diff options
Diffstat (limited to 'libavcodec/x86/cavsdsp.c')
-rw-r--r-- | libavcodec/x86/cavsdsp.c | 169 |
1 files changed, 101 insertions, 68 deletions
diff --git a/libavcodec/x86/cavsdsp.c b/libavcodec/x86/cavsdsp.c index b323a10..2ac3bb5 100644 --- a/libavcodec/x86/cavsdsp.c +++ b/libavcodec/x86/cavsdsp.c @@ -5,20 +5,20 @@ * MMX-optimized DSP functions, based on H.264 optimizations by * Michael Niedermayer and Loren Merritt * - * This file is part of Libav. + * This file is part of FFmpeg. * - * Libav is free software; you can redistribute it and/or + * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * - * Libav is distributed in the hope that it will be useful, + * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with Libav; if not, write to the Free Software + * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -139,11 +139,9 @@ static inline void cavs_idct8_1d(int16_t *block, uint64_t bias) static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride) { int i; - DECLARE_ALIGNED(8, int16_t, b2)[64]; + LOCAL_ALIGNED(16, int16_t, b2, [64]); for(i=0; i<2; i++){ - DECLARE_ALIGNED(8, uint64_t, tmp); - cavs_idct8_1d(block+4*i, ff_pw_4.a); __asm__ volatile( @@ -155,19 +153,19 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride) "psraw $3, %%mm2 \n\t" "psraw $3, %%mm1 \n\t" "psraw $3, %%mm0 \n\t" - "movq %%mm7, %0 \n\t" + "movq %%mm7, (%0) \n\t" TRANSPOSE4( %%mm0, %%mm2, %%mm4, %%mm6, %%mm7 ) - "movq %%mm0, 8(%1) \n\t" - "movq %%mm6, 24(%1) \n\t" - "movq %%mm7, 40(%1) \n\t" - "movq %%mm4, 56(%1) \n\t" - "movq %0, %%mm7 \n\t" + "movq %%mm0, 8(%0) \n\t" + "movq %%mm6, 24(%0) \n\t" + "movq %%mm7, 40(%0) \n\t" + "movq %%mm4, 56(%0) \n\t" + "movq (%0), %%mm7 \n\t" TRANSPOSE4( %%mm7, %%mm5, %%mm3, %%mm1, %%mm0 ) - "movq %%mm7, (%1) \n\t" - "movq %%mm1, 16(%1) \n\t" - "movq %%mm0, 32(%1) \n\t" - "movq %%mm3, 48(%1) \n\t" - : "=m"(tmp) + "movq %%mm7, (%0) \n\t" + "movq %%mm1, 16(%0) \n\t" + "movq %%mm0, 32(%0) \n\t" + "movq %%mm3, 48(%0) \n\t" + : : "r"(b2+32*i) : "memory" ); @@ -198,7 +196,7 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride) ); } - ff_add_pixels_clamped_mmx(b2, dst, stride); + ff_add_pixels_clamped(b2, dst, stride); } #endif /* HAVE_MMX_INLINE */ @@ -212,10 +210,10 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride) ****************************************************************************/ /* vertical filter [-1 -2 96 42 -7 0] */ -#define QPEL_CAVSV1(A,B,C,D,E,F,OP,MUL2) \ +#define QPEL_CAVSV1(A,B,C,D,E,F,OP,ADD, MUL1, MUL2) \ "movd (%0), "#F" \n\t"\ "movq "#C", %%mm6 \n\t"\ - "pmullw %5, %%mm6 \n\t"\ + "pmullw "MANGLE(MUL1)", %%mm6\n\t"\ "movq "#D", %%mm7 \n\t"\ "pmullw "MANGLE(MUL2)", %%mm7\n\t"\ "psllw $3, "#E" \n\t"\ @@ -230,35 +228,35 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride) "psubw "#B", %%mm6 \n\t"\ "psraw $1, "#B" \n\t"\ "psubw "#A", %%mm6 \n\t"\ - "paddw %4, %%mm6 \n\t"\ + "paddw "MANGLE(ADD)", %%mm6 \n\t"\ "psraw $7, %%mm6 \n\t"\ "packuswb %%mm6, %%mm6 \n\t"\ OP(%%mm6, (%1), A, d) \ "add %3, %1 \n\t" /* vertical filter [ 0 -1 5 5 -1 0] */ -#define QPEL_CAVSV2(A,B,C,D,E,F,OP,MUL2) \ +#define QPEL_CAVSV2(A,B,C,D,E,F,OP,ADD, MUL1, MUL2) \ "movd (%0), "#F" \n\t"\ "movq "#C", %%mm6 \n\t"\ "paddw "#D", %%mm6 \n\t"\ - "pmullw %5, %%mm6 \n\t"\ + "pmullw "MANGLE(MUL1)", %%mm6\n\t"\ "add %2, %0 \n\t"\ "punpcklbw %%mm7, "#F" \n\t"\ "psubw "#B", %%mm6 \n\t"\ "psubw "#E", %%mm6 \n\t"\ - "paddw %4, %%mm6 \n\t"\ + "paddw "MANGLE(ADD)", %%mm6 \n\t"\ "psraw $3, %%mm6 \n\t"\ "packuswb %%mm6, %%mm6 \n\t"\ OP(%%mm6, (%1), A, d) \ "add %3, %1 \n\t" /* vertical filter [ 0 -7 42 96 -2 -1] */ -#define QPEL_CAVSV3(A,B,C,D,E,F,OP,MUL2) \ +#define QPEL_CAVSV3(A,B,C,D,E,F,OP,ADD, MUL1, MUL2) \ "movd (%0), "#F" \n\t"\ "movq "#C", %%mm6 \n\t"\ "pmullw "MANGLE(MUL2)", %%mm6\n\t"\ "movq "#D", %%mm7 \n\t"\ - "pmullw %5, %%mm7 \n\t"\ + "pmullw "MANGLE(MUL1)", %%mm7\n\t"\ "psllw $3, "#B" \n\t"\ "psubw "#B", %%mm6 \n\t"\ "psraw $3, "#B" \n\t"\ @@ -271,7 +269,7 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride) "psubw "#E", %%mm6 \n\t"\ "psraw $1, "#E" \n\t"\ "psubw "#F", %%mm6 \n\t"\ - "paddw %4, %%mm6 \n\t"\ + "paddw "MANGLE(ADD)", %%mm6 \n\t"\ "psraw $7, %%mm6 \n\t"\ "packuswb %%mm6, %%mm6 \n\t"\ OP(%%mm6, (%1), A, d) \ @@ -300,32 +298,34 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride) "punpcklbw %%mm7, %%mm2 \n\t"\ "punpcklbw %%mm7, %%mm3 \n\t"\ "punpcklbw %%mm7, %%mm4 \n\t"\ - VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, MUL2)\ - VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, MUL2)\ - VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, MUL2)\ - VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, MUL2)\ - VOP(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP, MUL2)\ - VOP(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP, MUL2)\ - VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, MUL2)\ - VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, MUL2)\ + VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, ADD, MUL1, MUL2)\ + VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, ADD, MUL1, MUL2)\ + VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, ADD, MUL1, MUL2)\ + VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, ADD, MUL1, MUL2)\ + VOP(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP, ADD, MUL1, MUL2)\ + VOP(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP, ADD, MUL1, MUL2)\ + VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, ADD, MUL1, MUL2)\ + VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, ADD, MUL1, MUL2)\ \ : "+a"(src), "+c"(dst)\ - : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride), "m"(ADD), "m"(MUL1)\ + : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride)\ + NAMED_CONSTRAINTS_ADD(ADD,MUL1,MUL2)\ : "memory"\ );\ if(h==16){\ __asm__ volatile(\ - VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, MUL2)\ - VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, MUL2)\ - VOP(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP, MUL2)\ - VOP(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP, MUL2)\ - VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, MUL2)\ - VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, MUL2)\ - VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, MUL2)\ - VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, MUL2)\ + VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, ADD, MUL1, MUL2)\ + VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, ADD, MUL1, MUL2)\ + VOP(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP, ADD, MUL1, MUL2)\ + VOP(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP, ADD, MUL1, MUL2)\ + VOP(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP, ADD, MUL1, MUL2)\ + VOP(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP, ADD, MUL1, MUL2)\ + VOP(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP, ADD, MUL1, MUL2)\ + VOP(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP, ADD, MUL1, MUL2)\ \ : "+a"(src), "+c"(dst)\ - : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride), "m"(ADD), "m"(MUL1)\ + : "S"((x86_reg)srcStride), "r"((x86_reg)dstStride)\ + NAMED_CONSTRAINTS_ADD(ADD,MUL1,MUL2)\ : "memory"\ );\ }\ @@ -338,7 +338,7 @@ static void OPNAME ## cavs_qpel8_h_ ## MMX(uint8_t *dst, const uint8_t *src, int int h=8;\ __asm__ volatile(\ "pxor %%mm7, %%mm7 \n\t"\ - "movq %5, %%mm6 \n\t"\ + "movq "MANGLE(ff_pw_5)", %%mm6\n\t"\ "1: \n\t"\ "movq (%0), %%mm0 \n\t"\ "movq 1(%0), %%mm2 \n\t"\ @@ -364,7 +364,7 @@ static void OPNAME ## cavs_qpel8_h_ ## MMX(uint8_t *dst, const uint8_t *src, int "paddw %%mm3, %%mm5 \n\t"\ "psubw %%mm2, %%mm0 \n\t"\ "psubw %%mm5, %%mm1 \n\t"\ - "movq %6, %%mm5 \n\t"\ + "movq "MANGLE(ff_pw_4)", %%mm5\n\t"\ "paddw %%mm5, %%mm0 \n\t"\ "paddw %%mm5, %%mm1 \n\t"\ "psraw $3, %%mm0 \n\t"\ @@ -376,7 +376,8 @@ static void OPNAME ## cavs_qpel8_h_ ## MMX(uint8_t *dst, const uint8_t *src, int "decl %2 \n\t"\ " jnz 1b \n\t"\ : "+a"(src), "+c"(dst), "+m"(h)\ - : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride), "m"(ff_pw_5), "m"(ff_pw_4)\ + : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride)\ + NAMED_CONSTRAINTS_ADD(ff_pw_4,ff_pw_5)\ : "memory"\ );\ }\ @@ -386,7 +387,7 @@ static inline void OPNAME ## cavs_qpel8or16_v1_ ## MMX(uint8_t *dst, const uint8 }\ \ static inline void OPNAME ## cavs_qpel8or16_v2_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h){\ - QPEL_CAVSVNUM(QPEL_CAVSV2,OP,ff_pw_4,ff_pw_5,ff_pw_5) \ + QPEL_CAVSVNUM(QPEL_CAVSV2,OP,ff_pw_4,ff_pw_5,ff_pw_42) \ }\ \ static inline void OPNAME ## cavs_qpel8or16_v3_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h){\ @@ -459,7 +460,7 @@ static void OPNAME ## cavs_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, const uin #endif /* (HAVE_MMXEXT_INLINE || HAVE_AMD3DNOW_INLINE) */ -#if HAVE_MMX_INLINE +#if HAVE_MMX_EXTERNAL static void put_cavs_qpel8_mc00_mmx(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) { @@ -472,6 +473,12 @@ static void avg_cavs_qpel8_mc00_mmx(uint8_t *dst, const uint8_t *src, ff_avg_pixels8_mmx(dst, src, stride, 8); } +static void avg_cavs_qpel8_mc00_mmxext(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + ff_avg_pixels8_mmxext(dst, src, stride, 8); +} + static void put_cavs_qpel16_mc00_mmx(uint8_t *dst, const uint8_t *src, ptrdiff_t stride) { @@ -484,18 +491,40 @@ static void avg_cavs_qpel16_mc00_mmx(uint8_t *dst, const uint8_t *src, ff_avg_pixels16_mmx(dst, src, stride, 16); } +static void avg_cavs_qpel16_mc00_mmxext(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + ff_avg_pixels16_mmxext(dst, src, stride, 16); +} + +static void put_cavs_qpel16_mc00_sse2(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + ff_put_pixels16_sse2(dst, src, stride, 16); +} + +static void avg_cavs_qpel16_mc00_sse2(uint8_t *dst, const uint8_t *src, + ptrdiff_t stride) +{ + ff_avg_pixels16_sse2(dst, src, stride, 16); +} +#endif + static av_cold void cavsdsp_init_mmx(CAVSDSPContext *c, AVCodecContext *avctx) { +#if HAVE_MMX_EXTERNAL c->put_cavs_qpel_pixels_tab[0][0] = put_cavs_qpel16_mc00_mmx; c->put_cavs_qpel_pixels_tab[1][0] = put_cavs_qpel8_mc00_mmx; c->avg_cavs_qpel_pixels_tab[0][0] = avg_cavs_qpel16_mc00_mmx; c->avg_cavs_qpel_pixels_tab[1][0] = avg_cavs_qpel8_mc00_mmx; +#endif +#if HAVE_MMX_INLINE c->cavs_idct8_add = cavs_idct8_add_mmx; c->idct_perm = FF_IDCT_PERM_TRANSPOSE; -} #endif /* HAVE_MMX_INLINE */ +} #define DSPFUNC(PFX, IDX, NUM, EXT) \ c->PFX ## _cavs_qpel_pixels_tab[IDX][ 2] = PFX ## _cavs_qpel ## NUM ## _mc20_ ## EXT; \ @@ -511,15 +540,6 @@ CAVS_MC(put_, 8, mmxext) CAVS_MC(put_, 16, mmxext) CAVS_MC(avg_, 8, mmxext) CAVS_MC(avg_, 16, mmxext) - -static av_cold void cavsdsp_init_mmxext(CAVSDSPContext *c, - AVCodecContext *avctx) -{ - DSPFUNC(put, 0, 16, mmxext); - DSPFUNC(put, 1, 8, mmxext); - DSPFUNC(avg, 0, 16, mmxext); - DSPFUNC(avg, 1, 8, mmxext); -} #endif /* HAVE_MMXEXT_INLINE */ #if HAVE_AMD3DNOW_INLINE @@ -543,18 +563,31 @@ static av_cold void cavsdsp_init_3dnow(CAVSDSPContext *c, av_cold void ff_cavsdsp_init_x86(CAVSDSPContext *c, AVCodecContext *avctx) { -#if HAVE_MMX_INLINE int cpu_flags = av_get_cpu_flags(); - if (INLINE_MMX(cpu_flags)) - cavsdsp_init_mmx(c, avctx); -#endif /* HAVE_MMX_INLINE */ + cavsdsp_init_mmx(c, avctx); #if HAVE_AMD3DNOW_INLINE if (INLINE_AMD3DNOW(cpu_flags)) cavsdsp_init_3dnow(c, avctx); #endif /* HAVE_AMD3DNOW_INLINE */ #if HAVE_MMXEXT_INLINE - if (INLINE_MMXEXT(cpu_flags)) - cavsdsp_init_mmxext(c, avctx); -#endif /* HAVE_MMXEXT_INLINE */ + if (INLINE_MMXEXT(cpu_flags)) { + DSPFUNC(put, 0, 16, mmxext); + DSPFUNC(put, 1, 8, mmxext); + DSPFUNC(avg, 0, 16, mmxext); + DSPFUNC(avg, 1, 8, mmxext); + } +#endif +#if HAVE_MMX_EXTERNAL + if (EXTERNAL_MMXEXT(cpu_flags)) { + c->avg_cavs_qpel_pixels_tab[0][0] = avg_cavs_qpel16_mc00_mmxext; + c->avg_cavs_qpel_pixels_tab[1][0] = avg_cavs_qpel8_mc00_mmxext; + } +#endif +#if HAVE_SSE2_EXTERNAL + if (EXTERNAL_SSE2(cpu_flags)) { + c->put_cavs_qpel_pixels_tab[0][0] = put_cavs_qpel16_mc00_sse2; + c->avg_cavs_qpel_pixels_tab[0][0] = avg_cavs_qpel16_mc00_sse2; + } +#endif } |