put loop counter in a register if possible. makes some of the qpel functions 3% faster.

Originally committed as revision 11876 to svn://svn.ffmpeg.org/ffmpeg/trunk

diff --git a/libavcodec/i386/dsputil_mmx.c b/libavcodec/i386/dsputil_mmx.c
index 611ee37..5534318 100644 (file)
--- a/libavcodec/i386/dsputil_mmx.c
+++ b/libavcodec/i386/dsputil_mmx.c
@@ -2061,7 +2061,7 @@ static void OPNAME ## mpeg4_qpel16_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, in
         "add %4, %1                       \n\t"\
         "decl %2                          \n\t"\
         " jnz 1b                          \n\t"\
-        : "+a"(src), "+c"(dst), "+m"(h)\
+        : "+a"(src), "+c"(dst), "+g"(h)\
         : "d"((long)srcStride), "S"((long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\
         : "memory"\
     );\
@@ -2174,7 +2174,7 @@ static void OPNAME ## mpeg4_qpel8_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, int
         "add %4, %1                       \n\t"\
         "decl %2                          \n\t"\
         " jnz 1b                          \n\t"\
-        : "+a"(src), "+c"(dst), "+m"(h)\
+        : "+a"(src), "+c"(dst), "+g"(h)\
         : "S"((long)srcStride), "D"((long)dstStride), /*"m"(ff_pw_20), "m"(ff_pw_3),*/ "m"(temp), "m"(ROUNDER)\
         : "memory"\
     );\

diff --git a/libavcodec/i386/h264dsp_mmx.c b/libavcodec/i386/h264dsp_mmx.c
index 4dc5732..340d9fa 100644 (file)
--- a/libavcodec/i386/h264dsp_mmx.c
+++ b/libavcodec/i386/h264dsp_mmx.c
@@ -796,7 +796,7 @@ static av_noinline void OPNAME ## h264_qpel4_h_lowpass_ ## MMX(uint8_t *dst, uin
         "add %4, %1                 \n\t"\
         "decl %2                    \n\t"\
         " jnz 1b                    \n\t"\
-        : "+a"(src), "+c"(dst), "+m"(h)\
+        : "+a"(src), "+c"(dst), "+g"(h)\
         : "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
         : "memory"\
     );\
@@ -930,7 +930,7 @@ static av_noinline void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, in
         "add %3, %1                 \n\t"\
         "decl %2                    \n\t"\
         " jnz 1b                    \n\t"\
-        : "+a"(tmp), "+c"(dst), "+m"(h)\
+        : "+a"(tmp), "+c"(dst), "+g"(h)\
         : "S"((long)dstStride)\
         : "memory"\
     );\
@@ -987,7 +987,7 @@ static av_noinline void OPNAME ## h264_qpel8_h_lowpass_ ## MMX(uint8_t *dst, uin
         "add %4, %1                 \n\t"\
         "decl %2                    \n\t"\
         " jnz 1b                    \n\t"\
-        : "+a"(src), "+c"(dst), "+m"(h)\
+        : "+a"(src), "+c"(dst), "+g"(h)\
         : "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
         : "memory"\
     );\
@@ -1201,7 +1201,7 @@ static av_noinline void OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst
         "add %3, %1                 \n\t"\
         "decl %2                    \n\t"\
         " jnz 1b                    \n\t"\
-        : "+a"(tmp), "+c"(dst), "+m"(h)\
+        : "+a"(tmp), "+c"(dst), "+g"(h)\
         : "S"((long)dstStride)\
         : "memory"\
     );\