virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-23 12:43:46 +02:00
Code Issues Releases Activity

x86: build: replace mmx2 by mmxext

Browse Source 
Refactoring mmx2/mmxext YASM code with cpuflags will force renames.
So switching to a consistent naming scheme beforehand is sensible.
The name "mmxext" is more official and widespread and also the name
of the CPU flag, as reported e.g. by the Linux kernel.
This commit is contained in:
Diego Biurrun 2012-07-08 18:42:12 +02:00
parent 66adb7ce1b
commit 239fdf1b4a
38 changed files with 120 additions and 108 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Doxyfile
View File

@ -1378,7 +1378,7 @@ PREDEFINED = "__attribute__(x)=" \
"DEF(x)=x ## _TMPL" \ "DEF(x)=x ## _TMPL" \
HAVE_AV_CONFIG_H \ HAVE_AV_CONFIG_H \
HAVE_MMX \ HAVE_MMX \
HAVE_MMX2 \ HAVE_MMXEXT \
HAVE_AMD3DNOW \ HAVE_AMD3DNOW \
"DECLARE_ALIGNED(a,t,n)=t n" \ "DECLARE_ALIGNED(a,t,n)=t n" \
"offsetof(x,y)=0x42" "offsetof(x,y)=0x42"

12
configure vendored
View File

@ -243,7 +243,7 @@ Optimization options (experts only):
--disable-amd3dnow disable 3DNow! optimizations --disable-amd3dnow disable 3DNow! optimizations
--disable-amd3dnowext disable 3DNow! extended optimizations --disable-amd3dnowext disable 3DNow! extended optimizations
--disable-mmx disable MMX optimizations --disable-mmx disable MMX optimizations
--disable-mmx2 disable MMX2 optimizations --disable-mmxext disable MMXEXT optimizations
--disable-sse disable SSE optimizations --disable-sse disable SSE optimizations
--disable-ssse3 disable SSSE3 optimizations --disable-ssse3 disable SSSE3 optimizations
--disable-avx disable AVX optimizations --disable-avx disable AVX optimizations
@ -1054,7 +1054,7 @@ ARCH_EXT_LIST='
fma4 fma4
mmi mmi
mmx mmx
mmx2 mmxext
neon neon
ppc4xx ppc4xx
sse sse
@ -1302,7 +1302,7 @@ x86_64_suggest="cmov fast_cmov"
amd3dnow_deps="mmx" amd3dnow_deps="mmx"
amd3dnowext_deps="amd3dnow" amd3dnowext_deps="amd3dnow"
mmx_deps="x86" mmx_deps="x86"
mmx2_deps="mmx" mmxext_deps="mmx"
sse_deps="mmx" sse_deps="mmx"
ssse3_deps="sse" ssse3_deps="sse"
avx_deps="ssse3" avx_deps="ssse3"
@ -2861,9 +2861,9 @@ EOF
# check whether xmm clobbers are supported # check whether xmm clobbers are supported
check_asm xmm_clobbers '"":::"%xmm0"' check_asm xmm_clobbers '"":::"%xmm0"'
# check whether binutils is new enough to compile SSSE3/MMX2 # check whether binutils is new enough to compile SSSE3/MMXEXT
enabled ssse3 && check_asm ssse3 '"pabsw %xmm0, %xmm0"' enabled ssse3 && check_asm ssse3 '"pabsw %xmm0, %xmm0"'
enabled mmx2 && check_asm mmx2 '"pmaxub %mm0, %mm1"' enabled mmxext && check_asm mmxext '"pmaxub %mm0, %mm1"'
if ! disabled_any asm mmx yasm; then if ! disabled_any asm mmx yasm; then
if check_cmd $yasmexe --version; then if check_cmd $yasmexe --version; then
@ -3307,7 +3307,7 @@ echo "runtime cpu detection ${runtime_cpudetect-no}"
if enabled x86; then if enabled x86; then
echo "${yasmexe} ${yasm-no}" echo "${yasmexe} ${yasm-no}"
echo "MMX enabled ${mmx-no}" echo "MMX enabled ${mmx-no}"
echo "MMX2 enabled ${mmx2-no}" echo "MMXEXT enabled ${mmxext-no}"
echo "3DNow! enabled ${amd3dnow-no}" echo "3DNow! enabled ${amd3dnow-no}"
echo "3DNow! extended enabled ${amd3dnowext-no}" echo "3DNow! extended enabled ${amd3dnowext-no}"
echo "SSE enabled ${sse-no}" echo "SSE enabled ${sse-no}"

5
doc/APIchanges
View File

@ -13,6 +13,11 @@ libavutil: 2011-04-18
API changes, most recent first: API changes, most recent first:
2012-08-03 - xxxxxxx - lavu 51.37.1 - cpu.h
lsws 2.1.1 - swscale.h
Rename AV_CPU_FLAG_MMX2 ---> AV_CPU_FLAG_MMXEXT.
Rename SWS_CPU_CAPS_MMX2 ---> SWS_CPU_CAPS_MMXEXT.
2012-07-xx - xxxxxxx - lavf 54.13.0 - avformat.h 2012-07-xx - xxxxxxx - lavf 54.13.0 - avformat.h
Add AVFMT_FLAG_NOBUFFER for low latency use cases. Add AVFMT_FLAG_NOBUFFER for low latency use cases.

4
libavcodec/dct-test.c
View File

@ -87,7 +87,7 @@ static const struct algo fdct_tab[] = {
#if HAVE_MMX && HAVE_INLINE_ASM #if HAVE_MMX && HAVE_INLINE_ASM
{ "MMX", ff_fdct_mmx, NO_PERM, AV_CPU_FLAG_MMX }, { "MMX", ff_fdct_mmx, NO_PERM, AV_CPU_FLAG_MMX },
{ "MMX2", ff_fdct_mmx2, NO_PERM, AV_CPU_FLAG_MMX2 }, { "MMXEXT", ff_fdct_mmx2, NO_PERM, AV_CPU_FLAG_MMXEXT },
{ "SSE2", ff_fdct_sse2, NO_PERM, AV_CPU_FLAG_SSE2 }, { "SSE2", ff_fdct_sse2, NO_PERM, AV_CPU_FLAG_SSE2 },
#endif #endif
@ -111,7 +111,7 @@ static const struct algo idct_tab[] = {
#if HAVE_MMX && HAVE_INLINE_ASM #if HAVE_MMX && HAVE_INLINE_ASM
{ "SIMPLE-MMX", ff_simple_idct_mmx, MMX_SIMPLE_PERM, AV_CPU_FLAG_MMX }, { "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-MMX", ff_idct_xvid_mmx, NO_PERM, AV_CPU_FLAG_MMX, 1 },
{ "XVID-MMX2", ff_idct_xvid_mmx2, NO_PERM, AV_CPU_FLAG_MMX2, 1 }, { "XVID-MMXEXT", ff_idct_xvid_mmx2, NO_PERM, AV_CPU_FLAG_MMXEXT, 1 },
{ "XVID-SSE2", ff_idct_xvid_sse2, SSE2_PERM, AV_CPU_FLAG_SSE2, 1 }, { "XVID-SSE2", ff_idct_xvid_sse2, SSE2_PERM, AV_CPU_FLAG_SSE2, 1 },
#endif #endif

4
libavcodec/motion-test.c
View File

@ -116,8 +116,8 @@ int main(int argc, char **argv)
AVCodecContext *ctx; AVCodecContext *ctx;
int c; int c;
DSPContext cctx, mmxctx; DSPContext cctx, mmxctx;
int flags[2] = { AV_CPU_FLAG_MMX, AV_CPU_FLAG_MMX2 }; int flags[2] = { AV_CPU_FLAG_MMX, AV_CPU_FLAG_MMXEXT };
int flags_size = HAVE_MMX2 ? 2 : 1; int flags_size = HAVE_MMXEXT ? 2 : 1;
if (argc > 1) { if (argc > 1) {
help(); help();

2
libavcodec/x86/ac3dsp.asm
View File

@ -68,7 +68,7 @@ cglobal ac3_exponent_min_%1, 3,4,2, exp, reuse_blks, expn, offset
%define LOOP_ALIGN %define LOOP_ALIGN
INIT_MMX INIT_MMX
AC3_EXPONENT_MIN mmx AC3_EXPONENT_MIN mmx
%if HAVE_MMX2 %if HAVE_MMXEXT
%define PMINUB PMINUB_MMXEXT %define PMINUB PMINUB_MMXEXT
%define LOOP_ALIGN ALIGN 16 %define LOOP_ALIGN ALIGN 16
AC3_EXPONENT_MIN mmxext AC3_EXPONENT_MIN mmxext

2
libavcodec/x86/ac3dsp_mmx.c
View File

@ -65,7 +65,7 @@ av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact)
c->float_to_fixed24 = ff_float_to_fixed24_3dnow; c->float_to_fixed24 = ff_float_to_fixed24_3dnow;
} }
} }
if (mm_flags & AV_CPU_FLAG_MMX2 && HAVE_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT && HAVE_MMXEXT) {
c->ac3_exponent_min = ff_ac3_exponent_min_mmxext; c->ac3_exponent_min = ff_ac3_exponent_min_mmxext;
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx2; c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx2;
} }

2
libavcodec/x86/cavsdsp_mmx.c
View File

@ -486,7 +486,7 @@ void ff_cavsdsp_init_mmx(CAVSDSPContext *c, AVCodecContext *avctx)
int mm_flags = av_get_cpu_flags(); int mm_flags = av_get_cpu_flags();
#if HAVE_INLINE_ASM #if HAVE_INLINE_ASM
if (mm_flags & AV_CPU_FLAG_MMX2) ff_cavsdsp_init_mmx2 (c, avctx); if (mm_flags & AV_CPU_FLAG_MMXEXT) ff_cavsdsp_init_mmx2(c, avctx);
if (mm_flags & AV_CPU_FLAG_3DNOW) ff_cavsdsp_init_3dnow(c, avctx); if (mm_flags & AV_CPU_FLAG_3DNOW) ff_cavsdsp_init_3dnow(c, avctx);
#endif /* HAVE_INLINE_ASM */ #endif /* HAVE_INLINE_ASM */
} }

4
libavcodec/x86/dsputil_mmx.c
View File

@ -3028,7 +3028,7 @@ void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
c->idct_add = ff_idct_xvid_sse2_add; c->idct_add = ff_idct_xvid_sse2_add;
c->idct = ff_idct_xvid_sse2; c->idct = ff_idct_xvid_sse2;
c->idct_permutation_type = FF_SSE2_IDCT_PERM; c->idct_permutation_type = FF_SSE2_IDCT_PERM;
} else if (mm_flags & AV_CPU_FLAG_MMX2) { } else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->idct_put = ff_idct_xvid_mmx2_put; c->idct_put = ff_idct_xvid_mmx2_put;
c->idct_add = ff_idct_xvid_mmx2_add; c->idct_add = ff_idct_xvid_mmx2_add;
c->idct = ff_idct_xvid_mmx2; c->idct = ff_idct_xvid_mmx2;
@ -3044,7 +3044,7 @@ void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
dsputil_init_mmx(c, avctx, mm_flags); dsputil_init_mmx(c, avctx, mm_flags);
} }
if (mm_flags & AV_CPU_FLAG_MMX2) if (mm_flags & AV_CPU_FLAG_MMXEXT)
dsputil_init_mmx2(c, avctx, mm_flags); dsputil_init_mmx2(c, avctx, mm_flags);
if (mm_flags & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW) if (mm_flags & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW)

7
libavcodec/x86/dsputilenc_mmx.c
View File

@ -1111,7 +1111,7 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
(dct_algo==FF_DCT_AUTO || dct_algo==FF_DCT_MMX)) { (dct_algo==FF_DCT_AUTO || dct_algo==FF_DCT_MMX)) {
if(mm_flags & AV_CPU_FLAG_SSE2){ if(mm_flags & AV_CPU_FLAG_SSE2){
c->fdct = ff_fdct_sse2; c->fdct = ff_fdct_sse2;
}else if(mm_flags & AV_CPU_FLAG_MMX2){ } else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->fdct = ff_fdct_mmx2; c->fdct = ff_fdct_mmx2;
}else{ }else{
c->fdct = ff_fdct_mmx; c->fdct = ff_fdct_mmx;
@ -1144,8 +1144,7 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
c->ssd_int8_vs_int16 = ssd_int8_vs_int16_mmx; c->ssd_int8_vs_int16 = ssd_int8_vs_int16_mmx;
if (mm_flags & AV_CPU_FLAG_MMXEXT) {
if (mm_flags & AV_CPU_FLAG_MMX2) {
c->sum_abs_dctelem= sum_abs_dctelem_mmx2; c->sum_abs_dctelem= sum_abs_dctelem_mmx2;
c->vsad[4]= vsad_intra16_mmx2; c->vsad[4]= vsad_intra16_mmx2;
@ -1186,7 +1185,7 @@ void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx)
c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx; c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx;
c->hadamard8_diff[1] = ff_hadamard8_diff_mmx; c->hadamard8_diff[1] = ff_hadamard8_diff_mmx;
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx2; c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx2;
c->hadamard8_diff[1] = ff_hadamard8_diff_mmx2; c->hadamard8_diff[1] = ff_hadamard8_diff_mmx2;
} }

4
libavcodec/x86/h264_intrapred_init.c
View File

@ -198,7 +198,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
} }
} }
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_mmx2; h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_mmx2;
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_mmx2; h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_mmx2;
if (chroma_format_idc == 1) if (chroma_format_idc == 1)
@ -308,7 +308,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id, const int bit_depth
} }
} }
} else if (bit_depth == 10) { } else if (bit_depth == 10) {
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
h->pred4x4[DC_PRED ] = ff_pred4x4_dc_10_mmxext; h->pred4x4[DC_PRED ] = ff_pred4x4_dc_10_mmxext;
h->pred4x4[HOR_UP_PRED ] = ff_pred4x4_horizontal_up_10_mmxext; h->pred4x4[HOR_UP_PRED ] = ff_pred4x4_horizontal_up_10_mmxext;

6
libavcodec/x86/h264dsp_mmx.c
View File

@ -218,7 +218,7 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
#if HAVE_YASM #if HAVE_YASM
int mm_flags = av_get_cpu_flags(); int mm_flags = av_get_cpu_flags();
if (chroma_format_idc == 1 && mm_flags & AV_CPU_FLAG_MMX2) if (chroma_format_idc == 1 && mm_flags & AV_CPU_FLAG_MMXEXT)
c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmx2; c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmx2;
if (bit_depth == 8) { if (bit_depth == 8) {
@ -236,7 +236,7 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
if (mm_flags & AV_CPU_FLAG_CMOV) if (mm_flags & AV_CPU_FLAG_CMOV)
c->h264_luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_mmx; c->h264_luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_mmx;
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->h264_idct_dc_add = ff_h264_idct_dc_add_8_mmx2; c->h264_idct_dc_add = ff_h264_idct_dc_add_8_mmx2;
c->h264_idct8_dc_add = ff_h264_idct8_dc_add_8_mmx2; c->h264_idct8_dc_add = ff_h264_idct8_dc_add_8_mmx2;
c->h264_idct_add16 = ff_h264_idct_add16_8_mmx2; c->h264_idct_add16 = ff_h264_idct_add16_8_mmx2;
@ -304,7 +304,7 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
} }
} else if (bit_depth == 10) { } else if (bit_depth == 10) {
if (mm_flags & AV_CPU_FLAG_MMX) { if (mm_flags & AV_CPU_FLAG_MMX) {
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
#if ARCH_X86_32 #if ARCH_X86_32
c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_10_mmx2; c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_10_mmx2;
c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_mmx2; c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_mmx2;

2
libavcodec/x86/motion_est_mmx.c
View File

@ -444,7 +444,7 @@ void ff_dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx)
c->sad[0]= sad16_mmx; c->sad[0]= sad16_mmx;
c->sad[1]= sad8_mmx; c->sad[1]= sad8_mmx;
} }
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->pix_abs[0][0] = sad16_mmx2; c->pix_abs[0][0] = sad16_mmx2;
c->pix_abs[1][0] = sad8_mmx2; c->pix_abs[1][0] = sad8_mmx2;

10
libavcodec/x86/mpegvideo_mmx.c
View File

@ -595,15 +595,15 @@ static void denoise_dct_sse2(MpegEncContext *s, DCTELEM *block){
#define HAVE_SSSE3 0 #define HAVE_SSSE3 0
#undef HAVE_SSE2 #undef HAVE_SSE2
#undef HAVE_MMX2 #undef HAVE_MMXEXT
#define HAVE_SSE2 0 #define HAVE_SSE2 0
#define HAVE_MMX2 0 #define HAVE_MMXEXT 0
#define RENAME(a) a ## _MMX #define RENAME(a) a ## _MMX
#define RENAMEl(a) a ## _mmx #define RENAMEl(a) a ## _mmx
#include "mpegvideo_mmx_template.c" #include "mpegvideo_mmx_template.c"
#undef HAVE_MMX2 #undef HAVE_MMXEXT
#define HAVE_MMX2 1 #define HAVE_MMXEXT 1
#undef RENAME #undef RENAME
#undef RENAMEl #undef RENAMEl
#define RENAME(a) a ## _MMX2 #define RENAME(a) a ## _MMX2
@ -660,7 +660,7 @@ void ff_MPV_common_init_mmx(MpegEncContext *s)
#endif #endif
if(mm_flags & AV_CPU_FLAG_SSE2){ if(mm_flags & AV_CPU_FLAG_SSE2){
s->dct_quantize= dct_quantize_SSE2; s->dct_quantize= dct_quantize_SSE2;
} else if(mm_flags & AV_CPU_FLAG_MMX2){ } else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
s->dct_quantize= dct_quantize_MMX2; s->dct_quantize= dct_quantize_MMX2;
} else { } else {
s->dct_quantize= dct_quantize_MMX; s->dct_quantize= dct_quantize_MMX;

2
libavcodec/x86/mpegvideo_mmx_template.c
View File

@ -48,7 +48,7 @@
#define MMREG_WIDTH "8" #define MMREG_WIDTH "8"
#define MM "%%mm" #define MM "%%mm"
#define MOVQ "movq" #define MOVQ "movq"
#if HAVE_MMX2 #if HAVE_MMXEXT
#define SPREADW(a) "pshufw $0, "a", "a" \n\t" #define SPREADW(a) "pshufw $0, "a", "a" \n\t"
#define PMAXW(a,b) "pmaxsw "a", "b" \n\t" #define PMAXW(a,b) "pmaxsw "a", "b" \n\t"
#define PMAX(a,b) \ #define PMAX(a,b) \

2
libavcodec/x86/pngdsp-init.c
View File

@ -41,7 +41,7 @@ void ff_pngdsp_init_x86(PNGDSPContext *dsp)
if (flags & AV_CPU_FLAG_MMX) if (flags & AV_CPU_FLAG_MMX)
dsp->add_bytes_l2 = ff_add_bytes_l2_mmx; dsp->add_bytes_l2 = ff_add_bytes_l2_mmx;
#endif #endif
if (flags & AV_CPU_FLAG_MMX2) if (flags & AV_CPU_FLAG_MMXEXT)
dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmx2; dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmx2;
if (flags & AV_CPU_FLAG_SSE2) if (flags & AV_CPU_FLAG_SSE2)
dsp->add_bytes_l2 = ff_add_bytes_l2_sse2; dsp->add_bytes_l2 = ff_add_bytes_l2_sse2;

2
libavcodec/x86/rv34dsp_init.c
View File

@ -37,7 +37,7 @@ av_cold void ff_rv34dsp_init_x86(RV34DSPContext* c, DSPContext *dsp)
if (mm_flags & AV_CPU_FLAG_MMX) if (mm_flags & AV_CPU_FLAG_MMX)
c->rv34_idct_dc_add = ff_rv34_idct_dc_add_mmx; c->rv34_idct_dc_add = ff_rv34_idct_dc_add_mmx;
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->rv34_inv_transform_dc = ff_rv34_idct_dc_noround_mmx2; c->rv34_inv_transform_dc = ff_rv34_idct_dc_noround_mmx2;
c->rv34_idct_add = ff_rv34_idct_add_mmx2; c->rv34_idct_add = ff_rv34_idct_add_mmx2;
} }

2
libavcodec/x86/rv40dsp_init.c
View File

@ -200,7 +200,7 @@ void ff_rv40dsp_init_x86(RV34DSPContext *c, DSPContext *dsp)
QPEL_MC_SET(put_, _mmx) QPEL_MC_SET(put_, _mmx)
#endif #endif
} }
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_mmx2; c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_mmx2;
c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_mmx2; c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_mmx2;
c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_mmx2; c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_mmx2;

2
libavcodec/x86/snowdsp_mmx.c
View File

@ -889,7 +889,7 @@ void ff_dwt_init_x86(DWTContext *c)
c->inner_add_yblock = ff_snow_inner_add_yblock_sse2; c->inner_add_yblock = ff_snow_inner_add_yblock_sse2;
} }
else{ else{
if(mm_flags & AV_CPU_FLAG_MMX2){ if (mm_flags & AV_CPU_FLAG_MMXEXT) {
c->horizontal_compose97i = ff_snow_horizontal_compose97i_mmx; c->horizontal_compose97i = ff_snow_horizontal_compose97i_mmx;
#if HAVE_7REGS #if HAVE_7REGS
c->vertical_compose97i = ff_snow_vertical_compose97i_mmx; c->vertical_compose97i = ff_snow_vertical_compose97i_mmx;

4
libavcodec/x86/vc1dsp_mmx.c
View File

@ -757,7 +757,7 @@ void ff_vc1dsp_init_mmx(VC1DSPContext *dsp)
dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_mmx; dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_mmx;
} }
if (mm_flags & AV_CPU_FLAG_MMX2){ if (mm_flags & AV_CPU_FLAG_MMXEXT) {
dsp->avg_vc1_mspel_pixels_tab[ 0] = ff_avg_vc1_mspel_mc00_mmx2; 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[ 4] = avg_vc1_mspel_mc01_mmx2;
dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_mmx2; dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_mmx2;
@ -798,7 +798,7 @@ void ff_vc1dsp_init_mmx(VC1DSPContext *dsp)
dsp->put_no_rnd_vc1_chroma_pixels_tab[0]= ff_put_vc1_chroma_mc8_mmx_nornd; dsp->put_no_rnd_vc1_chroma_pixels_tab[0]= ff_put_vc1_chroma_mc8_mmx_nornd;
} }
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
ASSIGN_LF(mmx2); ASSIGN_LF(mmx2);
dsp->avg_no_rnd_vc1_chroma_pixels_tab[0]= ff_avg_vc1_chroma_mc8_mmx2_nornd; dsp->avg_no_rnd_vc1_chroma_pixels_tab[0]= ff_avg_vc1_chroma_mc8_mmx2_nornd;
} else if (mm_flags & AV_CPU_FLAG_3DNOW) { } else if (mm_flags & AV_CPU_FLAG_3DNOW) {

2
libavcodec/x86/vp3dsp_init.c
View File

@ -49,7 +49,7 @@ av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags)
} }
#endif #endif
if (HAVE_MMX2 && cpuflags & AV_CPU_FLAG_MMX2) { if (HAVE_MMXEXT && cpuflags & AV_CPU_FLAG_MMXEXT) {
c->idct_dc_add = ff_vp3_idct_dc_add_mmx2; c->idct_dc_add = ff_vp3_idct_dc_add_mmx2;
if (!(flags & CODEC_FLAG_BITEXACT)) { if (!(flags & CODEC_FLAG_BITEXACT)) {

2
libavcodec/x86/vp8dsp-init.c
View File

@ -350,7 +350,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
/* note that 4-tap width=16 functions are missing because w=16 /* note that 4-tap width=16 functions are missing because w=16
* is only used for luma, and luma is always a copy or sixtap. */ * is only used for luma, and luma is always a copy or sixtap. */
if (mm_flags & AV_CPU_FLAG_MMX2) { if (mm_flags & AV_CPU_FLAG_MMXEXT) {
VP8_MC_FUNC(2, 4, mmx2); VP8_MC_FUNC(2, 4, mmx2);
VP8_BILINEAR_MC_FUNC(2, 4, mmx2); VP8_BILINEAR_MC_FUNC(2, 4, mmx2);
#if ARCH_X86_32 #if ARCH_X86_32

6
libavfilter/x86/gradfun.c
View File

@ -28,7 +28,7 @@
DECLARE_ALIGNED(16, static const uint16_t, pw_7f)[8] = {0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F}; DECLARE_ALIGNED(16, static const uint16_t, pw_7f)[8] = {0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F,0x7F};
DECLARE_ALIGNED(16, static const uint16_t, pw_ff)[8] = {0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF}; DECLARE_ALIGNED(16, static const uint16_t, pw_ff)[8] = {0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF};
#if HAVE_MMX2 #if HAVE_MMXEXT
static void gradfun_filter_line_mmx2(uint8_t *dst, uint8_t *src, uint16_t *dc, int width, int thresh, const uint16_t *dithers) static void gradfun_filter_line_mmx2(uint8_t *dst, uint8_t *src, uint16_t *dc, int width, int thresh, const uint16_t *dithers)
{ {
intptr_t x; intptr_t x;
@ -173,8 +173,8 @@ av_cold void ff_gradfun_init_x86(GradFunContext *gf)
int cpu_flags = av_get_cpu_flags(); int cpu_flags = av_get_cpu_flags();
#if HAVE_INLINE_ASM #if HAVE_INLINE_ASM
#if HAVE_MMX2 #if HAVE_MMXEXT
if (cpu_flags & AV_CPU_FLAG_MMX2) if (cpu_flags & AV_CPU_FLAG_MMXEXT)
gf->filter_line = gradfun_filter_line_mmx2; gf->filter_line = gradfun_filter_line_mmx2;
#endif #endif
#if HAVE_SSSE3 #if HAVE_SSSE3

6
libavfilter/x86/yadif.c
View File

@ -45,7 +45,7 @@ DECLARE_ASM_CONST(16, const xmm_reg, pw_1) = {0x0001000100010001ULL, 0x000100010
#undef COMPILE_TEMPLATE_SSE #undef COMPILE_TEMPLATE_SSE
#endif #endif
#if HAVE_MMX2 #if HAVE_MMXEXT
#undef RENAME #undef RENAME
#define RENAME(a) a ## _mmx2 #define RENAME(a) a ## _mmx2
#include "yadif_template.c" #include "yadif_template.c"
@ -58,8 +58,8 @@ av_cold void ff_yadif_init_x86(YADIFContext *yadif)
int cpu_flags = av_get_cpu_flags(); int cpu_flags = av_get_cpu_flags();
#if HAVE_INLINE_ASM #if HAVE_INLINE_ASM
#if HAVE_MMX2 #if HAVE_MMXEXT
if (cpu_flags & AV_CPU_FLAG_MMX2) if (cpu_flags & AV_CPU_FLAG_MMXEXT)
yadif->filter_line = yadif_filter_line_mmx2; yadif->filter_line = yadif_filter_line_mmx2;
#endif #endif
#if HAVE_SSE #if HAVE_SSE

8
libavutil/cpu.c
View File

@ -47,10 +47,10 @@ void av_set_cpu_flags_mask(int mask)
int av_parse_cpu_flags(const char *s) int av_parse_cpu_flags(const char *s)
{ {
#define CPUFLAG_MMX2 (AV_CPU_FLAG_MMX | AV_CPU_FLAG_MMX2 | AV_CPU_FLAG_CMOV) #define CPUFLAG_MMXEXT (AV_CPU_FLAG_MMX | AV_CPU_FLAG_MMXEXT | AV_CPU_FLAG_CMOV)
#define CPUFLAG_3DNOW (AV_CPU_FLAG_3DNOW | AV_CPU_FLAG_MMX) #define CPUFLAG_3DNOW (AV_CPU_FLAG_3DNOW | AV_CPU_FLAG_MMX)
#define CPUFLAG_3DNOWEXT (AV_CPU_FLAG_3DNOWEXT | CPUFLAG_3DNOW) #define CPUFLAG_3DNOWEXT (AV_CPU_FLAG_3DNOWEXT | CPUFLAG_3DNOW)
#define CPUFLAG_SSE (AV_CPU_FLAG_SSE | CPUFLAG_MMX2) #define CPUFLAG_SSE (AV_CPU_FLAG_SSE | CPUFLAG_MMXEXT)
#define CPUFLAG_SSE2 (AV_CPU_FLAG_SSE2 | CPUFLAG_SSE) #define CPUFLAG_SSE2 (AV_CPU_FLAG_SSE2 | CPUFLAG_SSE)
#define CPUFLAG_SSE2SLOW (AV_CPU_FLAG_SSE2SLOW | CPUFLAG_SSE2) #define CPUFLAG_SSE2SLOW (AV_CPU_FLAG_SSE2SLOW | CPUFLAG_SSE2)
#define CPUFLAG_SSE3 (AV_CPU_FLAG_SSE3 | CPUFLAG_SSE2) #define CPUFLAG_SSE3 (AV_CPU_FLAG_SSE3 | CPUFLAG_SSE2)
@ -67,7 +67,7 @@ int av_parse_cpu_flags(const char *s)
{ "altivec" , NULL, 0, AV_OPT_TYPE_CONST, { AV_CPU_FLAG_ALTIVEC }, .unit = "flags" }, { "altivec" , NULL, 0, AV_OPT_TYPE_CONST, { AV_CPU_FLAG_ALTIVEC }, .unit = "flags" },
#elif ARCH_X86 #elif ARCH_X86
{ "mmx" , NULL, 0, AV_OPT_TYPE_CONST, { AV_CPU_FLAG_MMX }, .unit = "flags" }, { "mmx" , NULL, 0, AV_OPT_TYPE_CONST, { AV_CPU_FLAG_MMX }, .unit = "flags" },
{ "mmx2" , NULL, 0, AV_OPT_TYPE_CONST, { CPUFLAG_MMX2 }, .unit = "flags" }, { "mmxext" , NULL, 0, AV_OPT_TYPE_CONST, { CPUFLAG_MMXEXT }, .unit = "flags" },
{ "sse" , NULL, 0, AV_OPT_TYPE_CONST, { CPUFLAG_SSE }, .unit = "flags" }, { "sse" , NULL, 0, AV_OPT_TYPE_CONST, { CPUFLAG_SSE }, .unit = "flags" },
{ "sse2" , NULL, 0, AV_OPT_TYPE_CONST, { CPUFLAG_SSE2 }, .unit = "flags" }, { "sse2" , NULL, 0, AV_OPT_TYPE_CONST, { CPUFLAG_SSE2 }, .unit = "flags" },
{ "sse2slow", NULL, 0, AV_OPT_TYPE_CONST, { CPUFLAG_SSE2SLOW }, .unit = "flags" }, { "sse2slow", NULL, 0, AV_OPT_TYPE_CONST, { CPUFLAG_SSE2SLOW }, .unit = "flags" },
@ -129,7 +129,7 @@ static const struct {
{ AV_CPU_FLAG_ALTIVEC, "altivec" }, { AV_CPU_FLAG_ALTIVEC, "altivec" },
#elif ARCH_X86 #elif ARCH_X86
{ AV_CPU_FLAG_MMX, "mmx" }, { AV_CPU_FLAG_MMX, "mmx" },
{ AV_CPU_FLAG_MMX2, "mmx2" }, { AV_CPU_FLAG_MMXEXT, "mmxext" },
{ AV_CPU_FLAG_SSE, "sse" }, { AV_CPU_FLAG_SSE, "sse" },
{ AV_CPU_FLAG_SSE2, "sse2" }, { AV_CPU_FLAG_SSE2, "sse2" },
{ AV_CPU_FLAG_SSE2SLOW, "sse2(slow)" }, { AV_CPU_FLAG_SSE2SLOW, "sse2(slow)" },

5
libavutil/cpu.h
View File

@ -21,11 +21,16 @@
#ifndef AVUTIL_CPU_H #ifndef AVUTIL_CPU_H
#define AVUTIL_CPU_H #define AVUTIL_CPU_H
#include "version.h"
#define AV_CPU_FLAG_FORCE 0x80000000 /* force usage of selected flags (OR) */ #define AV_CPU_FLAG_FORCE 0x80000000 /* force usage of selected flags (OR) */
/* lower 16 bits - CPU features */ /* lower 16 bits - CPU features */
#define AV_CPU_FLAG_MMX 0x0001 ///< standard MMX #define AV_CPU_FLAG_MMX 0x0001 ///< standard MMX
#define AV_CPU_FLAG_MMXEXT 0x0002 ///< SSE integer functions or AMD MMX ext
#if LIBAVUTIL_VERSION_MAJOR < 52
#define AV_CPU_FLAG_MMX2 0x0002 ///< SSE integer functions or AMD MMX ext #define AV_CPU_FLAG_MMX2 0x0002 ///< SSE integer functions or AMD MMX ext
#endif
#define AV_CPU_FLAG_3DNOW 0x0004 ///< AMD 3DNOW #define AV_CPU_FLAG_3DNOW 0x0004 ///< AMD 3DNOW
#define AV_CPU_FLAG_SSE 0x0008 ///< SSE functions #define AV_CPU_FLAG_SSE 0x0008 ///< SSE functions
#define AV_CPU_FLAG_SSE2 0x0010 ///< PIV SSE2 functions #define AV_CPU_FLAG_SSE2 0x0010 ///< PIV SSE2 functions

2
libavutil/version.h
View File

@ -38,7 +38,7 @@
#define LIBAVUTIL_VERSION_MAJOR 51 #define LIBAVUTIL_VERSION_MAJOR 51
#define LIBAVUTIL_VERSION_MINOR 37 #define LIBAVUTIL_VERSION_MINOR 37
#define LIBAVUTIL_VERSION_MICRO 0 #define LIBAVUTIL_VERSION_MICRO 1
#define LIBAVUTIL_VERSION_INT AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \ #define LIBAVUTIL_VERSION_INT AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \
LIBAVUTIL_VERSION_MINOR, \ LIBAVUTIL_VERSION_MINOR, \

4
libavutil/x86/cpu.c
View File

@ -122,7 +122,7 @@ int ff_get_cpu_flags_x86(void)
if (std_caps & (1 << 23)) if (std_caps & (1 << 23))
rval |= AV_CPU_FLAG_MMX; rval |= AV_CPU_FLAG_MMX;
if (std_caps & (1 << 25)) if (std_caps & (1 << 25))
rval |= AV_CPU_FLAG_MMX2; rval |= AV_CPU_FLAG_MMXEXT;
#if HAVE_SSE #if HAVE_SSE
if (std_caps & (1 << 25)) if (std_caps & (1 << 25))
rval |= AV_CPU_FLAG_SSE; rval |= AV_CPU_FLAG_SSE;
@ -159,7 +159,7 @@ int ff_get_cpu_flags_x86(void)
if (ext_caps & (1 << 23)) if (ext_caps & (1 << 23))
rval |= AV_CPU_FLAG_MMX; rval |= AV_CPU_FLAG_MMX;
if (ext_caps & (1 << 22)) if (ext_caps & (1 << 22))
rval |= AV_CPU_FLAG_MMX2; rval |= AV_CPU_FLAG_MMXEXT;
/* Allow for selectively disabling SSE2 functions on AMD processors /* Allow for selectively disabling SSE2 functions on AMD processors
with SSE2 support but not SSE4a. This includes Athlon64, some with SSE2 support but not SSE4a. This includes Athlon64, some

4
libswscale/swscale.c
View File

@ -661,8 +661,8 @@ static int swScale(SwsContext *c, const uint8_t *src[],
if (isPlanar(dstFormat) && isALPHA(dstFormat) && !alpPixBuf) if (isPlanar(dstFormat) && isALPHA(dstFormat) && !alpPixBuf)
fillPlane(dst[3], dstStride[3], dstW, dstY - lastDstY, lastDstY, 255); fillPlane(dst[3], dstStride[3], dstW, dstY - lastDstY, lastDstY, 255);
#if HAVE_MMX2 && HAVE_INLINE_ASM #if HAVE_MMXEXT && HAVE_INLINE_ASM
if (av_get_cpu_flags() & AV_CPU_FLAG_MMX2) if (av_get_cpu_flags() & AV_CPU_FLAG_MMXEXT)
__asm__ volatile ("sfence" ::: "memory"); __asm__ volatile ("sfence" ::: "memory");
#endif #endif
emms_c(); emms_c();

3
libswscale/swscale.h
View File

@ -82,7 +82,10 @@ const char *swscale_license(void);
* are only provided for API compatibility. * are only provided for API compatibility.
*/ */
#define SWS_CPU_CAPS_MMX 0x80000000 #define SWS_CPU_CAPS_MMX 0x80000000
#define SWS_CPU_CAPS_MMXEXT 0x20000000
#if LIBSWSCALE_VERSION_MAJOR < 3
#define SWS_CPU_CAPS_MMX2 0x20000000 #define SWS_CPU_CAPS_MMX2 0x20000000
#endif
#define SWS_CPU_CAPS_3DNOW 0x40000000 #define SWS_CPU_CAPS_3DNOW 0x40000000
#define SWS_CPU_CAPS_ALTIVEC 0x10000000 #define SWS_CPU_CAPS_ALTIVEC 0x10000000
#define SWS_CPU_CAPS_BFIN 0x01000000 #define SWS_CPU_CAPS_BFIN 0x01000000

12
libswscale/utils.c
View File

@ -577,7 +577,7 @@ fail:
return ret; return ret;
} }
#if HAVE_MMX2 && HAVE_INLINE_ASM #if HAVE_MMXEXT && HAVE_INLINE_ASM
static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode,
int16_t *filter, int32_t *filterPos, int numSplits) int16_t *filter, int32_t *filterPos, int numSplits)
{ {
@ -740,7 +740,7 @@ static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode,
return fragmentPos + 1; return fragmentPos + 1;
} }
#endif /* HAVE_MMX2 && HAVE_INLINE_ASM */ #endif /* HAVE_MMXEXT && HAVE_INLINE_ASM */
static void getSubSampleFactors(int *h, int *v, enum PixelFormat format) static void getSubSampleFactors(int *h, int *v, enum PixelFormat format)
{ {
@ -973,7 +973,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FF_ALLOC_OR_GOTO(c, c->formatConvBuffer,
(FFALIGN(srcW, 16) * 2 * FFALIGN(c->srcBpc, 8) >> 3) + 16, (FFALIGN(srcW, 16) * 2 * FFALIGN(c->srcBpc, 8) >> 3) + 16,
fail); fail);
if (HAVE_MMX2 && HAVE_INLINE_ASM && cpu_flags & AV_CPU_FLAG_MMX2 && if (HAVE_MMXEXT && HAVE_INLINE_ASM && cpu_flags & AV_CPU_FLAG_MMXEXT &&
c->srcBpc == 8 && c->dstBpc <= 10) { c->srcBpc == 8 && c->dstBpc <= 10) {
c->canMMX2BeUsed = (dstW >= srcW && (dstW & 31) == 0 && c->canMMX2BeUsed = (dstW >= srcW && (dstW & 31) == 0 &&
(srcW & 15) == 0) ? 1 : 0; (srcW & 15) == 0) ? 1 : 0;
@ -1012,7 +1012,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
/* precalculate horizontal scaler filter coefficients */ /* precalculate horizontal scaler filter coefficients */
{ {
#if HAVE_MMX2 && HAVE_INLINE_ASM #if HAVE_MMXEXT && HAVE_INLINE_ASM
// can't downscale !!! // can't downscale !!!
if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) { if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {
c->lumMmx2FilterCodeSize = initMMX2HScaler(dstW, c->lumXInc, NULL, c->lumMmx2FilterCodeSize = initMMX2HScaler(dstW, c->lumXInc, NULL,
@ -1048,7 +1048,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ); mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
#endif #endif
} else } else
#endif /* HAVE_MMX2 && HAVE_INLINE_ASM */ #endif /* HAVE_MMXEXT && HAVE_INLINE_ASM */
{ {
const int filterAlign = const int filterAlign =
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 : (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 :
@ -1208,7 +1208,7 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
#endif #endif
sws_format_name(dstFormat)); sws_format_name(dstFormat));
if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2) if (HAVE_MMXEXT && cpu_flags & AV_CPU_FLAG_MMXEXT)
av_log(c, AV_LOG_INFO, "using MMX2\n"); av_log(c, AV_LOG_INFO, "using MMX2\n");
else if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW) else if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW)
av_log(c, AV_LOG_INFO, "using 3DNOW\n"); av_log(c, AV_LOG_INFO, "using 3DNOW\n");

2
libswscale/version.h
View File

@ -28,7 +28,7 @@
#define LIBSWSCALE_VERSION_MAJOR 2 #define LIBSWSCALE_VERSION_MAJOR 2
#define LIBSWSCALE_VERSION_MINOR 1 #define LIBSWSCALE_VERSION_MINOR 1
#define LIBSWSCALE_VERSION_MICRO 0 #define LIBSWSCALE_VERSION_MICRO 1
#define LIBSWSCALE_VERSION_INT AV_VERSION_INT(LIBSWSCALE_VERSION_MAJOR, \ #define LIBSWSCALE_VERSION_INT AV_VERSION_INT(LIBSWSCALE_VERSION_MAJOR, \
LIBSWSCALE_VERSION_MINOR, \ LIBSWSCALE_VERSION_MINOR, \

12
libswscale/x86/rgb2rgb.c
View File

@ -85,7 +85,7 @@ DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;
//Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one. //Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
#define COMPILE_TEMPLATE_MMX2 0 #define COMPILE_TEMPLATE_MMXEXT 0
#define COMPILE_TEMPLATE_AMD3DNOW 0 #define COMPILE_TEMPLATE_AMD3DNOW 0
#define COMPILE_TEMPLATE_SSE2 0 #define COMPILE_TEMPLATE_SSE2 0
@ -96,8 +96,8 @@ DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;
//MMX2 versions //MMX2 versions
#undef RENAME #undef RENAME
#undef COMPILE_TEMPLATE_MMX2 #undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMX2 1 #define COMPILE_TEMPLATE_MMXEXT 1
#define RENAME(a) a ## _MMX2 #define RENAME(a) a ## _MMX2
#include "rgb2rgb_template.c" #include "rgb2rgb_template.c"
@ -110,10 +110,10 @@ DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;
//3DNOW versions //3DNOW versions
#undef RENAME #undef RENAME
#undef COMPILE_TEMPLATE_MMX2 #undef COMPILE_TEMPLATE_MMXEXT
#undef COMPILE_TEMPLATE_SSE2 #undef COMPILE_TEMPLATE_SSE2
#undef COMPILE_TEMPLATE_AMD3DNOW #undef COMPILE_TEMPLATE_AMD3DNOW
#define COMPILE_TEMPLATE_MMX2 0 #define COMPILE_TEMPLATE_MMXEXT 0
#define COMPILE_TEMPLATE_SSE2 0 #define COMPILE_TEMPLATE_SSE2 0
#define COMPILE_TEMPLATE_AMD3DNOW 1 #define COMPILE_TEMPLATE_AMD3DNOW 1
#define RENAME(a) a ## _3DNOW #define RENAME(a) a ## _3DNOW
@ -137,7 +137,7 @@ av_cold void rgb2rgb_init_x86(void)
rgb2rgb_init_MMX(); rgb2rgb_init_MMX();
if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW) if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW)
rgb2rgb_init_3DNOW(); rgb2rgb_init_3DNOW();
if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2) if (HAVE_MMXEXT && cpu_flags & AV_CPU_FLAG_MMXEXT)
rgb2rgb_init_MMX2(); rgb2rgb_init_MMX2();
if (HAVE_SSE && cpu_flags & AV_CPU_FLAG_SSE2) if (HAVE_SSE && cpu_flags & AV_CPU_FLAG_SSE2)
rgb2rgb_init_SSE2(); rgb2rgb_init_SSE2();

18
libswscale/x86/rgb2rgb_template.c
View File

@ -35,7 +35,7 @@
#if COMPILE_TEMPLATE_AMD3DNOW #if COMPILE_TEMPLATE_AMD3DNOW
#define PREFETCH "prefetch" #define PREFETCH "prefetch"
#define PAVGB "pavgusb" #define PAVGB "pavgusb"
#elif COMPILE_TEMPLATE_MMX2 #elif COMPILE_TEMPLATE_MMXEXT
#define PREFETCH "prefetchnta" #define PREFETCH "prefetchnta"
#define PAVGB "pavgb" #define PAVGB "pavgb"
#else #else
@ -49,7 +49,7 @@
#define EMMS "emms" #define EMMS "emms"
#endif #endif
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
#define MOVNTQ "movntq" #define MOVNTQ "movntq"
#define SFENCE "sfence" #define SFENCE "sfence"
#else #else
@ -1148,7 +1148,7 @@ static inline void RENAME(shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst,
PREFETCH" 32(%1, %0) \n\t" PREFETCH" 32(%1, %0) \n\t"
"movq (%1, %0), %%mm0 \n\t" "movq (%1, %0), %%mm0 \n\t"
"movq 8(%1, %0), %%mm1 \n\t" "movq 8(%1, %0), %%mm1 \n\t"
# if COMPILE_TEMPLATE_MMX2 # if COMPILE_TEMPLATE_MMXEXT
"pshufw $177, %%mm0, %%mm3 \n\t" "pshufw $177, %%mm0, %%mm3 \n\t"
"pshufw $177, %%mm1, %%mm5 \n\t" "pshufw $177, %%mm1, %%mm5 \n\t"
"pand %%mm7, %%mm0 \n\t" "pand %%mm7, %%mm0 \n\t"
@ -1512,7 +1512,7 @@ static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t
} }
#endif /* !COMPILE_TEMPLATE_AMD3DNOW */ #endif /* !COMPILE_TEMPLATE_AMD3DNOW */
#if COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW #if COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW
static inline void RENAME(planar2x)(const uint8_t *src, uint8_t *dst, int srcWidth, int srcHeight, int srcStride, int dstStride) static inline void RENAME(planar2x)(const uint8_t *src, uint8_t *dst, int srcWidth, int srcHeight, int srcStride, int dstStride)
{ {
int x,y; int x,y;
@ -1602,7 +1602,7 @@ static inline void RENAME(planar2x)(const uint8_t *src, uint8_t *dst, int srcWid
SFENCE" \n\t" SFENCE" \n\t"
:::"memory"); :::"memory");
} }
#endif /* COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW */ #endif /* COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW */
#if !COMPILE_TEMPLATE_AMD3DNOW #if !COMPILE_TEMPLATE_AMD3DNOW
/** /**
@ -1810,7 +1810,7 @@ static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_
"1: \n\t" "1: \n\t"
PREFETCH" 64(%0, %%"REG_d") \n\t" PREFETCH" 64(%0, %%"REG_d") \n\t"
PREFETCH" 64(%1, %%"REG_d") \n\t" PREFETCH" 64(%1, %%"REG_d") \n\t"
#if COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW #if COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW
"movq (%0, %%"REG_d"), %%mm0 \n\t" "movq (%0, %%"REG_d"), %%mm0 \n\t"
"movq (%1, %%"REG_d"), %%mm1 \n\t" "movq (%1, %%"REG_d"), %%mm1 \n\t"
"movq 6(%0, %%"REG_d"), %%mm2 \n\t" "movq 6(%0, %%"REG_d"), %%mm2 \n\t"
@ -1871,7 +1871,7 @@ static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_
"packssdw %%mm1, %%mm0 \n\t" // V1 V0 U1 U0 "packssdw %%mm1, %%mm0 \n\t" // V1 V0 U1 U0
"psraw $7, %%mm0 \n\t" "psraw $7, %%mm0 \n\t"
#if COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW #if COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW
"movq 12(%0, %%"REG_d"), %%mm4 \n\t" "movq 12(%0, %%"REG_d"), %%mm4 \n\t"
"movq 12(%1, %%"REG_d"), %%mm1 \n\t" "movq 12(%1, %%"REG_d"), %%mm1 \n\t"
"movq 18(%0, %%"REG_d"), %%mm2 \n\t" "movq 18(%0, %%"REG_d"), %%mm2 \n\t"
@ -2592,9 +2592,9 @@ static inline void RENAME(rgb2rgb_init)(void)
yuyvtoyuv422 = RENAME(yuyvtoyuv422); yuyvtoyuv422 = RENAME(yuyvtoyuv422);
#endif /* !COMPILE_TEMPLATE_SSE2 */ #endif /* !COMPILE_TEMPLATE_SSE2 */
#if COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW #if COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW
planar2x = RENAME(planar2x); planar2x = RENAME(planar2x);
#endif /* COMPILE_TEMPLATE_MMX2 || COMPILE_TEMPLATE_AMD3DNOW */ #endif /* COMPILE_TEMPLATE_MMXEXT || COMPILE_TEMPLATE_AMD3DNOW */
rgb24toyv12 = RENAME(rgb24toyv12); rgb24toyv12 = RENAME(rgb24toyv12);
yuyvtoyuv420 = RENAME(yuyvtoyuv420); yuyvtoyuv420 = RENAME(yuyvtoyuv420);

16
libswscale/x86/swscale.c
View File

@ -72,16 +72,16 @@ DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
//MMX versions //MMX versions
#if HAVE_MMX #if HAVE_MMX
#undef RENAME #undef RENAME
#define COMPILE_TEMPLATE_MMX2 0 #define COMPILE_TEMPLATE_MMXEXT 0
#define RENAME(a) a ## _MMX #define RENAME(a) a ## _MMX
#include "swscale_template.c" #include "swscale_template.c"
#endif #endif
//MMX2 versions //MMX2 versions
#if HAVE_MMX2 #if HAVE_MMXEXT
#undef RENAME #undef RENAME
#undef COMPILE_TEMPLATE_MMX2 #undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMX2 1 #define COMPILE_TEMPLATE_MMXEXT 1
#define RENAME(a) a ## _MMX2 #define RENAME(a) a ## _MMX2
#include "swscale_template.c" #include "swscale_template.c"
#endif #endif
@ -308,8 +308,8 @@ av_cold void ff_sws_init_swScale_mmx(SwsContext *c)
#if HAVE_INLINE_ASM #if HAVE_INLINE_ASM
if (cpu_flags & AV_CPU_FLAG_MMX) if (cpu_flags & AV_CPU_FLAG_MMX)
sws_init_swScale_MMX(c); sws_init_swScale_MMX(c);
#if HAVE_MMX2 #if HAVE_MMXEXT
if (cpu_flags & AV_CPU_FLAG_MMX2) if (cpu_flags & AV_CPU_FLAG_MMXEXT)
sws_init_swScale_MMX2(c); sws_init_swScale_MMX2(c);
#endif #endif
#endif /* HAVE_INLINE_ASM */ #endif /* HAVE_INLINE_ASM */
@ -360,7 +360,7 @@ switch(c->dstBpc){ \
if (cpu_flags & AV_CPU_FLAG_MMX) { if (cpu_flags & AV_CPU_FLAG_MMX) {
ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx); ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx);
ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx); ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx);
ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMX2); ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmx2, cpu_flags & AV_CPU_FLAG_MMXEXT);
switch (c->srcFormat) { switch (c->srcFormat) {
case PIX_FMT_Y400A: case PIX_FMT_Y400A:
@ -392,7 +392,7 @@ switch(c->dstBpc){ \
break; break;
} }
} }
if (cpu_flags & AV_CPU_FLAG_MMX2) { if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmx2, , 1); ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmx2, , 1);
} }
#endif #endif

18
libswscale/x86/swscale_template.c
View File

@ -22,13 +22,13 @@
#undef MOVNTQ #undef MOVNTQ
#undef PREFETCH #undef PREFETCH
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
#define PREFETCH "prefetchnta" #define PREFETCH "prefetchnta"
#else #else
#define PREFETCH " # nop" #define PREFETCH " # nop"
#endif #endif
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
#define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t" #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
#else #else
#define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t" #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
@ -567,7 +567,7 @@ static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
"cmp "#dstw", "#index" \n\t"\ "cmp "#dstw", "#index" \n\t"\
" jb 1b \n\t" " jb 1b \n\t"
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
#undef WRITEBGR24 #undef WRITEBGR24
#define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX2(dst, dstw, index) #define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX2(dst, dstw, index)
#else #else
@ -1371,7 +1371,7 @@ static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
} }
} }
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst, static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
int dstWidth, const uint8_t *src, int dstWidth, const uint8_t *src,
int srcW, int xInc) int srcW, int xInc)
@ -1553,7 +1553,7 @@ static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
dst2[i] = src2[srcW-1]*128; dst2[i] = src2[srcW-1]*128;
} }
} }
#endif /* COMPILE_TEMPLATE_MMX2 */ #endif /* COMPILE_TEMPLATE_MMXEXT */
static av_cold void RENAME(sws_init_swScale)(SwsContext *c) static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
{ {
@ -1616,17 +1616,17 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
if (c->srcBpc == 8 && c->dstBpc <= 10) { if (c->srcBpc == 8 && c->dstBpc <= 10) {
// Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one). // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
if (c->flags & SWS_FAST_BILINEAR && c->canMMX2BeUsed) if (c->flags & SWS_FAST_BILINEAR && c->canMMX2BeUsed)
{ {
c->hyscale_fast = RENAME(hyscale_fast); c->hyscale_fast = RENAME(hyscale_fast);
c->hcscale_fast = RENAME(hcscale_fast); c->hcscale_fast = RENAME(hcscale_fast);
} else { } else {
#endif /* COMPILE_TEMPLATE_MMX2 */ #endif /* COMPILE_TEMPLATE_MMXEXT */
c->hyscale_fast = NULL; c->hyscale_fast = NULL;
c->hcscale_fast = NULL; c->hcscale_fast = NULL;
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
} }
#endif /* COMPILE_TEMPLATE_MMX2 */ #endif /* COMPILE_TEMPLATE_MMXEXT */
} }
} }

16
libswscale/x86/yuv2rgb.c
View File

@ -52,20 +52,20 @@ DECLARE_ASM_CONST(8, uint64_t, pb_07) = 0x0707070707070707ULL;
//MMX versions //MMX versions
#if HAVE_MMX #if HAVE_MMX
#undef RENAME #undef RENAME
#undef COMPILE_TEMPLATE_MMX2 #undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMX2 0 #define COMPILE_TEMPLATE_MMXEXT 0
#define RENAME(a) a ## _MMX #define RENAME(a) a ## _MMX
#include "yuv2rgb_template.c" #include "yuv2rgb_template.c"
#endif /* HAVE_MMX */ #endif /* HAVE_MMX */
//MMX2 versions //MMX2 versions
#if HAVE_MMX2 #if HAVE_MMXEXT
#undef RENAME #undef RENAME
#undef COMPILE_TEMPLATE_MMX2 #undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMX2 1 #define COMPILE_TEMPLATE_MMXEXT 1
#define RENAME(a) a ## _MMX2 #define RENAME(a) a ## _MMX2
#include "yuv2rgb_template.c" #include "yuv2rgb_template.c"
#endif /* HAVE_MMX2 */ #endif /* HAVE_MMXEXT */
#endif /* HAVE_INLINE_ASM */ #endif /* HAVE_INLINE_ASM */
@ -78,8 +78,8 @@ av_cold SwsFunc ff_yuv2rgb_init_mmx(SwsContext *c)
c->srcFormat != PIX_FMT_YUVA420P) c->srcFormat != PIX_FMT_YUVA420P)
return NULL; return NULL;
#if HAVE_MMX2 #if HAVE_MMXEXT
if (cpu_flags & AV_CPU_FLAG_MMX2) { if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
switch (c->dstFormat) { switch (c->dstFormat) {
case PIX_FMT_RGB24: return yuv420_rgb24_MMX2; case PIX_FMT_RGB24: return yuv420_rgb24_MMX2;
case PIX_FMT_BGR24: return yuv420_bgr24_MMX2; case PIX_FMT_BGR24: return yuv420_bgr24_MMX2;

12
libswscale/x86/yuv2rgb_template.c
View File

@ -25,7 +25,7 @@
#undef EMMS #undef EMMS
#undef SFENCE #undef SFENCE
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
#define MOVNTQ "movntq" #define MOVNTQ "movntq"
#define SFENCE "sfence" #define SFENCE "sfence"
#else #else
@ -182,7 +182,7 @@
"paddusb "GREEN_DITHER"(%4), %%mm2\n\t" \ "paddusb "GREEN_DITHER"(%4), %%mm2\n\t" \
"paddusb "RED_DITHER"(%4), %%mm1\n\t" \ "paddusb "RED_DITHER"(%4), %%mm1\n\t" \
#if !COMPILE_TEMPLATE_MMX2 #if !COMPILE_TEMPLATE_MMXEXT
static inline int RENAME(yuv420_rgb15)(SwsContext *c, const uint8_t *src[], static inline int RENAME(yuv420_rgb15)(SwsContext *c, const uint8_t *src[],
int srcStride[], int srcStride[],
int srcSliceY, int srcSliceH, int srcSliceY, int srcSliceH,
@ -238,7 +238,7 @@ static inline int RENAME(yuv420_rgb16)(SwsContext *c, const uint8_t *src[],
YUV2RGB_OPERANDS YUV2RGB_OPERANDS
YUV2RGB_ENDFUNC YUV2RGB_ENDFUNC
} }
#endif /* !COMPILE_TEMPLATE_MMX2 */ #endif /* !COMPILE_TEMPLATE_MMXEXT */
#define RGB_PACK24(blue, red)\ #define RGB_PACK24(blue, red)\
"packuswb %%mm3, %%mm0 \n" /* R0 R2 R4 R6 R1 R3 R5 R7 */\ "packuswb %%mm3, %%mm0 \n" /* R0 R2 R4 R6 R1 R3 R5 R7 */\
@ -255,7 +255,7 @@ static inline int RENAME(yuv420_rgb16)(SwsContext *c, const uint8_t *src[],
"punpckhwd %%mm6, %%mm5 \n" /* R4 G4 B4 R5 R6 G6 B6 R7 */\ "punpckhwd %%mm6, %%mm5 \n" /* R4 G4 B4 R5 R6 G6 B6 R7 */\
RGB_PACK24_B RGB_PACK24_B
#if COMPILE_TEMPLATE_MMX2 #if COMPILE_TEMPLATE_MMXEXT
DECLARE_ASM_CONST(8, int16_t, mask1101[4]) = {-1,-1, 0,-1}; DECLARE_ASM_CONST(8, int16_t, mask1101[4]) = {-1,-1, 0,-1};
DECLARE_ASM_CONST(8, int16_t, mask0010[4]) = { 0, 0,-1, 0}; DECLARE_ASM_CONST(8, int16_t, mask0010[4]) = { 0, 0,-1, 0};
DECLARE_ASM_CONST(8, int16_t, mask0110[4]) = { 0,-1,-1, 0}; DECLARE_ASM_CONST(8, int16_t, mask0110[4]) = { 0,-1,-1, 0};
@ -362,7 +362,7 @@ static inline int RENAME(yuv420_bgr24)(SwsContext *c, const uint8_t *src[],
MOVNTQ " %%mm5, 16(%1)\n\t" \ MOVNTQ " %%mm5, 16(%1)\n\t" \
MOVNTQ " %%mm"alpha", 24(%1)\n\t" \ MOVNTQ " %%mm"alpha", 24(%1)\n\t" \
#if !COMPILE_TEMPLATE_MMX2 #if !COMPILE_TEMPLATE_MMXEXT
static inline int RENAME(yuv420_rgb32)(SwsContext *c, const uint8_t *src[], static inline int RENAME(yuv420_rgb32)(SwsContext *c, const uint8_t *src[],
int srcStride[], int srcStride[],
int srcSliceY, int srcSliceH, int srcSliceY, int srcSliceH,
@ -449,4 +449,4 @@ static inline int RENAME(yuva420_bgr32)(SwsContext *c, const uint8_t *src[],
} }
#endif #endif
#endif /* !COMPILE_TEMPLATE_MMX2 */ #endif /* !COMPILE_TEMPLATE_MMXEXT */