Use constant offsets for memory operands since gcc is unable to

This fixes gcc failing to fit 6 memory locations into 7 registers on x86-32

Originally committed as revision 21337 to svn://svn.ffmpeg.org/ffmpeg/trunk
release/0.6
David Conrad 15 years ago
parent 613f2dc4ae
commit c4f2b6dce3
  1. 20
      libavcodec/x86/h264dsp_mmx.c

@ -834,9 +834,9 @@ static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40]
"por %%mm1, %%mm0 \n\t"
"movq %2, %%mm1 \n\t"
"movq %3, %%mm2 \n\t"
"psubw %4, %%mm1 \n\t"
"psubw %5, %%mm2 \n\t"
"movq 8+1*%2, %%mm2 \n\t"
"psubw %3, %%mm1 \n\t"
"psubw 8+1*%3, %%mm2 \n\t"
"packsswb %%mm2, %%mm1 \n\t"
"paddb %%mm5, %%mm1 \n\t"
"pminub %%mm4, %%mm1 \n\t"
@ -845,9 +845,7 @@ static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40]
::"m"(ref[l][b_idx]),
"m"(ref[l][b_idx+d_idx]),
"m"(mv[l][b_idx][0]),
"m"(mv[l][b_idx+2][0]),
"m"(mv[l][b_idx+d_idx][0]),
"m"(mv[l][b_idx+d_idx+2][0])
"m"(mv[l][b_idx+d_idx][0])
);
}
if(bidir==1){
@ -863,9 +861,9 @@ static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40]
"por %%mm1, %%mm3 \n\t"
"movq %2, %%mm1 \n\t"
"movq %3, %%mm2 \n\t"
"psubw %4, %%mm1 \n\t"
"psubw %5, %%mm2 \n\t"
"movq 8+1*%2, %%mm2 \n\t"
"psubw %3, %%mm1 \n\t"
"psubw 8+1*%3, %%mm2 \n\t"
"packsswb %%mm2, %%mm1 \n\t"
"paddb %%mm5, %%mm1 \n\t"
"pminub %%mm4, %%mm1 \n\t"
@ -874,9 +872,7 @@ static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40]
::"m"(ref[l][b_idx]),
"m"(ref[1-l][b_idx+d_idx]),
"m"(mv[l][b_idx][0]),
"m"(mv[l][b_idx+2][0]),
"m"(mv[1-l][b_idx+d_idx][0]),
"m"(mv[1-l][b_idx+d_idx+2][0])
"m"(mv[1-l][b_idx+d_idx][0])
);
}
__asm__ volatile(

Loading…
Cancel
Save