Merge remote-tracking branch 'origin/released'

* origin/released:
  pixman libxcb mesa xserver xkeyboard-config git update 8 oct 2012

Conflicts:
	xorg-server/Xext/xf86bigfont.c
	xorg-server/dix/dispatch.c
	xorg-server/include/dixstruct.h
	xorg-server/os/connection.c

79 files changed, 3085 insertions, 1667 deletions

diff --git a/libxcb/NEWS b/libxcb/NEWS
index 66d6253ae..533d72b87 100644
--- a/libxcb/NEWS
+++ b/libxcb/NEWS
@@ -1,3 +1,16 @@
+Release 1.9 (2012-10-05)
+========================
+
+* Always include "config.h" at the start of all C source files.
+* Add AC_USE_SYSTEM_EXTENSIONS to allow use of more system functionality
+* Return connection failure if display string specifies non-existent screen
+* c_client: Fix parallel-make issue creating 'man' directory
+* xcb_connect: launchd: Don't fall back on tcp if $DISPLAY is a path to a launchd socket
+* c_client.py: generate manpages
+* Allow xcb_send_request with >MAX_IOV iovecs
+* Add a .gitignore for src/man/
+* Fix a multi-thread deadlock
+
 Release 1.8.1 (2012-03-09)
 ==========================
 - Fix a busy loop on BSD and Mac OS
diff --git a/libxcb/configure.ac b/libxcb/configure.ac
index 3f44b534d..736438305 100644
--- a/libxcb/configure.ac
+++ b/libxcb/configure.ac
@@ -4,7 +4,7 @@
 # Initialize Autoconf
 AC_PREREQ(2.57)
 AC_INIT([libxcb],
-        1.8.1,
+        1.9,
         [xcb@lists.freedesktop.org])
 AC_CONFIG_SRCDIR([xcb.pc.in])
 # Set common system defines for POSIX extensions, such as _GNU_SOURCE
diff --git a/libxcb/src/Makefile.am b/libxcb/src/Makefile.am
index fed691447..6b5a2c3c4 100644
--- a/libxcb/src/Makefile.am
+++ b/libxcb/src/Makefile.am
@@ -227,7 +227,8 @@ noinst_HEADERS = xcbint.h
 
 STATIC_MANS = man/xcb-examples.3 man/xcb-requests.3
 BUILT_MANS = man/xcb_*.3
-man_MANS = = $(STATIC_MANS) $(BUILT_MANS)
+man_MANS = $(STATIC_MANS) $(BUILT_MANS)
+EXTRA_DIST = $(STATIC_MANS)
 
 BUILT_SOURCES = $(EXTSOURCES) $(BUILT_MANS)
 CLEANFILES = $(EXTSOURCES) $(EXTHEADERS) $(BUILT_MANS)
diff --git a/libxcb/xcb-proto/NEWS b/libxcb/xcb-proto/NEWS
index 3e50762ab..ebe184532 100644
--- a/libxcb/xcb-proto/NEWS
+++ b/libxcb/xcb-proto/NEWS
@@ -1,3 +1,9 @@
+Release 1.8 (2012-10-05)
+========================
+* dri2: Update to DRI2 1.4
+* XKB: Fix broken events
+* xproto: add doc tags, xcbgen: handle doc tags
+
 Release 1.7.1 (2012-03-23)
 ==========================
 * Add ge and xf86vidmode protocol descriptions
diff --git a/libxcb/xcb-proto/configure.ac b/libxcb/xcb-proto/configure.ac
index e20c9aa1e..ba6a90cd8 100644
--- a/libxcb/xcb-proto/configure.ac
+++ b/libxcb/xcb-proto/configure.ac
@@ -3,7 +3,7 @@
 
 AC_PREREQ(2.57)
 AC_INIT([XCB Proto],
-        1.7.1,
+        1.8,
         [xcb@lists.freedesktop.org])
 AC_CONFIG_SRCDIR([xcb-proto.pc.in])
 AM_INIT_AUTOMAKE([foreign dist-bzip2])
diff --git a/mesalib/src/mapi/glapi/gen/gl_gentable.py b/mesalib/src/mapi/glapi/gen/gl_gentable.py
index 5078485c0..72ec569db 100644
--- a/mesalib/src/mapi/glapi/gen/gl_gentable.py
+++ b/mesalib/src/mapi/glapi/gen/gl_gentable.py
@@ -42,7 +42,7 @@ header = """/* GLXEXT is the define used in the xserver when the GLX extension i
 #endif
 
 #if (defined(GLXEXT) && defined(HAVE_BACKTRACE)) \\
-	|| (!defined(GLXEXT) && defined(DEBUG) && !defined(_WIN32_WCE))
+	|| (!defined(GLXEXT) && defined(DEBUG) && !defined(_WIN32_WCE) && !defined(__CYGWIN__))
 #define USE_BACKTRACE
 #endif
 
diff --git a/mesalib/src/mesa/drivers/common/meta.c b/mesalib/src/mesa/drivers/common/meta.c
index d0bb5e0ae..24d8d485a 100644
--- a/mesalib/src/mesa/drivers/common/meta.c
+++ b/mesalib/src/mesa/drivers/common/meta.c
@@ -78,7 +78,6 @@
 #include "main/glformats.h"
 #include "../glsl/ralloc.h"
 
-static void *mem_ctx;
 /** Return offset in bytes of the field within a vertex struct */
 #define OFFSET(FIELD) ((void *) offsetof(struct vertex, FIELD))
 
@@ -3048,7 +3047,6 @@ setup_glsl_generate_mipmap(struct gl_context *ctx,
    };
    struct glsl_sampler *sampler;
    const char *vs_source;
-   const char *fs_template;
 
    static const char *vs_int_source =
       "#version 130\n"
@@ -3071,49 +3069,8 @@ setup_glsl_generate_mipmap(struct gl_context *ctx,
       "   out_color = texture(tex2d, texCoords.xy);\n"
       "}\n";
    char *fs_source;
-   const char *extension_mode;
    GLuint vs, fs;
-
-   if (ctx->Const.GLSLVersion < 130) {
-      vs_source =
-         "attribute vec2 position;\n"
-         "attribute vec3 textureCoords;\n"
-         "varying vec3 texCoords;\n"
-         "void main()\n"
-         "{\n"
-         "   texCoords = textureCoords;\n"
-         "   gl_Position = vec4(position, 0.0, 1.0);\n"
-         "}\n";
-      fs_template =
-         "#extension GL_EXT_texture_array : %s\n"
-         "uniform %s texSampler;\n"
-         "varying vec3 texCoords;\n"
-         "void main()\n"
-         "{\n"
-         "   gl_FragColor = %s(texSampler, %s);\n"
-         "}\n";
-   } else {
-      vs_source =
-         "#version 130\n"
-         "in vec2 position;\n"
-         "in vec3 textureCoords;\n"
-         "out vec3 texCoords;\n"
-         "void main()\n"
-         "{\n"
-         "   texCoords = textureCoords;\n"
-         "   gl_Position = vec4(position, 0.0, 1.0);\n"
-         "}\n";
-      fs_template =
-         "#version 130\n"
-         "uniform %s texSampler;\n"
-         "in vec3 texCoords;\n"
-         "out %s out_color;\n"
-         "\n"
-         "void main()\n"
-         "{\n"
-         "   out_color = texture(texSampler, %s);\n"
-         "}\n";
-    }
+   void *mem_ctx;
 
    /* Check if already initialized */
    if (mipmap->ArrayObj == 0) {
@@ -3143,18 +3100,50 @@ setup_glsl_generate_mipmap(struct gl_context *ctx,
 
    mem_ctx = ralloc_context(NULL);
 
-   if (ctx->Const.GLSLVersion < 130) {
-      extension_mode = ((target == GL_TEXTURE_1D_ARRAY) ||
-                        (target == GL_TEXTURE_2D_ARRAY)) ?
-                       "require" : "disable";
+   if (ctx->API == API_OPENGLES2 || ctx->Const.GLSLVersion < 130) {
+      vs_source =
+         "attribute vec2 position;\n"
+         "attribute vec3 textureCoords;\n"
+         "varying vec3 texCoords;\n"
+         "void main()\n"
+         "{\n"
+         "   texCoords = textureCoords;\n"
+         "   gl_Position = vec4(position, 0.0, 1.0);\n"
+         "}\n";
 
-      fs_source = ralloc_asprintf(mem_ctx, fs_template,
-                                  extension_mode, sampler->type,
+      fs_source = ralloc_asprintf(mem_ctx,
+                                  "#extension GL_EXT_texture_array : enable\n"
+                                  "uniform %s texSampler;\n"
+                                  "varying vec3 texCoords;\n"
+                                  "void main()\n"
+                                  "{\n"
+                                  "   gl_FragColor = %s(texSampler, %s);\n"
+                                  "}\n",
+                                  sampler->type,
                                   sampler->func, sampler->texcoords);
    }
    else {
-      fs_source = ralloc_asprintf(mem_ctx, fs_template,
-                                  sampler->type, "vec4",
+      vs_source =
+         "#version 130\n"
+         "in vec2 position;\n"
+         "in vec3 textureCoords;\n"
+         "out vec3 texCoords;\n"
+         "void main()\n"
+         "{\n"
+         "   texCoords = textureCoords;\n"
+         "   gl_Position = vec4(position, 0.0, 1.0);\n"
+         "}\n";
+      fs_source = ralloc_asprintf(mem_ctx,
+                                  "#version 130\n"
+                                  "uniform %s texSampler;\n"
+                                  "in vec3 texCoords;\n"
+                                  "out vec4 out_color;\n"
+                                  "\n"
+                                  "void main()\n"
+                                  "{\n"
+                                  "   out_color = texture(texSampler, %s);\n"
+                                  "}\n",
+                                  sampler->type,
                                   sampler->texcoords);
    }
 
diff --git a/mesalib/src/mesa/drivers/dri/swrast/swrast.c b/mesalib/src/mesa/drivers/dri/swrast/swrast.c
index dbd339bd8..5bdbe709b 100644
--- a/mesalib/src/mesa/drivers/dri/swrast/swrast.c
+++ b/mesalib/src/mesa/drivers/dri/swrast/swrast.c
@@ -761,7 +761,7 @@ dri_create_context(gl_api api,
     mesaCtx = &ctx->Base;
 
     /* basic context setup */
-    if (!_mesa_initialize_context(mesaCtx, api, visual, sharedCtx, &functions, (void *) cPriv)) {
+    if (!_mesa_initialize_context(mesaCtx, api, visual, sharedCtx, &functions)) {
 	*error = __DRI_CTX_ERROR_NO_MEMORY;
 	goto context_fail;
     }
diff --git a/mesalib/src/mesa/drivers/windows/gdi/wmesa.c b/mesalib/src/mesa/drivers/windows/gdi/wmesa.c
index c61020273..1580e5727 100644
--- a/mesalib/src/mesa/drivers/windows/gdi/wmesa.c
+++ b/mesalib/src/mesa/drivers/windows/gdi/wmesa.c
@@ -1085,7 +1085,7 @@ WMesaContext WMesaCreateContext(HDC hDC,
     /* initialize the Mesa context data */
     ctx = &c->gl_ctx;
     _mesa_initialize_context(ctx, API_OPENGL, visual,
-                             NULL, &functions, (void *)c);
+                             NULL, &functions);
 
     /* visual no longer needed - it was copied by _mesa_initialize_context() */
     _mesa_destroy_visual(visual);
diff --git a/mesalib/src/mesa/main/api_exec.c b/mesalib/src/mesa/main/api_exec.c
index ddf7c7fc1..f42da9438 100644
--- a/mesalib/src/mesa/main/api_exec.c
+++ b/mesalib/src/mesa/main/api_exec.c
@@ -335,8 +335,6 @@ _mesa_create_exec_table(struct gl_context *ctx)
       SET_NormalPointer(exec, _mesa_NormalPointer);
       SET_PrioritizeTextures(exec, _mesa_PrioritizeTextures);
       SET_TexCoordPointer(exec, _mesa_TexCoordPointer);
-   }
-   if (ctx->API != API_OPENGL_CORE) {
       SET_VertexPointer(exec, _mesa_VertexPointer);
    }
 #endif
diff --git a/mesalib/src/mesa/main/context.c b/mesalib/src/mesa/main/context.c
index c50504d19..d3fced946 100644
--- a/mesalib/src/mesa/main/context.c
+++ b/mesalib/src/mesa/main/context.c
@@ -897,20 +897,17 @@ _mesa_alloc_dispatch_table(int size)
  *        etc with, or NULL
  * \param driverFunctions table of device driver functions for this context
  *        to use
- * \param driverContext pointer to driver-specific context data
  */
 GLboolean
 _mesa_initialize_context(struct gl_context *ctx,
                          gl_api api,
                          const struct gl_config *visual,
                          struct gl_context *share_list,
-                         const struct dd_function_table *driverFunctions,
-                         void *driverContext)
+                         const struct dd_function_table *driverFunctions)
 {
    struct gl_shared_state *shared;
    int i;
 
-   /*ASSERT(driverContext);*/
    assert(driverFunctions->NewTextureObject);
    assert(driverFunctions->FreeTextureImageBuffer);
 
@@ -934,7 +931,6 @@ _mesa_initialize_context(struct gl_context *ctx,
     * textures.
     */
    ctx->Driver = *driverFunctions;
-   ctx->DriverCtx = driverContext;
 
    if (share_list) {
       /* share state with another context */
@@ -1049,7 +1045,6 @@ _mesa_initialize_context(struct gl_context *ctx,
  * \param share_list another context to share display lists with or NULL
  * \param driverFunctions points to the dd_function_table into which the
  *        driver has plugged in all its special functions.
- * \param driverContext points to the device driver's private context state
  * 
  * \return pointer to a new __struct gl_contextRec or NULL if error.
  */
@@ -1057,20 +1052,18 @@ struct gl_context *
 _mesa_create_context(gl_api api,
                      const struct gl_config *visual,
                      struct gl_context *share_list,
-                     const struct dd_function_table *driverFunctions,
-                     void *driverContext)
+                     const struct dd_function_table *driverFunctions)
 {
    struct gl_context *ctx;
 
    ASSERT(visual);
-   /*ASSERT(driverContext);*/
 
    ctx = calloc(1, sizeof(struct gl_context));
    if (!ctx)
       return NULL;
 
    if (_mesa_initialize_context(ctx, api, visual, share_list,
-                                driverFunctions, driverContext)) {
+                                driverFunctions)) {
       return ctx;
    }
    else {
diff --git a/mesalib/src/mesa/main/context.h b/mesalib/src/mesa/main/context.h
index f0b4471b1..a5f770a6e 100644
--- a/mesalib/src/mesa/main/context.h
+++ b/mesalib/src/mesa/main/context.h
@@ -109,15 +109,13 @@ _mesa_initialize_context( struct gl_context *ctx,
                           gl_api api,
                           const struct gl_config *visual,
                           struct gl_context *share_list,
-                          const struct dd_function_table *driverFunctions,
-                          void *driverContext );
+                          const struct dd_function_table *driverFunctions);
 
 extern struct gl_context *
 _mesa_create_context(gl_api api,
                      const struct gl_config *visual,
                      struct gl_context *share_list,
-                     const struct dd_function_table *driverFunctions,
-                     void *driverContext);
+                     const struct dd_function_table *driverFunctions);
 
 extern void
 _mesa_free_context_data( struct gl_context *ctx );
diff --git a/mesalib/src/mesa/main/mtypes.h b/mesalib/src/mesa/main/mtypes.h
index e790e18c4..40a802f52 100644
--- a/mesalib/src/mesa/main/mtypes.h
+++ b/mesalib/src/mesa/main/mtypes.h
@@ -3403,8 +3403,6 @@ struct gl_context
     */
    struct dd_function_table Driver;
 
-   void *DriverCtx;	/**< Points to device driver context/state */
-
    /** Core/Driver constants */
    struct gl_constants Const;
 
diff --git a/mesalib/src/mesa/main/state.c b/mesalib/src/mesa/main/state.c
index 76946bd93..fb8b71cfe 100644
--- a/mesalib/src/mesa/main/state.c
+++ b/mesalib/src/mesa/main/state.c
@@ -627,7 +627,8 @@ _mesa_set_varying_vp_inputs( struct gl_context *ctx,
        *
        * It's okay to check the VP pointer here, because this is called after
        * _mesa_update_state in the vbo module. */
-      if (ctx->VertexProgram._TnlProgram) {
+      if (ctx->VertexProgram._TnlProgram ||
+          ctx->FragmentProgram._TexEnvProgram) {
          ctx->NewState |= _NEW_VARYING_VP_INPUTS;
       }
       /*printf("%s %x\n", __FUNCTION__, varying_inputs);*/
diff --git a/mesalib/src/mesa/main/teximage.c b/mesalib/src/mesa/main/teximage.c
index 019516f00..225081578 100644
--- a/mesalib/src/mesa/main/teximage.c
+++ b/mesalib/src/mesa/main/teximage.c
@@ -2871,13 +2871,15 @@ teximage(struct gl_context *ctx, GLboolean compressed, GLuint dims,
                                        border, internalFormat, texFormat);
 
             /* Give the texture to the driver.  <pixels> may be null. */
-            if (compressed) {
-               ctx->Driver.CompressedTexImage(ctx, dims, texImage,
-                                              imageSize, pixels);
-            }
-            else {
-               ctx->Driver.TexImage(ctx, dims, texImage, format,
-                                    type, pixels, unpack);
+            if (width > 0 && height > 0 && depth > 0) {
+               if (compressed) {
+                  ctx->Driver.CompressedTexImage(ctx, dims, texImage,
+                                                 imageSize, pixels);
+               }
+               else {
+                  ctx->Driver.TexImage(ctx, dims, texImage, format,
+                                       type, pixels, unpack);
+               }
             }
 
             check_gen_mipmap(ctx, target, texObj, level);
diff --git a/mesalib/src/mesa/state_tracker/st_context.c b/mesalib/src/mesa/state_tracker/st_context.c
index aa3534125..da361d887 100644
--- a/mesalib/src/mesa/state_tracker/st_context.c
+++ b/mesalib/src/mesa/state_tracker/st_context.c
@@ -225,7 +225,7 @@ struct st_context *st_create_context(gl_api api, struct pipe_context *pipe,
    memset(&funcs, 0, sizeof(funcs));
    st_init_driver_functions(&funcs);
 
-   ctx = _mesa_create_context(api, visual, shareCtx, &funcs, NULL);
+   ctx = _mesa_create_context(api, visual, shareCtx, &funcs);
    if (!ctx) {
       return NULL;
    }
diff --git a/mesalib/src/mesa/x86/gen_matypes.c b/mesalib/src/mesa/x86/gen_matypes.c
index a1d8ee67d..61f181cd9 100644
--- a/mesalib/src/mesa/x86/gen_matypes.c
+++ b/mesalib/src/mesa/x86/gen_matypes.c
@@ -93,7 +93,6 @@ int main( int argc, char **argv )
     */
    OFFSET_HEADER( "struct gl_context" );
 
-   OFFSET( "CTX_DRIVER_CTX              ", struct gl_context, DriverCtx );
    printf( "\n" );
    OFFSET( "CTX_LIGHT_ENABLED           ", struct gl_context, Light.Enabled );
    OFFSET( "CTX_LIGHT_SHADE_MODEL       ", struct gl_context, Light.ShadeModel );
diff --git a/pixman/.gitignore b/pixman/.gitignore
index a4d9f99a7..2d089fc5e 100644
--- a/pixman/.gitignore
+++ b/pixman/.gitignore
@@ -39,10 +39,6 @@ demos/screen-test
 demos/srgb-test
 demos/trap-test
 demos/tri-test
-pixman/pixman-combine32.c
-pixman/pixman-combine32.h
-pixman/pixman-combine64.c
-pixman/pixman-combine64.h
 pixman/pixman-srgb.c
 pixman/pixman-version.h
 test/a1-trap-test
diff --git a/pixman/configure.ac b/pixman/configure.ac
index 5fda54700..c069b483c 100644
--- a/pixman/configure.ac
+++ b/pixman/configure.ac
@@ -182,12 +182,6 @@ AC_SUBST(LT_VERSION_INFO)
 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
diff --git a/pixman/demos/radial-test.c b/pixman/demos/radial-test.c
index 35e90d786..e64f3577f 100644
--- a/pixman/demos/radial-test.c
+++ b/pixman/demos/radial-test.c
@@ -133,7 +133,7 @@ main (int argc, char **argv)
     pixman_image_t *src_img, *dest_img;
     int i, j;
 
-    enable_fp_exceptions ();
+    enable_divbyzero_exceptions ();
 
     dest_img = pixman_image_create_bits (PIXMAN_a8r8g8b8,
 					 WIDTH, HEIGHT,
diff --git a/pixman/pixman/Makefile.am b/pixman/pixman/Makefile.am
index 270d65ec0..cbed0855a 100644
--- a/pixman/pixman/Makefile.am
+++ b/pixman/pixman/Makefile.am
@@ -12,16 +12,10 @@ noinst_LTLIBRARIES =
 
 EXTRA_DIST =				\
 	Makefile.win32			\
-	make-combine.pl			\
-	make-srgb.pl			\
-	pixman-combine.c.template	\
-	pixman-combine.h.template	\
 	pixman-region.c			\
 	solaris-hwcap.mapfile		\
 	$(NULL)
 
-DISTCLEANFILES = $(BUILT_SOURCES)
-
 # mmx code
 if USE_X86_MMX
 noinst_LTLIBRARIES += libpixman-mmx.la
@@ -127,5 +121,5 @@ libpixman_1_la_LDFLAGS += $(LS_LDFLAGS)
 libpixman_1_la_LIBADD += libpixman-loongson-mmi.la
 endif
 
-.c.s : $(libpixmaninclude_HEADERS) $(BUILT_SOURCES)
+.c.s : $(libpixmaninclude_HEADERS)
 	$(CC) $(CFLAGS) $(ASM_CFLAGS_$(@:pixman-%.s=%)) $(ASM_CFLAGS_$(@:pixman-arm-%.s=arm_%)) -DHAVE_CONFIG_H -I$(srcdir) -I$(builddir) -I$(top_builddir) -S -o $@ $<
diff --git a/pixman/pixman/Makefile.sources b/pixman/pixman/Makefile.sources
index cf7040f33..5351fb03d 100644
--- a/pixman/pixman/Makefile.sources
+++ b/pixman/pixman/Makefile.sources
@@ -4,7 +4,7 @@ libpixman_sources =			\
 	pixman-access-accessors.c	\
 	pixman-bits-image.c		\
 	pixman-combine32.c		\
-	pixman-combine64.c		\
+	pixman-combine-float.c		\
 	pixman-conical-gradient.c	\
 	pixman-x86.c			\
 	pixman-mips.c			\
@@ -25,7 +25,6 @@ libpixman_sources =			\
 	pixman-region16.c		\
 	pixman-region32.c		\
 	pixman-solid-fill.c		\
-	pixman-srgb.c			\
 	pixman-timer.c			\
 	pixman-trap.c			\
 	pixman-utils.c			\
@@ -35,30 +34,8 @@ libpixman_headers =			\
 	pixman.h			\
 	pixman-accessor.h		\
 	pixman-combine32.h		\
-	pixman-combine64.h		\
 	pixman-compiler.h		\
 	pixman-edge-imp.h		\
 	pixman-inlines.h		\
 	pixman-private.h		\
 	$(NULL)
-
-BUILT_SOURCES =				\
-	pixman-combine32.c		\
-	pixman-combine32.h		\
-	pixman-combine64.c		\
-	pixman-combine64.h		\
-	pixman-srgb.c			\
-	$(NULL)
-
-pixman-srgb.c: make-srgb.pl
-	$(PERL) $< > $@ || ($(RM) $@; exit 1)
-
-pixman-combine32.c: pixman-combine.c.template make-combine.pl
-	$(PERL) $(lastword $+) 8 < $< > $@ || ($(RM) $@; exit 1)
-pixman-combine32.h: pixman-combine.h.template make-combine.pl
-	$(PERL) $(lastword $+) 8 < $< > $@ || ($(RM) $@; exit 1)
-
-pixman-combine64.c: pixman-combine.c.template make-combine.pl
-	$(PERL) $(lastword $+) 16 < $< > $@ || ($(RM) $@; exit 1)
-pixman-combine64.h: pixman-combine.h.template make-combine.pl
-	$(PERL) $(lastword $+) 16 < $< > $@ || ($(RM) $@; exit 1)
diff --git a/pixman/pixman/make-combine.pl b/pixman/pixman/make-combine.pl
deleted file mode 100644
index 417bdf085..000000000
--- a/pixman/pixman/make-combine.pl
+++ /dev/null
@@ -1,86 +0,0 @@
-$usage = "Usage: combine.pl { 8 | 16 } < pixman-combine.c.template";

-

-$#ARGV == 0 or die $usage;

-

-# Get the component size.

-$size = int($ARGV[0]);

-$size == 8 or $size == 16 or die $usage;

-

-$pixel_size = $size * 4;

-$half_pixel_size = $size * 2;

-

-sub mask {

-    my $str = shift;

-    my $suffix;

-    $suffix = "ULL" if $size > 8;

-

-    return "0x" . $str . $suffix;

-}

-

-# Generate mask strings.

-$nibbles = $size / 4;

-$mask = "f" x $nibbles;

-$zero_mask = "0" x $nibbles;

-$one_half = "8" . "0" x ($nibbles - 1);

-

-print "/* WARNING: This file is generated by combine.pl from combine.inc.\n";

-print "   Please edit one of those files rather than this one. */\n";

-print "\n";

-

-print "#line 1 \"pixman-combine.c.template\"\n";

-

-$mask_ = mask($mask);

-$one_half_ = mask($one_half);

-$g_mask = mask($mask . $zero_mask);

-$b_mask = mask($mask . $zero_mask x 2);

-$a_mask = mask($mask . $zero_mask x 3);

-$rb_mask = mask($mask . $zero_mask . $mask);

-$ag_mask = mask($mask . $zero_mask . $mask . $zero_mask);

-$rb_one_half = mask($one_half . $zero_mask . $one_half);

-$rb_mask_plus_one = mask("1" . $zero_mask x 2 . "1" .  $zero_mask);

-

-while (<STDIN>) {

-    # Mask and 1/2 value for a single component.

-    s/#define COMPONENT_SIZE\b/$& $size/;

-    s/#define MASK\b/$& $mask_/;

-    s/#define ONE_HALF\b/$& $one_half_/;

-

-    # Shifts and masks for green, blue, and alpha.

-    s/#define G_SHIFT\b/$& $size/;

-    s/#define R_SHIFT\b/$& $size * 2/;

-    s/#define A_SHIFT\b/$& $size * 3/;

-    s/#define G_MASK\b/$& $g_mask/;

-    s/#define R_MASK\b/$& $b_mask/;

-    s/#define A_MASK\b/$& $a_mask/;

-

-    # Special values for dealing with red + blue at the same time.

-    s/#define RB_MASK\b/$& $rb_mask/;

-    s/#define AG_MASK\b/$& $ag_mask/;

-    s/#define RB_ONE_HALF\b/$& $rb_one_half/;

-    s/#define RB_MASK_PLUS_ONE\b/$& $rb_mask_plus_one/;

-

-    # Add 32/64 suffix to combining function types.

-    s/\bCombineFunc\b/CombineFunc$pixel_size/;

-    s/\bFbComposeFunctions\b/FbComposeFunctions$pixel_size/;

-    s/combine_width/combine_$pixel_size/;

-    s/_pixman_setup_combiner_functions_width/_pixman_setup_combiner_functions_$pixel_size/;

-    s/UNc/UN$size/g;

-    s/ALPHA_c/ALPHA_$size/g;

-    s/RED_c/RED_$size/g;

-    s/GREEN_c/GREEN_$size/g;

-    s/BLUE_c/BLUE_$size/g;

-

-    # Convert comp*_t values into the appropriate real types.

-    s/comp1_t/uint${size}_t/g;

-    s/comp2_t/uint${half_pixel_size}_t/g;

-    s/comp4_t/uint${pixel_size}_t/g;

-

-    # Change the function table name for the 64-bit version.

-    s/pixman_composeFunctions/pixman_composeFunctions64/ if $size == 16;

-

-    # Change the header for the 64-bit version

-    s/pixman-combine.h/pixman-combine64.h/ if $size == 16;

-    s/pixman-combine.h/pixman-combine32.h/ if $size == 8;

-

-    print;

-}

diff --git a/pixman/pixman/pixman-access.c b/pixman/pixman/pixman-access.c
index 9feafc49d..b5c8e4017 100644
--- a/pixman/pixman/pixman-access.c
+++ b/pixman/pixman/pixman-access.c
@@ -31,6 +31,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include <assert.h>
+#include <math.h>
 
 #include "pixman-accessor.h"
 #include "pixman-private.h"
@@ -506,20 +507,122 @@ MAKE_ACCESSORS(a1);
 MAKE_ACCESSORS(g1);
 
 /********************************** Fetch ************************************/
+/* Table mapping sRGB-encoded 8 bit numbers to linearly encoded
+ * floating point numbers. We assume that single precision
+ * floating point follows the IEEE 754 format.
+ */
+static const uint32_t to_linear_u[256] =
+{
+    0x00000000, 0x399f22b4, 0x3a1f22b4, 0x3a6eb40e, 0x3a9f22b4, 0x3ac6eb61,
+    0x3aeeb40e, 0x3b0b3e5d, 0x3b1f22b4, 0x3b33070b, 0x3b46eb61, 0x3b5b518a,
+    0x3b70f18a, 0x3b83e1c5, 0x3b8fe614, 0x3b9c87fb, 0x3ba9c9b5, 0x3bb7ad6d,
+    0x3bc63547, 0x3bd5635f, 0x3be539bd, 0x3bf5ba70, 0x3c0373b5, 0x3c0c6152,
+    0x3c15a703, 0x3c1f45bc, 0x3c293e68, 0x3c3391f4, 0x3c3e4149, 0x3c494d43,
+    0x3c54b6c7, 0x3c607eb1, 0x3c6ca5df, 0x3c792d22, 0x3c830aa8, 0x3c89af9e,
+    0x3c9085db, 0x3c978dc5, 0x3c9ec7c0, 0x3ca63432, 0x3cadd37d, 0x3cb5a601,
+    0x3cbdac20, 0x3cc5e639, 0x3cce54ab, 0x3cd6f7d2, 0x3cdfd00e, 0x3ce8ddb9,
+    0x3cf2212c, 0x3cfb9ac1, 0x3d02a569, 0x3d0798dc, 0x3d0ca7e4, 0x3d11d2ae,
+    0x3d171963, 0x3d1c7c2e, 0x3d21fb3a, 0x3d2796af, 0x3d2d4ebb, 0x3d332380,
+    0x3d39152b, 0x3d3f23e3, 0x3d454fd0, 0x3d4b991c, 0x3d51ffeb, 0x3d588466,
+    0x3d5f26b7, 0x3d65e6fe, 0x3d6cc564, 0x3d73c210, 0x3d7add25, 0x3d810b65,
+    0x3d84b793, 0x3d88732e, 0x3d8c3e48, 0x3d9018f4, 0x3d940343, 0x3d97fd48,
+    0x3d9c0714, 0x3da020b9, 0x3da44a48, 0x3da883d6, 0x3daccd70, 0x3db12728,
+    0x3db59110, 0x3dba0b38, 0x3dbe95b2, 0x3dc3308f, 0x3dc7dbe0, 0x3dcc97b4,
+    0x3dd1641c, 0x3dd6412a, 0x3ddb2eec, 0x3de02d75, 0x3de53cd3, 0x3dea5d16,
+    0x3def8e52, 0x3df4d091, 0x3dfa23e5, 0x3dff885e, 0x3e027f06, 0x3e05427f,
+    0x3e080ea2, 0x3e0ae376, 0x3e0dc104, 0x3e10a752, 0x3e139669, 0x3e168e50,
+    0x3e198f0e, 0x3e1c98ab, 0x3e1fab2e, 0x3e22c6a0, 0x3e25eb08, 0x3e29186a,
+    0x3e2c4ed0, 0x3e2f8e42, 0x3e32d6c4, 0x3e362861, 0x3e39831e, 0x3e3ce702,
+    0x3e405416, 0x3e43ca5e, 0x3e4749e4, 0x3e4ad2ae, 0x3e4e64c2, 0x3e520027,
+    0x3e55a4e6, 0x3e595303, 0x3e5d0a8a, 0x3e60cb7c, 0x3e6495e0, 0x3e6869bf,
+    0x3e6c4720, 0x3e702e08, 0x3e741e7f, 0x3e78188c, 0x3e7c1c34, 0x3e8014c0,
+    0x3e822039, 0x3e84308b, 0x3e8645b8, 0x3e885fc3, 0x3e8a7eb0, 0x3e8ca281,
+    0x3e8ecb3a, 0x3e90f8df, 0x3e932b72, 0x3e9562f6, 0x3e979f6f, 0x3e99e0e0,
+    0x3e9c274e, 0x3e9e72b8, 0x3ea0c322, 0x3ea31892, 0x3ea57308, 0x3ea7d28a,
+    0x3eaa3718, 0x3eaca0b7, 0x3eaf0f69, 0x3eb18332, 0x3eb3fc16, 0x3eb67a15,
+    0x3eb8fd34, 0x3ebb8576, 0x3ebe12de, 0x3ec0a56e, 0x3ec33d2a, 0x3ec5da14,
+    0x3ec87c30, 0x3ecb2380, 0x3ecdd008, 0x3ed081ca, 0x3ed338c9, 0x3ed5f508,
+    0x3ed8b68a, 0x3edb7d52, 0x3ede4962, 0x3ee11abe, 0x3ee3f168, 0x3ee6cd64,
+    0x3ee9aeb6, 0x3eec955d, 0x3eef815d, 0x3ef272ba, 0x3ef56976, 0x3ef86594,
+    0x3efb6717, 0x3efe6e02, 0x3f00bd2b, 0x3f02460c, 0x3f03d1a5, 0x3f055ff8,
+    0x3f06f105, 0x3f0884ce, 0x3f0a1b54, 0x3f0bb499, 0x3f0d509f, 0x3f0eef65,
+    0x3f1090ef, 0x3f12353c, 0x3f13dc50, 0x3f15862a, 0x3f1732cc, 0x3f18e237,
+    0x3f1a946d, 0x3f1c4970, 0x3f1e013f, 0x3f1fbbde, 0x3f21794c, 0x3f23398c,
+    0x3f24fca0, 0x3f26c286, 0x3f288b42, 0x3f2a56d3, 0x3f2c253d, 0x3f2df680,
+    0x3f2fca9d, 0x3f31a195, 0x3f337b6a, 0x3f35581e, 0x3f3737b1, 0x3f391a24,
+    0x3f3aff7a, 0x3f3ce7b2, 0x3f3ed2d0, 0x3f40c0d2, 0x3f42b1bc, 0x3f44a58e,
+    0x3f469c49, 0x3f4895ee, 0x3f4a9280, 0x3f4c91ff, 0x3f4e946c, 0x3f5099c8,
+    0x3f52a216, 0x3f54ad55, 0x3f56bb88, 0x3f58ccae, 0x3f5ae0cb, 0x3f5cf7de,
+    0x3f5f11ec, 0x3f612ef0, 0x3f634eef, 0x3f6571ea, 0x3f6797e1, 0x3f69c0d6,
+    0x3f6beccb, 0x3f6e1bc0, 0x3f704db6, 0x3f7282af, 0x3f74baac, 0x3f76f5ae,
+    0x3f7933b6, 0x3f7b74c6, 0x3f7db8de, 0x3f800000
+};
+
+static const float * const to_linear = (const float *)to_linear_u;
+
+static uint8_t
+to_srgb (float f)
+{
+    uint8_t low = 0;
+    uint8_t high = 255;
+
+    while (high - low > 1)
+    {
+	uint8_t mid = (low + high) / 2;
+
+	if (to_linear[mid] > f)
+	    high = mid;
+	else
+	    low = mid;
+    }
+
+    if (to_linear[high] - f < f - to_linear[low])
+	return high;
+    else
+	return low;
+}
 
-/* Expects a uint64_t buffer */
 static void
-fetch_scanline_a2r10g10b10 (pixman_image_t *image,
-                            int             x,
-                            int             y,
-                            int             width,
-                            uint32_t *      b,
-                            const uint32_t *mask)
+fetch_scanline_a8r8g8b8_sRGB_float (pixman_image_t *image,
+				    int             x,
+				    int             y,
+				    int             width,
+				    uint32_t *      b,
+				    const uint32_t *mask)
 {
     const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
     const uint32_t *pixel = bits + x;
     const uint32_t *end = pixel + width;
-    uint64_t *buffer = (uint64_t *)b;
+    argb_t *buffer = (argb_t *)b;
+
+    while (pixel < end)
+    {
+	uint32_t p = READ (image, pixel++);
+	argb_t *argb = buffer;
+
+	argb->a = pixman_unorm_to_float ((p >> 24) & 0xff, 8);
+
+	argb->r = to_linear [(p >> 16) & 0xff];
+	argb->g = to_linear [(p >>  8) & 0xff];
+	argb->b = to_linear [(p >>  0) & 0xff];
+
+	buffer++;
+    }
+}
+
+/* Expects a float buffer */
+static void
+fetch_scanline_a2r10g10b10_float (pixman_image_t *image,
+				  int             x,
+				  int             y,
+				  int             width,
+				  uint32_t *      b,
+				  const uint32_t *mask)
+{
+    const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
+    const uint32_t *pixel = bits + x;
+    const uint32_t *end = pixel + width;
+    argb_t *buffer = (argb_t *)b;
 
     while (pixel < end)
     {
@@ -529,62 +632,59 @@ fetch_scanline_a2r10g10b10 (pixman_image_t *image,
 	uint64_t g = (p >> 10) & 0x3ff;
 	uint64_t b = p & 0x3ff;
 
-	r = r << 6 | r >> 4;
-	g = g << 6 | g >> 4;
-	b = b << 6 | b >> 4;
-
-	a <<= 14;
-	a |= a >> 2;
-	a |= a >> 4;
-	a |= a >> 8;
+	buffer->a = pixman_unorm_to_float (a, 2);
+	buffer->r = pixman_unorm_to_float (r, 10);
+	buffer->g = pixman_unorm_to_float (g, 10);
+	buffer->b = pixman_unorm_to_float (b, 10);
 
-	*buffer++ = a << 48 | r << 32 | g << 16 | b;
+	buffer++;
     }
 }
 
-/* Expects a uint64_t buffer */
+/* Expects a float buffer */
 static void
-fetch_scanline_x2r10g10b10 (pixman_image_t *image,
-                            int             x,
-                            int             y,
-                            int             width,
-                            uint32_t *      b,
-                            const uint32_t *mask)
+fetch_scanline_x2r10g10b10_float (pixman_image_t *image,
+				  int             x,
+				  int             y,
+				  int             width,
+				  uint32_t *      b,
+				  const uint32_t *mask)
 {
     const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
     const uint32_t *pixel = (uint32_t *)bits + x;
     const uint32_t *end = pixel + width;
-    uint64_t *buffer = (uint64_t *)b;
-    
+    argb_t *buffer = (argb_t *)b;
+
     while (pixel < end)
     {
 	uint32_t p = READ (image, pixel++);
 	uint64_t r = (p >> 20) & 0x3ff;
 	uint64_t g = (p >> 10) & 0x3ff;
 	uint64_t b = p & 0x3ff;
-	
-	r = r << 6 | r >> 4;
-	g = g << 6 | g >> 4;
-	b = b << 6 | b >> 4;
-	
-	*buffer++ = 0xffffULL << 48 | r << 32 | g << 16 | b;
+
+	buffer->a = 1.0;
+	buffer->r = pixman_unorm_to_float (r, 10);
+	buffer->g = pixman_unorm_to_float (g, 10);
+	buffer->b = pixman_unorm_to_float (b, 10);
+
+	buffer++;
     }
 }
 
-/* Expects a uint64_t buffer */
+/* Expects a float buffer */
 static void
-fetch_scanline_a2b10g10r10 (pixman_image_t *image,
-                            int             x,
-                            int             y,
-                            int             width,
-                            uint32_t *      b,
-                            const uint32_t *mask)
+fetch_scanline_a2b10g10r10_float (pixman_image_t *image,
+				  int             x,
+				  int             y,
+				  int             width,
+				  uint32_t *      b,
+				  const uint32_t *mask)
 {
     const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
     const uint32_t *pixel = bits + x;
     const uint32_t *end = pixel + width;
-    uint64_t *buffer = (uint64_t *)b;
-    
+    argb_t *buffer = (argb_t *)b;
+
     while (pixel < end)
     {
 	uint32_t p = READ (image, pixel++);
@@ -592,46 +692,43 @@ fetch_scanline_a2b10g10r10 (pixman_image_t *image,
 	uint64_t b = (p >> 20) & 0x3ff;
 	uint64_t g = (p >> 10) & 0x3ff;
 	uint64_t r = p & 0x3ff;
-	
-	r = r << 6 | r >> 4;
-	g = g << 6 | g >> 4;
-	b = b << 6 | b >> 4;
-	
-	a <<= 14;
-	a |= a >> 2;
-	a |= a >> 4;
-	a |= a >> 8;
 
-	*buffer++ = a << 48 | r << 32 | g << 16 | b;
+	buffer->a = pixman_unorm_to_float (a, 2);
+	buffer->r = pixman_unorm_to_float (r, 10);
+	buffer->g = pixman_unorm_to_float (g, 10);
+	buffer->b = pixman_unorm_to_float (b, 10);
+
+	buffer++;
     }
 }
 
-/* Expects a uint64_t buffer */
+/* Expects a float buffer */
 static void
-fetch_scanline_x2b10g10r10 (pixman_image_t *image,
-                            int             x,
-                            int             y,
-                            int             width,
-                            uint32_t *      b,
-                            const uint32_t *mask)
+fetch_scanline_x2b10g10r10_float (pixman_image_t *image,
+				  int             x,
+				  int             y,
+				  int             width,
+				  uint32_t *      b,
+				  const uint32_t *mask)
 {
     const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
     const uint32_t *pixel = (uint32_t *)bits + x;
     const uint32_t *end = pixel + width;
-    uint64_t *buffer = (uint64_t *)b;
-    
+    argb_t *buffer = (argb_t *)b;
+
     while (pixel < end)
     {
 	uint32_t p = READ (image, pixel++);
 	uint64_t b = (p >> 20) & 0x3ff;
 	uint64_t g = (p >> 10) & 0x3ff;
 	uint64_t r = p & 0x3ff;
-	
-	r = r << 6 | r >> 4;
-	g = g << 6 | g >> 4;
-	b = b << 6 | b >> 4;
-	
-	*buffer++ = 0xffffULL << 48 | r << 32 | g << 16 | b;
+
+	buffer->a = 1.0;
+	buffer->r = pixman_unorm_to_float (r, 10);
+	buffer->g = pixman_unorm_to_float (g, 10);
+	buffer->b = pixman_unorm_to_float (b, 10);
+
+	buffer++;
     }
 }
 
@@ -708,55 +805,51 @@ fetch_scanline_yv12 (pixman_image_t *image,
 
 /**************************** Pixel wise fetching *****************************/
 
-/* Despite the type, expects a uint64_t buffer */
-static uint64_t
-fetch_pixel_a2r10g10b10 (bits_image_t *image,
-			 int		  offset,
-			 int           line)
+static argb_t
+fetch_pixel_x2r10g10b10_float (bits_image_t *image,
+			       int	   offset,
+			       int           line)
 {
     uint32_t *bits = image->bits + line * image->rowstride;
     uint32_t p = READ (image, bits + offset);
-    uint64_t a = p >> 30;
     uint64_t r = (p >> 20) & 0x3ff;
     uint64_t g = (p >> 10) & 0x3ff;
     uint64_t b = p & 0x3ff;
+    argb_t argb;
 
-    r = r << 6 | r >> 4;
-    g = g << 6 | g >> 4;
-    b = b << 6 | b >> 4;
+    argb.a = 1.0;
+    argb.r = pixman_unorm_to_float (r, 10);
+    argb.g = pixman_unorm_to_float (g, 10);
+    argb.b = pixman_unorm_to_float (b, 10);
 
-    a <<= 14;
-    a |= a >> 2;
-    a |= a >> 4;
-    a |= a >> 8;
-
-    return a << 48 | r << 32 | g << 16 | b;
+    return argb;
 }
 
-/* Despite the type, this function expects a uint64_t buffer */
-static uint64_t
-fetch_pixel_x2r10g10b10 (bits_image_t *image,
-			 int	   offset,
-			 int           line)
+static argb_t
+fetch_pixel_a2r10g10b10_float (bits_image_t *image,
+			       int	     offset,
+			       int           line)
 {
     uint32_t *bits = image->bits + line * image->rowstride;
     uint32_t p = READ (image, bits + offset);
+    uint64_t a = p >> 30;
     uint64_t r = (p >> 20) & 0x3ff;
     uint64_t g = (p >> 10) & 0x3ff;
     uint64_t b = p & 0x3ff;
-    
-    r = r << 6 | r >> 4;
-    g = g << 6 | g >> 4;
-    b = b << 6 | b >> 4;
-    
-    return 0xffffULL << 48 | r << 32 | g << 16 | b;
+    argb_t argb;
+
+    argb.a = pixman_unorm_to_float (a, 2);
+    argb.r = pixman_unorm_to_float (r, 10);
+    argb.g = pixman_unorm_to_float (g, 10);
+    argb.b = pixman_unorm_to_float (b, 10);
+
+    return argb;
 }
 
-/* Despite the type, expects a uint64_t buffer */
-static uint64_t
-fetch_pixel_a2b10g10r10 (bits_image_t *image,
-			 int           offset,
-			 int           line)
+static argb_t
+fetch_pixel_a2b10g10r10_float (bits_image_t *image,
+			       int           offset,
+			       int           line)
 {
     uint32_t *bits = image->bits + line * image->rowstride;
     uint32_t p = READ (image, bits + offset);
@@ -764,36 +857,52 @@ fetch_pixel_a2b10g10r10 (bits_image_t *image,
     uint64_t b = (p >> 20) & 0x3ff;
     uint64_t g = (p >> 10) & 0x3ff;
     uint64_t r = p & 0x3ff;
-    
-    r = r << 6 | r >> 4;
-    g = g << 6 | g >> 4;
-    b = b << 6 | b >> 4;
-    
-    a <<= 14;
-    a |= a >> 2;
-    a |= a >> 4;
-    a |= a >> 8;
-    
-    return a << 48 | r << 32 | g << 16 | b;
+    argb_t argb;
+
+    argb.a = pixman_unorm_to_float (a, 2);
+    argb.r = pixman_unorm_to_float (r, 10);
+    argb.g = pixman_unorm_to_float (g, 10);
+    argb.b = pixman_unorm_to_float (b, 10);
+
+    return argb;
 }
 
-/* Despite the type, this function expects a uint64_t buffer */
-static uint64_t
-fetch_pixel_x2b10g10r10 (bits_image_t *image,
-			 int           offset,
-			 int           line)
+static argb_t
+fetch_pixel_x2b10g10r10_float (bits_image_t *image,
+			       int           offset,
+			       int           line)
 {
     uint32_t *bits = image->bits + line * image->rowstride;
     uint32_t p = READ (image, bits + offset);
     uint64_t b = (p >> 20) & 0x3ff;
     uint64_t g = (p >> 10) & 0x3ff;
     uint64_t r = p & 0x3ff;
-    
-    r = r << 6 | r >> 4;
-    g = g << 6 | g >> 4;
-    b = b << 6 | b >> 4;
-    
-    return 0xffffULL << 48 | r << 32 | g << 16 | b;
+    argb_t argb;
+
+    argb.a = 1.0;
+    argb.r = pixman_unorm_to_float (r, 10);
+    argb.g = pixman_unorm_to_float (g, 10);
+    argb.b = pixman_unorm_to_float (b, 10);
+
+    return argb;
+}
+
+static argb_t
+fetch_pixel_a8r8g8b8_sRGB_float (bits_image_t *image,
+				 int	       offset,
+				 int           line)
+{
+    uint32_t *bits = image->bits + line * image->rowstride;
+    uint32_t p = READ (image, bits + offset);
+    argb_t argb;
+
+    argb.a = pixman_unorm_to_float ((p >> 24) & 0xff, 8);
+
+    argb.r = to_linear [(p >> 16) & 0xff];
+    argb.g = to_linear [(p >>  8) & 0xff];
+    argb.b = to_linear [(p >>  0) & 0xff];
+
+    return argb;
 }
 
 static uint32_t
@@ -854,178 +963,174 @@ fetch_pixel_yv12 (bits_image_t *image,
 /*********************************** Store ************************************/
 
 static void
-store_scanline_a2r10g10b10 (bits_image_t *  image,
-                            int             x,
-                            int             y,
-                            int             width,
-                            const uint32_t *v)
+store_scanline_a2r10g10b10_float (bits_image_t *  image,
+				  int             x,
+				  int             y,
+				  int             width,
+				  const uint32_t *v)
 {
     uint32_t *bits = image->bits + image->rowstride * y;
     uint32_t *pixel = bits + x;
-    uint64_t *values = (uint64_t *)v;
+    argb_t *values = (argb_t *)v;
     int i;
-    
+
     for (i = 0; i < width; ++i)
     {
+	uint16_t a, r, g, b;
+
+	a = pixman_float_to_unorm (values[i].a, 2);
+	r = pixman_float_to_unorm (values[i].r, 10);
+	g = pixman_float_to_unorm (values[i].g, 10);
+	b = pixman_float_to_unorm (values[i].b, 10);
+
 	WRITE (image, pixel++,
-	       ((values[i] >> 32) & 0xc0000000) |
-	       ((values[i] >> 18) & 0x3ff00000) |
-	       ((values[i] >> 12) & 0xffc00) | 
-	       ((values[i] >> 6) & 0x3ff));    
+	       (a << 30) | (r << 20) | (g << 10) | b);
     }
 }
 
 static void
-store_scanline_x2r10g10b10 (bits_image_t *  image,
-                            int             x,
-                            int             y,
-                            int             width,
-                            const uint32_t *v)
+store_scanline_x2r10g10b10_float (bits_image_t *  image,
+				  int             x,
+				  int             y,
+				  int             width,
+				  const uint32_t *v)
 {
     uint32_t *bits = image->bits + image->rowstride * y;
-    uint64_t *values = (uint64_t *)v;
     uint32_t *pixel = bits + x;
+    argb_t *values = (argb_t *)v;
     int i;
-    
+
     for (i = 0; i < width; ++i)
     {
+	uint16_t r, g, b;
+
+	r = pixman_float_to_unorm (values[i].r, 10);
+	g = pixman_float_to_unorm (values[i].g, 10);
+	b = pixman_float_to_unorm (values[i].b, 10);
+
 	WRITE (image, pixel++,
-	       ((values[i] >> 18) & 0x3ff00000) | 
-	       ((values[i] >> 12) & 0xffc00) |
-	       ((values[i] >> 6) & 0x3ff));
+	       (r << 20) | (g << 10) | b);
     }
 }
 
 static void
-store_scanline_a2b10g10r10 (bits_image_t *  image,
-                            int             x,
-                            int             y,
-                            int             width,
-                            const uint32_t *v)
+store_scanline_a2b10g10r10_float (bits_image_t *  image,
+				  int             x,
+				  int             y,
+				  int             width,
+				  const uint32_t *v)
 {
     uint32_t *bits = image->bits + image->rowstride * y;
     uint32_t *pixel = bits + x;
-    uint64_t *values = (uint64_t *)v;
+    argb_t *values = (argb_t *)v;
     int i;
-    
+
     for (i = 0; i < width; ++i)
     {
+	uint16_t a, r, g, b;
+
+	a = pixman_float_to_unorm (values[i].a, 2);
+	r = pixman_float_to_unorm (values[i].r, 10);
+	g = pixman_float_to_unorm (values[i].g, 10);
+	b = pixman_float_to_unorm (values[i].b, 10);
+
 	WRITE (image, pixel++,
-	       ((values[i] >> 32) & 0xc0000000) |
-	       ((values[i] >> 38) & 0x3ff) |
-	       ((values[i] >> 12) & 0xffc00) |
-	       ((values[i] << 14) & 0x3ff00000));
+	       (a << 30) | (b << 20) | (g << 10) | r);
     }
 }
 
 static void
-store_scanline_x2b10g10r10 (bits_image_t *  image,
-                            int             x,
-                            int             y,
-                            int             width,
-                            const uint32_t *v)
+store_scanline_x2b10g10r10_float (bits_image_t *  image,
+				  int             x,
+				  int             y,
+				  int             width,
+				  const uint32_t *v)
 {
     uint32_t *bits = image->bits + image->rowstride * y;
-    uint64_t *values = (uint64_t *)v;
     uint32_t *pixel = bits + x;
+    argb_t *values = (argb_t *)v;
     int i;
-    
+
+    for (i = 0; i < width; ++i)
+    {
+	uint16_t r, g, b;
+
+	r = pixman_float_to_unorm (values[i].r, 10);
+	g = pixman_float_to_unorm (values[i].g, 10);
+	b = pixman_float_to_unorm (values[i].b, 10);
+
+	WRITE (image, pixel++,
+	       (b << 20) | (g << 10) | r);
+    }
+}
+
+static void
+store_scanline_a8r8g8b8_sRGB_float (bits_image_t *  image,
+				    int             x,
+				    int             y,
+				    int             width,
+				    const uint32_t *v)
+{
+    uint32_t *bits = image->bits + image->rowstride * y;
+    uint32_t *pixel = bits + x;
+    argb_t *values = (argb_t *)v;
+    int i;
+
     for (i = 0; i < width; ++i)
     {
+	uint8_t a, r, g, b;
+
+	a = pixman_float_to_unorm (values[i].a, 8);
+	r = to_srgb (values[i].r);
+	g = to_srgb (values[i].g);
+	b = to_srgb (values[i].b);
+
 	WRITE (image, pixel++,
-	       ((values[i] >> 38) & 0x3ff) |
-	       ((values[i] >> 12) & 0xffc00) |
-	       ((values[i] << 14) & 0x3ff00000));
+	       (a << 24) | (r << 16) | (g << 8) | b);
     }
 }
 
 /*
- * Contracts a 64bpp image to 32bpp and then stores it using a regular 32-bit
- * store proc. Despite the type, this function expects a uint64_t buffer.
+ * Contracts a floating point image to 32bpp and then stores it using a
+ * regular 32-bit store proc. Despite the type, this function expects an
+ * argb_t buffer.
  */
 static void
-store_scanline_generic_64 (bits_image_t *  image,
-                           int             x,
-                           int             y,
-                           int             width,
-                           const uint32_t *values)
+store_scanline_generic_float (bits_image_t *  image,
+			      int             x,
+			      int             y,
+			      int             width,
+			      const uint32_t *values)
 {
     uint32_t *argb8_pixels;
-    
+
     assert (image->common.type == BITS);
-    
+
     argb8_pixels = pixman_malloc_ab (width, sizeof(uint32_t));
     if (!argb8_pixels)
 	return;
-    
+
     /* Contract the scanline.  We could do this in place if values weren't
      * const.
      */
-    pixman_contract (argb8_pixels, (uint64_t *)values, width);
-    
+    pixman_contract_from_float (argb8_pixels, (argb_t *)values, width);
+
     image->store_scanline_32 (image, x, y, width, argb8_pixels);
-    
+
     free (argb8_pixels);
 }
 
-/* Despite the type, this function expects both buffer
- * and mask to be uint64_t
- */
 static void
-fetch_scanline_generic_64 (pixman_image_t *image,
-                           int             x,
-                           int             y,
-                           int             width,
-                           uint32_t *      buffer,
-                           const uint32_t *mask)
+fetch_scanline_generic_float (pixman_image_t *image,
+			      int	      x,
+			      int	      y,
+			      int	      width,
+			      uint32_t *      buffer,
+			      const uint32_t *mask)
 {
-    pixman_format_code_t format;
-
-    /* Fetch the pixels into the first half of buffer and then expand them in
-     * place.
-     */
     image->bits.fetch_scanline_32 (image, x, y, width, buffer, NULL);
 
-    format = image->bits.format;
-    if (PIXMAN_FORMAT_TYPE (format) == PIXMAN_TYPE_COLOR	||
-	PIXMAN_FORMAT_TYPE (format) == PIXMAN_TYPE_GRAY)
-    {
-	/* Indexed formats are mapped to a8r8g8b8 with full
-	 * precision, so when expanding we shouldn't correct
-	 * for the width of the channels
-	 */
-
-	format = PIXMAN_a8r8g8b8;
-    }
-
-    pixman_expand ((uint64_t *)buffer, buffer, format, width);
-}
-
-/* Despite the type, this function expects a uint64_t *buffer */
-static uint64_t
-fetch_pixel_generic_64 (bits_image_t *image,
-			int	      offset,
-			int           line)
-{
-    uint32_t pixel32 = image->fetch_pixel_32 (image, offset, line);
-    uint64_t result;
-    pixman_format_code_t format;
-
-    format = image->format;
-    if (PIXMAN_FORMAT_TYPE (format) == PIXMAN_TYPE_COLOR	||
-	PIXMAN_FORMAT_TYPE (format) == PIXMAN_TYPE_GRAY)
-    {
-	/* Indexed formats are mapped to a8r8g8b8 with full
-	 * precision, so when expanding we shouldn't correct
-	 * for the width of the channels
-	 */
-
-	format = PIXMAN_a8r8g8b8;
-    }
-
-    pixman_expand ((uint64_t *)&result, &pixel32, format, 1);
-
-    return result;
+    pixman_expand_to_float ((argb_t *)buffer, buffer, image->bits.format, width);
 }
 
 /* The 32_sRGB paths should be deleted after narrow processing
@@ -1046,35 +1151,20 @@ fetch_scanline_a8r8g8b8_32_sRGB (pixman_image_t *image,
     
     while (pixel < end)
     {
-	tmp = READ (image, pixel++);
-	*buffer++ =                 (tmp >> 24)               << 24
-		  | (srgb_to_linear[(tmp >> 16) & 0xff] >> 8) << 16
-		  | (srgb_to_linear[(tmp >>  8) & 0xff] >> 8) <<  8
-		  | (srgb_to_linear[(tmp >>  0) & 0xff] >> 8) <<  0;
-    }
-}
+	uint8_t a, r, g, b;
 
-static void
-fetch_scanline_a8r8g8b8_64_sRGB (pixman_image_t *image,
-                                 int             x,
-                                 int             y,
-                                 int             width,
-                                 uint32_t       *b,
-                                 const uint32_t *mask)
-{
-    const uint32_t *bits = image->bits.bits + y * image->bits.rowstride;
-    const uint32_t *pixel = (uint32_t *)bits + x;
-    const uint32_t *end = pixel + width;
-    uint64_t *buffer = (uint64_t *)b;
-    uint32_t tmp;
-    
-    while (pixel < end)
-    {
 	tmp = READ (image, pixel++);
-	*buffer++ = (uint64_t)               ((tmp >> 24) * 257)  << 48
-		  | (uint64_t) srgb_to_linear[(tmp >> 16) & 0xff] << 32
-		  | (uint64_t) srgb_to_linear[(tmp >>  8) & 0xff] << 16
-		  | (uint64_t) srgb_to_linear[(tmp >>  0) & 0xff] <<  0;
+
+	a = (tmp >> 24) & 0xff;
+	r = (tmp >> 16) & 0xff;
+	g = (tmp >> 8) & 0xff;
+	b = (tmp >> 0) & 0xff;
+
+	r = to_linear[r] * 255.0f + 0.5f;
+	g = to_linear[g] * 255.0f + 0.5f;
+	b = to_linear[b] * 255.0f + 0.5f;
+
+	*buffer++ = (a << 24) | (r << 16) | (g << 8) | (b << 0);
     }
 }
 
@@ -1085,23 +1175,18 @@ fetch_pixel_a8r8g8b8_32_sRGB (bits_image_t *image,
 {
     uint32_t *bits = image->bits + line * image->rowstride;
     uint32_t tmp = READ (image, bits + offset);
-    return                 (tmp >> 24)               << 24
-	 | (srgb_to_linear[(tmp >> 16) & 0xff] >> 8) << 16
-	 | (srgb_to_linear[(tmp >>  8) & 0xff] >> 8) <<  8
-	 | (srgb_to_linear[(tmp >>  0) & 0xff] >> 8) <<  0;
-}
+    uint8_t a, r, g, b;
 
-static uint64_t
-fetch_pixel_a8r8g8b8_64_sRGB (bits_image_t *image,
-			      int           offset,
-			      int	    line)
-{
-    uint32_t *bits = image->bits + line * image->rowstride;
-    uint32_t tmp = READ (image, bits + offset);
-    return (uint64_t)               ((tmp >> 24) * 257)  << 48
-	 | (uint64_t) srgb_to_linear[(tmp >> 16) & 0xff] << 32
-	 | (uint64_t) srgb_to_linear[(tmp >>  8) & 0xff] << 16
-	 | (uint64_t) srgb_to_linear[(tmp >>  0) & 0xff] <<  0;
+    a = (tmp >> 24) & 0xff;
+    r = (tmp >> 16) & 0xff;
+    g = (tmp >> 8) & 0xff;
+    b = (tmp >> 0) & 0xff;
+
+    r = to_linear[r] * 255.0f + 0.5f;
+    g = to_linear[g] * 255.0f + 0.5f;
+    b = to_linear[b] * 255.0f + 0.5f;
+
+    return (a << 24) | (r << 16) | (g << 8) | (b << 0);
 }
 
 static void
@@ -1119,37 +1204,34 @@ store_scanline_a8r8g8b8_32_sRGB (bits_image_t   *image,
     
     for (i = 0; i < width; ++i)
     {
+	uint8_t a, r, g, b;
+
 	tmp = values[i];
-	WRITE (image, pixel++,
-		  ((uint32_t)     (tmp >> 24     )          << 24)
-		| (linear_to_srgb[(tmp >> 16 << 4) & 0xfff] << 16)
-		| (linear_to_srgb[(tmp >>  8 << 4) & 0xfff] <<  8)
-		| (linear_to_srgb[(tmp >>  0 << 4) & 0xfff] <<  0));
+
+	a = (tmp >> 24) & 0xff;
+	r = (tmp >> 16) & 0xff;
+	g = (tmp >> 8) & 0xff;
+	b = (tmp >> 0) & 0xff;
+
+	r = to_srgb (r * (1/255.0f));
+	g = to_srgb (g * (1/255.0f));
+	b = to_srgb (b * (1/255.0f));
+	
+	WRITE (image, pixel++, a | (r << 16) | (g << 8) | (b << 0));
     }
 }
 
-static void
-store_scanline_a8r8g8b8_64_sRGB (bits_image_t  *image,
-                                int             x,
-                                int             y,
-                                int             width,
-                                const uint32_t *v)
+static argb_t
+fetch_pixel_generic_float (bits_image_t *image,
+			   int		 offset,
+			   int           line)
 {
-    uint32_t *bits = image->bits + image->rowstride * y;
-    uint64_t *values = (uint64_t *)v;
-    uint32_t *pixel = bits + x;
-    uint64_t tmp;
-    int i;
-    
-    for (i = 0; i < width; ++i)
-    {
-	tmp = values[i];
-	WRITE (image, pixel++,
-		  ((uint32_t)     (tmp >> 56)          << 24)
-		| (linear_to_srgb[(tmp >> 36) & 0xfff] << 16)
-		| (linear_to_srgb[(tmp >> 20) & 0xfff] <<  8)
-		| (linear_to_srgb[(tmp >>  4) & 0xfff] <<  0));
-    }
+    uint32_t pixel32 = image->fetch_pixel_32 (image, offset, line);
+    argb_t f;
+
+    pixman_expand_to_float (&f, &pixel32, image->format, 1);
+
+    return f;
 }
 
 /*
@@ -1163,10 +1245,10 @@ fetch_pixel_generic_lossy_32 (bits_image_t *image,
 			      int           offset,
 			      int           line)
 {
-    uint64_t pixel64 = image->fetch_pixel_64 (image, offset, line);
+    argb_t pixel64 = image->fetch_pixel_float (image, offset, line);
     uint32_t result;
 
-    pixman_contract (&result, &pixel64, 1);
+    pixman_contract_from_float (&result, &pixel64, 1);
 
     return result;
 }
@@ -1175,20 +1257,22 @@ typedef struct
 {
     pixman_format_code_t	format;
     fetch_scanline_t		fetch_scanline_32;
-    fetch_scanline_t		fetch_scanline_64;
+    fetch_scanline_t		fetch_scanline_float;
     fetch_pixel_32_t		fetch_pixel_32;
-    fetch_pixel_64_t		fetch_pixel_64;
+    fetch_pixel_float_t		fetch_pixel_float;
     store_scanline_t		store_scanline_32;
-    store_scanline_t		store_scanline_64;
+    store_scanline_t		store_scanline_float;
 } format_info_t;
 
 #define FORMAT_INFO(format) 						\
     {									\
 	PIXMAN_ ## format,						\
 	    fetch_scanline_ ## format,					\
-	    fetch_scanline_generic_64,					\
-	    fetch_pixel_ ## format, fetch_pixel_generic_64,		\
-	    store_scanline_ ## format, store_scanline_generic_64	\
+	    fetch_scanline_generic_float,				\
+	    fetch_pixel_ ## format,					\
+	    fetch_pixel_generic_float,					\
+	    store_scanline_ ## format,					\
+	    store_scanline_generic_float				\
     }
 
 static const format_info_t accessors[] =
@@ -1206,10 +1290,10 @@ static const format_info_t accessors[] =
 
 /* sRGB formats */
   { PIXMAN_a8r8g8b8_sRGB,
-    fetch_scanline_a8r8g8b8_32_sRGB,
-    fetch_scanline_a8r8g8b8_64_sRGB,
-    fetch_pixel_a8r8g8b8_32_sRGB, fetch_pixel_a8r8g8b8_64_sRGB,
-    store_scanline_a8r8g8b8_32_sRGB, store_scanline_a8r8g8b8_64_sRGB },
+    fetch_scanline_a8r8g8b8_32_sRGB, fetch_scanline_a8r8g8b8_sRGB_float,
+    fetch_pixel_a8r8g8b8_32_sRGB, fetch_pixel_a8r8g8b8_sRGB_float,
+    store_scanline_a8r8g8b8_32_sRGB, store_scanline_a8r8g8b8_sRGB_float,
+  },
 
 /* 24bpp formats */
     FORMAT_INFO (r8g8b8),
@@ -1269,34 +1353,34 @@ static const format_info_t accessors[] =
 /* Wide formats */
     
     { PIXMAN_a2r10g10b10,
-      NULL, fetch_scanline_a2r10g10b10,
-      fetch_pixel_generic_lossy_32, fetch_pixel_a2r10g10b10,
-      NULL, store_scanline_a2r10g10b10 },
-    
+      NULL, fetch_scanline_a2r10g10b10_float,
+      fetch_pixel_generic_lossy_32, fetch_pixel_a2r10g10b10_float,
+      NULL, store_scanline_a2r10g10b10_float },
+
     { PIXMAN_x2r10g10b10,
-      NULL, fetch_scanline_x2r10g10b10,
-      fetch_pixel_generic_lossy_32, fetch_pixel_x2r10g10b10,
-      NULL, store_scanline_x2r10g10b10 },
-    
+      NULL, fetch_scanline_x2r10g10b10_float,
+      fetch_pixel_generic_lossy_32, fetch_pixel_x2r10g10b10_float,
+      NULL, store_scanline_x2r10g10b10_float },
+
     { PIXMAN_a2b10g10r10,
-      NULL, fetch_scanline_a2b10g10r10,
-      fetch_pixel_generic_lossy_32, fetch_pixel_a2b10g10r10,
-      NULL, store_scanline_a2b10g10r10 },
-    
+      NULL, fetch_scanline_a2b10g10r10_float,
+      fetch_pixel_generic_lossy_32, fetch_pixel_a2b10g10r10_float,
+      NULL, store_scanline_a2b10g10r10_float },
+
     { PIXMAN_x2b10g10r10,
-      NULL, fetch_scanline_x2b10g10r10,
-      fetch_pixel_generic_lossy_32, fetch_pixel_x2b10g10r10,
-      NULL, store_scanline_x2b10g10r10 },
-    
+      NULL, fetch_scanline_x2b10g10r10_float,
+      fetch_pixel_generic_lossy_32, fetch_pixel_x2b10g10r10_float,
+      NULL, store_scanline_x2b10g10r10_float },
+
 /* YUV formats */
     { PIXMAN_yuy2,
-      fetch_scanline_yuy2, fetch_scanline_generic_64,
-      fetch_pixel_yuy2, fetch_pixel_generic_64,
+      fetch_scanline_yuy2, fetch_scanline_generic_float,
+      fetch_pixel_yuy2, fetch_pixel_generic_float,
       NULL, NULL },
-    
+
     { PIXMAN_yv12,
-      fetch_scanline_yv12, fetch_scanline_generic_64,
-      fetch_pixel_yv12, fetch_pixel_generic_64,
+      fetch_scanline_yv12, fetch_scanline_generic_float,
+      fetch_pixel_yv12, fetch_pixel_generic_float,
       NULL, NULL },
     
     { PIXMAN_null },
@@ -1312,11 +1396,11 @@ setup_accessors (bits_image_t *image)
 	if (info->format == image->format)
 	{
 	    image->fetch_scanline_32 = info->fetch_scanline_32;
-	    image->fetch_scanline_64 = info->fetch_scanline_64;
+	    image->fetch_scanline_float = info->fetch_scanline_float;
 	    image->fetch_pixel_32 = info->fetch_pixel_32;
-	    image->fetch_pixel_64 = info->fetch_pixel_64;
+	    image->fetch_pixel_float = info->fetch_pixel_float;
 	    image->store_scanline_32 = info->store_scanline_32;
-	    image->store_scanline_64 = info->store_scanline_64;
+	    image->store_scanline_float = info->store_scanline_float;
 	    
 	    return;
 	}
diff --git a/pixman/pixman/pixman-bits-image.c b/pixman/pixman/pixman-bits-image.c
index c69e1518b..029093ddc 100644
--- a/pixman/pixman/pixman-bits-image.c
+++ b/pixman/pixman/pixman-bits-image.c
@@ -36,43 +36,18 @@
 #include "pixman-combine32.h"
 #include "pixman-inlines.h"
 
-/*
- * By default, just evaluate the image at 32bpp and expand.  Individual image
- * types can plug in a better scanline getter if they want to. For example
- * we  could produce smoother gradients by evaluating them at higher color
- * depth, but that's a project for the future.
- */
 static uint32_t *
-_pixman_image_get_scanline_generic_64 (pixman_iter_t * iter,
-                                       const uint32_t *mask)
+_pixman_image_get_scanline_generic_float (pixman_iter_t * iter,
+					  const uint32_t *mask)
 {
-    int             width  = iter->width;
-    uint32_t *      buffer = iter->buffer;
-
     pixman_iter_get_scanline_t fetch_32 = iter->data;
-    uint32_t *mask8 = NULL;
-
-    /* Contract the mask image, if one exists, so that the 32-bit fetch
-     * function can use it.
-     */
-    if (mask)
-    {
-	mask8 = pixman_malloc_ab (width, sizeof(uint32_t));
-	if (!mask8)
-	    return buffer;
-
-	pixman_contract (mask8, (uint64_t *)mask, width);
-    }
-
-    /* Fetch the source image into the first half of buffer. */
-    fetch_32 (iter, mask8);
+    uint32_t *buffer = iter->buffer;
 
-    /* Expand from 32bpp to 64bpp in place. */
-    pixman_expand ((uint64_t *)buffer, buffer, PIXMAN_a8r8g8b8, width);
+    fetch_32 (iter, NULL);
 
-    free (mask8);
+    pixman_expand_to_float ((argb_t *)buffer, buffer, PIXMAN_a8r8g8b8, iter->width);
 
-    return buffer;
+    return iter->buffer;
 }
 
 /* Fetch functions */
@@ -965,17 +940,17 @@ replicate_pixel_32 (bits_image_t *   bits,
 }
 
 static void
-replicate_pixel_64 (bits_image_t *   bits,
-		    int              x,
-		    int              y,
-		    int              width,
-		    uint32_t *       b)
+replicate_pixel_float (bits_image_t *   bits,
+		       int              x,
+		       int              y,
+		       int              width,
+		       uint32_t *       b)
 {
-    uint64_t color;
-    uint64_t *buffer = (uint64_t *)b;
-    uint64_t *end;
+    argb_t color;
+    argb_t *buffer = (argb_t *)b;
+    argb_t *end;
 
-    color = bits->fetch_pixel_64 (bits, x, y);
+    color = bits->fetch_pixel_float (bits, x, y);
 
     end = buffer + width;
     while (buffer < end)
@@ -994,7 +969,7 @@ bits_image_fetch_untransformed_repeat_none (bits_image_t *image,
 
     if (y < 0 || y >= image->height)
     {
-	memset (buffer, 0, width * (wide? 8 : 4));
+	memset (buffer, 0, width * (wide? sizeof (argb_t) : 4));
 	return;
     }
 
@@ -1002,10 +977,10 @@ bits_image_fetch_untransformed_repeat_none (bits_image_t *image,
     {
 	w = MIN (width, -x);
 
-	memset (buffer, 0, w * (wide ? 8 : 4));
+	memset (buffer, 0, w * (wide ? sizeof (argb_t) : 4));
 
 	width -= w;
-	buffer += w * (wide? 2 : 1);
+	buffer += w * (wide? 4 : 1);
 	x += w;
     }
 
@@ -1014,16 +989,16 @@ bits_image_fetch_untransformed_repeat_none (bits_image_t *image,
 	w = MIN (width, image->width - x);
 
 	if (wide)
-	    image->fetch_scanline_64 ((pixman_image_t *)image, x, y, w, buffer, NULL);
+	    image->fetch_scanline_float ((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);
+	buffer += w * (wide? 4 : 1);
 	x += w;
     }
 
-    memset (buffer, 0, width * (wide ? 8 : 4));
+    memset (buffer, 0, width * (wide ? sizeof (argb_t) : 4));
 }
 
 static void
@@ -1045,7 +1020,7 @@ bits_image_fetch_untransformed_repeat_normal (bits_image_t *image,
     if (image->width == 1)
     {
 	if (wide)
-	    replicate_pixel_64 (image, 0, y, width, buffer);
+	    replicate_pixel_float (image, 0, y, width, buffer);
 	else
 	    replicate_pixel_32 (image, 0, y, width, buffer);
 
@@ -1062,11 +1037,11 @@ bits_image_fetch_untransformed_repeat_normal (bits_image_t *image,
 	w = MIN (width, image->width - x);
 
 	if (wide)
-	    image->fetch_scanline_64 ((pixman_image_t *)image, x, y, w, buffer, NULL);
+	    image->fetch_scanline_float ((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);
+	buffer += w * (wide? 4 : 1);
 	x += w;
 	width -= w;
     }
@@ -1098,9 +1073,8 @@ bits_image_fetch_untransformed_32 (pixman_iter_t * iter,
 }
 
 static uint32_t *
-bits_image_fetch_untransformed_64 (pixman_iter_t * iter,
-                                   const uint32_t *mask)
-				   
+bits_image_fetch_untransformed_float (pixman_iter_t * iter,
+				      const uint32_t *mask)
 {
     pixman_image_t *image  = iter->image;
     int             x      = iter->x;
@@ -1128,7 +1102,7 @@ typedef struct
     pixman_format_code_t	format;
     uint32_t			flags;
     pixman_iter_get_scanline_t	get_scanline_32;
-    pixman_iter_get_scanline_t	get_scanline_64;
+    pixman_iter_get_scanline_t  get_scanline_float;
 } fetcher_info_t;
 
 static const fetcher_info_t fetcher_info[] =
@@ -1140,7 +1114,7 @@ static const fetcher_info_t fetcher_info[] =
        FAST_PATH_NO_PAD_REPEAT			|
        FAST_PATH_NO_REFLECT_REPEAT),
       bits_image_fetch_untransformed_32,
-      bits_image_fetch_untransformed_64
+      bits_image_fetch_untransformed_float
     },
 
 #define FAST_BILINEAR_FLAGS						\
@@ -1156,13 +1130,13 @@ static const fetcher_info_t fetcher_info[] =
     { PIXMAN_a8r8g8b8,
       FAST_BILINEAR_FLAGS,
       bits_image_fetch_bilinear_no_repeat_8888,
-      _pixman_image_get_scanline_generic_64
+      _pixman_image_get_scanline_generic_float
     },
 
     { PIXMAN_x8r8g8b8,
       FAST_BILINEAR_FLAGS,
       bits_image_fetch_bilinear_no_repeat_8888,
-      _pixman_image_get_scanline_generic_64
+      _pixman_image_get_scanline_generic_float
     },
 
 #define GENERAL_BILINEAR_FLAGS						\
@@ -1183,14 +1157,14 @@ static const fetcher_info_t fetcher_info[] =
     { PIXMAN_ ## format,						\
       GENERAL_BILINEAR_FLAGS | FAST_PATH_ ## repeat ## _REPEAT,		\
       bits_image_fetch_bilinear_affine_ ## name,			\
-      _pixman_image_get_scanline_generic_64				\
+      _pixman_image_get_scanline_generic_float				\
     },
 
 #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				\
+      bits_image_fetch_nearest_affine_ ## name,				\
+      _pixman_image_get_scanline_generic_float				\
     },
 
 #define AFFINE_FAST_PATHS(name, format, repeat)				\
@@ -1218,11 +1192,15 @@ static const fetcher_info_t fetcher_info[] =
     { 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
+      _pixman_image_get_scanline_generic_float
     },
 
     /* General */
-    { PIXMAN_any, 0, bits_image_fetch_general, _pixman_image_get_scanline_generic_64 },
+    { PIXMAN_any,
+      0,
+      bits_image_fetch_general,
+      _pixman_image_get_scanline_generic_float
+    },
 
     { PIXMAN_null },
 };
@@ -1252,7 +1230,7 @@ _pixman_bits_image_src_iter_init (pixman_image_t *image, pixman_iter_t *iter)
 	    else
 	    {
 		iter->data = info->get_scanline_32;
-		iter->get_scanline = info->get_scanline_64;
+		iter->get_scanline = info->get_scanline_float;
 	    }
 	    return;
 	}
@@ -1307,30 +1285,27 @@ dest_get_scanline_wide (pixman_iter_t *iter, const uint32_t *mask)
     int             x      = iter->x;
     int             y      = iter->y;
     int             width  = iter->width;
-    uint64_t *	    buffer = (uint64_t *)iter->buffer;
+    argb_t *	    buffer = (argb_t *)iter->buffer;
 
-    image->fetch_scanline_64 (
+    image->fetch_scanline_float (
 	(pixman_image_t *)image, x, y, width, (uint32_t *)buffer, mask);
     if (image->common.alpha_map)
     {
-	uint64_t *alpha;
+	argb_t *alpha;
 
-	if ((alpha = malloc (width * sizeof (uint64_t))))
+	if ((alpha = malloc (width * sizeof (argb_t))))
 	{
 	    int i;
 
 	    x -= image->common.alpha_origin_x;
 	    y -= image->common.alpha_origin_y;
 
-	    image->common.alpha_map->fetch_scanline_64 (
+	    image->common.alpha_map->fetch_scanline_float (
 		(pixman_image_t *)image->common.alpha_map,
 		x, y, width, (uint32_t *)alpha, mask);
 
 	    for (i = 0; i < width; ++i)
-	    {
-		buffer[i] &= ~0xffff000000000000ULL;
-		buffer[i] |= (alpha[i] & 0xffff000000000000ULL);
-	    }
+		buffer[i].a = alpha[i].a;
 
 	    free (alpha);
 	}
@@ -1371,14 +1346,14 @@ dest_write_back_wide (pixman_iter_t *iter)
     int             width  = iter->width;
     const uint32_t *buffer = iter->buffer;
 
-    image->store_scanline_64 (image, x, y, width, buffer);
+    image->store_scanline_float (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->store_scanline_float (
 	    image->common.alpha_map, x, y, width, buffer);
     }
 
diff --git a/pixman/pixman/pixman-combine-float.c b/pixman/pixman/pixman-combine-float.c
new file mode 100644
index 000000000..7bf5b5ae7
--- /dev/null
+++ b/pixman/pixman/pixman-combine-float.c
@@ -0,0 +1,1003 @@
+/* -*- Mode: c; c-basic-offset: 4; tab-width: 8; indent-tabs-mode: t; -*- */
+/*
+ * Copyright © 2010, 2012 Soren Sandmann Pedersen
+ * Copyright © 2010, 2012 Red Hat, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Author: Soren Sandmann Pedersen (sandmann@cs.au.dk)
+ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <math.h>
+#include <string.h>
+#include <float.h>
+
+#include "pixman-private.h"
+
+typedef float (* combine_channel_t) (float sa, float s, float da, float d);
+
+static force_inline void
+combine_inner (pixman_bool_t component,
+	       float *dest, const float *src, const float *mask, int n_pixels,
+	       combine_channel_t combine_a, combine_channel_t combine_c)
+{
+    int i;
+
+    if (!mask)
+    {
+	for (i = 0; i < 4 * n_pixels; i += 4)
+	{
+	    float sa = src[i + 0];
+	    float sr = src[i + 1];
+	    float sg = src[i + 2];
+	    float sb = src[i + 3];
+	    
+	    float da = dest[i + 0];
+	    float dr = dest[i + 1];
+	    float dg = dest[i + 2];
+	    float db = dest[i + 3];					
+	    
+	    dest[i + 0] = combine_a (sa, sa, da, da);
+	    dest[i + 1] = combine_c (sa, sr, da, dr);
+	    dest[i + 2] = combine_c (sa, sg, da, dg);
+	    dest[i + 3] = combine_c (sa, sb, da, db);
+	}
+    }
+    else
+    {
+	for (i = 0; i < 4 * n_pixels; i += 4)
+	{
+	    float sa, sr, sg, sb;
+	    float ma, mr, mg, mb;
+	    float da, dr, dg, db;
+	    
+	    sa = src[i + 0];
+	    sr = src[i + 1];
+	    sg = src[i + 2];
+	    sb = src[i + 3];
+	    
+	    if (component)
+	    {
+		ma = mask[i + 0];
+		mr = mask[i + 1];
+		mg = mask[i + 2];
+		mb = mask[i + 3];
+
+		sr *= mr;
+		sg *= mg;
+		sb *= mb;
+
+		ma *= sa;
+		mr *= sa;
+		mg *= sa;
+		mb *= sa;
+		
+		sa = ma;
+	    }
+	    else
+	    {
+		ma = mask[i + 0];
+
+		sa *= ma;
+		sr *= ma;
+		sg *= ma;
+		sb *= ma;
+
+		ma = mr = mg = mb = sa;
+	    }
+	    
+	    da = dest[i + 0];
+	    dr = dest[i + 1];
+	    dg = dest[i + 2];
+	    db = dest[i + 3];
+	    
+	    dest[i + 0] = combine_a (ma, sa, da, da);
+	    dest[i + 1] = combine_c (mr, sr, da, dr);
+	    dest[i + 2] = combine_c (mg, sg, da, dg);
+	    dest[i + 3] = combine_c (mb, sb, da, db);
+	}
+    }
+}
+
+#define MAKE_COMBINER(name, component, combine_a, combine_c)		\
+    static void								\
+    combine_ ## name ## _float (pixman_implementation_t *imp,		\
+				pixman_op_t              op,		\
+				float                   *dest,		\
+				const float             *src,		\
+				const float             *mask,		\
+				int		         n_pixels)	\
+    {									\
+	combine_inner (component, dest, src, mask, n_pixels,		\
+		       combine_a, combine_c);				\
+    }
+
+#define MAKE_COMBINERS(name, combine_a, combine_c)			\
+    MAKE_COMBINER(name ## _ca, TRUE, combine_a, combine_c)		\
+    MAKE_COMBINER(name ## _u, FALSE, combine_a, combine_c)
+
+
+/*
+ * Porter/Duff operators
+ */
+typedef enum
+{
+    ZERO,
+    ONE,
+    SRC_ALPHA,
+    DEST_ALPHA,
+    INV_SA,
+    INV_DA,
+    SA_OVER_DA,
+    DA_OVER_SA,
+    INV_SA_OVER_DA,
+    INV_DA_OVER_SA,
+    ONE_MINUS_SA_OVER_DA,
+    ONE_MINUS_DA_OVER_SA,
+    ONE_MINUS_INV_DA_OVER_SA,
+    ONE_MINUS_INV_SA_OVER_DA
+} combine_factor_t;
+
+#define CLAMP(f)					\
+    (((f) < 0)? 0 : (((f) > 1.0) ? 1.0 : (f)))
+
+static force_inline float
+get_factor (combine_factor_t factor, float sa, float da)
+{
+    float f = -1;
+
+    switch (factor)
+    {
+    case ZERO:
+	f = 0.0f;
+	break;
+
+    case ONE:
+	f = 1.0f;
+	break;
+
+    case SRC_ALPHA:
+	f = sa;
+	break;
+
+    case DEST_ALPHA:
+	f = da;
+	break;
+
+    case INV_SA:
+	f = 1 - sa;
+	break;
+
+    case INV_DA:
+	f = 1 - da;
+	break;
+
+    case SA_OVER_DA:
+	if (da == 0.0f)
+	    f = 1.0f;
+	else
+	    f = CLAMP (sa / da);
+	break;
+
+    case DA_OVER_SA:
+	if (sa == 0.0f)
+	    f = 1.0f;
+	else
+	    f = CLAMP (da / sa);
+	break;
+
+    case INV_SA_OVER_DA:
+	if (da == 0.0f)
+	    f = 1.0f;
+	else
+	    f = CLAMP ((1.0f - sa) / da);
+	break;
+
+    case INV_DA_OVER_SA:
+	if (sa == 0.0f)
+	    f = 1.0f;
+	else
+	    f = CLAMP ((1.0f - da) / sa);
+	break;
+
+    case ONE_MINUS_SA_OVER_DA:
+	if (da == 0.0f)
+	    f = 0.0f;
+	else
+	    f = CLAMP (1.0f - sa / da);
+	break;
+
+    case ONE_MINUS_DA_OVER_SA:
+	if (sa == 0.0f)
+	    f = 0.0f;
+	else
+	    f = CLAMP (1.0f - da / sa);
+	break;
+
+    case ONE_MINUS_INV_DA_OVER_SA:
+	if (sa == 0.0f)
+	    f = 0.0f;
+	else
+	    f = CLAMP (1.0f - (1.0f - da) / sa);
+	break;
+
+    case ONE_MINUS_INV_SA_OVER_DA:
+	if (da == 0.0f)
+	    f = 0.0f;
+	else
+	    f = CLAMP (1.0f - (1.0f - sa) / da);
+	break;
+    }
+
+    return f;
+}
+
+#define MAKE_PD_COMBINERS(name, a, b)					\
+    static float force_inline						\
+    pd_combine_ ## name (float sa, float s, float da, float d)		\
+    {									\
+	const float fa = get_factor (a, sa, da);			\
+	const float fb = get_factor (b, sa, da);			\
+									\
+	return MIN (1.0f, s * fa + d * fb);				\
+    }									\
+    									\
+    MAKE_COMBINERS(name, pd_combine_ ## name, pd_combine_ ## name)
+
+MAKE_PD_COMBINERS (clear,			ZERO,				ZERO)
+MAKE_PD_COMBINERS (src,				ONE,				ZERO)
+MAKE_PD_COMBINERS (dst,				ZERO,				ONE)
+MAKE_PD_COMBINERS (over,			ONE,				INV_SA)
+MAKE_PD_COMBINERS (over_reverse,		INV_DA,				ONE)
+MAKE_PD_COMBINERS (in,				DEST_ALPHA,			ZERO)
+MAKE_PD_COMBINERS (in_reverse,			ZERO,				SRC_ALPHA)
+MAKE_PD_COMBINERS (out,				INV_DA,				ZERO)
+MAKE_PD_COMBINERS (out_reverse,			ZERO,				INV_SA)
+MAKE_PD_COMBINERS (atop,			DEST_ALPHA,			INV_SA)
+MAKE_PD_COMBINERS (atop_reverse,		INV_DA,				SRC_ALPHA)
+MAKE_PD_COMBINERS (xor,				INV_DA,				INV_SA)
+MAKE_PD_COMBINERS (add,				ONE,				ONE)
+
+MAKE_PD_COMBINERS (saturate,			INV_DA_OVER_SA,			ONE)
+
+MAKE_PD_COMBINERS (disjoint_clear,		ZERO,				ZERO)
+MAKE_PD_COMBINERS (disjoint_src,		ONE,				ZERO)
+MAKE_PD_COMBINERS (disjoint_dst,		ZERO,				ONE)
+MAKE_PD_COMBINERS (disjoint_over,		ONE,				INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_over_reverse,	INV_DA_OVER_SA,			ONE)
+MAKE_PD_COMBINERS (disjoint_in,			ONE_MINUS_INV_DA_OVER_SA,	ZERO)
+MAKE_PD_COMBINERS (disjoint_in_reverse,		ZERO,				ONE_MINUS_INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_out,		INV_DA_OVER_SA,			ZERO)
+MAKE_PD_COMBINERS (disjoint_out_reverse,	ZERO,				INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_atop,		ONE_MINUS_INV_DA_OVER_SA,	INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_atop_reverse,	INV_DA_OVER_SA,			ONE_MINUS_INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_xor,		INV_DA_OVER_SA,			INV_SA_OVER_DA)
+
+MAKE_PD_COMBINERS (conjoint_clear,		ZERO,				ZERO)
+MAKE_PD_COMBINERS (conjoint_src,		ONE,				ZERO)
+MAKE_PD_COMBINERS (conjoint_dst,		ZERO,				ONE)
+MAKE_PD_COMBINERS (conjoint_over,		ONE,				ONE_MINUS_SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_over_reverse,	ONE_MINUS_DA_OVER_SA,		ONE)
+MAKE_PD_COMBINERS (conjoint_in,			DA_OVER_SA,			ZERO)
+MAKE_PD_COMBINERS (conjoint_in_reverse,		ZERO,				SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_out,		ONE_MINUS_DA_OVER_SA,		ZERO)
+MAKE_PD_COMBINERS (conjoint_out_reverse,	ZERO,				ONE_MINUS_SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_atop,		DA_OVER_SA,			ONE_MINUS_SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_atop_reverse,	ONE_MINUS_DA_OVER_SA,		SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_xor,		ONE_MINUS_DA_OVER_SA,		ONE_MINUS_SA_OVER_DA)
+
+/*
+ * PDF blend modes:
+ *
+ * The following blend modes have been taken from the PDF ISO 32000
+ * specification, which at this point in time is available from
+ * http://www.adobe.com/devnet/acrobat/pdfs/PDF32000_2008.pdf
+ * The relevant chapters are 11.3.5 and 11.3.6.
+ * The formula for computing the final pixel color given in 11.3.6 is:
+ * αr × Cr = (1 – αs) × αb × Cb + (1 – αb) × αs × Cs + αb × αs × B(Cb, Cs)
+ * with B() being the blend function.
+ * Note that OVER is a special case of this operation, using B(Cb, Cs) = Cs
+ *
+ * These blend modes should match the SVG filter draft specification, as
+ * it has been designed to mirror ISO 32000. Note that at the current point
+ * no released draft exists that shows this, as the formulas have not been
+ * updated yet after the release of ISO 32000.
+ *
+ * The default implementation here uses the PDF_SEPARABLE_BLEND_MODE and
+ * PDF_NON_SEPARABLE_BLEND_MODE macros, which take the blend function as an
+ * argument. Note that this implementation operates on premultiplied colors,
+ * while the PDF specification does not. Therefore the code uses the formula
+ * ar.Cra = (1 – as) . Dca + (1 – ad) . Sca + B(Dca, ad, Sca, as)
+ */
+
+#define MAKE_SEPARABLE_PDF_COMBINERS(name)				\
+    static force_inline float						\
+    combine_ ## name ## _a (float sa, float s, float da, float d)	\
+    {									\
+	return da + sa - da * sa;					\
+    }									\
+    									\
+    static force_inline float						\
+    combine_ ## name ## _c (float sa, float s, float da, float d)	\
+    {									\
+	float f = (1 - sa) * d + (1 - da) * s;				\
+									\
+	return f + blend_ ## name (sa, s, da, d);			\
+    }									\
+    									\
+    MAKE_COMBINERS (name, combine_ ## name ## _a, combine_ ## name ## _c)
+
+static force_inline float
+blend_multiply (float sa, float s, float da, float d)
+{
+    return d * s;
+}
+
+static force_inline float
+blend_screen (float sa, float s, float da, float d)
+{
+    return d * sa + s * da - s * d;
+}
+
+static force_inline float
+blend_overlay (float sa, float s, float da, float d)
+{
+    if (2 * d < da)
+	return 2 * s * d;
+    else
+	return sa * da - 2 * (da - d) * (sa - s);
+}
+
+static force_inline float
+blend_darken (float sa, float s, float da, float d)
+{
+    s = s * da;
+    d = d * sa;
+
+    if (s > d)
+	return d;
+    else
+	return s;
+}
+
+static force_inline float
+blend_lighten (float sa, float s, float da, float d)
+{
+    s = s * da;
+    d = d * sa;
+
+    if (s > d)
+	return s;
+    else
+	return d;
+}
+
+static force_inline float
+blend_color_dodge (float sa, float s, float da, float d)
+{
+    if (d == 0.0f)
+	return 0.0f;
+    else if (d * sa >= sa * da - s * da)
+	return sa * da;
+    else if (sa - s == 0.0f)
+	return sa * da;
+    else
+	return sa * sa * d / (sa - s);
+}
+
+static force_inline float
+blend_color_burn (float sa, float s, float da, float d)
+{
+    if (d >= da)
+	return sa * da;
+    else if (sa * (da - d) >= s * da)
+	return 0.0f;
+    else if (s == 0.0f)
+	return 0.0f;
+    else
+	return sa * (da - sa * (da - d) / s);
+}
+
+static force_inline float
+blend_hard_light (float sa, float s, float da, float d)
+{
+    if (2 * s < sa)
+	return 2 * s * d;
+    else
+	return sa * da - 2 * (da - d) * (sa - s);
+}
+
+static force_inline float
+blend_soft_light (float sa, float s, float da, float d)
+{
+    if (2 * s < sa)
+    {
+	if (da == 0.0f)
+	    return d * sa;
+	else
+	    return d * sa - d * (da - d) * (sa - 2 * s) / da;
+    }
+    else
+    {
+	if (da == 0.0f)
+	{
+	    return 0.0f;
+	}
+	else
+	{
+	    if (4 * d <= da)
+		return d * sa + (2 * s - sa) * d * ((16 * d / da - 12) * d / da + 3);
+	    else
+		return d * sa + (sqrtf (d * da) - d) * (2 * s - sa);
+	}
+    }
+}
+
+static force_inline float
+blend_difference (float sa, float s, float da, float d)
+{
+    float dsa = d * sa;
+    float sda = s * da;
+
+    if (sda < dsa)
+	return dsa - sda;
+    else
+	return sda - dsa;
+}
+
+static force_inline float
+blend_exclusion (float sa, float s, float da, float d)
+{
+    return s * da + d * sa - 2 * d * s;
+}
+
+MAKE_SEPARABLE_PDF_COMBINERS (multiply)
+MAKE_SEPARABLE_PDF_COMBINERS (screen)
+MAKE_SEPARABLE_PDF_COMBINERS (overlay)
+MAKE_SEPARABLE_PDF_COMBINERS (darken)
+MAKE_SEPARABLE_PDF_COMBINERS (lighten)
+MAKE_SEPARABLE_PDF_COMBINERS (color_dodge)
+MAKE_SEPARABLE_PDF_COMBINERS (color_burn)
+MAKE_SEPARABLE_PDF_COMBINERS (hard_light)
+MAKE_SEPARABLE_PDF_COMBINERS (soft_light)
+MAKE_SEPARABLE_PDF_COMBINERS (difference)
+MAKE_SEPARABLE_PDF_COMBINERS (exclusion)
+
+/*
+ * PDF nonseperable blend modes.
+ *
+ * These are implemented using the following functions to operate in Hsl
+ * space, with Cmax, Cmid, Cmin referring to the max, mid and min value
+ * of the red, green and blue components.
+ *
+ * LUM (C) = 0.3 × Cred + 0.59 × Cgreen + 0.11 × Cblue
+ *
+ * clip_color (C):
+ *   l = LUM (C)
+ *   min = Cmin
+ *   max = Cmax
+ *   if n < 0.0
+ *     C = l + (((C – l) × l) ⁄     (l – min))
+ *   if x > 1.0
+ *     C = l + (((C – l) × (1 – l)) (max – l))
+ *   return C
+ *
+ * set_lum (C, l):
+ *   d = l – LUM (C)
+ *   C += d
+ *   return clip_color (C)
+ *
+ * SAT (C) = CH_MAX (C) - CH_MIN (C)
+ *
+ * set_sat (C, s):
+ *  if Cmax > Cmin
+ *    Cmid = ( ( ( Cmid – Cmin ) × s ) ⁄ ( Cmax – Cmin ) )
+ *    Cmax = s
+ *  else
+ *    Cmid = Cmax = 0.0
+ *  Cmin = 0.0
+ *  return C
+ */
+
+/* For premultiplied colors, we need to know what happens when C is
+ * multiplied by a real number. LUM and SAT are linear:
+ *
+ *    LUM (r × C) = r × LUM (C)		SAT (r × C) = r × SAT (C)
+ *
+ * If we extend clip_color with an extra argument a and change
+ *
+ *        if x >= 1.0
+ *
+ * into
+ *
+ *        if x >= a
+ *
+ * then clip_color is also linear:
+ *
+ *     r * clip_color (C, a) = clip_color (r_c, ra);
+ *
+ * for positive r.
+ *
+ * Similarly, we can extend set_lum with an extra argument that is just passed
+ * on to clip_color:
+ *
+ *     r × set_lum ( C, l, a)
+ *
+ *   = r × clip_color ( C + l - LUM (C), a)
+ *
+ *   = clip_color ( r * C + r × l - LUM (r × C), r * a)
+ *
+ *   = set_lum ( r * C, r * l, r * a)
+ *
+ * Finally, set_sat:
+ *
+ *     r * set_sat (C, s) = set_sat (x * C, r * s)
+ *
+ * The above holds for all non-zero x because they x'es in the fraction for
+ * C_mid cancel out. Specifically, it holds for x = r:
+ *
+ *     r * set_sat (C, s) = set_sat (r_c, rs)
+ *
+ *
+ *
+ *
+ * So, for the non-separable PDF blend modes, we have (using s, d for
+ * non-premultiplied colors, and S, D for premultiplied:
+ *
+ *   Color:
+ *
+ *     a_s * a_d * B(s, d)
+ *   = a_s * a_d * set_lum (S/a_s, LUM (D/a_d), 1)
+ *   = set_lum (S * a_d, a_s * LUM (D), a_s * a_d)
+ *
+ *
+ *   Luminosity:
+ *
+ *     a_s * a_d * B(s, d)
+ *   = a_s * a_d * set_lum (D/a_d, LUM(S/a_s), 1)
+ *   = set_lum (a_s * D, a_d * LUM(S), a_s * a_d)
+ *
+ *
+ *   Saturation:
+ *
+ *     a_s * a_d * B(s, d)
+ *   = a_s * a_d * set_lum (set_sat (D/a_d, SAT (S/a_s)), LUM (D/a_d), 1)
+ *   = set_lum (a_s * a_d * set_sat (D/a_d, SAT (S/a_s)),
+ *                                        a_s * LUM (D), a_s * a_d)
+ *   = set_lum (set_sat (a_s * D, a_d * SAT (S), a_s * LUM (D), a_s * a_d))
+ *
+ *   Hue:
+ *
+ *     a_s * a_d * B(s, d)
+ *   = a_s * a_d * set_lum (set_sat (S/a_s, SAT (D/a_d)), LUM (D/a_d), 1)
+ *   = set_lum (set_sat (a_d * S, a_s * SAT (D)), a_s * LUM (D), a_s * a_d)
+ *
+ */
+
+typedef struct
+{
+    float	r;
+    float	g;
+    float	b;
+} rgb_t;
+
+static force_inline float
+minf (float a, float b)
+{
+    return a < b? a : b;
+}
+
+static force_inline float
+maxf (float a, float b)
+{
+    return a > b? a : b;
+}
+
+static force_inline float
+channel_min (const rgb_t *c)
+{
+    return minf (minf (c->r, c->g), c->b);
+}
+
+static force_inline float
+channel_max (const rgb_t *c)
+{
+    return maxf (maxf (c->r, c->g), c->b);
+}
+
+static force_inline float
+get_lum (const rgb_t *c)
+{
+    return c->r * 0.3f + c->g * 0.59f + c->b * 0.11f;
+}
+
+static force_inline float
+get_sat (const rgb_t *c)
+{
+    return channel_max (c) - channel_min (c);
+}
+
+static void
+clip_color (rgb_t *color, float a)
+{
+    float l = get_lum (color);
+    float n = channel_min (color);
+    float x = channel_max (color);
+
+    if (n < 0.0f)
+    {
+	if ((l - n) < 4 * FLT_EPSILON)
+	{
+	    color->r = 0.0f;
+	    color->g = 0.0f;
+	    color->b = 0.0f;
+	}
+	else
+	{
+	    color->r = l + (((color->r - l) * l) / (l - n));
+	    color->g = l + (((color->g - l) * l) / (l - n));
+	    color->b = l + (((color->b - l) * l) / (l - n));
+	}
+    }
+    if (x > a)
+    {
+	if ((x - l) < 4 * FLT_EPSILON)
+	{
+	    color->r = a;
+	    color->g = a;
+	    color->b = a;
+	}
+	else
+	{
+	    color->r = l + (((color->r - l) * (a - l) / (x - l)));
+	    color->g = l + (((color->g - l) * (a - l) / (x - l)));
+	    color->b = l + (((color->b - l) * (a - l) / (x - l)));
+	}
+    }
+}
+
+static void
+set_lum (rgb_t *color, float sa, float l)
+{
+    float d = l - get_lum (color);
+
+    color->r = color->r + d;
+    color->g = color->g + d;
+    color->b = color->b + d;
+
+    clip_color (color, sa);
+}
+
+static void
+set_sat (rgb_t *src, float sat)
+{
+    float *max, *mid, *min;
+
+    if (src->r > src->g)
+    {
+	if (src->r > src->b)
+	{
+	    max = &(src->r);
+
+	    if (src->g > src->b)
+	    {
+		mid = &(src->g);
+		min = &(src->b);
+	    }
+	    else
+	    {
+		mid = &(src->b);
+		min = &(src->g);
+	    }
+	}
+	else
+	{
+	    max = &(src->b);
+	    mid = &(src->r);
+	    min = &(src->g);
+	}
+    }
+    else
+    {
+	if (src->r > src->b)
+	{
+	    max = &(src->g);
+	    mid = &(src->r);
+	    min = &(src->b);
+	}
+	else
+	{
+	    min = &(src->r);
+
+	    if (src->g > src->b)
+	    {
+		max = &(src->g);
+		mid = &(src->b);
+	    }
+	    else
+	    {
+		max = &(src->b);
+		mid = &(src->g);
+	    }
+	}
+    }
+
+    if (*max > *min)
+    {
+	*mid = (((*mid - *min) * sat) / (*max - *min));
+	*max = sat;
+    }
+    else
+    {
+	*mid = *max = 0.0f;
+    }
+
+    *min = 0.0f;
+}
+
+/*
+ * Hue:
+ * B(Cb, Cs) = set_lum (set_sat (Cs, SAT (Cb)), LUM (Cb))
+ */
+static force_inline void
+blend_hsl_hue (rgb_t *res,
+	       const rgb_t *dest, float da,
+	       const rgb_t *src, float sa)
+{
+    res->r = src->r * da;
+    res->g = src->g * da;
+    res->b = src->b * da;
+
+    set_sat (res, get_sat (dest) * sa);
+    set_lum (res, sa * da, get_lum (dest) * sa);
+}
+
+/*
+ * Saturation:
+ * B(Cb, Cs) = set_lum (set_sat (Cb, SAT (Cs)), LUM (Cb))
+ */
+static force_inline void
+blend_hsl_saturation (rgb_t *res,
+		      const rgb_t *dest, float da,
+		      const rgb_t *src, float sa)
+{
+    res->r = dest->r * sa;
+    res->g = dest->g * sa;
+    res->b = dest->b * sa;
+
+    set_sat (res, get_sat (src) * da);
+    set_lum (res, sa * da, get_lum (dest) * sa);
+}
+
+/*
+ * Color:
+ * B(Cb, Cs) = set_lum (Cs, LUM (Cb))
+ */
+static force_inline void
+blend_hsl_color (rgb_t *res,
+		 const rgb_t *dest, float da,
+		 const rgb_t *src, float sa)
+{
+    res->r = src->r * da;
+    res->g = src->g * da;
+    res->b = src->b * da;
+
+    set_lum (res, sa * da, get_lum (dest) * sa);
+}
+
+/*
+ * Luminosity:
+ * B(Cb, Cs) = set_lum (Cb, LUM (Cs))
+ */
+static force_inline void
+blend_hsl_luminosity (rgb_t *res,
+		      const rgb_t *dest, float da,
+		      const rgb_t *src, float sa)
+{
+    res->r = dest->r * sa;
+    res->g = dest->g * sa;
+    res->b = dest->b * sa;
+
+    set_lum (res, sa * da, get_lum (src) * da);
+}
+
+#define MAKE_NON_SEPARABLE_PDF_COMBINERS(name)				\
+    static void								\
+    combine_ ## name ## _u_float (pixman_implementation_t *imp,		\
+				  pixman_op_t              op,		\
+				  float                   *dest,	\
+				  const float             *src,		\
+				  const float             *mask,	\
+				  int		           n_pixels)	\
+    {									\
+    	int i;								\
+									\
+	for (i = 0; i < 4 * n_pixels; i += 4)				\
+	{								\
+	    float sa, da;						\
+	    rgb_t sc, dc, rc;						\
+									\
+	    sa = src[i + 0];						\
+	    sc.r = src[i + 1];						\
+	    sc.g = src[i + 2];						\
+	    sc.b = src[i + 3];						\
+									\
+	    da = dest[i + 0];						\
+	    dc.r = dest[i + 1];						\
+	    dc.g = dest[i + 2];						\
+	    dc.b = dest[i + 3];						\
+									\
+	    if (mask)							\
+	    {								\
+		float ma = mask[i + 0];					\
+									\
+		/* Component alpha is not supported for HSL modes */	\
+		sa *= ma;						\
+		sc.r *= ma;						\
+		sc.g *= ma;						\
+		sc.g *= ma;						\
+	    }								\
+									\
+	    blend_ ## name (&rc, &dc, da, &sc, sa);			\
+									\
+	    dest[i + 0] = sa + da - sa * da;				\
+	    dest[i + 1] = (1 - sa) * dc.r + (1 - da) * sc.r + rc.r;	\
+	    dest[i + 2] = (1 - sa) * dc.g + (1 - da) * sc.g + rc.g;	\
+	    dest[i + 3] = (1 - sa) * dc.b + (1 - da) * sc.b + rc.b;	\
+	}								\
+    }
+
+MAKE_NON_SEPARABLE_PDF_COMBINERS(hsl_hue)
+MAKE_NON_SEPARABLE_PDF_COMBINERS(hsl_saturation)
+MAKE_NON_SEPARABLE_PDF_COMBINERS(hsl_color)
+MAKE_NON_SEPARABLE_PDF_COMBINERS(hsl_luminosity)
+
+void
+_pixman_setup_combiner_functions_float (pixman_implementation_t *imp)
+{
+    /* Unified alpha */
+    imp->combine_float[PIXMAN_OP_CLEAR] = combine_clear_u_float;
+    imp->combine_float[PIXMAN_OP_SRC] = combine_src_u_float;
+    imp->combine_float[PIXMAN_OP_DST] = combine_dst_u_float;
+    imp->combine_float[PIXMAN_OP_OVER] = combine_over_u_float;
+    imp->combine_float[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_IN] = combine_in_u_float;
+    imp->combine_float[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_OUT] = combine_out_u_float;
+    imp->combine_float[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_ATOP] = combine_atop_u_float;
+    imp->combine_float[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_XOR] = combine_xor_u_float;
+    imp->combine_float[PIXMAN_OP_ADD] = combine_add_u_float;
+    imp->combine_float[PIXMAN_OP_SATURATE] = combine_saturate_u_float;
+
+    /* Disjoint, unified */
+    imp->combine_float[PIXMAN_OP_DISJOINT_CLEAR] = combine_disjoint_clear_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_SRC] = combine_disjoint_src_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_DST] = combine_disjoint_dst_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_disjoint_over_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_u_float;
+
+    /* Conjoint, unified */
+    imp->combine_float[PIXMAN_OP_CONJOINT_CLEAR] = combine_conjoint_clear_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_SRC] = combine_conjoint_src_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_DST] = combine_conjoint_dst_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_u_float;
+
+    /* PDF operators, unified */
+    imp->combine_float[PIXMAN_OP_MULTIPLY] = combine_multiply_u_float;
+    imp->combine_float[PIXMAN_OP_SCREEN] = combine_screen_u_float;
+    imp->combine_float[PIXMAN_OP_OVERLAY] = combine_overlay_u_float;
+    imp->combine_float[PIXMAN_OP_DARKEN] = combine_darken_u_float;
+    imp->combine_float[PIXMAN_OP_LIGHTEN] = combine_lighten_u_float;
+    imp->combine_float[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_u_float;
+    imp->combine_float[PIXMAN_OP_COLOR_BURN] = combine_color_burn_u_float;
+    imp->combine_float[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_u_float;
+    imp->combine_float[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_u_float;
+    imp->combine_float[PIXMAN_OP_DIFFERENCE] = combine_difference_u_float;
+    imp->combine_float[PIXMAN_OP_EXCLUSION] = combine_exclusion_u_float;
+
+    imp->combine_float[PIXMAN_OP_HSL_HUE] = combine_hsl_hue_u_float;
+    imp->combine_float[PIXMAN_OP_HSL_SATURATION] = combine_hsl_saturation_u_float;
+    imp->combine_float[PIXMAN_OP_HSL_COLOR] = combine_hsl_color_u_float;
+    imp->combine_float[PIXMAN_OP_HSL_LUMINOSITY] = combine_hsl_luminosity_u_float;
+
+    /* Component alpha combiners */
+    imp->combine_float_ca[PIXMAN_OP_CLEAR] = combine_clear_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_SRC] = combine_src_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DST] = combine_dst_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OVER] = combine_over_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_IN] = combine_in_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OUT] = combine_out_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_ATOP] = combine_atop_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_XOR] = combine_xor_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_ADD] = combine_add_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_SATURATE] = combine_saturate_ca_float;
+
+    /* Disjoint CA */
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_CLEAR] = combine_disjoint_clear_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_SRC] = combine_disjoint_src_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_DST] = combine_disjoint_dst_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_disjoint_over_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_ca_float;
+
+    /* Conjoint CA */
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_CLEAR] = combine_conjoint_clear_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_SRC] = combine_conjoint_src_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_DST] = combine_conjoint_dst_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_ca_float;
+
+    /* PDF operators CA */
+    imp->combine_float_ca[PIXMAN_OP_MULTIPLY] = combine_multiply_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_SCREEN] = combine_screen_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OVERLAY] = combine_overlay_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DARKEN] = combine_darken_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_LIGHTEN] = combine_lighten_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_COLOR_BURN] = combine_color_burn_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DIFFERENCE] = combine_difference_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_EXCLUSION] = combine_exclusion_ca_float;
+
+    /* It is not clear that these make sense, so make them noops for now */
+    imp->combine_float_ca[PIXMAN_OP_HSL_HUE] = combine_dst_u_float;
+    imp->combine_float_ca[PIXMAN_OP_HSL_SATURATION] = combine_dst_u_float;
+    imp->combine_float_ca[PIXMAN_OP_HSL_COLOR] = combine_dst_u_float;
+    imp->combine_float_ca[PIXMAN_OP_HSL_LUMINOSITY] = combine_dst_u_float;
+}
diff --git a/pixman/pixman/pixman-combine.c.template b/pixman/pixman/pixman-combine32.c
index f405312d4..54cc8771b 100644
--- a/pixman/pixman/pixman-combine.c.template
+++ b/pixman/pixman/pixman-combine32.c
@@ -1,3 +1,26 @@
+/*
+ * Copyright © 2000 Keith Packard, member of The XFree86 Project, Inc.
+ *             2005 Lars Knoll & Zack Rusin, Trolltech
+ *
+ * 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
@@ -6,17 +29,17 @@
 #include <string.h>
 
 #include "pixman-private.h"
-#include "pixman-combine.h"
+#include "pixman-combine32.h"
 
 /* component alpha helper functions */
 
 static void
-combine_mask_ca (comp4_t *src, comp4_t *mask)
+combine_mask_ca (uint32_t *src, uint32_t *mask)
 {
-    comp4_t a = *mask;
+    uint32_t a = *mask;
 
-    comp4_t x;
-    comp2_t xa;
+    uint32_t x;
+    uint16_t xa;
 
     if (!a)
     {
@@ -35,18 +58,18 @@ combine_mask_ca (comp4_t *src, comp4_t *mask)
     }
 
     xa = x >> A_SHIFT;
-    UNcx4_MUL_UNcx4 (x, a);
+    UN8x4_MUL_UN8x4 (x, a);
     *(src) = x;
     
-    UNcx4_MUL_UNc (a, xa);
+    UN8x4_MUL_UN8 (a, xa);
     *(mask) = a;
 }
 
 static void
-combine_mask_value_ca (comp4_t *src, const comp4_t *mask)
+combine_mask_value_ca (uint32_t *src, const uint32_t *mask)
 {
-    comp4_t a = *mask;
-    comp4_t x;
+    uint32_t a = *mask;
+    uint32_t x;
 
     if (!a)
     {
@@ -58,15 +81,15 @@ combine_mask_value_ca (comp4_t *src, const comp4_t *mask)
 	return;
 
     x = *(src);
-    UNcx4_MUL_UNcx4 (x, a);
+    UN8x4_MUL_UN8x4 (x, a);
     *(src) = x;
 }
 
 static void
-combine_mask_alpha_ca (const comp4_t *src, comp4_t *mask)
+combine_mask_alpha_ca (const uint32_t *src, uint32_t *mask)
 {
-    comp4_t a = *(mask);
-    comp4_t x;
+    uint32_t a = *(mask);
+    uint32_t x;
 
     if (!a)
 	return;
@@ -83,7 +106,7 @@ combine_mask_alpha_ca (const comp4_t *src, comp4_t *mask)
 	return;
     }
 
-    UNcx4_MUL_UNc (a, x);
+    UN8x4_MUL_UN8 (a, x);
     *(mask) = a;
 }
 
@@ -95,10 +118,10 @@ combine_mask_alpha_ca (const comp4_t *src, comp4_t *mask)
  * this difference will have two versions using the same convention.
  */
 
-static force_inline comp4_t
-combine_mask (const comp4_t *src, const comp4_t *mask, int i)
+static force_inline uint32_t
+combine_mask (const uint32_t *src, const uint32_t *mask, int i)
 {
-    comp4_t s, m;
+    uint32_t s, m;
 
     if (mask)
     {
@@ -111,7 +134,7 @@ combine_mask (const comp4_t *src, const comp4_t *mask, int i)
     s = *(src + i);
 
     if (mask)
-	UNcx4_MUL_UNc (s, m);
+	UN8x4_MUL_UN8 (s, m);
 
     return s;
 }
@@ -119,20 +142,20 @@ combine_mask (const comp4_t *src, const comp4_t *mask, int i)
 static void
 combine_clear (pixman_implementation_t *imp,
                pixman_op_t              op,
-               comp4_t *                dest,
-               const comp4_t *          src,
-               const comp4_t *          mask,
+               uint32_t *                dest,
+               const uint32_t *          src,
+               const uint32_t *          mask,
                int                      width)
 {
-    memset (dest, 0, width * sizeof(comp4_t));
+    memset (dest, 0, width * sizeof(uint32_t));
 }
 
 static void
 combine_dst (pixman_implementation_t *imp,
 	     pixman_op_t	      op,
-	     comp4_t *		      dest,
-	     const comp4_t *	      src,
-	     const comp4_t *          mask,
+	     uint32_t *		      dest,
+	     const uint32_t *	      src,
+	     const uint32_t *          mask,
 	     int		      width)
 {
     return;
@@ -141,22 +164,22 @@ combine_dst (pixman_implementation_t *imp,
 static void
 combine_src_u (pixman_implementation_t *imp,
                pixman_op_t              op,
-               comp4_t *                dest,
-               const comp4_t *          src,
-               const comp4_t *          mask,
+               uint32_t *                dest,
+               const uint32_t *          src,
+               const uint32_t *          mask,
                int                      width)
 {
     int i;
 
     if (!mask)
     {
-	memcpy (dest, src, width * sizeof (comp4_t));
+	memcpy (dest, src, width * sizeof (uint32_t));
     }
     else
     {
 	for (i = 0; i < width; ++i)
 	{
-	    comp4_t s = combine_mask (src, mask, i);
+	    uint32_t s = combine_mask (src, mask, i);
 
 	    *(dest + i) = s;
 	}
@@ -166,20 +189,20 @@ combine_src_u (pixman_implementation_t *imp,
 static void
 combine_over_u (pixman_implementation_t *imp,
                 pixman_op_t              op,
-                comp4_t *                dest,
-                const comp4_t *          src,
-                const comp4_t *          mask,
+                uint32_t *                dest,
+                const uint32_t *          src,
+                const uint32_t *          mask,
                 int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t ia = ALPHA_c (~s);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t ia = ALPHA_8 (~s);
 
-	UNcx4_MUL_UNc_ADD_UNcx4 (d, ia, s);
+	UN8x4_MUL_UN8_ADD_UN8x4 (d, ia, s);
 	*(dest + i) = d;
     }
 }
@@ -187,19 +210,19 @@ combine_over_u (pixman_implementation_t *imp,
 static void
 combine_over_reverse_u (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t ia = ALPHA_c (~*(dest + i));
-	UNcx4_MUL_UNc_ADD_UNcx4 (s, ia, d);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t ia = ALPHA_8 (~*(dest + i));
+	UN8x4_MUL_UN8_ADD_UN8x4 (s, ia, d);
 	*(dest + i) = s;
     }
 }
@@ -207,18 +230,18 @@ combine_over_reverse_u (pixman_implementation_t *imp,
 static void
 combine_in_u (pixman_implementation_t *imp,
               pixman_op_t              op,
-              comp4_t *                dest,
-              const comp4_t *          src,
-              const comp4_t *          mask,
+              uint32_t *                dest,
+              const uint32_t *          src,
+              const uint32_t *          mask,
               int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t a = ALPHA_c (*(dest + i));
-	UNcx4_MUL_UNc (s, a);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t a = ALPHA_8 (*(dest + i));
+	UN8x4_MUL_UN8 (s, a);
 	*(dest + i) = s;
     }
 }
@@ -226,19 +249,19 @@ combine_in_u (pixman_implementation_t *imp,
 static void
 combine_in_reverse_u (pixman_implementation_t *imp,
                       pixman_op_t              op,
-                      comp4_t *                dest,
-                      const comp4_t *          src,
-                      const comp4_t *          mask,
+                      uint32_t *                dest,
+                      const uint32_t *          src,
+                      const uint32_t *          mask,
                       int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t a = ALPHA_c (s);
-	UNcx4_MUL_UNc (d, a);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t a = ALPHA_8 (s);
+	UN8x4_MUL_UN8 (d, a);
 	*(dest + i) = d;
     }
 }
@@ -246,18 +269,18 @@ combine_in_reverse_u (pixman_implementation_t *imp,
 static void
 combine_out_u (pixman_implementation_t *imp,
                pixman_op_t              op,
-               comp4_t *                dest,
-               const comp4_t *          src,
-               const comp4_t *          mask,
+               uint32_t *                dest,
+               const uint32_t *          src,
+               const uint32_t *          mask,
                int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t a = ALPHA_c (~*(dest + i));
-	UNcx4_MUL_UNc (s, a);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t a = ALPHA_8 (~*(dest + i));
+	UN8x4_MUL_UN8 (s, a);
 	*(dest + i) = s;
     }
 }
@@ -265,19 +288,19 @@ combine_out_u (pixman_implementation_t *imp,
 static void
 combine_out_reverse_u (pixman_implementation_t *imp,
                        pixman_op_t              op,
-                       comp4_t *                dest,
-                       const comp4_t *          src,
-                       const comp4_t *          mask,
+                       uint32_t *                dest,
+                       const uint32_t *          src,
+                       const uint32_t *          mask,
                        int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t a = ALPHA_c (~s);
-	UNcx4_MUL_UNc (d, a);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t a = ALPHA_8 (~s);
+	UN8x4_MUL_UN8 (d, a);
 	*(dest + i) = d;
     }
 }
@@ -285,21 +308,21 @@ combine_out_reverse_u (pixman_implementation_t *imp,
 static void
 combine_atop_u (pixman_implementation_t *imp,
                 pixman_op_t              op,
-                comp4_t *                dest,
-                const comp4_t *          src,
-                const comp4_t *          mask,
+                uint32_t *                dest,
+                const uint32_t *          src,
+                const uint32_t *          mask,
                 int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t dest_a = ALPHA_c (d);
-	comp4_t src_ia = ALPHA_c (~s);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t dest_a = ALPHA_8 (d);
+	uint32_t src_ia = ALPHA_8 (~s);
 
-	UNcx4_MUL_UNc_ADD_UNcx4_MUL_UNc (s, dest_a, d, src_ia);
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_a, d, src_ia);
 	*(dest + i) = s;
     }
 }
@@ -307,21 +330,21 @@ combine_atop_u (pixman_implementation_t *imp,
 static void
 combine_atop_reverse_u (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t src_a = ALPHA_c (s);
-	comp4_t dest_ia = ALPHA_c (~d);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t src_a = ALPHA_8 (s);
+	uint32_t dest_ia = ALPHA_8 (~d);
 
-	UNcx4_MUL_UNc_ADD_UNcx4_MUL_UNc (s, dest_ia, d, src_a);
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_ia, d, src_a);
 	*(dest + i) = s;
     }
 }
@@ -329,21 +352,21 @@ combine_atop_reverse_u (pixman_implementation_t *imp,
 static void
 combine_xor_u (pixman_implementation_t *imp,
                pixman_op_t              op,
-               comp4_t *                dest,
-               const comp4_t *          src,
-               const comp4_t *          mask,
+               uint32_t *                dest,
+               const uint32_t *          src,
+               const uint32_t *          mask,
                int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t src_ia = ALPHA_c (~s);
-	comp4_t dest_ia = ALPHA_c (~d);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t src_ia = ALPHA_8 (~s);
+	uint32_t dest_ia = ALPHA_8 (~d);
 
-	UNcx4_MUL_UNc_ADD_UNcx4_MUL_UNc (s, dest_ia, d, src_ia);
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_ia, d, src_ia);
 	*(dest + i) = s;
     }
 }
@@ -351,18 +374,18 @@ combine_xor_u (pixman_implementation_t *imp,
 static void
 combine_add_u (pixman_implementation_t *imp,
                pixman_op_t              op,
-               comp4_t *                dest,
-               const comp4_t *          src,
-               const comp4_t *          mask,
+               uint32_t *                dest,
+               const uint32_t *          src,
+               const uint32_t *          mask,
                int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	UNcx4_ADD_UNcx4 (d, s);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	UN8x4_ADD_UN8x4 (d, s);
 	*(dest + i) = d;
     }
 }
@@ -370,28 +393,28 @@ combine_add_u (pixman_implementation_t *imp,
 static void
 combine_saturate_u (pixman_implementation_t *imp,
                     pixman_op_t              op,
-                    comp4_t *                dest,
-                    const comp4_t *          src,
-                    const comp4_t *          mask,
+                    uint32_t *                dest,
+                    const uint32_t *          src,
+                    const uint32_t *          mask,
                     int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp2_t sa, da;
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint16_t sa, da;
 
 	sa = s >> A_SHIFT;
 	da = ~d >> A_SHIFT;
 	if (sa > da)
 	{
-	    sa = DIV_UNc (da, sa);
-	    UNcx4_MUL_UNc (s, sa);
+	    sa = DIV_UN8 (da, sa);
+	    UN8x4_MUL_UN8 (s, sa);
 	}
 	;
-	UNcx4_ADD_UNcx4 (d, s);
+	UN8x4_ADD_UN8x4 (d, s);
 	*(dest + i) = d;
     }
 }
@@ -426,24 +449,24 @@ combine_saturate_u (pixman_implementation_t *imp,
 static void
 combine_multiply_u (pixman_implementation_t *imp,
                     pixman_op_t              op,
-                    comp4_t *                dest,
-                    const comp4_t *          src,
-                    const comp4_t *          mask,
+                    uint32_t *                dest,
+                    const uint32_t *          src,
+                    const uint32_t *          mask,
                     int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t ss = s;
-	comp4_t src_ia = ALPHA_c (~s);
-	comp4_t dest_ia = ALPHA_c (~d);
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t ss = s;
+	uint32_t src_ia = ALPHA_8 (~s);
+	uint32_t dest_ia = ALPHA_8 (~d);
 
-	UNcx4_MUL_UNc_ADD_UNcx4_MUL_UNc (ss, dest_ia, d, src_ia);
-	UNcx4_MUL_UNcx4 (d, s);
-	UNcx4_ADD_UNcx4 (d, ss);
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (ss, dest_ia, d, src_ia);
+	UN8x4_MUL_UN8x4 (d, s);
+	UN8x4_ADD_UN8x4 (d, ss);
 
 	*(dest + i) = d;
     }
@@ -452,26 +475,26 @@ combine_multiply_u (pixman_implementation_t *imp,
 static void
 combine_multiply_ca (pixman_implementation_t *imp,
                      pixman_op_t              op,
-                     comp4_t *                dest,
-                     const comp4_t *          src,
-                     const comp4_t *          mask,
+                     uint32_t *                dest,
+                     const uint32_t *          src,
+                     const uint32_t *          mask,
                      int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t m = *(mask + i);
-	comp4_t s = *(src + i);
-	comp4_t d = *(dest + i);
-	comp4_t r = d;
-	comp4_t dest_ia = ALPHA_c (~d);
+	uint32_t m = *(mask + i);
+	uint32_t s = *(src + i);
+	uint32_t d = *(dest + i);
+	uint32_t r = d;
+	uint32_t dest_ia = ALPHA_8 (~d);
 
 	combine_mask_ca (&s, &m);
 
-	UNcx4_MUL_UNcx4_ADD_UNcx4_MUL_UNc (r, ~m, s, dest_ia);
-	UNcx4_MUL_UNcx4 (d, s);
-	UNcx4_ADD_UNcx4 (r, d);
+	UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (r, ~m, s, dest_ia);
+	UN8x4_MUL_UN8x4 (d, s);
+	UN8x4_ADD_UN8x4 (r, d);
 
 	*(dest + i) = r;
     }
@@ -481,59 +504,59 @@ combine_multiply_ca (pixman_implementation_t *imp,
     static void								\
     combine_ ## name ## _u (pixman_implementation_t *imp,		\
 			    pixman_op_t              op,		\
-                            comp4_t *                dest,		\
-			    const comp4_t *          src,		\
-			    const comp4_t *          mask,		\
+                            uint32_t *                dest,		\
+			    const uint32_t *          src,		\
+			    const uint32_t *          mask,		\
 			    int                      width)		\
     {									\
 	int i;								\
 	for (i = 0; i < width; ++i) {					\
-	    comp4_t s = combine_mask (src, mask, i);			\
-	    comp4_t d = *(dest + i);					\
-	    comp1_t sa = ALPHA_c (s);					\
-	    comp1_t isa = ~sa;						\
-	    comp1_t da = ALPHA_c (d);					\
-	    comp1_t ida = ~da;						\
-	    comp4_t result;						\
+	    uint32_t s = combine_mask (src, mask, i);			\
+	    uint32_t d = *(dest + i);					\
+	    uint8_t sa = ALPHA_8 (s);					\
+	    uint8_t isa = ~sa;						\
+	    uint8_t da = ALPHA_8 (d);					\
+	    uint8_t ida = ~da;						\
+	    uint32_t result;						\
 									\
 	    result = d;							\
-	    UNcx4_MUL_UNc_ADD_UNcx4_MUL_UNc (result, isa, s, ida);	\
+	    UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (result, isa, s, ida);	\
 	    								\
 	    *(dest + i) = result +					\
-		(DIV_ONE_UNc (sa * (comp4_t)da) << A_SHIFT) +		\
-		(blend_ ## name (RED_c (d), da, RED_c (s), sa) << R_SHIFT) + \
-		(blend_ ## name (GREEN_c (d), da, GREEN_c (s), sa) << G_SHIFT) + \
-		(blend_ ## name (BLUE_c (d), da, BLUE_c (s), sa));	\
+		(DIV_ONE_UN8 (sa * (uint32_t)da) << A_SHIFT) +		\
+		(blend_ ## name (RED_8 (d), da, RED_8 (s), sa) << R_SHIFT) + \
+		(blend_ ## name (GREEN_8 (d), da, GREEN_8 (s), sa) << G_SHIFT) + \
+		(blend_ ## name (BLUE_8 (d), da, BLUE_8 (s), sa));	\
 	}								\
     }									\
     									\
     static void								\
     combine_ ## name ## _ca (pixman_implementation_t *imp,		\
 			     pixman_op_t              op,		\
-                             comp4_t *                dest,		\
-			     const comp4_t *          src,		\
-			     const comp4_t *          mask,		\
+                             uint32_t *                dest,		\
+			     const uint32_t *          src,		\
+			     const uint32_t *          mask,		\
 			     int                     width)		\
     {									\
 	int i;								\
 	for (i = 0; i < width; ++i) {					\
-	    comp4_t m = *(mask + i);					\
-	    comp4_t s = *(src + i);					\
-	    comp4_t d = *(dest + i);					\
-	    comp1_t da = ALPHA_c (d);					\
-	    comp1_t ida = ~da;						\
-	    comp4_t result;						\
+	    uint32_t m = *(mask + i);					\
+	    uint32_t s = *(src + i);					\
+	    uint32_t d = *(dest + i);					\
+	    uint8_t da = ALPHA_8 (d);					\
+	    uint8_t ida = ~da;						\
+	    uint32_t result;						\
             								\
 	    combine_mask_ca (&s, &m);					\
             								\
 	    result = d;							\
-	    UNcx4_MUL_UNcx4_ADD_UNcx4_MUL_UNc (result, ~m, s, ida);     \
+	    UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (result, ~m, s, ida);     \
             								\
 	    result +=							\
-	        (DIV_ONE_UNc (ALPHA_c (m) * (comp4_t)da) << A_SHIFT) +	\
-	        (blend_ ## name (RED_c (d), da, RED_c (s), RED_c (m)) << R_SHIFT) + \
-	        (blend_ ## name (GREEN_c (d), da, GREEN_c (s), GREEN_c (m)) << G_SHIFT) + \
-	        (blend_ ## name (BLUE_c (d), da, BLUE_c (s), BLUE_c (m))); \
+	        (DIV_ONE_UN8 (ALPHA_8 (m) * (uint32_t)da) << A_SHIFT) +	\
+	        (blend_ ## name (RED_8 (d), da, RED_8 (s), RED_8 (m)) << R_SHIFT) + \
+	        (blend_ ## name (GREEN_8 (d), da, GREEN_8 (s), GREEN_8 (m)) << G_SHIFT) + \
+	        (blend_ ## name (BLUE_8 (d), da, BLUE_8 (s), BLUE_8 (m))); \
 	    								\
 	    *(dest + i) = result;					\
 	}								\
@@ -543,10 +566,10 @@ combine_multiply_ca (pixman_implementation_t *imp,
  * Screen
  * B(Dca, ad, Sca, as) = Dca.sa + Sca.da - Dca.Sca
  */
-static inline comp4_t
-blend_screen (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_screen (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
-    return DIV_ONE_UNc (sca * da + dca * sa - sca * dca);
+    return DIV_ONE_UN8 (sca * da + dca * sa - sca * dca);
 }
 
 PDF_SEPARABLE_BLEND_MODE (screen)
@@ -559,16 +582,16 @@ PDF_SEPARABLE_BLEND_MODE (screen)
  *   otherwise
  *     Sa.Da - 2.(Da - Dca).(Sa - Sca)
  */
-static inline comp4_t
-blend_overlay (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_overlay (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
-    comp4_t rca;
+    uint32_t rca;
 
     if (2 * dca < da)
 	rca = 2 * sca * dca;
     else
 	rca = sa * da - 2 * (da - dca) * (sa - sca);
-    return DIV_ONE_UNc (rca);
+    return DIV_ONE_UN8 (rca);
 }
 
 PDF_SEPARABLE_BLEND_MODE (overlay)
@@ -577,14 +600,14 @@ PDF_SEPARABLE_BLEND_MODE (overlay)
  * Darken
  * B(Dca, Da, Sca, Sa) = min (Sca.Da, Dca.Sa)
  */
-static inline comp4_t
-blend_darken (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_darken (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
-    comp4_t s, d;
+    uint32_t s, d;
 
     s = sca * da;
     d = dca * sa;
-    return DIV_ONE_UNc (s > d ? d : s);
+    return DIV_ONE_UN8 (s > d ? d : s);
 }
 
 PDF_SEPARABLE_BLEND_MODE (darken)
@@ -593,14 +616,14 @@ PDF_SEPARABLE_BLEND_MODE (darken)
  * Lighten
  * B(Dca, Da, Sca, Sa) = max (Sca.Da, Dca.Sa)
  */
-static inline comp4_t
-blend_lighten (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_lighten (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
-    comp4_t s, d;
+    uint32_t s, d;
 
     s = sca * da;
     d = dca * sa;
-    return DIV_ONE_UNc (s > d ? s : d);
+    return DIV_ONE_UN8 (s > d ? s : d);
 }
 
 PDF_SEPARABLE_BLEND_MODE (lighten)
@@ -615,17 +638,17 @@ PDF_SEPARABLE_BLEND_MODE (lighten)
  *   otherwise
  *     Sa.Da. min (1, Dca / Da / (1 - Sca/Sa))
  */
-static inline comp4_t
-blend_color_dodge (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_color_dodge (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
     if (sca >= sa)
     {
-	return dca == 0 ? 0 : DIV_ONE_UNc (sa * da);
+	return dca == 0 ? 0 : DIV_ONE_UN8 (sa * da);
     }
     else
     {
-	comp4_t rca = dca * sa / (sa - sca);
-	return DIV_ONE_UNc (sa * MIN (rca, da));
+	uint32_t rca = dca * sa / (sa - sca);
+	return DIV_ONE_UN8 (sa * MIN (rca, da));
     }
 }
 
@@ -641,17 +664,17 @@ PDF_SEPARABLE_BLEND_MODE (color_dodge)
  *   otherwise
  *     Sa.Da.(1 - min (1, (1 - Dca/Da).Sa / Sca))
  */
-static inline comp4_t
-blend_color_burn (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_color_burn (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
     if (sca == 0)
     {
-	return dca < da ? 0 : DIV_ONE_UNc (sa * da);
+	return dca < da ? 0 : DIV_ONE_UN8 (sa * da);
     }
     else
     {
-	comp4_t rca = (da - dca) * sa / sca;
-	return DIV_ONE_UNc (sa * (MAX (rca, da) - rca));
+	uint32_t rca = (da - dca) * sa / sca;
+	return DIV_ONE_UN8 (sa * (MAX (rca, da) - rca));
     }
 }
 
@@ -665,13 +688,13 @@ PDF_SEPARABLE_BLEND_MODE (color_burn)
  *   otherwise
  *     Sa.Da - 2.(Da - Dca).(Sa - Sca)
  */
-static inline comp4_t
-blend_hard_light (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_hard_light (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
     if (2 * sca < sa)
-	return DIV_ONE_UNc (2 * sca * dca);
+	return DIV_ONE_UN8 (2 * sca * dca);
     else
-	return DIV_ONE_UNc (sa * da - 2 * (da - dca) * (sa - sca));
+	return DIV_ONE_UN8 (sa * da - 2 * (da - dca) * (sa - sca));
 }
 
 PDF_SEPARABLE_BLEND_MODE (hard_light)
@@ -686,11 +709,11 @@ PDF_SEPARABLE_BLEND_MODE (hard_light)
  *   otherwise
  *     (Dca.Sa + (SQRT (Dca/Da).Da - Dca).(2.Sca - Sa))
  */
-static inline comp4_t
-blend_soft_light (comp4_t dca_org,
-		  comp4_t da_org,
-		  comp4_t sca_org,
-		  comp4_t sa_org)
+static inline uint32_t
+blend_soft_light (uint32_t dca_org,
+		  uint32_t da_org,
+		  uint32_t sca_org,
+		  uint32_t sa_org)
 {
     double dca = dca_org * (1.0 / MASK);
     double da = da_org * (1.0 / MASK);
@@ -727,16 +750,16 @@ PDF_SEPARABLE_BLEND_MODE (soft_light)
  * Difference
  * B(Dca, Da, Sca, Sa) = abs (Dca.Sa - Sca.Da)
  */
-static inline comp4_t
-blend_difference (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_difference (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
-    comp4_t dcasa = dca * sa;
-    comp4_t scada = sca * da;
+    uint32_t dcasa = dca * sa;
+    uint32_t scada = sca * da;
 
     if (scada < dcasa)
-	return DIV_ONE_UNc (dcasa - scada);
+	return DIV_ONE_UN8 (dcasa - scada);
     else
-	return DIV_ONE_UNc (scada - dcasa);
+	return DIV_ONE_UN8 (scada - dcasa);
 }
 
 PDF_SEPARABLE_BLEND_MODE (difference)
@@ -749,10 +772,10 @@ PDF_SEPARABLE_BLEND_MODE (difference)
 /* This can be made faster by writing it directly and not using
  * PDF_SEPARABLE_BLEND_MODE, but that's a performance optimization */
 
-static inline comp4_t
-blend_exclusion (comp4_t dca, comp4_t da, comp4_t sca, comp4_t sa)
+static inline uint32_t
+blend_exclusion (uint32_t dca, uint32_t da, uint32_t sca, uint32_t sa)
 {
-    return DIV_ONE_UNc (sca * da + dca * sa - 2 * dca * sca);
+    return DIV_ONE_UN8 (sca * da + dca * sa - 2 * dca * sca);
 }
 
 PDF_SEPARABLE_BLEND_MODE (exclusion)
@@ -876,43 +899,43 @@ PDF_SEPARABLE_BLEND_MODE (exclusion)
     static void								\
     combine_ ## name ## _u (pixman_implementation_t *imp,		\
 			    pixman_op_t op,				\
-                            comp4_t *dest,				\
-			    const comp4_t *src,				\
-			    const comp4_t *mask,			\
+                            uint32_t *dest,				\
+			    const uint32_t *src,				\
+			    const uint32_t *mask,			\
 			    int width)					\
     {									\
 	int i;								\
 	for (i = 0; i < width; ++i)					\
 	{								\
-	    comp4_t s = combine_mask (src, mask, i);			\
-	    comp4_t d = *(dest + i);					\
-	    comp1_t sa = ALPHA_c (s);					\
-	    comp1_t isa = ~sa;						\
-	    comp1_t da = ALPHA_c (d);					\
-	    comp1_t ida = ~da;						\
-	    comp4_t result;						\
-	    comp4_t sc[3], dc[3], c[3];					\
+	    uint32_t s = combine_mask (src, mask, i);			\
+	    uint32_t d = *(dest + i);					\
+	    uint8_t sa = ALPHA_8 (s);					\
+	    uint8_t isa = ~sa;						\
+	    uint8_t da = ALPHA_8 (d);					\
+	    uint8_t ida = ~da;						\
+	    uint32_t result;						\
+	    uint32_t sc[3], dc[3], c[3];					\
             								\
 	    result = d;							\
-	    UNcx4_MUL_UNc_ADD_UNcx4_MUL_UNc (result, isa, s, ida);	\
-	    dc[0] = RED_c (d);						\
-	    sc[0] = RED_c (s);						\
-	    dc[1] = GREEN_c (d);					\
-	    sc[1] = GREEN_c (s);					\
-	    dc[2] = BLUE_c (d);						\
-	    sc[2] = BLUE_c (s);						\
+	    UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (result, isa, s, ida);	\
+	    dc[0] = RED_8 (d);						\
+	    sc[0] = RED_8 (s);						\
+	    dc[1] = GREEN_8 (d);					\
+	    sc[1] = GREEN_8 (s);					\
+	    dc[2] = BLUE_8 (d);						\
+	    sc[2] = BLUE_8 (s);						\
 	    blend_ ## name (c, dc, da, sc, sa);				\
             								\
 	    *(dest + i) = result +					\
-		(DIV_ONE_UNc (sa * (comp4_t)da) << A_SHIFT) +		\
-		(DIV_ONE_UNc (c[0]) << R_SHIFT) +			\
-		(DIV_ONE_UNc (c[1]) << G_SHIFT) +			\
-		(DIV_ONE_UNc (c[2]));					\
+		(DIV_ONE_UN8 (sa * (uint32_t)da) << A_SHIFT) +		\
+		(DIV_ONE_UN8 (c[0]) << R_SHIFT) +			\
+		(DIV_ONE_UN8 (c[1]) << G_SHIFT) +			\
+		(DIV_ONE_UN8 (c[2]));					\
 	}								\
     }
 
 static void
-set_lum (comp4_t dest[3], comp4_t src[3], comp4_t sa, comp4_t lum)
+set_lum (uint32_t dest[3], uint32_t src[3], uint32_t sa, uint32_t lum)
 {
     double a, l, min, max;
     double tmp[3];
@@ -971,10 +994,10 @@ set_lum (comp4_t dest[3], comp4_t src[3], comp4_t sa, comp4_t lum)
 }
 
 static void
-set_sat (comp4_t dest[3], comp4_t src[3], comp4_t sat)
+set_sat (uint32_t dest[3], uint32_t src[3], uint32_t sat)
 {
     int id[3];
-    comp4_t min, max;
+    uint32_t min, max;
 
     if (src[0] > src[1])
     {
@@ -1042,11 +1065,11 @@ set_sat (comp4_t dest[3], comp4_t src[3], comp4_t sat)
  * B(Cb, Cs) = set_lum (set_sat (Cs, SAT (Cb)), LUM (Cb))
  */
 static inline void
-blend_hsl_hue (comp4_t c[3],
-               comp4_t dc[3],
-               comp4_t da,
-               comp4_t sc[3],
-               comp4_t sa)
+blend_hsl_hue (uint32_t c[3],
+               uint32_t dc[3],
+               uint32_t da,
+               uint32_t sc[3],
+               uint32_t sa)
 {
     c[0] = sc[0] * da;
     c[1] = sc[1] * da;
@@ -1062,11 +1085,11 @@ PDF_NON_SEPARABLE_BLEND_MODE (hsl_hue)
  * B(Cb, Cs) = set_lum (set_sat (Cb, SAT (Cs)), LUM (Cb))
  */
 static inline void
-blend_hsl_saturation (comp4_t c[3],
-                      comp4_t dc[3],
-                      comp4_t da,
-                      comp4_t sc[3],
-                      comp4_t sa)
+blend_hsl_saturation (uint32_t c[3],
+                      uint32_t dc[3],
+                      uint32_t da,
+                      uint32_t sc[3],
+                      uint32_t sa)
 {
     c[0] = dc[0] * sa;
     c[1] = dc[1] * sa;
@@ -1082,11 +1105,11 @@ PDF_NON_SEPARABLE_BLEND_MODE (hsl_saturation)
  * B(Cb, Cs) = set_lum (Cs, LUM (Cb))
  */
 static inline void
-blend_hsl_color (comp4_t c[3],
-                 comp4_t dc[3],
-                 comp4_t da,
-                 comp4_t sc[3],
-                 comp4_t sa)
+blend_hsl_color (uint32_t c[3],
+                 uint32_t dc[3],
+                 uint32_t da,
+                 uint32_t sc[3],
+                 uint32_t sa)
 {
     c[0] = sc[0] * da;
     c[1] = sc[1] * da;
@@ -1101,11 +1124,11 @@ PDF_NON_SEPARABLE_BLEND_MODE (hsl_color)
  * B(Cb, Cs) = set_lum (Cb, LUM (Cs))
  */
 static inline void
-blend_hsl_luminosity (comp4_t c[3],
-                      comp4_t dc[3],
-                      comp4_t da,
-                      comp4_t sc[3],
-                      comp4_t sa)
+blend_hsl_luminosity (uint32_t c[3],
+                      uint32_t dc[3],
+                      uint32_t da,
+                      uint32_t sc[3],
+                      uint32_t sa)
 {
     c[0] = dc[0] * sa;
     c[1] = dc[1] * sa;
@@ -1162,20 +1185,20 @@ PDF_NON_SEPARABLE_BLEND_MODE (hsl_luminosity)
 #define COMBINE_XOR     (COMBINE_A_OUT | COMBINE_B_OUT)
 
 /* portion covered by a but not b */
-static comp1_t
-combine_disjoint_out_part (comp1_t a, comp1_t b)
+static uint8_t
+combine_disjoint_out_part (uint8_t a, uint8_t b)
 {
     /* min (1, (1-b) / a) */
 
     b = ~b;                 /* 1 - b */
     if (b >= a)             /* 1 - b >= a -> (1-b)/a >= 1 */
 	return MASK;        /* 1 */
-    return DIV_UNc (b, a);     /* (1-b) / a */
+    return DIV_UN8 (b, a);     /* (1-b) / a */
 }
 
 /* portion covered by both a and b */
-static comp1_t
-combine_disjoint_in_part (comp1_t a, comp1_t b)
+static uint8_t
+combine_disjoint_in_part (uint8_t a, uint8_t b)
 {
     /* max (1-(1-b)/a,0) */
     /*  = - min ((1-b)/a - 1, 0) */
@@ -1184,12 +1207,12 @@ combine_disjoint_in_part (comp1_t a, comp1_t b)
     b = ~b;                 /* 1 - b */
     if (b >= a)             /* 1 - b >= a -> (1-b)/a >= 1 */
 	return 0;           /* 1 - 1 */
-    return ~DIV_UNc(b, a);    /* 1 - (1-b) / a */
+    return ~DIV_UN8(b, a);    /* 1 - (1-b) / a */
 }
 
 /* portion covered by a but not b */
-static comp1_t
-combine_conjoint_out_part (comp1_t a, comp1_t b)
+static uint8_t
+combine_conjoint_out_part (uint8_t a, uint8_t b)
 {
     /* max (1-b/a,0) */
     /* = 1-min(b/a,1) */
@@ -1198,49 +1221,49 @@ combine_conjoint_out_part (comp1_t a, comp1_t b)
 
     if (b >= a)             /* b >= a -> b/a >= 1 */
 	return 0x00;        /* 0 */
-    return ~DIV_UNc(b, a);    /* 1 - b/a */
+    return ~DIV_UN8(b, a);    /* 1 - b/a */
 }
 
 /* portion covered by both a and b */
-static comp1_t
-combine_conjoint_in_part (comp1_t a, comp1_t b)
+static uint8_t
+combine_conjoint_in_part (uint8_t a, uint8_t b)
 {
     /* min (1,b/a) */
 
     if (b >= a)             /* b >= a -> b/a >= 1 */
 	return MASK;        /* 1 */
-    return DIV_UNc (b, a);     /* b/a */
+    return DIV_UN8 (b, a);     /* b/a */
 }
 
-#define GET_COMP(v, i)   ((comp2_t) (comp1_t) ((v) >> i))
+#define GET_COMP(v, i)   ((uint16_t) (uint8_t) ((v) >> i))
 
 #define ADD(x, y, i, t)							\
     ((t) = GET_COMP (x, i) + GET_COMP (y, i),				\
-     (comp4_t) ((comp1_t) ((t) | (0 - ((t) >> G_SHIFT)))) << (i))
+     (uint32_t) ((uint8_t) ((t) | (0 - ((t) >> G_SHIFT)))) << (i))
 
 #define GENERIC(x, y, i, ax, ay, t, u, v)				\
-    ((t) = (MUL_UNc (GET_COMP (y, i), ay, (u)) +			\
-            MUL_UNc (GET_COMP (x, i), ax, (v))),			\
-     (comp4_t) ((comp1_t) ((t) |					\
+    ((t) = (MUL_UN8 (GET_COMP (y, i), ay, (u)) +			\
+            MUL_UN8 (GET_COMP (x, i), ax, (v))),			\
+     (uint32_t) ((uint8_t) ((t) |					\
                            (0 - ((t) >> G_SHIFT)))) << (i))
 
 static void
-combine_disjoint_general_u (comp4_t *      dest,
-                            const comp4_t *src,
-                            const comp4_t *mask,
+combine_disjoint_general_u (uint32_t *      dest,
+                            const uint32_t *src,
+                            const uint32_t *mask,
                             int            width,
-                            comp1_t        combine)
+                            uint8_t        combine)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t m, n, o, p;
-	comp2_t Fa, Fb, t, u, v;
-	comp1_t sa = s >> A_SHIFT;
-	comp1_t da = d >> A_SHIFT;
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t m, n, o, p;
+	uint16_t Fa, Fb, t, u, v;
+	uint8_t sa = s >> A_SHIFT;
+	uint8_t da = d >> A_SHIFT;
 
 	switch (combine & COMBINE_A)
 	{
@@ -1291,23 +1314,23 @@ combine_disjoint_general_u (comp4_t *      dest,
 static void
 combine_disjoint_over_u (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp2_t a = s >> A_SHIFT;
+	uint32_t s = combine_mask (src, mask, i);
+	uint16_t a = s >> A_SHIFT;
 
 	if (s != 0x00)
 	{
-	    comp4_t d = *(dest + i);
+	    uint32_t d = *(dest + i);
 	    a = combine_disjoint_out_part (d >> A_SHIFT, a);
-	    UNcx4_MUL_UNc_ADD_UNcx4 (d, a, s);
+	    UN8x4_MUL_UN8_ADD_UN8x4 (d, a, s);
 
 	    *(dest + i) = d;
 	}
@@ -1317,9 +1340,9 @@ combine_disjoint_over_u (pixman_implementation_t *imp,
 static void
 combine_disjoint_in_u (pixman_implementation_t *imp,
                        pixman_op_t              op,
-                       comp4_t *                dest,
-                       const comp4_t *          src,
-                       const comp4_t *          mask,
+                       uint32_t *                dest,
+                       const uint32_t *          src,
+                       const uint32_t *          mask,
                        int                      width)
 {
     combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_IN);
@@ -1328,9 +1351,9 @@ combine_disjoint_in_u (pixman_implementation_t *imp,
 static void
 combine_disjoint_in_reverse_u (pixman_implementation_t *imp,
                                pixman_op_t              op,
-                               comp4_t *                dest,
-                               const comp4_t *          src,
-                               const comp4_t *          mask,
+                               uint32_t *                dest,
+                               const uint32_t *          src,
+                               const uint32_t *          mask,
                                int                      width)
 {
     combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_IN);
@@ -1339,9 +1362,9 @@ combine_disjoint_in_reverse_u (pixman_implementation_t *imp,
 static void
 combine_disjoint_out_u (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_OUT);
@@ -1350,9 +1373,9 @@ combine_disjoint_out_u (pixman_implementation_t *imp,
 static void
 combine_disjoint_out_reverse_u (pixman_implementation_t *imp,
                                 pixman_op_t              op,
-                                comp4_t *                dest,
-                                const comp4_t *          src,
-                                const comp4_t *          mask,
+                                uint32_t *                dest,
+                                const uint32_t *          src,
+                                const uint32_t *          mask,
                                 int                      width)
 {
     combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_OUT);
@@ -1361,9 +1384,9 @@ combine_disjoint_out_reverse_u (pixman_implementation_t *imp,
 static void
 combine_disjoint_atop_u (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_ATOP);
@@ -1372,9 +1395,9 @@ combine_disjoint_atop_u (pixman_implementation_t *imp,
 static void
 combine_disjoint_atop_reverse_u (pixman_implementation_t *imp,
                                  pixman_op_t              op,
-                                 comp4_t *                dest,
-                                 const comp4_t *          src,
-                                 const comp4_t *          mask,
+                                 uint32_t *                dest,
+                                 const uint32_t *          src,
+                                 const uint32_t *          mask,
                                  int                      width)
 {
     combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_ATOP);
@@ -1383,31 +1406,31 @@ combine_disjoint_atop_reverse_u (pixman_implementation_t *imp,
 static void
 combine_disjoint_xor_u (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     combine_disjoint_general_u (dest, src, mask, width, COMBINE_XOR);
 }
 
 static void
-combine_conjoint_general_u (comp4_t *      dest,
-                            const comp4_t *src,
-                            const comp4_t *mask,
+combine_conjoint_general_u (uint32_t *      dest,
+                            const uint32_t *src,
+                            const uint32_t *mask,
                             int            width,
-                            comp1_t        combine)
+                            uint8_t        combine)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = combine_mask (src, mask, i);
-	comp4_t d = *(dest + i);
-	comp4_t m, n, o, p;
-	comp2_t Fa, Fb, t, u, v;
-	comp1_t sa = s >> A_SHIFT;
-	comp1_t da = d >> A_SHIFT;
+	uint32_t s = combine_mask (src, mask, i);
+	uint32_t d = *(dest + i);
+	uint32_t m, n, o, p;
+	uint16_t Fa, Fb, t, u, v;
+	uint8_t sa = s >> A_SHIFT;
+	uint8_t da = d >> A_SHIFT;
 
 	switch (combine & COMBINE_A)
 	{
@@ -1461,9 +1484,9 @@ combine_conjoint_general_u (comp4_t *      dest,
 static void
 combine_conjoint_over_u (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_OVER);
@@ -1472,9 +1495,9 @@ combine_conjoint_over_u (pixman_implementation_t *imp,
 static void
 combine_conjoint_over_reverse_u (pixman_implementation_t *imp,
                                  pixman_op_t              op,
-                                 comp4_t *                dest,
-                                 const comp4_t *          src,
-                                 const comp4_t *          mask,
+                                 uint32_t *                dest,
+                                 const uint32_t *          src,
+                                 const uint32_t *          mask,
                                  int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_OVER);
@@ -1483,9 +1506,9 @@ combine_conjoint_over_reverse_u (pixman_implementation_t *imp,
 static void
 combine_conjoint_in_u (pixman_implementation_t *imp,
                        pixman_op_t              op,
-                       comp4_t *                dest,
-                       const comp4_t *          src,
-                       const comp4_t *          mask,
+                       uint32_t *                dest,
+                       const uint32_t *          src,
+                       const uint32_t *          mask,
                        int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_IN);
@@ -1494,9 +1517,9 @@ combine_conjoint_in_u (pixman_implementation_t *imp,
 static void
 combine_conjoint_in_reverse_u (pixman_implementation_t *imp,
                                pixman_op_t              op,
-                               comp4_t *                dest,
-                               const comp4_t *          src,
-                               const comp4_t *          mask,
+                               uint32_t *                dest,
+                               const uint32_t *          src,
+                               const uint32_t *          mask,
                                int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_IN);
@@ -1505,9 +1528,9 @@ combine_conjoint_in_reverse_u (pixman_implementation_t *imp,
 static void
 combine_conjoint_out_u (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_OUT);
@@ -1516,9 +1539,9 @@ combine_conjoint_out_u (pixman_implementation_t *imp,
 static void
 combine_conjoint_out_reverse_u (pixman_implementation_t *imp,
                                 pixman_op_t              op,
-                                comp4_t *                dest,
-                                const comp4_t *          src,
-                                const comp4_t *          mask,
+                                uint32_t *                dest,
+                                const uint32_t *          src,
+                                const uint32_t *          mask,
                                 int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_OUT);
@@ -1527,9 +1550,9 @@ combine_conjoint_out_reverse_u (pixman_implementation_t *imp,
 static void
 combine_conjoint_atop_u (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_ATOP);
@@ -1538,9 +1561,9 @@ combine_conjoint_atop_u (pixman_implementation_t *imp,
 static void
 combine_conjoint_atop_reverse_u (pixman_implementation_t *imp,
                                  pixman_op_t              op,
-                                 comp4_t *                dest,
-                                 const comp4_t *          src,
-                                 const comp4_t *          mask,
+                                 uint32_t *                dest,
+                                 const uint32_t *          src,
+                                 const uint32_t *          mask,
                                  int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_ATOP);
@@ -1549,9 +1572,9 @@ combine_conjoint_atop_reverse_u (pixman_implementation_t *imp,
 static void
 combine_conjoint_xor_u (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     combine_conjoint_general_u (dest, src, mask, width, COMBINE_XOR);
@@ -1563,28 +1586,28 @@ combine_conjoint_xor_u (pixman_implementation_t *imp,
 static void
 combine_clear_ca (pixman_implementation_t *imp,
                   pixman_op_t              op,
-                  comp4_t *                dest,
-                  const comp4_t *          src,
-                  const comp4_t *          mask,
+                  uint32_t *                dest,
+                  const uint32_t *          src,
+                  const uint32_t *          mask,
                   int                      width)
 {
-    memset (dest, 0, width * sizeof(comp4_t));
+    memset (dest, 0, width * sizeof(uint32_t));
 }
 
 static void
 combine_src_ca (pixman_implementation_t *imp,
                 pixman_op_t              op,
-                comp4_t *                dest,
-                const comp4_t *          src,
-                const comp4_t *          mask,
+                uint32_t *                dest,
+                const uint32_t *          src,
+                const uint32_t *          mask,
                 int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = *(src + i);
-	comp4_t m = *(mask + i);
+	uint32_t s = *(src + i);
+	uint32_t m = *(mask + i);
 
 	combine_mask_value_ca (&s, &m);
 
@@ -1595,26 +1618,26 @@ combine_src_ca (pixman_implementation_t *imp,
 static void
 combine_over_ca (pixman_implementation_t *imp,
                  pixman_op_t              op,
-                 comp4_t *                dest,
-                 const comp4_t *          src,
-                 const comp4_t *          mask,
+                 uint32_t *                dest,
+                 const uint32_t *          src,
+                 const uint32_t *          mask,
                  int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = *(src + i);
-	comp4_t m = *(mask + i);
-	comp4_t a;
+	uint32_t s = *(src + i);
+	uint32_t m = *(mask + i);
+	uint32_t a;
 
 	combine_mask_ca (&s, &m);
 
 	a = ~m;
 	if (a)
 	{
-	    comp4_t d = *(dest + i);
-	    UNcx4_MUL_UNcx4_ADD_UNcx4 (d, a, s);
+	    uint32_t d = *(dest + i);
+	    UN8x4_MUL_UN8x4_ADD_UN8x4 (d, a, s);
 	    s = d;
 	}
 
@@ -1625,25 +1648,25 @@ combine_over_ca (pixman_implementation_t *imp,
 static void
 combine_over_reverse_ca (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t d = *(dest + i);
-	comp4_t a = ~d >> A_SHIFT;
+	uint32_t d = *(dest + i);
+	uint32_t a = ~d >> A_SHIFT;
 
 	if (a)
 	{
-	    comp4_t s = *(src + i);
-	    comp4_t m = *(mask + i);
+	    uint32_t s = *(src + i);
+	    uint32_t m = *(mask + i);
 
-	    UNcx4_MUL_UNcx4 (s, m);
-	    UNcx4_MUL_UNc_ADD_UNcx4 (s, a, d);
+	    UN8x4_MUL_UN8x4 (s, m);
+	    UN8x4_MUL_UN8_ADD_UN8x4 (s, a, d);
 
 	    *(dest + i) = s;
 	}
@@ -1653,28 +1676,28 @@ combine_over_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_in_ca (pixman_implementation_t *imp,
                pixman_op_t              op,
-               comp4_t *                dest,
-               const comp4_t *          src,
-               const comp4_t *          mask,
+               uint32_t *                dest,
+               const uint32_t *          src,
+               const uint32_t *          mask,
                int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t d = *(dest + i);
-	comp2_t a = d >> A_SHIFT;
-	comp4_t s = 0;
+	uint32_t d = *(dest + i);
+	uint16_t a = d >> A_SHIFT;
+	uint32_t s = 0;
 
 	if (a)
 	{
-	    comp4_t m = *(mask + i);
+	    uint32_t m = *(mask + i);
 
 	    s = *(src + i);
 	    combine_mask_value_ca (&s, &m);
 
 	    if (a != MASK)
-		UNcx4_MUL_UNc (s, a);
+		UN8x4_MUL_UN8 (s, a);
 	}
 
 	*(dest + i) = s;
@@ -1684,30 +1707,30 @@ combine_in_ca (pixman_implementation_t *imp,
 static void
 combine_in_reverse_ca (pixman_implementation_t *imp,
                        pixman_op_t              op,
-                       comp4_t *                dest,
-                       const comp4_t *          src,
-                       const comp4_t *          mask,
+                       uint32_t *                dest,
+                       const uint32_t *          src,
+                       const uint32_t *          mask,
                        int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = *(src + i);
-	comp4_t m = *(mask + i);
-	comp4_t a;
+	uint32_t s = *(src + i);
+	uint32_t m = *(mask + i);
+	uint32_t a;
 
 	combine_mask_alpha_ca (&s, &m);
 
 	a = m;
 	if (a != ~0)
 	{
-	    comp4_t d = 0;
+	    uint32_t d = 0;
 
 	    if (a)
 	    {
 		d = *(dest + i);
-		UNcx4_MUL_UNcx4 (d, a);
+		UN8x4_MUL_UN8x4 (d, a);
 	    }
 
 	    *(dest + i) = d;
@@ -1718,28 +1741,28 @@ combine_in_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_out_ca (pixman_implementation_t *imp,
                 pixman_op_t              op,
-                comp4_t *                dest,
-                const comp4_t *          src,
-                const comp4_t *          mask,
+                uint32_t *                dest,
+                const uint32_t *          src,
+                const uint32_t *          mask,
                 int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t d = *(dest + i);
-	comp2_t a = ~d >> A_SHIFT;
-	comp4_t s = 0;
+	uint32_t d = *(dest + i);
+	uint16_t a = ~d >> A_SHIFT;
+	uint32_t s = 0;
 
 	if (a)
 	{
-	    comp4_t m = *(mask + i);
+	    uint32_t m = *(mask + i);
 
 	    s = *(src + i);
 	    combine_mask_value_ca (&s, &m);
 
 	    if (a != MASK)
-		UNcx4_MUL_UNc (s, a);
+		UN8x4_MUL_UN8 (s, a);
 	}
 
 	*(dest + i) = s;
@@ -1749,30 +1772,30 @@ combine_out_ca (pixman_implementation_t *imp,
 static void
 combine_out_reverse_ca (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = *(src + i);
-	comp4_t m = *(mask + i);
-	comp4_t a;
+	uint32_t s = *(src + i);
+	uint32_t m = *(mask + i);
+	uint32_t a;
 
 	combine_mask_alpha_ca (&s, &m);
 
 	a = ~m;
 	if (a != ~0)
 	{
-	    comp4_t d = 0;
+	    uint32_t d = 0;
 
 	    if (a)
 	    {
 		d = *(dest + i);
-		UNcx4_MUL_UNcx4 (d, a);
+		UN8x4_MUL_UN8x4 (d, a);
 	    }
 
 	    *(dest + i) = d;
@@ -1783,26 +1806,26 @@ combine_out_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_atop_ca (pixman_implementation_t *imp,
                  pixman_op_t              op,
-                 comp4_t *                dest,
-                 const comp4_t *          src,
-                 const comp4_t *          mask,
+                 uint32_t *                dest,
+                 const uint32_t *          src,
+                 const uint32_t *          mask,
                  int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t d = *(dest + i);
-	comp4_t s = *(src + i);
-	comp4_t m = *(mask + i);
-	comp4_t ad;
-	comp2_t as = d >> A_SHIFT;
+	uint32_t d = *(dest + i);
+	uint32_t s = *(src + i);
+	uint32_t m = *(mask + i);
+	uint32_t ad;
+	uint16_t as = d >> A_SHIFT;
 
 	combine_mask_ca (&s, &m);
 
 	ad = ~m;
 
-	UNcx4_MUL_UNcx4_ADD_UNcx4_MUL_UNc (d, ad, s, as);
+	UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (d, ad, s, as);
 
 	*(dest + i) = d;
     }
@@ -1811,26 +1834,26 @@ combine_atop_ca (pixman_implementation_t *imp,
 static void
 combine_atop_reverse_ca (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t d = *(dest + i);
-	comp4_t s = *(src + i);
-	comp4_t m = *(mask + i);
-	comp4_t ad;
-	comp2_t as = ~d >> A_SHIFT;
+	uint32_t d = *(dest + i);
+	uint32_t s = *(src + i);
+	uint32_t m = *(mask + i);
+	uint32_t ad;
+	uint16_t as = ~d >> A_SHIFT;
 
 	combine_mask_ca (&s, &m);
 
 	ad = m;
 
-	UNcx4_MUL_UNcx4_ADD_UNcx4_MUL_UNc (d, ad, s, as);
+	UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (d, ad, s, as);
 
 	*(dest + i) = d;
     }
@@ -1839,26 +1862,26 @@ combine_atop_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_xor_ca (pixman_implementation_t *imp,
                 pixman_op_t              op,
-                comp4_t *                dest,
-                const comp4_t *          src,
-                const comp4_t *          mask,
+                uint32_t *                dest,
+                const uint32_t *          src,
+                const uint32_t *          mask,
                 int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t d = *(dest + i);
-	comp4_t s = *(src + i);
-	comp4_t m = *(mask + i);
-	comp4_t ad;
-	comp2_t as = ~d >> A_SHIFT;
+	uint32_t d = *(dest + i);
+	uint32_t s = *(src + i);
+	uint32_t m = *(mask + i);
+	uint32_t ad;
+	uint16_t as = ~d >> A_SHIFT;
 
 	combine_mask_ca (&s, &m);
 
 	ad = ~m;
 
-	UNcx4_MUL_UNcx4_ADD_UNcx4_MUL_UNc (d, ad, s, as);
+	UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (d, ad, s, as);
 
 	*(dest + i) = d;
     }
@@ -1867,22 +1890,22 @@ combine_xor_ca (pixman_implementation_t *imp,
 static void
 combine_add_ca (pixman_implementation_t *imp,
                 pixman_op_t              op,
-                comp4_t *                dest,
-                const comp4_t *          src,
-                const comp4_t *          mask,
+                uint32_t *                dest,
+                const uint32_t *          src,
+                const uint32_t *          mask,
                 int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s = *(src + i);
-	comp4_t m = *(mask + i);
-	comp4_t d = *(dest + i);
+	uint32_t s = *(src + i);
+	uint32_t m = *(mask + i);
+	uint32_t d = *(dest + i);
 
 	combine_mask_value_ca (&s, &m);
 
-	UNcx4_ADD_UNcx4 (d, s);
+	UN8x4_ADD_UN8x4 (d, s);
 
 	*(dest + i) = d;
     }
@@ -1891,19 +1914,19 @@ combine_add_ca (pixman_implementation_t *imp,
 static void
 combine_saturate_ca (pixman_implementation_t *imp,
                      pixman_op_t              op,
-                     comp4_t *                dest,
-                     const comp4_t *          src,
-                     const comp4_t *          mask,
+                     uint32_t *                dest,
+                     const uint32_t *          src,
+                     const uint32_t *          mask,
                      int                      width)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s, d;
-	comp2_t sa, sr, sg, sb, da;
-	comp2_t t, u, v;
-	comp4_t m, n, o, p;
+	uint32_t s, d;
+	uint16_t sa, sr, sg, sb, da;
+	uint16_t t, u, v;
+	uint32_t m, n, o, p;
 
 	d = *(dest + i);
 	s = *(src + i);
@@ -1942,22 +1965,22 @@ combine_saturate_ca (pixman_implementation_t *imp,
 }
 
 static void
-combine_disjoint_general_ca (comp4_t *      dest,
-                             const comp4_t *src,
-                             const comp4_t *mask,
+combine_disjoint_general_ca (uint32_t *      dest,
+                             const uint32_t *src,
+                             const uint32_t *mask,
                              int            width,
-                             comp1_t        combine)
+                             uint8_t        combine)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s, d;
-	comp4_t m, n, o, p;
-	comp4_t Fa, Fb;
-	comp2_t t, u, v;
-	comp4_t sa;
-	comp1_t da;
+	uint32_t s, d;
+	uint32_t m, n, o, p;
+	uint32_t Fa, Fb;
+	uint16_t t, u, v;
+	uint32_t sa;
+	uint8_t da;
 
 	s = *(src + i);
 	m = *(mask + i);
@@ -1975,18 +1998,18 @@ combine_disjoint_general_ca (comp4_t *      dest,
 	    break;
 
 	case COMBINE_A_OUT:
-	    m = (comp4_t)combine_disjoint_out_part ((comp1_t) (sa >> 0), da);
-	    n = (comp4_t)combine_disjoint_out_part ((comp1_t) (sa >> G_SHIFT), da) << G_SHIFT;
-	    o = (comp4_t)combine_disjoint_out_part ((comp1_t) (sa >> R_SHIFT), da) << R_SHIFT;
-	    p = (comp4_t)combine_disjoint_out_part ((comp1_t) (sa >> A_SHIFT), da) << A_SHIFT;
+	    m = (uint32_t)combine_disjoint_out_part ((uint8_t) (sa >> 0), da);
+	    n = (uint32_t)combine_disjoint_out_part ((uint8_t) (sa >> G_SHIFT), da) << G_SHIFT;
+	    o = (uint32_t)combine_disjoint_out_part ((uint8_t) (sa >> R_SHIFT), da) << R_SHIFT;
+	    p = (uint32_t)combine_disjoint_out_part ((uint8_t) (sa >> A_SHIFT), da) << A_SHIFT;
 	    Fa = m | n | o | p;
 	    break;
 
 	case COMBINE_A_IN:
-	    m = (comp4_t)combine_disjoint_in_part ((comp1_t) (sa >> 0), da);
-	    n = (comp4_t)combine_disjoint_in_part ((comp1_t) (sa >> G_SHIFT), da) << G_SHIFT;
-	    o = (comp4_t)combine_disjoint_in_part ((comp1_t) (sa >> R_SHIFT), da) << R_SHIFT;
-	    p = (comp4_t)combine_disjoint_in_part ((comp1_t) (sa >> A_SHIFT), da) << A_SHIFT;
+	    m = (uint32_t)combine_disjoint_in_part ((uint8_t) (sa >> 0), da);
+	    n = (uint32_t)combine_disjoint_in_part ((uint8_t) (sa >> G_SHIFT), da) << G_SHIFT;
+	    o = (uint32_t)combine_disjoint_in_part ((uint8_t) (sa >> R_SHIFT), da) << R_SHIFT;
+	    p = (uint32_t)combine_disjoint_in_part ((uint8_t) (sa >> A_SHIFT), da) << A_SHIFT;
 	    Fa = m | n | o | p;
 	    break;
 
@@ -2002,18 +2025,18 @@ combine_disjoint_general_ca (comp4_t *      dest,
 	    break;
 
 	case COMBINE_B_OUT:
-	    m = (comp4_t)combine_disjoint_out_part (da, (comp1_t) (sa >> 0));
-	    n = (comp4_t)combine_disjoint_out_part (da, (comp1_t) (sa >> G_SHIFT)) << G_SHIFT;
-	    o = (comp4_t)combine_disjoint_out_part (da, (comp1_t) (sa >> R_SHIFT)) << R_SHIFT;
-	    p = (comp4_t)combine_disjoint_out_part (da, (comp1_t) (sa >> A_SHIFT)) << A_SHIFT;
+	    m = (uint32_t)combine_disjoint_out_part (da, (uint8_t) (sa >> 0));
+	    n = (uint32_t)combine_disjoint_out_part (da, (uint8_t) (sa >> G_SHIFT)) << G_SHIFT;
+	    o = (uint32_t)combine_disjoint_out_part (da, (uint8_t) (sa >> R_SHIFT)) << R_SHIFT;
+	    p = (uint32_t)combine_disjoint_out_part (da, (uint8_t) (sa >> A_SHIFT)) << A_SHIFT;
 	    Fb = m | n | o | p;
 	    break;
 
 	case COMBINE_B_IN:
-	    m = (comp4_t)combine_disjoint_in_part (da, (comp1_t) (sa >> 0));
-	    n = (comp4_t)combine_disjoint_in_part (da, (comp1_t) (sa >> G_SHIFT)) << G_SHIFT;
-	    o = (comp4_t)combine_disjoint_in_part (da, (comp1_t) (sa >> R_SHIFT)) << R_SHIFT;
-	    p = (comp4_t)combine_disjoint_in_part (da, (comp1_t) (sa >> A_SHIFT)) << A_SHIFT;
+	    m = (uint32_t)combine_disjoint_in_part (da, (uint8_t) (sa >> 0));
+	    n = (uint32_t)combine_disjoint_in_part (da, (uint8_t) (sa >> G_SHIFT)) << G_SHIFT;
+	    o = (uint32_t)combine_disjoint_in_part (da, (uint8_t) (sa >> R_SHIFT)) << R_SHIFT;
+	    p = (uint32_t)combine_disjoint_in_part (da, (uint8_t) (sa >> A_SHIFT)) << A_SHIFT;
 	    Fb = m | n | o | p;
 	    break;
 
@@ -2035,9 +2058,9 @@ combine_disjoint_general_ca (comp4_t *      dest,
 static void
 combine_disjoint_over_ca (pixman_implementation_t *imp,
                           pixman_op_t              op,
-                          comp4_t *                dest,
-                          const comp4_t *          src,
-                          const comp4_t *          mask,
+                          uint32_t *                dest,
+                          const uint32_t *          src,
+                          const uint32_t *          mask,
                           int                      width)
 {
     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_OVER);
@@ -2046,9 +2069,9 @@ combine_disjoint_over_ca (pixman_implementation_t *imp,
 static void
 combine_disjoint_in_ca (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_IN);
@@ -2057,9 +2080,9 @@ combine_disjoint_in_ca (pixman_implementation_t *imp,
 static void
 combine_disjoint_in_reverse_ca (pixman_implementation_t *imp,
                                 pixman_op_t              op,
-                                comp4_t *                dest,
-                                const comp4_t *          src,
-                                const comp4_t *          mask,
+                                uint32_t *                dest,
+                                const uint32_t *          src,
+                                const uint32_t *          mask,
                                 int                      width)
 {
     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_IN);
@@ -2068,9 +2091,9 @@ combine_disjoint_in_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_disjoint_out_ca (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_OUT);
@@ -2079,9 +2102,9 @@ combine_disjoint_out_ca (pixman_implementation_t *imp,
 static void
 combine_disjoint_out_reverse_ca (pixman_implementation_t *imp,
                                  pixman_op_t              op,
-                                 comp4_t *                dest,
-                                 const comp4_t *          src,
-                                 const comp4_t *          mask,
+                                 uint32_t *                dest,
+                                 const uint32_t *          src,
+                                 const uint32_t *          mask,
                                  int                      width)
 {
     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_OUT);
@@ -2090,9 +2113,9 @@ combine_disjoint_out_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_disjoint_atop_ca (pixman_implementation_t *imp,
                           pixman_op_t              op,
-                          comp4_t *                dest,
-                          const comp4_t *          src,
-                          const comp4_t *          mask,
+                          uint32_t *                dest,
+                          const uint32_t *          src,
+                          const uint32_t *          mask,
                           int                      width)
 {
     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_ATOP);
@@ -2101,9 +2124,9 @@ combine_disjoint_atop_ca (pixman_implementation_t *imp,
 static void
 combine_disjoint_atop_reverse_ca (pixman_implementation_t *imp,
                                   pixman_op_t              op,
-                                  comp4_t *                dest,
-                                  const comp4_t *          src,
-                                  const comp4_t *          mask,
+                                  uint32_t *                dest,
+                                  const uint32_t *          src,
+                                  const uint32_t *          mask,
                                   int                      width)
 {
     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_ATOP);
@@ -2112,31 +2135,31 @@ combine_disjoint_atop_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_disjoint_xor_ca (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_XOR);
 }
 
 static void
-combine_conjoint_general_ca (comp4_t *      dest,
-                             const comp4_t *src,
-                             const comp4_t *mask,
+combine_conjoint_general_ca (uint32_t *      dest,
+                             const uint32_t *src,
+                             const uint32_t *mask,
                              int            width,
-                             comp1_t        combine)
+                             uint8_t        combine)
 {
     int i;
 
     for (i = 0; i < width; ++i)
     {
-	comp4_t s, d;
-	comp4_t m, n, o, p;
-	comp4_t Fa, Fb;
-	comp2_t t, u, v;
-	comp4_t sa;
-	comp1_t da;
+	uint32_t s, d;
+	uint32_t m, n, o, p;
+	uint32_t Fa, Fb;
+	uint16_t t, u, v;
+	uint32_t sa;
+	uint8_t da;
 
 	s = *(src + i);
 	m = *(mask + i);
@@ -2154,18 +2177,18 @@ combine_conjoint_general_ca (comp4_t *      dest,
 	    break;
 
 	case COMBINE_A_OUT:
-	    m = (comp4_t)combine_conjoint_out_part ((comp1_t) (sa >> 0), da);
-	    n = (comp4_t)combine_conjoint_out_part ((comp1_t) (sa >> G_SHIFT), da) << G_SHIFT;
-	    o = (comp4_t)combine_conjoint_out_part ((comp1_t) (sa >> R_SHIFT), da) << R_SHIFT;
-	    p = (comp4_t)combine_conjoint_out_part ((comp1_t) (sa >> A_SHIFT), da) << A_SHIFT;
+	    m = (uint32_t)combine_conjoint_out_part ((uint8_t) (sa >> 0), da);
+	    n = (uint32_t)combine_conjoint_out_part ((uint8_t) (sa >> G_SHIFT), da) << G_SHIFT;
+	    o = (uint32_t)combine_conjoint_out_part ((uint8_t) (sa >> R_SHIFT), da) << R_SHIFT;
+	    p = (uint32_t)combine_conjoint_out_part ((uint8_t) (sa >> A_SHIFT), da) << A_SHIFT;
 	    Fa = m | n | o | p;
 	    break;
 
 	case COMBINE_A_IN:
-	    m = (comp4_t)combine_conjoint_in_part ((comp1_t) (sa >> 0), da);
-	    n = (comp4_t)combine_conjoint_in_part ((comp1_t) (sa >> G_SHIFT), da) << G_SHIFT;
-	    o = (comp4_t)combine_conjoint_in_part ((comp1_t) (sa >> R_SHIFT), da) << R_SHIFT;
-	    p = (comp4_t)combine_conjoint_in_part ((comp1_t) (sa >> A_SHIFT), da) << A_SHIFT;
+	    m = (uint32_t)combine_conjoint_in_part ((uint8_t) (sa >> 0), da);
+	    n = (uint32_t)combine_conjoint_in_part ((uint8_t) (sa >> G_SHIFT), da) << G_SHIFT;
+	    o = (uint32_t)combine_conjoint_in_part ((uint8_t) (sa >> R_SHIFT), da) << R_SHIFT;
+	    p = (uint32_t)combine_conjoint_in_part ((uint8_t) (sa >> A_SHIFT), da) << A_SHIFT;
 	    Fa = m | n | o | p;
 	    break;
 
@@ -2181,18 +2204,18 @@ combine_conjoint_general_ca (comp4_t *      dest,
 	    break;
 
 	case COMBINE_B_OUT:
-	    m = (comp4_t)combine_conjoint_out_part (da, (comp1_t) (sa >> 0));
-	    n = (comp4_t)combine_conjoint_out_part (da, (comp1_t) (sa >> G_SHIFT)) << G_SHIFT;
-	    o = (comp4_t)combine_conjoint_out_part (da, (comp1_t) (sa >> R_SHIFT)) << R_SHIFT;
-	    p = (comp4_t)combine_conjoint_out_part (da, (comp1_t) (sa >> A_SHIFT)) << A_SHIFT;
+	    m = (uint32_t)combine_conjoint_out_part (da, (uint8_t) (sa >> 0));
+	    n = (uint32_t)combine_conjoint_out_part (da, (uint8_t) (sa >> G_SHIFT)) << G_SHIFT;
+	    o = (uint32_t)combine_conjoint_out_part (da, (uint8_t) (sa >> R_SHIFT)) << R_SHIFT;
+	    p = (uint32_t)combine_conjoint_out_part (da, (uint8_t) (sa >> A_SHIFT)) << A_SHIFT;
 	    Fb = m | n | o | p;
 	    break;
 
 	case COMBINE_B_IN:
-	    m = (comp4_t)combine_conjoint_in_part (da, (comp1_t) (sa >> 0));
-	    n = (comp4_t)combine_conjoint_in_part (da, (comp1_t) (sa >> G_SHIFT)) << G_SHIFT;
-	    o = (comp4_t)combine_conjoint_in_part (da, (comp1_t) (sa >> R_SHIFT)) << R_SHIFT;
-	    p = (comp4_t)combine_conjoint_in_part (da, (comp1_t) (sa >> A_SHIFT)) << A_SHIFT;
+	    m = (uint32_t)combine_conjoint_in_part (da, (uint8_t) (sa >> 0));
+	    n = (uint32_t)combine_conjoint_in_part (da, (uint8_t) (sa >> G_SHIFT)) << G_SHIFT;
+	    o = (uint32_t)combine_conjoint_in_part (da, (uint8_t) (sa >> R_SHIFT)) << R_SHIFT;
+	    p = (uint32_t)combine_conjoint_in_part (da, (uint8_t) (sa >> A_SHIFT)) << A_SHIFT;
 	    Fb = m | n | o | p;
 	    break;
 
@@ -2214,9 +2237,9 @@ combine_conjoint_general_ca (comp4_t *      dest,
 static void
 combine_conjoint_over_ca (pixman_implementation_t *imp,
                           pixman_op_t              op,
-                          comp4_t *                dest,
-                          const comp4_t *          src,
-                          const comp4_t *          mask,
+                          uint32_t *                dest,
+                          const uint32_t *          src,
+                          const uint32_t *          mask,
                           int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_OVER);
@@ -2225,9 +2248,9 @@ combine_conjoint_over_ca (pixman_implementation_t *imp,
 static void
 combine_conjoint_over_reverse_ca (pixman_implementation_t *imp,
                                   pixman_op_t              op,
-                                  comp4_t *                dest,
-                                  const comp4_t *          src,
-                                  const comp4_t *          mask,
+                                  uint32_t *                dest,
+                                  const uint32_t *          src,
+                                  const uint32_t *          mask,
                                   int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_OVER);
@@ -2236,9 +2259,9 @@ combine_conjoint_over_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_conjoint_in_ca (pixman_implementation_t *imp,
                         pixman_op_t              op,
-                        comp4_t *                dest,
-                        const comp4_t *          src,
-                        const comp4_t *          mask,
+                        uint32_t *                dest,
+                        const uint32_t *          src,
+                        const uint32_t *          mask,
                         int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_IN);
@@ -2247,9 +2270,9 @@ combine_conjoint_in_ca (pixman_implementation_t *imp,
 static void
 combine_conjoint_in_reverse_ca (pixman_implementation_t *imp,
                                 pixman_op_t              op,
-                                comp4_t *                dest,
-                                const comp4_t *          src,
-                                const comp4_t *          mask,
+                                uint32_t *                dest,
+                                const uint32_t *          src,
+                                const uint32_t *          mask,
                                 int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_IN);
@@ -2258,9 +2281,9 @@ combine_conjoint_in_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_conjoint_out_ca (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_OUT);
@@ -2269,9 +2292,9 @@ combine_conjoint_out_ca (pixman_implementation_t *imp,
 static void
 combine_conjoint_out_reverse_ca (pixman_implementation_t *imp,
                                  pixman_op_t              op,
-                                 comp4_t *                dest,
-                                 const comp4_t *          src,
-                                 const comp4_t *          mask,
+                                 uint32_t *                dest,
+                                 const uint32_t *          src,
+                                 const uint32_t *          mask,
                                  int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_OUT);
@@ -2280,9 +2303,9 @@ combine_conjoint_out_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_conjoint_atop_ca (pixman_implementation_t *imp,
                           pixman_op_t              op,
-                          comp4_t *                dest,
-                          const comp4_t *          src,
-                          const comp4_t *          mask,
+                          uint32_t *                dest,
+                          const uint32_t *          src,
+                          const uint32_t *          mask,
                           int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_ATOP);
@@ -2291,9 +2314,9 @@ combine_conjoint_atop_ca (pixman_implementation_t *imp,
 static void
 combine_conjoint_atop_reverse_ca (pixman_implementation_t *imp,
                                   pixman_op_t              op,
-                                  comp4_t *                dest,
-                                  const comp4_t *          src,
-                                  const comp4_t *          mask,
+                                  uint32_t *                dest,
+                                  const uint32_t *          src,
+                                  const uint32_t *          mask,
                                   int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_ATOP);
@@ -2302,136 +2325,136 @@ combine_conjoint_atop_reverse_ca (pixman_implementation_t *imp,
 static void
 combine_conjoint_xor_ca (pixman_implementation_t *imp,
                          pixman_op_t              op,
-                         comp4_t *                dest,
-                         const comp4_t *          src,
-                         const comp4_t *          mask,
+                         uint32_t *                dest,
+                         const uint32_t *          src,
+                         const uint32_t *          mask,
                          int                      width)
 {
     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_XOR);
 }
 
 void
-_pixman_setup_combiner_functions_width (pixman_implementation_t *imp)
+_pixman_setup_combiner_functions_32 (pixman_implementation_t *imp)
 {
     /* Unified alpha */
-    imp->combine_width[PIXMAN_OP_CLEAR] = combine_clear;
-    imp->combine_width[PIXMAN_OP_SRC] = combine_src_u;
-    imp->combine_width[PIXMAN_OP_DST] = combine_dst;
-    imp->combine_width[PIXMAN_OP_OVER] = combine_over_u;
-    imp->combine_width[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_u;
-    imp->combine_width[PIXMAN_OP_IN] = combine_in_u;
-    imp->combine_width[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_u;
-    imp->combine_width[PIXMAN_OP_OUT] = combine_out_u;
-    imp->combine_width[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_u;
-    imp->combine_width[PIXMAN_OP_ATOP] = combine_atop_u;
-    imp->combine_width[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_u;
-    imp->combine_width[PIXMAN_OP_XOR] = combine_xor_u;
-    imp->combine_width[PIXMAN_OP_ADD] = combine_add_u;
-    imp->combine_width[PIXMAN_OP_SATURATE] = combine_saturate_u;
+    imp->combine_32[PIXMAN_OP_CLEAR] = combine_clear;
+    imp->combine_32[PIXMAN_OP_SRC] = combine_src_u;
+    imp->combine_32[PIXMAN_OP_DST] = combine_dst;
+    imp->combine_32[PIXMAN_OP_OVER] = combine_over_u;
+    imp->combine_32[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_u;
+    imp->combine_32[PIXMAN_OP_IN] = combine_in_u;
+    imp->combine_32[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_u;
+    imp->combine_32[PIXMAN_OP_OUT] = combine_out_u;
+    imp->combine_32[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_u;
+    imp->combine_32[PIXMAN_OP_ATOP] = combine_atop_u;
+    imp->combine_32[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_u;
+    imp->combine_32[PIXMAN_OP_XOR] = combine_xor_u;
+    imp->combine_32[PIXMAN_OP_ADD] = combine_add_u;
+    imp->combine_32[PIXMAN_OP_SATURATE] = combine_saturate_u;
 
     /* Disjoint, unified */
-    imp->combine_width[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear;
-    imp->combine_width[PIXMAN_OP_DISJOINT_SRC] = combine_src_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_DST] = combine_dst;
-    imp->combine_width[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_u;
-    imp->combine_width[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear;
+    imp->combine_32[PIXMAN_OP_DISJOINT_SRC] = combine_src_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_DST] = combine_dst;
+    imp->combine_32[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_u;
+    imp->combine_32[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_u;
 
     /* Conjoint, unified */
-    imp->combine_width[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear;
-    imp->combine_width[PIXMAN_OP_CONJOINT_SRC] = combine_src_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_DST] = combine_dst;
-    imp->combine_width[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_u;
-    imp->combine_width[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_u;
-
-    imp->combine_width[PIXMAN_OP_MULTIPLY] = combine_multiply_u;
-    imp->combine_width[PIXMAN_OP_SCREEN] = combine_screen_u;
-    imp->combine_width[PIXMAN_OP_OVERLAY] = combine_overlay_u;
-    imp->combine_width[PIXMAN_OP_DARKEN] = combine_darken_u;
-    imp->combine_width[PIXMAN_OP_LIGHTEN] = combine_lighten_u;
-    imp->combine_width[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_u;
-    imp->combine_width[PIXMAN_OP_COLOR_BURN] = combine_color_burn_u;
-    imp->combine_width[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_u;
-    imp->combine_width[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_u;
-    imp->combine_width[PIXMAN_OP_DIFFERENCE] = combine_difference_u;
-    imp->combine_width[PIXMAN_OP_EXCLUSION] = combine_exclusion_u;
-    imp->combine_width[PIXMAN_OP_HSL_HUE] = combine_hsl_hue_u;
-    imp->combine_width[PIXMAN_OP_HSL_SATURATION] = combine_hsl_saturation_u;
-    imp->combine_width[PIXMAN_OP_HSL_COLOR] = combine_hsl_color_u;
-    imp->combine_width[PIXMAN_OP_HSL_LUMINOSITY] = combine_hsl_luminosity_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear;
+    imp->combine_32[PIXMAN_OP_CONJOINT_SRC] = combine_src_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_DST] = combine_dst;
+    imp->combine_32[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_u;
+    imp->combine_32[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_u;
+
+    imp->combine_32[PIXMAN_OP_MULTIPLY] = combine_multiply_u;
+    imp->combine_32[PIXMAN_OP_SCREEN] = combine_screen_u;
+    imp->combine_32[PIXMAN_OP_OVERLAY] = combine_overlay_u;
+    imp->combine_32[PIXMAN_OP_DARKEN] = combine_darken_u;
+    imp->combine_32[PIXMAN_OP_LIGHTEN] = combine_lighten_u;
+    imp->combine_32[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_u;
+    imp->combine_32[PIXMAN_OP_COLOR_BURN] = combine_color_burn_u;
+    imp->combine_32[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_u;
+    imp->combine_32[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_u;
+    imp->combine_32[PIXMAN_OP_DIFFERENCE] = combine_difference_u;
+    imp->combine_32[PIXMAN_OP_EXCLUSION] = combine_exclusion_u;
+    imp->combine_32[PIXMAN_OP_HSL_HUE] = combine_hsl_hue_u;
+    imp->combine_32[PIXMAN_OP_HSL_SATURATION] = combine_hsl_saturation_u;
+    imp->combine_32[PIXMAN_OP_HSL_COLOR] = combine_hsl_color_u;
+    imp->combine_32[PIXMAN_OP_HSL_LUMINOSITY] = combine_hsl_luminosity_u;
 
     /* Component alpha combiners */
-    imp->combine_width_ca[PIXMAN_OP_CLEAR] = combine_clear_ca;
-    imp->combine_width_ca[PIXMAN_OP_SRC] = combine_src_ca;
+    imp->combine_32_ca[PIXMAN_OP_CLEAR] = combine_clear_ca;
+    imp->combine_32_ca[PIXMAN_OP_SRC] = combine_src_ca;
     /* dest */
-    imp->combine_width_ca[PIXMAN_OP_OVER] = combine_over_ca;
-    imp->combine_width_ca[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_IN] = combine_in_ca;
-    imp->combine_width_ca[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_OUT] = combine_out_ca;
-    imp->combine_width_ca[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_ATOP] = combine_atop_ca;
-    imp->combine_width_ca[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_XOR] = combine_xor_ca;
-    imp->combine_width_ca[PIXMAN_OP_ADD] = combine_add_ca;
-    imp->combine_width_ca[PIXMAN_OP_SATURATE] = combine_saturate_ca;
+    imp->combine_32_ca[PIXMAN_OP_OVER] = combine_over_ca;
+    imp->combine_32_ca[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_IN] = combine_in_ca;
+    imp->combine_32_ca[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_OUT] = combine_out_ca;
+    imp->combine_32_ca[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_ATOP] = combine_atop_ca;
+    imp->combine_32_ca[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_XOR] = combine_xor_ca;
+    imp->combine_32_ca[PIXMAN_OP_ADD] = combine_add_ca;
+    imp->combine_32_ca[PIXMAN_OP_SATURATE] = combine_saturate_ca;
 
     /* Disjoint CA */
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_SRC] = combine_src_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_DST] = combine_dst;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_SRC] = combine_src_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_DST] = combine_dst;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_ca;
 
     /* Conjoint CA */
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_SRC] = combine_src_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_DST] = combine_dst;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_ca;
-    imp->combine_width_ca[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_ca;
-
-    imp->combine_width_ca[PIXMAN_OP_MULTIPLY] = combine_multiply_ca;
-    imp->combine_width_ca[PIXMAN_OP_SCREEN] = combine_screen_ca;
-    imp->combine_width_ca[PIXMAN_OP_OVERLAY] = combine_overlay_ca;
-    imp->combine_width_ca[PIXMAN_OP_DARKEN] = combine_darken_ca;
-    imp->combine_width_ca[PIXMAN_OP_LIGHTEN] = combine_lighten_ca;
-    imp->combine_width_ca[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_ca;
-    imp->combine_width_ca[PIXMAN_OP_COLOR_BURN] = combine_color_burn_ca;
-    imp->combine_width_ca[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_ca;
-    imp->combine_width_ca[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_ca;
-    imp->combine_width_ca[PIXMAN_OP_DIFFERENCE] = combine_difference_ca;
-    imp->combine_width_ca[PIXMAN_OP_EXCLUSION] = combine_exclusion_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_SRC] = combine_src_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_DST] = combine_dst;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_ca;
+    imp->combine_32_ca[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_ca;
+
+    imp->combine_32_ca[PIXMAN_OP_MULTIPLY] = combine_multiply_ca;
+    imp->combine_32_ca[PIXMAN_OP_SCREEN] = combine_screen_ca;
+    imp->combine_32_ca[PIXMAN_OP_OVERLAY] = combine_overlay_ca;
+    imp->combine_32_ca[PIXMAN_OP_DARKEN] = combine_darken_ca;
+    imp->combine_32_ca[PIXMAN_OP_LIGHTEN] = combine_lighten_ca;
+    imp->combine_32_ca[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_ca;
+    imp->combine_32_ca[PIXMAN_OP_COLOR_BURN] = combine_color_burn_ca;
+    imp->combine_32_ca[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_ca;
+    imp->combine_32_ca[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_ca;
+    imp->combine_32_ca[PIXMAN_OP_DIFFERENCE] = combine_difference_ca;
+    imp->combine_32_ca[PIXMAN_OP_EXCLUSION] = combine_exclusion_ca;
 
     /* It is not clear that these make sense, so make them noops for now */
-    imp->combine_width_ca[PIXMAN_OP_HSL_HUE] = combine_dst;
-    imp->combine_width_ca[PIXMAN_OP_HSL_SATURATION] = combine_dst;
-    imp->combine_width_ca[PIXMAN_OP_HSL_COLOR] = combine_dst;
-    imp->combine_width_ca[PIXMAN_OP_HSL_LUMINOSITY] = combine_dst;
+    imp->combine_32_ca[PIXMAN_OP_HSL_HUE] = combine_dst;
+    imp->combine_32_ca[PIXMAN_OP_HSL_SATURATION] = combine_dst;
+    imp->combine_32_ca[PIXMAN_OP_HSL_COLOR] = combine_dst;
+    imp->combine_32_ca[PIXMAN_OP_HSL_LUMINOSITY] = combine_dst;
 }
diff --git a/pixman/pixman/pixman-combine.h.template b/pixman/pixman/pixman-combine32.h
index 20f784b5b..875dde3cf 100644
--- a/pixman/pixman/pixman-combine.h.template
+++ b/pixman/pixman/pixman-combine32.h
@@ -1,40 +1,39 @@
-
-#define COMPONENT_SIZE
-#define MASK
-#define ONE_HALF
-
-#define A_SHIFT
-#define R_SHIFT
-#define G_SHIFT
-#define A_MASK
-#define R_MASK
-#define G_MASK
-
-#define RB_MASK
-#define AG_MASK
-#define RB_ONE_HALF
-#define RB_MASK_PLUS_ONE
-
-#define ALPHA_c(x) ((x) >> A_SHIFT)
-#define RED_c(x) (((x) >> R_SHIFT) & MASK)
-#define GREEN_c(x) (((x) >> G_SHIFT) & MASK)
-#define BLUE_c(x) ((x) & MASK)
+#define COMPONENT_SIZE 8
+#define MASK 0xff
+#define ONE_HALF 0x80
+
+#define A_SHIFT 8 * 3
+#define R_SHIFT 8 * 2
+#define G_SHIFT 8
+#define A_MASK 0xff000000
+#define R_MASK 0xff0000
+#define G_MASK 0xff00
+
+#define RB_MASK 0xff00ff
+#define AG_MASK 0xff00ff00
+#define RB_ONE_HALF 0x800080
+#define RB_MASK_PLUS_ONE 0x10000100
+
+#define ALPHA_8(x) ((x) >> A_SHIFT)
+#define RED_8(x) (((x) >> R_SHIFT) & MASK)
+#define GREEN_8(x) (((x) >> G_SHIFT) & MASK)
+#define BLUE_8(x) ((x) & MASK)
 
 /*
  * Helper macros.
  */
 
-#define MUL_UNc(a, b, t)						\
-    ((t) = (a) * (comp2_t)(b) + ONE_HALF, ((((t) >> G_SHIFT ) + (t) ) >> G_SHIFT ))
+#define MUL_UN8(a, b, t)						\
+    ((t) = (a) * (uint16_t)(b) + ONE_HALF, ((((t) >> G_SHIFT ) + (t) ) >> G_SHIFT ))
 
-#define DIV_UNc(a, b)							\
-    (((comp2_t) (a) * MASK + ((b) / 2)) / (b))
+#define DIV_UN8(a, b)							\
+    (((uint16_t) (a) * MASK + ((b) / 2)) / (b))
 
-#define ADD_UNc(x, y, t)				     \
+#define ADD_UN8(x, y, t)				     \
     ((t) = (x) + (y),					     \
-     (comp4_t) (comp1_t) ((t) | (0 - ((t) >> G_SHIFT))))
+     (uint32_t) (uint8_t) ((t) | (0 - ((t) >> G_SHIFT))))
 
-#define DIV_ONE_UNc(x)							\
+#define DIV_ONE_UN8(x)							\
     (((x) + ONE_HALF + (((x) + ONE_HALF) >> G_SHIFT)) >> G_SHIFT)
 
 /*
@@ -45,7 +44,7 @@
 /*
  * x_rb = (x_rb * a) / 255
  */
-#define UNc_rb_MUL_UNc(x, a, t)						\
+#define UN8_rb_MUL_UN8(x, a, t)						\
     do									\
     {									\
 	t  = ((x) & RB_MASK) * (a);					\
@@ -57,7 +56,7 @@
 /*
  * x_rb = min (x_rb + y_rb, 255)
  */
-#define UNc_rb_ADD_UNc_rb(x, y, t)					\
+#define UN8_rb_ADD_UN8_rb(x, y, t)					\
     do									\
     {									\
 	t = ((x) + (y));						\
@@ -68,7 +67,7 @@
 /*
  * x_rb = (x_rb * a_rb) / 255
  */
-#define UNc_rb_MUL_UNc_rb(x, a, t)					\
+#define UN8_rb_MUL_UN8_rb(x, a, t)					\
     do									\
     {									\
 	t  = (x & MASK) * (a & MASK);					\
@@ -81,16 +80,16 @@
 /*
  * x_c = (x_c * a) / 255
  */
-#define UNcx4_MUL_UNc(x, a)						\
+#define UN8x4_MUL_UN8(x, a)						\
     do									\
     {									\
-	comp4_t r1__, r2__, t__;					\
+	uint32_t r1__, r2__, t__;					\
 									\
 	r1__ = (x);							\
-	UNc_rb_MUL_UNc (r1__, (a), t__);				\
+	UN8_rb_MUL_UN8 (r1__, (a), t__);				\
 									\
 	r2__ = (x) >> G_SHIFT;						\
-	UNc_rb_MUL_UNc (r2__, (a), t__);				\
+	UN8_rb_MUL_UN8 (r2__, (a), t__);				\
 									\
 	(x) = r1__ | (r2__ << G_SHIFT);					\
     } while (0)
@@ -98,20 +97,20 @@
 /*
  * x_c = (x_c * a) / 255 + y_c
  */
-#define UNcx4_MUL_UNc_ADD_UNcx4(x, a, y)				\
+#define UN8x4_MUL_UN8_ADD_UN8x4(x, a, y)				\
     do									\
     {									\
-	comp4_t r1__, r2__, r3__, t__;					\
+	uint32_t r1__, r2__, r3__, t__;					\
 									\
 	r1__ = (x);							\
 	r2__ = (y) & RB_MASK;						\
-	UNc_rb_MUL_UNc (r1__, (a), t__);				\
-	UNc_rb_ADD_UNc_rb (r1__, r2__, t__);				\
+	UN8_rb_MUL_UN8 (r1__, (a), t__);				\
+	UN8_rb_ADD_UN8_rb (r1__, r2__, t__);				\
 									\
 	r2__ = (x) >> G_SHIFT;						\
 	r3__ = ((y) >> G_SHIFT) & RB_MASK;				\
-	UNc_rb_MUL_UNc (r2__, (a), t__);				\
-	UNc_rb_ADD_UNc_rb (r2__, r3__, t__);				\
+	UN8_rb_MUL_UN8 (r2__, (a), t__);				\
+	UN8_rb_ADD_UN8_rb (r2__, r3__, t__);				\
 									\
 	(x) = r1__ | (r2__ << G_SHIFT);					\
     } while (0)
@@ -119,22 +118,22 @@
 /*
  * x_c = (x_c * a + y_c * b) / 255
  */
-#define UNcx4_MUL_UNc_ADD_UNcx4_MUL_UNc(x, a, y, b)			\
+#define UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8(x, a, y, b)			\
     do									\
     {									\
-	comp4_t r1__, r2__, r3__, t__;					\
+	uint32_t r1__, r2__, r3__, t__;					\
 									\
 	r1__ = (x);							\
 	r2__ = (y);							\
-	UNc_rb_MUL_UNc (r1__, (a), t__);				\
-	UNc_rb_MUL_UNc (r2__, (b), t__);				\
-	UNc_rb_ADD_UNc_rb (r1__, r2__, t__);				\
+	UN8_rb_MUL_UN8 (r1__, (a), t__);				\
+	UN8_rb_MUL_UN8 (r2__, (b), t__);				\
+	UN8_rb_ADD_UN8_rb (r1__, r2__, t__);				\
 									\
 	r2__ = ((x) >> G_SHIFT);					\
 	r3__ = ((y) >> G_SHIFT);					\
-	UNc_rb_MUL_UNc (r2__, (a), t__);				\
-	UNc_rb_MUL_UNc (r3__, (b), t__);				\
-	UNc_rb_ADD_UNc_rb (r2__, r3__, t__);				\
+	UN8_rb_MUL_UN8 (r2__, (a), t__);				\
+	UN8_rb_MUL_UN8 (r3__, (b), t__);				\
+	UN8_rb_ADD_UN8_rb (r2__, r3__, t__);				\
 									\
 	(x) = r1__ | (r2__ << G_SHIFT);					\
     } while (0)
@@ -142,18 +141,18 @@
 /*
  * x_c = (x_c * a_c) / 255
  */
-#define UNcx4_MUL_UNcx4(x, a)						\
+#define UN8x4_MUL_UN8x4(x, a)						\
     do									\
     {									\
-	comp4_t r1__, r2__, r3__, t__;					\
+	uint32_t r1__, r2__, r3__, t__;					\
 									\
 	r1__ = (x);							\
 	r2__ = (a);							\
-	UNc_rb_MUL_UNc_rb (r1__, r2__, t__);				\
+	UN8_rb_MUL_UN8_rb (r1__, r2__, t__);				\
 									\
 	r2__ = (x) >> G_SHIFT;						\
 	r3__ = (a) >> G_SHIFT;						\
-	UNc_rb_MUL_UNc_rb (r2__, r3__, t__);				\
+	UN8_rb_MUL_UN8_rb (r2__, r3__, t__);				\
 									\
 	(x) = r1__ | (r2__ << G_SHIFT);					\
     } while (0)
@@ -161,22 +160,22 @@
 /*
  * x_c = (x_c * a_c) / 255 + y_c
  */
-#define UNcx4_MUL_UNcx4_ADD_UNcx4(x, a, y)				\
+#define UN8x4_MUL_UN8x4_ADD_UN8x4(x, a, y)				\
     do									\
     {									\
-	comp4_t r1__, r2__, r3__, t__;					\
+	uint32_t r1__, r2__, r3__, t__;					\
 									\
 	r1__ = (x);							\
 	r2__ = (a);							\
-	UNc_rb_MUL_UNc_rb (r1__, r2__, t__);				\
+	UN8_rb_MUL_UN8_rb (r1__, r2__, t__);				\
 	r2__ = (y) & RB_MASK;						\
-	UNc_rb_ADD_UNc_rb (r1__, r2__, t__);				\
+	UN8_rb_ADD_UN8_rb (r1__, r2__, t__);				\
 									\
 	r2__ = ((x) >> G_SHIFT);					\
 	r3__ = ((a) >> G_SHIFT);					\
-	UNc_rb_MUL_UNc_rb (r2__, r3__, t__);				\
+	UN8_rb_MUL_UN8_rb (r2__, r3__, t__);				\
 	r3__ = ((y) >> G_SHIFT) & RB_MASK;				\
-	UNc_rb_ADD_UNc_rb (r2__, r3__, t__);				\
+	UN8_rb_ADD_UN8_rb (r2__, r3__, t__);				\
 									\
 	(x) = r1__ | (r2__ << G_SHIFT);					\
     } while (0)
@@ -184,24 +183,24 @@
 /*
  * x_c = (x_c * a_c + y_c * b) / 255
  */
-#define UNcx4_MUL_UNcx4_ADD_UNcx4_MUL_UNc(x, a, y, b)			\
+#define UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8(x, a, y, b)			\
     do									\
     {									\
-	comp4_t r1__, r2__, r3__, t__;					\
+	uint32_t r1__, r2__, r3__, t__;					\
 									\
 	r1__ = (x);							\
 	r2__ = (a);							\
-	UNc_rb_MUL_UNc_rb (r1__, r2__, t__);				\
+	UN8_rb_MUL_UN8_rb (r1__, r2__, t__);				\
 	r2__ = (y);							\
-	UNc_rb_MUL_UNc (r2__, (b), t__);				\
-	UNc_rb_ADD_UNc_rb (r1__, r2__, t__);				\
+	UN8_rb_MUL_UN8 (r2__, (b), t__);				\
+	UN8_rb_ADD_UN8_rb (r1__, r2__, t__);				\
 									\
 	r2__ = (x) >> G_SHIFT;						\
 	r3__ = (a) >> G_SHIFT;						\
-	UNc_rb_MUL_UNc_rb (r2__, r3__, t__);				\
+	UN8_rb_MUL_UN8_rb (r2__, r3__, t__);				\
 	r3__ = (y) >> G_SHIFT;						\
-	UNc_rb_MUL_UNc (r3__, (b), t__);				\
-	UNc_rb_ADD_UNc_rb (r2__, r3__, t__);				\
+	UN8_rb_MUL_UN8 (r3__, (b), t__);				\
+	UN8_rb_ADD_UN8_rb (r2__, r3__, t__);				\
 									\
 	x = r1__ | (r2__ << G_SHIFT);					\
     } while (0)
@@ -209,18 +208,18 @@
 /*
   x_c = min(x_c + y_c, 255)
 */
-#define UNcx4_ADD_UNcx4(x, y)						\
+#define UN8x4_ADD_UN8x4(x, y)						\
     do									\
     {									\
-	comp4_t r1__, r2__, r3__, t__;					\
+	uint32_t r1__, r2__, r3__, t__;					\
 									\
 	r1__ = (x) & RB_MASK;						\
 	r2__ = (y) & RB_MASK;						\
-	UNc_rb_ADD_UNc_rb (r1__, r2__, t__);				\
+	UN8_rb_ADD_UN8_rb (r1__, r2__, t__);				\
 									\
 	r2__ = ((x) >> G_SHIFT) & RB_MASK;				\
 	r3__ = ((y) >> G_SHIFT) & RB_MASK;				\
-	UNc_rb_ADD_UNc_rb (r2__, r3__, t__);				\
+	UN8_rb_ADD_UN8_rb (r2__, r3__, t__);				\
 									\
 	x = r1__ | (r2__ << G_SHIFT);					\
     } while (0)
diff --git a/pixman/pixman/pixman-conical-gradient.c b/pixman/pixman/pixman-conical-gradient.c
index 05d3595b6..8bb46aecd 100644
--- a/pixman/pixman/pixman-conical-gradient.c
+++ b/pixman/pixman/pixman-conical-gradient.c
@@ -165,7 +165,8 @@ conical_get_scanline_wide (pixman_iter_t *iter, const uint32_t *mask)
 {
     uint32_t *buffer = conical_get_scanline_narrow (iter, NULL);
 
-    pixman_expand ((uint64_t *)buffer, buffer, PIXMAN_a8r8g8b8, iter->width);
+    pixman_expand_to_float (
+	(argb_t *)buffer, buffer, PIXMAN_a8r8g8b8, iter->width);
 
     return buffer;
 }
diff --git a/pixman/pixman/pixman-general.c b/pixman/pixman/pixman-general.c
index 42a84a0b4..0bf91e444 100644
--- a/pixman/pixman/pixman-general.c
+++ b/pixman/pixman/pixman-general.c
@@ -130,7 +130,7 @@ general_composite_rect  (pixman_implementation_t *imp,
     else
     {
 	narrow = 0;
-	Bpp = 8;
+	Bpp = 16;
     }
 
     if (width * Bpp > SCANLINE_BUFFER_LENGTH)
@@ -145,6 +145,14 @@ general_composite_rect  (pixman_implementation_t *imp,
     mask_buffer = src_buffer + width * Bpp;
     dest_buffer = mask_buffer + width * Bpp;
 
+    if (!narrow)
+    {
+	/* To make sure there aren't any NANs in the buffers */
+	memset (src_buffer, 0, width * Bpp);
+	memset (mask_buffer, 0, width * Bpp);
+	memset (dest_buffer, 0, width * Bpp);
+    }
+    
     /* src iter */
     src_iter_flags = narrow | op_flags[op].src;
 
@@ -212,7 +220,7 @@ _pixman_implementation_create_general (void)
     pixman_implementation_t *imp = _pixman_implementation_create (NULL, general_fast_path);
 
     _pixman_setup_combiner_functions_32 (imp);
-    _pixman_setup_combiner_functions_64 (imp);
+    _pixman_setup_combiner_functions_float (imp);
 
     imp->src_iter_init = general_src_iter_init;
     imp->dest_iter_init = general_dest_iter_init;
diff --git a/pixman/pixman/pixman-implementation.c b/pixman/pixman/pixman-implementation.c
index 5dd050199..a70892c75 100644
--- a/pixman/pixman/pixman-implementation.c
+++ b/pixman/pixman/pixman-implementation.c
@@ -177,11 +177,11 @@ _pixman_implementation_lookup_combiner (pixman_implementation_t *imp,
 	switch ((narrow << 1) | component_alpha)
 	{
 	case 0: /* not narrow, not component alpha */
-	    f = (pixman_combine_32_func_t)imp->combine_64[op];
+	    f = (pixman_combine_32_func_t)imp->combine_float[op];
 	    break;
 	    
 	case 1: /* not narrow, component_alpha */
-	    f = (pixman_combine_32_func_t)imp->combine_64_ca[op];
+	    f = (pixman_combine_32_func_t)imp->combine_float_ca[op];
 	    break;
 
 	case 2: /* narrow, not component alpha */
diff --git a/pixman/pixman/pixman-linear-gradient.c b/pixman/pixman/pixman-linear-gradient.c
index e51136827..40c8c9f37 100644
--- a/pixman/pixman/pixman-linear-gradient.c
+++ b/pixman/pixman/pixman-linear-gradient.c
@@ -227,7 +227,8 @@ linear_get_scanline_wide (pixman_iter_t *iter, const uint32_t *mask)
 {
     uint32_t *buffer = linear_get_scanline_narrow (iter, NULL);
 
-    pixman_expand ((uint64_t *)buffer, buffer, PIXMAN_a8r8g8b8, iter->width);
+    pixman_expand_to_float (
+	(argb_t *)buffer, buffer, PIXMAN_a8r8g8b8, iter->width);
 
     return buffer;
 }
diff --git a/pixman/pixman/pixman-noop.c b/pixman/pixman/pixman-noop.c
index 7b9759ffb..850caa192 100644
--- a/pixman/pixman/pixman-noop.c
+++ b/pixman/pixman/pixman-noop.c
@@ -93,9 +93,9 @@ noop_src_iter_init (pixman_implementation_t *imp, pixman_iter_t *iter)
 	}
 	else
 	{
-	    uint64_t color = bits->fetch_pixel_64 (bits, 0, 0);
-	    uint64_t *buffer = (uint64_t *)iter->buffer;
-	    uint64_t *end = buffer + iter->width;
+	    argb_t color = bits->fetch_pixel_float (bits, 0, 0);
+	    argb_t *buffer = (argb_t *)iter->buffer;
+	    argb_t *end = buffer + iter->width;
 
 	    while (buffer < end)
 		*(buffer++) = color;
diff --git a/pixman/pixman/pixman-private.h b/pixman/pixman/pixman-private.h
index b9c8319df..dd03a939b 100644
--- a/pixman/pixman/pixman-private.h
+++ b/pixman/pixman/pixman-private.h
@@ -45,6 +45,16 @@ typedef struct radial_gradient radial_gradient_t;
 typedef struct bits_image bits_image_t;
 typedef struct circle circle_t;
 
+typedef struct argb_t argb_t;
+
+struct argb_t
+{
+    float a;
+    float r;
+    float g;
+    float b;
+};
+
 typedef void (*fetch_scanline_t) (pixman_image_t *image,
 				  int             x,
 				  int             y,
@@ -56,9 +66,9 @@ typedef uint32_t (*fetch_pixel_32_t) (bits_image_t *image,
 				      int           x,
 				      int           y);
 
-typedef uint64_t (*fetch_pixel_64_t) (bits_image_t *image,
-				      int           x,
-				      int           y);
+typedef argb_t (*fetch_pixel_float_t) (bits_image_t *image,
+				       int           x,
+				       int           y);
 
 typedef void (*store_scanline_t) (bits_image_t *  image,
 				  int             x,
@@ -112,9 +122,9 @@ struct solid_fill
 {
     image_common_t common;
     pixman_color_t color;
-    
+
     uint32_t	   color_32;
-    uint64_t	   color_64;
+    argb_t	   color_float;
 };
 
 struct gradient
@@ -173,9 +183,9 @@ struct bits_image
     fetch_pixel_32_t	       fetch_pixel_32;
     store_scanline_t           store_scanline_32;
 
-    fetch_scanline_t           fetch_scanline_64;
-    fetch_pixel_64_t	       fetch_pixel_64;
-    store_scanline_t           store_scanline_64;
+    fetch_scanline_t	       fetch_scanline_float;
+    fetch_pixel_float_t	       fetch_pixel_float;
+    store_scanline_t           store_scanline_float;
 
     /* Used for indirect access to the bits */
     pixman_read_memory_func_t  read_func;
@@ -414,12 +424,12 @@ typedef void (*pixman_combine_32_func_t) (pixman_implementation_t *imp,
 					  const uint32_t *         mask,
 					  int                      width);
 
-typedef void (*pixman_combine_64_func_t) (pixman_implementation_t *imp,
-					  pixman_op_t              op,
-					  uint64_t *               dest,
-					  const uint64_t *         src,
-					  const uint64_t *         mask,
-					  int                      width);
+typedef void (*pixman_combine_float_func_t) (pixman_implementation_t *imp,
+					     pixman_op_t	      op,
+					     float *		      dest,
+					     const float *	      src,
+					     const float *	      mask,
+					     int		      n_pixels);
 
 typedef void (*pixman_composite_func_t) (pixman_implementation_t *imp,
 					 pixman_composite_info_t *info);
@@ -449,7 +459,7 @@ typedef pixman_bool_t (*pixman_iter_init_func_t) (pixman_implementation_t *imp,
 						  pixman_iter_t           *iter);
 
 void _pixman_setup_combiner_functions_32 (pixman_implementation_t *imp);
-void _pixman_setup_combiner_functions_64 (pixman_implementation_t *imp);
+void _pixman_setup_combiner_functions_float (pixman_implementation_t *imp);
 
 typedef struct
 {
@@ -476,8 +486,8 @@ struct pixman_implementation_t
 
     pixman_combine_32_func_t	combine_32[PIXMAN_N_OPERATORS];
     pixman_combine_32_func_t	combine_32_ca[PIXMAN_N_OPERATORS];
-    pixman_combine_64_func_t	combine_64[PIXMAN_N_OPERATORS];
-    pixman_combine_64_func_t	combine_64_ca[PIXMAN_N_OPERATORS];
+    pixman_combine_float_func_t	combine_float[PIXMAN_N_OPERATORS];
+    pixman_combine_float_func_t	combine_float_ca[PIXMAN_N_OPERATORS];
 };
 
 uint32_t
@@ -754,6 +764,12 @@ get_implementation (void)
     return global_implementation;
 }
 
+/* This function is exported for the sake of the test suite and not part
+ * of the ABI.
+ */
+PIXMAN_EXPORT pixman_implementation_t *
+_pixman_internal_only_get_implementation (void);
+
 /* Memory allocation helpers */
 void *
 pixman_malloc_ab (unsigned int n, unsigned int b);
@@ -772,15 +788,15 @@ _pixman_addition_overflows_int (unsigned int a, unsigned int b);
 
 /* Compositing utilities */
 void
-pixman_expand (uint64_t *           dst,
-               const uint32_t *     src,
-               pixman_format_code_t format,
-               int                  width);
+pixman_expand_to_float (argb_t               *dst,
+			const uint32_t       *src,
+			pixman_format_code_t  format,
+			int                   width);
 
 void
-pixman_contract (uint32_t *      dst,
-                 const uint64_t *src,
-                 int             width);
+pixman_contract_from_float (uint32_t     *dst,
+			    const argb_t *src,
+			    int           width);
 
 /* Region Helpers */
 pixman_bool_t
@@ -941,6 +957,9 @@ unorm_to_unorm (uint32_t val, int from_bits, int to_bits)
     return result;
 }
 
+uint16_t pixman_float_to_unorm (float f, int n_bits);
+float pixman_unorm_to_float (uint16_t u, int n_bits);
+
 /*
  * Various debugging code
  */
@@ -1089,18 +1108,6 @@ void pixman_timer_register (pixman_timer_t *timer);
 
 #endif /* PIXMAN_TIMERS */
 
-/* sRGB<->linear conversion tables. Linear color space is the same
- * as sRGB but the components are in linear light (gamma 1.0).
- *
- * linear_to_srgb maps linear value from 0 to 4095 ([0.0, 1.0])
- * and returns 8-bit sRGB value.
- *
- * srgb_to_linear maps 8-bit sRGB value to 16-bit linear value
- * with range 0 to 65535 ([0.0, 1.0]).
- */
-extern const uint8_t linear_to_srgb[4096];
-extern const uint16_t srgb_to_linear[256];
-
 #endif /* __ASSEMBLER__ */
 
 #endif /* PIXMAN_PRIVATE_H */
diff --git a/pixman/pixman/pixman-radial-gradient.c b/pixman/pixman/pixman-radial-gradient.c
index 41bb79d73..8d562468d 100644
--- a/pixman/pixman/pixman-radial-gradient.c
+++ b/pixman/pixman/pixman-radial-gradient.c
@@ -405,7 +405,8 @@ radial_get_scanline_wide (pixman_iter_t *iter, const uint32_t *mask)
 {
     uint32_t *buffer = radial_get_scanline_narrow (iter, NULL);
 
-    pixman_expand ((uint64_t *)buffer, buffer, PIXMAN_a8r8g8b8, iter->width);
+    pixman_expand_to_float (
+	(argb_t *)buffer, buffer, PIXMAN_a8r8g8b8, iter->width);
 
     return buffer;
 }
diff --git a/pixman/pixman/pixman-solid-fill.c b/pixman/pixman/pixman-solid-fill.c
index 26f85cea3..60d56d52a 100644
--- a/pixman/pixman/pixman-solid-fill.c
+++ b/pixman/pixman/pixman-solid-fill.c
@@ -40,9 +40,9 @@ _pixman_solid_fill_iter_init (pixman_image_t *image, pixman_iter_t  *iter)
     }
     else
     {
-	uint64_t *b = (uint64_t *)iter->buffer;
-	uint64_t *e = b + iter->width;
-	uint64_t color = image->solid.color_64;
+	argb_t *b = (argb_t *)iter->buffer;
+	argb_t *e = b + iter->width;
+	argb_t color = image->solid.color_float;
 
 	while (b < e)
 	    *(b++) = color;
@@ -61,14 +61,17 @@ color_to_uint32 (const pixman_color_t *color)
         (color->blue >> 8);
 }
 
-static uint64_t
-color_to_uint64 (const pixman_color_t *color)
+static argb_t
+color_to_float (const pixman_color_t *color)
 {
-    return
-        ((uint64_t)color->alpha << 48) |
-        ((uint64_t)color->red << 32) |
-        ((uint64_t)color->green << 16) |
-        ((uint64_t)color->blue);
+    argb_t result;
+
+    result.a = pixman_unorm_to_float (color->alpha, 16);
+    result.r = pixman_unorm_to_float (color->red, 16);
+    result.g = pixman_unorm_to_float (color->green, 16);
+    result.b = pixman_unorm_to_float (color->blue, 16);
+
+    return result;
 }
 
 PIXMAN_EXPORT pixman_image_t *
@@ -82,7 +85,7 @@ pixman_image_create_solid_fill (const pixman_color_t *color)
     img->type = SOLID;
     img->solid.color = *color;
     img->solid.color_32 = color_to_uint32 (color);
-    img->solid.color_64 = color_to_uint64 (color);
+    img->solid.color_float = color_to_float (color);
 
     return img;
 }
diff --git a/pixman/pixman/pixman-utils.c b/pixman/pixman/pixman-utils.c
index 5633f8f79..b1e9fb62d 100644
--- a/pixman/pixman/pixman-utils.c
+++ b/pixman/pixman/pixman-utils.c
@@ -71,59 +71,96 @@ pixman_malloc_abc (unsigned int a,
 	return malloc (a * b * c);
 }
 
-static void
-unorm_to_unorm_params (int in_width, int out_width, uint32_t *factor, int *shift)
+static force_inline uint16_t
+float_to_unorm (float f, int n_bits)
 {
-    int w = 0;
+    uint32_t u;
 
-    *factor = 0;
-    while (in_width != 0 && w < out_width)
-    {
-	*factor |= 1 << w;
-	w += in_width;
-    }
+    if (f > 1.0)
+	f = 1.0;
+    if (f < 0.0)
+	f = 0.0;
+
+    u = f * (1 << n_bits);
+    u -= (u >> n_bits);
+
+    return u;
+}
+
+static force_inline float
+unorm_to_float (uint16_t u, int n_bits)
+{
+    uint32_t m = ((1 << n_bits) - 1);
 
-    /* Did we generate too many bits? */
-    *shift = w - out_width;
+    return (u & m) * (1.f / (float)m);
 }
 
 /*
- * This function expands images from ARGB8 format to ARGB16.  To preserve
- * precision, it needs to know the original source format.  For example, if the
- * source was PIXMAN_x1r5g5b5 and the red component contained bits 12345, then
- * the expanded value is 12345123.  To correctly expand this to 16 bits, it
- * should be 1234512345123451 and not 1234512312345123.
+ * This function expands images from a8r8g8b8 to argb_t.  To preserve
+ * precision, it needs to know from which source format the a8r8g8b8 pixels
+ * originally came.
+ *
+ * For example, if the source was PIXMAN_x1r5g5b5 and the red component
+ * contained bits 12345, then the 8-bit value is 12345123.  To correctly
+ * expand this to floating point, it should be 12345 / 31.0 and not
+ * 12345123 / 255.0.
  */
 void
-pixman_expand (uint64_t *           dst,
-               const uint32_t *     src,
-               pixman_format_code_t format,
-               int                  width)
+pixman_expand_to_float (argb_t               *dst,
+			const uint32_t       *src,
+			pixman_format_code_t  format,
+			int                   width)
 {
+    static const float multipliers[16] = {
+	0.0f,
+	1.0f / ((1 <<  1) - 1),
+	1.0f / ((1 <<  2) - 1),
+	1.0f / ((1 <<  3) - 1),
+	1.0f / ((1 <<  4) - 1),
+	1.0f / ((1 <<  5) - 1),
+	1.0f / ((1 <<  6) - 1),
+	1.0f / ((1 <<  7) - 1),
+	1.0f / ((1 <<  8) - 1),
+	1.0f / ((1 <<  9) - 1),
+	1.0f / ((1 << 10) - 1),
+	1.0f / ((1 << 11) - 1),
+	1.0f / ((1 << 12) - 1),
+	1.0f / ((1 << 13) - 1),
+	1.0f / ((1 << 14) - 1),
+	1.0f / ((1 << 15) - 1),
+    };
+    int a_size, r_size, g_size, b_size;
+    int a_shift, r_shift, g_shift, b_shift;
+    float a_mul, r_mul, g_mul, b_mul;
+    uint32_t a_mask, r_mask, g_mask, b_mask;
+    int i;
+
+    if (!PIXMAN_FORMAT_VIS (format))
+	format = PIXMAN_a8r8g8b8;
+
     /*
      * Determine the sizes of each component and the masks and shifts
      * required to extract them from the source pixel.
      */
-    const int a_size = PIXMAN_FORMAT_A (format),
-              r_size = PIXMAN_FORMAT_R (format),
-              g_size = PIXMAN_FORMAT_G (format),
-              b_size = PIXMAN_FORMAT_B (format);
-    const int a_shift = 32 - a_size,
-              r_shift = 24 - r_size,
-              g_shift = 16 - g_size,
-              b_shift =  8 - b_size;
-    const uint8_t a_mask = ~(~0 << a_size),
-                  r_mask = ~(~0 << r_size),
-                  g_mask = ~(~0 << g_size),
-                  b_mask = ~(~0 << b_size);
-    uint32_t au_factor, ru_factor, gu_factor, bu_factor;
-    int au_shift, ru_shift, gu_shift, bu_shift;
-    int i;
-
-    unorm_to_unorm_params (a_size, 16, &au_factor, &au_shift);
-    unorm_to_unorm_params (r_size, 16, &ru_factor, &ru_shift);
-    unorm_to_unorm_params (g_size, 16, &gu_factor, &gu_shift);
-    unorm_to_unorm_params (b_size, 16, &bu_factor, &bu_shift);
+    a_size = PIXMAN_FORMAT_A (format);
+    r_size = PIXMAN_FORMAT_R (format);
+    g_size = PIXMAN_FORMAT_G (format);
+    b_size = PIXMAN_FORMAT_B (format);
+
+    a_shift = 32 - a_size;
+    r_shift = 24 - r_size;
+    g_shift = 16 - g_size;
+    b_shift =  8 - b_size;
+
+    a_mask = ((1 << a_size) - 1);
+    r_mask = ((1 << r_size) - 1);
+    g_mask = ((1 << g_size) - 1);
+    b_mask = ((1 << b_size) - 1);
+
+    a_mul = multipliers[a_size];
+    r_mul = multipliers[r_size];
+    g_mul = multipliers[g_size];
+    b_mul = multipliers[b_size];
 
     /* Start at the end so that we can do the expansion in place
      * when src == dst
@@ -131,59 +168,43 @@ pixman_expand (uint64_t *           dst,
     for (i = width - 1; i >= 0; i--)
     {
 	const uint32_t pixel = src[i];
-	uint8_t a, r, g, b;
-	uint64_t a16, r16, g16, b16;
-
-	if (a_size)
-	{
-	    a = (pixel >> a_shift) & a_mask;
-            a16 = a * au_factor >> au_shift;
-	}
-	else
-	{
-	    a16 = 0xffff;
-	}
-
-	if (r_size)
-	{
-	    r = (pixel >> r_shift) & r_mask;
-	    g = (pixel >> g_shift) & g_mask;
-	    b = (pixel >> b_shift) & b_mask;
-            r16 = r * ru_factor >> ru_shift;
-            g16 = g * gu_factor >> gu_shift;
-            b16 = b * bu_factor >> bu_shift;
-	}
-	else
-	{
-	    r16 = g16 = b16 = 0;
-	}
-	
-	dst[i] = a16 << 48 | r16 << 32 | g16 << 16 | b16;
+
+	dst[i].a = a_mask? ((pixel >> a_shift) & a_mask) * a_mul : 1.0f;
+	dst[i].r = ((pixel >> r_shift) & r_mask) * r_mul;
+	dst[i].g = ((pixel >> g_shift) & g_mask) * g_mul;
+	dst[i].b = ((pixel >> b_shift) & b_mask) * b_mul;
     }
 }
 
-/*
- * Contracting is easier than expanding.  We just need to truncate the
- * components.
- */
+uint16_t
+pixman_float_to_unorm (float f, int n_bits)
+{
+    return float_to_unorm (f, n_bits);
+}
+
+float
+pixman_unorm_to_float (uint16_t u, int n_bits)
+{
+    return unorm_to_float (u, n_bits);
+}
+
 void
-pixman_contract (uint32_t *      dst,
-                 const uint64_t *src,
-                 int             width)
+pixman_contract_from_float (uint32_t     *dst,
+			    const argb_t *src,
+			    int           width)
 {
     int i;
 
-    /* Start at the beginning so that we can do the contraction in
-     * place when src == dst
-     */
-    for (i = 0; i < width; i++)
+    for (i = 0; i < width; ++i)
     {
-	const uint8_t a = src[i] >> 56,
-	              r = src[i] >> 40,
-	              g = src[i] >> 24,
-	              b = src[i] >> 8;
+	uint8_t a, r, g, b;
 
-	dst[i] = a << 24 | r << 16 | g << 8 | b;
+	a = float_to_unorm (src[i].a, 8);
+	r = float_to_unorm (src[i].r, 8);
+	g = float_to_unorm (src[i].g, 8);
+	b = float_to_unorm (src[i].b, 8);
+
+	dst[i] = (a << 24) | (r << 16) | (g << 8) | (b << 0);
     }
 }
 
@@ -262,6 +283,15 @@ pixman_region32_copy_from_region16 (pixman_region32_t *dst,
     return retval;
 }
 
+/* This function is exported for the sake of the test suite and not part
+ * of the ABI.
+ */
+PIXMAN_EXPORT pixman_implementation_t *
+_pixman_internal_only_get_implementation (void)
+{
+    return get_implementation ();
+}
+
 #ifdef DEBUG
 
 void
diff --git a/pixman/test/Makefile.sources b/pixman/test/Makefile.sources
index 0f344116d..077897161 100644
--- a/pixman/test/Makefile.sources
+++ b/pixman/test/Makefile.sources
@@ -4,6 +4,7 @@ TESTPROGRAMS =			\
 	pdf-op-test		\
 	region-test		\
 	region-translate-test	\
+	combiner-test		\
 	fetch-test		\
 	rotate-test		\
 	oob-test		\
diff --git a/pixman/test/blitters-test.c b/pixman/test/blitters-test.c
index 8c46cef97..30d69124c 100644
--- a/pixman/test/blitters-test.c
+++ b/pixman/test/blitters-test.c
@@ -172,10 +172,12 @@ static pixman_format_code_t img_fmt_list[] = {
     PIXMAN_x14r6g6b6,
     PIXMAN_r8g8b8,
     PIXMAN_b8g8r8,
+#if 0 /* These are going to use floating point in the near future */
     PIXMAN_x2r10g10b10,
     PIXMAN_a2r10g10b10,
     PIXMAN_x2b10g10r10,
     PIXMAN_a2b10g10r10,
+#endif
     PIXMAN_a1r5g5b5,
     PIXMAN_x1r5g5b5,
     PIXMAN_a1b5g5r5,
@@ -395,6 +397,6 @@ main (int argc, const char *argv[])
     }
 
     return fuzzer_test_main("blitters", 2000000,
-			    0x3E1DD2E8,
+			    0x46136E0A,
 			    test_composite, argc, argv);
 }
diff --git a/pixman/test/combiner-test.c b/pixman/test/combiner-test.c
new file mode 100644
index 000000000..c438ae62e
--- /dev/null
+++ b/pixman/test/combiner-test.c
@@ -0,0 +1,151 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include "utils.h"
+#include <sys/types.h>
+#include "pixman-private.h"
+
+static const pixman_op_t op_list[] =
+{
+    PIXMAN_OP_SRC,
+    PIXMAN_OP_OVER,
+    PIXMAN_OP_ADD,
+    PIXMAN_OP_CLEAR,
+    PIXMAN_OP_SRC,
+    PIXMAN_OP_DST,
+    PIXMAN_OP_OVER,
+    PIXMAN_OP_OVER_REVERSE,
+    PIXMAN_OP_IN,
+    PIXMAN_OP_IN_REVERSE,
+    PIXMAN_OP_OUT,
+    PIXMAN_OP_OUT_REVERSE,
+    PIXMAN_OP_ATOP,
+    PIXMAN_OP_ATOP_REVERSE,
+    PIXMAN_OP_XOR,
+    PIXMAN_OP_ADD,
+    PIXMAN_OP_SATURATE,
+    PIXMAN_OP_DISJOINT_CLEAR,
+    PIXMAN_OP_DISJOINT_SRC,
+    PIXMAN_OP_DISJOINT_DST,
+    PIXMAN_OP_DISJOINT_OVER,
+    PIXMAN_OP_DISJOINT_OVER_REVERSE,
+    PIXMAN_OP_DISJOINT_IN,
+    PIXMAN_OP_DISJOINT_IN_REVERSE,
+    PIXMAN_OP_DISJOINT_OUT,
+    PIXMAN_OP_DISJOINT_OUT_REVERSE,
+    PIXMAN_OP_DISJOINT_ATOP,
+    PIXMAN_OP_DISJOINT_ATOP_REVERSE,
+    PIXMAN_OP_DISJOINT_XOR,
+    PIXMAN_OP_CONJOINT_CLEAR,
+    PIXMAN_OP_CONJOINT_SRC,
+    PIXMAN_OP_CONJOINT_DST,
+    PIXMAN_OP_CONJOINT_OVER,
+    PIXMAN_OP_CONJOINT_OVER_REVERSE,
+    PIXMAN_OP_CONJOINT_IN,
+    PIXMAN_OP_CONJOINT_IN_REVERSE,
+    PIXMAN_OP_CONJOINT_OUT,
+    PIXMAN_OP_CONJOINT_OUT_REVERSE,
+    PIXMAN_OP_CONJOINT_ATOP,
+    PIXMAN_OP_CONJOINT_ATOP_REVERSE,
+    PIXMAN_OP_CONJOINT_XOR,
+    PIXMAN_OP_MULTIPLY,
+    PIXMAN_OP_SCREEN,
+    PIXMAN_OP_OVERLAY,
+    PIXMAN_OP_DARKEN,
+    PIXMAN_OP_LIGHTEN,
+    PIXMAN_OP_COLOR_DODGE,
+    PIXMAN_OP_COLOR_BURN,
+    PIXMAN_OP_HARD_LIGHT,
+    PIXMAN_OP_DIFFERENCE,
+    PIXMAN_OP_EXCLUSION,
+    PIXMAN_OP_SOFT_LIGHT,
+    PIXMAN_OP_HSL_HUE,
+    PIXMAN_OP_HSL_SATURATION,
+    PIXMAN_OP_HSL_COLOR,
+    PIXMAN_OP_HSL_LUMINOSITY,
+};
+
+static float
+rand_float (void)
+{
+    uint32_t u = lcg_rand_u32();
+
+    return *(float *)&u;
+}
+
+static void
+random_floats (argb_t *argb, int width)
+{
+    int i;
+
+    for (i = 0; i < width; ++i)
+    {
+	argb_t *p = argb + i;
+
+	p->a = rand_float();
+	p->r = rand_float();
+	p->g = rand_float();
+	p->b = rand_float();
+    }
+}
+
+#define WIDTH	512
+
+static pixman_combine_float_func_t
+lookup_combiner (pixman_implementation_t *imp, pixman_op_t op,
+		 pixman_bool_t component_alpha)
+{
+    pixman_combine_float_func_t f;
+
+    do
+    {
+	if (component_alpha)
+	    f = imp->combine_float_ca[op];
+	else
+	    f = imp->combine_float[op];
+	
+	imp = imp->fallback;
+    }
+    while (!f);
+
+    return f;
+}
+
+int
+main ()
+{
+    pixman_implementation_t *impl;
+    argb_t *src_bytes = malloc (WIDTH * sizeof (argb_t));
+    argb_t *mask_bytes = malloc (WIDTH * sizeof (argb_t));
+    argb_t *dest_bytes = malloc (WIDTH * sizeof (argb_t));
+    int i;
+
+    enable_divbyzero_exceptions();
+    
+    impl = _pixman_internal_only_get_implementation();
+    
+    lcg_srand (0);
+
+    for (i = 0; i < ARRAY_LENGTH (op_list); ++i)
+    {
+	pixman_op_t op = op_list[i];
+	pixman_combine_float_func_t combiner;
+	int ca;
+
+	for (ca = 0; ca < 2; ++ca)
+	{
+	    combiner = lookup_combiner (impl, op, ca);
+
+	    random_floats (src_bytes, WIDTH);
+	    random_floats (mask_bytes, WIDTH);
+	    random_floats (dest_bytes, WIDTH);
+
+	    combiner (impl, op,
+		      (float *)dest_bytes,
+		      (float *)mask_bytes,
+		      (float *)src_bytes,
+		      WIDTH);
+	}
+    }	
+
+    return 0;
+}
diff --git a/pixman/test/glyph-test.c b/pixman/test/glyph-test.c
index 84de5aace..9dd5b41e4 100644
--- a/pixman/test/glyph-test.c
+++ b/pixman/test/glyph-test.c
@@ -30,10 +30,13 @@ static const pixman_format_code_t formats[] =
     PIXMAN_x14r6g6b6,
     PIXMAN_r8g8b8,
     PIXMAN_b8g8r8,
+#if 0
+    /* These use floating point */
     PIXMAN_x2r10g10b10,
     PIXMAN_a2r10g10b10,
     PIXMAN_x2b10g10r10,
     PIXMAN_a2b10g10r10,
+#endif
     PIXMAN_a1r5g5b5,
     PIXMAN_x1r5g5b5,
     PIXMAN_a1b5g5r5,
@@ -329,7 +332,7 @@ test_glyphs (int testnum, int verbose)
 int
 main (int argc, const char *argv[])
 {
-    return fuzzer_test_main ("glyph", 30000,
-			     0x741CB2DB,
+    return fuzzer_test_main ("glyph", 30000,	
+			     0x79E74996,
 			     test_glyphs, argc, argv);
 }
diff --git a/pixman/test/gradient-crash-test.c b/pixman/test/gradient-crash-test.c
index 73e5bbcd5..962d1cbe8 100644
--- a/pixman/test/gradient-crash-test.c
+++ b/pixman/test/gradient-crash-test.c
@@ -85,7 +85,7 @@ main (int argc, char **argv)
     pixman_fixed_t r_inner;
     pixman_fixed_t r_outer;
 
-    enable_fp_exceptions();
+    enable_divbyzero_exceptions();
     
     for (i = 0; i < WIDTH * HEIGHT; ++i)
 	dest[i] = 0x4f00004f; /* pale blue */
diff --git a/pixman/test/pdf-op-test.c b/pixman/test/pdf-op-test.c
index 99cb7dfea..dcb3a603a 100644
--- a/pixman/test/pdf-op-test.c
+++ b/pixman/test/pdf-op-test.c
@@ -36,7 +36,7 @@ main ()
 {
     int o, s, m, d;
 
-    enable_fp_exceptions();
+    enable_divbyzero_exceptions();
 
     for (o = 0; o < ARRAY_LENGTH (pdf_ops); ++o)
     {
diff --git a/pixman/test/stress-test.c b/pixman/test/stress-test.c
index edcfe0947..059250dd4 100644
--- a/pixman/test/stress-test.c
+++ b/pixman/test/stress-test.c
@@ -850,7 +850,7 @@ main (int argc, char **argv)
 
     pixman_disable_out_of_bounds_workaround ();
 
-    enable_fp_exceptions();
+    enable_divbyzero_exceptions();
 
     if (getenv ("VERBOSE") != NULL)
 	verbose = TRUE;
diff --git a/pixman/test/utils.c b/pixman/test/utils.c
index c922ae5d5..716bb7594 100644
--- a/pixman/test/utils.c
+++ b/pixman/test/utils.c
@@ -723,21 +723,11 @@ fail_after (int seconds, const char *msg)
 }
 
 void
-enable_fp_exceptions (void)
+enable_divbyzero_exceptions (void)
 {
 #ifdef HAVE_FENV_H
 #ifdef HAVE_FEENABLEEXCEPT
-    /* Note: we don't enable the FE_INEXACT trap because
-     * that happens quite commonly. It is possible that
-     * over- and underflow should similarly be considered
-     * okay, but for now the test suite passes with them
-     * enabled, and it's useful to know if they start
-     * occuring.
-     */
-    feenableexcept (FE_DIVBYZERO	|
-		    FE_INVALID		|
-		    FE_OVERFLOW		|
-		    FE_UNDERFLOW);
+    feenableexcept (FE_DIVBYZERO);
 #endif
 #endif
 }
diff --git a/pixman/test/utils.h b/pixman/test/utils.h
index faf427f0a..f7ea34c5f 100644
--- a/pixman/test/utils.h
+++ b/pixman/test/utils.h
@@ -110,7 +110,7 @@ void
 fail_after (int seconds, const char *msg);
 
 /* If possible, enable traps for floating point exceptions */
-void enable_fp_exceptions(void);
+void enable_divbyzero_exceptions(void);
 
 /* Converts a8r8g8b8 pixels to pixels that
  *  - are not premultiplied,
diff --git a/xorg-server/Xext/xf86bigfont.c b/xorg-server/Xext/xf86bigfont.c
index 2bc5fc93f..42f5549be 100644
--- a/xorg-server/Xext/xf86bigfont.c
+++ b/xorg-server/Xext/xf86bigfont.c
@@ -278,9 +278,10 @@ ProcXF86BigfontQueryVersion(ClientPtr client)
     xXF86BigfontQueryVersionReply reply;
 
     REQUEST_SIZE_MATCH(xXF86BigfontQueryVersionReq);
+
     reply.type = X_Reply;
-    reply.length = 0;
     reply.sequenceNumber = client->sequence;
+    reply.length = 0;
     reply.majorVersion = SERVER_XF86BIGFONT_MAJOR_VERSION;
     reply.minorVersion = SERVER_XF86BIGFONT_MINOR_VERSION;
     #ifdef WIN32
@@ -292,7 +293,8 @@ ProcXF86BigfontQueryVersion(ClientPtr client)
     #endif
 #ifdef HAS_SHM
     reply.signature = signature;
-    reply.capabilities = (LocalClient(client) && !client->swapped ? XF86Bigfont_CAP_LocalShm : 0);
+    reply.capabilities = (client->local && !client->swapped)
+                         ? XF86Bigfont_CAP_LocalShm : 0
 #else
     reply.signature = 0;        /* This is redundant. Avoids uninitialized memory. */
     reply.capabilities = 0;
@@ -360,7 +362,7 @@ ProcXF86BigfontQueryFont(ClientPtr client)
 #else
     switch (client->req_len) {
     case 2:                    /* client with version 1.0 libX11 */
-        stuff_flags = (LocalClient(client) &&
+        stuff_flags = (client->local &&
                        !client->swapped ? XF86Bigfont_FLAGS_Shm : 0);
         break;
     case 3:                    /* client with version 1.1 libX11 */
diff --git a/xorg-server/Xi/exevents.c b/xorg-server/Xi/exevents.c
index 9483180a7..2a097d74f 100644
--- a/xorg-server/Xi/exevents.c
+++ b/xorg-server/Xi/exevents.c
@@ -951,8 +951,6 @@ UpdateDeviceState(DeviceIntPtr device, DeviceEvent *event)
 
         if (!(event->flags & TOUCH_POINTER_EMULATED))
             return DONT_PROCESS;
-        if (!(event->flags & TOUCH_END))
-            return DONT_PROCESS;
 
         DecreaseButtonCount(device, key, &t->buttonsDown, &t->motionMask,
                             &t->state);
diff --git a/xorg-server/Xi/xiselectev.c b/xorg-server/Xi/xiselectev.c
index 0e753c60f..3040aa911 100644
--- a/xorg-server/Xi/xiselectev.c
+++ b/xorg-server/Xi/xiselectev.c
@@ -180,8 +180,13 @@ ProcXISelectEvents(ClientPtr client)
                     if (CLIENT_ID(iclient->resource) == client->index)
                         continue;
 
-                    dixLookupDevice(&tmp, evmask->deviceid, serverClient,
-                                    DixReadAccess);
+                    if (evmask->deviceid == XIAllDevices)
+                        tmp = inputInfo.all_devices;
+                    else if (evmask->deviceid == XIAllMasterDevices)
+                        tmp = inputInfo.all_master_devices;
+                    else
+                        dixLookupDevice(&tmp, evmask->deviceid, serverClient,
+                                        DixReadAccess);
                     if (!tmp)
                         return BadImplementation;       /* this shouldn't happen */
 
diff --git a/xorg-server/autogen.sh b/xorg-server/autogen.sh
index 588fef395..8b6f58f69 100644
--- a/xorg-server/autogen.sh
+++ b/xorg-server/autogen.sh
@@ -1,12 +1,12 @@
-#! /bin/sh

-

-srcdir=`dirname $0`

-test -z "$srcdir" && srcdir=.

-

-ORIGDIR=`pwd`

-cd $srcdir

-

-autoreconf --force -v --install || exit 1

-cd $ORIGDIR || exit $?

-

-$srcdir/configure --enable-maintainer-mode "$@"

+#! /bin/sh
+
+srcdir=`dirname $0`
+test -z "$srcdir" && srcdir=.
+
+ORIGDIR=`pwd`
+cd $srcdir
+
+autoreconf --force -v --install || exit 1
+cd $ORIGDIR || exit $?
+
+$srcdir/configure "$@"
diff --git a/xorg-server/configure.ac b/xorg-server/configure.ac
index ac3bf26f1..e6866140e 100644
--- a/xorg-server/configure.ac
+++ b/xorg-server/configure.ac
@@ -26,12 +26,11 @@ dnl
 dnl Process this file with autoconf to create configure.
 
 AC_PREREQ(2.60)
-AC_INIT([xorg-server], 1.13.0, [https://bugs.freedesktop.org/enter_bug.cgi?product=xorg], xorg-server)
-RELEASE_DATE="2012-09-05"
-RELEASE_NAME="Iced Tea"
+AC_INIT([xorg-server], 1.13.99.0, [https://bugs.freedesktop.org/enter_bug.cgi?product=xorg], xorg-server)
+RELEASE_DATE="2012-10-04"
+RELEASE_NAME="Horchata"
 AC_CONFIG_SRCDIR([Makefile.am])
 AM_INIT_AUTOMAKE([foreign dist-bzip2])
-AM_MAINTAINER_MODE
 
 # Require xorg-macros minimum of 1.14 for XORG_COMPILER_BRAND in XORG_DEFAULT_OPTIONS
 m4_ifndef([XORG_MACROS_VERSION],
diff --git a/xorg-server/dix/dispatch.c b/xorg-server/dix/dispatch.c
index 1221f044d..5c17d3750 100644
--- a/xorg-server/dix/dispatch.c
+++ b/xorg-server/dix/dispatch.c
@@ -3542,14 +3542,16 @@ ProcInitialConnection(ClientPtr client)
     REQUEST(xReq);

     xConnClientPrefix *prefix;

     int whichbyte = 1;

+    char order;

 

-    prefix = (xConnClientPrefix *) ((char *) stuff + sz_xReq);

-    if ((prefix->byteOrder != 'l') && (prefix->byteOrder != 'B'))

-        return client->noClientException = -1;

-    if (((*(char *) &whichbyte) && (prefix->byteOrder == 'B')) ||

-        (!(*(char *) &whichbyte) && (prefix->byteOrder == 'l'))) {

-        client->swapped = TRUE;

-        SwapConnClientPrefix(prefix);

+    prefix = (xConnClientPrefix *) ((char *)stuff + sz_xReq);

+    order = prefix->byteOrder;

+    if (order != 'l' && order != 'B' && order != 'r' && order != 'R')

+	return client->noClientException = -1;

+    if (((*(char *) &whichbyte) && (order == 'B' || order == 'R')) ||

+	(!(*(char *) &whichbyte) && (order == 'l' || order == 'r'))) {

+	client->swapped = TRUE;

+	SwapConnClientPrefix(prefix);

     }

     stuff->reqType = 2;

     stuff->length += bytes_to_int32(prefix->nbytesAuthProto) +

@@ -3557,6 +3559,9 @@ ProcInitialConnection(ClientPtr client)
     if (client->swapped) {

         swaps(&stuff->length);

     }

+    if (order == 'r' || order == 'R') {

+	client->local = FALSE;

+    }

     ResetCurrentRequest(client);

     return Success;

 }

diff --git a/xorg-server/dix/events.c b/xorg-server/dix/events.c
index 005db808f..dd9dce7bb 100644
--- a/xorg-server/dix/events.c
+++ b/xorg-server/dix/events.c
@@ -1558,11 +1558,13 @@ ActivateKeyboardGrab(DeviceIntPtr keybd, GrabPtr grab, TimeStamp time,
     WindowPtr oldWin;
 
     /* slave devices need to float for the duration of the grab. */
-    if (grab->grabtype == XI2 &&
+    if (grab->grabtype == XI2 && keybd->enabled &&
         !(passive & ImplicitGrabMask) && !IsMaster(keybd))
         DetachFromMaster(keybd);
 
-    if (grabinfo->grab)
+    if (!keybd->enabled)
+        oldWin = NULL;
+    else if (grabinfo->grab)
         oldWin = grabinfo->grab->window;
     else if (keybd->focus)
         oldWin = keybd->focus->win;
@@ -1572,7 +1574,8 @@ ActivateKeyboardGrab(DeviceIntPtr keybd, GrabPtr grab, TimeStamp time,
         oldWin = keybd->focus->win;
     if (keybd->valuator)
         keybd->valuator->motionHintWindow = NullWindow;
-    DoFocusEvents(keybd, oldWin, grab->window, NotifyGrab);
+    if (oldWin)
+        DoFocusEvents(keybd, oldWin, grab->window, NotifyGrab);
     if (syncEvents.playingEvents)
         grabinfo->grabTime = syncEvents.time;
     else
diff --git a/xorg-server/hw/kdrive/ephyr/ephyrdriext.c b/xorg-server/hw/kdrive/ephyr/ephyrdriext.c
index a42be07d5..1a98a2dff 100644
--- a/xorg-server/hw/kdrive/ephyr/ephyrdriext.c
+++ b/xorg-server/hw/kdrive/ephyr/ephyrdriext.c
@@ -561,7 +561,7 @@ ProcXF86DRIQueryDirectRenderingCapable(register ClientPtr client)
         return BadValue;
     }
 
-    if (!LocalClient(client) || client->swapped)
+    if (!client->local || client->swapped)
         isCapable = 0;
 
     rep = (xXF86DRIQueryDirectRenderingCapableReply) {
@@ -1229,7 +1229,7 @@ ProcXF86DRIDispatch(register ClientPtr client)
     }
     }
 
-    if (!LocalClient(client))
+    if (!client->local)
         return DRIErrorBase + XF86DRIClientNotLocal;
 
     switch (stuff->data) {
diff --git a/xorg-server/hw/kdrive/ephyr/ephyrglxext.c b/xorg-server/hw/kdrive/ephyr/ephyrglxext.c
index df285cfcb..22d510891 100644
--- a/xorg-server/hw/kdrive/ephyr/ephyrglxext.c
+++ b/xorg-server/hw/kdrive/ephyr/ephyrglxext.c
@@ -61,10 +61,16 @@ int ephyrGLXGetFBConfigsSGIX(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXGetFBConfigsSGIXSwap(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXCreateContext(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXCreateContextSwap(__GLXclientState * a_cl, GLbyte * a_pc);
+int ephyrGLXCreateNewContext(__GLXclientState * a_cl, GLbyte * a_pc);
+int ephyrGLXCreateNewContextSwap(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXDestroyContext(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXDestroyContextSwap(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXMakeCurrent(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXMakeCurrentSwap(__GLXclientState * a_cl, GLbyte * a_pc);
+int ephyrGLXMakeCurrentReadSGI(__GLXclientState * a_cl, GLbyte * a_pc);
+int ephyrGLXMakeCurrentReadSGISwap(__GLXclientState * a_cl, GLbyte * a_pc);
+int ephyrGLXMakeContextCurrent(__GLXclientState * a_cl, GLbyte * a_pc);
+int ephyrGLXMakeContextCurrentSwap(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXGetString(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXGetStringSwap(__GLXclientState * a_cl, GLbyte * a_pc);
 int ephyrGLXGetIntegerv(__GLXclientState * a_cl, GLbyte * a_pc);
@@ -108,6 +114,9 @@ ephyrHijackGLXExtension(void)
     dispatch_functions[X_GLXCreateContext][0] = ephyrGLXCreateContext;
     dispatch_functions[X_GLXCreateContext][1] = ephyrGLXCreateContextSwap;
 
+    dispatch_functions[X_GLXCreateNewContext][0] = ephyrGLXCreateNewContext;
+    dispatch_functions[X_GLXCreateNewContext][1] = ephyrGLXCreateNewContextSwap;
+
     dispatch_functions[X_GLXDestroyContext][0] = ephyrGLXDestroyContext;
     dispatch_functions[X_GLXDestroyContext][1] = ephyrGLXDestroyContextSwap;
 
@@ -123,14 +132,24 @@ ephyrHijackGLXExtension(void)
     dispatch_functions[61][0] = ephyrGLXGetIntegerv;
     dispatch_functions[61][1] = ephyrGLXGetIntegervSwap;
 
+    dispatch_functions[X_GLXMakeContextCurrent][0] =
+        ephyrGLXMakeContextCurrent;
+    dispatch_functions[X_GLXMakeContextCurrent][1] =
+        ephyrGLXMakeContextCurrentSwap;
+
     /*
      * hijack some vendor priv entry point dispatch functions
      */
     dispatch_functions = VendorPriv_dispatch_info.dispatch_functions;
     dispatch_functions[92][0] = ephyrGLXGetFBConfigsSGIX;
     dispatch_functions[92][1] = ephyrGLXGetFBConfigsSGIXSwap;
+
+    dispatch_functions[89][0] = ephyrGLXMakeCurrentReadSGI;
+    dispatch_functions[89][1] = ephyrGLXMakeCurrentReadSGISwap;
+
     EPHYR_LOG("hijacked glx entry points to forward requests to host X\n");
 
+
     return TRUE;
 }
 
@@ -445,7 +464,8 @@ ephyrGLXCreateContextReal(xGLXCreateContextReq * a_req, Bool a_do_swap)
     if (!ephyrHostGLXCreateContext(a_req->screen,
                                    host_w_attrs.visualid,
                                    a_req->context,
-                                   a_req->shareList, a_req->isDirect)) {
+                                   a_req->shareList, 0,
+                                   a_req->isDirect, X_GLXCreateContext)) {
         EPHYR_LOG_ERROR("ephyrHostGLXCreateContext() failed\n");
         goto out;
     }
@@ -455,6 +475,45 @@ ephyrGLXCreateContextReal(xGLXCreateContextReq * a_req, Bool a_do_swap)
     return res;
 }
 
+static int
+ephyrGLXCreateNewContextReal(xGLXCreateNewContextReq * a_req, Bool a_do_swap)
+{
+    int res = BadImplementation;
+
+    __GLX_DECLARE_SWAP_VARIABLES;
+
+    EPHYR_RETURN_VAL_IF_FAIL(a_req, BadValue);
+    EPHYR_LOG("enter\n");
+
+    if (a_do_swap) {
+        __GLX_SWAP_SHORT(&a_req->length);
+        __GLX_SWAP_INT(&a_req->context);
+        __GLX_SWAP_INT(&a_req->fbconfig);
+        __GLX_SWAP_INT(&a_req->screen);
+        __GLX_SWAP_INT(&a_req->renderType);
+        __GLX_SWAP_INT(&a_req->shareList);
+    }
+
+    EPHYR_LOG("context creation requested. localid:%d, "
+              "screen:%d, fbconfig:%d, renderType:%d, direct:%d\n",
+              (int) a_req->context, (int) a_req->screen,
+              (int) a_req->fbconfig, (int) a_req->renderType,
+              (int) a_req->isDirect);
+
+    if (!ephyrHostGLXCreateContext(a_req->screen,
+                                   a_req->fbconfig,
+                                   a_req->context,
+                                   a_req->shareList, a_req->renderType,
+                                   a_req->isDirect, X_GLXCreateNewContext)) {
+        EPHYR_LOG_ERROR("ephyrHostGLXCreateNewContext() failed\n");
+        goto out;
+    }
+    res = Success;
+ out:
+    EPHYR_LOG("leave\n");
+    return res;
+}
+
 int
 ephyrGLXCreateContext(__GLXclientState * cl, GLbyte * pc)
 {
@@ -471,6 +530,22 @@ ephyrGLXCreateContextSwap(__GLXclientState * cl, GLbyte * pc)
     return ephyrGLXCreateContextReal(req, TRUE);
 }
 
+int
+ephyrGLXCreateNewContext(__GLXclientState * cl, GLbyte * pc)
+{
+    xGLXCreateNewContextReq *req = (xGLXCreateNewContextReq *) pc;
+
+    return ephyrGLXCreateNewContextReal(req, FALSE);
+}
+
+int
+ephyrGLXCreateNewContextSwap(__GLXclientState * cl, GLbyte * pc)
+{
+    xGLXCreateNewContextReq *req = (xGLXCreateNewContextReq *) pc;
+
+    return ephyrGLXCreateNewContextReal(req, TRUE);
+}
+
 static int
 ephyrGLXDestroyContextReal(__GLXclientState * a_cl,
                            GLbyte * a_pc, Bool a_do_swap)
@@ -505,26 +580,34 @@ ephyrGLXDestroyContextSwap(__GLXclientState * a_cl, GLbyte * a_pc)
 }
 
 static int
-ephyrGLXMakeCurrentReal(__GLXclientState * a_cl, GLbyte * a_pc, Bool a_do_swap)
+ephyrGLXMakeCurrentReal(__GLXclientState * a_cl, GLXDrawable write,
+                        GLXDrawable read, GLXContextTag ctx,
+                        GLXContextTag old_ctx, Bool a_do_swap)
 {
     int res = BadImplementation;
-    xGLXMakeCurrentReq *req = (xGLXMakeCurrentReq *) a_pc;
     xGLXMakeCurrentReply reply;
-    DrawablePtr drawable = NULL;
-    GLXContextTag contextTag = 0;
-    int rc = 0;
+    DrawablePtr drawableR = NULL, drawableW = NULL;
+    GLXContextTag new_ctx = 0;
 
     EPHYR_LOG("enter\n");
-    rc = dixLookupDrawable(&drawable,
-                           req->drawable, a_cl->client, 0, DixReadAccess);
-    EPHYR_RETURN_VAL_IF_FAIL(drawable, BadValue);
-    EPHYR_RETURN_VAL_IF_FAIL(drawable->pScreen, BadValue);
-    EPHYR_LOG("screen nummber requested:%d\n", drawable->pScreen->myNum);
-
-    if (!ephyrHostGLXMakeCurrent(hostx_get_window(drawable->pScreen->myNum),
-                                 req->context,
-                                 req->oldContextTag,
-                                 (int *) &contextTag)) {
+    res = dixLookupDrawable(&drawableW, write, a_cl->client, 0, DixReadAccess);
+    EPHYR_RETURN_VAL_IF_FAIL(drawableW, BadValue);
+    EPHYR_RETURN_VAL_IF_FAIL(drawableW->pScreen, BadValue);
+    EPHYR_LOG("screen nummber requested:%d\n", drawableW->pScreen->myNum);
+
+    if (read != write) {
+        res = dixLookupDrawable(&drawableR, read, a_cl->client, 0,
+                                DixReadAccess);
+        EPHYR_RETURN_VAL_IF_FAIL(drawableR, BadValue);
+        EPHYR_RETURN_VAL_IF_FAIL(drawableR->pScreen, BadValue);
+    }
+    else {
+        drawableR = drawableW;
+    }
+
+    if (!ephyrHostGLXMakeCurrent(hostx_get_window(drawableW->pScreen->myNum),
+                                 hostx_get_window(drawableR->pScreen->myNum),
+                                 ctx, old_ctx, (int *) &new_ctx)) {
         EPHYR_LOG_ERROR("ephyrHostGLXMakeCurrent() failed\n");
         goto out;
     }
@@ -532,7 +615,7 @@ ephyrGLXMakeCurrentReal(__GLXclientState * a_cl, GLbyte * a_pc, Bool a_do_swap)
         .type = X_Reply,
         .sequenceNumber = a_cl->client->sequence,
         .length = 0,
-        .contextTag = contextTag
+        .contextTag = new_ctx
     };
     if (a_do_swap) {
         __GLX_DECLARE_SWAP_VARIABLES;
@@ -551,13 +634,71 @@ ephyrGLXMakeCurrentReal(__GLXclientState * a_cl, GLbyte * a_pc, Bool a_do_swap)
 int
 ephyrGLXMakeCurrent(__GLXclientState * a_cl, GLbyte * a_pc)
 {
-    return ephyrGLXMakeCurrentReal(a_cl, a_pc, FALSE);
+    xGLXMakeCurrentReq *req = (xGLXMakeCurrentReq *) a_pc;
+    return ephyrGLXMakeCurrentReal(a_cl, req->drawable, req->drawable,
+                                   req->context, req->oldContextTag, FALSE);
 }
 
 int
 ephyrGLXMakeCurrentSwap(__GLXclientState * a_cl, GLbyte * a_pc)
 {
-    return ephyrGLXMakeCurrentReal(a_cl, a_pc, TRUE);
+    xGLXMakeCurrentReq *req = (xGLXMakeCurrentReq *) a_pc;
+    __GLX_DECLARE_SWAP_VARIABLES;
+
+    __GLX_SWAP_INT(&req->drawable);
+    __GLX_SWAP_INT(&req->context);
+    __GLX_SWAP_INT(&req->oldContextTag);
+
+    return ephyrGLXMakeCurrentReal(a_cl, req->drawable, req->drawable,
+                                   req->context, req->oldContextTag, TRUE);
+}
+
+int
+ephyrGLXMakeCurrentReadSGI(__GLXclientState * a_cl, GLbyte * a_pc)
+{
+    xGLXMakeCurrentReadSGIReq *req = (xGLXMakeCurrentReadSGIReq *) a_pc;
+
+    return ephyrGLXMakeCurrentReal(a_cl, req->drawable, req->readable,
+                                   req->context, req->oldContextTag, FALSE);
+}
+
+int
+ephyrGLXMakeCurrentReadSGISwap(__GLXclientState * a_cl, GLbyte * a_pc)
+{
+    xGLXMakeCurrentReadSGIReq *req = (xGLXMakeCurrentReadSGIReq *) a_pc;
+    __GLX_DECLARE_SWAP_VARIABLES;
+
+    __GLX_SWAP_INT(&req->drawable);
+    __GLX_SWAP_INT(&req->readable);
+    __GLX_SWAP_INT(&req->context);
+    __GLX_SWAP_INT(&req->oldContextTag);
+
+    return ephyrGLXMakeCurrentReal(a_cl, req->drawable, req->readable,
+                                   req->context, req->oldContextTag, TRUE);
+}
+
+int
+ephyrGLXMakeContextCurrent(__GLXclientState * a_cl, GLbyte * a_pc)
+{
+    xGLXMakeContextCurrentReq *req = (xGLXMakeContextCurrentReq *) a_pc;
+
+    return ephyrGLXMakeCurrentReal(a_cl, req->drawable, req->readdrawable,
+                                   req->context, req->oldContextTag, FALSE);
+}
+
+int
+ephyrGLXMakeContextCurrentSwap(__GLXclientState * a_cl, GLbyte * a_pc)
+{
+    xGLXMakeContextCurrentReq *req = (xGLXMakeContextCurrentReq *) a_pc;
+    __GLX_DECLARE_SWAP_VARIABLES;
+
+    __GLX_SWAP_INT(&req->drawable);
+    __GLX_SWAP_INT(&req->readdrawable);
+    __GLX_SWAP_INT(&req->context);
+    __GLX_SWAP_INT(&req->oldContextTag);
+
+    return ephyrGLXMakeCurrentReal(a_cl, req->drawable, req->readdrawable,
+                                   req->context, req->oldContextTag, TRUE);
 }
 
 static int
diff --git a/xorg-server/hw/kdrive/ephyr/ephyrhostglx.c b/xorg-server/hw/kdrive/ephyr/ephyrhostglx.c
index 6b9da6fb1..392489a00 100644
--- a/xorg-server/hw/kdrive/ephyr/ephyrhostglx.c
+++ b/xorg-server/hw/kdrive/ephyr/ephyrhostglx.c
@@ -52,6 +52,8 @@
 #include "ephyrlog.h"
 #include "hostx.h"
 
+static int glx_major, glx_minor;
+
 enum VisualConfRequestType {
     EPHYR_GET_FB_CONFIG,
     EPHYR_VENDOR_PRIV_GET_FB_CONFIG_SGIX,
@@ -99,6 +101,12 @@ ephyrHostGLXQueryVersion(int *a_major, int *a_minor)
     EPHYR_RETURN_VAL_IF_FAIL(a_major && a_minor, FALSE);
     EPHYR_LOG("enter\n");
 
+    if (glx_major) {
+        *a_major = glx_major;
+        *a_minor = glx_minor;
+        return TRUE;
+    }
+
     if (!ephyrHostGLXGetMajorOpcode(&major_opcode)) {
         EPHYR_LOG_ERROR("failed to get major opcode\n");
         goto out;
@@ -117,8 +125,8 @@ ephyrHostGLXQueryVersion(int *a_major, int *a_minor)
     UnlockDisplay(dpy);
     SyncHandle();
 
-    *a_major = reply.majorVersion;
-    *a_minor = reply.minorVersion;
+    *a_major = glx_major = reply.majorVersion;
+    *a_minor = glx_minor = reply.minorVersion;
 
     EPHYR_LOG("major:%d, minor:%d\n", *a_major, *a_minor);
 
@@ -431,17 +439,20 @@ ephyrHostGLXSendClientInfo(int32_t a_major, int32_t a_minor,
 
 Bool
 ephyrHostGLXCreateContext(int a_screen,
-                          int a_visual_id,
+                          int a_generic_id,
                           int a_context_id,
-                          int a_share_list_ctxt_id, Bool a_direct)
+                          int a_share_list_ctxt_id,
+                          int a_render_type,
+                          Bool a_direct,
+                          int code)
 {
     Bool is_ok = FALSE;
     Display *dpy = hostx_get_display();
     int major_opcode = 0, remote_context_id = 0;
-    xGLXCreateContextReq *req;
 
-    EPHYR_LOG("enter. screen:%d, visual:%d, contextid:%d, direct:%d\n",
-              a_screen, a_visual_id, a_context_id, a_direct);
+    EPHYR_LOG("enter. screen:%d, generic_id:%d, contextid:%d, rendertype:%d, "
+                 "direct:%d\n", a_screen, a_generic_id, a_context_id,
+                 a_render_type, a_direct);
 
     if (!hostx_allocate_resource_id_peer(a_context_id, &remote_context_id)) {
         EPHYR_LOG_ERROR("failed to peer the context id %d host X",
@@ -456,15 +467,38 @@ ephyrHostGLXCreateContext(int a_screen,
 
     LockDisplay(dpy);
 
-    /* Send the glXCreateContext request */
-    GetReq(GLXCreateContext, req);
-    req->reqType = major_opcode;
-    req->glxCode = X_GLXCreateContext;
-    req->context = remote_context_id;
-    req->visual = a_visual_id;
-    req->screen = DefaultScreen(dpy);
-    req->shareList = a_share_list_ctxt_id;
-    req->isDirect = a_direct;
+    switch (code) {
+    case X_GLXCreateContext: {
+        /* Send the glXCreateContext request */
+        xGLXCreateContextReq *req;
+        GetReq(GLXCreateContext, req);
+        req->reqType = major_opcode;
+        req->glxCode = X_GLXCreateContext;
+        req->context = remote_context_id;
+        req->visual = a_generic_id;
+        req->screen = DefaultScreen(dpy);
+        req->shareList = a_share_list_ctxt_id;
+        req->isDirect = a_direct;
+    }
+
+    case X_GLXCreateNewContext: {
+        /* Send the glXCreateNewContext request */
+        xGLXCreateNewContextReq *req;
+        GetReq(GLXCreateNewContext, req);
+        req->reqType = major_opcode;
+        req->glxCode = X_GLXCreateNewContext;
+        req->context = remote_context_id;
+        req->fbconfig = a_generic_id;
+        req->screen = DefaultScreen(dpy);
+        req->renderType = a_render_type;
+        req->shareList = a_share_list_ctxt_id;
+        req->isDirect = a_direct;
+    }
+
+    default:
+        /* This should never be reached !*/
+        EPHYR_LOG("Internal error! Invalid CreateContext code!\n");
+    }
 
     UnlockDisplay(dpy);
     SyncHandle();
@@ -512,20 +546,19 @@ ephyrHostDestroyContext(int a_ctxt_id)
 }
 
 Bool
-ephyrHostGLXMakeCurrent(int a_drawable,
+ephyrHostGLXMakeCurrent(int a_drawable, int a_readable,
                         int a_glx_ctxt_id, int a_old_ctxt_tag, int *a_ctxt_tag)
 {
     Bool is_ok = FALSE;
     Display *dpy = hostx_get_display();
     int32_t major_opcode = 0;
     int remote_glx_ctxt_id = 0;
-    xGLXMakeCurrentReq *req;
     xGLXMakeCurrentReply reply;
 
     EPHYR_RETURN_VAL_IF_FAIL(a_ctxt_tag, FALSE);
 
-    EPHYR_LOG("enter. drawable:%d, context:%d, oldtag:%d\n",
-              a_drawable, a_glx_ctxt_id, a_old_ctxt_tag);
+    EPHYR_LOG("enter. drawable:%d, read:%d, context:%d, oldtag:%d\n",
+              a_drawable, a_readable, a_glx_ctxt_id, a_old_ctxt_tag);
 
     if (!ephyrHostGLXGetMajorOpcode(&major_opcode)) {
         EPHYR_LOG_ERROR("failed to get major opcode\n");
@@ -538,12 +571,48 @@ ephyrHostGLXMakeCurrent(int a_drawable,
 
     LockDisplay(dpy);
 
-    GetReq(GLXMakeCurrent, req);
-    req->reqType = major_opcode;
-    req->glxCode = X_GLXMakeCurrent;
-    req->drawable = a_drawable;
-    req->context = remote_glx_ctxt_id;
-    req->oldContextTag = a_old_ctxt_tag;
+    /* If both drawables are the same, use the old MakeCurrent request.
+     * Otherwise, if we have GLX 1.3 or higher, use the MakeContextCurrent
+     * request which supports separate read and draw targets.  Failing that,
+     * try the SGI MakeCurrentRead extension.  Logic cribbed from Mesa. */
+    if (a_drawable == a_readable) {
+        xGLXMakeCurrentReq *req;
+
+        GetReq(GLXMakeCurrent, req);
+        req->reqType = major_opcode;
+        req->glxCode = X_GLXMakeCurrent;
+        req->drawable = a_drawable;
+        req->context = remote_glx_ctxt_id;
+        req->oldContextTag = a_old_ctxt_tag;
+    }
+    else if (glx_major > 1 || glx_minor >= 3) {
+        xGLXMakeContextCurrentReq *req;
+
+        GetReq(GLXMakeContextCurrent, req);
+        req->reqType = major_opcode;
+        req->glxCode = X_GLXMakeContextCurrent;
+        req->drawable = a_drawable;
+        req->readdrawable = a_readable;
+        req->context = remote_glx_ctxt_id;
+        req->oldContextTag = a_old_ctxt_tag;
+    }
+    else {
+        xGLXVendorPrivateWithReplyReq *vpreq;
+        xGLXMakeCurrentReadSGIReq *req;
+
+        GetReqExtra(GLXVendorPrivateWithReply,
+                    (sz_xGLXMakeCurrentReadSGIReq -
+                     sz_xGLXVendorPrivateWithReplyReq),
+                    vpreq);
+        req = (xGLXMakeCurrentReadSGIReq *) vpreq;
+        req->reqType = major_opcode;
+        req->glxCode = X_GLXVendorPrivateWithReply;
+        req->vendorCode = X_GLXvop_MakeCurrentReadSGI;
+        req->drawable = a_drawable;
+        req->readable = a_readable;
+        req->context = remote_glx_ctxt_id;
+        req->oldContextTag = a_old_ctxt_tag;
+    }
 
     memset(&reply, 0, sizeof(reply));
     if (!_XReply(dpy, (xReply *) & reply, 0, False)) {
diff --git a/xorg-server/hw/kdrive/ephyr/ephyrhostglx.h b/xorg-server/hw/kdrive/ephyr/ephyrhostglx.h
index 9c6012070..7ff515dc2 100644
--- a/xorg-server/hw/kdrive/ephyr/ephyrhostglx.h
+++ b/xorg-server/hw/kdrive/ephyr/ephyrhostglx.h
@@ -55,13 +55,16 @@ Bool ephyrHostGLXGetMajorOpcode(int32_t * a_opcode);
 Bool ephyrHostGLXSendClientInfo(int32_t a_major, int32_t a_minor,
                                 const char *a_extension_list);
 Bool ephyrHostGLXCreateContext(int a_screen,
-                               int a_visual_id,
+                               int a_generic_id,
                                int a_context_id,
-                               int a_shared_list_ctx_id, Bool a_direct);
+                               int a_share_list_ctxt_id,
+                               int a_render_type,
+                               Bool a_direct,
+                               int code);
 
 Bool ephyrHostDestroyContext(int a_ctxt_id);
 
-Bool ephyrHostGLXMakeCurrent(int a_drawable, int a_glx_ctxt_id,
+Bool ephyrHostGLXMakeCurrent(int a_drawable, int a_readable, int a_glx_ctxt_id,
                              int a_olg_ctxt_tag, int *a_ctxt_tag);
 
 Bool ephyrHostGetIntegerValue(int a_current_context_tag, int a_int, int *a_val);
diff --git a/xorg-server/hw/kdrive/src/kinput.c b/xorg-server/hw/kdrive/src/kinput.c
index 2a6fec9a7..76424c1c5 100644
--- a/xorg-server/hw/kdrive/src/kinput.c
+++ b/xorg-server/hw/kdrive/src/kinput.c
@@ -1057,7 +1057,7 @@ KdGetOptions(InputOption **options, char *string)
 
     if (strchr(string, '=')) {
         tam_key = (strchr(string, '=') - string);
-        key = strndup(string, tam_key + 1);
+        key = strndup(string, tam_key);
         if (!key)
             goto out;
 
diff --git a/xorg-server/hw/xfree86/common/xf86DGA.c b/xorg-server/hw/xfree86/common/xf86DGA.c
index a441dee99..c25a2747b 100644
--- a/xorg-server/hw/xfree86/common/xf86DGA.c
+++ b/xorg-server/hw/xfree86/common/xf86DGA.c
@@ -2095,7 +2095,7 @@ ProcXDGADispatch(ClientPtr client)
 {
     REQUEST(xReq);
 
-    if (!LocalClient(client))
+    if (!client->local)
         return DGAErrorBase + XF86DGAClientNotLocal;
 
 #ifdef DGA_REQ_DEBUG
diff --git a/xorg-server/hw/xfree86/common/xf86Module.h b/xorg-server/hw/xfree86/common/xf86Module.h
index 83f9790d2..fd90aac54 100644
--- a/xorg-server/hw/xfree86/common/xf86Module.h
+++ b/xorg-server/hw/xfree86/common/xf86Module.h
@@ -80,7 +80,7 @@ typedef enum {
  * mask is 0xFFFF0000.
  */
 #define ABI_ANSIC_VERSION	SET_ABI_VERSION(0, 4)
-#define ABI_VIDEODRV_VERSION	SET_ABI_VERSION(13, 0)
+#define ABI_VIDEODRV_VERSION	SET_ABI_VERSION(14, 0)
 #define ABI_XINPUT_VERSION	SET_ABI_VERSION(18, 0)
 #define ABI_EXTENSION_VERSION	SET_ABI_VERSION(7, 0)
 #define ABI_FONT_VERSION	SET_ABI_VERSION(0, 6)
diff --git a/xorg-server/hw/xfree86/common/xf86Opt.h b/xorg-server/hw/xfree86/common/xf86Opt.h
index 0d808de14..c3ba2d75c 100644
--- a/xorg-server/hw/xfree86/common/xf86Opt.h
+++ b/xorg-server/hw/xfree86/common/xf86Opt.h
@@ -85,7 +85,7 @@ extern _X_EXPORT int xf86CheckIntOption(XF86OptionPtr optlist, const char *name,
 extern _X_EXPORT double xf86CheckRealOption(XF86OptionPtr optlist,
                                             const char *name, double deflt);
 extern _X_EXPORT char *xf86CheckStrOption(XF86OptionPtr optlist,
-                                          const char *name, char *deflt);
+                                          const char *name, const char *deflt);
 extern _X_EXPORT int xf86CheckBoolOption(XF86OptionPtr list, const char *name,
                                          int deflt);
 extern _X_EXPORT double xf86CheckPercentOption(XF86OptionPtr list,
diff --git a/xorg-server/hw/xfree86/common/xf86Option.c b/xorg-server/hw/xfree86/common/xf86Option.c
index 7cc744029..c2ec79a53 100644
--- a/xorg-server/hw/xfree86/common/xf86Option.c
+++ b/xorg-server/hw/xfree86/common/xf86Option.c
@@ -276,7 +276,7 @@ xf86CheckRealOption(XF86OptionPtr optlist, const char *name, double deflt)
 }
 
 char *
-xf86CheckStrOption(XF86OptionPtr optlist, const char *name, char *deflt)
+xf86CheckStrOption(XF86OptionPtr optlist, const char *name, const char *deflt)
 {
     return LookupStrOption(optlist, name, deflt, FALSE);
 }
diff --git a/xorg-server/hw/xfree86/common/xf86vmode.c b/xorg-server/hw/xfree86/common/xf86vmode.c
index 7cd2eedb3..cad0e6a21 100644
--- a/xorg-server/hw/xfree86/common/xf86vmode.c
+++ b/xorg-server/hw/xfree86/common/xf86vmode.c
@@ -1590,7 +1590,7 @@ ProcXF86VidModeGetPermissions(ClientPtr client)
         return BadValue;
 
     if (xf86GetVidModeEnabled() &&
-        (xf86GetVidModeAllowNonLocal() || LocalClient(client))) {
+        (xf86GetVidModeAllowNonLocal() || client->local)) {
         rep.permissions |= XF86VM_WRITE_PERMISSION;
     }
     if (client->swapped) {
@@ -1659,7 +1659,7 @@ ProcXF86VidModeDispatch(ClientPtr client)
     default:
         if (!xf86GetVidModeEnabled())
             return VidModeErrorBase + XF86VidModeExtensionDisabled;
-        if (xf86GetVidModeAllowNonLocal() || LocalClient(client)) {
+        if (xf86GetVidModeAllowNonLocal() || client->local) {
             switch (stuff->data) {
             case X_XF86VidModeAddModeLine:
                 return ProcXF86VidModeAddModeLine(client);
@@ -2083,7 +2083,7 @@ SProcXF86VidModeDispatch(ClientPtr client)
     default:
         if (!xf86GetVidModeEnabled())
             return VidModeErrorBase + XF86VidModeExtensionDisabled;
-        if (xf86GetVidModeAllowNonLocal() || LocalClient(client)) {
+        if (xf86GetVidModeAllowNonLocal() || client->local) {
             switch (stuff->data) {
             case X_XF86VidModeAddModeLine:
                 return SProcXF86VidModeAddModeLine(client);
diff --git a/xorg-server/hw/xfree86/dri/xf86dri.c b/xorg-server/hw/xfree86/dri/xf86dri.c
index ba74bb04f..0b2e8fa1b 100644
--- a/xorg-server/hw/xfree86/dri/xf86dri.c
+++ b/xorg-server/hw/xfree86/dri/xf86dri.c
@@ -118,7 +118,7 @@ ProcXF86DRIQueryDirectRenderingCapable(register ClientPtr client)
         return BadValue;
     }
 
-    if (!LocalClient(client) || client->swapped)
+    if (!client->local || client->swapped)
         isCapable = 0;
 
     rep = (xXF86DRIQueryDirectRenderingCapableReply) {
@@ -528,7 +528,7 @@ ProcXF86DRIDispatch(register ClientPtr client)
         return ProcXF86DRIQueryDirectRenderingCapable(client);
     }
 
-    if (!LocalClient(client))
+    if (!client->local)
         return DRIErrorBase + XF86DRIClientNotLocal;
 
     switch (stuff->data) {
diff --git a/xorg-server/hw/xfree86/dri2/dri2ext.c b/xorg-server/hw/xfree86/dri2/dri2ext.c
index ee610c0ec..e1decec9e 100644
--- a/xorg-server/hw/xfree86/dri2/dri2ext.c
+++ b/xorg-server/hw/xfree86/dri2/dri2ext.c
@@ -588,7 +588,7 @@ ProcDRI2Dispatch(ClientPtr client)
         return ProcDRI2QueryVersion(client);
     }
 
-    if (!LocalClient(client))
+    if (!client->local)
         return BadRequest;
 
     switch (stuff->data) {
diff --git a/xorg-server/hw/xfree86/int10/generic.c b/xorg-server/hw/xfree86/int10/generic.c
index 4633120fd..d7594def2 100644
--- a/xorg-server/hw/xfree86/int10/generic.c
+++ b/xorg-server/hw/xfree86/int10/generic.c
@@ -178,7 +178,8 @@ xf86ExtendedInitInt10(int entityIndex, int Flags)
      */
     vbiosMem = (char *) base + V_BIOS;
     memset(vbiosMem, 0, 2 * V_BIOS_SIZE);
-    if (pci_device_read_rom(pInt->dev, vbiosMem) < V_BIOS_SIZE) {
+    if (pci_device_read_rom(pInt->dev, vbiosMem) != 0
+        || pInt->dev->rom_size < V_BIOS_SIZE) {
         xf86DrvMsg(pScrn->scrnIndex, X_WARNING,
                    "Unable to retrieve all of segment 0x0C0000.\n");
     }
diff --git a/xorg-server/hw/xquartz/applewm.c b/xorg-server/hw/xquartz/applewm.c
index d41a81db3..aea0a45f2 100644
--- a/xorg-server/hw/xquartz/applewm.c
+++ b/xorg-server/hw/xquartz/applewm.c
@@ -612,7 +612,7 @@ ProcAppleWMDispatch(register ClientPtr client)
         return ProcAppleWMQueryVersion(client);
     }
 
-    if (!LocalClient(client))
+    if (!client->local)
         return WMErrorBase + AppleWMClientNotLocal;
 
     switch (stuff->data) {
@@ -684,7 +684,7 @@ SProcAppleWMDispatch(register ClientPtr client)
     REQUEST(xReq);
 
     /* It is bound to be non-local when there is byte swapping */
-    if (!LocalClient(client))
+    if (!client->local)
         return WMErrorBase + AppleWMClientNotLocal;
 
     /* only local clients are allowed WM access */
diff --git a/xorg-server/hw/xquartz/xpr/appledri.c b/xorg-server/hw/xquartz/xpr/appledri.c
index f77848f13..9aac07240 100644
--- a/xorg-server/hw/xquartz/xpr/appledri.c
+++ b/xorg-server/hw/xquartz/xpr/appledri.c
@@ -129,7 +129,7 @@ ProcAppleDRIQueryDirectRenderingCapable(register ClientPtr client)
     }
     rep.isCapable = isCapable;
 
-    if (!LocalClient(client))
+    if (!client->local)
         rep.isCapable = 0;
 
     if (client->swapped) {
@@ -354,7 +354,7 @@ ProcAppleDRIDispatch(register ClientPtr client)
         return ProcAppleDRIQueryDirectRenderingCapable(client);
     }
 
-    if (!LocalClient(client))
+    if (!client->local)
         return DRIErrorBase + AppleDRIClientNotLocal;
 
     switch (stuff->data) {
@@ -469,7 +469,7 @@ SProcAppleDRIDispatch(register ClientPtr client)
         return SProcAppleDRIQueryDirectRenderingCapable(client);
     }
 
-    if (!LocalClient(client))
+    if (!client->local)
         return DRIErrorBase + AppleDRIClientNotLocal;
 
     switch (stuff->data) {
diff --git a/xorg-server/hw/xwin/winwindowswm.c b/xorg-server/hw/xwin/winwindowswm.c
index 0389c7e22..19d5e95a9 100644
--- a/xorg-server/hw/xwin/winwindowswm.c
+++ b/xorg-server/hw/xwin/winwindowswm.c
@@ -499,7 +499,7 @@ ProcWindowsWMDispatch(ClientPtr client)
         return ProcWindowsWMQueryVersion(client);
     }
 
-    if (!LocalClient(client))
+    if (!client->local)
         return WMErrorBase + WindowsWMClientNotLocal;
 
     switch (stuff->data) {
@@ -549,7 +549,7 @@ SProcWindowsWMDispatch(ClientPtr client)
     REQUEST(xReq);
 
     /* It is bound to be non-local when there is byte swapping */
-    if (!LocalClient(client))
+    if (!client->local)
         return WMErrorBase + WindowsWMClientNotLocal;
 
     /* only local clients are allowed WM access */
diff --git a/xorg-server/include/dixstruct.h b/xorg-server/include/dixstruct.h
index 5286f1691..e351fcf92 100644
--- a/xorg-server/include/dixstruct.h
+++ b/xorg-server/include/dixstruct.h
@@ -56,8 +56,7 @@ ReplyNotSwappd(ClientPtr /* pClient */ ,
                void * /* pbuf */ ) _X_NORETURN;
 
 typedef enum { ClientStateInitial,
-    /* 1 is unused now, was ClientStateAuthenticating */
-    ClientStateRunning = 2,
+    ClientStateRunning,
     ClientStateRetained,
     ClientStateGone
 } ClientState;
@@ -86,26 +85,29 @@ typedef struct _Window *SaveSetElt;
 #endif
 
 typedef struct _Client {
-    int index;
-    Mask clientAsMask;
     pointer requestBuffer;
     pointer osPrivate;          /* for OS layer, including scheduler */
-    Bool swapped;
+    Mask clientAsMask;
+    short index;
+    unsigned char majorOp, minorOp;
+    int swapped:1;
+    int local:1;
+    int big_requests:1;          /* supports large requests */
+    int clientGone:1;
+    int closeDownMode:2;
+    int clientState:2;
+    char smart_priority;
+    short noClientException;      /* this client died or needs to be killed */
+    int priority;
     ReplySwapPtr pSwapReplyFunc;
     XID errorValue;
     unsigned short sequence;    //MH
-    int closeDownMode;
-    int clientGone;
-    int noClientException;      /* this client died or needs to be
-                                 * killed */
     int ignoreCount;            /* count for Attend/IgnoreClient */
-    SaveSetElt *saveSet;
     int numSaved;
+    SaveSetElt *saveSet;
     int (**requestVector) (ClientPtr /* pClient */ );
     CARD32 req_len;             /* length of current request */
-    Bool big_requests;          /* supports large requests */
-    int priority;
-    ClientState clientState;
+    unsigned int replyBytesRemaining;
     PrivateRec *devPrivates;
     unsigned short xkbClientFlags;
     unsigned short mapNotifyMask;
@@ -113,15 +115,12 @@ typedef struct _Client {
     unsigned short vMajor, vMinor;
     KeyCode minKC, maxKC;
 
-    unsigned long replyBytesRemaining;
-    int smart_priority;
-    long smart_start_tick;
-    long smart_stop_tick;
-    long smart_check_tick;
+    int smart_start_tick;
+    int smart_stop_tick;
+    int smart_check_tick;
 
     DeviceIntPtr clientPtr;
     ClientIdPtr clientIds;
-    unsigned short majorOp, minorOp;
 } ClientRec;
 
 /*
diff --git a/xorg-server/include/os.h b/xorg-server/include/os.h
index 015800089..d906834f1 100644
--- a/xorg-server/include/os.h
+++ b/xorg-server/include/os.h
@@ -409,9 +409,6 @@ extern _X_EXPORT int
 InvalidHost(sockaddrPtr /*saddr */ , int /*len */ , ClientPtr client);
 
 extern _X_EXPORT int
-LocalClient(ClientPtr /* client */ );
-
-extern _X_EXPORT int
 LocalClientCred(ClientPtr, int *, int *);
 
 #define LCC_UID_SET	(1 << 0)
diff --git a/xorg-server/include/windowstr.h b/xorg-server/include/windowstr.h
index 79a957e58..15a83f056 100644
--- a/xorg-server/include/windowstr.h
+++ b/xorg-server/include/windowstr.h
@@ -86,8 +86,8 @@ typedef struct _WindowOpt {
     struct _OtherClients *otherClients; /* default: NULL */
     struct _GrabRec *passiveGrabs;      /* default: NULL */
     PropertyPtr userProps;      /* default: NULL */
-    unsigned long backingBitPlanes;     /* default: ~0L */
-    unsigned long backingPixel; /* default: 0 */
+    CARD32 backingBitPlanes;    /* default: ~0L */
+    CARD32 backingPixel;        /* default: 0 */
     RegionPtr boundingShape;    /* default: NULL */
     RegionPtr clipShape;        /* default: NULL */
     RegionPtr inputShape;       /* default: NULL */
diff --git a/xorg-server/mi/mipointer.c b/xorg-server/mi/mipointer.c
index 4defaf5ec..f34506326 100644
--- a/xorg-server/mi/mipointer.c
+++ b/xorg-server/mi/mipointer.c
@@ -575,13 +575,8 @@ miPointerSetPosition(DeviceIntPtr pDev, int mode, double *screenx,
 
     miPointerPtr pPointer;
 
-    if (!pDev)
-        return NULL;
-
     pPointer = MIPOINTER(pDev);
     pScreen = pPointer->pScreen;
-    if (!pScreen)
-        return NULL;            /* called before ready */
 
     x = trunc(*screenx);
     y = trunc(*screeny);
diff --git a/xorg-server/os/access.c b/xorg-server/os/access.c
index 71af6e7c9..97845994f 100644
--- a/xorg-server/os/access.c
+++ b/xorg-server/os/access.c
@@ -1182,14 +1182,6 @@ ComputeLocalClient(ClientPtr client)
     return FALSE;
 }
 
-Bool
-LocalClient(ClientPtr client)
-{
-    if (!client->osPrivate)
-        return FALSE;
-    return ((OsCommPtr) client->osPrivate)->local_client;
-}
-
 /*
  * Return the uid and gid of a connected local client
  * 
@@ -1351,7 +1343,7 @@ AuthorizedClient(ClientPtr client)
     if (rc != Success)
         return rc;
 
-    return LocalClient(client) ? Success : BadAccess;
+    return client->local ? Success : BadAccess;
 }
 
 /* Add a host to the access control list.  This is the external interface
diff --git a/xorg-server/os/connection.c b/xorg-server/os/connection.c
index fa471dd6c..d073c2f91 100644
--- a/xorg-server/os/connection.c
+++ b/xorg-server/os/connection.c
@@ -801,7 +801,7 @@ AllocNewConnection(XtransConnInfo trans_conn, int fd, CARD32 conn_time)
         free(oc);
         return NullClient;
     }
-    oc->local_client = ComputeLocalClient(client);
+    client->local = ComputeLocalClient(client);
 #if !defined(WIN32)
     ConnectionTranslation[fd] = client->index;
 #else
@@ -931,7 +931,7 @@ ErrorConnMax(XtransConnInfo trans_conn)
     xConnSetupPrefix csp;
     char pad[3] = { 0, 0, 0 };
     struct iovec iov[3];
-    char byteOrder = 0;
+    char order = 0;
     int whichbyte = 1;
     struct timeval waittime;
     fd_set mask;
@@ -944,15 +944,15 @@ ErrorConnMax(XtransConnInfo trans_conn)
     FD_SET(fd, &mask);
     (void) Select(fd + 1, &mask, NULL, NULL, &waittime);
     /* try to read the byte-order of the connection */
-    (void) _XSERVTransRead(trans_conn, &byteOrder, 1);
-    if ((byteOrder == 'l') || (byteOrder == 'B')) {
+    (void) _XSERVTransRead(trans_conn, &order, 1);
+    if (order == 'l' || order == 'B' || order == 'r' || order == 'R') {
         csp.success = xFalse;
         csp.lengthReason = sizeof(NOROOM) - 1;
         csp.length = (sizeof(NOROOM) + 2) >> 2;
         csp.majorVersion = X_PROTOCOL;
         csp.minorVersion = X_PROTOCOL_REVISION;
-        if (((*(char *) &whichbyte) && (byteOrder == 'B')) ||
-            (!(*(char *) &whichbyte) && (byteOrder == 'l'))) {
+	if (((*(char *) &whichbyte) && (order == 'B' || order == 'R')) ||
+	    (!(*(char *) &whichbyte) && (order == 'l' || order == 'r'))) {
             swaps(&csp.majorVersion);
             swaps(&csp.minorVersion);
             swaps(&csp.length);
@@ -1085,7 +1085,7 @@ CloseDownConnection(ClientPtr client)
     ErrorF("CloseDownConnection: client index = %d, socket fd = %d\n",
            client->index, oc->fd);
 #endif
-    if (oc->output && oc->output->count)
+    if (oc->output)
         FlushClient(client, oc, (char *) NULL, 0);
 #ifdef XDMCP
     XdmcpCloseDisplay(oc->fd);
diff --git a/xorg-server/os/io.c b/xorg-server/os/io.c
index 9ccfdba62..8bb54ff74 100644
--- a/xorg-server/os/io.c
+++ b/xorg-server/os/io.c
@@ -82,6 +82,23 @@ SOFTWARE.
 CallbackListPtr ReplyCallback;
 CallbackListPtr FlushCallback;
 
+typedef struct _connectionInput {
+    struct _connectionInput *next;
+    char *buffer;               /* contains current client input */
+    char *bufptr;               /* pointer to current start of data */
+    int bufcnt;                 /* count of bytes in buffer */
+    int lenLastReq;
+    int size;
+    unsigned int ignoreBytes;   /* bytes to ignore before the next request */
+} ConnectionInput;
+
+typedef struct _connectionOutput {
+    struct _connectionOutput *next;
+    unsigned char *buf;
+    int size;
+    int count;
+} ConnectionOutput;
+
 static ConnectionInputPtr AllocateInputBuffer(void);
 static ConnectionOutputPtr AllocateOutputBuffer(void);
 
@@ -846,10 +863,13 @@ FlushClient(ClientPtr who, OsCommPtr oc, const void *__extraBuf, int extraCount)
     long todo;
 
     if (!oco)
-        return 0;
+	return 0;
     written = 0;
     padsize = padding_for_int32(extraCount);
     notWritten = oco->count + extraCount + padsize;
+    if (!notWritten)
+        return 0;
+
     todo = notWritten;
     while (notWritten) {
         long before = written;  /* amount of whole thing written */
diff --git a/xorg-server/os/osdep.h b/xorg-server/os/osdep.h
index fff088701..0ccdb62ef 100644
--- a/xorg-server/os/osdep.h
+++ b/xorg-server/os/osdep.h
@@ -107,22 +107,8 @@ typedef Bool (*AddAuthorFunc) (unsigned name_length, const char *name,
                                unsigned data_length, char *data);
 #endif
 
-typedef struct _connectionInput {
-    struct _connectionInput *next;
-    char *buffer;               /* contains current client input */
-    char *bufptr;               /* pointer to current start of data */
-    int bufcnt;                 /* count of bytes in buffer */
-    int lenLastReq;
-    int size;
-    unsigned int ignoreBytes;   /* bytes to ignore before the next request */
-} ConnectionInput, *ConnectionInputPtr;
-
-typedef struct _connectionOutput {
-    struct _connectionOutput *next;
-    int size;
-    unsigned char *buf;
-    int count;
-} ConnectionOutput, *ConnectionOutputPtr;
+typedef struct _connectionInput *ConnectionInputPtr;
+typedef struct _connectionOutput *ConnectionOutputPtr;
 
 struct _osComm;
 
@@ -162,7 +148,6 @@ typedef struct _osComm {
     XID auth_id;                /* authorization id */
     CARD32 conn_time;           /* timestamp if not established, else 0  */
     struct _XtransConnInfo *trans_conn; /* transport connection object */
-    Bool local_client;
 } OsCommRec, *OsCommPtr;
 
 extern int FlushClient(ClientPtr /*who */ ,
diff --git a/xorg-server/xkeyboard-config/rules/base.xml.in b/xorg-server/xkeyboard-config/rules/base.xml.in
index 87943048e..712ddabf6 100644
--- a/xorg-server/xkeyboard-config/rules/base.xml.in
+++ b/xorg-server/xkeyboard-config/rules/base.xml.in
@@ -3433,6 +3433,12 @@
             </languageList>
           </configItem>
         </variant>
+        <variant>
+          <configItem>
+            <name>ibm</name>
+            <_description>Italian (IBM 142)</_description>
+          </configItem>
+        </variant>
       </variantList>
     </layout>
     <layout>
diff --git a/xorg-server/xkeyboard-config/symbols/it b/xorg-server/xkeyboard-config/symbols/it
index 1d26a044b..d70e49755 100644
--- a/xorg-server/xkeyboard-config/symbols/it
+++ b/xorg-server/xkeyboard-config/symbols/it
@@ -255,3 +255,46 @@ xkb_symbols "htcdream" {
 
         include "level3(alt_switch)"
 };
+
+partial alphanumeric_keys
+xkb_symbols "ibm" {
+
+    // Modifies the basic italian layout to match IBM 142 variant
+
+    include "us"
+
+    name[Group1]="Italian (IBM 142)";
+
+    key <TLDE>	{ [ backslash,        bar                             ] };
+
+    key <AE02>	{ [         2,   quotedbl                             ] };
+    key <AE03>	{ [         3,   sterling,   numbersign               ] };
+    key <AE06>	{ [         6,  ampersand                             ] };
+    key <AE07>	{ [         7,      slash,    braceleft               ] };
+    key <AE08>	{ [         8,  parenleft,  bracketleft               ] };
+    key <AE09>	{ [         9, parenright, bracketright               ] };
+    key <AE10>	{ [         0,      equal,   braceright               ] };
+    key <AE11>	{ [apostrophe,   question                             ] };
+    key <AE12>	{ [    igrave, asciicircum                            ] };
+
+    key <AD01>	{ [         q,           Q,          at               ] };
+    key <AD03>	{ [         e,           E,    EuroSign               ] };
+    key <AD11>	{ [    egrave,     eacute                             ] };
+    key <AD12>	{ [      plus,   asterisk,   asciitilde               ] };
+
+    key <AC10>	{ [    ograve,   ccedilla                             ] };
+    key <AC11>	{ [    agrave,     degree                             ] };
+
+    key <AB08>	{ [     comma,  semicolon                             ] };
+    key <AB09>	{ [    period,      colon                             ] };
+    key <AB10>	{ [     minus, underscore                             ] };
+
+    key <BKSL>	{ [    ugrave,    section,       grave                ] };
+
+    key <LSGT>	{ [      less,    greater                             ] };
+
+    include "level3(ralt_switch)"
+    include "kpdl(comma)"
+
+};
+