Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

Move static inline function to a macro, so that constant propagation in

Browse Source 
inline asm works for gcc-3.x also (hopefully). Should fix gcc-3.x FATE
breakage after r25254.

Originally committed as revision 25262 to svn://svn.ffmpeg.org/ffmpeg/trunk
oldabi
Ronald S. Bultje 15 years ago
parent fc7c40c2bd
commit a52ffc3f54
1 changed files with 113 additions and 117 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 230
      libavcodec/x86/h264dsp_mmx.c

230
libavcodec/x86/h264dsp_mmx.c
Unescape View File

@ -63,123 +63,119 @@ void ff_h264_idct_add8_sse2 (uint8_t **dest, const int *block_offset, DCTEL
/***********************************/ /***********************************/
/* deblocking */ /* deblocking */
static av_always_inline void h264_loop_filter_strength_iteration_mmx2(int16_t bS[2][4][4], uint8_t nnz[40], #define h264_loop_filter_strength_iteration_mmx2(bS, nz, ref, mv, bidir, edges, step, mask_mv, dir, d_idx, mask_dir) \
int8_t ref[2][40], int16_t mv[2][40][2], do { \
int bidir, int edges, int step, x86_reg b_idx; \
int mask_mv, int dir, const int d_idx, mask_mv <<= 3; \
const uint64_t mask_dir) for( b_idx=0; b_idx<edges; b_idx+=step ) { \
{ if (!mask_dir) \
x86_reg b_idx; __asm__ volatile( \
mask_mv <<= 3; "pxor %%mm0, %%mm0 \n\t" \
for( b_idx=0; b_idx<edges; b_idx+=step ) { :: \
if (!mask_dir) ); \
__asm__ volatile( if(!(mask_mv & b_idx)) { \
"pxor %%mm0, %%mm0 \n\t" if(bidir) { \
:: __asm__ volatile( \
); "movd %a3(%0,%2), %%mm2 \n" \
if(!(mask_mv & b_idx)) { "punpckldq %a4(%0,%2), %%mm2 \n" /* { ref0[bn], ref1[bn] } */ \
if(bidir) { "pshufw $0x44, 12(%0,%2), %%mm0 \n" /* { ref0[b], ref0[b] } */ \
__asm__ volatile( "pshufw $0x44, 52(%0,%2), %%mm1 \n" /* { ref1[b], ref1[b] } */ \
"movd %a3(%0,%2), %%mm2 \n" "pshufw $0x4E, %%mm2, %%mm3 \n" \
"punpckldq %a4(%0,%2), %%mm2 \n" // { ref0[bn], ref1[bn] } "psubb %%mm2, %%mm0 \n" /* { ref0[b]!=ref0[bn], ref0[b]!=ref1[bn] } */ \
"pshufw $0x44, 12(%0,%2), %%mm0 \n" // { ref0[b], ref0[b] } "psubb %%mm3, %%mm1 \n" /* { ref1[b]!=ref1[bn], ref1[b]!=ref0[bn] } */ \
"pshufw $0x44, 52(%0,%2), %%mm1 \n" // { ref1[b], ref1[b] } \
"pshufw $0x4E, %%mm2, %%mm3 \n" "por %%mm1, %%mm0 \n" \
"psubb %%mm2, %%mm0 \n" // { ref0[b]!=ref0[bn], ref0[b]!=ref1[bn] } "movq %a5(%1,%2,4), %%mm1 \n" \
"psubb %%mm3, %%mm1 \n" // { ref1[b]!=ref1[bn], ref1[b]!=ref0[bn] } "movq %a6(%1,%2,4), %%mm2 \n" \
"movq %%mm1, %%mm3 \n" \
"por %%mm1, %%mm0 \n" "movq %%mm2, %%mm4 \n" \
"movq %a5(%1,%2,4), %%mm1 \n" "psubw 48(%1,%2,4), %%mm1 \n" \
"movq %a6(%1,%2,4), %%mm2 \n" "psubw 56(%1,%2,4), %%mm2 \n" \
"movq %%mm1, %%mm3 \n" "psubw 208(%1,%2,4), %%mm3 \n" \
"movq %%mm2, %%mm4 \n" "psubw 216(%1,%2,4), %%mm4 \n" \
"psubw 48(%1,%2,4), %%mm1 \n" "packsswb %%mm2, %%mm1 \n" \
"psubw 56(%1,%2,4), %%mm2 \n" "packsswb %%mm4, %%mm3 \n" \
"psubw 208(%1,%2,4), %%mm3 \n" "paddb %%mm6, %%mm1 \n" \
"psubw 216(%1,%2,4), %%mm4 \n" "paddb %%mm6, %%mm3 \n" \
"packsswb %%mm2, %%mm1 \n" "psubusb %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \
"packsswb %%mm4, %%mm3 \n" "psubusb %%mm5, %%mm3 \n" \
"paddb %%mm6, %%mm1 \n" "packsswb %%mm3, %%mm1 \n" \
"paddb %%mm6, %%mm3 \n" \
"psubusb %%mm5, %%mm1 \n" // abs(mv[b] - mv[bn]) >= limit "por %%mm1, %%mm0 \n" \
"psubusb %%mm5, %%mm3 \n" "movq %a7(%1,%2,4), %%mm1 \n" \
"packsswb %%mm3, %%mm1 \n" "movq %a8(%1,%2,4), %%mm2 \n" \
"movq %%mm1, %%mm3 \n" \
"por %%mm1, %%mm0 \n" "movq %%mm2, %%mm4 \n" \
"movq %a7(%1,%2,4), %%mm1 \n" "psubw 48(%1,%2,4), %%mm1 \n" \
"movq %a8(%1,%2,4), %%mm2 \n" "psubw 56(%1,%2,4), %%mm2 \n" \
"movq %%mm1, %%mm3 \n" "psubw 208(%1,%2,4), %%mm3 \n" \
"movq %%mm2, %%mm4 \n" "psubw 216(%1,%2,4), %%mm4 \n" \
"psubw 48(%1,%2,4), %%mm1 \n" "packsswb %%mm2, %%mm1 \n" \
"psubw 56(%1,%2,4), %%mm2 \n" "packsswb %%mm4, %%mm3 \n" \
"psubw 208(%1,%2,4), %%mm3 \n" "paddb %%mm6, %%mm1 \n" \
"psubw 216(%1,%2,4), %%mm4 \n" "paddb %%mm6, %%mm3 \n" \
"packsswb %%mm2, %%mm1 \n" "psubusb %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \
"packsswb %%mm4, %%mm3 \n" "psubusb %%mm5, %%mm3 \n" \
"paddb %%mm6, %%mm1 \n" "packsswb %%mm3, %%mm1 \n" \
"paddb %%mm6, %%mm3 \n" \
"psubusb %%mm5, %%mm1 \n" // abs(mv[b] - mv[bn]) >= limit "pshufw $0x4E, %%mm1, %%mm1 \n" \
"psubusb %%mm5, %%mm3 \n" "por %%mm1, %%mm0 \n" \
"packsswb %%mm3, %%mm1 \n" "pshufw $0x4E, %%mm0, %%mm1 \n" \
"pminub %%mm1, %%mm0 \n" \
"pshufw $0x4E, %%mm1, %%mm1 \n" ::"r"(ref), \
"por %%mm1, %%mm0 \n" "r"(mv), \
"pshufw $0x4E, %%mm0, %%mm1 \n" "r"(b_idx), \
"pminub %%mm1, %%mm0 \n" "i"(d_idx+12), \
::"r"(ref), "i"(d_idx+52), \
"r"(mv), "i"(d_idx*4+48), \
"r"(b_idx), "i"(d_idx*4+56), \
"i"(d_idx+12), "i"(d_idx*4+208), \
"i"(d_idx+52), "i"(d_idx*4+216) \
"i"(d_idx*4+48), ); \
"i"(d_idx*4+56), } else { \
"i"(d_idx*4+208), __asm__ volatile( \
"i"(d_idx*4+216) "movd 12(%0,%2), %%mm0 \n" \
); "psubb %a3(%0,%2), %%mm0 \n" /* ref[b] != ref[bn] */ \
} else { "movq 48(%1,%2,4), %%mm1 \n" \
__asm__ volatile( "movq 56(%1,%2,4), %%mm2 \n" \
"movd 12(%0,%2), %%mm0 \n" "psubw %a4(%1,%2,4), %%mm1 \n" \
"psubb %a3(%0,%2), %%mm0 \n" // ref[b] != ref[bn] "psubw %a5(%1,%2,4), %%mm2 \n" \
"movq 48(%1,%2,4), %%mm1 \n" "packsswb %%mm2, %%mm1 \n" \
"movq 56(%1,%2,4), %%mm2 \n" "paddb %%mm6, %%mm1 \n" \
"psubw %a4(%1,%2,4), %%mm1 \n" "psubusb %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \
"psubw %a5(%1,%2,4), %%mm2 \n" "packsswb %%mm1, %%mm1 \n" \
"packsswb %%mm2, %%mm1 \n" "por %%mm1, %%mm0 \n" \
"paddb %%mm6, %%mm1 \n" ::"r"(ref), \
"psubusb %%mm5, %%mm1 \n" // abs(mv[b] - mv[bn]) >= limit "r"(mv), \
"packsswb %%mm1, %%mm1 \n" "r"(b_idx), \
"por %%mm1, %%mm0 \n" "i"(d_idx+12), \
::"r"(ref), "i"(d_idx*4+48), \
"r"(mv), "i"(d_idx*4+56) \
"r"(b_idx), ); \
"i"(d_idx+12), } \
"i"(d_idx*4+48), } \
"i"(d_idx*4+56) __asm__ volatile( \
); "movd 12(%0,%1), %%mm1 \n" \
} "por %a2(%0,%1), %%mm1 \n" /* nnz[b] || nnz[bn] */ \
} ::"r"(nnz), \
__asm__ volatile( "r"(b_idx), \
"movd 12(%0,%1), %%mm1 \n" "i"(d_idx+12) \
"por %a2(%0,%1), %%mm1 \n" // nnz[b] || nnz[bn] ); \
::"r"(nnz), __asm__ volatile( \
"r"(b_idx), "pminub %%mm7, %%mm1 \n" \
"i"(d_idx+12) "pminub %%mm7, %%mm0 \n" \
); "psllw $1, %%mm1 \n" \
__asm__ volatile( "pxor %%mm2, %%mm2 \n" \
"pminub %%mm7, %%mm1 \n" "pmaxub %%mm0, %%mm1 \n" \
"pminub %%mm7, %%mm0 \n" "punpcklbw %%mm2, %%mm1 \n" \
"psllw $1, %%mm1 \n" "movq %%mm1, %a1(%0,%2) \n" \
"pxor %%mm2, %%mm2 \n" ::"r"(bS), \
"pmaxub %%mm0, %%mm1 \n" "i"(32*dir), \
"punpcklbw %%mm2, %%mm1 \n" "r"(b_idx) \
"movq %%mm1, %a1(%0,%2) \n" :"memory" \
::"r"(bS), ); \
"i"(32*dir), } \
"r"(b_idx) } while (0)
:"memory"
);
}
}
static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2], static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2],
int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field ) { int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field ) {

Write Preview
Loading…
Cancel
Save