falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
Browse Source

Merge commit 'fa8fcab1e0d31074c0644c4ac5194474c6c26415' 

* commit 'fa8fcab1e0d31074c0644c4ac5194474c6c26415':
  x86: h264_chromamc_10bit: drop pointless PAVG %define
  x86: mmx2 ---> mmxext in function names
  swscale: do not forget to swap data in formats with different endianness

Conflicts:
	libavcodec/x86/dsputil_mmx.c
	libavfilter/x86/gradfun.c
	libswscale/input.c
	libswscale/utils.c
	libswscale/x86/swscale.c
	tests/ref/lavfi/pixfmts_scale

Merged-by: Michael Niedermayer <michaelni@gmx.at>
tags/n1.1
Michael Niedermayer 13 years ago
parent
14f1fa56b2 fa8fcab1e0
commit
add7513e64
24 changed files with 256 additions and 226 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +2
    -2
      libavcodec/dct-test.c
  2. +1
    -1
      libavcodec/dsputil.h
  3. +15
    -14
      libavcodec/x86/cavsdsp.c
  4. +1
    -1
      libavcodec/x86/diracdsp_mmx.c
  5. +1
    -1
      libavcodec/x86/diracdsp_mmx.h
  6. +73
    -73
      libavcodec/x86/dsputil_mmx.c
  7. +5
    -5
      libavcodec/x86/dsputil_mmx.h
  8. +16
    -9
      libavcodec/x86/dsputilenc_mmx.c
  9. +4
    -3
      libavcodec/x86/fdct.c
  10. +1
    -2
      libavcodec/x86/h264_chromamc_10bit.asm
  11. +27
    -27
      libavcodec/x86/h264_qpel.c
  12. +6
    -7
      libavcodec/x86/h264dsp_init.c
  13. +6
    -5
      libavcodec/x86/idct_mmx_xvid.c
  14. +3
    -3
      libavcodec/x86/idct_xvid.h
  15. +20
    -16
      libavcodec/x86/motion_est.c
  16. +3
    -3
      libavcodec/x86/mpegvideoenc.c
  17. +36
    -29
      libavcodec/x86/vc1dsp_mmx.c
  18. +4
    -2
      libavfilter/x86/gradfun.c
  19. +2
    -2
      libavfilter/x86/yadif.c
  20. +9
    -3
      libswscale/input.c
  21. +11
    -10
      libswscale/utils.c
  22. +2
    -2
      libswscale/x86/rgb2rgb.c
  23. +3
    -3
      libswscale/x86/swscale.c
  24. +5
    -3
      libswscale/x86/yuv2rgb.c

+ 2
- 2
libavcodec/dct-test.c View File

@@ -86,7 +86,7 @@ static const struct algo fdct_tab[] = {

#if HAVE_MMX_INLINE
{ "MMX", ff_fdct_mmx, NO_PERM, AV_CPU_FLAG_MMX },
{ "MMXEXT", ff_fdct_mmx2, NO_PERM, AV_CPU_FLAG_MMXEXT },
{ "MMXEXT", ff_fdct_mmxext, NO_PERM, AV_CPU_FLAG_MMXEXT },
{ "SSE2", ff_fdct_sse2, NO_PERM, AV_CPU_FLAG_SSE2 },
#endif

@@ -131,7 +131,7 @@ static const struct algo idct_tab[] = {
#endif
{ "SIMPLE-MMX", ff_simple_idct_mmx, MMX_SIMPLE_PERM, AV_CPU_FLAG_MMX },
{ "XVID-MMX", ff_idct_xvid_mmx, NO_PERM, AV_CPU_FLAG_MMX, 1 },
{ "XVID-MMXEXT", ff_idct_xvid_mmx2, NO_PERM, AV_CPU_FLAG_MMXEXT, 1 },
{ "XVID-MMXEXT", ff_idct_xvid_mmxext, NO_PERM, AV_CPU_FLAG_MMXEXT, 1 },
{ "XVID-SSE2", ff_idct_xvid_sse2, SSE2_PERM, AV_CPU_FLAG_SSE2, 1 },
#if ARCH_X86_64 && HAVE_YASM
{ "PR-SSE2", ff_prores_idct_put_10_sse2_wrap, TRANSPOSE_PERM, AV_CPU_FLAG_SSE2, 1 },


+ 1
- 1
libavcodec/dsputil.h View File

@@ -52,7 +52,7 @@ void ff_j_rev_dct1 (DCTELEM *data);
void ff_wmv2_idct_c(DCTELEM *data);

void ff_fdct_mmx(DCTELEM *block);
void ff_fdct_mmx2(DCTELEM *block);
void ff_fdct_mmxext(DCTELEM *block);
void ff_fdct_sse2(DCTELEM *block);

#define H264_IDCT(depth) \


+ 15
- 14
libavcodec/x86/cavsdsp.c View File

@@ -438,21 +438,22 @@ static void ff_ ## OPNAME ## cavs_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, ui
#endif /* (HAVE_MMXEXT_INLINE || HAVE_AMD3DNOW_INLINE) */

#if HAVE_MMXEXT_INLINE
QPEL_CAVS(put_, PUT_OP, mmx2)
QPEL_CAVS(avg_,AVG_MMXEXT_OP, mmx2)
QPEL_CAVS(put_, PUT_OP, mmxext)
QPEL_CAVS(avg_, AVG_MMXEXT_OP, mmxext)

CAVS_MC(put_, 8, mmx2)
CAVS_MC(put_, 16,mmx2)
CAVS_MC(avg_, 8, mmx2)
CAVS_MC(avg_, 16,mmx2)
CAVS_MC(put_, 8, mmxext)
CAVS_MC(put_, 16, mmxext)
CAVS_MC(avg_, 8, mmxext)
CAVS_MC(avg_, 16, mmxext)

static void ff_cavsdsp_init_mmx2(CAVSDSPContext* c, AVCodecContext *avctx) {
static void ff_cavsdsp_init_mmxext(CAVSDSPContext *c, AVCodecContext *avctx)
{
#define dspfunc(PFX, IDX, NUM) \
c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_mmx2; \
c->PFX ## _pixels_tab[IDX][ 2] = ff_ ## PFX ## NUM ## _mc20_mmx2; \
c->PFX ## _pixels_tab[IDX][ 4] = ff_ ## PFX ## NUM ## _mc01_mmx2; \
c->PFX ## _pixels_tab[IDX][ 8] = ff_ ## PFX ## NUM ## _mc02_mmx2; \
c->PFX ## _pixels_tab[IDX][12] = ff_ ## PFX ## NUM ## _mc03_mmx2; \
c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_mmxext; \
c->PFX ## _pixels_tab[IDX][ 2] = ff_ ## PFX ## NUM ## _mc20_mmxext; \
c->PFX ## _pixels_tab[IDX][ 4] = ff_ ## PFX ## NUM ## _mc01_mmxext; \
c->PFX ## _pixels_tab[IDX][ 8] = ff_ ## PFX ## NUM ## _mc02_mmxext; \
c->PFX ## _pixels_tab[IDX][12] = ff_ ## PFX ## NUM ## _mc03_mmxext; \

dspfunc(put_cavs_qpel, 0, 16);
dspfunc(put_cavs_qpel, 1, 8);
@@ -475,7 +476,7 @@ CAVS_MC(avg_, 16,3dnow)

static void ff_cavsdsp_init_3dnow(CAVSDSPContext* c, AVCodecContext *avctx) {
#define dspfunc(PFX, IDX, NUM) \
c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_mmx2; \
c->PFX ## _pixels_tab[IDX][ 0] = ff_ ## PFX ## NUM ## _mc00_mmxext; \
c->PFX ## _pixels_tab[IDX][ 2] = ff_ ## PFX ## NUM ## _mc20_3dnow; \
c->PFX ## _pixels_tab[IDX][ 4] = ff_ ## PFX ## NUM ## _mc01_3dnow; \
c->PFX ## _pixels_tab[IDX][ 8] = ff_ ## PFX ## NUM ## _mc02_3dnow; \
@@ -496,7 +497,7 @@ av_cold void ff_cavsdsp_init_x86(CAVSDSPContext *c, AVCodecContext *avctx)
int mm_flags = av_get_cpu_flags();

#if HAVE_MMXEXT_INLINE
if (mm_flags & AV_CPU_FLAG_MMXEXT) ff_cavsdsp_init_mmx2(c, avctx);
if (mm_flags & AV_CPU_FLAG_MMXEXT) ff_cavsdsp_init_mmxext(c, avctx);
#endif /* HAVE_MMXEXT_INLINE */
#if HAVE_AMD3DNOW_INLINE
if (mm_flags & AV_CPU_FLAG_3DNOW) ff_cavsdsp_init_3dnow(c, avctx);


+ 1
- 1
libavcodec/x86/diracdsp_mmx.c View File

@@ -78,7 +78,7 @@ void ff_diracdsp_init_mmx(DiracDSPContext* c)

#if HAVE_MMXEXT_INLINE
if (mm_flags & AV_CPU_FLAG_MMX2) {
PIXFUNC(avg, 0, mmx2);
PIXFUNC(avg, 0, mmxext);
}
#endif



+ 1
- 1
libavcodec/x86/diracdsp_mmx.h View File

@@ -27,7 +27,7 @@ void ff_diracdsp_init_mmx(DiracDSPContext* c);

DECL_DIRAC_PIXOP(put, mmx);
DECL_DIRAC_PIXOP(avg, mmx);
DECL_DIRAC_PIXOP(avg, mmx2);
DECL_DIRAC_PIXOP(avg, mmxext);

void ff_put_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h);
void ff_avg_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h);


+ 73
- 73
libavcodec/x86/dsputil_mmx.c View File

@@ -208,7 +208,7 @@ DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 };
/***********************************/
/* MMXEXT specific */

#define DEF(x) x ## _mmx2
#define DEF(x) x ## _mmxext

/* Introduced only in MMXEXT set */
#define PAVGB "pavgb"
@@ -222,11 +222,11 @@ DECLARE_ALIGNED(16, const double, ff_pd_2)[2] = { 2.0, 2.0 };

#define put_no_rnd_pixels16_mmx put_pixels16_mmx
#define put_no_rnd_pixels8_mmx put_pixels8_mmx
#define put_pixels16_mmx2 put_pixels16_mmx
#define put_pixels8_mmx2 put_pixels8_mmx
#define put_pixels4_mmx2 put_pixels4_mmx
#define put_no_rnd_pixels16_mmx2 put_no_rnd_pixels16_mmx
#define put_no_rnd_pixels8_mmx2 put_no_rnd_pixels8_mmx
#define put_pixels16_mmxext put_pixels16_mmx
#define put_pixels8_mmxext put_pixels8_mmx
#define put_pixels4_mmxext put_pixels4_mmx
#define put_no_rnd_pixels16_mmxext put_no_rnd_pixels16_mmx
#define put_no_rnd_pixels8_mmxext put_no_rnd_pixels8_mmx
#define put_pixels16_3dnow put_pixels16_mmx
#define put_pixels8_3dnow put_pixels8_mmx
#define put_pixels4_3dnow put_pixels4_mmx
@@ -944,11 +944,11 @@ static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height,
OP(%%mm5, out, %%mm7, d)

#define QPEL_BASE(OPNAME, ROUNDER, RND, OP_MMXEXT, OP_3DNOW) \
static void OPNAME ## mpeg4_qpel16_h_lowpass_mmx2(uint8_t *dst, \
uint8_t *src, \
int dstStride, \
int srcStride, \
int h) \
static void OPNAME ## mpeg4_qpel16_h_lowpass_mmxext(uint8_t *dst, \
uint8_t *src, \
int dstStride, \
int srcStride, \
int h) \
{ \
uint64_t temp; \
\
@@ -1138,11 +1138,11 @@ static void OPNAME ## mpeg4_qpel16_h_lowpass_3dnow(uint8_t *dst, \
} \
} \
\
static void OPNAME ## mpeg4_qpel8_h_lowpass_mmx2(uint8_t *dst, \
uint8_t *src, \
int dstStride, \
int srcStride, \
int h) \
static void OPNAME ## mpeg4_qpel8_h_lowpass_mmxext(uint8_t *dst, \
uint8_t *src, \
int dstStride, \
int srcStride, \
int h) \
{ \
__asm__ volatile ( \
"pxor %%mm7, %%mm7 \n\t" \
@@ -1775,9 +1775,9 @@ QPEL_BASE(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, PUT_OP)
QPEL_OP(put_, ff_pw_16, _, PUT_OP, 3dnow)
QPEL_OP(avg_, ff_pw_16, _, AVG_3DNOW_OP, 3dnow)
QPEL_OP(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, 3dnow)
QPEL_OP(put_, ff_pw_16, _, PUT_OP, mmx2)
QPEL_OP(avg_, ff_pw_16, _, AVG_MMXEXT_OP, mmx2)
QPEL_OP(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, mmx2)
QPEL_OP(put_, ff_pw_16, _, PUT_OP, mmxext)
QPEL_OP(avg_, ff_pw_16, _, AVG_MMXEXT_OP, mmxext)
QPEL_OP(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, mmxext)

/***********************************/
/* bilinear qpel: not compliant to any spec, only for -lavdopts fast */
@@ -1831,10 +1831,10 @@ QPEL_2TAP_L3(OPNAME, SIZE, MMX, 31, 1, stride, -1) \
QPEL_2TAP_L3(OPNAME, SIZE, MMX, 13, stride, -stride, 1) \
QPEL_2TAP_L3(OPNAME, SIZE, MMX, 33, stride + 1, -stride, -1) \

QPEL_2TAP(put_, 16, mmx2)
QPEL_2TAP(avg_, 16, mmx2)
QPEL_2TAP(put_, 8, mmx2)
QPEL_2TAP(avg_, 8, mmx2)
QPEL_2TAP(put_, 16, mmxext)
QPEL_2TAP(avg_, 16, mmxext)
QPEL_2TAP(put_, 8, mmxext)
QPEL_2TAP(avg_, 8, mmxext)
QPEL_2TAP(put_, 16, 3dnow)
QPEL_2TAP(avg_, 16, 3dnow)
QPEL_2TAP(put_, 8, 3dnow)
@@ -2099,7 +2099,7 @@ static void name(void *mem, int stride, int h) \
} while (--h); \
}

PREFETCH(prefetch_mmx2, prefetcht0)
PREFETCH(prefetch_mmxext, prefetcht0)
PREFETCH(prefetch_3dnow, prefetch)
#undef PREFETCH

@@ -2153,22 +2153,22 @@ CHROMA_MC(avg, 8, 10, avx)
#if HAVE_INLINE_ASM

/* CAVS-specific */
void ff_put_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride)
void ff_put_cavs_qpel8_mc00_mmxext(uint8_t *dst, uint8_t *src, int stride)
{
put_pixels8_mmx(dst, src, stride, 8);
}

void ff_avg_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride)
void ff_avg_cavs_qpel8_mc00_mmxext(uint8_t *dst, uint8_t *src, int stride)
{
avg_pixels8_mmx(dst, src, stride, 8);
}

void ff_put_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride)
void ff_put_cavs_qpel16_mc00_mmxext(uint8_t *dst, uint8_t *src, int stride)
{
put_pixels16_mmx(dst, src, stride, 16);
}

void ff_avg_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride)
void ff_avg_cavs_qpel16_mc00_mmxext(uint8_t *dst, uint8_t *src, int stride)
{
avg_pixels16_mmx(dst, src, stride, 16);
}
@@ -2180,10 +2180,10 @@ void ff_put_vc1_mspel_mc00_mmx(uint8_t *dst, const uint8_t *src,
put_pixels8_mmx(dst, src, stride, 8);
}

void ff_avg_vc1_mspel_mc00_mmx2(uint8_t *dst, const uint8_t *src,
int stride, int rnd)
void ff_avg_vc1_mspel_mc00_mmxext(uint8_t *dst, const uint8_t *src,
int stride, int rnd)
{
avg_pixels8_mmx2(dst, src, stride, 8);
avg_pixels8_mmxext(dst, src, stride, 8);
}

/* only used in VP3/5/6 */
@@ -2242,7 +2242,7 @@ void ff_ ## OPNAME ## _dirac_pixels32_ ## EXT(uint8_t *dst, const uint8_t *src[5

DIRAC_PIXOP(put, mmx)
DIRAC_PIXOP(avg, mmx)
DIRAC_PIXOP(avg, mmx2)
DIRAC_PIXOP(avg, mmxext)

void ff_put_dirac_pixels16_sse2(uint8_t *dst, const uint8_t *src[5], int stride, int h)
{
@@ -2620,68 +2620,68 @@ static void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx, int mm_flags)

}

static void dsputil_init_mmx2(DSPContext *c, AVCodecContext *avctx,
int mm_flags)
static void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
int mm_flags)
{
const int bit_depth = avctx->bits_per_raw_sample;
const int high_bit_depth = bit_depth > 8;

#if HAVE_INLINE_ASM
c->prefetch = prefetch_mmx2;
c->prefetch = prefetch_mmxext;

if (!high_bit_depth) {
c->put_pixels_tab[0][1] = put_pixels16_x2_mmx2;
c->put_pixels_tab[0][2] = put_pixels16_y2_mmx2;
c->put_pixels_tab[0][1] = put_pixels16_x2_mmxext;
c->put_pixels_tab[0][2] = put_pixels16_y2_mmxext;

c->avg_pixels_tab[0][0] = avg_pixels16_mmx2;
c->avg_pixels_tab[0][1] = avg_pixels16_x2_mmx2;
c->avg_pixels_tab[0][2] = avg_pixels16_y2_mmx2;
c->avg_pixels_tab[0][0] = avg_pixels16_mmxext;
c->avg_pixels_tab[0][1] = avg_pixels16_x2_mmxext;
c->avg_pixels_tab[0][2] = avg_pixels16_y2_mmxext;

c->put_pixels_tab[1][1] = put_pixels8_x2_mmx2;
c->put_pixels_tab[1][2] = put_pixels8_y2_mmx2;
c->put_pixels_tab[1][1] = put_pixels8_x2_mmxext;
c->put_pixels_tab[1][2] = put_pixels8_y2_mmxext;

c->avg_pixels_tab[1][0] = avg_pixels8_mmx2;
c->avg_pixels_tab[1][1] = avg_pixels8_x2_mmx2;
c->avg_pixels_tab[1][2] = avg_pixels8_y2_mmx2;
c->avg_pixels_tab[1][0] = avg_pixels8_mmxext;
c->avg_pixels_tab[1][1] = avg_pixels8_x2_mmxext;
c->avg_pixels_tab[1][2] = avg_pixels8_y2_mmxext;
}

if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
if (!high_bit_depth) {
c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmx2;
c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmx2;
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_mmx2;
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_mmx2;
c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmxext;
c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmxext;
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_mmxext;
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_mmxext;

c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmx2;
c->avg_pixels_tab[1][3] = avg_pixels8_xy2_mmx2;
c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmxext;
c->avg_pixels_tab[1][3] = avg_pixels8_xy2_mmxext;
}
}

if (CONFIG_VP3_DECODER && (avctx->codec_id == AV_CODEC_ID_VP3 ||
avctx->codec_id == AV_CODEC_ID_THEORA)) {
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_exact_mmx2;
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_exact_mmx2;
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_exact_mmxext;
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_exact_mmxext;
}
#endif /* HAVE_INLINE_ASM */

if (CONFIG_H264QPEL) {
#if HAVE_INLINE_ASM
SET_QPEL_FUNCS(put_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(put_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(avg_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(avg_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(put_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(put_qpel, 1, 8, mmxext, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmxext, );
SET_QPEL_FUNCS(avg_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(avg_qpel, 1, 8, mmxext, );
#endif /* HAVE_INLINE_ASM */

if (!high_bit_depth) {
#if HAVE_INLINE_ASM
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(put_h264_qpel, 2, 4, mmx2, );
SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmx2, );
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmxext, );
SET_QPEL_FUNCS(put_h264_qpel, 2, 4, mmxext, );
SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmxext, );
SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmxext, );
#endif /* HAVE_INLINE_ASM */
} else if (bit_depth == 10) {
#if HAVE_YASM
@@ -2697,10 +2697,10 @@ static void dsputil_init_mmx2(DSPContext *c, AVCodecContext *avctx,
}

#if HAVE_INLINE_ASM
SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, mmxext, );
SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, mmxext, );
#endif /* HAVE_INLINE_ASM */
}

@@ -3041,9 +3041,9 @@ void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
c->idct = ff_idct_xvid_sse2;
c->idct_permutation_type = FF_SSE2_IDCT_PERM;
} else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->idct_put = ff_idct_xvid_mmx2_put;
c->idct_add = ff_idct_xvid_mmx2_add;
c->idct = ff_idct_xvid_mmx2;
c->idct_put = ff_idct_xvid_mmxext_put;
c->idct_add = ff_idct_xvid_mmxext_add;
c->idct = ff_idct_xvid_mmxext;
} else {
c->idct_put = ff_idct_xvid_mmx_put;
c->idct_add = ff_idct_xvid_mmx_add;
@@ -3057,7 +3057,7 @@ void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
}

if (mm_flags & AV_CPU_FLAG_MMXEXT)
dsputil_init_mmx2(c, avctx, mm_flags);
dsputil_init_mmxext(c, avctx, mm_flags);

if (mm_flags & AV_CPU_FLAG_3DNOW)
dsputil_init_3dnow(c, avctx, mm_flags);


+ 5
- 5
libavcodec/x86/dsputil_mmx.h View File

@@ -89,13 +89,13 @@ void ff_add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_s
void ff_put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
void ff_put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);

void ff_put_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
void ff_avg_cavs_qpel8_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
void ff_put_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
void ff_avg_cavs_qpel16_mc00_mmx2(uint8_t *dst, uint8_t *src, int stride);
void ff_put_cavs_qpel8_mc00_mmxext(uint8_t *dst, uint8_t *src, int stride);
void ff_avg_cavs_qpel8_mc00_mmxext(uint8_t *dst, uint8_t *src, int stride);
void ff_put_cavs_qpel16_mc00_mmxext(uint8_t *dst, uint8_t *src, int stride);
void ff_avg_cavs_qpel16_mc00_mmxext(uint8_t *dst, uint8_t *src, int stride);

void ff_put_vc1_mspel_mc00_mmx(uint8_t *dst, const uint8_t *src, int stride, int rnd);
void ff_avg_vc1_mspel_mc00_mmx2(uint8_t *dst, const uint8_t *src, int stride, int rnd);
void ff_avg_vc1_mspel_mc00_mmxext(uint8_t *dst, const uint8_t *src, int stride, int rnd);

void ff_put_rv40_qpel8_mc33_mmx(uint8_t *block, uint8_t *pixels, int line_size);
void ff_put_rv40_qpel16_mc33_mmx(uint8_t *block, uint8_t *pixels, int line_size);


+ 16
- 9
libavcodec/x86/dsputilenc_mmx.c View File

@@ -647,7 +647,9 @@ static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_si
}
#undef SUM

static int vsad_intra16_mmx2(void *v, uint8_t * pix, uint8_t * dummy, int line_size, int h) {
static int vsad_intra16_mmxext(void *v, uint8_t *pix, uint8_t *dummy,
int line_size, int h)
{
int tmp;

assert( (((int)pix) & 7) == 0);
@@ -765,7 +767,9 @@ static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, in
}
#undef SUM

static int vsad16_mmx2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h) {
static int vsad16_mmxext(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
int tmp;

assert( (((int)pix1) & 7) == 0);
@@ -845,7 +849,10 @@ static void diff_bytes_mmx(uint8_t *dst, const uint8_t *src1, const uint8_t *src
dst[i+0] = src1[i+0]-src2[i+0];
}

static void sub_hfyu_median_prediction_mmx2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top){
static void sub_hfyu_median_prediction_mmxext(uint8_t *dst, const uint8_t *src1,
const uint8_t *src2, int w,
int *left, int *left_top)
{
x86_reg i=0;
uint8_t l, lt;

@@ -977,7 +984,7 @@ DCT_SAD_FUNC(mmx)

#define HSUM(a,t,dst) HSUM_MMXEXT(a,t,dst)
#define MMABS(a,z) MMABS_MMXEXT(a,z)
DCT_SAD_FUNC(mmx2)
DCT_SAD_FUNC(mmxext)
#undef HSUM
#undef DCT_SAD

@@ -1116,7 +1123,7 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
if(mm_flags & AV_CPU_FLAG_SSE2){
c->fdct = ff_fdct_sse2;
} else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->fdct = ff_fdct_mmx2;
c->fdct = ff_fdct_mmxext;
}else{
c->fdct = ff_fdct_mmx;
}
@@ -1149,14 +1156,14 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
c->ssd_int8_vs_int16 = ssd_int8_vs_int16_mmx;

if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->sum_abs_dctelem= sum_abs_dctelem_mmx2;
c->vsad[4]= vsad_intra16_mmx2;
c->sum_abs_dctelem = sum_abs_dctelem_mmxext;
c->vsad[4] = vsad_intra16_mmxext;

if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->vsad[0] = vsad16_mmx2;
c->vsad[0] = vsad16_mmxext;
}

c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_mmx2;
c->sub_hfyu_median_prediction = sub_hfyu_median_prediction_mmxext;
}

if(mm_flags & AV_CPU_FLAG_SSE2){


+ 4
- 3
libavcodec/x86/fdct.c View File

@@ -440,7 +440,8 @@ static av_always_inline void fdct_row_sse2(const int16_t *in, int16_t *out)
);
}

static av_always_inline void fdct_row_mmx2(const int16_t *in, int16_t *out, const int16_t *table)
static av_always_inline void fdct_row_mmxext(const int16_t *in, int16_t *out,
const int16_t *table)
{
__asm__ volatile (
"pshufw $0x1B, 8(%0), %%mm5 \n\t"
@@ -555,7 +556,7 @@ void ff_fdct_mmx(int16_t *block)
}
}

void ff_fdct_mmx2(int16_t *block)
void ff_fdct_mmxext(int16_t *block)
{
DECLARE_ALIGNED(8, int64_t, align_tmp)[16];
int16_t *block1= (int16_t*)align_tmp;
@@ -566,7 +567,7 @@ void ff_fdct_mmx2(int16_t *block)
fdct_col_mmx(block, block1, 4);

for(i=8;i>0;i--) {
fdct_row_mmx2(block1, block, table);
fdct_row_mmxext(block1, block, table);
block1 += 8;
table += 32;
block += 8;


+ 1
- 2
libavcodec/x86/h264_chromamc_10bit.asm View File

@@ -245,7 +245,7 @@ cglobal %1_h264_chroma_mc2_10, 6,7
%if %0==3
movq %2, %3
%endif
PAVG %1, %2
pavgw %1, %2
%endmacro

%define CHROMAMC_AVG NOTHING
@@ -260,7 +260,6 @@ CHROMA_MC4 put
CHROMA_MC2 put

%define CHROMAMC_AVG AVG
%define PAVG pavgw
INIT_XMM sse2
CHROMA_MC8 avg
%if HAVE_AVX_EXTERNAL


+ 27
- 27
libavcodec/x86/h264_qpel.c View File

@@ -1002,36 +1002,36 @@ static void OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp,
OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 16);\
}\

#define put_pixels8_l2_sse2 put_pixels8_l2_mmx2
#define avg_pixels8_l2_sse2 avg_pixels8_l2_mmx2
#define put_pixels16_l2_sse2 put_pixels16_l2_mmx2
#define avg_pixels16_l2_sse2 avg_pixels16_l2_mmx2
#define put_pixels8_l2_ssse3 put_pixels8_l2_mmx2
#define avg_pixels8_l2_ssse3 avg_pixels8_l2_mmx2
#define put_pixels16_l2_ssse3 put_pixels16_l2_mmx2
#define avg_pixels16_l2_ssse3 avg_pixels16_l2_mmx2
#define put_pixels8_l2_sse2 put_pixels8_l2_mmxext
#define avg_pixels8_l2_sse2 avg_pixels8_l2_mmxext
#define put_pixels16_l2_sse2 put_pixels16_l2_mmxext
#define avg_pixels16_l2_sse2 avg_pixels16_l2_mmxext
#define put_pixels8_l2_ssse3 put_pixels8_l2_mmxext
#define avg_pixels8_l2_ssse3 avg_pixels8_l2_mmxext
#define put_pixels16_l2_ssse3 put_pixels16_l2_mmxext
#define avg_pixels16_l2_ssse3 avg_pixels16_l2_mmxext

#define put_pixels8_l2_shift5_sse2 put_pixels8_l2_shift5_mmx2
#define avg_pixels8_l2_shift5_sse2 avg_pixels8_l2_shift5_mmx2
#define put_pixels16_l2_shift5_sse2 put_pixels16_l2_shift5_mmx2
#define avg_pixels16_l2_shift5_sse2 avg_pixels16_l2_shift5_mmx2
#define put_pixels8_l2_shift5_ssse3 put_pixels8_l2_shift5_mmx2
#define avg_pixels8_l2_shift5_ssse3 avg_pixels8_l2_shift5_mmx2
#define put_pixels16_l2_shift5_ssse3 put_pixels16_l2_shift5_mmx2
#define avg_pixels16_l2_shift5_ssse3 avg_pixels16_l2_shift5_mmx2
#define put_pixels8_l2_shift5_sse2 put_pixels8_l2_shift5_mmxext
#define avg_pixels8_l2_shift5_sse2 avg_pixels8_l2_shift5_mmxext
#define put_pixels16_l2_shift5_sse2 put_pixels16_l2_shift5_mmxext
#define avg_pixels16_l2_shift5_sse2 avg_pixels16_l2_shift5_mmxext
#define put_pixels8_l2_shift5_ssse3 put_pixels8_l2_shift5_mmxext
#define avg_pixels8_l2_shift5_ssse3 avg_pixels8_l2_shift5_mmxext
#define put_pixels16_l2_shift5_ssse3 put_pixels16_l2_shift5_mmxext
#define avg_pixels16_l2_shift5_ssse3 avg_pixels16_l2_shift5_mmxext

#define put_h264_qpel8_h_lowpass_l2_sse2 put_h264_qpel8_h_lowpass_l2_mmx2
#define avg_h264_qpel8_h_lowpass_l2_sse2 avg_h264_qpel8_h_lowpass_l2_mmx2
#define put_h264_qpel16_h_lowpass_l2_sse2 put_h264_qpel16_h_lowpass_l2_mmx2
#define avg_h264_qpel16_h_lowpass_l2_sse2 avg_h264_qpel16_h_lowpass_l2_mmx2
#define put_h264_qpel8_h_lowpass_l2_sse2 put_h264_qpel8_h_lowpass_l2_mmxext
#define avg_h264_qpel8_h_lowpass_l2_sse2 avg_h264_qpel8_h_lowpass_l2_mmxext
#define put_h264_qpel16_h_lowpass_l2_sse2 put_h264_qpel16_h_lowpass_l2_mmxext
#define avg_h264_qpel16_h_lowpass_l2_sse2 avg_h264_qpel16_h_lowpass_l2_mmxext

#define put_h264_qpel8_v_lowpass_ssse3 put_h264_qpel8_v_lowpass_sse2
#define avg_h264_qpel8_v_lowpass_ssse3 avg_h264_qpel8_v_lowpass_sse2
#define put_h264_qpel16_v_lowpass_ssse3 put_h264_qpel16_v_lowpass_sse2
#define avg_h264_qpel16_v_lowpass_ssse3 avg_h264_qpel16_v_lowpass_sse2

#define put_h264_qpel8or16_hv2_lowpass_sse2 put_h264_qpel8or16_hv2_lowpass_mmx2
#define avg_h264_qpel8or16_hv2_lowpass_sse2 avg_h264_qpel8or16_hv2_lowpass_mmx2
#define put_h264_qpel8or16_hv2_lowpass_sse2 put_h264_qpel8or16_hv2_lowpass_mmxext
#define avg_h264_qpel8or16_hv2_lowpass_sse2 avg_h264_qpel8or16_hv2_lowpass_mmxext

#define H264_MC(OPNAME, SIZE, MMX, ALIGN) \
H264_MC_C(OPNAME, SIZE, MMX, ALIGN)\
@@ -1045,8 +1045,8 @@ static void put_h264_qpel16_mc00_sse2 (uint8_t *dst, uint8_t *src, int stride){
static void avg_h264_qpel16_mc00_sse2 (uint8_t *dst, uint8_t *src, int stride){
avg_pixels16_sse2(dst, src, stride, 16);
}
#define put_h264_qpel8_mc00_sse2 put_h264_qpel8_mc00_mmx2
#define avg_h264_qpel8_mc00_sse2 avg_h264_qpel8_mc00_mmx2
#define put_h264_qpel8_mc00_sse2 put_h264_qpel8_mc00_mmxext
#define avg_h264_qpel8_mc00_sse2 avg_h264_qpel8_mc00_mmxext

#define H264_MC_C(OPNAME, SIZE, MMX, ALIGN) \
static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## MMX (uint8_t *dst, uint8_t *src, int stride){\
@@ -1168,8 +1168,8 @@ QPEL_H264(put_, PUT_OP, 3dnow)
QPEL_H264(avg_, AVG_3DNOW_OP, 3dnow)
#undef PAVGB
#define PAVGB "pavgb"
QPEL_H264(put_, PUT_OP, mmx2)
QPEL_H264(avg_,AVG_MMXEXT_OP, mmx2)
QPEL_H264(put_, PUT_OP, mmxext)
QPEL_H264(avg_, AVG_MMXEXT_OP, mmxext)
QPEL_H264_V_XMM(put_, PUT_OP, sse2)
QPEL_H264_V_XMM(avg_,AVG_MMXEXT_OP, sse2)
QPEL_H264_HV_XMM(put_, PUT_OP, sse2)
@@ -1185,7 +1185,7 @@ QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, ssse3)
#undef PAVGB

H264_MC_4816(3dnow)
H264_MC_4816(mmx2)
H264_MC_4816(mmxext)
H264_MC_816(H264_MC_V, sse2)
H264_MC_816(H264_MC_HV, sse2)
#if HAVE_SSSE3_INLINE


+ 6
- 7
libavcodec/x86/h264dsp_init.c View File

@@ -130,18 +130,17 @@ LF_FUNCS(uint16_t, 10)

#if ARCH_X86_32 && HAVE_YASM
LF_FUNC(v8, luma, 8, mmx2)
static void ff_deblock_v_luma_8_mmx2(uint8_t *pix, int stride, int alpha,
int beta, int8_t *tc0)
static void ff_deblock_v_luma_8_mmxext(uint8_t *pix, int stride, int alpha,
int beta, int8_t *tc0)
{
if ((tc0[0] & tc0[1]) >= 0)
ff_deblock_v8_luma_8_mmx2(pix + 0, stride, alpha, beta, tc0);
if ((tc0[2] & tc0[3]) >= 0)
ff_deblock_v8_luma_8_mmx2(pix + 8, stride, alpha, beta, tc0 + 2);
}

LF_IFUNC(v8, luma_intra, 8, mmx2)
static void ff_deblock_v_luma_intra_8_mmx2(uint8_t *pix, int stride,
int alpha, int beta)
static void ff_deblock_v_luma_intra_8_mmxext(uint8_t *pix, int stride,
int alpha, int beta)
{
ff_deblock_v8_luma_intra_8_mmx2(pix + 0, stride, alpha, beta);
ff_deblock_v8_luma_intra_8_mmx2(pix + 8, stride, alpha, beta);
@@ -247,9 +246,9 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
c->h264_h_loop_filter_chroma_intra = ff_deblock_h_chroma_intra_8_mmx2;
}
#if ARCH_X86_32
c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_mmx2;
c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_mmxext;
c->h264_h_loop_filter_luma = ff_deblock_h_luma_8_mmx2;
c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_mmx2;
c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_mmxext;
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_mmx2;
#endif /* ARCH_X86_32 */
c->weight_h264_pixels_tab[0] = ff_h264_weight_16_mmx2;


+ 6
- 5
libavcodec/x86/idct_mmx_xvid.c View File

@@ -512,7 +512,8 @@ __asm__ volatile(
//-----------------------------------------------------------------------------


void ff_idct_xvid_mmx2(short *block){
void ff_idct_xvid_mmxext(short *block)
{
__asm__ volatile(
//# Process each row
DCT_8_INV_ROW_XMM(0*16(%0), 0*16(%0), 64*0(%2), 8*0(%1))
@@ -542,15 +543,15 @@ void ff_idct_xvid_mmx_add(uint8_t *dest, int line_size, DCTELEM *block)
ff_add_pixels_clamped_mmx(block, dest, line_size);
}

void ff_idct_xvid_mmx2_put(uint8_t *dest, int line_size, DCTELEM *block)
void ff_idct_xvid_mmxext_put(uint8_t *dest, int line_size, DCTELEM *block)
{
ff_idct_xvid_mmx2(block);
ff_idct_xvid_mmxext(block);
ff_put_pixels_clamped_mmx(block, dest, line_size);
}

void ff_idct_xvid_mmx2_add(uint8_t *dest, int line_size, DCTELEM *block)
void ff_idct_xvid_mmxext_add(uint8_t *dest, int line_size, DCTELEM *block)
{
ff_idct_xvid_mmx2(block);
ff_idct_xvid_mmxext(block);
ff_add_pixels_clamped_mmx(block, dest, line_size);
}



+ 3
- 3
libavcodec/x86/idct_xvid.h View File

@@ -34,9 +34,9 @@ void ff_idct_xvid_mmx(short *block);
void ff_idct_xvid_mmx_put(uint8_t *dest, int line_size, DCTELEM *block);
void ff_idct_xvid_mmx_add(uint8_t *dest, int line_size, DCTELEM *block);

void ff_idct_xvid_mmx2(short *block);
void ff_idct_xvid_mmx2_put(uint8_t *dest, int line_size, DCTELEM *block);
void ff_idct_xvid_mmx2_add(uint8_t *dest, int line_size, DCTELEM *block);
void ff_idct_xvid_mmxext(short *block);
void ff_idct_xvid_mmxext_put(uint8_t *dest, int line_size, DCTELEM *block);
void ff_idct_xvid_mmxext_add(uint8_t *dest, int line_size, DCTELEM *block);

void ff_idct_xvid_sse2(short *block);
void ff_idct_xvid_sse2_put(uint8_t *dest, int line_size, short *block);


+ 20
- 16
libavcodec/x86/motion_est.c View File

@@ -74,7 +74,8 @@ static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
);
}

static inline void sad8_1_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
static inline void sad8_1_mmxext(uint8_t *blk1, uint8_t *blk2,
int stride, int h)
{
__asm__ volatile(
".p2align 4 \n\t"
@@ -120,7 +121,8 @@ static int sad16_sse2(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)
return ret;
}

static inline void sad8_x2a_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
static inline void sad8_x2a_mmxext(uint8_t *blk1, uint8_t *blk2,
int stride, int h)
{
__asm__ volatile(
".p2align 4 \n\t"
@@ -142,7 +144,8 @@ static inline void sad8_x2a_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h
);
}

static inline void sad8_y2a_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
static inline void sad8_y2a_mmxext(uint8_t *blk1, uint8_t *blk2,
int stride, int h)
{
__asm__ volatile(
"movq (%1), %%mm0 \n\t"
@@ -167,7 +170,8 @@ static inline void sad8_y2a_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h
);
}

static inline void sad8_4_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
static inline void sad8_4_mmxext(uint8_t *blk1, uint8_t *blk2,
int stride, int h)
{
__asm__ volatile(
"movq "MANGLE(bone)", %%mm5 \n\t"
@@ -304,7 +308,7 @@ static inline int sum_mmx(void)
return ret&0xFFFF;
}

static inline int sum_mmx2(void)
static inline int sum_mmxext(void)
{
int ret;
__asm__ volatile(
@@ -424,7 +428,7 @@ static int sad16_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride,
}\

PIX_SAD(mmx)
PIX_SAD(mmx2)
PIX_SAD(mmxext)

#endif /* HAVE_INLINE_ASM */

@@ -447,19 +451,19 @@ void ff_dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx)
c->sad[1]= sad8_mmx;
}
if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->pix_abs[0][0] = sad16_mmx2;
c->pix_abs[1][0] = sad8_mmx2;
c->pix_abs[0][0] = sad16_mmxext;
c->pix_abs[1][0] = sad8_mmxext;

c->sad[0]= sad16_mmx2;
c->sad[1]= sad8_mmx2;
c->sad[0] = sad16_mmxext;
c->sad[1] = sad8_mmxext;

if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->pix_abs[0][1] = sad16_x2_mmx2;
c->pix_abs[0][2] = sad16_y2_mmx2;
c->pix_abs[0][3] = sad16_xy2_mmx2;
c->pix_abs[1][1] = sad8_x2_mmx2;
c->pix_abs[1][2] = sad8_y2_mmx2;
c->pix_abs[1][3] = sad8_xy2_mmx2;
c->pix_abs[0][1] = sad16_x2_mmxext;
c->pix_abs[0][2] = sad16_y2_mmxext;
c->pix_abs[0][3] = sad16_xy2_mmxext;
c->pix_abs[1][1] = sad8_x2_mmxext;
c->pix_abs[1][2] = sad8_y2_mmxext;
c->pix_abs[1][3] = sad8_xy2_mmxext;
}
}
if ((mm_flags & AV_CPU_FLAG_SSE2) && !(mm_flags & AV_CPU_FLAG_3DNOW) && avctx->codec_id != AV_CODEC_ID_SNOW) {


+ 3
- 3
libavcodec/x86/mpegvideoenc.c View File

@@ -47,8 +47,8 @@ extern uint16_t ff_inv_zigzag_direct16[64];
#define COMPILE_TEMPLATE_SSSE3 0
#undef RENAME
#undef RENAMEl
#define RENAME(a) a ## _MMX2
#define RENAMEl(a) a ## _mmx2
#define RENAME(a) a ## _MMXEXT
#define RENAMEl(a) a ## _mmxext
#include "mpegvideoenc_template.c"
#endif /* HAVE_MMXEXT_INLINE */

@@ -92,7 +92,7 @@ void ff_dct_encode_init_x86(MpegEncContext *s)
#endif
#if HAVE_MMXEXT_INLINE
if (INLINE_MMXEXT(mm_flags))
s->dct_quantize = dct_quantize_MMX2;
s->dct_quantize = dct_quantize_MMXEXT;
#endif
#if HAVE_SSE2_INLINE
if (INLINE_SSE2(mm_flags))


+ 36
- 29
libavcodec/x86/vc1dsp_mmx.c View File

@@ -466,7 +466,10 @@ VC1_MSPEL_MC(avg_)
static void put_vc1_mspel_mc ## a ## b ## _mmx(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
put_vc1_mspel_mc(dst, src, stride, a, b, rnd); \
}\
static void avg_vc1_mspel_mc ## a ## b ## _mmx2(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
static void avg_vc1_mspel_mc ## a ## b ## _mmxext(uint8_t *dst, \
const uint8_t *src, \
int stride, int rnd) \
{ \
avg_vc1_mspel_mc(dst, src, stride, a, b, rnd); \
}

@@ -489,7 +492,8 @@ DECLARE_FUNCTION(3, 1)
DECLARE_FUNCTION(3, 2)
DECLARE_FUNCTION(3, 3)

static void vc1_inv_trans_4x4_dc_mmx2(uint8_t *dest, int linesize, DCTELEM *block)
static void vc1_inv_trans_4x4_dc_mmxext(uint8_t *dest, int linesize,
DCTELEM *block)
{
int dc = block[0];
dc = (17 * dc + 4) >> 3;
@@ -527,7 +531,8 @@ static void vc1_inv_trans_4x4_dc_mmx2(uint8_t *dest, int linesize, DCTELEM *bloc
);
}

static void vc1_inv_trans_4x8_dc_mmx2(uint8_t *dest, int linesize, DCTELEM *block)
static void vc1_inv_trans_4x8_dc_mmxext(uint8_t *dest, int linesize,
DCTELEM *block)
{
int dc = block[0];
dc = (17 * dc + 4) >> 3;
@@ -588,7 +593,8 @@ static void vc1_inv_trans_4x8_dc_mmx2(uint8_t *dest, int linesize, DCTELEM *bloc
);
}

static void vc1_inv_trans_8x4_dc_mmx2(uint8_t *dest, int linesize, DCTELEM *block)
static void vc1_inv_trans_8x4_dc_mmxext(uint8_t *dest, int linesize,
DCTELEM *block)
{
int dc = block[0];
dc = ( 3 * dc + 1) >> 1;
@@ -626,7 +632,8 @@ static void vc1_inv_trans_8x4_dc_mmx2(uint8_t *dest, int linesize, DCTELEM *bloc
);
}

static void vc1_inv_trans_8x8_dc_mmx2(uint8_t *dest, int linesize, DCTELEM *block)
static void vc1_inv_trans_8x8_dc_mmxext(uint8_t *dest, int linesize,
DCTELEM *block)
{
int dc = block[0];
dc = (3 * dc + 1) >> 1;
@@ -712,29 +719,29 @@ av_cold void ff_vc1dsp_init_mmx(VC1DSPContext *dsp)

av_cold void ff_vc1dsp_init_mmxext(VC1DSPContext *dsp)
{
dsp->avg_vc1_mspel_pixels_tab[ 0] = ff_avg_vc1_mspel_mc00_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_mmx2;
dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 1] = avg_vc1_mspel_mc10_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 5] = avg_vc1_mspel_mc11_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 9] = avg_vc1_mspel_mc12_mmx2;
dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 2] = avg_vc1_mspel_mc20_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 6] = avg_vc1_mspel_mc21_mmx2;
dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_mmx2;
dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 3] = avg_vc1_mspel_mc30_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 7] = avg_vc1_mspel_mc31_mmx2;
dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_mmx2;
dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_mmx2;
dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_mmx2;
dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_mmx2;
dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_mmx2;
dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 0] = ff_avg_vc1_mspel_mc00_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_mmxext;
dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 1] = avg_vc1_mspel_mc10_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 5] = avg_vc1_mspel_mc11_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 9] = avg_vc1_mspel_mc12_mmxext;
dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 2] = avg_vc1_mspel_mc20_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 6] = avg_vc1_mspel_mc21_mmxext;
dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_mmxext;
dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 3] = avg_vc1_mspel_mc30_mmxext;
dsp->avg_vc1_mspel_pixels_tab[ 7] = avg_vc1_mspel_mc31_mmxext;
dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_mmxext;
dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_mmxext;
dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_mmxext;
dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_mmxext;
dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_mmxext;
dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_mmxext;
}
#endif /* HAVE_INLINE_ASM */

+ 4
- 2
libavfilter/x86/gradfun.c View File

@@ -30,7 +30,9 @@ DECLARE_ALIGNED(16, static const uint16_t, pw_7f)[8] = {0x7F,0x7F,0x7F,0x7F,0x7F
DECLARE_ALIGNED(16, static const uint16_t, pw_ff)[8] = {0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF};

#if HAVE_MMXEXT_INLINE
static void gradfun_filter_line_mmx2(uint8_t *dst, const uint8_t *src, const uint16_t *dc, int width, int thresh, const uint16_t *dithers)
static void gradfun_filter_line_mmxext(uint8_t *dst, const uint8_t *src, const uint16_t *dc,
int width, int thresh,
const uint16_t *dithers)
{
intptr_t x;
if (width & 3) {
@@ -175,7 +177,7 @@ av_cold void ff_gradfun_init_x86(GradFunContext *gf)

#if HAVE_MMXEXT_INLINE
if (cpu_flags & AV_CPU_FLAG_MMXEXT)
gf->filter_line = gradfun_filter_line_mmx2;
gf->filter_line = gradfun_filter_line_mmxext;
#endif
#if HAVE_SSSE3_INLINE
if (cpu_flags & AV_CPU_FLAG_SSSE3)


+ 2
- 2
libavfilter/x86/yadif.c View File

@@ -48,7 +48,7 @@ DECLARE_ASM_CONST(16, const xmm_reg, pw_1) = {0x0001000100010001ULL, 0x000100010

#if HAVE_MMXEXT_INLINE
#undef RENAME
#define RENAME(a) a ## _mmx2
#define RENAME(a) a ## _mmxext
#include "yadif_template.c"
#endif

@@ -60,7 +60,7 @@ av_cold void ff_yadif_init_x86(YADIFContext *yadif)

#if HAVE_MMXEXT_INLINE
if (cpu_flags & AV_CPU_FLAG_MMXEXT)
yadif->filter_line = yadif_filter_line_mmx2;
yadif->filter_line = yadif_filter_line_mmxext;
#endif
#if HAVE_SSE2_INLINE
if (cpu_flags & AV_CPU_FLAG_SSE2)


+ 9
- 3
libswscale/input.c View File

@@ -935,6 +935,7 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
case AV_PIX_FMT_YUV420P14LE:
case AV_PIX_FMT_YUV420P16LE:
case AV_PIX_FMT_YUV422P16LE:
case AV_PIX_FMT_YUV444P16LE:

case AV_PIX_FMT_YUVA444P9LE:
case AV_PIX_FMT_YUVA422P9LE:
@@ -945,7 +946,6 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
case AV_PIX_FMT_YUVA420P16LE:
case AV_PIX_FMT_YUVA422P16LE:
case AV_PIX_FMT_YUVA444P16LE:
case AV_PIX_FMT_YUV444P16LE:
c->chrToYV12 = bswap16UV_c;
break;
#else
@@ -1186,6 +1186,9 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
case AV_PIX_FMT_YUV422P16LE:
case AV_PIX_FMT_YUV444P16LE:

case AV_PIX_FMT_GRAY16LE:
c->lumToYV12 = bswap16Y_c;
break;
case AV_PIX_FMT_YUVA444P9LE:
case AV_PIX_FMT_YUVA422P9LE:
case AV_PIX_FMT_YUVA420P9LE:
@@ -1195,8 +1198,8 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
case AV_PIX_FMT_YUVA420P16LE:
case AV_PIX_FMT_YUVA422P16LE:
case AV_PIX_FMT_YUVA444P16LE:
case AV_PIX_FMT_GRAY16LE:
c->lumToYV12 = bswap16Y_c;
c->alpToYV12 = bswap16Y_c;
break;
#else
case AV_PIX_FMT_YUV444P9BE:
@@ -1215,6 +1218,9 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
case AV_PIX_FMT_YUV422P16BE:
case AV_PIX_FMT_YUV444P16BE:

case AV_PIX_FMT_GRAY16BE:
c->lumToYV12 = bswap16Y_c;
break;
case AV_PIX_FMT_YUVA444P9BE:
case AV_PIX_FMT_YUVA422P9BE:
case AV_PIX_FMT_YUVA420P9BE:
@@ -1224,8 +1230,8 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
case AV_PIX_FMT_YUVA420P16BE:
case AV_PIX_FMT_YUVA422P16BE:
case AV_PIX_FMT_YUVA444P16BE:
case AV_PIX_FMT_GRAY16BE:
c->lumToYV12 = bswap16Y_c;
c->alpToYV12 = bswap16Y_c;
break;
#endif
case AV_PIX_FMT_YUYV422:


+ 11
- 10
libswscale/utils.c View File

@@ -627,8 +627,9 @@ fail:
}

#if HAVE_MMXEXT_INLINE
static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode,
int16_t *filter, int32_t *filterPos, int numSplits)
static int init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode,
int16_t *filter, int32_t *filterPos,
int numSplits)
{
uint8_t *fragmentA;
x86_reg imm8OfPShufW1A;
@@ -1107,10 +1108,10 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
#if HAVE_MMXEXT_INLINE
// can't downscale !!!
if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) {
c->lumMmxextFilterCodeSize = initMMX2HScaler(dstW, c->lumXInc, NULL,
NULL, NULL, 8);
c->chrMmxextFilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc,
NULL, NULL, NULL, 4);
c->lumMmxextFilterCodeSize = init_hscaler_mmxext(dstW, c->lumXInc, NULL,
NULL, NULL, 8);
c->chrMmxextFilterCodeSize = init_hscaler_mmxext(c->chrDstW, c->chrXInc,
NULL, NULL, NULL, 4);

#if USE_MMAP
c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize,
@@ -1150,10 +1151,10 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW / 2 / 8 + 8) * sizeof(int32_t), fail);
FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail);

initMMX2HScaler( dstW, c->lumXInc, c->lumMmxextFilterCode,
c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
init_hscaler_mmxext( dstW, c->lumXInc, c->lumMmxextFilterCode,
c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);

#if USE_MMAP
mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ);


+ 2
- 2
libswscale/x86/rgb2rgb.c View File

@@ -102,7 +102,7 @@ DECLARE_ASM_CONST(8, uint64_t, mul16_mid) = 0x2080208020802080ULL;
#undef RENAME
#undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMXEXT 1
#define RENAME(a) a ## _MMX2
#define RENAME(a) a ## _MMXEXT
#include "rgb2rgb_template.c"

//SSE2 versions
@@ -142,7 +142,7 @@ av_cold void rgb2rgb_init_x86(void)
if (INLINE_AMD3DNOW(cpu_flags))
rgb2rgb_init_3DNOW();
if (INLINE_MMXEXT(cpu_flags))
rgb2rgb_init_MMX2();
rgb2rgb_init_MMXEXT();
if (INLINE_SSE2(cpu_flags))
rgb2rgb_init_SSE2();
#endif /* HAVE_INLINE_ASM */


+ 3
- 3
libswscale/x86/swscale.c View File

@@ -85,7 +85,7 @@ DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
#undef RENAME
#undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMXEXT 1
#define RENAME(a) a ## _MMX2
#define RENAME(a) a ## _MMXEXT
#include "swscale_template.c"
#endif

@@ -211,7 +211,7 @@ static void yuv2yuvX_sse3(const int16_t *filter, int filterSize,
const uint8_t *dither, int offset)
{
if(((int)dest) & 15){
return yuv2yuvX_MMX2(filter, filterSize, src, dest, dstW, dither, offset);
return yuv2yuvX_MMXEXT(filter, filterSize, src, dest, dstW, dither, offset);
}
if (offset) {
__asm__ volatile("movq (%0), %%xmm3\n\t"
@@ -381,7 +381,7 @@ av_cold void ff_sws_init_swScale_mmx(SwsContext *c)
sws_init_swScale_MMX(c);
#if HAVE_MMXEXT_INLINE
if (cpu_flags & AV_CPU_FLAG_MMXEXT)
sws_init_swScale_MMX2(c);
sws_init_swScale_MMXEXT(c);
if (cpu_flags & AV_CPU_FLAG_SSE3){
if(c->use_mmx_vfilter && !(c->flags & SWS_ACCURATE_RND))
c->yuv2planeX = yuv2yuvX_sse3;


+ 5
- 3
libswscale/x86/yuv2rgb.c View File

@@ -63,7 +63,7 @@ DECLARE_ASM_CONST(8, uint64_t, pb_07) = 0x0707070707070707ULL;
#undef RENAME
#undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMXEXT 1
#define RENAME(a) a ## _MMX2
#define RENAME(a) a ## _MMXEXT
#include "yuv2rgb_template.c"
#endif /* HAVE_MMXEXT_INLINE */

@@ -77,8 +77,10 @@ av_cold SwsFunc ff_yuv2rgb_init_mmx(SwsContext *c)
#if HAVE_MMXEXT_INLINE
if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
switch (c->dstFormat) {
case AV_PIX_FMT_RGB24: return yuv420_rgb24_MMX2;
case AV_PIX_FMT_BGR24: return yuv420_bgr24_MMX2;
case AV_PIX_FMT_RGB24:
return yuv420_rgb24_MMXEXT;
case AV_PIX_FMT_BGR24:
return yuv420_bgr24_MMXEXT;
}
}
#endif


Write Preview
Loading…
Cancel
Save