replace a few and/sub/... by cmov

this is faster on P3, should be faster on AMD, and should be slower on P4
its disabled by default (benchmarks welcome so we know when to enable it)

Originally committed as revision 6615 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/cabac.h

@@ -459,6 +459,14 @@ static int get_cabac(CABACContext *c, uint8_t * const state){
         "movl "LOW      "(%2), %%ebx            \n\t"
 //eax:state ebx:low, edx:range, esi:RangeLPS
         "subl %%esi, %%edx                      \n\t"
+#ifdef CMOV_IS_FAST //FIXME actually define this somewhere
+        "cmpl %%ebx, %%edx                      \n\t"
+        "cmova %%edx, %%esi                     \n\t"
+        "sbbl %%ecx, %%ecx                      \n\t"
+        "andl %%ecx, %%edx                      \n\t"
+        "subl %%edx, %%ebx                      \n\t"
+        "xorl %%ecx, %%eax                      \n\t"
+#else
         "movl %%edx, %%ecx                      \n\t"
         "subl %%ebx, %%edx                      \n\t"
         "sarl $31, %%edx                        \n\t" //lps_mask
@@ -467,9 +475,10 @@ static int get_cabac(CABACContext *c, uint8_t * const state){
         "addl %%ecx, %%esi                      \n\t" //new range
         "andl %%edx, %%ecx                      \n\t"
         "subl %%ecx, %%ebx                      \n\t"
+        "xorl %%edx, %%eax                      \n\t"
+#endif
 
 //eax:state ebx:low edx:mask esi:range
-        "xorl %%edx, %%eax                      \n\t"
         "movzbl "MPS_STATE"(%2, %%eax), %%ecx   \n\t"
         "movb %%cl, (%1)                        \n\t"