (__m128i*)pm);
+ xmm_dst_hi = load_128_aligned ((__m128i*) pd);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ negate_2x128 (xmm_src_lo, xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+
+ pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_src_lo, &xmm_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ if (pm)
+ pm += 4;
+
+ w -= 4;
+ }
+
+ while (w)
+ {
+ uint32_t s = combine1 (ps, pm);
+ uint32_t d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (d), negate_1x64 (
+ expand_alpha_1x64 (unpack_32_1x64 (s)))));
+ ps++;
+ if (pm)
+ pm++;
+ w--;
+ }
+}
+
+static force_inline void
+core_combine_out_u_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t* pm,
+ int w)
+{
+ while (w && ((unsigned long) pd & 15))
+ {
+ uint32_t s = combine1 (ps, pm);
+ uint32_t d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (s), negate_1x64 (
+ expand_alpha_1x64 (unpack_32_1x64 (d)))));
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+
+ while (w >= 4)
+ {
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+
+ xmm_src_hi = combine4 ((__m128i*) ps, (__m128i*)pm);
+ xmm_dst_hi = load_128_aligned ((__m128i*) pd);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ negate_2x128 (xmm_dst_lo, xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ w -= 4;
+ if (pm)
+ pm += 4;
+ }
+
+ while (w)
+ {
+ uint32_t s = combine1 (ps, pm);
+ uint32_t d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (s), negate_1x64 (
+ expand_alpha_1x64 (unpack_32_1x64 (d)))));
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline uint32_t
+core_combine_atop_u_pixel_sse2 (uint32_t src,
+ uint32_t dst)
+{
+ __m64 s = unpack_32_1x64 (src);
+ __m64 d = unpack_32_1x64 (dst);
+
+ __m64 sa = negate_1x64 (expand_alpha_1x64 (s));
+ __m64 da = expand_alpha_1x64 (d);
+
+ return pack_1x64_32 (pix_add_multiply_1x64 (&s, &da, &d, &sa));
+}
+
+static force_inline void
+core_combine_atop_u_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t* pm,
+ int w)
+{
+ uint32_t s, d;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_alpha_src_lo, xmm_alpha_src_hi;
+ __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi;
+
+ while (w && ((unsigned long) pd & 15))
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_atop_u_pixel_sse2 (s, d);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+
+ while (w >= 4)
+ {
+ xmm_src_hi = combine4 ((__m128i*)ps, (__m128i*)pm);
+ xmm_dst_hi = load_128_aligned ((__m128i*) pd);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi);
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ negate_2x128 (xmm_alpha_src_lo, xmm_alpha_src_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi);
+
+ pix_add_multiply_2x128 (
+ &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi, &xmm_alpha_src_lo, &xmm_alpha_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ w -= 4;
+ if (pm)
+ pm += 4;
+ }
+
+ while (w)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_atop_u_pixel_sse2 (s, d);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline uint32_t
+core_combine_reverse_atop_u_pixel_sse2 (uint32_t src,
+ uint32_t dst)
+{
+ __m64 s = unpack_32_1x64 (src);
+ __m64 d = unpack_32_1x64 (dst);
+
+ __m64 sa = expand_alpha_1x64 (s);
+ __m64 da = negate_1x64 (expand_alpha_1x64 (d));
+
+ return pack_1x64_32 (pix_add_multiply_1x64 (&s, &da, &d, &sa));
+}
+
+static force_inline void
+core_combine_reverse_atop_u_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t* pm,
+ int w)
+{
+ uint32_t s, d;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_alpha_src_lo, xmm_alpha_src_hi;
+ __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi;
+
+ while (w && ((unsigned long) pd & 15))
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_reverse_atop_u_pixel_sse2 (s, d);
+ ps++;
+ w--;
+ if (pm)
+ pm++;
+ }
+
+ while (w >= 4)
+ {
+ xmm_src_hi = combine4 ((__m128i*)ps, (__m128i*)pm);
+ xmm_dst_hi = load_128_aligned ((__m128i*) pd);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi);
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ negate_2x128 (xmm_alpha_dst_lo, xmm_alpha_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ pix_add_multiply_2x128 (
+ &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi, &xmm_alpha_src_lo, &xmm_alpha_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ w -= 4;
+ if (pm)
+ pm += 4;
+ }
+
+ while (w)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_reverse_atop_u_pixel_sse2 (s, d);
+ ps++;
+ w--;
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline uint32_t
+core_combine_xor_u_pixel_sse2 (uint32_t src,
+ uint32_t dst)
+{
+ __m64 s = unpack_32_1x64 (src);
+ __m64 d = unpack_32_1x64 (dst);
+
+ __m64 neg_d = negate_1x64 (expand_alpha_1x64 (d));
+ __m64 neg_s = negate_1x64 (expand_alpha_1x64 (s));
+
+ return pack_1x64_32 (pix_add_multiply_1x64 (&s, &neg_d, &d, &neg_s));
+}
+
+static force_inline void
+core_combine_xor_u_sse2 (uint32_t* dst,
+ const uint32_t* src,
+ const uint32_t *mask,
+ int width)
+{
+ int w = width;
+ uint32_t s, d;
+ uint32_t* pd = dst;
+ const uint32_t* ps = src;
+ const uint32_t* pm = mask;
+
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_alpha_src_lo, xmm_alpha_src_hi;
+ __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi;
+
+ while (w && ((unsigned long) pd & 15))
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_xor_u_pixel_sse2 (s, d);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+
+ while (w >= 4)
+ {
+ xmm_src = combine4 ((__m128i*) ps, (__m128i*) pm);
+ xmm_dst = load_128_aligned ((__m128i*) pd);
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi);
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ negate_2x128 (xmm_alpha_src_lo, xmm_alpha_src_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi);
+ negate_2x128 (xmm_alpha_dst_lo, xmm_alpha_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ pix_add_multiply_2x128 (
+ &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi, &xmm_alpha_src_lo, &xmm_alpha_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ w -= 4;
+ if (pm)
+ pm += 4;
+ }
+
+ while (w)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_xor_u_pixel_sse2 (s, d);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline void
+core_combine_add_u_sse2 (uint32_t* dst,
+ const uint32_t* src,
+ const uint32_t* mask,
+ int width)
+{
+ int w = width;
+ uint32_t s, d;
+ uint32_t* pd = dst;
+ const uint32_t* ps = src;
+ const uint32_t* pm = mask;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ ps++;
+ if (pm)
+ pm++;
+ *pd++ = _mm_cvtsi64_si32 (
+ _mm_adds_pu8 (_mm_cvtsi32_si64 (s), _mm_cvtsi32_si64 (d)));
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ __m128i s;
+
+ s = combine4 ((__m128i*)ps, (__m128i*)pm);
+
+ save_128_aligned (
+ (__m128i*)pd, _mm_adds_epu8 (s, load_128_aligned ((__m128i*)pd)));
+
+ pd += 4;
+ ps += 4;
+ if (pm)
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w--)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ ps++;
+ *pd++ = _mm_cvtsi64_si32 (
+ _mm_adds_pu8 (_mm_cvtsi32_si64 (s), _mm_cvtsi32_si64 (d)));
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline uint32_t
+core_combine_saturate_u_pixel_sse2 (uint32_t src,
+ uint32_t dst)
+{
+ __m64 ms = unpack_32_1x64 (src);
+ __m64 md = unpack_32_1x64 (dst);
+ uint32_t sa = src >> 24;
+ uint32_t da = ~dst >> 24;
+
+ if (sa > da)
+ {
+ ms = pix_multiply_1x64 (
+ ms, expand_alpha_1x64 (unpack_32_1x64 (DIV_UN8 (da, sa) << 24)));
+ }
+
+ return pack_1x64_32 (_mm_adds_pu16 (md, ms));
+}
+
+static force_inline void
+core_combine_saturate_u_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, d;
+
+ uint32_t pack_cmp;
+ __m128i xmm_src, xmm_dst;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_saturate_u_pixel_sse2 (s, d);
+ w--;
+ ps++;
+ if (pm)
+ pm++;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst = load_128_aligned ((__m128i*)pd);
+ xmm_src = combine4 ((__m128i*)ps, (__m128i*)pm);
+
+ pack_cmp = _mm_movemask_epi8 (
+ _mm_cmpgt_epi32 (
+ _mm_srli_epi32 (xmm_src, 24),
+ _mm_srli_epi32 (_mm_xor_si128 (xmm_dst, mask_ff000000), 24)));
+
+ /* if some alpha src is grater than respective ~alpha dst */
+ if (pack_cmp)
+ {
+ s = combine1 (ps++, pm);
+ d = *pd;
+ *pd++ = core_combine_saturate_u_pixel_sse2 (s, d);
+ if (pm)
+ pm++;
+
+ s = combine1 (ps++, pm);
+ d = *pd;
+ *pd++ = core_combine_saturate_u_pixel_sse2 (s, d);
+ if (pm)
+ pm++;
+
+ s = combine1 (ps++, pm);
+ d = *pd;
+ *pd++ = core_combine_saturate_u_pixel_sse2 (s, d);
+ if (pm)
+ pm++;
+
+ s = combine1 (ps++, pm);
+ d = *pd;
+ *pd++ = core_combine_saturate_u_pixel_sse2 (s, d);
+ if (pm)
+ pm++;
+ }
+ else
+ {
+ save_128_aligned ((__m128i*)pd, _mm_adds_epu8 (xmm_dst, xmm_src));
+
+ pd += 4;
+ ps += 4;
+ if (pm)
+ pm += 4;
+ }
+
+ w -= 4;
+ }
+
+ while (w--)
+ {
+ s = combine1 (ps, pm);
+ d = *pd;
+
+ *pd++ = core_combine_saturate_u_pixel_sse2 (s, d);
+ ps++;
+ if (pm)
+ pm++;
+ }
+}
+
+static force_inline void
+core_combine_src_ca_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m)));
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m)));
+ w--;
+ }
+}
+
+static force_inline uint32_t
+core_combine_over_ca_pixel_sse2 (uint32_t src,
+ uint32_t mask,
+ uint32_t dst)
+{
+ __m64 s = unpack_32_1x64 (src);
+ __m64 expAlpha = expand_alpha_1x64 (s);
+ __m64 unpk_mask = unpack_32_1x64 (mask);
+ __m64 unpk_dst = unpack_32_1x64 (dst);
+
+ return pack_1x64_32 (in_over_1x64 (&s, &expAlpha, &unpk_mask, &unpk_dst));
+}
+
+static force_inline void
+core_combine_over_ca_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_over_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ in_over_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_over_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+}
+
+static force_inline uint32_t
+core_combine_over_reverse_ca_pixel_sse2 (uint32_t src,
+ uint32_t mask,
+ uint32_t dst)
+{
+ __m64 d = unpack_32_1x64 (dst);
+
+ return pack_1x64_32 (
+ over_1x64 (d, expand_alpha_1x64 (d),
+ pix_multiply_1x64 (unpack_32_1x64 (src),
+ unpack_32_1x64 (mask))));
+}
+
+static force_inline void
+core_combine_over_reverse_ca_sse2 (uint32_t* pd,
+ const uint32_t* ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_over_reverse_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ over_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_mask_lo, xmm_mask_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_over_reverse_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+}
+
+static force_inline void
+core_combine_in_ca_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (m)),
+ expand_alpha_1x64 (unpack_32_1x64 (d))));
+
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (s), unpack_32_1x64 (m)),
+ expand_alpha_1x64 (unpack_32_1x64 (d))));
+
+ w--;
+ }
+}
+
+static force_inline void
+core_combine_in_reverse_ca_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (d),
+ pix_multiply_1x64 (unpack_32_1x64 (m),
+ expand_alpha_1x64 (unpack_32_1x64 (s)))));
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+ pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (d),
+ pix_multiply_1x64 (unpack_32_1x64 (m),
+ expand_alpha_1x64 (unpack_32_1x64 (s)))));
+ w--;
+ }
+}
+
+static force_inline void
+core_combine_out_ca_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (s), unpack_32_1x64 (m)),
+ negate_1x64 (expand_alpha_1x64 (unpack_32_1x64 (d)))));
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+ negate_2x128 (xmm_alpha_lo, xmm_alpha_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+ pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (s), unpack_32_1x64 (m)),
+ negate_1x64 (expand_alpha_1x64 (unpack_32_1x64 (d)))));
+
+ w--;
+ }
+}
+
+static force_inline void
+core_combine_out_reverse_ca_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (d),
+ negate_1x64 (pix_multiply_1x64 (
+ unpack_32_1x64 (m),
+ expand_alpha_1x64 (unpack_32_1x64 (s))))));
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ negate_2x128 (xmm_mask_lo, xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ pix_multiply_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (d),
+ negate_1x64 (pix_multiply_1x64 (
+ unpack_32_1x64 (m),
+ expand_alpha_1x64 (unpack_32_1x64 (s))))));
+ w--;
+ }
+}
+
+static force_inline uint32_t
+core_combine_atop_ca_pixel_sse2 (uint32_t src,
+ uint32_t mask,
+ uint32_t dst)
+{
+ __m64 m = unpack_32_1x64 (mask);
+ __m64 s = unpack_32_1x64 (src);
+ __m64 d = unpack_32_1x64 (dst);
+ __m64 sa = expand_alpha_1x64 (s);
+ __m64 da = expand_alpha_1x64 (d);
+
+ s = pix_multiply_1x64 (s, m);
+ m = negate_1x64 (pix_multiply_1x64 (m, sa));
+
+ return pack_1x64_32 (pix_add_multiply_1x64 (&d, &m, &s, &da));
+}
+
+static force_inline void
+core_combine_atop_ca_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_alpha_src_lo, xmm_alpha_src_hi;
+ __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_atop_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi);
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_src_lo, &xmm_src_hi);
+ pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ negate_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ pix_add_multiply_2x128 (
+ &xmm_dst_lo, &xmm_dst_hi, &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_atop_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+}
+
+static force_inline uint32_t
+core_combine_reverse_atop_ca_pixel_sse2 (uint32_t src,
+ uint32_t mask,
+ uint32_t dst)
+{
+ __m64 m = unpack_32_1x64 (mask);
+ __m64 s = unpack_32_1x64 (src);
+ __m64 d = unpack_32_1x64 (dst);
+
+ __m64 da = negate_1x64 (expand_alpha_1x64 (d));
+ __m64 sa = expand_alpha_1x64 (s);
+
+ s = pix_multiply_1x64 (s, m);
+ m = pix_multiply_1x64 (m, sa);
+
+ return pack_1x64_32 (pix_add_multiply_1x64 (&d, &m, &s, &da));
+}
+
+static force_inline void
+core_combine_reverse_atop_ca_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_alpha_src_lo, xmm_alpha_src_hi;
+ __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_reverse_atop_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi);
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_src_lo, &xmm_src_hi);
+ pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ negate_2x128 (xmm_alpha_dst_lo, xmm_alpha_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ pix_add_multiply_2x128 (
+ &xmm_dst_lo, &xmm_dst_hi, &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_reverse_atop_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+}
+
+static force_inline uint32_t
+core_combine_xor_ca_pixel_sse2 (uint32_t src,
+ uint32_t mask,
+ uint32_t dst)
+{
+ __m64 a = unpack_32_1x64 (mask);
+ __m64 s = unpack_32_1x64 (src);
+ __m64 d = unpack_32_1x64 (dst);
+
+ __m64 alpha_dst = negate_1x64 (pix_multiply_1x64 (
+ a, expand_alpha_1x64 (s)));
+ __m64 dest = pix_multiply_1x64 (s, a);
+ __m64 alpha_src = negate_1x64 (expand_alpha_1x64 (d));
+
+ return pack_1x64_32 (pix_add_multiply_1x64 (&d,
+ &alpha_dst,
+ &dest,
+ &alpha_src));
+}
+
+static force_inline void
+core_combine_xor_ca_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_alpha_src_lo, xmm_alpha_src_hi;
+ __m128i xmm_alpha_dst_lo, xmm_alpha_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_xor_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi);
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_src_lo, &xmm_src_hi);
+ pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi,
+ &xmm_alpha_src_lo, &xmm_alpha_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ negate_2x128 (xmm_alpha_dst_lo, xmm_alpha_dst_hi,
+ &xmm_alpha_dst_lo, &xmm_alpha_dst_hi);
+ negate_2x128 (xmm_mask_lo, xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ pix_add_multiply_2x128 (
+ &xmm_dst_lo, &xmm_dst_hi, &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_src_lo, &xmm_src_hi, &xmm_alpha_dst_lo, &xmm_alpha_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = core_combine_xor_ca_pixel_sse2 (s, m, d);
+ w--;
+ }
+}
+
+static force_inline void
+core_combine_add_ca_sse2 (uint32_t * pd,
+ const uint32_t *ps,
+ const uint32_t *pm,
+ int w)
+{
+ uint32_t s, m, d;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask_lo, xmm_mask_hi;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ _mm_adds_pu8 (pix_multiply_1x64 (unpack_32_1x64 (s),
+ unpack_32_1x64 (m)),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_src_hi = load_128_unaligned ((__m128i*)ps);
+ xmm_mask_hi = load_128_unaligned ((__m128i*)pm);
+ xmm_dst_hi = load_128_aligned ((__m128i*)pd);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_src_lo, &xmm_src_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (
+ _mm_adds_epu8 (xmm_src_lo, xmm_dst_lo),
+ _mm_adds_epu8 (xmm_src_hi, xmm_dst_hi)));
+
+ ps += 4;
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ s = *ps++;
+ m = *pm++;
+ d = *pd;
+
+ *pd++ = pack_1x64_32 (
+ _mm_adds_pu8 (pix_multiply_1x64 (unpack_32_1x64 (s),
+ unpack_32_1x64 (m)),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+}
+
+/* ---------------------------------------------------
+ * fb_compose_setup_sSE2
+ */
+static force_inline __m64
+create_mask_16_64 (uint16_t mask)
+{
+ return _mm_set1_pi16 (mask);
+}
+
+static force_inline __m128i
+create_mask_16_128 (uint16_t mask)
+{
+ return _mm_set1_epi16 (mask);
+}
+
+static force_inline __m64
+create_mask_2x32_64 (uint32_t mask0,
+ uint32_t mask1)
+{
+ return _mm_set_pi32 (mask0, mask1);
+}
+
+/* Work around a code generation bug in Sun Studio 12. */
+#if defined(__SUNPRO_C) && (__SUNPRO_C >= 0x590)
+# define create_mask_2x32_128(mask0, mask1) \
+ (_mm_set_epi32 ((mask0), (mask1), (mask0), (mask1)))
+#else
+static force_inline __m128i
+create_mask_2x32_128 (uint32_t mask0,
+ uint32_t mask1)
+{
+ return _mm_set_epi32 (mask0, mask1, mask0, mask1);
+}
+#endif
+
+/* SSE2 code patch for fbcompose.c */
+
+static void
+sse2_combine_over_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_over_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_over_reverse_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_over_reverse_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_in_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_in_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_in_reverse_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_reverse_in_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_out_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_out_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_out_reverse_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_reverse_out_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_atop_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_atop_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_atop_reverse_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_reverse_atop_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_xor_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_xor_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_add_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_add_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_saturate_u (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_saturate_u_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_src_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_src_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_over_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_over_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_over_reverse_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_over_reverse_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_in_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_in_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_in_reverse_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_in_reverse_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_out_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_out_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_out_reverse_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_out_reverse_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_atop_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_atop_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_atop_reverse_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_reverse_atop_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_xor_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_xor_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+static void
+sse2_combine_add_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ uint32_t * dst,
+ const uint32_t * src,
+ const uint32_t * mask,
+ int width)
+{
+ core_combine_add_ca_sse2 (dst, src, mask, width);
+ _mm_empty ();
+}
+
+/* -------------------------------------------------------------------
+ * composite_over_n_8888
+ */
+
+static void
+sse2_composite_over_n_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src;
+ uint32_t *dst_line, *dst, d;
+ int32_t w;
+ int dst_stride;
+ __m128i xmm_src, xmm_alpha;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+
+ xmm_src = expand_pixel_32_1x128 (src);
+ xmm_alpha = expand_alpha_1x128 (xmm_src);
+
+ while (height--)
+ {
+ dst = dst_line;
+
+ dst_line += dst_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ d = *dst;
+ *dst++ = pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src),
+ _mm_movepi64_pi64 (xmm_alpha),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ /* rebuid the 4 pixel data and save*/
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ w -= 4;
+ dst += 4;
+ }
+
+ while (w)
+ {
+ d = *dst;
+ *dst++ = pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src),
+ _mm_movepi64_pi64 (xmm_alpha),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+
+ }
+ _mm_empty ();
+}
+
+/* ---------------------------------------------------------------------
+ * composite_over_n_0565
+ */
+static void
+sse2_composite_over_n_0565 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src;
+ uint16_t *dst_line, *dst, d;
+ int32_t w;
+ int dst_stride;
+ __m128i xmm_src, xmm_alpha;
+ __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+
+ xmm_src = expand_pixel_32_1x128 (src);
+ xmm_alpha = expand_alpha_1x128 (xmm_src);
+
+ while (height--)
+ {
+ dst = dst_line;
+
+ dst_line += dst_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ d = *dst;
+
+ *dst++ = pack_565_32_16 (
+ pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src),
+ _mm_movepi64_pi64 (xmm_alpha),
+ expand565_16_1x64 (d))));
+ w--;
+ }
+
+ while (w >= 8)
+ {
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_565_128_4x128 (xmm_dst,
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3);
+
+ over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_dst0, &xmm_dst1);
+ over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_dst2, &xmm_dst3);
+
+ xmm_dst = pack_565_4x128_128 (
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3);
+
+ save_128_aligned ((__m128i*)dst, xmm_dst);
+
+ dst += 8;
+ w -= 8;
+ }
+
+ while (w--)
+ {
+ d = *dst;
+ *dst++ = pack_565_32_16 (
+ pack_1x64_32 (over_1x64 (_mm_movepi64_pi64 (xmm_src),
+ _mm_movepi64_pi64 (xmm_alpha),
+ expand565_16_1x64 (d))));
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* ------------------------------
+ * composite_add_n_8888_8888_ca
+ */
+static void
+sse2_composite_add_n_8888_8888_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint32_t *dst_line, d;
+ uint32_t *mask_line, m;
+ uint32_t pack_cmp;
+ int dst_stride, mask_stride;
+
+ __m128i xmm_src, xmm_alpha;
+ __m128i xmm_dst;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+
+ __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+ srca = src >> 24;
+
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1);
+
+ xmm_src = _mm_unpacklo_epi8 (
+ create_mask_2x32_128 (src, src), _mm_setzero_si128 ());
+ xmm_alpha = expand_alpha_1x128 (xmm_src);
+ mmx_src = _mm_movepi64_pi64 (xmm_src);
+ mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+
+ while (height--)
+ {
+ int w = width;
+ const uint32_t *pm = (uint32_t *)mask_line;
+ uint32_t *pd = (uint32_t *)dst_line;
+
+ dst_line += dst_stride;
+ mask_line += mask_stride;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ m = *pm++;
+
+ if (m)
+ {
+ d = *pd;
+
+ mmx_mask = unpack_32_1x64 (m);
+ mmx_dest = unpack_32_1x64 (d);
+
+ *pd = pack_1x64_32 (
+ _mm_adds_pu8 (pix_multiply_1x64 (mmx_mask, mmx_src), mmx_dest));
+ }
+
+ pd++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_mask = load_128_unaligned ((__m128i*)pm);
+
+ pack_cmp =
+ _mm_movemask_epi8 (
+ _mm_cmpeq_epi32 (xmm_mask, _mm_setzero_si128 ()));
+
+ /* if all bits in mask are zero, pack_cmp are equal to 0xffff */
+ if (pack_cmp != 0xffff)
+ {
+ xmm_dst = load_128_aligned ((__m128i*)pd);
+
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+
+ pix_multiply_2x128 (&xmm_src, &xmm_src,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+ xmm_mask_hi = pack_2x128_128 (xmm_mask_lo, xmm_mask_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, _mm_adds_epu8 (xmm_mask_hi, xmm_dst));
+ }
+
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ m = *pm++;
+
+ if (m)
+ {
+ d = *pd;
+
+ mmx_mask = unpack_32_1x64 (m);
+ mmx_dest = unpack_32_1x64 (d);
+
+ *pd = pack_1x64_32 (
+ _mm_adds_pu8 (pix_multiply_1x64 (mmx_mask, mmx_src), mmx_dest));
+ }
+
+ pd++;
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* ---------------------------------------------------------------------------
+ * composite_over_n_8888_8888_ca
+ */
+
+static void
+sse2_composite_over_n_8888_8888_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src;
+ uint32_t *dst_line, d;
+ uint32_t *mask_line, m;
+ uint32_t pack_cmp;
+ int dst_stride, mask_stride;
+
+ __m128i xmm_src, xmm_alpha;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+
+ __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1);
+
+ xmm_src = _mm_unpacklo_epi8 (
+ create_mask_2x32_128 (src, src), _mm_setzero_si128 ());
+ xmm_alpha = expand_alpha_1x128 (xmm_src);
+ mmx_src = _mm_movepi64_pi64 (xmm_src);
+ mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+
+ while (height--)
+ {
+ int w = width;
+ const uint32_t *pm = (uint32_t *)mask_line;
+ uint32_t *pd = (uint32_t *)dst_line;
+
+ dst_line += dst_stride;
+ mask_line += mask_stride;
+
+ while (w && (unsigned long)pd & 15)
+ {
+ m = *pm++;
+
+ if (m)
+ {
+ d = *pd;
+ mmx_mask = unpack_32_1x64 (m);
+ mmx_dest = unpack_32_1x64 (d);
+
+ *pd = pack_1x64_32 (in_over_1x64 (&mmx_src,
+ &mmx_alpha,
+ &mmx_mask,
+ &mmx_dest));
+ }
+
+ pd++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_mask = load_128_unaligned ((__m128i*)pm);
+
+ pack_cmp =
+ _mm_movemask_epi8 (
+ _mm_cmpeq_epi32 (xmm_mask, _mm_setzero_si128 ()));
+
+ /* if all bits in mask are zero, pack_cmp are equal to 0xffff */
+ if (pack_cmp != 0xffff)
+ {
+ xmm_dst = load_128_aligned ((__m128i*)pd);
+
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ in_over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)pd, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+
+ pd += 4;
+ pm += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ m = *pm++;
+
+ if (m)
+ {
+ d = *pd;
+ mmx_mask = unpack_32_1x64 (m);
+ mmx_dest = unpack_32_1x64 (d);
+
+ *pd = pack_1x64_32 (
+ in_over_1x64 (&mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest));
+ }
+
+ pd++;
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/*---------------------------------------------------------------------
+ * composite_over_8888_n_8888
+ */
+
+static void
+sse2_composite_over_8888_n_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src;
+ uint32_t mask;
+ int32_t w;
+ int dst_stride, src_stride;
+
+ __m128i xmm_mask;
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ mask = _pixman_image_get_solid (imp, mask_image, PIXMAN_a8r8g8b8);
+
+ xmm_mask = create_mask_16_128 (mask >> 24);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ uint32_t s = *src++;
+
+ if (s)
+ {
+ uint32_t d = *dst;
+
+ __m64 ms = unpack_32_1x64 (s);
+ __m64 alpha = expand_alpha_1x64 (ms);
+ __m64 dest = _mm_movepi64_pi64 (xmm_mask);
+ __m64 alpha_dst = unpack_32_1x64 (d);
+
+ *dst = pack_1x64_32 (
+ in_over_1x64 (&ms, &alpha, &dest, &alpha_dst));
+ }
+ dst++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_src = load_128_unaligned ((__m128i*)src);
+
+ if (!is_zero (xmm_src))
+ {
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ in_over_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_mask, &xmm_mask,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+
+ dst += 4;
+ src += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ uint32_t s = *src++;
+
+ if (s)
+ {
+ uint32_t d = *dst;
+
+ __m64 ms = unpack_32_1x64 (s);
+ __m64 alpha = expand_alpha_1x64 (ms);
+ __m64 mask = _mm_movepi64_pi64 (xmm_mask);
+ __m64 dest = unpack_32_1x64 (d);
+
+ *dst = pack_1x64_32 (
+ in_over_1x64 (&ms, &alpha, &mask, &dest));
+ }
+
+ dst++;
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/*---------------------------------------------------------------------
+ * composite_over_8888_n_8888
+ */
+
+static void
+sse2_composite_src_x888_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src;
+ int32_t w;
+ int dst_stride, src_stride;
+
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ *dst++ = *src++ | 0xff000000;
+ w--;
+ }
+
+ while (w >= 16)
+ {
+ __m128i xmm_src1, xmm_src2, xmm_src3, xmm_src4;
+
+ xmm_src1 = load_128_unaligned ((__m128i*)src + 0);
+ xmm_src2 = load_128_unaligned ((__m128i*)src + 1);
+ xmm_src3 = load_128_unaligned ((__m128i*)src + 2);
+ xmm_src4 = load_128_unaligned ((__m128i*)src + 3);
+
+ save_128_aligned ((__m128i*)dst + 0, _mm_or_si128 (xmm_src1, mask_ff000000));
+ save_128_aligned ((__m128i*)dst + 1, _mm_or_si128 (xmm_src2, mask_ff000000));
+ save_128_aligned ((__m128i*)dst + 2, _mm_or_si128 (xmm_src3, mask_ff000000));
+ save_128_aligned ((__m128i*)dst + 3, _mm_or_si128 (xmm_src4, mask_ff000000));
+
+ dst += 16;
+ src += 16;
+ w -= 16;
+ }
+
+ while (w)
+ {
+ *dst++ = *src++ | 0xff000000;
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* ---------------------------------------------------------------------
+ * composite_over_x888_n_8888
+ */
+static void
+sse2_composite_over_x888_n_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src;
+ uint32_t mask;
+ int dst_stride, src_stride;
+ int32_t w;
+
+ __m128i xmm_mask, xmm_alpha;
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ mask = _pixman_image_get_solid (imp, mask_image, PIXMAN_a8r8g8b8);
+
+ xmm_mask = create_mask_16_128 (mask >> 24);
+ xmm_alpha = mask_00ff;
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ uint32_t s = (*src++) | 0xff000000;
+ uint32_t d = *dst;
+
+ __m64 src = unpack_32_1x64 (s);
+ __m64 alpha = _mm_movepi64_pi64 (xmm_alpha);
+ __m64 mask = _mm_movepi64_pi64 (xmm_mask);
+ __m64 dest = unpack_32_1x64 (d);
+
+ *dst++ = pack_1x64_32 (
+ in_over_1x64 (&src, &alpha, &mask, &dest));
+
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_src = _mm_or_si128 (
+ load_128_unaligned ((__m128i*)src), mask_ff000000);
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ in_over_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_mask, &xmm_mask,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ dst += 4;
+ src += 4;
+ w -= 4;
+
+ }
+
+ while (w)
+ {
+ uint32_t s = (*src++) | 0xff000000;
+ uint32_t d = *dst;
+
+ __m64 src = unpack_32_1x64 (s);
+ __m64 alpha = _mm_movepi64_pi64 (xmm_alpha);
+ __m64 mask = _mm_movepi64_pi64 (xmm_mask);
+ __m64 dest = unpack_32_1x64 (d);
+
+ *dst++ = pack_1x64_32 (
+ in_over_1x64 (&src, &alpha, &mask, &dest));
+
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* --------------------------------------------------------------------
+ * composite_over_8888_8888
+ */
+static void
+sse2_composite_over_8888_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ int dst_stride, src_stride;
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ dst = dst_line;
+ src = src_line;
+
+ while (height--)
+ {
+ core_combine_over_u_sse2 (dst, src, NULL, width);
+
+ dst += dst_stride;
+ src += src_stride;
+ }
+ _mm_empty ();
+}
+
+/* ------------------------------------------------------------------
+ * composite_over_8888_0565
+ */
+static force_inline uint16_t
+composite_over_8888_0565pixel (uint32_t src, uint16_t dst)
+{
+ __m64 ms;
+
+ ms = unpack_32_1x64 (src);
+ return pack_565_32_16 (
+ pack_1x64_32 (
+ over_1x64 (
+ ms, expand_alpha_1x64 (ms), expand565_16_1x64 (dst))));
+}
+
+static void
+sse2_composite_over_8888_0565 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint16_t *dst_line, *dst, d;
+ uint32_t *src_line, *src, s;
+ int dst_stride, src_stride;
+ int32_t w;
+
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+#if 0
+ /* FIXME
+ *
+ * I copy the code from MMX one and keep the fixme.
+ * If it's a problem there, probably is a problem here.
+ */
+ assert (src_image->drawable == mask_image->drawable);
+#endif
+
+ while (height--)
+ {
+ dst = dst_line;
+ src = src_line;
+
+ dst_line += dst_stride;
+ src_line += src_stride;
+ w = width;
+
+ /* Align dst on a 16-byte boundary */
+ while (w &&
+ ((unsigned long)dst & 15))
+ {
+ s = *src++;
+ d = *dst;
+
+ *dst++ = composite_over_8888_0565pixel (s, d);
+ w--;
+ }
+
+ /* It's a 8 pixel loop */
+ while (w >= 8)
+ {
+ /* I'm loading unaligned because I'm not sure
+ * about the address alignment.
+ */
+ xmm_src = load_128_unaligned ((__m128i*) src);
+ xmm_dst = load_128_aligned ((__m128i*) dst);
+
+ /* Unpacking */
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_565_128_4x128 (xmm_dst,
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3);
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ /* I'm loading next 4 pixels from memory
+ * before to optimze the memory read.
+ */
+ xmm_src = load_128_unaligned ((__m128i*) (src + 4));
+
+ over_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_dst0, &xmm_dst1);
+
+ /* Unpacking */
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ over_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_dst2, &xmm_dst3);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_565_4x128_128 (
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3));
+
+ w -= 8;
+ dst += 8;
+ src += 8;
+ }
+
+ while (w--)
+ {
+ s = *src++;
+ d = *dst;
+
+ *dst++ = composite_over_8888_0565pixel (s, d);
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* -----------------------------------------------------------------
+ * composite_over_n_8_8888
+ */
+
+static void
+sse2_composite_over_n_8_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint32_t *dst_line, *dst;
+ uint8_t *mask_line, *mask;
+ int dst_stride, mask_stride;
+ int32_t w;
+ uint32_t m, d;
+
+ __m128i xmm_src, xmm_alpha, xmm_def;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+
+ __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ xmm_def = create_mask_2x32_128 (src, src);
+ xmm_src = expand_pixel_32_1x128 (src);
+ xmm_alpha = expand_alpha_1x128 (xmm_src);
+ mmx_src = _mm_movepi64_pi64 (xmm_src);
+ mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ uint8_t m = *mask++;
+
+ if (m)
+ {
+ d = *dst;
+ mmx_mask = expand_pixel_8_1x64 (m);
+ mmx_dest = unpack_32_1x64 (d);
+
+ *dst = pack_1x64_32 (in_over_1x64 (&mmx_src,
+ &mmx_alpha,
+ &mmx_mask,
+ &mmx_dest));
+ }
+
+ w--;
+ dst++;
+ }
+
+ while (w >= 4)
+ {
+ m = *((uint32_t*)mask);
+
+ if (srca == 0xff && m == 0xffffffff)
+ {
+ save_128_aligned ((__m128i*)dst, xmm_def);
+ }
+ else if (m)
+ {
+ xmm_dst = load_128_aligned ((__m128i*) dst);
+ xmm_mask = unpack_32_1x128 (m);
+ xmm_mask = _mm_unpacklo_epi8 (xmm_mask, _mm_setzero_si128 ());
+
+ /* Unpacking */
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ in_over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+
+ w -= 4;
+ dst += 4;
+ mask += 4;
+ }
+
+ while (w)
+ {
+ uint8_t m = *mask++;
+
+ if (m)
+ {
+ d = *dst;
+ mmx_mask = expand_pixel_8_1x64 (m);
+ mmx_dest = unpack_32_1x64 (d);
+
+ *dst = pack_1x64_32 (in_over_1x64 (&mmx_src,
+ &mmx_alpha,
+ &mmx_mask,
+ &mmx_dest));
+ }
+
+ w--;
+ dst++;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* ----------------------------------------------------------------
+ * composite_over_n_8_8888
+ */
+
+pixman_bool_t
+pixman_fill_sse2 (uint32_t *bits,
+ int stride,
+ int bpp,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t data)
+{
+ uint32_t byte_width;
+ uint8_t *byte_line;
+
+ __m128i xmm_def;
+
+ if (bpp == 8)
+ {
+ uint8_t b;
+ uint16_t w;
+
+ stride = stride * (int) sizeof (uint32_t) / 1;
+ byte_line = (uint8_t *)(((uint8_t *)bits) + stride * y + x);
+ byte_width = width;
+ stride *= 1;
+
+ b = data & 0xff;
+ w = (b << 8) | b;
+ data = (w << 16) | w;
+ }
+ else if (bpp == 16)
+ {
+ stride = stride * (int) sizeof (uint32_t) / 2;
+ byte_line = (uint8_t *)(((uint16_t *)bits) + stride * y + x);
+ byte_width = 2 * width;
+ stride *= 2;
+
+ data = (data & 0xffff) * 0x00010001;
+ }
+ else if (bpp == 32)
+ {
+ stride = stride * (int) sizeof (uint32_t) / 4;
+ byte_line = (uint8_t *)(((uint32_t *)bits) + stride * y + x);
+ byte_width = 4 * width;
+ stride *= 4;
+ }
+ else
+ {
+ return FALSE;
+ }
+
+ xmm_def = create_mask_2x32_128 (data, data);
+
+ while (height--)
+ {
+ int w;
+ uint8_t *d = byte_line;
+ byte_line += stride;
+ w = byte_width;
+
+ while (w >= 1 && ((unsigned long)d & 1))
+ {
+ *(uint8_t *)d = data;
+ w -= 1;
+ d += 1;
+ }
+
+ while (w >= 2 && ((unsigned long)d & 3))
+ {
+ *(uint16_t *)d = data;
+ w -= 2;
+ d += 2;
+ }
+
+ while (w >= 4 && ((unsigned long)d & 15))
+ {
+ *(uint32_t *)d = data;
+
+ w -= 4;
+ d += 4;
+ }
+
+ while (w >= 128)
+ {
+ save_128_aligned ((__m128i*)(d), xmm_def);
+ save_128_aligned ((__m128i*)(d + 16), xmm_def);
+ save_128_aligned ((__m128i*)(d + 32), xmm_def);
+ save_128_aligned ((__m128i*)(d + 48), xmm_def);
+ save_128_aligned ((__m128i*)(d + 64), xmm_def);
+ save_128_aligned ((__m128i*)(d + 80), xmm_def);
+ save_128_aligned ((__m128i*)(d + 96), xmm_def);
+ save_128_aligned ((__m128i*)(d + 112), xmm_def);
+
+ d += 128;
+ w -= 128;
+ }
+
+ if (w >= 64)
+ {
+ save_128_aligned ((__m128i*)(d), xmm_def);
+ save_128_aligned ((__m128i*)(d + 16), xmm_def);
+ save_128_aligned ((__m128i*)(d + 32), xmm_def);
+ save_128_aligned ((__m128i*)(d + 48), xmm_def);
+
+ d += 64;
+ w -= 64;
+ }
+
+ if (w >= 32)
+ {
+ save_128_aligned ((__m128i*)(d), xmm_def);
+ save_128_aligned ((__m128i*)(d + 16), xmm_def);
+
+ d += 32;
+ w -= 32;
+ }
+
+ if (w >= 16)
+ {
+ save_128_aligned ((__m128i*)(d), xmm_def);
+
+ d += 16;
+ w -= 16;
+ }
+
+ while (w >= 4)
+ {
+ *(uint32_t *)d = data;
+
+ w -= 4;
+ d += 4;
+ }
+
+ if (w >= 2)
+ {
+ *(uint16_t *)d = data;
+ w -= 2;
+ d += 2;
+ }
+
+ if (w >= 1)
+ {
+ *(uint8_t *)d = data;
+ w -= 1;
+ d += 1;
+ }
+ }
+
+ _mm_empty ();
+ return TRUE;
+}
+
+static void
+sse2_composite_src_n_8_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint32_t *dst_line, *dst;
+ uint8_t *mask_line, *mask;
+ int dst_stride, mask_stride;
+ int32_t w;
+ uint32_t m;
+
+ __m128i xmm_src, xmm_def;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ srca = src >> 24;
+ if (src == 0)
+ {
+ pixman_fill_sse2 (dst_image->bits.bits, dst_image->bits.rowstride,
+ PIXMAN_FORMAT_BPP (dst_image->bits.format),
+ dest_x, dest_y, width, height, 0);
+ return;
+ }
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ xmm_def = create_mask_2x32_128 (src, src);
+ xmm_src = expand_pixel_32_1x128 (src);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ uint8_t m = *mask++;
+
+ if (m)
+ {
+ *dst = pack_1x64_32 (
+ pix_multiply_1x64 (
+ _mm_movepi64_pi64 (xmm_src), expand_pixel_8_1x64 (m)));
+ }
+ else
+ {
+ *dst = 0;
+ }
+
+ w--;
+ dst++;
+ }
+
+ while (w >= 4)
+ {
+ m = *((uint32_t*)mask);
+
+ if (srca == 0xff && m == 0xffffffff)
+ {
+ save_128_aligned ((__m128i*)dst, xmm_def);
+ }
+ else if (m)
+ {
+ xmm_mask = unpack_32_1x128 (m);
+ xmm_mask = _mm_unpacklo_epi8 (xmm_mask, _mm_setzero_si128 ());
+
+ /* Unpacking */
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ pix_multiply_2x128 (&xmm_src, &xmm_src,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_mask_lo, xmm_mask_hi));
+ }
+ else
+ {
+ save_128_aligned ((__m128i*)dst, _mm_setzero_si128 ());
+ }
+
+ w -= 4;
+ dst += 4;
+ mask += 4;
+ }
+
+ while (w)
+ {
+ uint8_t m = *mask++;
+
+ if (m)
+ {
+ *dst = pack_1x64_32 (
+ pix_multiply_1x64 (
+ _mm_movepi64_pi64 (xmm_src), expand_pixel_8_1x64 (m)));
+ }
+ else
+ {
+ *dst = 0;
+ }
+
+ w--;
+ dst++;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/*-----------------------------------------------------------------------
+ * composite_over_n_8_0565
+ */
+
+static void
+sse2_composite_over_n_8_0565 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src, srca;
+ uint16_t *dst_line, *dst, d;
+ uint8_t *mask_line, *mask;
+ int dst_stride, mask_stride;
+ int32_t w;
+ uint32_t m;
+ __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+
+ __m128i xmm_src, xmm_alpha;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+ __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ srca = src >> 24;
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ xmm_src = expand_pixel_32_1x128 (src);
+ xmm_alpha = expand_alpha_1x128 (xmm_src);
+ mmx_src = _mm_movepi64_pi64 (xmm_src);
+ mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ m = *mask++;
+
+ if (m)
+ {
+ d = *dst;
+ mmx_mask = expand_alpha_rev_1x64 (unpack_32_1x64 (m));
+ mmx_dest = expand565_16_1x64 (d);
+
+ *dst = pack_565_32_16 (
+ pack_1x64_32 (
+ in_over_1x64 (
+ &mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)));
+ }
+
+ w--;
+ dst++;
+ }
+
+ while (w >= 8)
+ {
+ xmm_dst = load_128_aligned ((__m128i*) dst);
+ unpack_565_128_4x128 (xmm_dst,
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3);
+
+ m = *((uint32_t*)mask);
+ mask += 4;
+
+ if (m)
+ {
+ xmm_mask = unpack_32_1x128 (m);
+ xmm_mask = _mm_unpacklo_epi8 (xmm_mask, _mm_setzero_si128 ());
+
+ /* Unpacking */
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ in_over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst0, &xmm_dst1);
+ }
+
+ m = *((uint32_t*)mask);
+ mask += 4;
+
+ if (m)
+ {
+ xmm_mask = unpack_32_1x128 (m);
+ xmm_mask = _mm_unpacklo_epi8 (xmm_mask, _mm_setzero_si128 ());
+
+ /* Unpacking */
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+
+ expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+ in_over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst2, &xmm_dst3);
+ }
+
+ save_128_aligned (
+ (__m128i*)dst, pack_565_4x128_128 (
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3));
+
+ w -= 8;
+ dst += 8;
+ }
+
+ while (w)
+ {
+ m = *mask++;
+
+ if (m)
+ {
+ d = *dst;
+ mmx_mask = expand_alpha_rev_1x64 (unpack_32_1x64 (m));
+ mmx_dest = expand565_16_1x64 (d);
+
+ *dst = pack_565_32_16 (
+ pack_1x64_32 (
+ in_over_1x64 (
+ &mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)));
+ }
+
+ w--;
+ dst++;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* -----------------------------------------------------------------------
+ * composite_over_pixbuf_0565
+ */
+
+static void
+sse2_composite_over_pixbuf_0565 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint16_t *dst_line, *dst, d;
+ uint32_t *src_line, *src, s;
+ int dst_stride, src_stride;
+ int32_t w;
+ uint32_t opaque, zero;
+
+ __m64 ms;
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+#if 0
+ /* FIXME
+ *
+ * I copy the code from MMX one and keep the fixme.
+ * If it's a problem there, probably is a problem here.
+ */
+ assert (src_image->drawable == mask_image->drawable);
+#endif
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ s = *src++;
+ d = *dst;
+
+ ms = unpack_32_1x64 (s);
+
+ *dst++ = pack_565_32_16 (
+ pack_1x64_32 (
+ over_rev_non_pre_1x64 (ms, expand565_16_1x64 (d))));
+ w--;
+ }
+
+ while (w >= 8)
+ {
+ /* First round */
+ xmm_src = load_128_unaligned ((__m128i*)src);
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ opaque = is_opaque (xmm_src);
+ zero = is_zero (xmm_src);
+
+ unpack_565_128_4x128 (xmm_dst,
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3);
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+
+ /* preload next round*/
+ xmm_src = load_128_unaligned ((__m128i*)(src + 4));
+
+ if (opaque)
+ {
+ invert_colors_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_dst0, &xmm_dst1);
+ }
+ else if (!zero)
+ {
+ over_rev_non_pre_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_dst0, &xmm_dst1);
+ }
+
+ /* Second round */
+ opaque = is_opaque (xmm_src);
+ zero = is_zero (xmm_src);
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+
+ if (opaque)
+ {
+ invert_colors_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_dst2, &xmm_dst3);
+ }
+ else if (!zero)
+ {
+ over_rev_non_pre_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_dst2, &xmm_dst3);
+ }
+
+ save_128_aligned (
+ (__m128i*)dst, pack_565_4x128_128 (
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3));
+
+ w -= 8;
+ src += 8;
+ dst += 8;
+ }
+
+ while (w)
+ {
+ s = *src++;
+ d = *dst;
+
+ ms = unpack_32_1x64 (s);
+
+ *dst++ = pack_565_32_16 (
+ pack_1x64_32 (
+ over_rev_non_pre_1x64 (ms, expand565_16_1x64 (d))));
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* -------------------------------------------------------------------------
+ * composite_over_pixbuf_8888
+ */
+
+static void
+sse2_composite_over_pixbuf_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst, d;
+ uint32_t *src_line, *src, s;
+ int dst_stride, src_stride;
+ int32_t w;
+ uint32_t opaque, zero;
+
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst_lo, xmm_dst_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+#if 0
+ /* FIXME
+ *
+ * I copy the code from MMX one and keep the fixme.
+ * If it's a problem there, probably is a problem here.
+ */
+ assert (src_image->drawable == mask_image->drawable);
+#endif
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ s = *src++;
+ d = *dst;
+
+ *dst++ = pack_1x64_32 (
+ over_rev_non_pre_1x64 (
+ unpack_32_1x64 (s), unpack_32_1x64 (d)));
+
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_src_hi = load_128_unaligned ((__m128i*)src);
+
+ opaque = is_opaque (xmm_src_hi);
+ zero = is_zero (xmm_src_hi);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+
+ if (opaque)
+ {
+ invert_colors_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+ else if (!zero)
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ over_rev_non_pre_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+
+ w -= 4;
+ dst += 4;
+ src += 4;
+ }
+
+ while (w)
+ {
+ s = *src++;
+ d = *dst;
+
+ *dst++ = pack_1x64_32 (
+ over_rev_non_pre_1x64 (
+ unpack_32_1x64 (s), unpack_32_1x64 (d)));
+
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* -------------------------------------------------------------------------------------------------
+ * composite_over_n_8888_0565_ca
+ */
+
+static void
+sse2_composite_over_n_8888_0565_ca (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src;
+ uint16_t *dst_line, *dst, d;
+ uint32_t *mask_line, *mask, m;
+ int dst_stride, mask_stride;
+ int w;
+ uint32_t pack_cmp;
+
+ __m128i xmm_src, xmm_alpha;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+ __m128i xmm_dst, xmm_dst0, xmm_dst1, xmm_dst2, xmm_dst3;
+
+ __m64 mmx_src, mmx_alpha, mmx_mask, mmx_dest;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint16_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1);
+
+ xmm_src = expand_pixel_32_1x128 (src);
+ xmm_alpha = expand_alpha_1x128 (xmm_src);
+ mmx_src = _mm_movepi64_pi64 (xmm_src);
+ mmx_alpha = _mm_movepi64_pi64 (xmm_alpha);
+
+ while (height--)
+ {
+ w = width;
+ mask = mask_line;
+ dst = dst_line;
+ mask_line += mask_stride;
+ dst_line += dst_stride;
+
+ while (w && ((unsigned long)dst & 15))
+ {
+ m = *(uint32_t *) mask;
+
+ if (m)
+ {
+ d = *dst;
+ mmx_mask = unpack_32_1x64 (m);
+ mmx_dest = expand565_16_1x64 (d);
+
+ *dst = pack_565_32_16 (
+ pack_1x64_32 (
+ in_over_1x64 (
+ &mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)));
+ }
+
+ w--;
+ dst++;
+ mask++;
+ }
+
+ while (w >= 8)
+ {
+ /* First round */
+ xmm_mask = load_128_unaligned ((__m128i*)mask);
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ pack_cmp = _mm_movemask_epi8 (
+ _mm_cmpeq_epi32 (xmm_mask, _mm_setzero_si128 ()));
+
+ unpack_565_128_4x128 (xmm_dst,
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3);
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+
+ /* preload next round */
+ xmm_mask = load_128_unaligned ((__m128i*)(mask + 4));
+
+ /* preload next round */
+ if (pack_cmp != 0xffff)
+ {
+ in_over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst0, &xmm_dst1);
+ }
+
+ /* Second round */
+ pack_cmp = _mm_movemask_epi8 (
+ _mm_cmpeq_epi32 (xmm_mask, _mm_setzero_si128 ()));
+
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+
+ if (pack_cmp != 0xffff)
+ {
+ in_over_2x128 (&xmm_src, &xmm_src,
+ &xmm_alpha, &xmm_alpha,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst2, &xmm_dst3);
+ }
+
+ save_128_aligned (
+ (__m128i*)dst, pack_565_4x128_128 (
+ &xmm_dst0, &xmm_dst1, &xmm_dst2, &xmm_dst3));
+
+ w -= 8;
+ dst += 8;
+ mask += 8;
+ }
+
+ while (w)
+ {
+ m = *(uint32_t *) mask;
+
+ if (m)
+ {
+ d = *dst;
+ mmx_mask = unpack_32_1x64 (m);
+ mmx_dest = expand565_16_1x64 (d);
+
+ *dst = pack_565_32_16 (
+ pack_1x64_32 (
+ in_over_1x64 (
+ &mmx_src, &mmx_alpha, &mmx_mask, &mmx_dest)));
+ }
+
+ w--;
+ dst++;
+ mask++;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* -----------------------------------------------------------------------
+ * composite_in_n_8_8
+ */
+
+static void
+sse2_composite_in_n_8_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ uint8_t *mask_line, *mask;
+ int dst_stride, mask_stride;
+ uint32_t d, m;
+ uint32_t src;
+ uint8_t sa;
+ int32_t w;
+
+ __m128i xmm_alpha;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ sa = src >> 24;
+
+ xmm_alpha = expand_alpha_1x128 (expand_pixel_32_1x128 (src));
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w && ((unsigned long)dst & 15))
+ {
+ m = (uint32_t) *mask++;
+ d = (uint32_t) *dst;
+
+ *dst++ = (uint8_t) pack_1x64_32 (
+ pix_multiply_1x64 (
+ pix_multiply_1x64 (_mm_movepi64_pi64 (xmm_alpha),
+ unpack_32_1x64 (m)),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+
+ while (w >= 16)
+ {
+ xmm_mask = load_128_unaligned ((__m128i*)mask);
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ pix_multiply_2x128 (&xmm_alpha, &xmm_alpha,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ pix_multiply_2x128 (&xmm_mask_lo, &xmm_mask_hi,
+ &xmm_dst_lo, &xmm_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ mask += 16;
+ dst += 16;
+ w -= 16;
+ }
+
+ while (w)
+ {
+ m = (uint32_t) *mask++;
+ d = (uint32_t) *dst;
+
+ *dst++ = (uint8_t) pack_1x64_32 (
+ pix_multiply_1x64 (
+ pix_multiply_1x64 (
+ _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* -----------------------------------------------------------------------
+ * composite_in_n_8
+ */
+
+static void
+sse2_composite_in_n_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ int dst_stride;
+ uint32_t d;
+ uint32_t src;
+ int32_t w;
+
+ __m128i xmm_alpha;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ xmm_alpha = expand_alpha_1x128 (expand_pixel_32_1x128 (src));
+
+ src = src >> 24;
+
+ if (src == 0xff)
+ return;
+
+ if (src == 0x00)
+ {
+ pixman_fill (dst_image->bits.bits, dst_image->bits.rowstride,
+ 8, dest_x, dest_y, width, height, src);
+
+ return;
+ }
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ w = width;
+
+ while (w && ((unsigned long)dst & 15))
+ {
+ d = (uint32_t) *dst;
+
+ *dst++ = (uint8_t) pack_1x64_32 (
+ pix_multiply_1x64 (
+ _mm_movepi64_pi64 (xmm_alpha),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+
+ while (w >= 16)
+ {
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ pix_multiply_2x128 (&xmm_alpha, &xmm_alpha,
+ &xmm_dst_lo, &xmm_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ dst += 16;
+ w -= 16;
+ }
+
+ while (w)
+ {
+ d = (uint32_t) *dst;
+
+ *dst++ = (uint8_t) pack_1x64_32 (
+ pix_multiply_1x64 (
+ _mm_movepi64_pi64 (xmm_alpha),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* ---------------------------------------------------------------------------
+ * composite_in_8_8
+ */
+
+static void
+sse2_composite_in_8_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ uint8_t *src_line, *src;
+ int src_stride, dst_stride;
+ int32_t w;
+ uint32_t s, d;
+
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint8_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+ w = width;
+
+ while (w && ((unsigned long)dst & 15))
+ {
+ s = (uint32_t) *src++;
+ d = (uint32_t) *dst;
+
+ *dst++ = (uint8_t) pack_1x64_32 (
+ pix_multiply_1x64 (
+ unpack_32_1x64 (s), unpack_32_1x64 (d)));
+ w--;
+ }
+
+ while (w >= 16)
+ {
+ xmm_src = load_128_unaligned ((__m128i*)src);
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ pix_multiply_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_dst_lo, &xmm_dst_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ src += 16;
+ dst += 16;
+ w -= 16;
+ }
+
+ while (w)
+ {
+ s = (uint32_t) *src++;
+ d = (uint32_t) *dst;
+
+ *dst++ = (uint8_t) pack_1x64_32 (
+ pix_multiply_1x64 (unpack_32_1x64 (s), unpack_32_1x64 (d)));
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* -------------------------------------------------------------------------
+ * composite_add_n_8_8
+ */
+
+static void
+sse2_composite_add_n_8_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ uint8_t *mask_line, *mask;
+ int dst_stride, mask_stride;
+ int32_t w;
+ uint32_t src;
+ uint8_t sa;
+ uint32_t m, d;
+
+ __m128i xmm_alpha;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ sa = src >> 24;
+
+ xmm_alpha = expand_alpha_1x128 (expand_pixel_32_1x128 (src));
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+ w = width;
+
+ while (w && ((unsigned long)dst & 15))
+ {
+ m = (uint32_t) *mask++;
+ d = (uint32_t) *dst;
+
+ *dst++ = (uint8_t) pack_1x64_32 (
+ _mm_adds_pu16 (
+ pix_multiply_1x64 (
+ _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)),
+ unpack_32_1x64 (d)));
+ w--;
+ }
+
+ while (w >= 16)
+ {
+ xmm_mask = load_128_unaligned ((__m128i*)mask);
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ pix_multiply_2x128 (&xmm_alpha, &xmm_alpha,
+ &xmm_mask_lo, &xmm_mask_hi,
+ &xmm_mask_lo, &xmm_mask_hi);
+
+ xmm_dst_lo = _mm_adds_epu16 (xmm_mask_lo, xmm_dst_lo);
+ xmm_dst_hi = _mm_adds_epu16 (xmm_mask_hi, xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+
+ mask += 16;
+ dst += 16;
+ w -= 16;
+ }
+
+ while (w)
+ {
+ m = (uint32_t) *mask++;
+ d = (uint32_t) *dst;
+
+ *dst++ = (uint8_t) pack_1x64_32 (
+ _mm_adds_pu16 (
+ pix_multiply_1x64 (
+ _mm_movepi64_pi64 (xmm_alpha), unpack_32_1x64 (m)),
+ unpack_32_1x64 (d)));
+
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* -------------------------------------------------------------------------
+ * composite_add_n_8_8
+ */
+
+static void
+sse2_composite_add_n_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ int dst_stride;
+ int32_t w;
+ uint32_t src;
+
+ __m128i xmm_src;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ src >>= 24;
+
+ if (src == 0x00)
+ return;
+
+ if (src == 0xff)
+ {
+ pixman_fill (dst_image->bits.bits, dst_image->bits.rowstride,
+ 8, dest_x, dest_y, width, height, 0xff);
+
+ return;
+ }
+
+ src = (src << 24) | (src << 16) | (src << 8) | src;
+ xmm_src = _mm_set_epi32 (src, src, src, src);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ w = width;
+
+ while (w && ((unsigned long)dst & 15))
+ {
+ *dst = (uint8_t)_mm_cvtsi64_si32 (
+ _mm_adds_pu8 (
+ _mm_movepi64_pi64 (xmm_src),
+ _mm_cvtsi32_si64 (*dst)));
+
+ w--;
+ dst++;
+ }
+
+ while (w >= 16)
+ {
+ save_128_aligned (
+ (__m128i*)dst, _mm_adds_epu8 (xmm_src, load_128_aligned ((__m128i*)dst)));
+
+ dst += 16;
+ w -= 16;
+ }
+
+ while (w)
+ {
+ *dst = (uint8_t)_mm_cvtsi64_si32 (
+ _mm_adds_pu8 (
+ _mm_movepi64_pi64 (xmm_src),
+ _mm_cvtsi32_si64 (*dst)));
+
+ w--;
+ dst++;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* ----------------------------------------------------------------------
+ * composite_add_8_8
+ */
+
+static void
+sse2_composite_add_8_8 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint8_t *dst_line, *dst;
+ uint8_t *src_line, *src;
+ int dst_stride, src_stride;
+ int32_t w;
+ uint16_t t;
+
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint8_t, src_stride, src_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint8_t, dst_stride, dst_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ src = src_line;
+
+ dst_line += dst_stride;
+ src_line += src_stride;
+ w = width;
+
+ /* Small head */
+ while (w && (unsigned long)dst & 3)
+ {
+ t = (*dst) + (*src++);
+ *dst++ = t | (0 - (t >> 8));
+ w--;
+ }
+
+ core_combine_add_u_sse2 ((uint32_t*)dst, (uint32_t*)src, NULL, w >> 2);
+
+ /* Small tail */
+ dst += w & 0xfffc;
+ src += w & 0xfffc;
+
+ w &= 3;
+
+ while (w)
+ {
+ t = (*dst) + (*src++);
+ *dst++ = t | (0 - (t >> 8));
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* ---------------------------------------------------------------------
+ * composite_add_8888_8888
+ */
+static void
+sse2_composite_add_8888_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *dst_line, *dst;
+ uint32_t *src_line, *src;
+ int dst_stride, src_stride;
+
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+
+ while (height--)
+ {
+ dst = dst_line;
+ dst_line += dst_stride;
+ src = src_line;
+ src_line += src_stride;
+
+ core_combine_add_u_sse2 (dst, src, NULL, width);
+ }
+
+ _mm_empty ();
+}
+
+/* -------------------------------------------------------------------------------------------------
+ * sse2_composite_copy_area
+ */
+
+static pixman_bool_t
+pixman_blt_sse2 (uint32_t *src_bits,
+ uint32_t *dst_bits,
+ int src_stride,
+ int dst_stride,
+ int src_bpp,
+ int dst_bpp,
+ int src_x,
+ int src_y,
+ int dst_x,
+ int dst_y,
+ int width,
+ int height)
+{
+ uint8_t * src_bytes;
+ uint8_t * dst_bytes;
+ int byte_width;
+
+ if (src_bpp != dst_bpp)
+ return FALSE;
+
+ if (src_bpp == 16)
+ {
+ src_stride = src_stride * (int) sizeof (uint32_t) / 2;
+ dst_stride = dst_stride * (int) sizeof (uint32_t) / 2;
+ src_bytes =(uint8_t *)(((uint16_t *)src_bits) + src_stride * (src_y) + (src_x));
+ dst_bytes = (uint8_t *)(((uint16_t *)dst_bits) + dst_stride * (dst_y) + (dst_x));
+ byte_width = 2 * width;
+ src_stride *= 2;
+ dst_stride *= 2;
+ }
+ else if (src_bpp == 32)
+ {
+ src_stride = src_stride * (int) sizeof (uint32_t) / 4;
+ dst_stride = dst_stride * (int) sizeof (uint32_t) / 4;
+ src_bytes = (uint8_t *)(((uint32_t *)src_bits) + src_stride * (src_y) + (src_x));
+ dst_bytes = (uint8_t *)(((uint32_t *)dst_bits) + dst_stride * (dst_y) + (dst_x));
+ byte_width = 4 * width;
+ src_stride *= 4;
+ dst_stride *= 4;
+ }
+ else
+ {
+ return FALSE;
+ }
+
+ while (height--)
+ {
+ int w;
+ uint8_t *s = src_bytes;
+ uint8_t *d = dst_bytes;
+ src_bytes += src_stride;
+ dst_bytes += dst_stride;
+ w = byte_width;
+
+ while (w >= 2 && ((unsigned long)d & 3))
+ {
+ *(uint16_t *)d = *(uint16_t *)s;
+ w -= 2;
+ s += 2;
+ d += 2;
+ }
+
+ while (w >= 4 && ((unsigned long)d & 15))
+ {
+ *(uint32_t *)d = *(uint32_t *)s;
+
+ w -= 4;
+ s += 4;
+ d += 4;
+ }
+
+ while (w >= 64)
+ {
+ __m128i xmm0, xmm1, xmm2, xmm3;
+
+ xmm0 = load_128_unaligned ((__m128i*)(s));
+ xmm1 = load_128_unaligned ((__m128i*)(s + 16));
+ xmm2 = load_128_unaligned ((__m128i*)(s + 32));
+ xmm3 = load_128_unaligned ((__m128i*)(s + 48));
+
+ save_128_aligned ((__m128i*)(d), xmm0);
+ save_128_aligned ((__m128i*)(d + 16), xmm1);
+ save_128_aligned ((__m128i*)(d + 32), xmm2);
+ save_128_aligned ((__m128i*)(d + 48), xmm3);
+
+ s += 64;
+ d += 64;
+ w -= 64;
+ }
+
+ while (w >= 16)
+ {
+ save_128_aligned ((__m128i*)d, load_128_unaligned ((__m128i*)s) );
+
+ w -= 16;
+ d += 16;
+ s += 16;
+ }
+
+ while (w >= 4)
+ {
+ *(uint32_t *)d = *(uint32_t *)s;
+
+ w -= 4;
+ s += 4;
+ d += 4;
+ }
+
+ if (w >= 2)
+ {
+ *(uint16_t *)d = *(uint16_t *)s;
+ w -= 2;
+ s += 2;
+ d += 2;
+ }
+ }
+
+ _mm_empty ();
+
+ return TRUE;
+}
+
+static void
+sse2_composite_copy_area (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ pixman_blt_sse2 (src_image->bits.bits,
+ dst_image->bits.bits,
+ src_image->bits.rowstride,
+ dst_image->bits.rowstride,
+ PIXMAN_FORMAT_BPP (src_image->bits.format),
+ PIXMAN_FORMAT_BPP (dst_image->bits.format),
+ src_x, src_y, dest_x, dest_y, width, height);
+}
+
+static void
+sse2_composite_over_x888_8_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *src, *src_line, s;
+ uint32_t *dst, *dst_line, d;
+ uint8_t *mask, *mask_line;
+ uint32_t m;
+ int src_stride, mask_stride, dst_stride;
+ int32_t w;
+ __m64 ms;
+
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ src = src_line;
+ src_line += src_stride;
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ s = 0xff000000 | *src++;
+ m = (uint32_t) *mask++;
+ d = *dst;
+ ms = unpack_32_1x64 (s);
+
+ if (m != 0xff)
+ {
+ __m64 ma = expand_alpha_rev_1x64 (unpack_32_1x64 (m));
+ __m64 md = unpack_32_1x64 (d);
+
+ ms = in_over_1x64 (&ms, &mask_x00ff, &ma, &md);
+ }
+
+ *dst++ = pack_1x64_32 (ms);
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ m = *(uint32_t*) mask;
+ xmm_src = _mm_or_si128 (load_128_unaligned ((__m128i*)src), mask_ff000000);
+
+ if (m == 0xffffffff)
+ {
+ save_128_aligned ((__m128i*)dst, xmm_src);
+ }
+ else
+ {
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ xmm_mask = _mm_unpacklo_epi16 (unpack_32_1x128 (m), _mm_setzero_si128());
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ in_over_2x128 (&xmm_src_lo, &xmm_src_hi, &mask_00ff, &mask_00ff, &xmm_mask_lo, &xmm_mask_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned ((__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+
+ src += 4;
+ dst += 4;
+ mask += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ m = (uint32_t) *mask++;
+
+ if (m)
+ {
+ s = 0xff000000 | *src;
+
+ if (m == 0xff)
+ {
+ *dst = s;
+ }
+ else
+ {
+ __m64 ma, md, ms;
+
+ d = *dst;
+
+ ma = expand_alpha_rev_1x64 (unpack_32_1x64 (m));
+ md = unpack_32_1x64 (d);
+ ms = unpack_32_1x64 (s);
+
+ *dst = pack_1x64_32 (in_over_1x64 (&ms, &mask_x00ff, &ma, &md));
+ }
+
+ }
+
+ src++;
+ dst++;
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+static void
+sse2_composite_over_8888_8_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *src, *src_line, s;
+ uint32_t *dst, *dst_line, d;
+ uint8_t *mask, *mask_line;
+ uint32_t m;
+ int src_stride, mask_stride, dst_stride;
+ int32_t w;
+
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi, xmm_srca_lo, xmm_srca_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint8_t, mask_stride, mask_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ src = src_line;
+ src_line += src_stride;
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ uint32_t sa;
+
+ s = *src++;
+ m = (uint32_t) *mask++;
+ d = *dst;
+
+ sa = s >> 24;
+
+ if (m)
+ {
+ if (sa == 0xff && m == 0xff)
+ {
+ *dst = s;
+ }
+ else
+ {
+ __m64 ms, md, ma, msa;
+
+ ma = expand_alpha_rev_1x64 (load_32_1x64 (m));
+ ms = unpack_32_1x64 (s);
+ md = unpack_32_1x64 (d);
+
+ msa = expand_alpha_rev_1x64 (load_32_1x64 (sa));
+
+ *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md));
+ }
+ }
+
+ dst++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ m = *(uint32_t *) mask;
+
+ if (m)
+ {
+ xmm_src = load_128_unaligned ((__m128i*)src);
+
+ if (m == 0xffffffff && is_opaque (xmm_src))
+ {
+ save_128_aligned ((__m128i *)dst, xmm_src);
+ }
+ else
+ {
+ xmm_dst = load_128_aligned ((__m128i *)dst);
+
+ xmm_mask = _mm_unpacklo_epi16 (unpack_32_1x128 (m), _mm_setzero_si128());
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, &xmm_srca_lo, &xmm_srca_hi);
+ expand_alpha_rev_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ in_over_2x128 (&xmm_src_lo, &xmm_src_hi, &xmm_srca_lo, &xmm_srca_hi,
+ &xmm_mask_lo, &xmm_mask_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned ((__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+ }
+
+ src += 4;
+ dst += 4;
+ mask += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ uint32_t sa;
+
+ s = *src++;
+ m = (uint32_t) *mask++;
+ d = *dst;
+
+ sa = s >> 24;
+
+ if (m)
+ {
+ if (sa == 0xff && m == 0xff)
+ {
+ *dst = s;
+ }
+ else
+ {
+ __m64 ms, md, ma, msa;
+
+ ma = expand_alpha_rev_1x64 (load_32_1x64 (m));
+ ms = unpack_32_1x64 (s);
+ md = unpack_32_1x64 (d);
+
+ msa = expand_alpha_rev_1x64 (load_32_1x64 (sa));
+
+ *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md));
+ }
+ }
+
+ dst++;
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+static void
+sse2_composite_over_reverse_n_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t src;
+ uint32_t *dst_line, *dst;
+ __m128i xmm_src;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_dsta_hi, xmm_dsta_lo;
+ int dst_stride;
+ int32_t w;
+
+ src = _pixman_image_get_solid (imp, src_image, dst_image->bits.format);
+
+ if (src == 0)
+ return;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+
+ xmm_src = expand_pixel_32_1x128 (src);
+
+ while (height--)
+ {
+ dst = dst_line;
+
+ dst_line += dst_stride;
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ __m64 vd;
+
+ vd = unpack_32_1x64 (*dst);
+
+ *dst = pack_1x64_32 (over_1x64 (vd, expand_alpha_1x64 (vd),
+ _mm_movepi64_pi64 (xmm_src)));
+ w--;
+ dst++;
+ }
+
+ while (w >= 4)
+ {
+ __m128i tmp_lo, tmp_hi;
+
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+ expand_alpha_2x128 (xmm_dst_lo, xmm_dst_hi, &xmm_dsta_lo, &xmm_dsta_hi);
+
+ tmp_lo = xmm_src;
+ tmp_hi = xmm_src;
+
+ over_2x128 (&xmm_dst_lo, &xmm_dst_hi,
+ &xmm_dsta_lo, &xmm_dsta_hi,
+ &tmp_lo, &tmp_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (tmp_lo, tmp_hi));
+
+ w -= 4;
+ dst += 4;
+ }
+
+ while (w)
+ {
+ __m64 vd;
+
+ vd = unpack_32_1x64 (*dst);
+
+ *dst = pack_1x64_32 (over_1x64 (vd, expand_alpha_1x64 (vd),
+ _mm_movepi64_pi64 (xmm_src)));
+ w--;
+ dst++;
+ }
+
+ }
+
+ _mm_empty ();
+}
+
+static void
+sse2_composite_over_8888_8888_8888 (pixman_implementation_t *imp,
+ pixman_op_t op,
+ pixman_image_t * src_image,
+ pixman_image_t * mask_image,
+ pixman_image_t * dst_image,
+ int32_t src_x,
+ int32_t src_y,
+ int32_t mask_x,
+ int32_t mask_y,
+ int32_t dest_x,
+ int32_t dest_y,
+ int32_t width,
+ int32_t height)
+{
+ uint32_t *src, *src_line, s;
+ uint32_t *dst, *dst_line, d;
+ uint32_t *mask, *mask_line;
+ uint32_t m;
+ int src_stride, mask_stride, dst_stride;
+ int32_t w;
+
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi, xmm_srca_lo, xmm_srca_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_mask, xmm_mask_lo, xmm_mask_hi;
+
+ PIXMAN_IMAGE_GET_LINE (
+ dst_image, dest_x, dest_y, uint32_t, dst_stride, dst_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ mask_image, mask_x, mask_y, uint32_t, mask_stride, mask_line, 1);
+ PIXMAN_IMAGE_GET_LINE (
+ src_image, src_x, src_y, uint32_t, src_stride, src_line, 1);
+
+ while (height--)
+ {
+ src = src_line;
+ src_line += src_stride;
+ dst = dst_line;
+ dst_line += dst_stride;
+ mask = mask_line;
+ mask_line += mask_stride;
+
+ w = width;
+
+ while (w && (unsigned long)dst & 15)
+ {
+ uint32_t sa;
+
+ s = *src++;
+ m = (*mask++) >> 24;
+ d = *dst;
+
+ sa = s >> 24;
+
+ if (m)
+ {
+ if (sa == 0xff && m == 0xff)
+ {
+ *dst = s;
+ }
+ else
+ {
+ __m64 ms, md, ma, msa;
+
+ ma = expand_alpha_rev_1x64 (load_32_1x64 (m));
+ ms = unpack_32_1x64 (s);
+ md = unpack_32_1x64 (d);
+
+ msa = expand_alpha_rev_1x64 (load_32_1x64 (sa));
+
+ *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md));
+ }
+ }
+
+ dst++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ xmm_mask = load_128_unaligned ((__m128i*)mask);
+
+ if (!is_transparent (xmm_mask))
+ {
+ xmm_src = load_128_unaligned ((__m128i*)src);
+
+ if (is_opaque (xmm_mask) && is_opaque (xmm_src))
+ {
+ save_128_aligned ((__m128i *)dst, xmm_src);
+ }
+ else
+ {
+ xmm_dst = load_128_aligned ((__m128i *)dst);
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_mask, &xmm_mask_lo, &xmm_mask_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi, &xmm_srca_lo, &xmm_srca_hi);
+ expand_alpha_2x128 (xmm_mask_lo, xmm_mask_hi, &xmm_mask_lo, &xmm_mask_hi);
+
+ in_over_2x128 (&xmm_src_lo, &xmm_src_hi, &xmm_srca_lo, &xmm_srca_hi,
+ &xmm_mask_lo, &xmm_mask_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned ((__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+ }
+
+ src += 4;
+ dst += 4;
+ mask += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ uint32_t sa;
+
+ s = *src++;
+ m = (*mask++) >> 24;
+ d = *dst;
+
+ sa = s >> 24;
+
+ if (m)
+ {
+ if (sa == 0xff && m == 0xff)
+ {
+ *dst = s;
+ }
+ else
+ {
+ __m64 ms, md, ma, msa;
+
+ ma = expand_alpha_rev_1x64 (load_32_1x64 (m));
+ ms = unpack_32_1x64 (s);
+ md = unpack_32_1x64 (d);
+
+ msa = expand_alpha_rev_1x64 (load_32_1x64 (sa));
+
+ *dst = pack_1x64_32 (in_over_1x64 (&ms, &msa, &ma, &md));
+ }
+ }
+
+ dst++;
+ w--;
+ }
+ }
+
+ _mm_empty ();
+}
+
+/* A variant of 'core_combine_over_u_sse2' with minor tweaks */
+static force_inline void
+scaled_nearest_scanline_sse2_8888_8888_OVER (uint32_t* pd,
+ const uint32_t* ps,
+ int32_t w,
+ pixman_fixed_t vx,
+ pixman_fixed_t unit_x,
+ pixman_fixed_t max_vx,
+ pixman_bool_t fully_transparent_src)
+{
+ uint32_t s, d;
+ const uint32_t* pm = NULL;
+
+ __m128i xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_src_lo, xmm_src_hi;
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+
+ if (fully_transparent_src)
+ return;
+
+ /* Align dst on a 16-byte boundary */
+ while (w && ((unsigned long)pd & 15))
+ {
+ d = *pd;
+ s = combine1 (ps + (vx >> 16), pm);
+ vx += unit_x;
+
+ *pd++ = core_combine_over_u_pixel_sse2 (s, d);
+ if (pm)
+ pm++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ __m128i tmp;
+ uint32_t tmp1, tmp2, tmp3, tmp4;
+
+ tmp1 = ps[vx >> 16];
+ vx += unit_x;
+ tmp2 = ps[vx >> 16];
+ vx += unit_x;
+ tmp3 = ps[vx >> 16];
+ vx += unit_x;
+ tmp4 = ps[vx >> 16];
+ vx += unit_x;
+
+ tmp = _mm_set_epi32 (tmp4, tmp3, tmp2, tmp1);
+
+ xmm_src_hi = combine4 ((__m128i*)&tmp, (__m128i*)pm);
+
+ if (is_opaque (xmm_src_hi))
+ {
+ save_128_aligned ((__m128i*)pd, xmm_src_hi);
+ }
+ else if (!is_zero (xmm_src_hi))
+ {
+ xmm_dst_hi = load_128_aligned ((__m128i*) pd);
+
+ unpack_128_2x128 (xmm_src_hi, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst_hi, &xmm_dst_lo, &xmm_dst_hi);
+
+ expand_alpha_2x128 (
+ xmm_src_lo, xmm_src_hi, &xmm_alpha_lo, &xmm_alpha_hi);
+
+ over_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ /* rebuid the 4 pixel data and save*/
+ save_128_aligned ((__m128i*)pd,
+ pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+
+ w -= 4;
+ pd += 4;
+ if (pm)
+ pm += 4;
+ }
+
+ while (w)
+ {
+ d = *pd;
+ s = combine1 (ps + (vx >> 16), pm);
+ vx += unit_x;
+
+ *pd++ = core_combine_over_u_pixel_sse2 (s, d);
+ if (pm)
+ pm++;
+
+ w--;
+ }
+ _mm_empty ();
+}
+
+FAST_NEAREST_MAINLOOP (sse2_8888_8888_cover_OVER,
+ scaled_nearest_scanline_sse2_8888_8888_OVER,
+ uint32_t, uint32_t, COVER)
+FAST_NEAREST_MAINLOOP (sse2_8888_8888_none_OVER,
+ scaled_nearest_scanline_sse2_8888_8888_OVER,
+ uint32_t, uint32_t, NONE)
+FAST_NEAREST_MAINLOOP (sse2_8888_8888_pad_OVER,
+ scaled_nearest_scanline_sse2_8888_8888_OVER,
+ uint32_t, uint32_t, PAD)
+
+static force_inline void
+scaled_nearest_scanline_sse2_8888_n_8888_OVER (const uint32_t * mask,
+ uint32_t * dst,
+ const uint32_t * src,
+ int32_t w,
+ pixman_fixed_t vx,
+ pixman_fixed_t unit_x,
+ pixman_fixed_t max_vx,
+ pixman_bool_t zero_src)
+{
+ __m128i xmm_mask;
+ __m128i xmm_src, xmm_src_lo, xmm_src_hi;
+ __m128i xmm_dst, xmm_dst_lo, xmm_dst_hi;
+ __m128i xmm_alpha_lo, xmm_alpha_hi;
+
+ if (zero_src || (*mask >> 24) == 0)
+ return;
+
+ xmm_mask = create_mask_16_128 (*mask >> 24);
+
+ while (w && (unsigned long)dst & 15)
+ {
+ uint32_t s = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+
+ if (s)
+ {
+ uint32_t d = *dst;
+
+ __m64 ms = unpack_32_1x64 (s);
+ __m64 alpha = expand_alpha_1x64 (ms);
+ __m64 dest = _mm_movepi64_pi64 (xmm_mask);
+ __m64 alpha_dst = unpack_32_1x64 (d);
+
+ *dst = pack_1x64_32 (
+ in_over_1x64 (&ms, &alpha, &dest, &alpha_dst));
+ }
+ dst++;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ uint32_t tmp1, tmp2, tmp3, tmp4;
+
+ tmp1 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp2 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp3 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+ tmp4 = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+
+ xmm_src = _mm_set_epi32 (tmp4, tmp3, tmp2, tmp1);
+
+ if (!is_zero (xmm_src))
+ {
+ xmm_dst = load_128_aligned ((__m128i*)dst);
+
+ unpack_128_2x128 (xmm_src, &xmm_src_lo, &xmm_src_hi);
+ unpack_128_2x128 (xmm_dst, &xmm_dst_lo, &xmm_dst_hi);
+ expand_alpha_2x128 (xmm_src_lo, xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi);
+
+ in_over_2x128 (&xmm_src_lo, &xmm_src_hi,
+ &xmm_alpha_lo, &xmm_alpha_hi,
+ &xmm_mask, &xmm_mask,
+ &xmm_dst_lo, &xmm_dst_hi);
+
+ save_128_aligned (
+ (__m128i*)dst, pack_2x128_128 (xmm_dst_lo, xmm_dst_hi));
+ }
+
+ dst += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ uint32_t s = src[pixman_fixed_to_int (vx)];
+ vx += unit_x;
+
+ if (s)
+ {
+ uint32_t d = *dst;
+
+ __m64 ms = unpack_32_1x64 (s);
+ __m64 alpha = expand_alpha_1x64 (ms);
+ __m64 mask = _mm_movepi64_pi64 (xmm_mask);
+ __m64 dest = unpack_32_1x64 (d);
+
+ *dst = pack_1x64_32 (
+ in_over_1x64 (&ms, &alpha, &mask, &dest));
+ }
+
+ dst++;
+ w--;
+ }
+
+ _mm_empty ();
+}
+
+FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_cover_OVER,
+ scaled_nearest_scanline_sse2_8888_n_8888_OVER,
+ uint32_t, uint32_t, uint32_t, COVER, TRUE, TRUE)
+FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_pad_OVER,
+ scaled_nearest_scanline_sse2_8888_n_8888_OVER,
+ uint32_t, uint32_t, uint32_t, PAD, TRUE, TRUE)
+FAST_NEAREST_MAINLOOP_COMMON (sse2_8888_n_8888_none_OVER,
+ scaled_nearest_scanline_sse2_8888_n_8888_OVER,
+ uint32_t, uint32_t, uint32_t, NONE, TRUE, TRUE)
+
+static const pixman_fast_path_t sse2_fast_paths[] =
+{
+ /* PIXMAN_OP_OVER */
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, r5g6b5, sse2_composite_over_n_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, b5g6r5, sse2_composite_over_n_8_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, null, a8r8g8b8, sse2_composite_over_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, null, x8r8g8b8, sse2_composite_over_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, null, r5g6b5, sse2_composite_over_n_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, a8r8g8b8, sse2_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, x8r8g8b8, sse2_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, a8b8g8r8, sse2_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, x8b8g8r8, sse2_composite_over_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, null, r5g6b5, sse2_composite_over_8888_0565),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, null, b5g6r5, sse2_composite_over_8888_0565),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8r8g8b8, sse2_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8r8g8b8, sse2_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, a8b8g8r8, sse2_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, solid, a8, x8b8g8r8, sse2_composite_over_n_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, a8r8g8b8, sse2_composite_over_8888_8888_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, x8r8g8b8, sse2_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, a8, a8r8g8b8, sse2_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, x8b8g8r8, sse2_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, a8, a8b8g8r8, sse2_composite_over_8888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, a8, x8r8g8b8, sse2_composite_over_x888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, a8, a8r8g8b8, sse2_composite_over_x888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, a8, x8b8g8r8, sse2_composite_over_x888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, a8, a8b8g8r8, sse2_composite_over_x888_8_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, solid, a8r8g8b8, sse2_composite_over_x888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, solid, x8r8g8b8, sse2_composite_over_x888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, solid, a8b8g8r8, sse2_composite_over_x888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, solid, x8b8g8r8, sse2_composite_over_x888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, a8r8g8b8, sse2_composite_over_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8r8g8b8, solid, x8r8g8b8, sse2_composite_over_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, solid, a8b8g8r8, sse2_composite_over_8888_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER, a8b8g8r8, solid, x8b8g8r8, sse2_composite_over_8888_n_8888),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, a8r8g8b8, sse2_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, x8r8g8b8, sse2_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, a8b8g8r8, sse2_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, x8b8g8r8, sse2_composite_over_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8r8g8b8, r5g6b5, sse2_composite_over_n_8888_0565_ca),
+ PIXMAN_STD_FAST_PATH_CA (OVER, solid, a8b8g8r8, b5g6r5, sse2_composite_over_n_8888_0565_ca),
+ PIXMAN_STD_FAST_PATH (OVER, pixbuf, pixbuf, a8r8g8b8, sse2_composite_over_pixbuf_8888),
+ PIXMAN_STD_FAST_PATH (OVER, pixbuf, pixbuf, x8r8g8b8, sse2_composite_over_pixbuf_8888),
+ PIXMAN_STD_FAST_PATH (OVER, rpixbuf, rpixbuf, a8b8g8r8, sse2_composite_over_pixbuf_8888),
+ PIXMAN_STD_FAST_PATH (OVER, rpixbuf, rpixbuf, x8b8g8r8, sse2_composite_over_pixbuf_8888),
+ PIXMAN_STD_FAST_PATH (OVER, pixbuf, pixbuf, r5g6b5, sse2_composite_over_pixbuf_0565),
+ PIXMAN_STD_FAST_PATH (OVER, rpixbuf, rpixbuf, b5g6r5, sse2_composite_over_pixbuf_0565),
+ PIXMAN_STD_FAST_PATH (OVER, x8r8g8b8, null, x8r8g8b8, sse2_composite_copy_area),
+ PIXMAN_STD_FAST_PATH (OVER, x8b8g8r8, null, x8b8g8r8, sse2_composite_copy_area),
+
+ /* PIXMAN_OP_OVER_REVERSE */
+ PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8r8g8b8, sse2_composite_over_reverse_n_8888),
+ PIXMAN_STD_FAST_PATH (OVER_REVERSE, solid, null, a8b8g8r8, sse2_composite_over_reverse_n_8888),
+
+ /* PIXMAN_OP_ADD */
+ PIXMAN_STD_FAST_PATH_CA (ADD, solid, a8r8g8b8, a8r8g8b8, sse2_composite_add_n_8888_8888_ca),
+ PIXMAN_STD_FAST_PATH (ADD, a8, null, a8, sse2_composite_add_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, a8r8g8b8, null, a8r8g8b8, sse2_composite_add_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, a8b8g8r8, null, a8b8g8r8, sse2_composite_add_8888_8888),
+ PIXMAN_STD_FAST_PATH (ADD, solid, a8, a8, sse2_composite_add_n_8_8),
+ PIXMAN_STD_FAST_PATH (ADD, solid, null, a8, sse2_composite_add_n_8),
+
+ /* PIXMAN_OP_SRC */
+ PIXMAN_STD_FAST_PATH (SRC, solid, a8, a8r8g8b8, sse2_composite_src_n_8_8888),
+ PIXMAN_STD_FAST_PATH (SRC, solid, a8, x8r8g8b8, sse2_composite_src_n_8_8888),
+ PIXMAN_STD_FAST_PATH (SRC, solid, a8, a8b8g8r8, sse2_composite_src_n_8_8888),
+ PIXMAN_STD_FAST_PATH (SRC, solid, a8, x8b8g8r8, sse2_composite_src_n_8_8888),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, a8r8g8b8, sse2_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, a8b8g8r8, sse2_composite_src_x888_8888),
+ PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, a8r8g8b8, sse2_composite_copy_area),
+ PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, a8b8g8r8, sse2_composite_copy_area),
+ PIXMAN_STD_FAST_PATH (SRC, a8r8g8b8, null, x8r8g8b8, sse2_composite_copy_area),
+ PIXMAN_STD_FAST_PATH (SRC, a8b8g8r8, null, x8b8g8r8, sse2_composite_copy_area),
+ PIXMAN_STD_FAST_PATH (SRC, x8r8g8b8, null, x8r8g8b8, sse2_composite_copy_area),
+ PIXMAN_STD_FAST_PATH (SRC, x8b8g8r8, null, x8b8g8r8, sse2_composite_copy_area),
+ PIXMAN_STD_FAST_PATH (SRC, r5g6b5, null, r5g6b5, sse2_composite_copy_area),
+ PIXMAN_STD_FAST_PATH (SRC, b5g6r5, null, b5g6r5, sse2_composite_copy_area),
+
+ /* PIXMAN_OP_IN */
+ PIXMAN_STD_FAST_PATH (IN, a8, null, a8, sse2_composite_in_8_8),
+ PIXMAN_STD_FAST_PATH (IN, solid, a8, a8, sse2_composite_in_n_8_8),
+ PIXMAN_STD_FAST_PATH (IN, solid, null, a8, sse2_composite_in_n_8),
+
+ SIMPLE_NEAREST_FAST_PATH_COVER (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_COVER (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_COVER (OVER, a8r8g8b8, a8r8g8b8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_COVER (OVER, a8b8g8r8, a8b8g8r8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_NONE (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_NONE (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_NONE (OVER, a8r8g8b8, a8r8g8b8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_NONE (OVER, a8b8g8r8, a8b8g8r8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_PAD (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_PAD (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_PAD (OVER, a8r8g8b8, a8r8g8b8, sse2_8888_8888),
+ SIMPLE_NEAREST_FAST_PATH_PAD (OVER, a8b8g8r8, a8b8g8r8, sse2_8888_8888),
+
+ SIMPLE_NEAREST_SOLID_MASK_FAST_PATH (OVER, a8r8g8b8, a8r8g8b8, sse2_8888_n_8888),
+ SIMPLE_NEAREST_SOLID_MASK_FAST_PATH (OVER, a8b8g8r8, a8b8g8r8, sse2_8888_n_8888),
+ SIMPLE_NEAREST_SOLID_MASK_FAST_PATH (OVER, a8r8g8b8, x8r8g8b8, sse2_8888_n_8888),
+ SIMPLE_NEAREST_SOLID_MASK_FAST_PATH (OVER, a8b8g8r8, x8b8g8r8, sse2_8888_n_8888),
+
+ { PIXMAN_OP_NONE },
+};
+
+static pixman_bool_t
+sse2_blt (pixman_implementation_t *imp,
+ uint32_t * src_bits,
+ uint32_t * dst_bits,
+ int src_stride,
+ int dst_stride,
+ int src_bpp,
+ int dst_bpp,
+ int src_x,
+ int src_y,
+ int dst_x,
+ int dst_y,
+ int width,
+ int height)
+{
+ if (!pixman_blt_sse2 (
+ src_bits, dst_bits, src_stride, dst_stride, src_bpp, dst_bpp,
+ src_x, src_y, dst_x, dst_y, width, height))
+
+ {
+ return _pixman_implementation_blt (
+ imp->delegate,
+ src_bits, dst_bits, src_stride, dst_stride, src_bpp, dst_bpp,
+ src_x, src_y, dst_x, dst_y, width, height);
+ }
+
+ return TRUE;
+}
+
+#if defined(__GNUC__) && !defined(__x86_64__) && !defined(__amd64__)
+__attribute__((__force_align_arg_pointer__))
+#endif
+static pixman_bool_t
+sse2_fill (pixman_implementation_t *imp,
+ uint32_t * bits,
+ int stride,
+ int bpp,
+ int x,
+ int y,
+ int width,
+ int height,
+ uint32_t xor)
+{
+ if (!pixman_fill_sse2 (bits, stride, bpp, x, y, width, height, xor))
+ {
+ return _pixman_implementation_fill (
+ imp->delegate, bits, stride, bpp, x, y, width, height, xor);
+ }
+
+ return TRUE;
+}
+
+static uint32_t *
+sse2_fetch_x8r8g8b8 (pixman_iter_t *iter, const uint32_t *mask)
+{
+ int w = iter->width;
+ __m128i ff000000 = mask_ff000000;
+ uint32_t *dst = iter->buffer;
+ uint32_t *src = (uint32_t *)iter->bits;
+
+ iter->bits += iter->stride;
+
+ while (w && ((unsigned long)dst) & 0x0f)
+ {
+ *dst++ = (*src++) | 0xff000000;
+ w--;
+ }
+
+ while (w >= 4)
+ {
+ save_128_aligned (
+ (__m128i *)dst, _mm_or_si128 (
+ load_128_unaligned ((__m128i *)src), ff000000));
+
+ dst += 4;
+ src += 4;
+ w -= 4;
+ }
+
+ while (w)
+ {
+ *dst++ = (*src++) | 0xff000000;
+ w--;
+ }
+
+ return iter->buffer;
+}
+
+static uint32_t *
+sse2_fetch_r5g6b5 (pixman_iter_t *iter, const uint32_t *mask)
+{
+ int w = iter->width;
+ uint32_t *dst = iter->buffer;
+ uint16_t *src = (uint16_t *)iter->bits;
+ __m128i ff000000 = mask_ff000000;
+
+ iter->bits += iter->stride;
+
+ while (w && ((unsigned long)dst) & 0x0f)
+ {
+ uint16_t s = *src++;
+
+ *dst++ = CONVERT_0565_TO_8888 (s);
+ w--;
+ }
+
+ while (w >= 8)
+ {
+ __m128i lo, hi, s;
+
+ s = _mm_loadu_si128 ((__m128i *)src);
+
+ lo = unpack_565_to_8888 (_mm_unpacklo_epi16 (s, _mm_setzero_si128 ()));
+ hi = unpack_565_to_8888 (_mm_unpackhi_epi16 (s, _mm_setzero_si128 ()));
+
+ save_128_aligned ((__m128i *)(dst + 0), _mm_or_si128 (lo, ff000000));
+ save_128_aligned ((__m128i *)(dst + 4), _mm_or_si128 (hi, ff000000));
+
+ dst += 8;
+ src += 8;
+ w -= 8;
+ }
+
+ while (w)
+ {
+ uint16_t s = *src++;
+
+ *dst++ = CONVERT_0565_TO_8888 (s);
+ w--;
+ }
+
+ return iter->buffer;
+}
+
+static uint32_t *
+sse2_fetch_a8 (pixman_iter_t *iter, const uint32_t *mask)
+{
+ int w = iter->width;
+ uint32_t *dst = iter->buffer;
+ uint8_t *src = iter->bits;
+ __m128i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6;
+
+ iter->bits += iter->stride;
+
+ while (w && (((unsigned long)dst) & 15))
+ {
+ *dst++ = *(src++) << 24;
+ w--;
+ }
+
+ while (w >= 16)
+ {
+ xmm0 = _mm_loadu_si128((__m128i *)src);
+
+ xmm1 = _mm_unpacklo_epi8 (_mm_setzero_si128(), xmm0);
+ xmm2 = _mm_unpackhi_epi8 (_mm_setzero_si128(), xmm0);
+ xmm3 = _mm_unpacklo_epi16 (_mm_setzero_si128(), xmm1);
+ xmm4 = _mm_unpackhi_epi16 (_mm_setzero_si128(), xmm1);
+ xmm5 = _mm_unpacklo_epi16 (_mm_setzero_si128(), xmm2);
+ xmm6 = _mm_unpackhi_epi16 (_mm_setzero_si128(), xmm2);
+
+ _mm_store_si128(((__m128i *)(dst + 0)), xmm3);
+ _mm_store_si128(((__m128i *)(dst + 4)), xmm4);
+ _mm_store_si128(((__m128i *)(dst + 8)), xmm5);
+ _mm_store_si128(((__m128i *)(dst + 12)), xmm6);
+
+ dst += 16;
+ src += 16;
+ w -= 16;
+ }
+
+ while (w)
+ {
+ *dst++ = *(src++) << 24;
+ w--;
+ }
+
+ return iter->buffer;
+}
+
+typedef struct
+{
+ pixman_format_code_t format;
+ pixman_iter_get_scanline_t get_scanline;
+} fetcher_info_t;
+
+static const fetcher_info_t fetchers[] =
+{
+ { PIXMAN_x8r8g8b8, sse2_fetch_x8r8g8b8 },
+ { PIXMAN_r5g6b5, sse2_fetch_r5g6b5 },
+ { PIXMAN_a8, sse2_fetch_a8 },
+ { PIXMAN_null }
+};
+
+static void
+sse2_src_iter_init (pixman_implementation_t *imp,
+ pixman_iter_t *iter,
+ pixman_image_t *image,
+ int x, int y, int width, int height,
+ uint8_t *buffer, iter_flags_t flags)
+{
+#define FLAGS \
+ (FAST_PATH_STANDARD_FLAGS | FAST_PATH_ID_TRANSFORM)
+
+ if ((flags & ITER_NARROW) &&
+ (image->common.flags & FLAGS) == FLAGS &&
+ x >= 0 && y >= 0 &&
+ x + width <= image->bits.width &&
+ y + height <= image->bits.height)
+ {
+ const fetcher_info_t *f;
+
+ for (f = &fetchers[0]; f->format != PIXMAN_null; f++)
+ {
+ if (image->common.extended_format_code == f->format)
+ {
+ uint8_t *b = (uint8_t *)image->bits.bits;
+ int s = image->bits.rowstride * 4;
+
+ iter->bits = b + s * y + x * PIXMAN_FORMAT_BPP (f->format) / 8;
+ iter->stride = s;
+ iter->width = width;
+ iter->buffer = (uint32_t *)buffer;
+
+ iter->get_scanline = f->get_scanline;
+ return;
+ }
+ }
+ }
+
+ _pixman_implementation_src_iter_init (
+ imp->delegate, iter, image, x, y, width, height, buffer, flags);
+}
+
+#if defined(__GNUC__) && !defined(__x86_64__) && !defined(__amd64__)
+__attribute__((__force_align_arg_pointer__))
+#endif
+pixman_implementation_t *
+_pixman_implementation_create_sse2 (pixman_implementation_t *fallback)
+{
+ pixman_implementation_t *imp = _pixman_implementation_create (fallback, sse2_fast_paths);
+
+ /* SSE2 constants */
+ mask_565_r = create_mask_2x32_128 (0x00f80000, 0x00f80000);
+ mask_565_g1 = create_mask_2x32_128 (0x00070000, 0x00070000);
+ mask_565_g2 = create_mask_2x32_128 (0x000000e0, 0x000000e0);
+ mask_565_b = create_mask_2x32_128 (0x0000001f, 0x0000001f);
+ mask_red = create_mask_2x32_128 (0x00f80000, 0x00f80000);
+ mask_green = create_mask_2x32_128 (0x0000fc00, 0x0000fc00);
+ mask_blue = create_mask_2x32_128 (0x000000f8, 0x000000f8);
+ mask_565_fix_rb = create_mask_2x32_128 (0x00e000e0, 0x00e000e0);
+ mask_565_fix_g = create_mask_2x32_128 (0x0000c000, 0x0000c000);
+ mask_0080 = create_mask_16_128 (0x0080);
+ mask_00ff = create_mask_16_128 (0x00ff);
+ mask_0101 = create_mask_16_128 (0x0101);
+ mask_ffff = create_mask_16_128 (0xffff);
+ mask_ff000000 = create_mask_2x32_128 (0xff000000, 0xff000000);
+ mask_alpha = create_mask_2x32_128 (0x00ff0000, 0x00000000);
+
+ /* MMX constants */
+ mask_x565_rgb = create_mask_2x32_64 (0x000001f0, 0x003f001f);
+ mask_x565_unpack = create_mask_2x32_64 (0x00000084, 0x04100840);
+
+ mask_x0080 = create_mask_16_64 (0x0080);
+ mask_x00ff = create_mask_16_64 (0x00ff);
+ mask_x0101 = create_mask_16_64 (0x0101);
+ mask_x_alpha = create_mask_2x32_64 (0x00ff0000, 0x00000000);
+
+ _mm_empty ();
+
+ /* Set up function pointers */
+
+ /* SSE code patch for fbcompose.c */
+ imp->combine_32[PIXMAN_OP_OVER] = sse2_combine_over_u;
+ imp->combine_32[PIXMAN_OP_OVER_REVERSE] = sse2_combine_over_reverse_u;
+ imp->combine_32[PIXMAN_OP_IN] = sse2_combine_in_u;
+ imp->combine_32[PIXMAN_OP_IN_REVERSE] = sse2_combine_in_reverse_u;
+ imp->combine_32[PIXMAN_OP_OUT] = sse2_combine_out_u;
+ imp->combine_32[PIXMAN_OP_OUT_REVERSE] = sse2_combine_out_reverse_u;
+ imp->combine_32[PIXMAN_OP_ATOP] = sse2_combine_atop_u;
+ imp->combine_32[PIXMAN_OP_ATOP_REVERSE] = sse2_combine_atop_reverse_u;
+ imp->combine_32[PIXMAN_OP_XOR] = sse2_combine_xor_u;
+ imp->combine_32[PIXMAN_OP_ADD] = sse2_combine_add_u;
+
+ imp->combine_32[PIXMAN_OP_SATURATE] = sse2_combine_saturate_u;
+
+ imp->combine_32_ca[PIXMAN_OP_SRC] = sse2_combine_src_ca;
+ imp->combine_32_ca[PIXMAN_OP_OVER] = sse2_combine_over_ca;
+ imp->combine_32_ca[PIXMAN_OP_OVER_REVERSE] = sse2_combine_over_reverse_ca;
+ imp->combine_32_ca[PIXMAN_OP_IN] = sse2_combine_in_ca;
+ imp->combine_32_ca[PIXMAN_OP_IN_REVERSE] = sse2_combine_in_reverse_ca;
+ imp->combine_32_ca[PIXMAN_OP_OUT] = sse2_combine_out_ca;
+ imp->combine_32_ca[PIXMAN_OP_OUT_REVERSE] = sse2_combine_out_reverse_ca;
+ imp->combine_32_ca[PIXMAN_OP_ATOP] = sse2_combine_atop_ca;
+ imp->combine_32_ca[PIXMAN_OP_ATOP_REVERSE] = sse2_combine_atop_reverse_ca;
+ imp->combine_32_ca[PIXMAN_OP_XOR] = sse2_combine_xor_ca;
+ imp->combine_32_ca[PIXMAN_OP_ADD] = sse2_combine_add_ca;
+
+ imp->blt = sse2_blt;
+ imp->fill = sse2_fill;
+
+ imp->src_iter_init = sse2_src_iter_init;
+
+ return imp;
+}
+
+#endif /* USE_SSE2 */
diff --git a/pixman/pixman/pixman-trap.c b/pixman/pixman/pixman-trap.c index 787093a85..4973143b5 100644 --- a/pixman/pixman/pixman-trap.c +++ b/pixman/pixman/pixman-trap.c @@ -136,8 +136,8 @@ _pixman_edge_multi_init (pixman_edge_t * e, if (ne > 0)
{
- int nx = ne / e->dy;
- ne -= nx * e->dy;
+ pixman_fixed_48_16_t nx = ne / e->dy;
+ ne -= nx * (pixman_fixed_48_16_t)e->dy;
stepx += nx * e->signdx;
}
diff --git a/pixman/pixman/pixman-version.h b/pixman/pixman/pixman-version.h new file mode 100644 index 000000000..70642e961 --- /dev/null +++ b/pixman/pixman/pixman-version.h @@ -0,0 +1,50 @@ +/* + * Copyright © 2008 Red Hat, Inc. + * + * Permission is hereby granted, free of charge, to any person + * obtaining a copy of this software and associated documentation + * files (the "Software"), to deal in the Software without + * restriction, including without limitation the rights to use, copy, + * modify, merge, publish, distribute, sublicense, and/or sell copies + * of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be + * included in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Author: Carl D. Worth <cworth@cworth.org> + */ + +#ifndef PIXMAN_VERSION_H__ +#define PIXMAN_VERSION_H__ + +#ifndef PIXMAN_H__ +# error pixman-version.h should only be included by pixman.h +#endif + +#define PIXMAN_VERSION_MAJOR 0 +#define PIXMAN_VERSION_MINOR 19 +#define PIXMAN_VERSION_MICRO 1 + +#define PIXMAN_VERSION_STRING "0.19.1" + +#define PIXMAN_VERSION_ENCODE(major, minor, micro) ( \ + ((major) * 10000) \ + + ((minor) * 100) \ + + ((micro) * 1)) + +#define PIXMAN_VERSION PIXMAN_VERSION_ENCODE( \ + PIXMAN_VERSION_MAJOR, \ + PIXMAN_VERSION_MINOR, \ + PIXMAN_VERSION_MICRO) + +#endif /* PIXMAN_VERSION_H__ */ diff --git a/pixman/test/Makefile.am b/pixman/test/Makefile.am index 3ce466eec..05cb83d06 100644 --- a/pixman/test/Makefile.am +++ b/pixman/test/Makefile.am @@ -1,44 +1,44 @@ -AM_CFLAGS = @OPENMP_CFLAGS@ -AM_LDFLAGS = @OPENMP_CFLAGS@ @TESTPROGS_EXTRA_LDFLAGS@ -LDADD = $(top_builddir)/pixman/libpixman-1.la -lm -INCLUDES = -I$(top_srcdir)/pixman -I$(top_builddir)/pixman - -TESTPROGRAMS = \ - a1-trap-test \ - pdf-op-test \ - region-test \ - region-translate-test \ - fetch-test \ - oob-test \ - trap-crasher \ - alpha-loop \ - scaling-crash-test \ - gradient-crash-test \ - alphamap \ - stress-test \ - blitters-test \ - scaling-test \ - affine-test \ - composite - -pdf_op_test_SOURCES = pdf-op-test.c utils.c utils.h -region_test_SOURCES = region-test.c utils.c utils.h -blitters_test_SOURCES = blitters-test.c utils.c utils.h -scaling_test_SOURCES = scaling-test.c utils.c utils.h -affine_test_SOURCES = affine-test.c utils.c utils.h -alphamap_SOURCES = alphamap.c utils.c utils.h -alpha_loop_SOURCES = alpha-loop.c utils.c utils.h -composite_SOURCES = composite.c utils.c utils.h -gradient_crash_test_SOURCES = gradient-crash-test.c utils.c utils.h -stress_test_SOURCES = stress-test.c utils.c utils.h - -# Benchmarks - -BENCHMARKS = \ - lowlevel-blt-bench - -lowlevel_blt_bench_SOURCES = lowlevel-blt-bench.c utils.c utils.h - -noinst_PROGRAMS = $(TESTPROGRAMS) $(BENCHMARKS) - -TESTS = $(TESTPROGRAMS) +AM_CFLAGS = @OPENMP_CFLAGS@
+AM_LDFLAGS = @OPENMP_CFLAGS@ @TESTPROGS_EXTRA_LDFLAGS@
+LDADD = $(top_builddir)/pixman/libpixman-1.la -lm
+INCLUDES = -I$(top_srcdir)/pixman -I$(top_builddir)/pixman
+
+TESTPROGRAMS = \
+ a1-trap-test \
+ pdf-op-test \
+ region-test \
+ region-translate-test \
+ fetch-test \
+ oob-test \
+ trap-crasher \
+ alpha-loop \
+ scaling-crash-test \
+ gradient-crash-test \
+ alphamap \
+ stress-test \
+ blitters-test \
+ scaling-test \
+ affine-test \
+ composite
+
+pdf_op_test_SOURCES = pdf-op-test.c utils.c utils.h
+region_test_SOURCES = region-test.c utils.c utils.h
+blitters_test_SOURCES = blitters-test.c utils.c utils.h
+scaling_test_SOURCES = scaling-test.c utils.c utils.h
+affine_test_SOURCES = affine-test.c utils.c utils.h
+alphamap_SOURCES = alphamap.c utils.c utils.h
+alpha_loop_SOURCES = alpha-loop.c utils.c utils.h
+composite_SOURCES = composite.c utils.c utils.h
+gradient_crash_test_SOURCES = gradient-crash-test.c utils.c utils.h
+stress_test_SOURCES = stress-test.c utils.c utils.h
+
+# Benchmarks
+
+BENCHMARKS = \
+ lowlevel-blt-bench
+
+lowlevel_blt_bench_SOURCES = lowlevel-blt-bench.c utils.c utils.h
+
+noinst_PROGRAMS = $(TESTPROGRAMS) $(BENCHMARKS)
+
+TESTS = $(TESTPROGRAMS)
diff --git a/pixman/test/scaling-test.c b/pixman/test/scaling-test.c index dbb9d39b0..2a0826a46 100644 --- a/pixman/test/scaling-test.c +++ b/pixman/test/scaling-test.c @@ -1,368 +1,368 @@ -/* - * Test program, which can detect some problems with nearest neighbour - * and bilinear scaling in pixman. Testing is done by running lots - * of random SRC and OVER compositing operations a8r8g8b8, x8a8r8g8b8 - * and r5g6b5 color formats. - * - * Script 'fuzzer-find-diff.pl' can be used to narrow down the problem in - * the case of test failure. - */ -#include <assert.h> -#include <stdlib.h> -#include <stdio.h> -#include "utils.h" - -#define MAX_SRC_WIDTH 48 -#define MAX_SRC_HEIGHT 8 -#define MAX_DST_WIDTH 48 -#define MAX_DST_HEIGHT 8 -#define MAX_STRIDE 4 - -/* - * Composite operation with pseudorandom images - */ -uint32_t -test_composite (int testnum, - int verbose) -{ - int i; - pixman_image_t * src_img; - pixman_image_t * mask_img; - pixman_image_t * dst_img; - pixman_transform_t transform; - pixman_region16_t clip; - int src_width, src_height; - int mask_width, mask_height; - int dst_width, dst_height; - int src_stride, mask_stride, dst_stride; - int src_x, src_y; - int mask_x, mask_y; - int dst_x, dst_y; - int src_bpp; - int mask_bpp = 1; - int dst_bpp; - int w, h; - pixman_fixed_t scale_x = 65536, scale_y = 65536; - pixman_fixed_t translate_x = 0, translate_y = 0; - pixman_fixed_t mask_scale_x = 65536, mask_scale_y = 65536; - pixman_fixed_t mask_translate_x = 0, mask_translate_y = 0; - pixman_op_t op; - pixman_repeat_t repeat = PIXMAN_REPEAT_NONE; - pixman_repeat_t mask_repeat = PIXMAN_REPEAT_NONE; - pixman_format_code_t src_fmt, dst_fmt; - uint32_t * srcbuf; - uint32_t * dstbuf; - uint32_t * maskbuf; - uint32_t crc32; - FLOAT_REGS_CORRUPTION_DETECTOR_START (); - - lcg_srand (testnum); - - src_bpp = (lcg_rand_n (2) == 0) ? 2 : 4; - dst_bpp = (lcg_rand_n (2) == 0) ? 2 : 4; - switch (lcg_rand_n (3)) - { - case 0: - op = PIXMAN_OP_SRC; - break; - case 1: - op = PIXMAN_OP_OVER; - break; - default: - op = PIXMAN_OP_ADD; - break; - } - - src_width = lcg_rand_n (MAX_SRC_WIDTH) + 1; - src_height = lcg_rand_n (MAX_SRC_HEIGHT) + 1; - - if (lcg_rand_n (2)) - { - mask_width = lcg_rand_n (MAX_SRC_WIDTH) + 1; - mask_height = lcg_rand_n (MAX_SRC_HEIGHT) + 1; - } - else - { - mask_width = mask_height = 1; - } - - dst_width = lcg_rand_n (MAX_DST_WIDTH) + 1; - dst_height = lcg_rand_n (MAX_DST_HEIGHT) + 1; - src_stride = src_width * src_bpp + lcg_rand_n (MAX_STRIDE) * src_bpp; - mask_stride = mask_width * mask_bpp + lcg_rand_n (MAX_STRIDE) * mask_bpp; - dst_stride = dst_width * dst_bpp + lcg_rand_n (MAX_STRIDE) * dst_bpp; - - if (src_stride & 3) - src_stride += 2; - - if (mask_stride & 1) - mask_stride += 1; - if (mask_stride & 2) - mask_stride += 2; - - if (dst_stride & 3) - dst_stride += 2; - - src_x = -(src_width / 4) + lcg_rand_n (src_width * 3 / 2); - src_y = -(src_height / 4) + lcg_rand_n (src_height * 3 / 2); - mask_x = -(mask_width / 4) + lcg_rand_n (mask_width * 3 / 2); - mask_y = -(mask_height / 4) + lcg_rand_n (mask_height * 3 / 2); - dst_x = -(dst_width / 4) + lcg_rand_n (dst_width * 3 / 2); - dst_y = -(dst_height / 4) + lcg_rand_n (dst_height * 3 / 2); - w = lcg_rand_n (dst_width * 3 / 2 - dst_x); - h = lcg_rand_n (dst_height * 3 / 2 - dst_y); - - srcbuf = (uint32_t *)malloc (src_stride * src_height); - maskbuf = (uint32_t *)malloc (mask_stride * mask_height); - dstbuf = (uint32_t *)malloc (dst_stride * dst_height); - - for (i = 0; i < src_stride * src_height; i++) - *((uint8_t *)srcbuf + i) = lcg_rand_n (256); - - for (i = 0; i < mask_stride * mask_height; i++) - *((uint8_t *)maskbuf + i) = lcg_rand_n (256); - - for (i = 0; i < dst_stride * dst_height; i++) - *((uint8_t *)dstbuf + i) = lcg_rand_n (256); - - src_fmt = src_bpp == 4 ? (lcg_rand_n (2) == 0 ? - PIXMAN_a8r8g8b8 : PIXMAN_x8r8g8b8) : PIXMAN_r5g6b5; - - dst_fmt = dst_bpp == 4 ? (lcg_rand_n (2) == 0 ? - PIXMAN_a8r8g8b8 : PIXMAN_x8r8g8b8) : PIXMAN_r5g6b5; - - src_img = pixman_image_create_bits ( - src_fmt, src_width, src_height, srcbuf, src_stride); - - mask_img = pixman_image_create_bits ( - PIXMAN_a8, mask_width, mask_height, maskbuf, mask_stride); - - dst_img = pixman_image_create_bits ( - dst_fmt, dst_width, dst_height, dstbuf, dst_stride); - - image_endian_swap (src_img, src_bpp * 8); - image_endian_swap (dst_img, dst_bpp * 8); - - if (lcg_rand_n (4) > 0) - { - scale_x = -32768 * 3 + lcg_rand_N (65536 * 5); - scale_y = -32768 * 3 + lcg_rand_N (65536 * 5); - translate_x = lcg_rand_N (65536); - translate_y = lcg_rand_N (65536); - pixman_transform_init_scale (&transform, scale_x, scale_y); - pixman_transform_translate (&transform, NULL, translate_x, translate_y); - pixman_image_set_transform (src_img, &transform); - } - - if (lcg_rand_n (2) > 0) - { - mask_scale_x = -32768 * 3 + lcg_rand_N (65536 * 5); - mask_scale_y = -32768 * 3 + lcg_rand_N (65536 * 5); - mask_translate_x = lcg_rand_N (65536); - mask_translate_y = lcg_rand_N (65536); - pixman_transform_init_scale (&transform, mask_scale_x, mask_scale_y); - pixman_transform_translate (&transform, NULL, mask_translate_x, mask_translate_y); - pixman_image_set_transform (mask_img, &transform); - } - - switch (lcg_rand_n (4)) - { - case 0: - mask_repeat = PIXMAN_REPEAT_NONE; - break; - - case 1: - mask_repeat = PIXMAN_REPEAT_NORMAL; - break; - - case 2: - mask_repeat = PIXMAN_REPEAT_PAD; - break; - - case 3: - mask_repeat = PIXMAN_REPEAT_REFLECT; - break; - - default: - break; - } - pixman_image_set_repeat (mask_img, mask_repeat); - - switch (lcg_rand_n (4)) - { - case 0: - repeat = PIXMAN_REPEAT_NONE; - break; - - case 1: - repeat = PIXMAN_REPEAT_NORMAL; - break; - - case 2: - repeat = PIXMAN_REPEAT_PAD; - break; - - case 3: - repeat = PIXMAN_REPEAT_REFLECT; - break; - - default: - break; - } - pixman_image_set_repeat (src_img, repeat); - - if (lcg_rand_n (2)) - pixman_image_set_filter (src_img, PIXMAN_FILTER_NEAREST, NULL, 0); - else - pixman_image_set_filter (src_img, PIXMAN_FILTER_BILINEAR, NULL, 0); - - if (lcg_rand_n (2)) - pixman_image_set_filter (mask_img, PIXMAN_FILTER_NEAREST, NULL, 0); - else - pixman_image_set_filter (mask_img, PIXMAN_FILTER_BILINEAR, NULL, 0); - - if (verbose) - { - printf ("src_fmt=%08X, dst_fmt=%08X\n", src_fmt, dst_fmt); - printf ("op=%d, scale_x=%d, scale_y=%d, repeat=%d\n", - op, scale_x, scale_y, repeat); - printf ("translate_x=%d, translate_y=%d\n", - translate_x, translate_y); - printf ("src_width=%d, src_height=%d, dst_width=%d, dst_height=%d\n", - src_width, src_height, dst_width, dst_height); - printf ("src_x=%d, src_y=%d, dst_x=%d, dst_y=%d\n", - src_x, src_y, dst_x, dst_y); - printf ("w=%d, h=%d\n", w, h); - } - - if (lcg_rand_n (8) == 0) - { - pixman_box16_t clip_boxes[2]; - int n = lcg_rand_n (2) + 1; - - for (i = 0; i < n; i++) - { - clip_boxes[i].x1 = lcg_rand_n (src_width); - clip_boxes[i].y1 = lcg_rand_n (src_height); - clip_boxes[i].x2 = - clip_boxes[i].x1 + lcg_rand_n (src_width - clip_boxes[i].x1); - clip_boxes[i].y2 = - clip_boxes[i].y1 + lcg_rand_n (src_height - clip_boxes[i].y1); - - if (verbose) - { - printf ("source clip box: [%d,%d-%d,%d]\n", - clip_boxes[i].x1, clip_boxes[i].y1, - clip_boxes[i].x2, clip_boxes[i].y2); - } - } - - pixman_region_init_rects (&clip, clip_boxes, n); - pixman_image_set_clip_region (src_img, &clip); - pixman_image_set_source_clipping (src_img, 1); - pixman_region_fini (&clip); - } - - if (lcg_rand_n (8) == 0) - { - pixman_box16_t clip_boxes[2]; - int n = lcg_rand_n (2) + 1; - - for (i = 0; i < n; i++) - { - clip_boxes[i].x1 = lcg_rand_n (mask_width); - clip_boxes[i].y1 = lcg_rand_n (mask_height); - clip_boxes[i].x2 = - clip_boxes[i].x1 + lcg_rand_n (mask_width - clip_boxes[i].x1); - clip_boxes[i].y2 = - clip_boxes[i].y1 + lcg_rand_n (mask_height - clip_boxes[i].y1); - - if (verbose) - { - printf ("mask clip box: [%d,%d-%d,%d]\n", - clip_boxes[i].x1, clip_boxes[i].y1, - clip_boxes[i].x2, clip_boxes[i].y2); - } - } - - pixman_region_init_rects (&clip, clip_boxes, n); - pixman_image_set_clip_region (mask_img, &clip); - pixman_image_set_source_clipping (mask_img, 1); - pixman_region_fini (&clip); - } - - if (lcg_rand_n (8) == 0) - { - pixman_box16_t clip_boxes[2]; - int n = lcg_rand_n (2) + 1; - for (i = 0; i < n; i++) - { - clip_boxes[i].x1 = lcg_rand_n (dst_width); - clip_boxes[i].y1 = lcg_rand_n (dst_height); - clip_boxes[i].x2 = - clip_boxes[i].x1 + lcg_rand_n (dst_width - clip_boxes[i].x1); - clip_boxes[i].y2 = - clip_boxes[i].y1 + lcg_rand_n (dst_height - clip_boxes[i].y1); - - if (verbose) - { - printf ("destination clip box: [%d,%d-%d,%d]\n", - clip_boxes[i].x1, clip_boxes[i].y1, - clip_boxes[i].x2, clip_boxes[i].y2); - } - } - pixman_region_init_rects (&clip, clip_boxes, n); - pixman_image_set_clip_region (dst_img, &clip); - pixman_region_fini (&clip); - } - - if (lcg_rand_n (2) == 0) - pixman_image_composite (op, src_img, NULL, dst_img, - src_x, src_y, 0, 0, dst_x, dst_y, w, h); - else - pixman_image_composite (op, src_img, mask_img, dst_img, - src_x, src_y, mask_x, mask_y, dst_x, dst_y, w, h); - - if (dst_fmt == PIXMAN_x8r8g8b8) - { - /* ignore unused part */ - for (i = 0; i < dst_stride * dst_height / 4; i++) - dstbuf[i] &= 0xFFFFFF; - } - - image_endian_swap (dst_img, dst_bpp * 8); - - if (verbose) - { - int j; - - for (i = 0; i < dst_height; i++) - { - for (j = 0; j < dst_stride; j++) - printf ("%02X ", *((uint8_t *)dstbuf + i * dst_stride + j)); - - printf ("\n"); - } - } - - pixman_image_unref (src_img); - pixman_image_unref (mask_img); - pixman_image_unref (dst_img); - - crc32 = compute_crc32 (0, dstbuf, dst_stride * dst_height); - free (srcbuf); - free (maskbuf); - free (dstbuf); - - FLOAT_REGS_CORRUPTION_DETECTOR_FINISH (); - return crc32; -} - -int -main (int argc, const char *argv[]) -{ - pixman_disable_out_of_bounds_workaround (); - - return fuzzer_test_main("scaling", 8000000, 0x80DF1CB2, - test_composite, argc, argv); -} +/*
+ * Test program, which can detect some problems with nearest neighbour
+ * and bilinear scaling in pixman. Testing is done by running lots
+ * of random SRC and OVER compositing operations a8r8g8b8, x8a8r8g8b8
+ * and r5g6b5 color formats.
+ *
+ * Script 'fuzzer-find-diff.pl' can be used to narrow down the problem in
+ * the case of test failure.
+ */
+#include <assert.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include "utils.h"
+
+#define MAX_SRC_WIDTH 48
+#define MAX_SRC_HEIGHT 8
+#define MAX_DST_WIDTH 48
+#define MAX_DST_HEIGHT 8
+#define MAX_STRIDE 4
+
+/*
+ * Composite operation with pseudorandom images
+ */
+uint32_t
+test_composite (int testnum,
+ int verbose)
+{
+ int i;
+ pixman_image_t * src_img;
+ pixman_image_t * mask_img;
+ pixman_image_t * dst_img;
+ pixman_transform_t transform;
+ pixman_region16_t clip;
+ int src_width, src_height;
+ int mask_width, mask_height;
+ int dst_width, dst_height;
+ int src_stride, mask_stride, dst_stride;
+ int src_x, src_y;
+ int mask_x, mask_y;
+ int dst_x, dst_y;
+ int src_bpp;
+ int mask_bpp = 1;
+ int dst_bpp;
+ int w, h;
+ pixman_fixed_t scale_x = 65536, scale_y = 65536;
+ pixman_fixed_t translate_x = 0, translate_y = 0;
+ pixman_fixed_t mask_scale_x = 65536, mask_scale_y = 65536;
+ pixman_fixed_t mask_translate_x = 0, mask_translate_y = 0;
+ pixman_op_t op;
+ pixman_repeat_t repeat = PIXMAN_REPEAT_NONE;
+ pixman_repeat_t mask_repeat = PIXMAN_REPEAT_NONE;
+ pixman_format_code_t src_fmt, dst_fmt;
+ uint32_t * srcbuf;
+ uint32_t * dstbuf;
+ uint32_t * maskbuf;
+ uint32_t crc32;
+ FLOAT_REGS_CORRUPTION_DETECTOR_START ();
+
+ lcg_srand (testnum);
+
+ src_bpp = (lcg_rand_n (2) == 0) ? 2 : 4;
+ dst_bpp = (lcg_rand_n (2) == 0) ? 2 : 4;
+ switch (lcg_rand_n (3))
+ {
+ case 0:
+ op = PIXMAN_OP_SRC;
+ break;
+ case 1:
+ op = PIXMAN_OP_OVER;
+ break;
+ default:
+ op = PIXMAN_OP_ADD;
+ break;
+ }
+
+ src_width = lcg_rand_n (MAX_SRC_WIDTH) + 1;
+ src_height = lcg_rand_n (MAX_SRC_HEIGHT) + 1;
+
+ if (lcg_rand_n (2))
+ {
+ mask_width = lcg_rand_n (MAX_SRC_WIDTH) + 1;
+ mask_height = lcg_rand_n (MAX_SRC_HEIGHT) + 1;
+ }
+ else
+ {
+ mask_width = mask_height = 1;
+ }
+
+ dst_width = lcg_rand_n (MAX_DST_WIDTH) + 1;
+ dst_height = lcg_rand_n (MAX_DST_HEIGHT) + 1;
+ src_stride = src_width * src_bpp + lcg_rand_n (MAX_STRIDE) * src_bpp;
+ mask_stride = mask_width * mask_bpp + lcg_rand_n (MAX_STRIDE) * mask_bpp;
+ dst_stride = dst_width * dst_bpp + lcg_rand_n (MAX_STRIDE) * dst_bpp;
+
+ if (src_stride & 3)
+ src_stride += 2;
+
+ if (mask_stride & 1)
+ mask_stride += 1;
+ if (mask_stride & 2)
+ mask_stride += 2;
+
+ if (dst_stride & 3)
+ dst_stride += 2;
+
+ src_x = -(src_width / 4) + lcg_rand_n (src_width * 3 / 2);
+ src_y = -(src_height / 4) + lcg_rand_n (src_height * 3 / 2);
+ mask_x = -(mask_width / 4) + lcg_rand_n (mask_width * 3 / 2);
+ mask_y = -(mask_height / 4) + lcg_rand_n (mask_height * 3 / 2);
+ dst_x = -(dst_width / 4) + lcg_rand_n (dst_width * 3 / 2);
+ dst_y = -(dst_height / 4) + lcg_rand_n (dst_height * 3 / 2);
+ w = lcg_rand_n (dst_width * 3 / 2 - dst_x);
+ h = lcg_rand_n (dst_height * 3 / 2 - dst_y);
+
+ srcbuf = (uint32_t *)malloc (src_stride * src_height);
+ maskbuf = (uint32_t *)malloc (mask_stride * mask_height);
+ dstbuf = (uint32_t *)malloc (dst_stride * dst_height);
+
+ for (i = 0; i < src_stride * src_height; i++)
+ *((uint8_t *)srcbuf + i) = lcg_rand_n (256);
+
+ for (i = 0; i < mask_stride * mask_height; i++)
+ *((uint8_t *)maskbuf + i) = lcg_rand_n (256);
+
+ for (i = 0; i < dst_stride * dst_height; i++)
+ *((uint8_t *)dstbuf + i) = lcg_rand_n (256);
+
+ src_fmt = src_bpp == 4 ? (lcg_rand_n (2) == 0 ?
+ PIXMAN_a8r8g8b8 : PIXMAN_x8r8g8b8) : PIXMAN_r5g6b5;
+
+ dst_fmt = dst_bpp == 4 ? (lcg_rand_n (2) == 0 ?
+ PIXMAN_a8r8g8b8 : PIXMAN_x8r8g8b8) : PIXMAN_r5g6b5;
+
+ src_img = pixman_image_create_bits (
+ src_fmt, src_width, src_height, srcbuf, src_stride);
+
+ mask_img = pixman_image_create_bits (
+ PIXMAN_a8, mask_width, mask_height, maskbuf, mask_stride);
+
+ dst_img = pixman_image_create_bits (
+ dst_fmt, dst_width, dst_height, dstbuf, dst_stride);
+
+ image_endian_swap (src_img, src_bpp * 8);
+ image_endian_swap (dst_img, dst_bpp * 8);
+
+ if (lcg_rand_n (4) > 0)
+ {
+ scale_x = -32768 * 3 + lcg_rand_N (65536 * 5);
+ scale_y = -32768 * 3 + lcg_rand_N (65536 * 5);
+ translate_x = lcg_rand_N (65536);
+ translate_y = lcg_rand_N (65536);
+ pixman_transform_init_scale (&transform, scale_x, scale_y);
+ pixman_transform_translate (&transform, NULL, translate_x, translate_y);
+ pixman_image_set_transform (src_img, &transform);
+ }
+
+ if (lcg_rand_n (2) > 0)
+ {
+ mask_scale_x = -32768 * 3 + lcg_rand_N (65536 * 5);
+ mask_scale_y = -32768 * 3 + lcg_rand_N (65536 * 5);
+ mask_translate_x = lcg_rand_N (65536);
+ mask_translate_y = lcg_rand_N (65536);
+ pixman_transform_init_scale (&transform, mask_scale_x, mask_scale_y);
+ pixman_transform_translate (&transform, NULL, mask_translate_x, mask_translate_y);
+ pixman_image_set_transform (mask_img, &transform);
+ }
+
+ switch (lcg_rand_n (4))
+ {
+ case 0:
+ mask_repeat = PIXMAN_REPEAT_NONE;
+ break;
+
+ case 1:
+ mask_repeat = PIXMAN_REPEAT_NORMAL;
+ break;
+
+ case 2:
+ mask_repeat = PIXMAN_REPEAT_PAD;
+ break;
+
+ case 3:
+ mask_repeat = PIXMAN_REPEAT_REFLECT;
+ break;
+
+ default:
+ break;
+ }
+ pixman_image_set_repeat (mask_img, mask_repeat);
+
+ switch (lcg_rand_n (4))
+ {
+ case 0:
+ repeat = PIXMAN_REPEAT_NONE;
+ break;
+
+ case 1:
+ repeat = PIXMAN_REPEAT_NORMAL;
+ break;
+
+ case 2:
+ repeat = PIXMAN_REPEAT_PAD;
+ break;
+
+ case 3:
+ repeat = PIXMAN_REPEAT_REFLECT;
+ break;
+
+ default:
+ break;
+ }
+ pixman_image_set_repeat (src_img, repeat);
+
+ if (lcg_rand_n (2))
+ pixman_image_set_filter (src_img, PIXMAN_FILTER_NEAREST, NULL, 0);
+ else
+ pixman_image_set_filter (src_img, PIXMAN_FILTER_BILINEAR, NULL, 0);
+
+ if (lcg_rand_n (2))
+ pixman_image_set_filter (mask_img, PIXMAN_FILTER_NEAREST, NULL, 0);
+ else
+ pixman_image_set_filter (mask_img, PIXMAN_FILTER_BILINEAR, NULL, 0);
+
+ if (verbose)
+ {
+ printf ("src_fmt=%08X, dst_fmt=%08X\n", src_fmt, dst_fmt);
+ printf ("op=%d, scale_x=%d, scale_y=%d, repeat=%d\n",
+ op, scale_x, scale_y, repeat);
+ printf ("translate_x=%d, translate_y=%d\n",
+ translate_x, translate_y);
+ printf ("src_width=%d, src_height=%d, dst_width=%d, dst_height=%d\n",
+ src_width, src_height, dst_width, dst_height);
+ printf ("src_x=%d, src_y=%d, dst_x=%d, dst_y=%d\n",
+ src_x, src_y, dst_x, dst_y);
+ printf ("w=%d, h=%d\n", w, h);
+ }
+
+ if (lcg_rand_n (8) == 0)
+ {
+ pixman_box16_t clip_boxes[2];
+ int n = lcg_rand_n (2) + 1;
+
+ for (i = 0; i < n; i++)
+ {
+ clip_boxes[i].x1 = lcg_rand_n (src_width);
+ clip_boxes[i].y1 = lcg_rand_n (src_height);
+ clip_boxes[i].x2 =
+ clip_boxes[i].x1 + lcg_rand_n (src_width - clip_boxes[i].x1);
+ clip_boxes[i].y2 =
+ clip_boxes[i].y1 + lcg_rand_n (src_height - clip_boxes[i].y1);
+
+ if (verbose)
+ {
+ printf ("source clip box: [%d,%d-%d,%d]\n",
+ clip_boxes[i].x1, clip_boxes[i].y1,
+ clip_boxes[i].x2, clip_boxes[i].y2);
+ }
+ }
+
+ pixman_region_init_rects (&clip, clip_boxes, n);
+ pixman_image_set_clip_region (src_img, &clip);
+ pixman_image_set_source_clipping (src_img, 1);
+ pixman_region_fini (&clip);
+ }
+
+ if (lcg_rand_n (8) == 0)
+ {
+ pixman_box16_t clip_boxes[2];
+ int n = lcg_rand_n (2) + 1;
+
+ for (i = 0; i < n; i++)
+ {
+ clip_boxes[i].x1 = lcg_rand_n (mask_width);
+ clip_boxes[i].y1 = lcg_rand_n (mask_height);
+ clip_boxes[i].x2 =
+ clip_boxes[i].x1 + lcg_rand_n (mask_width - clip_boxes[i].x1);
+ clip_boxes[i].y2 =
+ clip_boxes[i].y1 + lcg_rand_n (mask_height - clip_boxes[i].y1);
+
+ if (verbose)
+ {
+ printf ("mask clip box: [%d,%d-%d,%d]\n",
+ clip_boxes[i].x1, clip_boxes[i].y1,
+ clip_boxes[i].x2, clip_boxes[i].y2);
+ }
+ }
+
+ pixman_region_init_rects (&clip, clip_boxes, n);
+ pixman_image_set_clip_region (mask_img, &clip);
+ pixman_image_set_source_clipping (mask_img, 1);
+ pixman_region_fini (&clip);
+ }
+
+ if (lcg_rand_n (8) == 0)
+ {
+ pixman_box16_t clip_boxes[2];
+ int n = lcg_rand_n (2) + 1;
+ for (i = 0; i < n; i++)
+ {
+ clip_boxes[i].x1 = lcg_rand_n (dst_width);
+ clip_boxes[i].y1 = lcg_rand_n (dst_height);
+ clip_boxes[i].x2 =
+ clip_boxes[i].x1 + lcg_rand_n (dst_width - clip_boxes[i].x1);
+ clip_boxes[i].y2 =
+ clip_boxes[i].y1 + lcg_rand_n (dst_height - clip_boxes[i].y1);
+
+ if (verbose)
+ {
+ printf ("destination clip box: [%d,%d-%d,%d]\n",
+ clip_boxes[i].x1, clip_boxes[i].y1,
+ clip_boxes[i].x2, clip_boxes[i].y2);
+ }
+ }
+ pixman_region_init_rects (&clip, clip_boxes, n);
+ pixman_image_set_clip_region (dst_img, &clip);
+ pixman_region_fini (&clip);
+ }
+
+ if (lcg_rand_n (2) == 0)
+ pixman_image_composite (op, src_img, NULL, dst_img,
+ src_x, src_y, 0, 0, dst_x, dst_y, w, h);
+ else
+ pixman_image_composite (op, src_img, mask_img, dst_img,
+ src_x, src_y, mask_x, mask_y, dst_x, dst_y, w, h);
+
+ if (dst_fmt == PIXMAN_x8r8g8b8)
+ {
+ /* ignore unused part */
+ for (i = 0; i < dst_stride * dst_height / 4; i++)
+ dstbuf[i] &= 0xFFFFFF;
+ }
+
+ image_endian_swap (dst_img, dst_bpp * 8);
+
+ if (verbose)
+ {
+ int j;
+
+ for (i = 0; i < dst_height; i++)
+ {
+ for (j = 0; j < dst_stride; j++)
+ printf ("%02X ", *((uint8_t *)dstbuf + i * dst_stride + j));
+
+ printf ("\n");
+ }
+ }
+
+ pixman_image_unref (src_img);
+ pixman_image_unref (mask_img);
+ pixman_image_unref (dst_img);
+
+ crc32 = compute_crc32 (0, dstbuf, dst_stride * dst_height);
+ free (srcbuf);
+ free (maskbuf);
+ free (dstbuf);
+
+ FLOAT_REGS_CORRUPTION_DETECTOR_FINISH ();
+ return crc32;
+}
+
+int
+main (int argc, const char *argv[])
+{
+ pixman_disable_out_of_bounds_workaround ();
+
+ return fuzzer_test_main("scaling", 8000000, 0x80DF1CB2,
+ test_composite, argc, argv);
+}
|