Altivec-optimized blitters!

Vast majority of this work is compliments of Bob Ippolito.

http://www.devolution.com/pipermail/sdl/2005-February/067466.html and many
 other posts.

--HG--
extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%401048

configure.in

@@ -1839,17 +1839,18 @@ CheckAltivec()
 {
     AC_MSG_CHECKING(for GCC Altivec instruction support)
     have_gcc_altivec=no
+    save_CFLAGS="${CFLAGS}"
+    CFLAGS="${CFLAGS} -DGCC_ALTIVEC -DUSE_ALTIVEC_BLITTERS -faltivec"
     AC_TRY_COMPILE([
+    vector unsigned int vzero() {
+        return vec_splat_u32(0);
+    }
     ],[
-        asm volatile ("mtspr 256, %0\n\t"
-                      "vand %%v0, %%v0, %%v0"
-                      :
-                      : "r" (-1));
     ],[
     have_gcc_altivec=yes
     ])
-    if test x$have_gcc_altivec = xyes; then
-        CFLAGS="$CFLAGS -DGCC_ALTIVEC"
+    if test x$have_gcc_altivec = xno; then
+        CFLAGS="${save_CFLAGS}"
     fi
     AC_MSG_RESULT($have_gcc_altivec)
 }
@@ -2564,6 +2565,7 @@ case "$target" in
         CheckMacGL
         CheckPTHREAD
         CheckSIGACTION
+        CheckAltivec
         # If either the audio or CD driver is used, add the AudioUnit framework
         if test x$enable_audio = xyes -o x$enable_cdrom = xyes; then
             SYSTEM_LIBS="$SYSTEM_LIBS -framework AudioToolbox -framework AudioUnit"

src/video/SDL_blit.h

@@ -374,6 +374,20 @@ do {						\
 	dB = (((sB-dB)*(A))>>8)+dB;		\
 } while(0)
 
+/* Blend the RGB values of two pixels based on a source alpha value */
+#define ACCURATE_ALPHA_BLEND(sR, sG, sB, sA, dR, dG, dB)	\
+do {						\
+    unsigned tR, tG, tB, tA; \
+    tA = 255 - sA; \
+    tR = 1 + (sR * sA) + (dR * tA); \
+    dR = (tR + (tR >> 8)) >> 8; \
+    tG = 1 + (sG * sA) + (dG * tA); \
+    dG = (tG + (tG >> 8)) >> 8; \
+    tB = 1 + (sB * sA) + (dB * tA); \
+    dB = (tB + (tB >> 8)) >> 8; \
+} while(0)
+
+
 /* This is a very useful loop for optimizing blitters */
 #if defined(_MSC_VER) && (_MSC_VER == 1300)
 /* There's a bug in the Visual C++ 7 optimizer when compiling this code */