svn merge ^/branches/released .

9 files changed, 11496 insertions, 11454 deletions

diff --git a/pixman/configure.ac b/pixman/configure.ac
index ac0d16158..6552f1270 100644
--- a/pixman/configure.ac
+++ b/pixman/configure.ac
@@ -1,783 +1,797 @@
-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], 3)

-

-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 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

-	   test/Makefile])

+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], 3)
+
+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
+	   test/Makefile])
diff --git a/pixman/pixman/pixman-bits-image.c b/pixman/pixman/pixman-bits-image.c
index 85ff2a339..c453e0ee6 100644
--- a/pixman/pixman/pixman-bits-image.c
+++ b/pixman/pixman/pixman-bits-image.c
@@ -1,1443 +1,1441 @@
-/*

- * Copyright © 2000 Keith Packard, member of The XFree86 Project, Inc.

- *             2005 Lars Knoll & Zack Rusin, Trolltech

- *             2008 Aaron Plattner, NVIDIA Corporation

- * Copyright © 2000 SuSE, Inc.

- * Copyright © 2007, 2009 Red Hat, Inc.

- * Copyright © 2008 André Tupinambá <andrelrt@gmail.com>

- *

- * 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 Keith Packard not be used in

- * advertising or publicity pertaining to distribution of the software without

- * specific, written prior permission.  Keith Packard 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.

- */

-

-#ifdef HAVE_CONFIG_H

-#include <config.h>

-#endif

-#include <stdio.h>

-#include <stdlib.h>

-#include <string.h>

-#include "pixman-private.h"

-#include "pixman-combine32.h"

-

-/* Store functions */

-void

-_pixman_image_store_scanline_32 (bits_image_t *  image,

-                                 int             x,

-                                 int             y,

-                                 int             width,

-                                 const uint32_t *buffer)

-{

-    image->store_scanline_32 (image, x, y, width, buffer);

-

-    if (image->common.alpha_map)

-    {

-	x -= image->common.alpha_origin_x;

-	y -= image->common.alpha_origin_y;

-

-	image->common.alpha_map->store_scanline_32 (

-	    image->common.alpha_map, x, y, width, buffer);

-    }

-}

-

-void

-_pixman_image_store_scanline_64 (bits_image_t *  image,

-                                 int             x,

-                                 int             y,

-                                 int             width,

-                                 const uint32_t *buffer)

-{

-    image->store_scanline_64 (image, x, y, width, buffer);

-

-    if (image->common.alpha_map)

-    {

-	x -= image->common.alpha_origin_x;

-	y -= image->common.alpha_origin_y;

-

-	image->common.alpha_map->store_scanline_64 (

-	    image->common.alpha_map, x, y, width, buffer);

-    }

-}

-

-/* Fetch functions */

-

-static force_inline uint32_t

-fetch_pixel_no_alpha (bits_image_t *image,

-		      int x, int y, pixman_bool_t check_bounds)

-{

-    if (check_bounds &&

-	(x < 0 || x >= image->width || y < 0 || y >= image->height))

-    {

-	return 0;

-    }

-

-    return image->fetch_pixel_32 (image, x, y);

-}

-

-typedef uint32_t (* get_pixel_t) (bits_image_t *image,

-				  int x, int y, pixman_bool_t check_bounds);

-

-static force_inline void

-repeat (pixman_repeat_t repeat, int size, int *coord)

-{

-    switch (repeat)

-    {

-    case PIXMAN_REPEAT_NORMAL:

-	*coord = MOD (*coord, size);

-	break;

-

-    case PIXMAN_REPEAT_PAD:

-	*coord = CLIP (*coord, 0, size - 1);

-	break;

-

-    case PIXMAN_REPEAT_REFLECT:

-	*coord = MOD (*coord, size * 2);

-

-	if (*coord >= size)

-	    *coord = size * 2 - *coord - 1;

-	break;

-

-    case PIXMAN_REPEAT_NONE:

-	break;

-

-    default:

-        break;

-    }

-}

-

-static force_inline uint32_t

-bits_image_fetch_pixel_nearest (bits_image_t   *image,

-				pixman_fixed_t  x,

-				pixman_fixed_t  y,

-				get_pixel_t	get_pixel)

-{

-    int x0 = pixman_fixed_to_int (x - pixman_fixed_e);

-    int y0 = pixman_fixed_to_int (y - pixman_fixed_e);

-

-    if (image->common.repeat != PIXMAN_REPEAT_NONE)

-    {

-	repeat (image->common.repeat, image->width, &x0);

-	repeat (image->common.repeat, image->height, &y0);

-

-	return get_pixel (image, x0, y0, FALSE);

-    }

-    else

-    {

-	return get_pixel (image, x0, y0, TRUE);

-    }

-}

-

-#if SIZEOF_LONG > 4

-

-static force_inline uint32_t

-bilinear_interpolation (uint32_t tl, uint32_t tr,

-			uint32_t bl, uint32_t br,

-			int distx, int disty)

-{

-    uint64_t distxy, distxiy, distixy, distixiy;

-    uint64_t tl64, tr64, bl64, br64;

-    uint64_t f, r;

-

-    distxy = distx * disty;

-    distxiy = distx * (256 - disty);

-    distixy = (256 - distx) * disty;

-    distixiy = (256 - distx) * (256 - disty);

-

-    /* Alpha and Blue */

-    tl64 = tl & 0xff0000ff;

-    tr64 = tr & 0xff0000ff;

-    bl64 = bl & 0xff0000ff;

-    br64 = br & 0xff0000ff;

-

-    f = tl64 * distixiy + tr64 * distxiy + bl64 * distixy + br64 * distxy;

-    r = f & 0x0000ff0000ff0000ull;

-

-    /* Red and Green */

-    tl64 = tl;

-    tl64 = ((tl64 << 16) & 0x000000ff00000000ull) | (tl64 & 0x0000ff00ull);

-

-    tr64 = tr;

-    tr64 = ((tr64 << 16) & 0x000000ff00000000ull) | (tr64 & 0x0000ff00ull);

-

-    bl64 = bl;

-    bl64 = ((bl64 << 16) & 0x000000ff00000000ull) | (bl64 & 0x0000ff00ull);

-

-    br64 = br;

-    br64 = ((br64 << 16) & 0x000000ff00000000ull) | (br64 & 0x0000ff00ull);

-

-    f = tl64 * distixiy + tr64 * distxiy + bl64 * distixy + br64 * distxy;

-    r |= ((f >> 16) & 0x000000ff00000000ull) | (f & 0xff000000ull);

-

-    return (uint32_t)(r >> 16);

-}

-

-#else

-

-static force_inline uint32_t

-bilinear_interpolation (uint32_t tl, uint32_t tr,

-			uint32_t bl, uint32_t br,

-			int distx, int disty)

-{

-    int distxy, distxiy, distixy, distixiy;

-    uint32_t f, r;

-

-    distxy = distx * disty;

-    distxiy = (distx << 8) - distxy;	/* distx * (256 - disty) */

-    distixy = (disty << 8) - distxy;	/* disty * (256 - distx) */

-    distixiy =

-	256 * 256 - (disty << 8) -

-	(distx << 8) + distxy;		/* (256 - distx) * (256 - disty) */

-

-    /* Blue */

-    r = (tl & 0x000000ff) * distixiy + (tr & 0x000000ff) * distxiy

-      + (bl & 0x000000ff) * distixy  + (br & 0x000000ff) * distxy;

-

-    /* Green */

-    f = (tl & 0x0000ff00) * distixiy + (tr & 0x0000ff00) * distxiy

-      + (bl & 0x0000ff00) * distixy  + (br & 0x0000ff00) * distxy;

-    r |= f & 0xff000000;

-

-    tl >>= 16;

-    tr >>= 16;

-    bl >>= 16;

-    br >>= 16;

-    r >>= 16;

-

-    /* Red */

-    f = (tl & 0x000000ff) * distixiy + (tr & 0x000000ff) * distxiy

-      + (bl & 0x000000ff) * distixy  + (br & 0x000000ff) * distxy;

-    r |= f & 0x00ff0000;

-

-    /* Alpha */

-    f = (tl & 0x0000ff00) * distixiy + (tr & 0x0000ff00) * distxiy

-      + (bl & 0x0000ff00) * distixy  + (br & 0x0000ff00) * distxy;

-    r |= f & 0xff000000;

-

-    return r;

-}

-

-#endif

-

-static force_inline uint32_t

-bits_image_fetch_pixel_bilinear (bits_image_t   *image,

-				 pixman_fixed_t  x,

-				 pixman_fixed_t  y,

-				 get_pixel_t	 get_pixel)

-{

-    pixman_repeat_t repeat_mode = image->common.repeat;

-    int width = image->width;

-    int height = image->height;

-    int x1, y1, x2, y2;

-    uint32_t tl, tr, bl, br;

-    int32_t distx, disty;

-

-    x1 = x - pixman_fixed_1 / 2;

-    y1 = y - pixman_fixed_1 / 2;

-

-    distx = (x1 >> 8) & 0xff;

-    disty = (y1 >> 8) & 0xff;

-

-    x1 = pixman_fixed_to_int (x1);

-    y1 = pixman_fixed_to_int (y1);

-    x2 = x1 + 1;

-    y2 = y1 + 1;

-

-    if (repeat_mode != PIXMAN_REPEAT_NONE)

-    {

-	repeat (repeat_mode, width, &x1);

-	repeat (repeat_mode, height, &y1);

-	repeat (repeat_mode, width, &x2);

-	repeat (repeat_mode, height, &y2);

-

-	tl = get_pixel (image, x1, y1, FALSE);

-	bl = get_pixel (image, x1, y2, FALSE);

-	tr = get_pixel (image, x2, y1, FALSE);

-	br = get_pixel (image, x2, y2, FALSE);

-    }

-    else

-    {

-	tl = get_pixel (image, x1, y1, TRUE);

-	tr = get_pixel (image, x2, y1, TRUE);

-	bl = get_pixel (image, x1, y2, TRUE);

-	br = get_pixel (image, x2, y2, TRUE);

-    }

-

-    return bilinear_interpolation (tl, tr, bl, br, distx, disty);

-}

-

-static void

-bits_image_fetch_bilinear_no_repeat_8888 (pixman_image_t * ima,

-					  int              offset,

-					  int              line,

-					  int              width,

-					  uint32_t *       buffer,

-					  const uint32_t * mask)

-{

-    bits_image_t *bits = &ima->bits;

-    pixman_fixed_t x_top, x_bottom, x;

-    pixman_fixed_t ux_top, ux_bottom, ux;

-    pixman_vector_t v;

-    uint32_t top_mask, bottom_mask;

-    uint32_t *top_row;

-    uint32_t *bottom_row;

-    uint32_t *end;

-    uint32_t zero[2] = { 0, 0 };

-    int y, y1, y2;

-    int disty;

-    int mask_inc;

-    int w;

-

-    /* reference point is the center of the pixel */

-    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;

-    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;

-    v.vector[2] = pixman_fixed_1;

-

-    if (!pixman_transform_point_3d (bits->common.transform, &v))

-	return;

-

-    ux = ux_top = ux_bottom = bits->common.transform->matrix[0][0];

-    x = x_top = x_bottom = v.vector[0] - pixman_fixed_1/2;

-

-    y = v.vector[1] - pixman_fixed_1/2;

-    disty = (y >> 8) & 0xff;

-

-    /* Load the pointers to the first and second lines from the source

-     * image that bilinear code must read.

-     *

-     * The main trick in this code is about the check if any line are

-     * outside of the image;

-     *

-     * When I realize that a line (any one) is outside, I change

-     * the pointer to a dummy area with zeros. Once I change this, I

-     * must be sure the pointer will not change, so I set the

-     * variables to each pointer increments inside the loop.

-     */

-    y1 = pixman_fixed_to_int (y);

-    y2 = y1 + 1;

-

-    if (y1 < 0 || y1 >= bits->height)

-    {

-	top_row = zero;

-	x_top = 0;

-	ux_top = 0;

-    }

-    else

-    {

-	top_row = bits->bits + y1 * bits->rowstride;

-	x_top = x;

-	ux_top = ux;

-    }

-

-    if (y2 < 0 || y2 >= bits->height)

-    {

-	bottom_row = zero;

-	x_bottom = 0;

-	ux_bottom = 0;

-    }

-    else

-    {

-	bottom_row = bits->bits + y2 * bits->rowstride;

-	x_bottom = x;

-	ux_bottom = ux;

-    }

-

-    /* Instead of checking whether the operation uses the mast in

-     * each loop iteration, verify this only once and prepare the

-     * variables to make the code smaller inside the loop.

-     */

-    if (!mask)

-    {

-	uint32_t mask_bits = 1;

-

-        mask_inc = 0;

-        mask = &mask_bits;

-    }

-    else

-    {

-        /* If have a mask, prepare the variables to check it */

-        mask_inc = 1;

-    }

-

-    /* If both are zero, then the whole thing is zero */

-    if (top_row == zero && bottom_row == zero)

-    {

-	memset (buffer, 0, width * sizeof (uint32_t));

-	return;

-    }

-    else if (bits->format == PIXMAN_x8r8g8b8)

-    {

-	if (top_row == zero)

-	{

-	    top_mask = 0;

-	    bottom_mask = 0xff000000;

-	}

-	else if (bottom_row == zero)

-	{

-	    top_mask = 0xff000000;

-	    bottom_mask = 0;

-	}

-	else

-	{

-	    top_mask = 0xff000000;

-	    bottom_mask = 0xff000000;

-	}

-    }

-    else

-    {

-	top_mask = 0;

-	bottom_mask = 0;

-    }

-

-    end = buffer + width;

-

-    /* Zero fill to the left of the image */

-    while (buffer < end && x < pixman_fixed_minus_1)

-    {

-	*buffer++ = 0;

-	x += ux;

-	x_top += ux_top;

-	x_bottom += ux_bottom;

-	mask += mask_inc;

-    }

-

-    /* Left edge

-     */

-    while (buffer < end && x < 0)

-    {

-	uint32_t tr, br;

-	int32_t distx;

-

-	tr = top_row[pixman_fixed_to_int (x_top) + 1] | top_mask;

-	br = bottom_row[pixman_fixed_to_int (x_bottom) + 1] | bottom_mask;

-

-	distx = (x >> 8) & 0xff;

-

-	*buffer++ = bilinear_interpolation (0, tr, 0, br, distx, disty);

-

-	x += ux;

-	x_top += ux_top;

-	x_bottom += ux_bottom;

-	mask += mask_inc;

-    }

-

-    /* Main part */

-    w = pixman_int_to_fixed (bits->width - 1);

-

-    while (buffer < end  &&  x < w)

-    {

-	if (*mask)

-	{

-	    uint32_t tl, tr, bl, br;

-	    int32_t distx;

-

-	    tl = top_row [pixman_fixed_to_int (x_top)] | top_mask;

-	    tr = top_row [pixman_fixed_to_int (x_top) + 1] | top_mask;

-	    bl = bottom_row [pixman_fixed_to_int (x_bottom)] | bottom_mask;

-	    br = bottom_row [pixman_fixed_to_int (x_bottom) + 1] | bottom_mask;

-

-	    distx = (x >> 8) & 0xff;

-

-	    *buffer = bilinear_interpolation (tl, tr, bl, br, distx, disty);

-	}

-

-	buffer++;

-	x += ux;

-	x_top += ux_top;

-	x_bottom += ux_bottom;

-	mask += mask_inc;

-    }

-

-    /* Right Edge */

-    w = pixman_int_to_fixed (bits->width);

-    while (buffer < end  &&  x < w)

-    {

-	if (*mask)

-	{

-	    uint32_t tl, bl;

-	    int32_t distx;

-

-	    tl = top_row [pixman_fixed_to_int (x_top)] | top_mask;

-	    bl = bottom_row [pixman_fixed_to_int (x_bottom)] | bottom_mask;

-

-	    distx = (x >> 8) & 0xff;

-

-	    *buffer = bilinear_interpolation (tl, 0, bl, 0, distx, disty);

-	}

-

-	buffer++;

-	x += ux;

-	x_top += ux_top;

-	x_bottom += ux_bottom;

-	mask += mask_inc;

-    }

-

-    /* Zero fill to the left of the image */

-    while (buffer < end)

-	*buffer++ = 0;

-}

-

-static force_inline uint32_t

-bits_image_fetch_pixel_convolution (bits_image_t   *image,

-				    pixman_fixed_t  x,

-				    pixman_fixed_t  y,

-				    get_pixel_t     get_pixel)

-{

-    pixman_fixed_t *params = image->common.filter_params;

-    int x_off = (params[0] - pixman_fixed_1) >> 1;

-    int y_off = (params[1] - pixman_fixed_1) >> 1;

-    int32_t cwidth = pixman_fixed_to_int (params[0]);

-    int32_t cheight = pixman_fixed_to_int (params[1]);

-    int32_t srtot, sgtot, sbtot, satot;

-    int32_t i, j, x1, x2, y1, y2;

-    pixman_repeat_t repeat_mode = image->common.repeat;

-    int width = image->width;

-    int height = image->height;

-

-    params += 2;

-

-    x1 = pixman_fixed_to_int (x - pixman_fixed_e - x_off);

-    y1 = pixman_fixed_to_int (y - pixman_fixed_e - y_off);

-    x2 = x1 + cwidth;

-    y2 = y1 + cheight;

-

-    srtot = sgtot = sbtot = satot = 0;

-

-    for (i = y1; i < y2; ++i)

-    {

-	for (j = x1; j < x2; ++j)

-	{

-	    int rx = j;

-	    int ry = i;

-

-	    pixman_fixed_t f = *params;

-

-	    if (f)

-	    {

-		uint32_t pixel;

-

-		if (repeat_mode != PIXMAN_REPEAT_NONE)

-		{

-		    repeat (repeat_mode, width, &rx);

-		    repeat (repeat_mode, height, &ry);

-

-		    pixel = get_pixel (image, rx, ry, FALSE);

-		}

-		else

-		{

-		    pixel = get_pixel (image, rx, ry, TRUE);

-		}

-

-		srtot += RED_8 (pixel) * f;

-		sgtot += GREEN_8 (pixel) * f;

-		sbtot += BLUE_8 (pixel) * f;

-		satot += ALPHA_8 (pixel) * f;

-	    }

-

-	    params++;

-	}

-    }

-

-    satot >>= 16;

-    srtot >>= 16;

-    sgtot >>= 16;

-    sbtot >>= 16;

-

-    satot = CLIP (satot, 0, 0xff);

-    srtot = CLIP (srtot, 0, 0xff);

-    sgtot = CLIP (sgtot, 0, 0xff);

-    sbtot = CLIP (sbtot, 0, 0xff);

-

-    return ((satot << 24) | (srtot << 16) | (sgtot <<  8) | (sbtot));

-}

-

-static force_inline uint32_t

-bits_image_fetch_pixel_filtered (bits_image_t *image,

-				 pixman_fixed_t x,

-				 pixman_fixed_t y,

-				 get_pixel_t    get_pixel)

-{

-    switch (image->common.filter)

-    {

-    case PIXMAN_FILTER_NEAREST:

-    case PIXMAN_FILTER_FAST:

-	return bits_image_fetch_pixel_nearest (image, x, y, get_pixel);

-	break;

-

-    case PIXMAN_FILTER_BILINEAR:

-    case PIXMAN_FILTER_GOOD:

-    case PIXMAN_FILTER_BEST:

-	return bits_image_fetch_pixel_bilinear (image, x, y, get_pixel);

-	break;

-

-    case PIXMAN_FILTER_CONVOLUTION:

-	return bits_image_fetch_pixel_convolution (image, x, y, get_pixel);

-	break;

-

-    default:

-        break;

-    }

-

-    return 0;

-}

-

-static void

-bits_image_fetch_affine_no_alpha (pixman_image_t * image,

-				  int              offset,

-				  int              line,

-				  int              width,

-				  uint32_t *       buffer,

-				  const uint32_t * mask)

-{

-    pixman_fixed_t x, y;

-    pixman_fixed_t ux, uy;

-    pixman_vector_t v;

-    int i;

-

-    /* reference point is the center of the pixel */

-    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;

-    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;

-    v.vector[2] = pixman_fixed_1;

-

-    if (image->common.transform)

-    {

-	if (!pixman_transform_point_3d (image->common.transform, &v))

-	    return;

-

-	ux = image->common.transform->matrix[0][0];

-	uy = image->common.transform->matrix[1][0];

-    }

-    else

-    {

-	ux = pixman_fixed_1;

-	uy = 0;

-    }

-

-    x = v.vector[0];

-    y = v.vector[1];

-

-    for (i = 0; i < width; ++i)

-    {

-	if (!mask || mask[i])

-	{

-	    buffer[i] = bits_image_fetch_pixel_filtered (

-		&image->bits, x, y, fetch_pixel_no_alpha);

-	}

-

-	x += ux;

-	y += uy;

-    }

-}

-

-/* General fetcher */

-static force_inline uint32_t

-fetch_pixel_general (bits_image_t *image, int x, int y, pixman_bool_t check_bounds)

-{

-    uint32_t pixel;

-

-    if (check_bounds &&

-	(x < 0 || x >= image->width || y < 0 || y >= image->height))

-    {

-	return 0;

-    }

-

-    pixel = image->fetch_pixel_32 (image, x, y);

-

-    if (image->common.alpha_map)

-    {

-	uint32_t pixel_a;

-

-	x -= image->common.alpha_origin_x;

-	y -= image->common.alpha_origin_y;

-

-	if (x < 0 || x >= image->common.alpha_map->width ||

-	    y < 0 || y >= image->common.alpha_map->height)

-	{

-	    pixel_a = 0;

-	}

-	else

-	{

-	    pixel_a = image->common.alpha_map->fetch_pixel_32 (

-		image->common.alpha_map, x, y);

-

-	    pixel_a = ALPHA_8 (pixel_a);

-	}

-

-	pixel &= 0x00ffffff;

-	pixel |= (pixel_a << 24);

-    }

-

-    return pixel;

-}

-

-static void

-bits_image_fetch_general (pixman_image_t * image,

-			  int              offset,

-			  int              line,

-			  int              width,

-			  uint32_t *       buffer,

-			  const uint32_t * mask)

-{

-    pixman_fixed_t x, y, w;

-    pixman_fixed_t ux, uy, uw;

-    pixman_vector_t v;

-    int i;

-

-    /* reference point is the center of the pixel */

-    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;

-    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;

-    v.vector[2] = pixman_fixed_1;

-

-    if (image->common.transform)

-    {

-	if (!pixman_transform_point_3d (image->common.transform, &v))

-	    return;

-

-	ux = image->common.transform->matrix[0][0];

-	uy = image->common.transform->matrix[1][0];

-	uw = image->common.transform->matrix[2][0];

-    }

-    else

-    {

-	ux = pixman_fixed_1;

-	uy = 0;

-	uw = 0;

-    }

-

-    x = v.vector[0];

-    y = v.vector[1];

-    w = v.vector[2];

-

-    for (i = 0; i < width; ++i)

-    {

-	pixman_fixed_t x0, y0;

-

-	if (!mask || mask[i])

-	{

-	    if (w != 0)

-	    {

-		x0 = ((pixman_fixed_48_16_t)x << 16) / w;

-		y0 = ((pixman_fixed_48_16_t)y << 16) / w;

-	    }

-	    else

-	    {

-		x0 = 0;

-		y0 = 0;

-	    }

-

-	    buffer[i] = bits_image_fetch_pixel_filtered (

-		&image->bits, x0, y0, fetch_pixel_general);

-	}

-

-	x += ux;

-	y += uy;

-	w += uw;

-    }

-}

-

-static const uint8_t zero[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };

-

-typedef uint32_t (* convert_pixel_t) (const uint8_t *row, int x);

-

-static force_inline void

-bits_image_fetch_bilinear_affine (pixman_image_t * image,

-				  int              offset,

-				  int              line,

-				  int              width,

-				  uint32_t *       buffer,

-				  const uint32_t * mask,

-

-				  convert_pixel_t	convert_pixel,

-				  pixman_format_code_t	format,

-				  pixman_repeat_t	repeat_mode)

-{

-    pixman_fixed_t x, y;

-    pixman_fixed_t ux, uy;

-    pixman_vector_t v;

-    bits_image_t *bits = &image->bits;

-    int i;

-

-    /* reference point is the center of the pixel */

-    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;

-    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;

-    v.vector[2] = pixman_fixed_1;

-

-    if (!pixman_transform_point_3d (image->common.transform, &v))

-	return;

-

-    ux = image->common.transform->matrix[0][0];

-    uy = image->common.transform->matrix[1][0];

-

-    x = v.vector[0];

-    y = v.vector[1];

-

-    for (i = 0; i < width; ++i)

-    {

-	int x1, y1, x2, y2;

-	uint32_t tl, tr, bl, br;

-	int32_t distx, disty;

-	int width = image->bits.width;

-	int height = image->bits.height;

-	const uint8_t *row1;

-	const uint8_t *row2;

-

-	if (mask && !mask[i])

-	    goto next;

-

-	x1 = x - pixman_fixed_1 / 2;

-	y1 = y - pixman_fixed_1 / 2;

-

-	distx = (x1 >> 8) & 0xff;

-	disty = (y1 >> 8) & 0xff;

-

-	y1 = pixman_fixed_to_int (y1);

-	y2 = y1 + 1;

-	x1 = pixman_fixed_to_int (x1);

-	x2 = x1 + 1;

-

-	if (repeat_mode != PIXMAN_REPEAT_NONE)

-	{

-	    uint32_t mask;

-

-	    mask = PIXMAN_FORMAT_A (format)? 0 : 0xff000000;

-

-	    repeat (repeat_mode, width, &x1);

-	    repeat (repeat_mode, height, &y1);

-	    repeat (repeat_mode, width, &x2);

-	    repeat (repeat_mode, height, &y2);

-

-	    row1 = (uint8_t *)bits->bits + bits->rowstride * 4 * y1;

-	    row2 = (uint8_t *)bits->bits + bits->rowstride * 4 * y2;

-

-	    tl = convert_pixel (row1, x1) | mask;

-	    tr = convert_pixel (row1, x2) | mask;

-	    bl = convert_pixel (row2, x1) | mask;

-	    br = convert_pixel (row2, x2) | mask;

-	}

-	else

-	{

-	    uint32_t mask1, mask2;

-	    int bpp;

-

-	    /* Note: PIXMAN_FORMAT_BPP() returns an unsigned value,

-	     * which means if you use it in expressions, those

-	     * expressions become unsigned themselves. Since

-	     * the variables below can be negative in some cases,

-	     * that will lead to crashes on 64 bit architectures.

-	     *

-	     * So this line makes sure bpp is signed

-	     */

-	    bpp = PIXMAN_FORMAT_BPP (format);

-

-	    if (x1 >= width || x2 < 0 || y1 >= height || y2 < 0)

-	    {

-		buffer[i] = 0;

-		goto next;

-	    }

-

-	    if (y2 == 0)

-	    {

-		row1 = zero;

-		mask1 = 0;

-	    }

-	    else

-	    {

-		row1 = (uint8_t *)bits->bits + bits->rowstride * 4 * y1;

-		row1 += bpp / 8 * x1;

-

-		mask1 = PIXMAN_FORMAT_A (format)? 0 : 0xff000000;

-	    }

-

-	    if (y1 == height - 1)

-	    {

-		row2 = zero;

-		mask2 = 0;

-	    }

-	    else

-	    {

-		row2 = (uint8_t *)bits->bits + bits->rowstride * 4 * y2;

-		row2 += bpp / 8 * x1;

-

-		mask2 = PIXMAN_FORMAT_A (format)? 0 : 0xff000000;

-	    }

-

-	    if (x2 == 0)

-	    {

-		tl = 0;

-		bl = 0;

-	    }

-	    else

-	    {

-		tl = convert_pixel (row1, 0) | mask1;

-		bl = convert_pixel (row2, 0) | mask2;

-	    }

-

-	    if (x1 == width - 1)

-	    {

-		tr = 0;

-		br = 0;

-	    }

-	    else

-	    {

-		tr = convert_pixel (row1, 1) | mask1;

-		br = convert_pixel (row2, 1) | mask2;

-	    }

-	}

-

-	buffer[i] = bilinear_interpolation (

-	    tl, tr, bl, br, distx, disty);

-

-    next:

-	x += ux;

-	y += uy;

-    }

-}

-

-static force_inline void

-bits_image_fetch_nearest_affine (pixman_image_t * image,

-				 int              offset,

-				 int              line,

-				 int              width,

-				 uint32_t *       buffer,

-				 const uint32_t * mask,

-				 

-				 convert_pixel_t	convert_pixel,

-				 pixman_format_code_t	format,

-				 pixman_repeat_t	repeat_mode)

-{

-    pixman_fixed_t x, y;

-    pixman_fixed_t ux, uy;

-    pixman_vector_t v;

-    bits_image_t *bits = &image->bits;

-    int i;

-

-    /* reference point is the center of the pixel */

-    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;

-    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;

-    v.vector[2] = pixman_fixed_1;

-

-    if (!pixman_transform_point_3d (image->common.transform, &v))

-	return;

-

-    ux = image->common.transform->matrix[0][0];

-    uy = image->common.transform->matrix[1][0];

-

-    x = v.vector[0];

-    y = v.vector[1];

-

-    for (i = 0; i < width; ++i)

-    {

-	int width, height, x0, y0;

-	const uint8_t *row;

-

-	if (mask && !mask[i])

-	    goto next;

-	

-	width = image->bits.width;

-	height = image->bits.height;

-	x0 = pixman_fixed_to_int (x - pixman_fixed_e);

-	y0 = pixman_fixed_to_int (y - pixman_fixed_e);

-

-	if (repeat_mode == PIXMAN_REPEAT_NONE &&

-	    (y0 < 0 || y0 >= height || x0 < 0 || x0 >= width))

-	{

-	    buffer[i] = 0;

-	}

-	else

-	{

-	    uint32_t mask = PIXMAN_FORMAT_A (format)? 0 : 0xff000000;

-

-	    if (repeat_mode != PIXMAN_REPEAT_NONE)

-	    {

-		repeat (repeat_mode, width, &x0);

-		repeat (repeat_mode, height, &y0);

-	    }

-

-	    row = (uint8_t *)bits->bits + bits->rowstride * 4 * y0;

-

-	    buffer[i] = convert_pixel (row, x0) | mask;

-	}

-

-    next:

-	x += ux;

-	y += uy;

-    }

-}

-

-static force_inline uint32_t

-convert_a8r8g8b8 (const uint8_t *row, int x)

-{

-    return *(((uint32_t *)row) + x);

-}

-

-static force_inline uint32_t

-convert_x8r8g8b8 (const uint8_t *row, int x)

-{

-    return *(((uint32_t *)row) + x);

-}

-

-static force_inline uint32_t

-convert_a8 (const uint8_t *row, int x)

-{

-    return *(row + x) << 24;

-}

-

-static force_inline uint32_t

-convert_r5g6b5 (const uint8_t *row, int x)

-{

-    return CONVERT_0565_TO_0888 (*((uint16_t *)row + x));

-}

-

-#define MAKE_BILINEAR_FETCHER(name, format, repeat_mode)		\

-    static void								\

-    bits_image_fetch_bilinear_affine_ ## name (pixman_image_t *image,	\

-					       int              offset,	\

-					       int              line,	\

-					       int              width,	\

-					       uint32_t *       buffer,	\

-					       const uint32_t * mask)	\

-    {									\

-	bits_image_fetch_bilinear_affine (image, offset, line,		\

-					  width, buffer, mask,		\

-					  convert_ ## format,		\

-					  PIXMAN_ ## format,		\

-					  repeat_mode);			\

-    }									\

-    extern int no_such_variable

-

-#define MAKE_NEAREST_FETCHER(name, format, repeat_mode)			\

-    static void								\

-    bits_image_fetch_nearest_affine_ ## name (pixman_image_t *image,	\

-					      int              offset,	\

-					      int              line,	\

-					      int              width,	\

-					      uint32_t *       buffer,	\

-					      const uint32_t * mask)	\

-    {									\

-	bits_image_fetch_nearest_affine (image, offset, line,		\

-					 width, buffer, mask,		\

-					 convert_ ## format,		\

-					 PIXMAN_ ## format,		\

-					 repeat_mode);			\

-    }									\

-    extern int no_such_variable

-

-#define MAKE_FETCHERS(name, format, repeat_mode)			\

-    MAKE_NEAREST_FETCHER (name, format, repeat_mode);			\

-    MAKE_BILINEAR_FETCHER (name, format, repeat_mode);

-

-MAKE_FETCHERS (pad_a8r8g8b8,     a8r8g8b8, PIXMAN_REPEAT_PAD);

-MAKE_FETCHERS (none_a8r8g8b8,    a8r8g8b8, PIXMAN_REPEAT_NONE);

-MAKE_FETCHERS (reflect_a8r8g8b8, a8r8g8b8, PIXMAN_REPEAT_REFLECT);

-MAKE_FETCHERS (normal_a8r8g8b8,  a8r8g8b8, PIXMAN_REPEAT_NORMAL);

-MAKE_FETCHERS (pad_x8r8g8b8,     x8r8g8b8, PIXMAN_REPEAT_PAD);

-MAKE_FETCHERS (none_x8r8g8b8,    x8r8g8b8, PIXMAN_REPEAT_NONE);

-MAKE_FETCHERS (reflect_x8r8g8b8, x8r8g8b8, PIXMAN_REPEAT_REFLECT);

-MAKE_FETCHERS (normal_x8r8g8b8,  x8r8g8b8, PIXMAN_REPEAT_NORMAL);

-MAKE_FETCHERS (pad_a8,           a8,       PIXMAN_REPEAT_PAD);

-MAKE_FETCHERS (none_a8,          a8,       PIXMAN_REPEAT_NONE);

-MAKE_FETCHERS (reflect_a8,	 a8,       PIXMAN_REPEAT_REFLECT);

-MAKE_FETCHERS (normal_a8,	 a8,       PIXMAN_REPEAT_NORMAL);

-MAKE_FETCHERS (pad_r5g6b5,       r5g6b5,   PIXMAN_REPEAT_PAD);

-MAKE_FETCHERS (none_r5g6b5,      r5g6b5,   PIXMAN_REPEAT_NONE);

-MAKE_FETCHERS (reflect_r5g6b5,   r5g6b5,   PIXMAN_REPEAT_REFLECT);

-MAKE_FETCHERS (normal_r5g6b5,    r5g6b5,   PIXMAN_REPEAT_NORMAL);

-

-static void

-bits_image_fetch_solid_32 (pixman_image_t * image,

-                           int              x,

-                           int              y,

-                           int              width,

-                           uint32_t *       buffer,

-                           const uint32_t * mask)

-{

-    uint32_t color;

-    uint32_t *end;

-

-    color = image->bits.fetch_pixel_32 (&image->bits, 0, 0);

-

-    end = buffer + width;

-    while (buffer < end)

-	*(buffer++) = color;

-}

-

-static void

-bits_image_fetch_solid_64 (pixman_image_t * image,

-                           int              x,

-                           int              y,

-                           int              width,

-                           uint32_t *       b,

-                           const uint32_t * unused)

-{

-    uint64_t color;

-    uint64_t *buffer = (uint64_t *)b;

-    uint64_t *end;

-

-    color = image->bits.fetch_pixel_64 (&image->bits, 0, 0);

-

-    end = buffer + width;

-    while (buffer < end)

-	*(buffer++) = color;

-}

-

-static void

-bits_image_fetch_untransformed_repeat_none (bits_image_t *image,

-                                            pixman_bool_t wide,

-                                            int           x,

-                                            int           y,

-                                            int           width,

-                                            uint32_t *    buffer)

-{

-    uint32_t w;

-

-    if (y < 0 || y >= image->height)

-    {

-	memset (buffer, 0, width * (wide? 8 : 4));

-	return;

-    }

-

-    if (x < 0)

-    {

-	w = MIN (width, -x);

-

-	memset (buffer, 0, w * (wide ? 8 : 4));

-

-	width -= w;

-	buffer += w * (wide? 2 : 1);

-	x += w;

-    }

-

-    if (x < image->width)

-    {

-	w = MIN (width, image->width - x);

-

-	if (wide)

-	    image->fetch_scanline_64 ((pixman_image_t *)image, x, y, w, buffer, NULL);

-	else

-	    image->fetch_scanline_32 ((pixman_image_t *)image, x, y, w, buffer, NULL);

-

-	width -= w;

-	buffer += w * (wide? 2 : 1);

-	x += w;

-    }

-

-    memset (buffer, 0, width * (wide ? 8 : 4));

-}

-

-static void

-bits_image_fetch_untransformed_repeat_normal (bits_image_t *image,

-                                              pixman_bool_t wide,

-                                              int           x,

-                                              int           y,

-                                              int           width,

-                                              uint32_t *    buffer)

-{

-    uint32_t w;

-

-    while (y < 0)

-	y += image->height;

-

-    while (y >= image->height)

-	y -= image->height;

-

-    while (width)

-    {

-	while (x < 0)

-	    x += image->width;

-	while (x >= image->width)

-	    x -= image->width;

-

-	w = MIN (width, image->width - x);

-

-	if (wide)

-	    image->fetch_scanline_64 ((pixman_image_t *)image, x, y, w, buffer, NULL);

-	else

-	    image->fetch_scanline_32 ((pixman_image_t *)image, x, y, w, buffer, NULL);

-

-	buffer += w * (wide? 2 : 1);

-	x += w;

-	width -= w;

-    }

-}

-

-static void

-bits_image_fetch_untransformed_32 (pixman_image_t * image,

-                                   int              x,

-                                   int              y,

-                                   int              width,

-                                   uint32_t *       buffer,

-                                   const uint32_t * mask)

-{

-    if (image->common.repeat == PIXMAN_REPEAT_NONE)

-    {

-	bits_image_fetch_untransformed_repeat_none (

-	    &image->bits, FALSE, x, y, width, buffer);

-    }

-    else

-    {

-	bits_image_fetch_untransformed_repeat_normal (

-	    &image->bits, FALSE, x, y, width, buffer);

-    }

-}

-

-static void

-bits_image_fetch_untransformed_64 (pixman_image_t * image,

-                                   int              x,

-                                   int              y,

-                                   int              width,

-                                   uint32_t *       buffer,

-                                   const uint32_t * unused)

-{

-    if (image->common.repeat == PIXMAN_REPEAT_NONE)

-    {

-	bits_image_fetch_untransformed_repeat_none (

-	    &image->bits, TRUE, x, y, width, buffer);

-    }

-    else

-    {

-	bits_image_fetch_untransformed_repeat_normal (

-	    &image->bits, TRUE, x, y, width, buffer);

-    }

-}

-

-typedef struct

-{

-    pixman_format_code_t	format;

-    uint32_t			flags;

-    fetch_scanline_t		fetch_32;

-    fetch_scanline_t		fetch_64;

-} fetcher_info_t;

-

-static const fetcher_info_t fetcher_info[] =

-{

-    { PIXMAN_solid,

-      FAST_PATH_NO_ALPHA_MAP,

-      bits_image_fetch_solid_32,

-      bits_image_fetch_solid_64

-    },

-

-    { PIXMAN_any,

-      (FAST_PATH_NO_ALPHA_MAP			|

-       FAST_PATH_ID_TRANSFORM			|

-       FAST_PATH_NO_CONVOLUTION_FILTER		|

-       FAST_PATH_NO_PAD_REPEAT			|

-       FAST_PATH_NO_REFLECT_REPEAT),

-      bits_image_fetch_untransformed_32,

-      bits_image_fetch_untransformed_64

-    },

-

-#define FAST_BILINEAR_FLAGS						\

-    (FAST_PATH_NO_ALPHA_MAP		|				\

-     FAST_PATH_NO_ACCESSORS		|				\

-     FAST_PATH_HAS_TRANSFORM		|				\

-     FAST_PATH_AFFINE_TRANSFORM		|				\

-     FAST_PATH_X_UNIT_POSITIVE		|				\

-     FAST_PATH_Y_UNIT_ZERO		|				\

-     FAST_PATH_NONE_REPEAT		|				\

-     FAST_PATH_BILINEAR_FILTER)

-

-    { PIXMAN_a8r8g8b8,

-      FAST_BILINEAR_FLAGS,

-      bits_image_fetch_bilinear_no_repeat_8888,

-      _pixman_image_get_scanline_generic_64

-    },

-

-    { PIXMAN_x8r8g8b8,

-      FAST_BILINEAR_FLAGS,

-      bits_image_fetch_bilinear_no_repeat_8888,

-      _pixman_image_get_scanline_generic_64

-    },

-

-#define GENERAL_BILINEAR_FLAGS						\

-    (FAST_PATH_NO_ALPHA_MAP		|				\

-     FAST_PATH_NO_ACCESSORS		|				\

-     FAST_PATH_HAS_TRANSFORM		|				\

-     FAST_PATH_AFFINE_TRANSFORM		|				\

-     FAST_PATH_BILINEAR_FILTER)

-

-#define GENERAL_NEAREST_FLAGS						\

-    (FAST_PATH_NO_ALPHA_MAP		|				\

-     FAST_PATH_NO_ACCESSORS		|				\

-     FAST_PATH_HAS_TRANSFORM		|				\

-     FAST_PATH_AFFINE_TRANSFORM		|				\

-     FAST_PATH_NEAREST_FILTER)

-

-#define BILINEAR_AFFINE_FAST_PATH(name, format, repeat)			\

-    { PIXMAN_ ## format,						\

-      GENERAL_BILINEAR_FLAGS | FAST_PATH_ ## repeat ## _REPEAT,		\

-      bits_image_fetch_bilinear_affine_ ## name,			\

-      _pixman_image_get_scanline_generic_64				\

-    },

-

-#define NEAREST_AFFINE_FAST_PATH(name, format, repeat)			\

-    { PIXMAN_ ## format,						\

-      GENERAL_NEAREST_FLAGS | FAST_PATH_ ## repeat ## _REPEAT,		\

-      bits_image_fetch_nearest_affine_ ## name,			\

-      _pixman_image_get_scanline_generic_64				\

-    },

-

-#define AFFINE_FAST_PATHS(name, format, repeat)				\

-    BILINEAR_AFFINE_FAST_PATH(name, format, repeat)			\

-    NEAREST_AFFINE_FAST_PATH(name, format, repeat)

-    

-    AFFINE_FAST_PATHS (pad_a8r8g8b8, a8r8g8b8, PAD)

-    AFFINE_FAST_PATHS (none_a8r8g8b8, a8r8g8b8, NONE)

-    AFFINE_FAST_PATHS (reflect_a8r8g8b8, a8r8g8b8, REFLECT)

-    AFFINE_FAST_PATHS (normal_a8r8g8b8, a8r8g8b8, NORMAL)

-    AFFINE_FAST_PATHS (pad_x8r8g8b8, x8r8g8b8, PAD)

-    AFFINE_FAST_PATHS (none_x8r8g8b8, x8r8g8b8, NONE)

-    AFFINE_FAST_PATHS (reflect_x8r8g8b8, x8r8g8b8, REFLECT)

-    AFFINE_FAST_PATHS (normal_x8r8g8b8, x8r8g8b8, NORMAL)

-    AFFINE_FAST_PATHS (pad_a8, a8, PAD)

-    AFFINE_FAST_PATHS (none_a8, a8, NONE)

-    AFFINE_FAST_PATHS (reflect_a8, a8, REFLECT)

-    AFFINE_FAST_PATHS (normal_a8, a8, NORMAL)

-    AFFINE_FAST_PATHS (pad_r5g6b5, r5g6b5, PAD)

-    AFFINE_FAST_PATHS (none_r5g6b5, r5g6b5, NONE)

-    AFFINE_FAST_PATHS (reflect_r5g6b5, r5g6b5, REFLECT)

-    AFFINE_FAST_PATHS (normal_r5g6b5, r5g6b5, NORMAL)

-

-    /* Affine, no alpha */

-    { PIXMAN_any,

-      (FAST_PATH_NO_ALPHA_MAP | FAST_PATH_HAS_TRANSFORM | FAST_PATH_AFFINE_TRANSFORM),

-      bits_image_fetch_affine_no_alpha,

-      _pixman_image_get_scanline_generic_64

-    },

-

-    /* General */

-    { PIXMAN_any, 0, bits_image_fetch_general, _pixman_image_get_scanline_generic_64 },

-

-    { PIXMAN_null },

-};

-

-static void

-bits_image_property_changed (pixman_image_t *image)

-{

-    uint32_t flags = image->common.flags;

-    pixman_format_code_t format = image->common.extended_format_code;

-    const fetcher_info_t *info;

-

-    _pixman_bits_image_setup_accessors (&image->bits);

-

-    info = fetcher_info;

-    while (info->format != PIXMAN_null)

-    {

-	if ((info->format == format || info->format == PIXMAN_any)	&&

-	    (info->flags & flags) == info->flags)

-	{

-	    image->common.get_scanline_32 = info->fetch_32;

-	    image->common.get_scanline_64 = info->fetch_64;

-	    break;

-	}

-

-	info++;

-    }

-}

-

-static uint32_t *

-create_bits (pixman_format_code_t format,

-             int                  width,

-             int                  height,

-             int *                rowstride_bytes)

-{

-    int stride;

-    int buf_size;

-    int bpp;

-

-    /* what follows is a long-winded way, avoiding any possibility of integer

-     * overflows, of saying:

-     * stride = ((width * bpp + 0x1f) >> 5) * sizeof (uint32_t);

-     */

-

-    bpp = PIXMAN_FORMAT_BPP (format);

-    if (pixman_multiply_overflows_int (width, bpp))

-	return NULL;

-

-    stride = width * bpp;

-    if (pixman_addition_overflows_int (stride, 0x1f))

-	return NULL;

-

-    stride += 0x1f;

-    stride >>= 5;

-

-    stride *= sizeof (uint32_t);

-

-    if (pixman_multiply_overflows_int (height, stride))

-	return NULL;

-

-    buf_size = height * stride;

-

-    if (rowstride_bytes)

-	*rowstride_bytes = stride;

-

-    return calloc (buf_size, 1);

-}

-

-PIXMAN_EXPORT pixman_image_t *

-pixman_image_create_bits (pixman_format_code_t format,

-                          int                  width,

-                          int                  height,

-                          uint32_t *           bits,

-                          int                  rowstride_bytes)

-{

-    pixman_image_t *image;

-    uint32_t *free_me = NULL;

-

-    /* must be a whole number of uint32_t's

-     */

-    return_val_if_fail (

-	bits == NULL || (rowstride_bytes % sizeof (uint32_t)) == 0, NULL);

-

-    return_val_if_fail (PIXMAN_FORMAT_BPP (format) >= PIXMAN_FORMAT_DEPTH (format), NULL);

-

-    if (!bits && width && height)

-    {

-	free_me = bits = create_bits (format, width, height, &rowstride_bytes);

-	if (!bits)

-	    return NULL;

-    }

-

-    image = _pixman_image_allocate ();

-

-    if (!image)

-    {

-	if (free_me)

-	    free (free_me);

-

-	return NULL;

-    }

-

-    image->type = BITS;

-    image->bits.format = format;

-    image->bits.width = width;

-    image->bits.height = height;

-    image->bits.bits = bits;

-    image->bits.free_me = free_me;

-    image->bits.read_func = NULL;

-    image->bits.write_func = NULL;

-

-    /* The rowstride is stored in number of uint32_t */

-    image->bits.rowstride = rowstride_bytes / (int) sizeof (uint32_t);

-

-    image->bits.indexed = NULL;

-

-    image->common.property_changed = bits_image_property_changed;

-

-    _pixman_image_reset_clip_region (image);

-

-    return image;

-}

+/*
+ * Copyright © 2000 Keith Packard, member of The XFree86 Project, Inc.
+ *             2005 Lars Knoll & Zack Rusin, Trolltech
+ *             2008 Aaron Plattner, NVIDIA Corporation
+ * Copyright © 2000 SuSE, Inc.
+ * Copyright © 2007, 2009 Red Hat, Inc.
+ * Copyright © 2008 André Tupinambá <andrelrt@gmail.com>
+ *
+ * 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 Keith Packard not be used in
+ * advertising or publicity pertaining to distribution of the software without
+ * specific, written prior permission.  Keith Packard 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.
+ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include "pixman-private.h"
+#include "pixman-combine32.h"
+
+/* Store functions */
+void
+_pixman_image_store_scanline_32 (bits_image_t *  image,
+                                 int             x,
+                                 int             y,
+                                 int             width,
+                                 const uint32_t *buffer)
+{
+    image->store_scanline_32 (image, x, y, width, buffer);
+
+    if (image->common.alpha_map)
+    {
+	x -= image->common.alpha_origin_x;
+	y -= image->common.alpha_origin_y;
+
+	image->common.alpha_map->store_scanline_32 (
+	    image->common.alpha_map, x, y, width, buffer);
+    }
+}
+
+void
+_pixman_image_store_scanline_64 (bits_image_t *  image,
+                                 int             x,
+                                 int             y,
+                                 int             width,
+                                 const uint32_t *buffer)
+{
+    image->store_scanline_64 (image, x, y, width, buffer);
+
+    if (image->common.alpha_map)
+    {
+	x -= image->common.alpha_origin_x;
+	y -= image->common.alpha_origin_y;
+
+	image->common.alpha_map->store_scanline_64 (
+	    image->common.alpha_map, x, y, width, buffer);
+    }
+}
+
+/* Fetch functions */
+
+static force_inline uint32_t
+fetch_pixel_no_alpha (bits_image_t *image,
+		      int x, int y, pixman_bool_t check_bounds)
+{
+    if (check_bounds &&
+	(x < 0 || x >= image->width || y < 0 || y >= image->height))
+    {
+	return 0;
+    }
+
+    return image->fetch_pixel_32 (image, x, y);
+}
+
+typedef uint32_t (* get_pixel_t) (bits_image_t *image,
+				  int x, int y, pixman_bool_t check_bounds);
+
+static force_inline void
+repeat (pixman_repeat_t repeat, int size, int *coord)
+{
+    switch (repeat)
+    {
+    case PIXMAN_REPEAT_NORMAL:
+	*coord = MOD (*coord, size);
+	break;
+
+    case PIXMAN_REPEAT_PAD:
+	*coord = CLIP (*coord, 0, size - 1);
+	break;
+
+    case PIXMAN_REPEAT_REFLECT:
+	*coord = MOD (*coord, size * 2);
+
+	if (*coord >= size)
+	    *coord = size * 2 - *coord - 1;
+	break;
+
+    case PIXMAN_REPEAT_NONE:
+	break;
+
+    default:
+        break;
+    }
+}
+
+static force_inline uint32_t
+bits_image_fetch_pixel_nearest (bits_image_t   *image,
+				pixman_fixed_t  x,
+				pixman_fixed_t  y,
+				get_pixel_t	get_pixel)
+{
+    int x0 = pixman_fixed_to_int (x - pixman_fixed_e);
+    int y0 = pixman_fixed_to_int (y - pixman_fixed_e);
+
+    if (image->common.repeat != PIXMAN_REPEAT_NONE)
+    {
+	repeat (image->common.repeat, image->width, &x0);
+	repeat (image->common.repeat, image->height, &y0);
+
+	return get_pixel (image, x0, y0, FALSE);
+    }
+    else
+    {
+	return get_pixel (image, x0, y0, TRUE);
+    }
+}
+
+#if SIZEOF_LONG > 4
+
+static force_inline uint32_t
+bilinear_interpolation (uint32_t tl, uint32_t tr,
+			uint32_t bl, uint32_t br,
+			int distx, int disty)
+{
+    uint64_t distxy, distxiy, distixy, distixiy;
+    uint64_t tl64, tr64, bl64, br64;
+    uint64_t f, r;
+
+    distxy = distx * disty;
+    distxiy = distx * (256 - disty);
+    distixy = (256 - distx) * disty;
+    distixiy = (256 - distx) * (256 - disty);
+
+    /* Alpha and Blue */
+    tl64 = tl & 0xff0000ff;
+    tr64 = tr & 0xff0000ff;
+    bl64 = bl & 0xff0000ff;
+    br64 = br & 0xff0000ff;
+
+    f = tl64 * distixiy + tr64 * distxiy + bl64 * distixy + br64 * distxy;
+    r = f & 0x0000ff0000ff0000ull;
+
+    /* Red and Green */
+    tl64 = tl;
+    tl64 = ((tl64 << 16) & 0x000000ff00000000ull) | (tl64 & 0x0000ff00ull);
+
+    tr64 = tr;
+    tr64 = ((tr64 << 16) & 0x000000ff00000000ull) | (tr64 & 0x0000ff00ull);
+
+    bl64 = bl;
+    bl64 = ((bl64 << 16) & 0x000000ff00000000ull) | (bl64 & 0x0000ff00ull);
+
+    br64 = br;
+    br64 = ((br64 << 16) & 0x000000ff00000000ull) | (br64 & 0x0000ff00ull);
+
+    f = tl64 * distixiy + tr64 * distxiy + bl64 * distixy + br64 * distxy;
+    r |= ((f >> 16) & 0x000000ff00000000ull) | (f & 0xff000000ull);
+
+    return (uint32_t)(r >> 16);
+}
+
+#else
+
+static force_inline uint32_t
+bilinear_interpolation (uint32_t tl, uint32_t tr,
+			uint32_t bl, uint32_t br,
+			int distx, int disty)
+{
+    int distxy, distxiy, distixy, distixiy;
+    uint32_t f, r;
+
+    distxy = distx * disty;
+    distxiy = (distx << 8) - distxy;	/* distx * (256 - disty) */
+    distixy = (disty << 8) - distxy;	/* disty * (256 - distx) */
+    distixiy =
+	256 * 256 - (disty << 8) -
+	(distx << 8) + distxy;		/* (256 - distx) * (256 - disty) */
+
+    /* Blue */
+    r = (tl & 0x000000ff) * distixiy + (tr & 0x000000ff) * distxiy
+      + (bl & 0x000000ff) * distixy  + (br & 0x000000ff) * distxy;
+
+    /* Green */
+    f = (tl & 0x0000ff00) * distixiy + (tr & 0x0000ff00) * distxiy
+      + (bl & 0x0000ff00) * distixy  + (br & 0x0000ff00) * distxy;
+    r |= f & 0xff000000;
+
+    tl >>= 16;
+    tr >>= 16;
+    bl >>= 16;
+    br >>= 16;
+    r >>= 16;
+
+    /* Red */
+    f = (tl & 0x000000ff) * distixiy + (tr & 0x000000ff) * distxiy
+      + (bl & 0x000000ff) * distixy  + (br & 0x000000ff) * distxy;
+    r |= f & 0x00ff0000;
+
+    /* Alpha */
+    f = (tl & 0x0000ff00) * distixiy + (tr & 0x0000ff00) * distxiy
+      + (bl & 0x0000ff00) * distixy  + (br & 0x0000ff00) * distxy;
+    r |= f & 0xff000000;
+
+    return r;
+}
+
+#endif
+
+static force_inline uint32_t
+bits_image_fetch_pixel_bilinear (bits_image_t   *image,
+				 pixman_fixed_t  x,
+				 pixman_fixed_t  y,
+				 get_pixel_t	 get_pixel)
+{
+    pixman_repeat_t repeat_mode = image->common.repeat;
+    int width = image->width;
+    int height = image->height;
+    int x1, y1, x2, y2;
+    uint32_t tl, tr, bl, br;
+    int32_t distx, disty;
+
+    x1 = x - pixman_fixed_1 / 2;
+    y1 = y - pixman_fixed_1 / 2;
+
+    distx = (x1 >> 8) & 0xff;
+    disty = (y1 >> 8) & 0xff;
+
+    x1 = pixman_fixed_to_int (x1);
+    y1 = pixman_fixed_to_int (y1);
+    x2 = x1 + 1;
+    y2 = y1 + 1;
+
+    if (repeat_mode != PIXMAN_REPEAT_NONE)
+    {
+	repeat (repeat_mode, width, &x1);
+	repeat (repeat_mode, height, &y1);
+	repeat (repeat_mode, width, &x2);
+	repeat (repeat_mode, height, &y2);
+
+	tl = get_pixel (image, x1, y1, FALSE);
+	bl = get_pixel (image, x1, y2, FALSE);
+	tr = get_pixel (image, x2, y1, FALSE);
+	br = get_pixel (image, x2, y2, FALSE);
+    }
+    else
+    {
+	tl = get_pixel (image, x1, y1, TRUE);
+	tr = get_pixel (image, x2, y1, TRUE);
+	bl = get_pixel (image, x1, y2, TRUE);
+	br = get_pixel (image, x2, y2, TRUE);
+    }
+
+    return bilinear_interpolation (tl, tr, bl, br, distx, disty);
+}
+
+static void
+bits_image_fetch_bilinear_no_repeat_8888 (pixman_image_t * ima,
+					  int              offset,
+					  int              line,
+					  int              width,
+					  uint32_t *       buffer,
+					  const uint32_t * mask)
+{
+    bits_image_t *bits = &ima->bits;
+    pixman_fixed_t x_top, x_bottom, x;
+    pixman_fixed_t ux_top, ux_bottom, ux;
+    pixman_vector_t v;
+    uint32_t top_mask, bottom_mask;
+    uint32_t *top_row;
+    uint32_t *bottom_row;
+    uint32_t *end;
+    uint32_t zero[2] = { 0, 0 };
+    int y, y1, y2;
+    int disty;
+    int mask_inc;
+    int w;
+
+    /* reference point is the center of the pixel */
+    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;
+    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;
+    v.vector[2] = pixman_fixed_1;
+
+    if (!pixman_transform_point_3d (bits->common.transform, &v))
+	return;
+
+    ux = ux_top = ux_bottom = bits->common.transform->matrix[0][0];
+    x = x_top = x_bottom = v.vector[0] - pixman_fixed_1/2;
+
+    y = v.vector[1] - pixman_fixed_1/2;
+    disty = (y >> 8) & 0xff;
+
+    /* Load the pointers to the first and second lines from the source
+     * image that bilinear code must read.
+     *
+     * The main trick in this code is about the check if any line are
+     * outside of the image;
+     *
+     * When I realize that a line (any one) is outside, I change
+     * the pointer to a dummy area with zeros. Once I change this, I
+     * must be sure the pointer will not change, so I set the
+     * variables to each pointer increments inside the loop.
+     */
+    y1 = pixman_fixed_to_int (y);
+    y2 = y1 + 1;
+
+    if (y1 < 0 || y1 >= bits->height)
+    {
+	top_row = zero;
+	x_top = 0;
+	ux_top = 0;
+    }
+    else
+    {
+	top_row = bits->bits + y1 * bits->rowstride;
+	x_top = x;
+	ux_top = ux;
+    }
+
+    if (y2 < 0 || y2 >= bits->height)
+    {
+	bottom_row = zero;
+	x_bottom = 0;
+	ux_bottom = 0;
+    }
+    else
+    {
+	bottom_row = bits->bits + y2 * bits->rowstride;
+	x_bottom = x;
+	ux_bottom = ux;
+    }
+
+    /* Instead of checking whether the operation uses the mast in
+     * each loop iteration, verify this only once and prepare the
+     * variables to make the code smaller inside the loop.
+     */
+    if (!mask)
+    {
+	uint32_t mask_bits = 1;
+
+        mask_inc = 0;
+        mask = &mask_bits;
+    }
+    else
+    {
+        /* If have a mask, prepare the variables to check it */
+        mask_inc = 1;
+    }
+
+    /* If both are zero, then the whole thing is zero */
+    if (top_row == zero && bottom_row == zero)
+    {
+	memset (buffer, 0, width * sizeof (uint32_t));
+	return;
+    }
+    else if (bits->format == PIXMAN_x8r8g8b8)
+    {
+	if (top_row == zero)
+	{
+	    top_mask = 0;
+	    bottom_mask = 0xff000000;
+	}
+	else if (bottom_row == zero)
+	{
+	    top_mask = 0xff000000;
+	    bottom_mask = 0;
+	}
+	else
+	{
+	    top_mask = 0xff000000;
+	    bottom_mask = 0xff000000;
+	}
+    }
+    else
+    {
+	top_mask = 0;
+	bottom_mask = 0;
+    }
+
+    end = buffer + width;
+
+    /* Zero fill to the left of the image */
+    while (buffer < end && x < pixman_fixed_minus_1)
+    {
+	*buffer++ = 0;
+	x += ux;
+	x_top += ux_top;
+	x_bottom += ux_bottom;
+	mask += mask_inc;
+    }
+
+    /* Left edge
+     */
+    while (buffer < end && x < 0)
+    {
+	uint32_t tr, br;
+	int32_t distx;
+
+	tr = top_row[pixman_fixed_to_int (x_top) + 1] | top_mask;
+	br = bottom_row[pixman_fixed_to_int (x_bottom) + 1] | bottom_mask;
+
+	distx = (x >> 8) & 0xff;
+
+	*buffer++ = bilinear_interpolation (0, tr, 0, br, distx, disty);
+
+	x += ux;
+	x_top += ux_top;
+	x_bottom += ux_bottom;
+	mask += mask_inc;
+    }
+
+    /* Main part */
+    w = pixman_int_to_fixed (bits->width - 1);
+
+    while (buffer < end  &&  x < w)
+    {
+	if (*mask)
+	{
+	    uint32_t tl, tr, bl, br;
+	    int32_t distx;
+
+	    tl = top_row [pixman_fixed_to_int (x_top)] | top_mask;
+	    tr = top_row [pixman_fixed_to_int (x_top) + 1] | top_mask;
+	    bl = bottom_row [pixman_fixed_to_int (x_bottom)] | bottom_mask;
+	    br = bottom_row [pixman_fixed_to_int (x_bottom) + 1] | bottom_mask;
+
+	    distx = (x >> 8) & 0xff;
+
+	    *buffer = bilinear_interpolation (tl, tr, bl, br, distx, disty);
+	}
+
+	buffer++;
+	x += ux;
+	x_top += ux_top;
+	x_bottom += ux_bottom;
+	mask += mask_inc;
+    }
+
+    /* Right Edge */
+    w = pixman_int_to_fixed (bits->width);
+    while (buffer < end  &&  x < w)
+    {
+	if (*mask)
+	{
+	    uint32_t tl, bl;
+	    int32_t distx;
+
+	    tl = top_row [pixman_fixed_to_int (x_top)] | top_mask;
+	    bl = bottom_row [pixman_fixed_to_int (x_bottom)] | bottom_mask;
+
+	    distx = (x >> 8) & 0xff;
+
+	    *buffer = bilinear_interpolation (tl, 0, bl, 0, distx, disty);
+	}
+
+	buffer++;
+	x += ux;
+	x_top += ux_top;
+	x_bottom += ux_bottom;
+	mask += mask_inc;
+    }
+
+    /* Zero fill to the left of the image */
+    while (buffer < end)
+	*buffer++ = 0;
+}
+
+static force_inline uint32_t
+bits_image_fetch_pixel_convolution (bits_image_t   *image,
+				    pixman_fixed_t  x,
+				    pixman_fixed_t  y,
+				    get_pixel_t     get_pixel)
+{
+    pixman_fixed_t *params = image->common.filter_params;
+    int x_off = (params[0] - pixman_fixed_1) >> 1;
+    int y_off = (params[1] - pixman_fixed_1) >> 1;
+    int32_t cwidth = pixman_fixed_to_int (params[0]);
+    int32_t cheight = pixman_fixed_to_int (params[1]);
+    int32_t srtot, sgtot, sbtot, satot;
+    int32_t i, j, x1, x2, y1, y2;
+    pixman_repeat_t repeat_mode = image->common.repeat;
+    int width = image->width;
+    int height = image->height;
+
+    params += 2;
+
+    x1 = pixman_fixed_to_int (x - pixman_fixed_e - x_off);
+    y1 = pixman_fixed_to_int (y - pixman_fixed_e - y_off);
+    x2 = x1 + cwidth;
+    y2 = y1 + cheight;
+
+    srtot = sgtot = sbtot = satot = 0;
+
+    for (i = y1; i < y2; ++i)
+    {
+	for (j = x1; j < x2; ++j)
+	{
+	    int rx = j;
+	    int ry = i;
+
+	    pixman_fixed_t f = *params;
+
+	    if (f)
+	    {
+		uint32_t pixel;
+
+		if (repeat_mode != PIXMAN_REPEAT_NONE)
+		{
+		    repeat (repeat_mode, width, &rx);
+		    repeat (repeat_mode, height, &ry);
+
+		    pixel = get_pixel (image, rx, ry, FALSE);
+		}
+		else
+		{
+		    pixel = get_pixel (image, rx, ry, TRUE);
+		}
+
+		srtot += RED_8 (pixel) * f;
+		sgtot += GREEN_8 (pixel) * f;
+		sbtot += BLUE_8 (pixel) * f;
+		satot += ALPHA_8 (pixel) * f;
+	    }
+
+	    params++;
+	}
+    }
+
+    satot >>= 16;
+    srtot >>= 16;
+    sgtot >>= 16;
+    sbtot >>= 16;
+
+    satot = CLIP (satot, 0, 0xff);
+    srtot = CLIP (srtot, 0, 0xff);
+    sgtot = CLIP (sgtot, 0, 0xff);
+    sbtot = CLIP (sbtot, 0, 0xff);
+
+    return ((satot << 24) | (srtot << 16) | (sgtot <<  8) | (sbtot));
+}
+
+static force_inline uint32_t
+bits_image_fetch_pixel_filtered (bits_image_t *image,
+				 pixman_fixed_t x,
+				 pixman_fixed_t y,
+				 get_pixel_t    get_pixel)
+{
+    switch (image->common.filter)
+    {
+    case PIXMAN_FILTER_NEAREST:
+    case PIXMAN_FILTER_FAST:
+	return bits_image_fetch_pixel_nearest (image, x, y, get_pixel);
+	break;
+
+    case PIXMAN_FILTER_BILINEAR:
+    case PIXMAN_FILTER_GOOD:
+    case PIXMAN_FILTER_BEST:
+	return bits_image_fetch_pixel_bilinear (image, x, y, get_pixel);
+	break;
+
+    case PIXMAN_FILTER_CONVOLUTION:
+	return bits_image_fetch_pixel_convolution (image, x, y, get_pixel);
+	break;
+
+    default:
+        break;
+    }
+
+    return 0;
+}
+
+static void
+bits_image_fetch_affine_no_alpha (pixman_image_t * image,
+				  int              offset,
+				  int              line,
+				  int              width,
+				  uint32_t *       buffer,
+				  const uint32_t * mask)
+{
+    pixman_fixed_t x, y;
+    pixman_fixed_t ux, uy;
+    pixman_vector_t v;
+    int i;
+
+    /* reference point is the center of the pixel */
+    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;
+    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;
+    v.vector[2] = pixman_fixed_1;
+
+    if (image->common.transform)
+    {
+	if (!pixman_transform_point_3d (image->common.transform, &v))
+	    return;
+
+	ux = image->common.transform->matrix[0][0];
+	uy = image->common.transform->matrix[1][0];
+    }
+    else
+    {
+	ux = pixman_fixed_1;
+	uy = 0;
+    }
+
+    x = v.vector[0];
+    y = v.vector[1];
+
+    for (i = 0; i < width; ++i)
+    {
+	if (!mask || mask[i])
+	{
+	    buffer[i] = bits_image_fetch_pixel_filtered (
+		&image->bits, x, y, fetch_pixel_no_alpha);
+	}
+
+	x += ux;
+	y += uy;
+    }
+}
+
+/* General fetcher */
+static force_inline uint32_t
+fetch_pixel_general (bits_image_t *image, int x, int y, pixman_bool_t check_bounds)
+{
+    uint32_t pixel;
+
+    if (check_bounds &&
+	(x < 0 || x >= image->width || y < 0 || y >= image->height))
+    {
+	return 0;
+    }
+
+    pixel = image->fetch_pixel_32 (image, x, y);
+
+    if (image->common.alpha_map)
+    {
+	uint32_t pixel_a;
+
+	x -= image->common.alpha_origin_x;
+	y -= image->common.alpha_origin_y;
+
+	if (x < 0 || x >= image->common.alpha_map->width ||
+	    y < 0 || y >= image->common.alpha_map->height)
+	{
+	    pixel_a = 0;
+	}
+	else
+	{
+	    pixel_a = image->common.alpha_map->fetch_pixel_32 (
+		image->common.alpha_map, x, y);
+
+	    pixel_a = ALPHA_8 (pixel_a);
+	}
+
+	pixel &= 0x00ffffff;
+	pixel |= (pixel_a << 24);
+    }
+
+    return pixel;
+}
+
+static void
+bits_image_fetch_general (pixman_image_t * image,
+			  int              offset,
+			  int              line,
+			  int              width,
+			  uint32_t *       buffer,
+			  const uint32_t * mask)
+{
+    pixman_fixed_t x, y, w;
+    pixman_fixed_t ux, uy, uw;
+    pixman_vector_t v;
+    int i;
+
+    /* reference point is the center of the pixel */
+    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;
+    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;
+    v.vector[2] = pixman_fixed_1;
+
+    if (image->common.transform)
+    {
+	if (!pixman_transform_point_3d (image->common.transform, &v))
+	    return;
+
+	ux = image->common.transform->matrix[0][0];
+	uy = image->common.transform->matrix[1][0];
+	uw = image->common.transform->matrix[2][0];
+    }
+    else
+    {
+	ux = pixman_fixed_1;
+	uy = 0;
+	uw = 0;
+    }
+
+    x = v.vector[0];
+    y = v.vector[1];
+    w = v.vector[2];
+
+    for (i = 0; i < width; ++i)
+    {
+	pixman_fixed_t x0, y0;
+
+	if (!mask || mask[i])
+	{
+	    if (w != 0)
+	    {
+		x0 = ((pixman_fixed_48_16_t)x << 16) / w;
+		y0 = ((pixman_fixed_48_16_t)y << 16) / w;
+	    }
+	    else
+	    {
+		x0 = 0;
+		y0 = 0;
+	    }
+
+	    buffer[i] = bits_image_fetch_pixel_filtered (
+		&image->bits, x0, y0, fetch_pixel_general);
+	}
+
+	x += ux;
+	y += uy;
+	w += uw;
+    }
+}
+
+static const uint8_t zero[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
+
+typedef uint32_t (* convert_pixel_t) (const uint8_t *row, int x);
+
+static force_inline void
+bits_image_fetch_bilinear_affine (pixman_image_t * image,
+				  int              offset,
+				  int              line,
+				  int              width,
+				  uint32_t *       buffer,
+				  const uint32_t * mask,
+
+				  convert_pixel_t	convert_pixel,
+				  pixman_format_code_t	format,
+				  pixman_repeat_t	repeat_mode)
+{
+    pixman_fixed_t x, y;
+    pixman_fixed_t ux, uy;
+    pixman_vector_t v;
+    bits_image_t *bits = &image->bits;
+    int i;
+
+    /* reference point is the center of the pixel */
+    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;
+    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;
+    v.vector[2] = pixman_fixed_1;
+
+    if (!pixman_transform_point_3d (image->common.transform, &v))
+	return;
+
+    ux = image->common.transform->matrix[0][0];
+    uy = image->common.transform->matrix[1][0];
+
+    x = v.vector[0];
+    y = v.vector[1];
+
+    for (i = 0; i < width; ++i)
+    {
+	int x1, y1, x2, y2;
+	uint32_t tl, tr, bl, br;
+	int32_t distx, disty;
+	int width = image->bits.width;
+	int height = image->bits.height;
+	const uint8_t *row1;
+	const uint8_t *row2;
+
+	if (mask && !mask[i])
+	    goto next;
+
+	x1 = x - pixman_fixed_1 / 2;
+	y1 = y - pixman_fixed_1 / 2;
+
+	distx = (x1 >> 8) & 0xff;
+	disty = (y1 >> 8) & 0xff;
+
+	y1 = pixman_fixed_to_int (y1);
+	y2 = y1 + 1;
+	x1 = pixman_fixed_to_int (x1);
+	x2 = x1 + 1;
+
+	if (repeat_mode != PIXMAN_REPEAT_NONE)
+	{
+	    uint32_t mask;
+
+	    mask = PIXMAN_FORMAT_A (format)? 0 : 0xff000000;
+
+	    repeat (repeat_mode, width, &x1);
+	    repeat (repeat_mode, height, &y1);
+	    repeat (repeat_mode, width, &x2);
+	    repeat (repeat_mode, height, &y2);
+
+	    row1 = (uint8_t *)bits->bits + bits->rowstride * 4 * y1;
+	    row2 = (uint8_t *)bits->bits + bits->rowstride * 4 * y2;
+
+	    tl = convert_pixel (row1, x1) | mask;
+	    tr = convert_pixel (row1, x2) | mask;
+	    bl = convert_pixel (row2, x1) | mask;
+	    br = convert_pixel (row2, x2) | mask;
+	}
+	else
+	{
+	    uint32_t mask1, mask2;
+	    int bpp;
+
+	    /* Note: PIXMAN_FORMAT_BPP() returns an unsigned value,
+	     * which means if you use it in expressions, those
+	     * expressions become unsigned themselves. Since
+	     * the variables below can be negative in some cases,
+	     * that will lead to crashes on 64 bit architectures.
+	     *
+	     * So this line makes sure bpp is signed
+	     */
+	    bpp = PIXMAN_FORMAT_BPP (format);
+
+	    if (x1 >= width || x2 < 0 || y1 >= height || y2 < 0)
+	    {
+		buffer[i] = 0;
+		goto next;
+	    }
+
+	    if (y2 == 0)
+	    {
+		row1 = zero;
+		mask1 = 0;
+	    }
+	    else
+	    {
+		row1 = (uint8_t *)bits->bits + bits->rowstride * 4 * y1;
+		row1 += bpp / 8 * x1;
+
+		mask1 = PIXMAN_FORMAT_A (format)? 0 : 0xff000000;
+	    }
+
+	    if (y1 == height - 1)
+	    {
+		row2 = zero;
+		mask2 = 0;
+	    }
+	    else
+	    {
+		row2 = (uint8_t *)bits->bits + bits->rowstride * 4 * y2;
+		row2 += bpp / 8 * x1;
+
+		mask2 = PIXMAN_FORMAT_A (format)? 0 : 0xff000000;
+	    }
+
+	    if (x2 == 0)
+	    {
+		tl = 0;
+		bl = 0;
+	    }
+	    else
+	    {
+		tl = convert_pixel (row1, 0) | mask1;
+		bl = convert_pixel (row2, 0) | mask2;
+	    }
+
+	    if (x1 == width - 1)
+	    {
+		tr = 0;
+		br = 0;
+	    }
+	    else
+	    {
+		tr = convert_pixel (row1, 1) | mask1;
+		br = convert_pixel (row2, 1) | mask2;
+	    }
+	}
+
+	buffer[i] = bilinear_interpolation (
+	    tl, tr, bl, br, distx, disty);
+
+    next:
+	x += ux;
+	y += uy;
+    }
+}
+
+static force_inline void
+bits_image_fetch_nearest_affine (pixman_image_t * image,
+				 int              offset,
+				 int              line,
+				 int              width,
+				 uint32_t *       buffer,
+				 const uint32_t * mask,
+				 
+				 convert_pixel_t	convert_pixel,
+				 pixman_format_code_t	format,
+				 pixman_repeat_t	repeat_mode)
+{
+    pixman_fixed_t x, y;
+    pixman_fixed_t ux, uy;
+    pixman_vector_t v;
+    bits_image_t *bits = &image->bits;
+    int i;
+
+    /* reference point is the center of the pixel */
+    v.vector[0] = pixman_int_to_fixed (offset) + pixman_fixed_1 / 2;
+    v.vector[1] = pixman_int_to_fixed (line) + pixman_fixed_1 / 2;
+    v.vector[2] = pixman_fixed_1;
+
+    if (!pixman_transform_point_3d (image->common.transform, &v))
+	return;
+
+    ux = image->common.transform->matrix[0][0];
+    uy = image->common.transform->matrix[1][0];
+
+    x = v.vector[0];
+    y = v.vector[1];
+
+    for (i = 0; i < width; ++i)
+    {
+	int width, height, x0, y0;
+	const uint8_t *row;
+
+	if (mask && !mask[i])
+	    goto next;
+	
+	width = image->bits.width;
+	height = image->bits.height;
+	x0 = pixman_fixed_to_int (x - pixman_fixed_e);
+	y0 = pixman_fixed_to_int (y - pixman_fixed_e);
+
+	if (repeat_mode == PIXMAN_REPEAT_NONE &&
+	    (y0 < 0 || y0 >= height || x0 < 0 || x0 >= width))
+	{
+	    buffer[i] = 0;
+	}
+	else
+	{
+	    uint32_t mask = PIXMAN_FORMAT_A (format)? 0 : 0xff000000;
+
+	    if (repeat_mode != PIXMAN_REPEAT_NONE)
+	    {
+		repeat (repeat_mode, width, &x0);
+		repeat (repeat_mode, height, &y0);
+	    }
+
+	    row = (uint8_t *)bits->bits + bits->rowstride * 4 * y0;
+
+	    buffer[i] = convert_pixel (row, x0) | mask;
+	}
+
+    next:
+	x += ux;
+	y += uy;
+    }
+}
+
+static force_inline uint32_t
+convert_a8r8g8b8 (const uint8_t *row, int x)
+{
+    return *(((uint32_t *)row) + x);
+}
+
+static force_inline uint32_t
+convert_x8r8g8b8 (const uint8_t *row, int x)
+{
+    return *(((uint32_t *)row) + x);
+}
+
+static force_inline uint32_t
+convert_a8 (const uint8_t *row, int x)
+{
+    return *(row + x) << 24;
+}
+
+static force_inline uint32_t
+convert_r5g6b5 (const uint8_t *row, int x)
+{
+    return CONVERT_0565_TO_0888 (*((uint16_t *)row + x));
+}
+
+#define MAKE_BILINEAR_FETCHER(name, format, repeat_mode)		\
+    static void								\
+    bits_image_fetch_bilinear_affine_ ## name (pixman_image_t *image,	\
+					       int              offset,	\
+					       int              line,	\
+					       int              width,	\
+					       uint32_t *       buffer,	\
+					       const uint32_t * mask)	\
+    {									\
+	bits_image_fetch_bilinear_affine (image, offset, line,		\
+					  width, buffer, mask,		\
+					  convert_ ## format,		\
+					  PIXMAN_ ## format,		\
+					  repeat_mode);			\
+    }
+
+#define MAKE_NEAREST_FETCHER(name, format, repeat_mode)			\
+    static void								\
+    bits_image_fetch_nearest_affine_ ## name (pixman_image_t *image,	\
+					      int              offset,	\
+					      int              line,	\
+					      int              width,	\
+					      uint32_t *       buffer,	\
+					      const uint32_t * mask)	\
+    {									\
+	bits_image_fetch_nearest_affine (image, offset, line,		\
+					 width, buffer, mask,		\
+					 convert_ ## format,		\
+					 PIXMAN_ ## format,		\
+					 repeat_mode);			\
+    }
+
+#define MAKE_FETCHERS(name, format, repeat_mode)			\
+    MAKE_NEAREST_FETCHER (name, format, repeat_mode)			\
+    MAKE_BILINEAR_FETCHER (name, format, repeat_mode)
+
+MAKE_FETCHERS (pad_a8r8g8b8,     a8r8g8b8, PIXMAN_REPEAT_PAD)
+MAKE_FETCHERS (none_a8r8g8b8,    a8r8g8b8, PIXMAN_REPEAT_NONE)
+MAKE_FETCHERS (reflect_a8r8g8b8, a8r8g8b8, PIXMAN_REPEAT_REFLECT)
+MAKE_FETCHERS (normal_a8r8g8b8,  a8r8g8b8, PIXMAN_REPEAT_NORMAL)
+MAKE_FETCHERS (pad_x8r8g8b8,     x8r8g8b8, PIXMAN_REPEAT_PAD)
+MAKE_FETCHERS (none_x8r8g8b8,    x8r8g8b8, PIXMAN_REPEAT_NONE)
+MAKE_FETCHERS (reflect_x8r8g8b8, x8r8g8b8, PIXMAN_REPEAT_REFLECT)
+MAKE_FETCHERS (normal_x8r8g8b8,  x8r8g8b8, PIXMAN_REPEAT_NORMAL)
+MAKE_FETCHERS (pad_a8,           a8,       PIXMAN_REPEAT_PAD)
+MAKE_FETCHERS (none_a8,          a8,       PIXMAN_REPEAT_NONE)
+MAKE_FETCHERS (reflect_a8,	 a8,       PIXMAN_REPEAT_REFLECT)
+MAKE_FETCHERS (normal_a8,	 a8,       PIXMAN_REPEAT_NORMAL)
+MAKE_FETCHERS (pad_r5g6b5,       r5g6b5,   PIXMAN_REPEAT_PAD)
+MAKE_FETCHERS (none_r5g6b5,      r5g6b5,   PIXMAN_REPEAT_NONE)
+MAKE_FETCHERS (reflect_r5g6b5,   r5g6b5,   PIXMAN_REPEAT_REFLECT)
+MAKE_FETCHERS (normal_r5g6b5,    r5g6b5,   PIXMAN_REPEAT_NORMAL)
+
+static void
+bits_image_fetch_solid_32 (pixman_image_t * image,
+                           int              x,
+                           int              y,
+                           int              width,
+                           uint32_t *       buffer,
+                           const uint32_t * mask)
+{
+    uint32_t color;
+    uint32_t *end;
+
+    color = image->bits.fetch_pixel_32 (&image->bits, 0, 0);
+
+    end = buffer + width;
+    while (buffer < end)
+	*(buffer++) = color;
+}
+
+static void
+bits_image_fetch_solid_64 (pixman_image_t * image,
+                           int              x,
+                           int              y,
+                           int              width,
+                           uint32_t *       b,
+                           const uint32_t * unused)
+{
+    uint64_t color;
+    uint64_t *buffer = (uint64_t *)b;
+    uint64_t *end;
+
+    color = image->bits.fetch_pixel_64 (&image->bits, 0, 0);
+
+    end = buffer + width;
+    while (buffer < end)
+	*(buffer++) = color;
+}
+
+static void
+bits_image_fetch_untransformed_repeat_none (bits_image_t *image,
+                                            pixman_bool_t wide,
+                                            int           x,
+                                            int           y,
+                                            int           width,
+                                            uint32_t *    buffer)
+{
+    uint32_t w;
+
+    if (y < 0 || y >= image->height)
+    {
+	memset (buffer, 0, width * (wide? 8 : 4));
+	return;
+    }
+
+    if (x < 0)
+    {
+	w = MIN (width, -x);
+
+	memset (buffer, 0, w * (wide ? 8 : 4));
+
+	width -= w;
+	buffer += w * (wide? 2 : 1);
+	x += w;
+    }
+
+    if (x < image->width)
+    {
+	w = MIN (width, image->width - x);
+
+	if (wide)
+	    image->fetch_scanline_64 ((pixman_image_t *)image, x, y, w, buffer, NULL);
+	else
+	    image->fetch_scanline_32 ((pixman_image_t *)image, x, y, w, buffer, NULL);
+
+	width -= w;
+	buffer += w * (wide? 2 : 1);
+	x += w;
+    }
+
+    memset (buffer, 0, width * (wide ? 8 : 4));
+}
+
+static void
+bits_image_fetch_untransformed_repeat_normal (bits_image_t *image,
+                                              pixman_bool_t wide,
+                                              int           x,
+                                              int           y,
+                                              int           width,
+                                              uint32_t *    buffer)
+{
+    uint32_t w;
+
+    while (y < 0)
+	y += image->height;
+
+    while (y >= image->height)
+	y -= image->height;
+
+    while (width)
+    {
+	while (x < 0)
+	    x += image->width;
+	while (x >= image->width)
+	    x -= image->width;
+
+	w = MIN (width, image->width - x);
+
+	if (wide)
+	    image->fetch_scanline_64 ((pixman_image_t *)image, x, y, w, buffer, NULL);
+	else
+	    image->fetch_scanline_32 ((pixman_image_t *)image, x, y, w, buffer, NULL);
+
+	buffer += w * (wide? 2 : 1);
+	x += w;
+	width -= w;
+    }
+}
+
+static void
+bits_image_fetch_untransformed_32 (pixman_image_t * image,
+                                   int              x,
+                                   int              y,
+                                   int              width,
+                                   uint32_t *       buffer,
+                                   const uint32_t * mask)
+{
+    if (image->common.repeat == PIXMAN_REPEAT_NONE)
+    {
+	bits_image_fetch_untransformed_repeat_none (
+	    &image->bits, FALSE, x, y, width, buffer);
+    }
+    else
+    {
+	bits_image_fetch_untransformed_repeat_normal (
+	    &image->bits, FALSE, x, y, width, buffer);
+    }
+}
+
+static void
+bits_image_fetch_untransformed_64 (pixman_image_t * image,
+                                   int              x,
+                                   int              y,
+                                   int              width,
+                                   uint32_t *       buffer,
+                                   const uint32_t * unused)
+{
+    if (image->common.repeat == PIXMAN_REPEAT_NONE)
+    {
+	bits_image_fetch_untransformed_repeat_none (
+	    &image->bits, TRUE, x, y, width, buffer);
+    }
+    else
+    {
+	bits_image_fetch_untransformed_repeat_normal (
+	    &image->bits, TRUE, x, y, width, buffer);
+    }
+}
+
+typedef struct
+{
+    pixman_format_code_t	format;
+    uint32_t			flags;
+    fetch_scanline_t		fetch_32;
+    fetch_scanline_t		fetch_64;
+} fetcher_info_t;
+
+static const fetcher_info_t fetcher_info[] =
+{
+    { PIXMAN_solid,
+      FAST_PATH_NO_ALPHA_MAP,
+      bits_image_fetch_solid_32,
+      bits_image_fetch_solid_64
+    },
+
+    { PIXMAN_any,
+      (FAST_PATH_NO_ALPHA_MAP			|
+       FAST_PATH_ID_TRANSFORM			|
+       FAST_PATH_NO_CONVOLUTION_FILTER		|
+       FAST_PATH_NO_PAD_REPEAT			|
+       FAST_PATH_NO_REFLECT_REPEAT),
+      bits_image_fetch_untransformed_32,
+      bits_image_fetch_untransformed_64
+    },
+
+#define FAST_BILINEAR_FLAGS						\
+    (FAST_PATH_NO_ALPHA_MAP		|				\
+     FAST_PATH_NO_ACCESSORS		|				\
+     FAST_PATH_HAS_TRANSFORM		|				\
+     FAST_PATH_AFFINE_TRANSFORM		|				\
+     FAST_PATH_X_UNIT_POSITIVE		|				\
+     FAST_PATH_Y_UNIT_ZERO		|				\
+     FAST_PATH_NONE_REPEAT		|				\
+     FAST_PATH_BILINEAR_FILTER)
+
+    { PIXMAN_a8r8g8b8,
+      FAST_BILINEAR_FLAGS,
+      bits_image_fetch_bilinear_no_repeat_8888,
+      _pixman_image_get_scanline_generic_64
+    },
+
+    { PIXMAN_x8r8g8b8,
+      FAST_BILINEAR_FLAGS,
+      bits_image_fetch_bilinear_no_repeat_8888,
+      _pixman_image_get_scanline_generic_64
+    },
+
+#define GENERAL_BILINEAR_FLAGS						\
+    (FAST_PATH_NO_ALPHA_MAP		|				\
+     FAST_PATH_NO_ACCESSORS		|				\
+     FAST_PATH_HAS_TRANSFORM		|				\
+     FAST_PATH_AFFINE_TRANSFORM		|				\
+     FAST_PATH_BILINEAR_FILTER)
+
+#define GENERAL_NEAREST_FLAGS						\
+    (FAST_PATH_NO_ALPHA_MAP		|				\
+     FAST_PATH_NO_ACCESSORS		|				\
+     FAST_PATH_HAS_TRANSFORM		|				\
+     FAST_PATH_AFFINE_TRANSFORM		|				\
+     FAST_PATH_NEAREST_FILTER)
+
+#define BILINEAR_AFFINE_FAST_PATH(name, format, repeat)			\
+    { PIXMAN_ ## format,						\
+      GENERAL_BILINEAR_FLAGS | FAST_PATH_ ## repeat ## _REPEAT,		\
+      bits_image_fetch_bilinear_affine_ ## name,			\
+      _pixman_image_get_scanline_generic_64				\
+    },
+
+#define NEAREST_AFFINE_FAST_PATH(name, format, repeat)			\
+    { PIXMAN_ ## format,						\
+      GENERAL_NEAREST_FLAGS | FAST_PATH_ ## repeat ## _REPEAT,		\
+      bits_image_fetch_nearest_affine_ ## name,			\
+      _pixman_image_get_scanline_generic_64				\
+    },
+
+#define AFFINE_FAST_PATHS(name, format, repeat)				\
+    BILINEAR_AFFINE_FAST_PATH(name, format, repeat)			\
+    NEAREST_AFFINE_FAST_PATH(name, format, repeat)
+    
+    AFFINE_FAST_PATHS (pad_a8r8g8b8, a8r8g8b8, PAD)
+    AFFINE_FAST_PATHS (none_a8r8g8b8, a8r8g8b8, NONE)
+    AFFINE_FAST_PATHS (reflect_a8r8g8b8, a8r8g8b8, REFLECT)
+    AFFINE_FAST_PATHS (normal_a8r8g8b8, a8r8g8b8, NORMAL)
+    AFFINE_FAST_PATHS (pad_x8r8g8b8, x8r8g8b8, PAD)
+    AFFINE_FAST_PATHS (none_x8r8g8b8, x8r8g8b8, NONE)
+    AFFINE_FAST_PATHS (reflect_x8r8g8b8, x8r8g8b8, REFLECT)
+    AFFINE_FAST_PATHS (normal_x8r8g8b8, x8r8g8b8, NORMAL)
+    AFFINE_FAST_PATHS (pad_a8, a8, PAD)
+    AFFINE_FAST_PATHS (none_a8, a8, NONE)
+    AFFINE_FAST_PATHS (reflect_a8, a8, REFLECT)
+    AFFINE_FAST_PATHS (normal_a8, a8, NORMAL)
+    AFFINE_FAST_PATHS (pad_r5g6b5, r5g6b5, PAD)
+    AFFINE_FAST_PATHS (none_r5g6b5, r5g6b5, NONE)
+    AFFINE_FAST_PATHS (reflect_r5g6b5, r5g6b5, REFLECT)
+    AFFINE_FAST_PATHS (normal_r5g6b5, r5g6b5, NORMAL)
+
+    /* Affine, no alpha */
+    { PIXMAN_any,
+      (FAST_PATH_NO_ALPHA_MAP | FAST_PATH_HAS_TRANSFORM | FAST_PATH_AFFINE_TRANSFORM),
+      bits_image_fetch_affine_no_alpha,
+      _pixman_image_get_scanline_generic_64
+    },
+
+    /* General */
+    { PIXMAN_any, 0, bits_image_fetch_general, _pixman_image_get_scanline_generic_64 },
+
+    { PIXMAN_null },
+};
+
+static void
+bits_image_property_changed (pixman_image_t *image)
+{
+    uint32_t flags = image->common.flags;
+    pixman_format_code_t format = image->common.extended_format_code;
+    const fetcher_info_t *info;
+
+    _pixman_bits_image_setup_accessors (&image->bits);
+
+    info = fetcher_info;
+    while (info->format != PIXMAN_null)
+    {
+	if ((info->format == format || info->format == PIXMAN_any)	&&
+	    (info->flags & flags) == info->flags)
+	{
+	    image->common.get_scanline_32 = info->fetch_32;
+	    image->common.get_scanline_64 = info->fetch_64;
+	    break;
+	}
+
+	info++;
+    }
+}
+
+static uint32_t *
+create_bits (pixman_format_code_t format,
+             int                  width,
+             int                  height,
+             int *                rowstride_bytes)
+{
+    int stride;
+    int buf_size;
+    int bpp;
+
+    /* what follows is a long-winded way, avoiding any possibility of integer
+     * overflows, of saying:
+     * stride = ((width * bpp + 0x1f) >> 5) * sizeof (uint32_t);
+     */
+
+    bpp = PIXMAN_FORMAT_BPP (format);
+    if (pixman_multiply_overflows_int (width, bpp))
+	return NULL;
+
+    stride = width * bpp;
+    if (pixman_addition_overflows_int (stride, 0x1f))
+	return NULL;
+
+    stride += 0x1f;
+    stride >>= 5;
+
+    stride *= sizeof (uint32_t);
+
+    if (pixman_multiply_overflows_int (height, stride))
+	return NULL;
+
+    buf_size = height * stride;
+
+    if (rowstride_bytes)
+	*rowstride_bytes = stride;
+
+    return calloc (buf_size, 1);
+}
+
+PIXMAN_EXPORT pixman_image_t *
+pixman_image_create_bits (pixman_format_code_t format,
+                          int                  width,
+                          int                  height,
+                          uint32_t *           bits,
+                          int                  rowstride_bytes)
+{
+    pixman_image_t *image;
+    uint32_t *free_me = NULL;
+
+    /* must be a whole number of uint32_t's
+     */
+    return_val_if_fail (
+	bits == NULL || (rowstride_bytes % sizeof (uint32_t)) == 0, NULL);
+
+    return_val_if_fail (PIXMAN_FORMAT_BPP (format) >= PIXMAN_FORMAT_DEPTH (format), NULL);
+
+    if (!bits && width && height)
+    {
+	free_me = bits = create_bits (format, width, height, &rowstride_bytes);
+	if (!bits)
+	    return NULL;
+    }
+
+    image = _pixman_image_allocate ();
+
+    if (!image)
+    {
+	if (free_me)
+	    free (free_me);
+
+	return NULL;
+    }
+
+    image->type = BITS;
+    image->bits.format = format;
+    image->bits.width = width;
+    image->bits.height = height;
+    image->bits.bits = bits;
+    image->bits.free_me = free_me;
+    image->bits.read_func = NULL;
+    image->bits.write_func = NULL;
+
+    /* The rowstride is stored in number of uint32_t */
+    image->bits.rowstride = rowstride_bytes / (int) sizeof (uint32_t);
+
+    image->bits.indexed = NULL;
+
+    image->common.property_changed = bits_image_property_changed;
+
+    _pixman_image_reset_clip_region (image);
+
+    return image;
+}
diff --git a/pixman/pixman/pixman-compiler.h b/pixman/pixman/pixman-compiler.h
index 340b402a3..ebbffc3f3 100644
--- a/pixman/pixman/pixman-compiler.h
+++ b/pixman/pixman/pixman-compiler.h
@@ -1,216 +1,215 @@
-/* Pixman uses some non-standard compiler features. This file ensures

- * they exist

- *

- * The features are:

- *

- *    FUNC	     must be defined to expand to the current function

- *    PIXMAN_EXPORT  should be defined to whatever is required to

- *                   export functions from a shared library

- *    limits	     limits for various types must be defined

- *    inline         must be defined

- *    force_inline   must be defined

- */

-#if defined (__GNUC__)

-#  define FUNC     ((const char*) (__PRETTY_FUNCTION__))

-#elif defined (__sun) || (defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L)

-#  define FUNC     ((const char*) (__func__))

-#else

-#  define FUNC     ((const char*) ("???"))

-#endif

-

-#ifndef INT16_MIN

-# define INT16_MIN              (-32767-1)

-#endif

-

-#ifndef INT16_MAX

-# define INT16_MAX              (32767)

-#endif

-

-#ifndef INT32_MIN

-# define INT32_MIN              (-2147483647-1)

-#endif

-

-#ifndef INT32_MAX

-# define INT32_MAX              (2147483647)

-#endif

-

-#ifndef UINT32_MIN

-# define UINT32_MIN             (0)

-#endif

-

-#ifndef UINT32_MAX

-# define UINT32_MAX             (4294967295U)

-#endif

-

-#ifndef M_PI

-# define M_PI			3.14159265358979323846

-#endif

-

-#ifdef _MSC_VER

-/* 'inline' is available only in C++ in MSVC */

-#   define inline __inline

-#   define force_inline __forceinline

-#   define noinline __declspec(noinline)

-#elif defined __GNUC__ || (defined(__SUNPRO_C) && (__SUNPRO_C >= 0x590))

-#   define inline __inline__

-#   define force_inline __inline__ __attribute__ ((__always_inline__))

-#   define noinline __attribute__((noinline))

-#else

-#   ifndef force_inline

-#      define force_inline inline

-#   endif

-#   ifndef noinline

-#      define noinline

-#   endif

-#endif

-

-/* GCC visibility */

-#if defined(__GNUC__) && __GNUC__ >= 4 && !defined(_WIN32)

-#   define PIXMAN_EXPORT __attribute__ ((visibility("default")))

-/* Sun Studio 8 visibility */

-#elif defined(__SUNPRO_C) && (__SUNPRO_C >= 0x550)

-#   define PIXMAN_EXPORT __global

-#else

-#   define PIXMAN_EXPORT

-#endif

-

-/* TLS */

-#if defined(PIXMAN_NO_TLS)

-

-#   define PIXMAN_DEFINE_THREAD_LOCAL(type, name)			\

-    static type name

-#   define PIXMAN_GET_THREAD_LOCAL(name)				\

-    (&name)

-

-#elif defined(TOOLCHAIN_SUPPORTS__THREAD)

-

-#   define PIXMAN_DEFINE_THREAD_LOCAL(type, name)			\

-    static __thread type name

-#   define PIXMAN_GET_THREAD_LOCAL(name)				\

-    (&name)

-

-#elif defined(__MINGW32__) && !defined(__WIN64)

-

-/* 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;		\

-    static void *tls_ ## name ## _mutex = NULL;				\

-    static unsigned tls_ ## name ## _index;				\

-									\

-    static type *							\

-    tls_ ## name ## _alloc (void)					\

-    {									\

-        type *value = calloc (1, sizeof (type));			\

-        if (value)							\

-            TlsSetValue (tls_ ## name ## _index, value);		\

-        return value;							\

-    }									\

-									\

-    static force_inline type *						\

-    tls_ ## name ## _get (void)						\

-    {									\

-	type *value;							\

-	if (!tls_ ## name ## _initialized)				\

-	{								\

-	    if (!tls_ ## name ## _mutex)				\

-	    {								\

-		void *mutex = CreateMutexA (NULL, 0, NULL);		\

-		if (InterlockedCompareExchangePointer (			\

-			&tls_ ## name ## _mutex, mutex, NULL) != NULL)	\

-		{							\

-		    CloseHandle (mutex);				\

-		}							\

-	    }								\

-	    WaitForSingleObject (tls_ ## name ## _mutex, 0xFFFFFFFF);	\

-	    if (!tls_ ## name ## _initialized)				\

-	    {								\

-		tls_ ## name ## _index = TlsAlloc ();			\

-		tls_ ## name ## _initialized = 1;			\

-	    }								\

-	    ReleaseMutex (tls_ ## name ## _mutex);			\

-	}								\

-	if (tls_ ## name ## _index == 0xFFFFFFFF)			\

-	    return NULL;						\

-	value = TlsGetValue (tls_ ## name ## _index);			\

-	if (!value)							\

-	    value = tls_ ## name ## _alloc ();				\

-	return value;							\

-    }

-

-#   define PIXMAN_GET_THREAD_LOCAL(name)				\

-    tls_ ## name ## _get ()

-

-#elif defined(_MSC_VER)

-

-#   define PIXMAN_DEFINE_THREAD_LOCAL(type, name)			\

-    static __declspec(thread) type name

-#   define PIXMAN_GET_THREAD_LOCAL(name)				\

-    (&name)

-

-#elif defined(HAVE_PTHREAD_SETSPECIFIC)

-

-#include <pthread.h>

-

-#  define PIXMAN_DEFINE_THREAD_LOCAL(type, name)			\

-    static pthread_once_t tls_ ## name ## _once_control = PTHREAD_ONCE_INIT; \

-    static pthread_key_t tls_ ## name ## _key;				\

-									\

-    static void								\

-    tls_ ## name ## _destroy_value (void *value)			\

-    {									\

-	free (value);							\

-    }									\

-									\

-    static void								\

-    tls_ ## name ## _make_key (void)					\

-    {									\

-	pthread_key_create (&tls_ ## name ## _key,			\

-			    tls_ ## name ## _destroy_value);		\

-    }									\

-									\

-    static type *							\

-    tls_ ## name ## _alloc (void)					\

-    {									\

-	type *value = calloc (1, sizeof (type));			\

-	if (value)							\

-	    pthread_setspecific (tls_ ## name ## _key, value);		\

-	return value;							\

-    }									\

-									\

-    static force_inline type *						\

-    tls_ ## name ## _get (void)						\

-    {									\

-	type *value = NULL;						\

-	if (pthread_once (&tls_ ## name ## _once_control,		\

-			  tls_ ## name ## _make_key) == 0)		\

-	{								\

-	    value = pthread_getspecific (tls_ ## name ## _key);		\

-	    if (!value)							\

-		value = tls_ ## name ## _alloc ();			\

-	}								\

-	return value;							\

-    }									\

-    extern int no_such_variable						

-

-#   define PIXMAN_GET_THREAD_LOCAL(name)				\

-    tls_ ## name ## _get ()

-

-#else

-

-#    error "Unknown thread local support for this system. Pixman will not work with multiple threads. Define PIXMAN_NO_TLS to acknowledge and accept this limitation and compile pixman without thread-safety support."

-

-#endif

+/* Pixman uses some non-standard compiler features. This file ensures
+ * they exist
+ *
+ * The features are:
+ *
+ *    FUNC	     must be defined to expand to the current function
+ *    PIXMAN_EXPORT  should be defined to whatever is required to
+ *                   export functions from a shared library
+ *    limits	     limits for various types must be defined
+ *    inline         must be defined
+ *    force_inline   must be defined
+ */
+#if defined (__GNUC__)
+#  define FUNC     ((const char*) (__PRETTY_FUNCTION__))
+#elif defined (__sun) || (defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L)
+#  define FUNC     ((const char*) (__func__))
+#else
+#  define FUNC     ((const char*) ("???"))
+#endif
+
+#ifndef INT16_MIN
+# define INT16_MIN              (-32767-1)
+#endif
+
+#ifndef INT16_MAX
+# define INT16_MAX              (32767)
+#endif
+
+#ifndef INT32_MIN
+# define INT32_MIN              (-2147483647-1)
+#endif
+
+#ifndef INT32_MAX
+# define INT32_MAX              (2147483647)
+#endif
+
+#ifndef UINT32_MIN
+# define UINT32_MIN             (0)
+#endif
+
+#ifndef UINT32_MAX
+# define UINT32_MAX             (4294967295U)
+#endif
+
+#ifndef M_PI
+# define M_PI			3.14159265358979323846
+#endif
+
+#ifdef _MSC_VER
+/* 'inline' is available only in C++ in MSVC */
+#   define inline __inline
+#   define force_inline __forceinline
+#   define noinline __declspec(noinline)
+#elif defined __GNUC__ || (defined(__SUNPRO_C) && (__SUNPRO_C >= 0x590))
+#   define inline __inline__
+#   define force_inline __inline__ __attribute__ ((__always_inline__))
+#   define noinline __attribute__((noinline))
+#else
+#   ifndef force_inline
+#      define force_inline inline
+#   endif
+#   ifndef noinline
+#      define noinline
+#   endif
+#endif
+
+/* GCC visibility */
+#if defined(__GNUC__) && __GNUC__ >= 4 && !defined(_WIN32)
+#   define PIXMAN_EXPORT __attribute__ ((visibility("default")))
+/* Sun Studio 8 visibility */
+#elif defined(__SUNPRO_C) && (__SUNPRO_C >= 0x550)
+#   define PIXMAN_EXPORT __global
+#else
+#   define PIXMAN_EXPORT
+#endif
+
+/* TLS */
+#if defined(PIXMAN_NO_TLS)
+
+#   define PIXMAN_DEFINE_THREAD_LOCAL(type, name)			\
+    static type name
+#   define PIXMAN_GET_THREAD_LOCAL(name)				\
+    (&name)
+
+#elif defined(TOOLCHAIN_SUPPORTS__THREAD)
+
+#   define PIXMAN_DEFINE_THREAD_LOCAL(type, name)			\
+    static __thread type name
+#   define PIXMAN_GET_THREAD_LOCAL(name)				\
+    (&name)
+
+#elif defined(__MINGW32__) && !defined(__WIN64)
+
+/* 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;		\
+    static void *tls_ ## name ## _mutex = NULL;				\
+    static unsigned tls_ ## name ## _index;				\
+									\
+    static type *							\
+    tls_ ## name ## _alloc (void)					\
+    {									\
+        type *value = calloc (1, sizeof (type));			\
+        if (value)							\
+            TlsSetValue (tls_ ## name ## _index, value);		\
+        return value;							\
+    }									\
+									\
+    static force_inline type *						\
+    tls_ ## name ## _get (void)						\
+    {									\
+	type *value;							\
+	if (!tls_ ## name ## _initialized)				\
+	{								\
+	    if (!tls_ ## name ## _mutex)				\
+	    {								\
+		void *mutex = CreateMutexA (NULL, 0, NULL);		\
+		if (InterlockedCompareExchangePointer (			\
+			&tls_ ## name ## _mutex, mutex, NULL) != NULL)	\
+		{							\
+		    CloseHandle (mutex);				\
+		}							\
+	    }								\
+	    WaitForSingleObject (tls_ ## name ## _mutex, 0xFFFFFFFF);	\
+	    if (!tls_ ## name ## _initialized)				\
+	    {								\
+		tls_ ## name ## _index = TlsAlloc ();			\
+		tls_ ## name ## _initialized = 1;			\
+	    }								\
+	    ReleaseMutex (tls_ ## name ## _mutex);			\
+	}								\
+	if (tls_ ## name ## _index == 0xFFFFFFFF)			\
+	    return NULL;						\
+	value = TlsGetValue (tls_ ## name ## _index);			\
+	if (!value)							\
+	    value = tls_ ## name ## _alloc ();				\
+	return value;							\
+    }
+
+#   define PIXMAN_GET_THREAD_LOCAL(name)				\
+    tls_ ## name ## _get ()
+
+#elif defined(_MSC_VER)
+
+#   define PIXMAN_DEFINE_THREAD_LOCAL(type, name)			\
+    static __declspec(thread) type name
+#   define PIXMAN_GET_THREAD_LOCAL(name)				\
+    (&name)
+
+#elif defined(HAVE_PTHREAD_SETSPECIFIC)
+
+#include <pthread.h>
+
+#  define PIXMAN_DEFINE_THREAD_LOCAL(type, name)			\
+    static pthread_once_t tls_ ## name ## _once_control = PTHREAD_ONCE_INIT; \
+    static pthread_key_t tls_ ## name ## _key;				\
+									\
+    static void								\
+    tls_ ## name ## _destroy_value (void *value)			\
+    {									\
+	free (value);							\
+    }									\
+									\
+    static void								\
+    tls_ ## name ## _make_key (void)					\
+    {									\
+	pthread_key_create (&tls_ ## name ## _key,			\
+			    tls_ ## name ## _destroy_value);		\
+    }									\
+									\
+    static type *							\
+    tls_ ## name ## _alloc (void)					\
+    {									\
+	type *value = calloc (1, sizeof (type));			\
+	if (value)							\
+	    pthread_setspecific (tls_ ## name ## _key, value);		\
+	return value;							\
+    }									\
+									\
+    static force_inline type *						\
+    tls_ ## name ## _get (void)						\
+    {									\
+	type *value = NULL;						\
+	if (pthread_once (&tls_ ## name ## _once_control,		\
+			  tls_ ## name ## _make_key) == 0)		\
+	{								\
+	    value = pthread_getspecific (tls_ ## name ## _key);		\
+	    if (!value)							\
+		value = tls_ ## name ## _alloc ();			\
+	}								\
+	return value;							\
+    }
+
+#   define PIXMAN_GET_THREAD_LOCAL(name)				\
+    tls_ ## name ## _get ()
+
+#else
+
+#    error "Unknown thread local support for this system. Pixman will not work with multiple threads. Define PIXMAN_NO_TLS to acknowledge and accept this limitation and compile pixman without thread-safety support."
+
+#endif
diff --git a/pixman/pixman/pixman-cpu.c b/pixman/pixman/pixman-cpu.c
index 8032fb42c..70253d1ea 100644
--- a/pixman/pixman/pixman-cpu.c
+++ b/pixman/pixman/pixman-cpu.c
@@ -1,598 +1,603 @@
-/*

- * 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.

- */

-#ifdef HAVE_CONFIG_H

-#include <config.h>

-#endif

-

-#include <string.h>

-

-#if defined(USE_ARM_SIMD) && defined(_MSC_VER)

-/* Needed for EXCEPTION_ILLEGAL_INSTRUCTION */

-#include <windows.h>

-#endif

-

-#include "pixman-private.h"

-

-#ifdef USE_VMX

-

-/* The CPU detection code needs to be in a file not compiled with

- * "-maltivec -mabi=altivec", as gcc would try to save vector register

- * across function calls causing SIGILL on cpus without Altivec/vmx.

- */

-static pixman_bool_t initialized = FALSE;

-static volatile pixman_bool_t have_vmx = TRUE;

-

-#ifdef __APPLE__

-#include <sys/sysctl.h>

-

-static pixman_bool_t

-pixman_have_vmx (void)

-{

-    if (!initialized)

-    {

-	size_t length = sizeof(have_vmx);

-	int error =

-	    sysctlbyname ("hw.optional.altivec", &have_vmx, &length, NULL, 0);

-

-	if (error)

-	    have_vmx = FALSE;

-

-	initialized = TRUE;

-    }

-    return have_vmx;

-}

-

-#elif defined (__OpenBSD__)

-#include <sys/param.h>

-#include <sys/sysctl.h>

-#include <machine/cpu.h>

-

-static pixman_bool_t

-pixman_have_vmx (void)

-{

-    if (!initialized)

-    {

-	int mib[2] = { CTL_MACHDEP, CPU_ALTIVEC };

-	size_t length = sizeof(have_vmx);

-	int error =

-	    sysctl (mib, 2, &have_vmx, &length, NULL, 0);

-

-	if (error != 0)

-	    have_vmx = FALSE;

-

-	initialized = TRUE;

-    }

-    return have_vmx;

-}

-

-#elif defined (__linux__)

-#include <sys/types.h>

-#include <sys/stat.h>

-#include <fcntl.h>

-#include <unistd.h>

-#include <stdio.h>

-#include <linux/auxvec.h>

-#include <asm/cputable.h>

-

-static pixman_bool_t

-pixman_have_vmx (void)

-{

-    if (!initialized)

-    {

-	char fname[64];

-	unsigned long buf[64];

-	ssize_t count = 0;

-	pid_t pid;

-	int fd, i;

-

-	pid = getpid ();

-	snprintf (fname, sizeof(fname) - 1, "/proc/%d/auxv", pid);

-

-	fd = open (fname, O_RDONLY);

-	if (fd >= 0)

-	{

-	    for (i = 0; i <= (count / sizeof(unsigned long)); i += 2)

-	    {

-		/* Read more if buf is empty... */

-		if (i == (count / sizeof(unsigned long)))

-		{

-		    count = read (fd, buf, sizeof(buf));

-		    if (count <= 0)

-			break;

-		    i = 0;

-		}

-

-		if (buf[i] == AT_HWCAP)

-		{

-		    have_vmx = !!(buf[i + 1] & PPC_FEATURE_HAS_ALTIVEC);

-		    initialized = TRUE;

-		    break;

-		}

-		else if (buf[i] == AT_NULL)

-		{

-		    break;

-		}

-	    }

-	    close (fd);

-	}

-    }

-    if (!initialized)

-    {

-	/* Something went wrong. Assume 'no' rather than playing

-	   fragile tricks with catching SIGILL. */

-	have_vmx = FALSE;

-	initialized = TRUE;

-    }

-

-    return have_vmx;

-}

-

-#else /* !__APPLE__ && !__OpenBSD__ && !__linux__ */

-#include <signal.h>

-#include <setjmp.h>

-

-static jmp_buf jump_env;

-

-static void

-vmx_test (int        sig,

-	  siginfo_t *si,

-	  void *     unused)

-{

-    longjmp (jump_env, 1);

-}

-

-static pixman_bool_t

-pixman_have_vmx (void)

-{

-    struct sigaction sa, osa;

-    int jmp_result;

-

-    if (!initialized)

-    {

-	sa.sa_flags = SA_SIGINFO;

-	sigemptyset (&sa.sa_mask);

-	sa.sa_sigaction = vmx_test;

-	sigaction (SIGILL, &sa, &osa);

-	jmp_result = setjmp (jump_env);

-	if (jmp_result == 0)

-	{

-	    asm volatile ( "vor 0, 0, 0" );

-	}

-	sigaction (SIGILL, &osa, NULL);

-	have_vmx = (jmp_result == 0);

-	initialized = TRUE;

-    }

-    return have_vmx;

-}

-

-#endif /* __APPLE__ */

-#endif /* USE_VMX */

-

-#if defined(USE_ARM_SIMD) || defined(USE_ARM_NEON)

-

-#if defined(_MSC_VER)

-

-#if defined(USE_ARM_SIMD)

-extern int pixman_msvc_try_arm_simd_op ();

-

-pixman_bool_t

-pixman_have_arm_simd (void)

-{

-    static pixman_bool_t initialized = FALSE;

-    static pixman_bool_t have_arm_simd = FALSE;

-

-    if (!initialized)

-    {

-	__try {

-	    pixman_msvc_try_arm_simd_op ();

-	    have_arm_simd = TRUE;

-	} __except (GetExceptionCode () == EXCEPTION_ILLEGAL_INSTRUCTION) {

-	    have_arm_simd = FALSE;

-	}

-	initialized = TRUE;

-    }

-

-    return have_arm_simd;

-}

-

-#endif /* USE_ARM_SIMD */

-

-#if defined(USE_ARM_NEON)

-extern int pixman_msvc_try_arm_neon_op ();

-

-pixman_bool_t

-pixman_have_arm_neon (void)

-{

-    static pixman_bool_t initialized = FALSE;

-    static pixman_bool_t have_arm_neon = FALSE;

-

-    if (!initialized)

-    {

-	__try

-	{

-	    pixman_msvc_try_arm_neon_op ();

-	    have_arm_neon = TRUE;

-	}

-	__except (GetExceptionCode () == EXCEPTION_ILLEGAL_INSTRUCTION)

-	{

-	    have_arm_neon = FALSE;

-	}

-	initialized = TRUE;

-    }

-

-    return have_arm_neon;

-}

-

-#endif /* USE_ARM_NEON */

-

-#else /* linux ELF */

-

-#include <stdlib.h>

-#include <unistd.h>

-#include <sys/types.h>

-#include <sys/stat.h>

-#include <sys/mman.h>

-#include <fcntl.h>

-#include <string.h>

-#include <elf.h>

-

-static pixman_bool_t arm_has_v7 = FALSE;

-static pixman_bool_t arm_has_v6 = FALSE;

-static pixman_bool_t arm_has_vfp = FALSE;

-static pixman_bool_t arm_has_neon = FALSE;

-static pixman_bool_t arm_has_iwmmxt = FALSE;

-static pixman_bool_t arm_tests_initialized = FALSE;

-

-static void

-pixman_arm_read_auxv ()

-{

-    int fd;

-    Elf32_auxv_t aux;

-

-    fd = open ("/proc/self/auxv", O_RDONLY);

-    if (fd >= 0)

-    {

-	while (read (fd, &aux, sizeof(Elf32_auxv_t)) == sizeof(Elf32_auxv_t))

-	{

-	    if (aux.a_type == AT_HWCAP)

-	    {

-		uint32_t hwcap = aux.a_un.a_val;

-		/* hardcode these values to avoid depending on specific

-		 * versions of the hwcap header, e.g. HWCAP_NEON

-		 */

-		arm_has_vfp = (hwcap & 64) != 0;

-		arm_has_iwmmxt = (hwcap & 512) != 0;

-		/* this flag is only present on kernel 2.6.29 */

-		arm_has_neon = (hwcap & 4096) != 0;

-	    }

-	    else if (aux.a_type == AT_PLATFORM)

-	    {

-		const char *plat = (const char*) aux.a_un.a_val;

-		if (strncmp (plat, "v7l", 3) == 0)

-		{

-		    arm_has_v7 = TRUE;

-		    arm_has_v6 = TRUE;

-		}

-		else if (strncmp (plat, "v6l", 3) == 0)

-		{

-		    arm_has_v6 = TRUE;

-		}

-	    }

-	}

-	close (fd);

-    }

-

-    arm_tests_initialized = TRUE;

-}

-

-#if defined(USE_ARM_SIMD)

-pixman_bool_t

-pixman_have_arm_simd (void)

-{

-    if (!arm_tests_initialized)

-	pixman_arm_read_auxv ();

-

-    return arm_has_v6;

-}

-

-#endif /* USE_ARM_SIMD */

-

-#if defined(USE_ARM_NEON)

-pixman_bool_t

-pixman_have_arm_neon (void)

-{

-    if (!arm_tests_initialized)

-	pixman_arm_read_auxv ();

-

-    return arm_has_neon;

-}

-

-#endif /* USE_ARM_NEON */

-

-#endif /* linux */

-

-#endif /* USE_ARM_SIMD || USE_ARM_NEON */

-

-#if defined(USE_MMX) || defined(USE_SSE2)

-/* The CPU detection code needs to be in a file not compiled with

- * "-mmmx -msse", as gcc would generate CMOV instructions otherwise

- * that would lead to SIGILL instructions on old CPUs that don't have

- * it.

- */

-#if !defined(__amd64__) && !defined(__x86_64__) && !defined(_M_AMD64)

-

-#ifdef HAVE_GETISAX

-#include <sys/auxv.h>

-#endif

-

-typedef enum

-{

-    NO_FEATURES = 0,

-    MMX = 0x1,

-    MMX_EXTENSIONS = 0x2,

-    SSE = 0x6,

-    SSE2 = 0x8,

-    CMOV = 0x10

-} cpu_features_t;

-

-

-static unsigned int

-detect_cpu_features (void)

-{

-    unsigned int features = 0;

-    unsigned int result = 0;

-

-#ifdef HAVE_GETISAX

-    if (getisax (&result, 1))

-    {

-	if (result & AV_386_CMOV)

-	    features |= CMOV;

-	if (result & AV_386_MMX)

-	    features |= MMX;

-	if (result & AV_386_AMD_MMX)

-	    features |= MMX_EXTENSIONS;

-	if (result & AV_386_SSE)

-	    features |= SSE;

-	if (result & AV_386_SSE2)

-	    features |= SSE2;

-    }

-#else

-    char vendor[13];

-#ifdef _MSC_VER

-    int vendor0 = 0, vendor1, vendor2;

-#endif

-    vendor[0] = 0;

-    vendor[12] = 0;

-

-#ifdef __GNUC__

-    /* see p. 118 of amd64 instruction set manual Vol3 */

-    /* We need to be careful about the handling of %ebx and

-     * %esp here. We can't declare either one as clobbered

-     * since they are special registers (%ebx is the "PIC

-     * register" holding an offset to global data, %esp the

-     * stack pointer), so we need to make sure they have their

-     * original values when we access the output operands.

-     */

-    __asm__ (

-        "pushf\n"

-        "pop %%eax\n"

-        "mov %%eax, %%ecx\n"

-        "xor $0x00200000, %%eax\n"

-        "push %%eax\n"

-        "popf\n"

-        "pushf\n"

-        "pop %%eax\n"

-        "mov $0x0, %%edx\n"

-        "xor %%ecx, %%eax\n"

-        "jz 1f\n"

-

-        "mov $0x00000000, %%eax\n"

-        "push %%ebx\n"

-        "cpuid\n"

-        "mov %%ebx, %%eax\n"

-        "pop %%ebx\n"

-        "mov %%eax, %1\n"

-        "mov %%edx, %2\n"

-        "mov %%ecx, %3\n"

-        "mov $0x00000001, %%eax\n"

-        "push %%ebx\n"

-        "cpuid\n"

-        "pop %%ebx\n"

-        "1:\n"

-        "mov %%edx, %0\n"

-	: "=r" (result),

-        "=m" (vendor[0]),

-        "=m" (vendor[4]),

-        "=m" (vendor[8])

-	:

-	: "%eax", "%ecx", "%edx"

-        );

-

-#elif defined (_MSC_VER)

-

-    _asm {

-	pushfd

-	pop eax

-	mov ecx, eax

-	xor eax, 00200000h

-	push eax

-	popfd

-	pushfd

-	pop eax

-	mov edx, 0

-	xor eax, ecx

-	jz nocpuid

-

-	mov eax, 0

-	push ebx

-	cpuid

-	mov eax, ebx

-	pop ebx

-	mov vendor0, eax

-	mov vendor1, edx

-	mov vendor2, ecx

-	mov eax, 1

-	push ebx

-	cpuid

-	pop ebx

-    nocpuid:

-	mov result, edx

-    }

-    memmove (vendor + 0, &vendor0, 4);

-    memmove (vendor + 4, &vendor1, 4);

-    memmove (vendor + 8, &vendor2, 4);

-

-#else

-#   error unsupported compiler

-#endif

-

-    features = 0;

-    if (result)

-    {

-	/* result now contains the standard feature bits */

-	if (result & (1 << 15))

-	    features |= CMOV;

-	if (result & (1 << 23))

-	    features |= MMX;

-	if (result & (1 << 25))

-	    features |= SSE;

-	if (result & (1 << 26))

-	    features |= SSE2;

-	if ((features & MMX) && !(features & SSE) &&

-	    (strcmp (vendor, "AuthenticAMD") == 0 ||

-	     strcmp (vendor, "Geode by NSC") == 0))

-	{

-	    /* check for AMD MMX extensions */

-#ifdef __GNUC__

-	    __asm__ (

-	        "	push %%ebx\n"

-	        "	mov $0x80000000, %%eax\n"

-	        "	cpuid\n"

-	        "	xor %%edx, %%edx\n"

-	        "	cmp $0x1, %%eax\n"

-	        "	jge 2f\n"

-	        "	mov $0x80000001, %%eax\n"

-	        "	cpuid\n"

-	        "2:\n"

-	        "	pop %%ebx\n"

-	        "	mov %%edx, %0\n"

-		: "=r" (result)

-		:

-		: "%eax", "%ecx", "%edx"

-	        );

-#elif defined _MSC_VER

-	    _asm {

-		push ebx

-		mov eax, 80000000h

-		cpuid

-		xor edx, edx

-		cmp eax, 1

-		jge notamd

-		mov eax, 80000001h

-		cpuid

-	    notamd:

-		pop ebx

-		mov result, edx

-	    }

-#endif

-	    if (result & (1 << 22))

-		features |= MMX_EXTENSIONS;

-	}

-    }

-#endif /* HAVE_GETISAX */

-

-    return features;

-}

-

-static pixman_bool_t

-pixman_have_mmx (void)

-{

-    static pixman_bool_t initialized = FALSE;

-    static pixman_bool_t mmx_present;

-

-    if (!initialized)

-    {

-	unsigned int features = detect_cpu_features ();

-	mmx_present = (features & (MMX | MMX_EXTENSIONS)) == (MMX | MMX_EXTENSIONS);

-	initialized = TRUE;

-    }

-

-    return mmx_present;

-}

-

-#ifdef USE_SSE2

-static pixman_bool_t

-pixman_have_sse2 (void)

-{

-    static pixman_bool_t initialized = FALSE;

-    static pixman_bool_t sse2_present;

-

-    if (!initialized)

-    {

-	unsigned int features = detect_cpu_features ();

-	sse2_present = (features & (MMX | MMX_EXTENSIONS | SSE | SSE2)) == (MMX | MMX_EXTENSIONS | SSE | SSE2);

-	initialized = TRUE;

-    }

-

-    return sse2_present;

-}

-

-#endif

-

-#else /* __amd64__ */

-#ifdef USE_MMX

-#define pixman_have_mmx() TRUE

-#endif

-#ifdef USE_SSE2

-#define pixman_have_sse2() TRUE

-#endif

-#endif /* __amd64__ */

-#endif

-

-pixman_implementation_t *

-_pixman_choose_implementation (void)

-{

-#ifdef USE_SSE2

-    if (pixman_have_sse2 ())

-	return _pixman_implementation_create_sse2 ();

-#endif

-#ifdef USE_MMX

-    if (pixman_have_mmx ())

-	return _pixman_implementation_create_mmx ();

-#endif

-

-#ifdef USE_ARM_NEON

-    if (pixman_have_arm_neon ())

-	return _pixman_implementation_create_arm_neon ();

-#endif

-#ifdef USE_ARM_SIMD

-    if (pixman_have_arm_simd ())

-	return _pixman_implementation_create_arm_simd ();

-#endif

-#ifdef USE_VMX

-    if (pixman_have_vmx ())

-	return _pixman_implementation_create_vmx ();

-#endif

-

-    return _pixman_implementation_create_fast_path ();

-}

-

+/*
+ * 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.
+ */
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <string.h>
+
+#if defined(USE_ARM_SIMD) && defined(_MSC_VER)
+/* Needed for EXCEPTION_ILLEGAL_INSTRUCTION */
+#include <windows.h>
+#endif
+
+#include "pixman-private.h"
+
+#ifdef USE_VMX
+
+/* The CPU detection code needs to be in a file not compiled with
+ * "-maltivec -mabi=altivec", as gcc would try to save vector register
+ * across function calls causing SIGILL on cpus without Altivec/vmx.
+ */
+static pixman_bool_t initialized = FALSE;
+static volatile pixman_bool_t have_vmx = TRUE;
+
+#ifdef __APPLE__
+#include <sys/sysctl.h>
+
+static pixman_bool_t
+pixman_have_vmx (void)
+{
+    if (!initialized)
+    {
+	size_t length = sizeof(have_vmx);
+	int error =
+	    sysctlbyname ("hw.optional.altivec", &have_vmx, &length, NULL, 0);
+
+	if (error)
+	    have_vmx = FALSE;
+
+	initialized = TRUE;
+    }
+    return have_vmx;
+}
+
+#elif defined (__OpenBSD__)
+#include <sys/param.h>
+#include <sys/sysctl.h>
+#include <machine/cpu.h>
+
+static pixman_bool_t
+pixman_have_vmx (void)
+{
+    if (!initialized)
+    {
+	int mib[2] = { CTL_MACHDEP, CPU_ALTIVEC };
+	size_t length = sizeof(have_vmx);
+	int error =
+	    sysctl (mib, 2, &have_vmx, &length, NULL, 0);
+
+	if (error != 0)
+	    have_vmx = FALSE;
+
+	initialized = TRUE;
+    }
+    return have_vmx;
+}
+
+#elif defined (__linux__)
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <stdio.h>
+#include <linux/auxvec.h>
+#include <asm/cputable.h>
+
+static pixman_bool_t
+pixman_have_vmx (void)
+{
+    if (!initialized)
+    {
+	char fname[64];
+	unsigned long buf[64];
+	ssize_t count = 0;
+	pid_t pid;
+	int fd, i;
+
+	pid = getpid ();
+	snprintf (fname, sizeof(fname) - 1, "/proc/%d/auxv", pid);
+
+	fd = open (fname, O_RDONLY);
+	if (fd >= 0)
+	{
+	    for (i = 0; i <= (count / sizeof(unsigned long)); i += 2)
+	    {
+		/* Read more if buf is empty... */
+		if (i == (count / sizeof(unsigned long)))
+		{
+		    count = read (fd, buf, sizeof(buf));
+		    if (count <= 0)
+			break;
+		    i = 0;
+		}
+
+		if (buf[i] == AT_HWCAP)
+		{
+		    have_vmx = !!(buf[i + 1] & PPC_FEATURE_HAS_ALTIVEC);
+		    initialized = TRUE;
+		    break;
+		}
+		else if (buf[i] == AT_NULL)
+		{
+		    break;
+		}
+	    }
+	    close (fd);
+	}
+    }
+    if (!initialized)
+    {
+	/* Something went wrong. Assume 'no' rather than playing
+	   fragile tricks with catching SIGILL. */
+	have_vmx = FALSE;
+	initialized = TRUE;
+    }
+
+    return have_vmx;
+}
+
+#else /* !__APPLE__ && !__OpenBSD__ && !__linux__ */
+#include <signal.h>
+#include <setjmp.h>
+
+static jmp_buf jump_env;
+
+static void
+vmx_test (int        sig,
+	  siginfo_t *si,
+	  void *     unused)
+{
+    longjmp (jump_env, 1);
+}
+
+static pixman_bool_t
+pixman_have_vmx (void)
+{
+    struct sigaction sa, osa;
+    int jmp_result;
+
+    if (!initialized)
+    {
+	sa.sa_flags = SA_SIGINFO;
+	sigemptyset (&sa.sa_mask);
+	sa.sa_sigaction = vmx_test;
+	sigaction (SIGILL, &sa, &osa);
+	jmp_result = setjmp (jump_env);
+	if (jmp_result == 0)
+	{
+	    asm volatile ( "vor 0, 0, 0" );
+	}
+	sigaction (SIGILL, &osa, NULL);
+	have_vmx = (jmp_result == 0);
+	initialized = TRUE;
+    }
+    return have_vmx;
+}
+
+#endif /* __APPLE__ */
+#endif /* USE_VMX */
+
+#if defined(USE_ARM_SIMD) || defined(USE_ARM_NEON)
+
+#if defined(_MSC_VER)
+
+#if defined(USE_ARM_SIMD)
+extern int pixman_msvc_try_arm_simd_op ();
+
+pixman_bool_t
+pixman_have_arm_simd (void)
+{
+    static pixman_bool_t initialized = FALSE;
+    static pixman_bool_t have_arm_simd = FALSE;
+
+    if (!initialized)
+    {
+	__try {
+	    pixman_msvc_try_arm_simd_op ();
+	    have_arm_simd = TRUE;
+	} __except (GetExceptionCode () == EXCEPTION_ILLEGAL_INSTRUCTION) {
+	    have_arm_simd = FALSE;
+	}
+	initialized = TRUE;
+    }
+
+    return have_arm_simd;
+}
+
+#endif /* USE_ARM_SIMD */
+
+#if defined(USE_ARM_NEON)
+extern int pixman_msvc_try_arm_neon_op ();
+
+pixman_bool_t
+pixman_have_arm_neon (void)
+{
+    static pixman_bool_t initialized = FALSE;
+    static pixman_bool_t have_arm_neon = FALSE;
+
+    if (!initialized)
+    {
+	__try
+	{
+	    pixman_msvc_try_arm_neon_op ();
+	    have_arm_neon = TRUE;
+	}
+	__except (GetExceptionCode () == EXCEPTION_ILLEGAL_INSTRUCTION)
+	{
+	    have_arm_neon = FALSE;
+	}
+	initialized = TRUE;
+    }
+
+    return have_arm_neon;
+}
+
+#endif /* USE_ARM_NEON */
+
+#elif defined (__linux__) /* linux ELF */
+
+#include <stdlib.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+#include <fcntl.h>
+#include <string.h>
+#include <elf.h>
+
+static pixman_bool_t arm_has_v7 = FALSE;
+static pixman_bool_t arm_has_v6 = FALSE;
+static pixman_bool_t arm_has_vfp = FALSE;
+static pixman_bool_t arm_has_neon = FALSE;
+static pixman_bool_t arm_has_iwmmxt = FALSE;
+static pixman_bool_t arm_tests_initialized = FALSE;
+
+static void
+pixman_arm_read_auxv ()
+{
+    int fd;
+    Elf32_auxv_t aux;
+
+    fd = open ("/proc/self/auxv", O_RDONLY);
+    if (fd >= 0)
+    {
+	while (read (fd, &aux, sizeof(Elf32_auxv_t)) == sizeof(Elf32_auxv_t))
+	{
+	    if (aux.a_type == AT_HWCAP)
+	    {
+		uint32_t hwcap = aux.a_un.a_val;
+		/* hardcode these values to avoid depending on specific
+		 * versions of the hwcap header, e.g. HWCAP_NEON
+		 */
+		arm_has_vfp = (hwcap & 64) != 0;
+		arm_has_iwmmxt = (hwcap & 512) != 0;
+		/* this flag is only present on kernel 2.6.29 */
+		arm_has_neon = (hwcap & 4096) != 0;
+	    }
+	    else if (aux.a_type == AT_PLATFORM)
+	    {
+		const char *plat = (const char*) aux.a_un.a_val;
+		if (strncmp (plat, "v7l", 3) == 0)
+		{
+		    arm_has_v7 = TRUE;
+		    arm_has_v6 = TRUE;
+		}
+		else if (strncmp (plat, "v6l", 3) == 0)
+		{
+		    arm_has_v6 = TRUE;
+		}
+	    }
+	}
+	close (fd);
+    }
+
+    arm_tests_initialized = TRUE;
+}
+
+#if defined(USE_ARM_SIMD)
+pixman_bool_t
+pixman_have_arm_simd (void)
+{
+    if (!arm_tests_initialized)
+	pixman_arm_read_auxv ();
+
+    return arm_has_v6;
+}
+
+#endif /* USE_ARM_SIMD */
+
+#if defined(USE_ARM_NEON)
+pixman_bool_t
+pixman_have_arm_neon (void)
+{
+    if (!arm_tests_initialized)
+	pixman_arm_read_auxv ();
+
+    return arm_has_neon;
+}
+
+#endif /* USE_ARM_NEON */
+
+#else /* linux ELF */
+
+#define pixman_have_arm_simd() FALSE
+#define pixman_have_arm_neon() FALSE
+
+#endif
+
+#endif /* USE_ARM_SIMD || USE_ARM_NEON */
+
+#if defined(USE_MMX) || defined(USE_SSE2)
+/* The CPU detection code needs to be in a file not compiled with
+ * "-mmmx -msse", as gcc would generate CMOV instructions otherwise
+ * that would lead to SIGILL instructions on old CPUs that don't have
+ * it.
+ */
+#if !defined(__amd64__) && !defined(__x86_64__) && !defined(_M_AMD64)
+
+#ifdef HAVE_GETISAX
+#include <sys/auxv.h>
+#endif
+
+typedef enum
+{
+    NO_FEATURES = 0,
+    MMX = 0x1,
+    MMX_EXTENSIONS = 0x2,
+    SSE = 0x6,
+    SSE2 = 0x8,
+    CMOV = 0x10
+} cpu_features_t;
+
+
+static unsigned int
+detect_cpu_features (void)
+{
+    unsigned int features = 0;
+    unsigned int result = 0;
+
+#ifdef HAVE_GETISAX
+    if (getisax (&result, 1))
+    {
+	if (result & AV_386_CMOV)
+	    features |= CMOV;
+	if (result & AV_386_MMX)
+	    features |= MMX;
+	if (result & AV_386_AMD_MMX)
+	    features |= MMX_EXTENSIONS;
+	if (result & AV_386_SSE)
+	    features |= SSE;
+	if (result & AV_386_SSE2)
+	    features |= SSE2;
+    }
+#else
+    char vendor[13];
+#ifdef _MSC_VER
+    int vendor0 = 0, vendor1, vendor2;
+#endif
+    vendor[0] = 0;
+    vendor[12] = 0;
+
+#ifdef __GNUC__
+    /* see p. 118 of amd64 instruction set manual Vol3 */
+    /* We need to be careful about the handling of %ebx and
+     * %esp here. We can't declare either one as clobbered
+     * since they are special registers (%ebx is the "PIC
+     * register" holding an offset to global data, %esp the
+     * stack pointer), so we need to make sure they have their
+     * original values when we access the output operands.
+     */
+    __asm__ (
+        "pushf\n"
+        "pop %%eax\n"
+        "mov %%eax, %%ecx\n"
+        "xor $0x00200000, %%eax\n"
+        "push %%eax\n"
+        "popf\n"
+        "pushf\n"
+        "pop %%eax\n"
+        "mov $0x0, %%edx\n"
+        "xor %%ecx, %%eax\n"
+        "jz 1f\n"
+
+        "mov $0x00000000, %%eax\n"
+        "push %%ebx\n"
+        "cpuid\n"
+        "mov %%ebx, %%eax\n"
+        "pop %%ebx\n"
+        "mov %%eax, %1\n"
+        "mov %%edx, %2\n"
+        "mov %%ecx, %3\n"
+        "mov $0x00000001, %%eax\n"
+        "push %%ebx\n"
+        "cpuid\n"
+        "pop %%ebx\n"
+        "1:\n"
+        "mov %%edx, %0\n"
+	: "=r" (result),
+        "=m" (vendor[0]),
+        "=m" (vendor[4]),
+        "=m" (vendor[8])
+	:
+	: "%eax", "%ecx", "%edx"
+        );
+
+#elif defined (_MSC_VER)
+
+    _asm {
+	pushfd
+	pop eax
+	mov ecx, eax
+	xor eax, 00200000h
+	push eax
+	popfd
+	pushfd
+	pop eax
+	mov edx, 0
+	xor eax, ecx
+	jz nocpuid
+
+	mov eax, 0
+	push ebx
+	cpuid
+	mov eax, ebx
+	pop ebx
+	mov vendor0, eax
+	mov vendor1, edx
+	mov vendor2, ecx
+	mov eax, 1
+	push ebx
+	cpuid
+	pop ebx
+    nocpuid:
+	mov result, edx
+    }
+    memmove (vendor + 0, &vendor0, 4);
+    memmove (vendor + 4, &vendor1, 4);
+    memmove (vendor + 8, &vendor2, 4);
+
+#else
+#   error unsupported compiler
+#endif
+
+    features = 0;
+    if (result)
+    {
+	/* result now contains the standard feature bits */
+	if (result & (1 << 15))
+	    features |= CMOV;
+	if (result & (1 << 23))
+	    features |= MMX;
+	if (result & (1 << 25))
+	    features |= SSE;
+	if (result & (1 << 26))
+	    features |= SSE2;
+	if ((features & MMX) && !(features & SSE) &&
+	    (strcmp (vendor, "AuthenticAMD") == 0 ||
+	     strcmp (vendor, "Geode by NSC") == 0))
+	{
+	    /* check for AMD MMX extensions */
+#ifdef __GNUC__
+	    __asm__ (
+	        "	push %%ebx\n"
+	        "	mov $0x80000000, %%eax\n"
+	        "	cpuid\n"
+	        "	xor %%edx, %%edx\n"
+	        "	cmp $0x1, %%eax\n"
+	        "	jge 2f\n"
+	        "	mov $0x80000001, %%eax\n"
+	        "	cpuid\n"
+	        "2:\n"
+	        "	pop %%ebx\n"
+	        "	mov %%edx, %0\n"
+		: "=r" (result)
+		:
+		: "%eax", "%ecx", "%edx"
+	        );
+#elif defined _MSC_VER
+	    _asm {
+		push ebx
+		mov eax, 80000000h
+		cpuid
+		xor edx, edx
+		cmp eax, 1
+		jge notamd
+		mov eax, 80000001h
+		cpuid
+	    notamd:
+		pop ebx
+		mov result, edx
+	    }
+#endif
+	    if (result & (1 << 22))
+		features |= MMX_EXTENSIONS;
+	}
+    }
+#endif /* HAVE_GETISAX */
+
+    return features;
+}
+
+static pixman_bool_t
+pixman_have_mmx (void)
+{
+    static pixman_bool_t initialized = FALSE;
+    static pixman_bool_t mmx_present;
+
+    if (!initialized)
+    {
+	unsigned int features = detect_cpu_features ();
+	mmx_present = (features & (MMX | MMX_EXTENSIONS)) == (MMX | MMX_EXTENSIONS);
+	initialized = TRUE;
+    }
+
+    return mmx_present;
+}
+
+#ifdef USE_SSE2
+static pixman_bool_t
+pixman_have_sse2 (void)
+{
+    static pixman_bool_t initialized = FALSE;
+    static pixman_bool_t sse2_present;
+
+    if (!initialized)
+    {
+	unsigned int features = detect_cpu_features ();
+	sse2_present = (features & (MMX | MMX_EXTENSIONS | SSE | SSE2)) == (MMX | MMX_EXTENSIONS | SSE | SSE2);
+	initialized = TRUE;
+    }
+
+    return sse2_present;
+}
+
+#endif
+
+#else /* __amd64__ */
+#ifdef USE_MMX
+#define pixman_have_mmx() TRUE
+#endif
+#ifdef USE_SSE2
+#define pixman_have_sse2() TRUE
+#endif
+#endif /* __amd64__ */
+#endif
+
+pixman_implementation_t *
+_pixman_choose_implementation (void)
+{
+#ifdef USE_SSE2
+    if (pixman_have_sse2 ())
+	return _pixman_implementation_create_sse2 ();
+#endif
+#ifdef USE_MMX
+    if (pixman_have_mmx ())
+	return _pixman_implementation_create_mmx ();
+#endif
+
+#ifdef USE_ARM_NEON
+    if (pixman_have_arm_neon ())
+	return _pixman_implementation_create_arm_neon ();
+#endif
+#ifdef USE_ARM_SIMD
+    if (pixman_have_arm_simd ())
+	return _pixman_implementation_create_arm_simd ();
+#endif
+#ifdef USE_VMX
+    if (pixman_have_vmx ())
+	return _pixman_implementation_create_vmx ();
+#endif
+
+    return _pixman_implementation_create_fast_path ();
+}
+
diff --git a/pixman/pixman/pixman-fast-path.c b/pixman/pixman/pixman-fast-path.c
index 01a3017c7..f103b4cf1 100644
--- a/pixman/pixman/pixman-fast-path.c
+++ b/pixman/pixman/pixman-fast-path.c
@@ -1,1937 +1,1937 @@
-/* -*- 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 (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 (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, 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 (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)

-	    {

-		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 (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 (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 (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 (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 (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 (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 (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 (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,

-				     uint16_t *      src,

-				     int32_t         w,

-				     pixman_fixed_t  vx,

-				     pixman_fixed_t  unit_x,

-				     pixman_fixed_t  max_vx)

-{

-    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++);

-	    }

-	}

-    }

-}

-

-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),

-

-    {   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 (void)

-{

-    pixman_implementation_t *general = _pixman_implementation_create_general ();

-    pixman_implementation_t *imp = _pixman_implementation_create (general, 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 (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 (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, 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 (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)
+	    {
+		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 (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 (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 (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 (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 (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 (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 (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 (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,
+				     uint16_t *      src,
+				     int32_t         w,
+				     pixman_fixed_t  vx,
+				     pixman_fixed_t  unit_x,
+				     pixman_fixed_t  max_vx)
+{
+    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++);
+	    }
+	}
+    }
+}
+
+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),
+
+    {   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 (void)
+{
+    pixman_implementation_t *general = _pixman_implementation_create_general ();
+    pixman_implementation_t *imp = _pixman_implementation_create (general, 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 0273cd33b..b46937a28 100644
--- a/pixman/pixman/pixman-fast-path.h
+++ b/pixman/pixman/pixman-fast-path.h
@@ -1,445 +1,449 @@
-/* -*- 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,							\

-		    src_type_t     *src,							\

-		    int32_t         w,								\

-		    pixman_fixed_t  vx,								\

-		    pixman_fixed_t  unit_x,							\

-		    pixman_fixed_t  max_vx)							\

-{												\

-	uint32_t   d;										\

-	src_type_t s1, s2;									\

-	uint8_t    a1, a2;									\

-	int        x1, x2;									\

-												\

-	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(scale_func_name, scanline_func, src_type_t, dst_type_t,		\

-			      repeat_mode)							\

-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;									\

-    src_type_t *src_first_line;									\

-    int       y;										\

-    pixman_fixed_t max_vx = 0; /* 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;										\

-    int       src_stride, dst_stride;								\

-												\

-    PIXMAN_IMAGE_GET_LINE (dst_image, dst_x, dst_y, dst_type_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, 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;									\

-												\

-	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 (dst, src, left_pad, 0, 0, 0);					\

-	    }											\

-	    if (width > 0)									\

-	    {											\

-		scanline_func (dst + left_pad, src, width, vx, unit_x, 0);			\

-	    }											\

-	    if (right_pad > 0)									\

-	    {											\

-		scanline_func (dst + left_pad + width, src + src_image->bits.width - 1,		\

-			        right_pad, 0, 0, 0);						\

-	    }											\

-	}											\

-	else if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NONE)				\

-	{											\

-	    static src_type_t zero = 0;								\

-	    if (y < 0 || y >= src_image->bits.height)						\

-	    {											\

-		scanline_func (dst, &zero, left_pad + width + right_pad, 0, 0, 0);		\

-		continue;									\

-	    }											\

-	    src = src_first_line + src_stride * y;						\

-	    if (left_pad > 0)									\

-	    {											\

-		scanline_func (dst, &zero, left_pad, 0, 0, 0);					\

-	    }											\

-	    if (width > 0)									\

-	    {											\

-		scanline_func (dst + left_pad, src, width, vx, unit_x, 0);			\

-	    }											\

-	    if (right_pad > 0)									\

-	    {											\

-		scanline_func (dst + left_pad + width, &zero, right_pad, 0, 0, 0);		\

-	    }											\

-	}											\

-	else											\

-	{											\

-	    src = src_first_line + src_stride * y;						\

-	    scanline_func (dst, src, width, vx, unit_x, max_vx);				\

-	}											\

-    }												\

-}

-

-#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(scale_func_name##_##OP,					\

-			  scaled_nearest_scanline_ ## scale_func_name ## _ ## OP,	\

-			  src_type_t, dst_type_t, repeat_mode)				\

-											\

-    extern int no_such_variable

-

-

-#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,	\

-    }

-

-/* 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)

-

-#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,							\
+		    src_type_t     *src,							\
+		    int32_t         w,								\
+		    pixman_fixed_t  vx,								\
+		    pixman_fixed_t  unit_x,							\
+		    pixman_fixed_t  max_vx)							\
+{												\
+	uint32_t   d;										\
+	src_type_t s1, s2;									\
+	uint8_t    a1, a2;									\
+	int        x1, x2;									\
+												\
+	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, dst_type_t,	\
+				  repeat_mode)							\
+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;									\
+    src_type_t *src_first_line;									\
+    int       y;										\
+    pixman_fixed_t max_vx = 0; /* 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;										\
+    int       src_stride, dst_stride;								\
+												\
+    PIXMAN_IMAGE_GET_LINE (dst_image, dst_x, dst_y, dst_type_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, 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;									\
+												\
+	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 (dst, src, left_pad, 0, 0, 0);					\
+	    }											\
+	    if (width > 0)									\
+	    {											\
+		scanline_func (dst + left_pad, src, width, vx, unit_x, 0);			\
+	    }											\
+	    if (right_pad > 0)									\
+	    {											\
+		scanline_func (dst + left_pad + width, src + src_image->bits.width - 1,		\
+			        right_pad, 0, 0, 0);						\
+	    }											\
+	}											\
+	else if (PIXMAN_REPEAT_ ## repeat_mode == PIXMAN_REPEAT_NONE)				\
+	{											\
+	    static src_type_t zero[1] = { 0 };							\
+	    if (y < 0 || y >= src_image->bits.height)						\
+	    {											\
+		scanline_func (dst, zero, left_pad + width + right_pad, 0, 0, 0);		\
+		continue;									\
+	    }											\
+	    src = src_first_line + src_stride * y;						\
+	    if (left_pad > 0)									\
+	    {											\
+		scanline_func (dst, zero, left_pad, 0, 0, 0);					\
+	    }											\
+	    if (width > 0)									\
+	    {											\
+		scanline_func (dst + left_pad, src, width, vx, unit_x, 0);			\
+	    }											\
+	    if (right_pad > 0)									\
+	    {											\
+		scanline_func (dst + left_pad + width, zero, right_pad, 0, 0, 0);		\
+	    }											\
+	}											\
+	else											\
+	{											\
+	    src = src_first_line + src_stride * y;						\
+	    scanline_func (dst, src, width, vx, unit_x, max_vx);				\
+	}											\
+    }												\
+}
+
+/* A workaround for old sun studio, see: https://bugs.freedesktop.org/show_bug.cgi?id=32764 */
+#define FAST_NEAREST_MAINLOOP(scale_func_name, scanline_func, src_type_t, dst_type_t,		\
+			      repeat_mode)							\
+	FAST_NEAREST_MAINLOOP_INT(_ ## 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_INT(_ ## 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,	\
+    }
+
+/* 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)
+
+#endif
diff --git a/pixman/pixman/pixman-matrix.c b/pixman/pixman/pixman-matrix.c
index abdfa0525..f2f67ab41 100644
--- a/pixman/pixman/pixman-matrix.c
+++ b/pixman/pixman/pixman-matrix.c
@@ -425,7 +425,8 @@ pixman_transform_is_inverse (const struct pixman_transform *a,
 {
     struct pixman_transform t;
 
-    pixman_transform_multiply (&t, a, b);
+    if (!pixman_transform_multiply (&t, a, b))
+	return FALSE;
 
     return pixman_transform_is_identity (&t);
 }
diff --git a/pixman/pixman/pixman-sse2.c b/pixman/pixman/pixman-sse2.c
index b19043bbd..94ba54cbf 100644
--- a/pixman/pixman/pixman-sse2.c
+++ b/pixman/pixman/pixman-sse2.c
@@ -1,6031 +1,6031 @@
-/*

- * 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 (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 (s, d);

-	ps++;

-	if (pm)

-	    pm++;

-	w--;

-    }

-

-    while (w >= 4)

-    {

-	/* I'm loading unaligned because I'm not sure about

-	 * the address alignment.

-	 */

-	xmm_src_hi = combine4 ((__m128i*)ps, (__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;

-	ps += 4;

-	pd += 4;

-	if (pm)

-	    pm += 4;

-    }

-

-    while (w)

-    {

-	d = *pd;

-	s = combine1 (ps, pm);

-

-	*pd++ = core_combine_over_u_pixel_sse2 (s, d);

-	ps++;

-	if (pm)

-	    pm++;

-

-	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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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)

-{

-    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;

-

-    /* 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 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),

-

-    { 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;

-}

-

-#if defined(__GNUC__) && !defined(__x86_64__) && !defined(__amd64__)

-__attribute__((__force_align_arg_pointer__))

-#endif

-pixman_implementation_t *

-_pixman_implementation_create_sse2 (void)

-{

-#ifdef USE_MMX

-    pixman_implementation_t *fallback = _pixman_implementation_create_mmx ();

-#else

-    pixman_implementation_t *fallback = _pixman_implementation_create_fast_path ();

-#endif

-    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;

-

-    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 (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 (s, d);
+	ps++;
+	if (pm)
+	    pm++;
+	w--;
+    }
+
+    while (w >= 4)
+    {
+	/* I'm loading unaligned because I'm not sure about
+	 * the address alignment.
+	 */
+	xmm_src_hi = combine4 ((__m128i*)ps, (__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;
+	ps += 4;
+	pd += 4;
+	if (pm)
+	    pm += 4;
+    }
+
+    while (w)
+    {
+	d = *pd;
+	s = combine1 (ps, pm);
+
+	*pd++ = core_combine_over_u_pixel_sse2 (s, d);
+	ps++;
+	if (pm)
+	    pm++;
+
+	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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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 (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)
+{
+    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;
+
+    /* 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 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),
+
+    { 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;
+}
+
+#if defined(__GNUC__) && !defined(__x86_64__) && !defined(__amd64__)
+__attribute__((__force_align_arg_pointer__))
+#endif
+pixman_implementation_t *
+_pixman_implementation_create_sse2 (void)
+{
+#ifdef USE_MMX
+    pixman_implementation_t *fallback = _pixman_implementation_create_mmx ();
+#else
+    pixman_implementation_t *fallback = _pixman_implementation_create_fast_path ();
+#endif
+    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;
+
+    return imp;
+}
+
+#endif /* USE_SSE2 */
diff --git a/pixman/test/Makefile.am b/pixman/test/Makefile.am
index 19c4f8006..71e535374 100644
--- a/pixman/test/Makefile.am
+++ b/pixman/test/Makefile.am
@@ -23,40 +23,61 @@ TESTPROGRAMS =			\
 	composite
 
 a1_trap_test_LDADD = $(TEST_LDADD)
+a1_trap_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
+
 fetch_test_LDADD = $(TEST_LDADD)
+fetch_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
+
 trap_crasher_LDADD = $(TEST_LDADD)
+trap_crasher_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
+
 oob_test_LDADD = $(TEST_LDADD)
+oob_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
+
 scaling_crash_test_LDADD = $(TEST_LDADD)
+scaling_crash_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
+
 region_translate_test_LDADD = $(TEST_LDADD)
+region_translate_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 
 pdf_op_test_LDADD = $(TEST_LDADD)
+pdf_op_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 pdf_op_test_SOURCES = pdf-op-test.c utils.c utils.h
 
 region_test_LDADD = $(TEST_LDADD)
+region_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 region_test_SOURCES = region-test.c utils.c utils.h
 
 blitters_test_LDADD = $(TEST_LDADD)
+blitters_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 blitters_test_SOURCES = blitters-test.c utils.c utils.h
 
 scaling_test_LDADD = $(TEST_LDADD)
+scaling_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 scaling_test_SOURCES = scaling-test.c utils.c utils.h
 
 affine_test_LDADD = $(TEST_LDADD)
+affine_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 affine_test_SOURCES = affine-test.c utils.c utils.h
 
 alphamap_LDADD = $(TEST_LDADD)
+alphamap_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 alphamap_SOURCES = alphamap.c utils.c utils.h
 
 alpha_loop_LDADD = $(TEST_LDADD)
+alpha_loop_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 alpha_loop_SOURCES = alpha-loop.c utils.c utils.h
 
 composite_LDADD = $(TEST_LDADD)
+composite_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 composite_SOURCES = composite.c utils.c utils.h
 
 gradient_crash_test_LDADD = $(TEST_LDADD)
+gradient_crash_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 gradient_crash_test_SOURCES = gradient-crash-test.c utils.c utils.h
 
 stress_test_LDADD = $(TEST_LDADD)
+stress_test_LDFLAGS = @TESTPROGS_EXTRA_LDFLAGS@
 stress_test_SOURCES = stress-test.c utils.c utils.h
 
 # GTK using test programs