fontconfig mesa pixman xkeyboard-config git update 2 Okt 2013

xkeyboard-config commit e5a53229a9914235921911f05b31d6092e844ea1
pixman           commit 7d05a7f4dc825f9c778e534fdabb749199c2e439
fontconfig       commit 0203055520206028eecee5d261887cdc91500e15
mesa             commit 848c0e72f36d0e1e460193a2d30b2f631529156f

6 files changed, 675 insertions, 99 deletions

diff --git a/pixman/configure.ac b/pixman/configure.ac
index 263c63edf..2dd477663 100644
--- a/pixman/configure.ac
+++ b/pixman/configure.ac
@@ -961,37 +961,38 @@ main ()
 ]]))
 
 AC_DEFUN([PIXMAN_CHECK_PTHREAD],[dnl
-    if test "z$support_for_pthread_setspecific" != "zyes"; then
+    if test "z$support_for_pthreads" != "zyes"; then
 	PIXMAN_LINK_WITH_ENV(
 		[$1], [pthread_test_program],
 		[PTHREAD_CFLAGS="$CFLAGS"
 		 PTHREAD_LIBS="$LIBS"
 		 PTHREAD_LDFLAGS="$LDFLAGS"
-		 support_for_pthread_setspecific=yes])
+		 support_for_pthreads=yes])
     fi
 ])
 
-if test $ac_cv_tls = none ; then
-    support_for_pthread_setspecific=no
+support_for_pthreads=no
 
-    AC_MSG_CHECKING(for pthread_setspecific)
+AC_MSG_CHECKING(for pthreads)
 
-    PIXMAN_CHECK_PTHREAD([CFLAGS="-pthread"; LDFLAGS="-pthread"])
-    PIXMAN_CHECK_PTHREAD([CFLAGS="-D_REENTRANT"; LIBS="-lpthread"])
-    PIXMAN_CHECK_PTHREAD([CFLAGS="-D_REENTRANT"; LDFLAGS="-lroot"])
+PIXMAN_CHECK_PTHREAD([CFLAGS="-pthread"; LDFLAGS="-pthread"])
+PIXMAN_CHECK_PTHREAD([CFLAGS="-D_REENTRANT"; LIBS="-lpthread"])
+PIXMAN_CHECK_PTHREAD([CFLAGS="-D_REENTRANT"; LDFLAGS="-lroot"])
     
-    if test $support_for_pthread_setspecific = yes; then
-	CFLAGS="$CFLAGS $PTHREAD_CFLAGS"
-	AC_DEFINE([HAVE_PTHREAD_SETSPECIFIC], [], [Whether pthread_setspecific() is supported])
+if test $support_for_pthreads = yes; then
+    AC_DEFINE([HAVE_PTHREADS], [], [Whether pthreads is supported])
+    if test $ac_cv_tls = none ; then
+        CFLAGS="$CFLAGS $PTHREAD_CFLAGS"
     fi
-
-    AC_MSG_RESULT($support_for_pthread_setspecific);
 fi
 
+AC_MSG_RESULT($support_for_pthreads)
+
 AC_SUBST(TOOLCHAIN_SUPPORTS__THREAD)
-AC_SUBST(HAVE_PTHREAD_SETSPECIFIC)
+AC_SUBST(HAVE_PTHREADS)
 AC_SUBST(PTHREAD_LDFLAGS)
 AC_SUBST(PTHREAD_LIBS)
+AC_SUBST(PTHREAD_CFLAGS)
 
 dnl =====================================
 dnl __attribute__((constructor))
diff --git a/pixman/pixman/pixman-compiler.h b/pixman/pixman/pixman-compiler.h
index 9b190b422..2489adc38 100644
--- a/pixman/pixman/pixman-compiler.h
+++ b/pixman/pixman/pixman-compiler.h
@@ -178,7 +178,7 @@
 #   define PIXMAN_GET_THREAD_LOCAL(name)				\
     (&name)
 
-#elif defined(HAVE_PTHREAD_SETSPECIFIC)
+#elif defined(HAVE_PTHREADS)
 
 #include <pthread.h>
 
diff --git a/pixman/pixman/pixman-vmx.c b/pixman/pixman/pixman-vmx.c
index f629003ab..c33631c0e 100644
--- a/pixman/pixman/pixman-vmx.c
+++ b/pixman/pixman/pixman-vmx.c
@@ -134,15 +134,11 @@ over (vector unsigned int src,
     source ## _mask = vec_lvsl (0, source);
 
 #define COMPUTE_SHIFT_MASKS(dest, source)				\
-    dest ## _mask = vec_lvsl (0, dest);					\
-    source ## _mask = vec_lvsl (0, source);				\
-    store_mask = vec_lvsr (0, dest);
+    source ## _mask = vec_lvsl (0, source);
 
 #define COMPUTE_SHIFT_MASKC(dest, source, mask)				\
     mask ## _mask = vec_lvsl (0, mask);					\
-    dest ## _mask = vec_lvsl (0, dest);					\
-    source ## _mask = vec_lvsl (0, source);				\
-    store_mask = vec_lvsr (0, dest);
+    source ## _mask = vec_lvsl (0, source);
 
 /* notice you have to declare temp vars...
  * Note: tmp3 and tmp4 must remain untouched!
@@ -151,23 +147,17 @@ over (vector unsigned int src,
 #define LOAD_VECTORS(dest, source)			  \
     tmp1 = (typeof(tmp1))vec_ld (0, source);		  \
     tmp2 = (typeof(tmp2))vec_ld (15, source);		  \
-    tmp3 = (typeof(tmp3))vec_ld (0, dest);		  \
     v ## source = (typeof(v ## source))			  \
 	vec_perm (tmp1, tmp2, source ## _mask);		  \
-    tmp4 = (typeof(tmp4))vec_ld (15, dest);		  \
-    v ## dest = (typeof(v ## dest))			  \
-	vec_perm (tmp3, tmp4, dest ## _mask);
+    v ## dest = (typeof(v ## dest))vec_ld (0, dest);
 
 #define LOAD_VECTORSC(dest, source, mask)		  \
     tmp1 = (typeof(tmp1))vec_ld (0, source);		  \
     tmp2 = (typeof(tmp2))vec_ld (15, source);		  \
-    tmp3 = (typeof(tmp3))vec_ld (0, dest);		  \
     v ## source = (typeof(v ## source))			  \
 	vec_perm (tmp1, tmp2, source ## _mask);		  \
-    tmp4 = (typeof(tmp4))vec_ld (15, dest);		  \
     tmp1 = (typeof(tmp1))vec_ld (0, mask);		  \
-    v ## dest = (typeof(v ## dest))			  \
-	vec_perm (tmp3, tmp4, dest ## _mask);		  \
+    v ## dest = (typeof(v ## dest))vec_ld (0, dest);	  \
     tmp2 = (typeof(tmp2))vec_ld (15, mask);		  \
     v ## mask = (typeof(v ## mask))			  \
 	vec_perm (tmp1, tmp2, mask ## _mask);
@@ -178,11 +168,7 @@ over (vector unsigned int src,
                                 splat_alpha (v ## mask));
 
 #define STORE_VECTOR(dest)						\
-    edges = vec_perm (tmp4, tmp3, dest ## _mask);			\
-    tmp3 = vec_perm ((vector unsigned char)v ## dest, edges, store_mask); \
-    tmp1 = vec_perm (edges, (vector unsigned char)v ## dest, store_mask); \
-    vec_st ((vector unsigned int) tmp3, 15, dest);			\
-    vec_st ((vector unsigned int) tmp1, 0, dest);
+    vec_st ((vector unsigned int) v ## dest, 0, dest);
 
 static void
 vmx_combine_over_u_no_mask (uint32_t *      dest,
@@ -191,8 +177,19 @@ vmx_combine_over_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t ia = ALPHA_8 (~s);
+
+	UN8x4_MUL_UN8_ADD_UN8x4 (d, ia, s);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -230,8 +227,23 @@ vmx_combine_over_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t ia;
+
+	UN8x4_MUL_UN8 (s, m);
+
+	ia = ALPHA_8 (~s);
+
+	UN8x4_MUL_UN8_ADD_UN8x4 (d, ia, s);
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -286,8 +298,18 @@ vmx_combine_over_reverse_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t ia = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8_ADD_UN8x4 (s, ia, d);
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -324,8 +346,21 @@ vmx_combine_over_reverse_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t ia = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8 (s, m);
+
+	UN8x4_MUL_UN8_ADD_UN8x4 (s, ia, d);
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -379,8 +414,17 @@ vmx_combine_in_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t s = *src++;
+	uint32_t a = ALPHA_8 (*dest);
+
+	UN8x4_MUL_UN8 (s, a);
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -415,8 +459,20 @@ vmx_combine_in_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t s = *src++;
+	uint32_t a = ALPHA_8 (*dest);
+
+	UN8x4_MUL_UN8 (s, m);
+	UN8x4_MUL_UN8 (s, a);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -468,8 +524,18 @@ vmx_combine_in_reverse_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t d = *dest;
+	uint32_t a = ALPHA_8 (*src++);
+
+	UN8x4_MUL_UN8 (d, a);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -505,8 +571,21 @@ vmx_combine_in_reverse_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t d = *dest;
+	uint32_t a = *src++;
+
+	UN8x4_MUL_UN8 (a, m);
+	a = ALPHA_8 (a);
+	UN8x4_MUL_UN8 (d, a);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -559,8 +638,18 @@ vmx_combine_out_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t s = *src++;
+	uint32_t a = ALPHA_8 (~(*dest));
+
+	UN8x4_MUL_UN8 (s, a);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -596,8 +685,20 @@ vmx_combine_out_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t s = *src++;
+	uint32_t a = ALPHA_8 (~(*dest));
+
+	UN8x4_MUL_UN8 (s, m);
+	UN8x4_MUL_UN8 (s, a);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -649,8 +750,18 @@ vmx_combine_out_reverse_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t d = *dest;
+	uint32_t a = ALPHA_8 (~(*src++));
+
+	UN8x4_MUL_UN8 (d, a);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -687,8 +798,21 @@ vmx_combine_out_reverse_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t d = *dest;
+	uint32_t a = *src++;
+
+	UN8x4_MUL_UN8 (a, m);
+	a = ALPHA_8 (~a);
+	UN8x4_MUL_UN8 (d, a);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -741,8 +865,20 @@ vmx_combine_atop_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t dest_a = ALPHA_8 (d);
+	uint32_t src_ia = ALPHA_8 (~s);
+
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_a, d, src_ia);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -781,8 +917,25 @@ vmx_combine_atop_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t dest_a = ALPHA_8 (d);
+	uint32_t src_ia;
+
+	UN8x4_MUL_UN8 (s, m);
+
+	src_ia = ALPHA_8 (~s);
+
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_a, d, src_ia);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -840,8 +993,20 @@ vmx_combine_atop_reverse_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t src_a = ALPHA_8 (s);
+	uint32_t dest_ia = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_ia, d, src_a);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -880,8 +1045,25 @@ vmx_combine_atop_reverse_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t src_a;
+	uint32_t dest_ia = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8 (s, m);
+
+	src_a = ALPHA_8 (s);
+
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_ia, d, src_a);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -939,8 +1121,20 @@ vmx_combine_xor_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t src_ia = ALPHA_8 (~s);
+	uint32_t dest_ia = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_ia, d, src_ia);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
 
@@ -979,8 +1173,25 @@ vmx_combine_xor_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t src_ia;
+	uint32_t dest_ia = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8 (s, m);
+
+	src_ia = ALPHA_8 (~s);
+
+	UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (s, dest_ia, d, src_ia);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1038,8 +1249,18 @@ vmx_combine_add_u_no_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+
+	UN8x4_ADD_UN8x4 (d, s);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKS (dest, src);
     /* printf ("%s\n",__PRETTY_FUNCTION__); */
@@ -1074,8 +1295,20 @@ vmx_combine_add_u_mask (uint32_t *      dest,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, src_mask, mask_mask, store_mask;
+    vector unsigned char tmp1, tmp2, src_mask, mask_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t m = ALPHA_8 (*mask++);
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+
+	UN8x4_MUL_UN8 (s, m);
+	UN8x4_ADD_UN8x4 (d, s);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1130,8 +1363,18 @@ vmx_combine_src_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+
+	UN8x4_MUL_UN8x4 (s, a);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1170,8 +1413,22 @@ vmx_combine_over_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t sa = ALPHA_8 (s);
+
+	UN8x4_MUL_UN8x4 (s, a);
+	UN8x4_MUL_UN8 (a, sa);
+	UN8x4_MUL_UN8x4_ADD_UN8x4 (d, ~a, s);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1214,8 +1471,21 @@ vmx_combine_over_reverse_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t ida = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8x4 (s, a);
+	UN8x4_MUL_UN8_ADD_UN8x4 (s, ida, d);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1257,8 +1527,20 @@ vmx_combine_in_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t da = ALPHA_8 (*dest);
+
+	UN8x4_MUL_UN8x4 (s, a);
+	UN8x4_MUL_UN8 (s, da);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1299,8 +1581,20 @@ vmx_combine_in_reverse_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t d = *dest;
+	uint32_t sa = ALPHA_8 (*src++);
+
+	UN8x4_MUL_UN8 (a, sa);
+	UN8x4_MUL_UN8x4 (d, a);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1342,8 +1636,21 @@ vmx_combine_out_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t da = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8x4 (s, a);
+	UN8x4_MUL_UN8 (s, da);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1386,8 +1693,21 @@ vmx_combine_out_reverse_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t sa = ALPHA_8 (s);
+
+	UN8x4_MUL_UN8 (a, sa);
+	UN8x4_MUL_UN8x4 (d, ~a);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1430,8 +1750,23 @@ vmx_combine_atop_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask, vsrca;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t sa = ALPHA_8 (s);
+	uint32_t da = ALPHA_8 (d);
+
+	UN8x4_MUL_UN8x4 (s, a);
+	UN8x4_MUL_UN8 (a, sa);
+	UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (d, ~a, s, da);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1481,8 +1816,23 @@ vmx_combine_atop_reverse_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t sa = ALPHA_8 (s);
+	uint32_t da = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8x4 (s, a);
+	UN8x4_MUL_UN8 (a, sa);
+	UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (d, a, s, da);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1529,8 +1879,23 @@ vmx_combine_xor_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+	uint32_t sa = ALPHA_8 (s);
+	uint32_t da = ALPHA_8 (~d);
+
+	UN8x4_MUL_UN8x4 (s, a);
+	UN8x4_MUL_UN8 (a, sa);
+	UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (d, ~a, s, da);
+
+	*dest++ = d;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
@@ -1577,8 +1942,20 @@ vmx_combine_add_ca (pixman_implementation_t *imp,
 {
     int i;
     vector unsigned int vdest, vsrc, vmask;
-    vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
-	dest_mask, mask_mask, src_mask, store_mask;
+    vector unsigned char tmp1, tmp2, mask_mask, src_mask;
+
+    while (width && ((uintptr_t)dest & 15))
+    {
+	uint32_t a = *mask++;
+	uint32_t s = *src++;
+	uint32_t d = *dest;
+
+	UN8x4_MUL_UN8x4 (s, a);
+	UN8x4_ADD_UN8x4 (s, d);
+
+	*dest++ = s;
+	width--;
+    }
 
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
diff --git a/pixman/test/Makefile.am b/pixman/test/Makefile.am
index 5d901d572..88dc36d2a 100644
--- a/pixman/test/Makefile.am
+++ b/pixman/test/Makefile.am
@@ -1,8 +1,8 @@
 include $(top_srcdir)/test/Makefile.sources
 
-AM_CFLAGS = $(OPENMP_CFLAGS)
-AM_LDFLAGS = $(OPENMP_CFLAGS) $(TESTPROGS_EXTRA_LDFLAGS)
-LDADD = libutils.la $(top_builddir)/pixman/libpixman-1.la -lm  $(PNG_LIBS)
+AM_CFLAGS = $(OPENMP_CFLAGS) $(PTHREAD_CFLAGS)
+AM_LDFLAGS = $(OPENMP_CFLAGS) $(TESTPROGS_EXTRA_LDFLAGS) $(PTHREAD_LDFLAGS)
+LDADD = libutils.la $(top_builddir)/pixman/libpixman-1.la -lm  $(PNG_LIBS) $(PTHREAD_LIBS)
 AM_CPPFLAGS = -I$(top_srcdir)/pixman -I$(top_builddir)/pixman $(PNG_CFLAGS)
 
 libutils_la_SOURCES = $(libutils_sources) $(libutils_headers)
diff --git a/pixman/test/Makefile.sources b/pixman/test/Makefile.sources
index 2fabdb574..2ae5d9f8d 100644
--- a/pixman/test/Makefile.sources
+++ b/pixman/test/Makefile.sources
@@ -13,6 +13,7 @@ TESTPROGRAMS =			\
 	infinite-loop		\
 	trap-crasher		\
 	alpha-loop		\
+	thread-test		\
 	scaling-crash-test	\
 	scaling-helpers-test	\
 	gradient-crash-test	\
diff --git a/pixman/test/thread-test.c b/pixman/test/thread-test.c
new file mode 100644
index 000000000..f24c31d06
--- /dev/null
+++ b/pixman/test/thread-test.c
@@ -0,0 +1,197 @@
+#include <config.h>
+
+#ifndef HAVE_PTHREADS
+
+int main ()
+{
+    printf ("Skipped thread-test - pthreads not supported\n");
+    return 0;
+}
+
+#else
+
+#include <stdlib.h>
+#include <pthread.h>
+#include "utils.h"
+
+typedef struct
+{
+    int       thread_no;
+    uint32_t *dst_buf;
+} info_t;
+
+static const pixman_op_t operators[] = 
+{
+    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,
+};
+
+static const pixman_format_code_t formats[] =
+{
+    PIXMAN_a8r8g8b8,
+    PIXMAN_r5g6b5,
+    PIXMAN_a8,
+    PIXMAN_a4,
+    PIXMAN_a1,
+    PIXMAN_b5g6r5,
+    PIXMAN_r8g8b8a8,
+    PIXMAN_a4r4g4b4
+};
+
+#define N_ROUNDS 8192
+
+#define RAND_ELT(arr)							\
+    arr[prng_rand() % ARRAY_LENGTH (arr)]
+
+#define DEST_WIDTH (7)
+
+static void *
+thread (void *data)
+{
+    info_t *info = data;
+    uint32_t crc32 = 0x0;
+    uint32_t src_buf[64];
+    pixman_image_t *dst_img, *src_img;
+    int i;
+
+    prng_srand (info->thread_no);
+
+    for (i = 0; i < N_ROUNDS; ++i)
+    {
+	pixman_op_t op;
+	int rand1, rand2;
+
+	prng_randmemset (info->dst_buf, DEST_WIDTH * sizeof (uint32_t), 0);
+	prng_randmemset (src_buf, sizeof (src_buf), 0);
+
+	src_img = pixman_image_create_bits (
+	    RAND_ELT (formats), 4, 4, src_buf, 16);
+	dst_img = pixman_image_create_bits (
+	    RAND_ELT (formats), DEST_WIDTH, 1, info->dst_buf,
+	    DEST_WIDTH * sizeof (uint32_t));
+
+	image_endian_swap (src_img);
+	image_endian_swap (dst_img);
+	
+	rand2 = prng_rand() % 4;
+	rand1 = prng_rand() % 4;
+	op = RAND_ELT (operators);
+
+	pixman_image_composite32 (
+	    op,
+	    src_img, NULL, dst_img,
+	    rand1, rand2, 0, 0, 0, 0, DEST_WIDTH, 1);
+
+	crc32 = compute_crc32_for_image (crc32, dst_img);
+
+	pixman_image_unref (src_img);
+	pixman_image_unref (dst_img);
+    }
+
+    return (void *)(uintptr_t)crc32;
+}
+
+static inline uint32_t
+byteswap32 (uint32_t x)
+{
+    return ((x & ((uint32_t)0xFF << 24)) >> 24) |
+           ((x & ((uint32_t)0xFF << 16)) >>  8) |
+           ((x & ((uint32_t)0xFF <<  8)) <<  8) |
+           ((x & ((uint32_t)0xFF <<  0)) << 24);
+}
+
+int
+main (void)
+{
+    uint32_t dest[16 * DEST_WIDTH];
+    info_t info[16] = { { 0 } };
+    pthread_t threads[16];
+    void *retvals[16];
+    uint32_t crc32s[16], crc32;
+    int i;
+
+    for (i = 0; i < 16; ++i)
+    {
+	info[i].thread_no = i;
+	info[i].dst_buf = &dest[i * DEST_WIDTH];
+    }
+
+    for (i = 0; i < 16; ++i)
+	pthread_create (&threads[i], NULL, thread, &info[i]);
+
+    for (i = 0; i < 16; ++i)
+	pthread_join (threads[i], &retvals[i]);
+
+    for (i = 0; i < 16; ++i)
+    {
+	crc32s[i] = (uintptr_t)retvals[i];
+
+	if (is_little_endian())
+	    crc32s[i] = byteswap32 (crc32s[i]);
+    }
+
+    crc32 = compute_crc32 (0, crc32s, sizeof crc32s);
+
+#define EXPECTED 0xFD497D8D
+
+    if (crc32 != EXPECTED)
+    {
+	printf ("thread-test failed. Got checksum 0x%08X, expected 0x%08X\n",
+		crc32, EXPECTED);
+	return 1;
+    }
+
+    return 0;
+}
+
+#endif
+