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

Consistently use "cpu_flags" as variable/parameter name for CPU flags

tags/n2.1
Diego Biurrun 12 years ago
parent
b6293e2798
commit
3ac7fa81b2
34 changed files with 235 additions and 237 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +1
    -3
      libavcodec/ppc/hpeldsp_altivec.c
  2. +10
    -10
      libavcodec/x86/ac3dsp_init.c
  3. +4
    -4
      libavcodec/x86/cavsdsp.c
  4. +26
    -26
      libavcodec/x86/dsputil_init.c
  5. +14
    -14
      libavcodec/x86/dsputilenc_mmx.c
  6. +11
    -9
      libavcodec/x86/fft_init.c
  7. +6
    -6
      libavcodec/x86/fmtconvert_init.c
  8. +11
    -11
      libavcodec/x86/h264_intrapred_init.c
  9. +6
    -6
      libavcodec/x86/h264_qpel.c
  10. +8
    -8
      libavcodec/x86/h264chroma_init.c
  11. +13
    -13
      libavcodec/x86/h264dsp_init.c
  12. +14
    -14
      libavcodec/x86/hpeldsp_init.c
  13. +2
    -2
      libavcodec/x86/lpc.c
  14. +4
    -4
      libavcodec/x86/motion_est.c
  15. +7
    -7
      libavcodec/x86/mpegaudiodsp.c
  16. +3
    -3
      libavcodec/x86/mpegvideo.c
  17. +5
    -5
      libavcodec/x86/mpegvideoenc.c
  18. +5
    -5
      libavcodec/x86/pngdsp_init.c
  19. +4
    -4
      libavcodec/x86/proresdsp_init.c
  20. +4
    -4
      libavcodec/x86/rv34dsp_init.c
  21. +6
    -6
      libavcodec/x86/rv40dsp_init.c
  22. +3
    -3
      libavcodec/x86/sbrdsp_init.c
  23. +9
    -9
      libavcodec/x86/vc1dsp_init.c
  24. +5
    -5
      libavcodec/x86/videodsp_init.c
  25. +3
    -3
      libavcodec/x86/vorbisdsp_init.c
  26. +4
    -4
      libavcodec/x86/vp3dsp_init.c
  27. +3
    -3
      libavcodec/x86/vp56dsp_init.c
  28. +8
    -8
      libavcodec/x86/vp8dsp_init.c
  29. +5
    -5
      libavfilter/x86/af_volume_init.c
  30. +8
    -8
      libavresample/x86/audio_convert_init.c
  31. +10
    -10
      libavresample/x86/audio_mix_init.c
  32. +6
    -6
      libavresample/x86/dither_init.c
  33. +1
    -3
      libavutil/ppc/float_dsp_init.c
  34. +6
    -6
      libavutil/x86/float_dsp_init.c

+ 1
- 3
libavcodec/ppc/hpeldsp_altivec.c View File

@@ -449,9 +449,7 @@ static void avg_pixels8_xy2_altivec(uint8_t *block, const uint8_t *pixels, ptrdi
av_cold void ff_hpeldsp_init_ppc(HpelDSPContext *c, int flags)
{
#if HAVE_ALTIVEC
int mm_flags = av_get_cpu_flags();

if (mm_flags & AV_CPU_FLAG_ALTIVEC) {
if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) {
c->avg_pixels_tab[0][0] = ff_avg_pixels16_altivec;
c->avg_pixels_tab[1][0] = avg_pixels8_altivec;
c->avg_pixels_tab[1][3] = avg_pixels8_xy2_altivec;


+ 10
- 10
libavcodec/x86/ac3dsp_init.c View File

@@ -180,46 +180,46 @@ static void ac3_downmix_sse(float **samples, float (*matrix)[2],

av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
c->ac3_exponent_min = ff_ac3_exponent_min_mmx;
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx;
c->ac3_lshift_int16 = ff_ac3_lshift_int16_mmx;
c->ac3_rshift_int32 = ff_ac3_rshift_int32_mmx;
}
if (EXTERNAL_AMD3DNOW(mm_flags)) {
if (EXTERNAL_AMD3DNOW(cpu_flags)) {
if (!bit_exact) {
c->float_to_fixed24 = ff_float_to_fixed24_3dnow;
}
}
if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->ac3_exponent_min = ff_ac3_exponent_min_mmxext;
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmxext;
}
if (EXTERNAL_SSE(mm_flags)) {
if (EXTERNAL_SSE(cpu_flags)) {
c->float_to_fixed24 = ff_float_to_fixed24_sse;
}
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
c->ac3_exponent_min = ff_ac3_exponent_min_sse2;
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_sse2;
c->float_to_fixed24 = ff_float_to_fixed24_sse2;
c->compute_mantissa_size = ff_ac3_compute_mantissa_size_sse2;
c->extract_exponents = ff_ac3_extract_exponents_sse2;
if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) {
c->ac3_lshift_int16 = ff_ac3_lshift_int16_sse2;
c->ac3_rshift_int32 = ff_ac3_rshift_int32_sse2;
}
}
if (EXTERNAL_SSSE3(mm_flags)) {
if (EXTERNAL_SSSE3(cpu_flags)) {
c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_ssse3;
if (!(mm_flags & AV_CPU_FLAG_ATOM)) {
if (!(cpu_flags & AV_CPU_FLAG_ATOM)) {
c->extract_exponents = ff_ac3_extract_exponents_ssse3;
}
}

#if HAVE_SSE_INLINE && HAVE_7REGS
if (INLINE_SSE(mm_flags)) {
if (INLINE_SSE(cpu_flags)) {
c->downmix = ac3_downmix_sse;
}
#endif


+ 4
- 4
libavcodec/x86/cavsdsp.c View File

@@ -542,17 +542,17 @@ 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 mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (mm_flags & AV_CPU_FLAG_MMX)
if (cpu_flags & AV_CPU_FLAG_MMX)
cavsdsp_init_mmx(c, avctx);
#endif /* HAVE_MMX_INLINE */
#if HAVE_MMXEXT_INLINE
if (mm_flags & AV_CPU_FLAG_MMXEXT)
if (cpu_flags & AV_CPU_FLAG_MMXEXT)
cavsdsp_init_mmxext(c, avctx);
#endif /* HAVE_MMXEXT_INLINE */
#if HAVE_AMD3DNOW_INLINE
if (mm_flags & AV_CPU_FLAG_3DNOW)
if (cpu_flags & AV_CPU_FLAG_3DNOW)
cavsdsp_init_3dnow(c, avctx);
#endif /* HAVE_AMD3DNOW_INLINE */
}

+ 26
- 26
libavcodec/x86/dsputil_init.c View File

@@ -529,7 +529,7 @@ QPEL_OP(put_no_rnd_, _no_rnd_, mmxext)
} while (0)

static av_cold void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx,
int mm_flags)
int cpu_flags)
{
#if HAVE_MMX_INLINE
const int high_bit_depth = avctx->bits_per_raw_sample > 8;
@@ -575,7 +575,7 @@ static av_cold void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx,
}

static av_cold void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
int mm_flags)
int cpu_flags)
{
#if HAVE_MMXEXT_INLINE
const int high_bit_depth = avctx->bits_per_raw_sample > 8;
@@ -597,7 +597,7 @@ static av_cold void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmxext, );

/* slower than cmov version on AMD */
if (!(mm_flags & AV_CPU_FLAG_3DNOW))
if (!(cpu_flags & AV_CPU_FLAG_3DNOW))
c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmxext;

c->scalarproduct_int16 = ff_scalarproduct_int16_mmxext;
@@ -612,7 +612,7 @@ static av_cold void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx,
}

static av_cold void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx,
int mm_flags)
int cpu_flags)
{
#if HAVE_SSE_INLINE
const int high_bit_depth = avctx->bits_per_raw_sample > 8;
@@ -630,7 +630,7 @@ static av_cold void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx,
}

static av_cold void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
int mm_flags)
int cpu_flags)
{
#if HAVE_SSE2_INLINE
const int high_bit_depth = avctx->bits_per_raw_sample > 8;
@@ -646,14 +646,14 @@ static av_cold void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
#if HAVE_SSE2_EXTERNAL
c->scalarproduct_int16 = ff_scalarproduct_int16_sse2;
c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_sse2;
if (mm_flags & AV_CPU_FLAG_ATOM) {
if (cpu_flags & AV_CPU_FLAG_ATOM) {
c->vector_clip_int32 = ff_vector_clip_int32_int_sse2;
} else {
c->vector_clip_int32 = ff_vector_clip_int32_sse2;
}
if (avctx->flags & CODEC_FLAG_BITEXACT) {
c->apply_window_int16 = ff_apply_window_int16_sse2;
} else if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
} else if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) {
c->apply_window_int16 = ff_apply_window_int16_round_sse2;
}
c->bswap_buf = ff_bswap32_buf_sse2;
@@ -661,25 +661,25 @@ static av_cold void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
}

static av_cold void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx,
int mm_flags)
int cpu_flags)
{
#if HAVE_SSSE3_EXTERNAL
c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_ssse3;
if (mm_flags & AV_CPU_FLAG_SSE4) // not really sse4, just slow on Conroe
if (cpu_flags & AV_CPU_FLAG_SSE4) // not really SSE4, just slow on Conroe
c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_sse4;

if (mm_flags & AV_CPU_FLAG_ATOM)
if (cpu_flags & AV_CPU_FLAG_ATOM)
c->apply_window_int16 = ff_apply_window_int16_ssse3_atom;
else
c->apply_window_int16 = ff_apply_window_int16_ssse3;
if (!(mm_flags & (AV_CPU_FLAG_SSE42|AV_CPU_FLAG_3DNOW))) // cachesplit
if (!(cpu_flags & (AV_CPU_FLAG_SSE42 | AV_CPU_FLAG_3DNOW))) // cachesplit
c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_ssse3;
c->bswap_buf = ff_bswap32_buf_ssse3;
#endif /* HAVE_SSSE3_EXTERNAL */
}

static av_cold void dsputil_init_sse4(DSPContext *c, AVCodecContext *avctx,
int mm_flags)
int cpu_flags)
{
#if HAVE_SSE4_EXTERNAL
c->vector_clip_int32 = ff_vector_clip_int32_sse4;
@@ -688,30 +688,30 @@ static av_cold void dsputil_init_sse4(DSPContext *c, AVCodecContext *avctx,

av_cold void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

#if HAVE_7REGS && HAVE_INLINE_ASM
if (mm_flags & AV_CPU_FLAG_CMOV)
if (cpu_flags & AV_CPU_FLAG_CMOV)
c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_cmov;
#endif

if (mm_flags & AV_CPU_FLAG_MMX)
dsputil_init_mmx(c, avctx, mm_flags);
if (cpu_flags & AV_CPU_FLAG_MMX)
dsputil_init_mmx(c, avctx, cpu_flags);

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

if (mm_flags & AV_CPU_FLAG_SSE)
dsputil_init_sse(c, avctx, mm_flags);
if (cpu_flags & AV_CPU_FLAG_SSE)
dsputil_init_sse(c, avctx, cpu_flags);

if (mm_flags & AV_CPU_FLAG_SSE2)
dsputil_init_sse2(c, avctx, mm_flags);
if (cpu_flags & AV_CPU_FLAG_SSE2)
dsputil_init_sse2(c, avctx, cpu_flags);

if (mm_flags & AV_CPU_FLAG_SSSE3)
dsputil_init_ssse3(c, avctx, mm_flags);
if (cpu_flags & AV_CPU_FLAG_SSSE3)
dsputil_init_ssse3(c, avctx, cpu_flags);

if (mm_flags & AV_CPU_FLAG_SSE4)
dsputil_init_sse4(c, avctx, mm_flags);
if (cpu_flags & AV_CPU_FLAG_SSE4)
dsputil_init_sse4(c, avctx, cpu_flags);

if (CONFIG_ENCODERS)
ff_dsputilenc_init_mmx(c, avctx);


+ 14
- 14
libavcodec/x86/dsputilenc_mmx.c View File

@@ -945,12 +945,12 @@ hadamard_func(ssse3)

av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

#if HAVE_YASM
int bit_depth = avctx->bits_per_raw_sample;

if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
if (bit_depth <= 8)
c->get_pixels = ff_get_pixels_mmx;
c->diff_pixels = ff_diff_pixels_mmx;
@@ -958,19 +958,19 @@ av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx)

c->pix_norm1 = ff_pix_norm1_mmx;
}
if (EXTERNAL_SSE2(mm_flags))
if (EXTERNAL_SSE2(cpu_flags))
if (bit_depth <= 8)
c->get_pixels = ff_get_pixels_sse2;
#endif /* HAVE_YASM */

#if HAVE_INLINE_ASM
if (mm_flags & AV_CPU_FLAG_MMX) {
if (cpu_flags & AV_CPU_FLAG_MMX) {
const int dct_algo = avctx->dct_algo;
if (avctx->bits_per_raw_sample <= 8 &&
(dct_algo==FF_DCT_AUTO || dct_algo==FF_DCT_MMX)) {
if(mm_flags & AV_CPU_FLAG_SSE2){
if (cpu_flags & AV_CPU_FLAG_SSE2) {
c->fdct = ff_fdct_sse2;
} else if (mm_flags & AV_CPU_FLAG_MMXEXT) {
} else if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
c->fdct = ff_fdct_mmxext;
}else{
c->fdct = ff_fdct_mmx;
@@ -998,7 +998,7 @@ av_cold 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) {
if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
c->sum_abs_dctelem = sum_abs_dctelem_mmxext;
c->vsad[4] = vsad_intra16_mmxext;

@@ -1009,12 +1009,12 @@ av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx)
c->sub_hfyu_median_prediction = sub_hfyu_median_prediction_mmxext;
}

if(mm_flags & AV_CPU_FLAG_SSE2){
if (cpu_flags & AV_CPU_FLAG_SSE2) {
c->sum_abs_dctelem= sum_abs_dctelem_sse2;
}

#if HAVE_SSSE3_INLINE
if(mm_flags & AV_CPU_FLAG_SSSE3){
if (cpu_flags & AV_CPU_FLAG_SSSE3) {
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->try_8x8basis= try_8x8basis_ssse3;
}
@@ -1023,7 +1023,7 @@ av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx)
}
#endif

if(mm_flags & AV_CPU_FLAG_3DNOW){
if (cpu_flags & AV_CPU_FLAG_3DNOW) {
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->try_8x8basis= try_8x8basis_3dnow;
}
@@ -1032,16 +1032,16 @@ av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx)
}
#endif /* HAVE_INLINE_ASM */

if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx;
c->hadamard8_diff[1] = ff_hadamard8_diff_mmx;

if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->hadamard8_diff[0] = ff_hadamard8_diff16_mmxext;
c->hadamard8_diff[1] = ff_hadamard8_diff_mmxext;
}

if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
c->sse[0] = ff_sse16_sse2;

#if HAVE_ALIGNED_STACK
@@ -1050,7 +1050,7 @@ av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx)
#endif
}

if (EXTERNAL_SSSE3(mm_flags) && HAVE_ALIGNED_STACK) {
if (EXTERNAL_SSSE3(cpu_flags) && HAVE_ALIGNED_STACK) {
c->hadamard8_diff[0] = ff_hadamard8_diff16_ssse3;
c->hadamard8_diff[1] = ff_hadamard8_diff_ssse3;
}


+ 11
- 9
libavcodec/x86/fft_init.c View File

@@ -23,22 +23,23 @@

av_cold void ff_fft_init_x86(FFTContext *s)
{
int has_vectors = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

#if ARCH_X86_32
if (EXTERNAL_AMD3DNOW(has_vectors)) {
if (EXTERNAL_AMD3DNOW(cpu_flags)) {
/* 3DNow! for K6-2/3 */
s->imdct_calc = ff_imdct_calc_3dnow;
s->imdct_half = ff_imdct_half_3dnow;
s->fft_calc = ff_fft_calc_3dnow;
}
if (EXTERNAL_AMD3DNOWEXT(has_vectors)) {
if (EXTERNAL_AMD3DNOWEXT(cpu_flags)) {
/* 3DNowEx for K7 */
s->imdct_calc = ff_imdct_calc_3dnowext;
s->imdct_half = ff_imdct_half_3dnowext;
s->fft_calc = ff_fft_calc_3dnowext;
}
#endif
if (EXTERNAL_SSE(has_vectors)) {
if (EXTERNAL_SSE(cpu_flags)) {
/* SSE for P3/P4/K8 */
s->imdct_calc = ff_imdct_calc_sse;
s->imdct_half = ff_imdct_half_sse;
@@ -46,7 +47,7 @@ av_cold void ff_fft_init_x86(FFTContext *s)
s->fft_calc = ff_fft_calc_sse;
s->fft_permutation = FF_FFT_PERM_SWAP_LSBS;
}
if (EXTERNAL_AVX(has_vectors) && s->nbits >= 5) {
if (EXTERNAL_AVX(cpu_flags) && s->nbits >= 5) {
/* AVX for SB */
s->imdct_half = ff_imdct_half_avx;
s->fft_calc = ff_fft_calc_avx;
@@ -57,12 +58,13 @@ av_cold void ff_fft_init_x86(FFTContext *s)
#if CONFIG_DCT
av_cold void ff_dct_init_x86(DCTContext *s)
{
int has_vectors = av_get_cpu_flags();
if (EXTERNAL_SSE(has_vectors))
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_SSE(cpu_flags))
s->dct32 = ff_dct32_float_sse;
if (EXTERNAL_SSE2(has_vectors))
if (EXTERNAL_SSE2(cpu_flags))
s->dct32 = ff_dct32_float_sse2;
if (EXTERNAL_AVX(has_vectors))
if (EXTERNAL_AVX(cpu_flags))
s->dct32 = ff_dct32_float_avx;
}
#endif

+ 6
- 6
libavcodec/x86/fmtconvert_init.c View File

@@ -116,29 +116,29 @@ static void float_interleave_sse(float *dst, const float **src,
av_cold void ff_fmt_convert_init_x86(FmtConvertContext *c, AVCodecContext *avctx)
{
#if HAVE_YASM
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
c->float_interleave = float_interleave_mmx;

if (EXTERNAL_AMD3DNOW(mm_flags)) {
if (EXTERNAL_AMD3DNOW(cpu_flags)) {
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->float_to_int16 = ff_float_to_int16_3dnow;
c->float_to_int16_interleave = float_to_int16_interleave_3dnow;
}
}
if (EXTERNAL_AMD3DNOWEXT(mm_flags)) {
if (EXTERNAL_AMD3DNOWEXT(cpu_flags)) {
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->float_to_int16_interleave = float_to_int16_interleave_3dnowext;
}
}
if (EXTERNAL_SSE(mm_flags)) {
if (EXTERNAL_SSE(cpu_flags)) {
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse;
c->float_to_int16 = ff_float_to_int16_sse;
c->float_to_int16_interleave = float_to_int16_interleave_sse;
c->float_interleave = float_interleave_sse;
}
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_sse2;
c->float_to_int16 = ff_float_to_int16_sse2;
c->float_to_int16_interleave = float_to_int16_interleave_sse2;


+ 11
- 11
libavcodec/x86/h264_intrapred_init.c View File

@@ -185,10 +185,10 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
const int bit_depth,
const int chroma_format_idc)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (bit_depth == 8) {
if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
h->pred16x16[VERT_PRED8x8 ] = ff_pred16x16_vertical_8_mmx;
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_8_mmx;
if (chroma_format_idc == 1) {
@@ -203,7 +203,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
if (chroma_format_idc == 1)
h->pred8x8 [PLANE_PRED8x8] = ff_pred8x8_plane_8_mmx;
if (codec_id == AV_CODEC_ID_SVQ3) {
if (mm_flags & AV_CPU_FLAG_CMOV)
if (cpu_flags & AV_CPU_FLAG_CMOV)
h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_plane_svq3_8_mmx;
} else if (codec_id == AV_CODEC_ID_RV40) {
h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_plane_rv40_8_mmx;
@@ -213,7 +213,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
}
}

if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_8_mmxext;
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_8_mmxext;
if (chroma_format_idc == 1)
@@ -265,11 +265,11 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
}
}

if (EXTERNAL_SSE(mm_flags)) {
if (EXTERNAL_SSE(cpu_flags)) {
h->pred16x16[VERT_PRED8x8] = ff_pred16x16_vertical_8_sse;
}

if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_8_sse2;
h->pred8x8l [DIAG_DOWN_LEFT_PRED ] = ff_pred8x8l_down_left_8_sse2;
h->pred8x8l [DIAG_DOWN_RIGHT_PRED ] = ff_pred8x8l_down_right_8_sse2;
@@ -292,7 +292,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
}
}

if (EXTERNAL_SSSE3(mm_flags)) {
if (EXTERNAL_SSSE3(cpu_flags)) {
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_8_ssse3;
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_8_ssse3;
if (chroma_format_idc == 1)
@@ -323,7 +323,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
}
}
} else if (bit_depth == 10) {
if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
h->pred4x4[DC_PRED ] = ff_pred4x4_dc_10_mmxext;
h->pred4x4[HOR_UP_PRED ] = ff_pred4x4_horizontal_up_10_mmxext;

@@ -339,7 +339,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
h->pred16x16[VERT_PRED8x8 ] = ff_pred16x16_vertical_10_mmxext;
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_10_mmxext;
}
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
h->pred4x4[DIAG_DOWN_LEFT_PRED ] = ff_pred4x4_down_left_10_sse2;
h->pred4x4[DIAG_DOWN_RIGHT_PRED] = ff_pred4x4_down_right_10_sse2;
h->pred4x4[VERT_LEFT_PRED ] = ff_pred4x4_vertical_left_10_sse2;
@@ -371,7 +371,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
h->pred16x16[VERT_PRED8x8 ] = ff_pred16x16_vertical_10_sse2;
h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_10_sse2;
}
if (EXTERNAL_SSSE3(mm_flags)) {
if (EXTERNAL_SSSE3(cpu_flags)) {
h->pred4x4[DIAG_DOWN_RIGHT_PRED] = ff_pred4x4_down_right_10_ssse3;
h->pred4x4[VERT_RIGHT_PRED ] = ff_pred4x4_vertical_right_10_ssse3;
h->pred4x4[HOR_DOWN_PRED ] = ff_pred4x4_horizontal_down_10_ssse3;
@@ -382,7 +382,7 @@ av_cold void ff_h264_pred_init_x86(H264PredContext *h, int codec_id,
h->pred8x8l[VERT_RIGHT_PRED ] = ff_pred8x8l_vertical_right_10_ssse3;
h->pred8x8l[HOR_UP_PRED ] = ff_pred8x8l_horizontal_up_10_ssse3;
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
h->pred4x4[DIAG_DOWN_LEFT_PRED ] = ff_pred4x4_down_left_10_avx;
h->pred4x4[DIAG_DOWN_RIGHT_PRED] = ff_pred4x4_down_right_10_avx;
h->pred4x4[VERT_LEFT_PRED ] = ff_pred4x4_vertical_left_10_avx;


+ 6
- 6
libavcodec/x86/h264_qpel.c View File

@@ -542,9 +542,9 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
{
#if HAVE_YASM
int high_bit_depth = bit_depth > 8;
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
if (!high_bit_depth) {
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmxext, );
@@ -564,8 +564,8 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
}
}

if (EXTERNAL_SSE2(mm_flags)) {
if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW) && !high_bit_depth) {
if (EXTERNAL_SSE2(cpu_flags)) {
if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW) && !high_bit_depth) {
// these functions are slower than mmx on AMD, but faster on Intel
H264_QPEL_FUNCS(0, 0, sse2);
}
@@ -596,7 +596,7 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
}
}

if (EXTERNAL_SSSE3(mm_flags)) {
if (EXTERNAL_SSSE3(cpu_flags)) {
if (!high_bit_depth) {
H264_QPEL_FUNCS(1, 0, ssse3);
H264_QPEL_FUNCS(1, 1, ssse3);
@@ -619,7 +619,7 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
}
}

if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
/* AVX implies 64 byte cache lines without the need to avoid unaligned
* memory accesses that cross the boundary between two cache lines.
* TODO: Port X264_CPU_CACHELINE_32/64 detection from x264 to avoid


+ 8
- 8
libavcodec/x86/h264chroma_init.c View File

@@ -71,45 +71,45 @@ av_cold void ff_h264chroma_init_x86(H264ChromaContext *c, int bit_depth)
{
#if HAVE_YASM
int high_bit_depth = bit_depth > 8;
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_MMX(mm_flags) && !high_bit_depth) {
if (EXTERNAL_MMX(cpu_flags) && !high_bit_depth) {
c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_rnd_mmx;
c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_mmx;
}

if (EXTERNAL_AMD3DNOW(mm_flags) && !high_bit_depth) {
if (EXTERNAL_AMD3DNOW(cpu_flags) && !high_bit_depth) {
c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_3dnow;
c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_3dnow;
}

if (EXTERNAL_MMXEXT(mm_flags) && !high_bit_depth) {
if (EXTERNAL_MMXEXT(cpu_flags) && !high_bit_depth) {
c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_mmxext;
c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_mmxext;
c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_mmxext;
c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_mmxext;
}

if (EXTERNAL_MMXEXT(mm_flags) && bit_depth > 8 && bit_depth <= 10) {
if (EXTERNAL_MMXEXT(cpu_flags) && bit_depth > 8 && bit_depth <= 10) {
c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_10_mmxext;
c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_10_mmxext;
c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_10_mmxext;
c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_10_mmxext;
}

if (EXTERNAL_SSE2(mm_flags) && bit_depth > 8 && bit_depth <= 10) {
if (EXTERNAL_SSE2(cpu_flags) && bit_depth > 8 && bit_depth <= 10) {
c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_sse2;
c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_sse2;
}

if (EXTERNAL_SSSE3(mm_flags) && !high_bit_depth) {
if (EXTERNAL_SSSE3(cpu_flags) && !high_bit_depth) {
c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_rnd_ssse3;
c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_ssse3;
c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_ssse3;
c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_ssse3;
}

if (EXTERNAL_AVX(mm_flags) && bit_depth > 8 && bit_depth <= 10) {
if (EXTERNAL_AVX(cpu_flags) && bit_depth > 8 && bit_depth <= 10) {
// AVX implies !cache64.
// TODO: Port cache(32|64) detection from x264.
c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_avx;


+ 13
- 13
libavcodec/x86/h264dsp_init.c View File

@@ -211,13 +211,13 @@ H264_BIWEIGHT_10_SSE(4, 10)
av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
const int chroma_format_idc)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (chroma_format_idc == 1 && EXTERNAL_MMXEXT(mm_flags))
if (chroma_format_idc == 1 && EXTERNAL_MMXEXT(cpu_flags))
c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmxext;

if (bit_depth == 8) {
if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
c->h264_idct_dc_add =
c->h264_idct_add = ff_h264_idct_add_8_mmx;
c->h264_idct8_dc_add =
@@ -228,10 +228,10 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
if (chroma_format_idc == 1)
c->h264_idct_add8 = ff_h264_idct_add8_8_mmx;
c->h264_idct_add16intra = ff_h264_idct_add16intra_8_mmx;
if (mm_flags & AV_CPU_FLAG_CMOV)
if (cpu_flags & AV_CPU_FLAG_CMOV)
c->h264_luma_dc_dequant_idct = ff_h264_luma_dc_dequant_idct_mmx;

if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->h264_idct_dc_add = ff_h264_idct_dc_add_8_mmxext;
c->h264_idct8_dc_add = ff_h264_idct8_dc_add_8_mmxext;
c->h264_idct_add16 = ff_h264_idct_add16_8_mmxext;
@@ -260,7 +260,7 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_mmxext;
c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_mmxext;

if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
c->h264_idct8_add = ff_h264_idct8_add_8_sse2;

c->h264_idct_add16 = ff_h264_idct_add16_8_sse2;
@@ -281,11 +281,11 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_sse2;
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_8_sse2;
}
if (EXTERNAL_SSSE3(mm_flags)) {
if (EXTERNAL_SSSE3(cpu_flags)) {
c->biweight_h264_pixels_tab[0] = ff_h264_biweight_16_ssse3;
c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_ssse3;
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
c->h264_v_loop_filter_luma = ff_deblock_v_luma_8_avx;
c->h264_h_loop_filter_luma = ff_deblock_h_luma_8_avx;
c->h264_v_loop_filter_luma_intra = ff_deblock_v_luma_intra_8_avx;
@@ -294,8 +294,8 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
}
}
} else if (bit_depth == 10) {
if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
#if ARCH_X86_32
c->h264_v_loop_filter_chroma = ff_deblock_v_chroma_10_mmxext;
c->h264_v_loop_filter_chroma_intra = ff_deblock_v_chroma_intra_10_mmxext;
@@ -305,7 +305,7 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_mmxext;
#endif /* ARCH_X86_32 */
c->h264_idct_dc_add = ff_h264_idct_dc_add_10_mmxext;
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
c->h264_idct_add = ff_h264_idct_add_10_sse2;
c->h264_idct8_dc_add = ff_h264_idct8_dc_add_10_sse2;

@@ -335,7 +335,7 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
c->h264_h_loop_filter_luma_intra = ff_deblock_h_luma_intra_10_sse2;
#endif /* HAVE_ALIGNED_STACK */
}
if (EXTERNAL_SSE4(mm_flags)) {
if (EXTERNAL_SSE4(cpu_flags)) {
c->weight_h264_pixels_tab[0] = ff_h264_weight_16_10_sse4;
c->weight_h264_pixels_tab[1] = ff_h264_weight_8_10_sse4;
c->weight_h264_pixels_tab[2] = ff_h264_weight_4_10_sse4;
@@ -344,7 +344,7 @@ av_cold void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth,
c->biweight_h264_pixels_tab[1] = ff_h264_biweight_8_10_sse4;
c->biweight_h264_pixels_tab[2] = ff_h264_biweight_4_10_sse4;
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
c->h264_idct_dc_add =
c->h264_idct_add = ff_h264_idct_add_10_avx;
c->h264_idct8_dc_add = ff_h264_idct8_dc_add_10_avx;


+ 14
- 14
libavcodec/x86/hpeldsp_init.c View File

@@ -157,7 +157,7 @@ HPELDSP_AVG_PIXELS16(_mmxext)
c->PFX ## _pixels_tab IDX [3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU; \
} while (0)

static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int mm_flags)
static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int cpu_flags)
{
#if HAVE_MMX_INLINE
SET_HPEL_FUNCS(put, [0], 16, mmx);
@@ -170,7 +170,7 @@ static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int mm_flags)
#endif /* HAVE_MMX_INLINE */
}

static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int mm_flags)
static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags)
{
#if HAVE_MMXEXT_EXTERNAL
c->put_pixels_tab[0][1] = ff_put_pixels16_x2_mmxext;
@@ -204,7 +204,7 @@ static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int mm_flags)
#endif /* HAVE_MMXEXT_EXTERNAL */
}

static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int mm_flags)
static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags)
{
#if HAVE_AMD3DNOW_EXTERNAL
c->put_pixels_tab[0][1] = ff_put_pixels16_x2_3dnow;
@@ -238,10 +238,10 @@ static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int mm_flags)
#endif /* HAVE_AMD3DNOW_EXTERNAL */
}

static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int mm_flags)
static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int cpu_flags)
{
#if HAVE_SSE2_EXTERNAL
if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) {
// these functions are slower than mmx on AMD, but faster on Intel
c->put_pixels_tab[0][0] = ff_put_pixels16_sse2;
c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2;
@@ -252,17 +252,17 @@ static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int mm_flags)

void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (mm_flags & AV_CPU_FLAG_MMX)
hpeldsp_init_mmx(c, flags, mm_flags);
if (cpu_flags & AV_CPU_FLAG_MMX)
hpeldsp_init_mmx(c, flags, cpu_flags);

if (mm_flags & AV_CPU_FLAG_MMXEXT)
hpeldsp_init_mmxext(c, flags, mm_flags);
if (cpu_flags & AV_CPU_FLAG_MMXEXT)
hpeldsp_init_mmxext(c, flags, cpu_flags);

if (mm_flags & AV_CPU_FLAG_3DNOW)
hpeldsp_init_3dnow(c, flags, mm_flags);
if (cpu_flags & AV_CPU_FLAG_3DNOW)
hpeldsp_init_3dnow(c, flags, cpu_flags);

if (mm_flags & AV_CPU_FLAG_SSE2)
hpeldsp_init_sse2(c, flags, mm_flags);
if (cpu_flags & AV_CPU_FLAG_SSE2)
hpeldsp_init_sse2(c, flags, cpu_flags);
}

+ 2
- 2
libavcodec/x86/lpc.c View File

@@ -149,9 +149,9 @@ static void lpc_compute_autocorr_sse2(const double *data, int len, int lag,
av_cold void ff_lpc_init_x86(LPCContext *c)
{
#if HAVE_SSE2_INLINE
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (mm_flags & (AV_CPU_FLAG_SSE2|AV_CPU_FLAG_SSE2SLOW)) {
if (cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) {
c->lpc_apply_welch_window = lpc_apply_welch_window_sse2;
c->lpc_compute_autocorr = lpc_compute_autocorr_sse2;
}


+ 4
- 4
libavcodec/x86/motion_est.c View File

@@ -435,9 +435,9 @@ PIX_SAD(mmxext)
av_cold void ff_dsputil_init_pix_mmx(DSPContext *c, AVCodecContext *avctx)
{
#if HAVE_INLINE_ASM
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (mm_flags & AV_CPU_FLAG_MMX) {
if (cpu_flags & AV_CPU_FLAG_MMX) {
c->pix_abs[0][0] = sad16_mmx;
c->pix_abs[0][1] = sad16_x2_mmx;
c->pix_abs[0][2] = sad16_y2_mmx;
@@ -450,7 +450,7 @@ av_cold void ff_dsputil_init_pix_mmx(DSPContext *c, AVCodecContext *avctx)
c->sad[0]= sad16_mmx;
c->sad[1]= sad8_mmx;
}
if (mm_flags & AV_CPU_FLAG_MMXEXT) {
if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
c->pix_abs[0][0] = sad16_mmxext;
c->pix_abs[1][0] = sad8_mmxext;

@@ -466,7 +466,7 @@ av_cold void ff_dsputil_init_pix_mmx(DSPContext *c, AVCodecContext *avctx)
c->pix_abs[1][3] = sad8_xy2_mmxext;
}
}
if ((mm_flags & AV_CPU_FLAG_SSE2) && !(mm_flags & AV_CPU_FLAG_3DNOW)) {
if ((cpu_flags & AV_CPU_FLAG_SSE2) && !(cpu_flags & AV_CPU_FLAG_3DNOW)) {
c->sad[0]= sad16_sse2;
}
#endif /* HAVE_INLINE_ASM */


+ 7
- 7
libavcodec/x86/mpegaudiodsp.c View File

@@ -226,7 +226,7 @@ DECL_IMDCT_BLOCKS(avx,avx)

av_cold void ff_mpadsp_init_x86(MPADSPContext *s)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

int i, j;
for (j = 0; j < 4; j++) {
@@ -243,21 +243,21 @@ av_cold void ff_mpadsp_init_x86(MPADSPContext *s)
}

#if HAVE_SSE2_INLINE
if (mm_flags & AV_CPU_FLAG_SSE2) {
if (cpu_flags & AV_CPU_FLAG_SSE2) {
s->apply_window_float = apply_window_mp3;
}
#endif /* HAVE_SSE2_INLINE */

#if HAVE_YASM
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
s->imdct36_blocks_float = imdct36_blocks_avx;
} else if (EXTERNAL_SSSE3(mm_flags)) {
} else if (EXTERNAL_SSSE3(cpu_flags)) {
s->imdct36_blocks_float = imdct36_blocks_ssse3;
} else if (EXTERNAL_SSE3(mm_flags)) {
} else if (EXTERNAL_SSE3(cpu_flags)) {
s->imdct36_blocks_float = imdct36_blocks_sse3;
} else if (EXTERNAL_SSE2(mm_flags)) {
} else if (EXTERNAL_SSE2(cpu_flags)) {
s->imdct36_blocks_float = imdct36_blocks_sse2;
} else if (EXTERNAL_SSE(mm_flags)) {
} else if (EXTERNAL_SSE(cpu_flags)) {
s->imdct36_blocks_float = imdct36_blocks_sse;
}
#endif /* HAVE_YASM */


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

@@ -557,9 +557,9 @@ static void denoise_dct_sse2(MpegEncContext *s, int16_t *block){
av_cold void ff_MPV_common_init_x86(MpegEncContext *s)
{
#if HAVE_INLINE_ASM
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (mm_flags & AV_CPU_FLAG_MMX) {
if (cpu_flags & AV_CPU_FLAG_MMX) {
s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_mmx;
s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_mmx;
s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_mmx;
@@ -568,7 +568,7 @@ av_cold void ff_MPV_common_init_x86(MpegEncContext *s)
s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_mmx;
s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_mmx;

if (mm_flags & AV_CPU_FLAG_SSE2) {
if (cpu_flags & AV_CPU_FLAG_SSE2) {
s->denoise_dct= denoise_dct_sse2;
} else {
s->denoise_dct= denoise_dct_mmx;


+ 5
- 5
libavcodec/x86/mpegvideoenc.c View File

@@ -87,20 +87,20 @@ av_cold void ff_MPV_encode_init_x86(MpegEncContext *s)

if (dct_algo == FF_DCT_AUTO || dct_algo == FF_DCT_MMX) {
#if HAVE_MMX_INLINE
int mm_flags = av_get_cpu_flags();
if (INLINE_MMX(mm_flags))
int cpu_flags = av_get_cpu_flags();
if (INLINE_MMX(cpu_flags))
s->dct_quantize = dct_quantize_MMX;
#endif
#if HAVE_MMXEXT_INLINE
if (INLINE_MMXEXT(mm_flags))
if (INLINE_MMXEXT(cpu_flags))
s->dct_quantize = dct_quantize_MMXEXT;
#endif
#if HAVE_SSE2_INLINE
if (INLINE_SSE2(mm_flags))
if (INLINE_SSE2(cpu_flags))
s->dct_quantize = dct_quantize_SSE2;
#endif
#if HAVE_SSSE3_INLINE
if (INLINE_SSSE3(mm_flags))
if (INLINE_SSSE3(cpu_flags))
s->dct_quantize = dct_quantize_SSSE3;
#endif
}


+ 5
- 5
libavcodec/x86/pngdsp_init.c View File

@@ -35,16 +35,16 @@ void ff_add_bytes_l2_sse2(uint8_t *dst, uint8_t *src1,

av_cold void ff_pngdsp_init_x86(PNGDSPContext *dsp)
{
int flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

#if ARCH_X86_32
if (EXTERNAL_MMX(flags))
if (EXTERNAL_MMX(cpu_flags))
dsp->add_bytes_l2 = ff_add_bytes_l2_mmx;
#endif
if (EXTERNAL_MMXEXT(flags))
if (EXTERNAL_MMXEXT(cpu_flags))
dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmxext;
if (EXTERNAL_SSE2(flags))
if (EXTERNAL_SSE2(cpu_flags))
dsp->add_bytes_l2 = ff_add_bytes_l2_sse2;
if (EXTERNAL_SSSE3(flags))
if (EXTERNAL_SSSE3(cpu_flags))
dsp->add_paeth_prediction = ff_add_png_paeth_prediction_ssse3;
}

+ 4
- 4
libavcodec/x86/proresdsp_init.c View File

@@ -35,19 +35,19 @@ void ff_prores_idct_put_10_avx (uint16_t *dst, int linesize,
av_cold void ff_proresdsp_x86_init(ProresDSPContext *dsp)
{
#if ARCH_X86_64
int flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_SSE2(flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
dsp->idct_put = ff_prores_idct_put_10_sse2;
}

if (EXTERNAL_SSE4(flags)) {
if (EXTERNAL_SSE4(cpu_flags)) {
dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
dsp->idct_put = ff_prores_idct_put_10_sse4;
}

if (EXTERNAL_AVX(flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
dsp->idct_permutation_type = FF_TRANSPOSE_IDCT_PERM;
dsp->idct_put = ff_prores_idct_put_10_avx;
}


+ 4
- 4
libavcodec/x86/rv34dsp_init.c View File

@@ -32,14 +32,14 @@ void ff_rv34_idct_add_mmxext(uint8_t *dst, ptrdiff_t stride, int16_t *block);

av_cold void ff_rv34dsp_init_x86(RV34DSPContext* c)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_MMX(mm_flags))
if (EXTERNAL_MMX(cpu_flags))
c->rv34_idct_dc_add = ff_rv34_idct_dc_add_mmx;
if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->rv34_inv_transform_dc = ff_rv34_idct_dc_noround_mmxext;
c->rv34_idct_add = ff_rv34_idct_add_mmxext;
}
if (EXTERNAL_SSE4(mm_flags))
if (EXTERNAL_SSE4(cpu_flags))
c->rv34_idct_dc_add = ff_rv34_idct_dc_add_sse4;
}

+ 6
- 6
libavcodec/x86/rv40dsp_init.c View File

@@ -214,9 +214,9 @@ static void avg_rv40_qpel16_mc33_mmx(uint8_t *dst, uint8_t *src,
av_cold void ff_rv40dsp_init_x86(RV34DSPContext *c)
{
#if HAVE_YASM
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
c->put_chroma_pixels_tab[0] = ff_put_rv40_chroma_mc8_mmx;
c->put_chroma_pixels_tab[1] = ff_put_rv40_chroma_mc4_mmx;
#if HAVE_MMX_INLINE
@@ -229,7 +229,7 @@ av_cold void ff_rv40dsp_init_x86(RV34DSPContext *c)
QPEL_MC_SET(put_, _mmx)
#endif
}
if (EXTERNAL_MMXEXT(mm_flags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_mmxext;
c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_mmxext;
c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_mmxext;
@@ -239,14 +239,14 @@ av_cold void ff_rv40dsp_init_x86(RV34DSPContext *c)
#if ARCH_X86_32
QPEL_MC_SET(avg_, _mmxext)
#endif
} else if (EXTERNAL_AMD3DNOW(mm_flags)) {
} else if (EXTERNAL_AMD3DNOW(cpu_flags)) {
c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_3dnow;
c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_3dnow;
#if ARCH_X86_32
QPEL_MC_SET(avg_, _3dnow)
#endif
}
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_sse2;
c->rv40_weight_pixels_tab[0][1] = ff_rv40_weight_func_rnd_8_sse2;
c->rv40_weight_pixels_tab[1][0] = ff_rv40_weight_func_nornd_16_sse2;
@@ -254,7 +254,7 @@ av_cold void ff_rv40dsp_init_x86(RV34DSPContext *c)
QPEL_MC_SET(put_, _sse2)
QPEL_MC_SET(avg_, _sse2)
}
if (EXTERNAL_SSSE3(mm_flags)) {
if (EXTERNAL_SSSE3(cpu_flags)) {
c->rv40_weight_pixels_tab[0][0] = ff_rv40_weight_func_rnd_16_ssse3;
c->rv40_weight_pixels_tab[0][1] = ff_rv40_weight_func_rnd_8_ssse3;
c->rv40_weight_pixels_tab[1][0] = ff_rv40_weight_func_nornd_16_ssse3;


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

@@ -39,9 +39,9 @@ void ff_sbr_qmf_pre_shuffle_sse2(float *z);

av_cold void ff_sbrdsp_init_x86(SBRDSPContext *s)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_SSE(mm_flags)) {
if (EXTERNAL_SSE(cpu_flags)) {
s->neg_odd_64 = ff_sbr_neg_odd_64_sse;
s->sum_square = ff_sbr_sum_square_sse;
s->sum64x5 = ff_sbr_sum64x5_sse;
@@ -50,7 +50,7 @@ av_cold void ff_sbrdsp_init_x86(SBRDSPContext *s)
s->qmf_post_shuffle = ff_sbr_qmf_post_shuffle_sse;
}

if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
s->qmf_deint_bfly = ff_sbr_qmf_deint_bfly_sse2;
s->qmf_pre_shuffle = ff_sbr_qmf_pre_shuffle_sse2;
}


+ 9
- 9
libavcodec/x86/vc1dsp_init.c View File

@@ -83,12 +83,12 @@ void ff_avg_vc1_chroma_mc8_nornd_ssse3(uint8_t *dst, uint8_t *src,

av_cold void ff_vc1dsp_init_x86(VC1DSPContext *dsp)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (INLINE_MMX(mm_flags))
if (INLINE_MMX(cpu_flags))
ff_vc1dsp_init_mmx(dsp);

if (INLINE_MMXEXT(mm_flags))
if (INLINE_MMXEXT(cpu_flags))
ff_vc1dsp_init_mmxext(dsp);

#define ASSIGN_LF(EXT) \
@@ -100,31 +100,31 @@ av_cold void ff_vc1dsp_init_x86(VC1DSPContext *dsp)
dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_ ## EXT

#if HAVE_YASM
if (mm_flags & AV_CPU_FLAG_MMX) {
if (cpu_flags & AV_CPU_FLAG_MMX) {
dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = ff_put_vc1_chroma_mc8_nornd_mmx;
}

if (mm_flags & AV_CPU_FLAG_MMXEXT) {
if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
ASSIGN_LF(mmxext);
dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_mmxext;

dsp->avg_vc1_mspel_pixels_tab[0] = avg_vc1_mspel_mc00_mmxext;
} else if (mm_flags & AV_CPU_FLAG_3DNOW) {
} else if (cpu_flags & AV_CPU_FLAG_3DNOW) {
dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_3dnow;
}

if (mm_flags & AV_CPU_FLAG_SSE2) {
if (cpu_flags & AV_CPU_FLAG_SSE2) {
dsp->vc1_v_loop_filter8 = ff_vc1_v_loop_filter8_sse2;
dsp->vc1_h_loop_filter8 = ff_vc1_h_loop_filter8_sse2;
dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_sse2;
dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_sse2;
}
if (mm_flags & AV_CPU_FLAG_SSSE3) {
if (cpu_flags & AV_CPU_FLAG_SSSE3) {
ASSIGN_LF(ssse3);
dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = ff_put_vc1_chroma_mc8_nornd_ssse3;
dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_ssse3;
}
if (mm_flags & AV_CPU_FLAG_SSE4) {
if (cpu_flags & AV_CPU_FLAG_SSE4) {
dsp->vc1_h_loop_filter8 = ff_vc1_h_loop_filter8_sse4;
dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_sse4;
}


+ 5
- 5
libavcodec/x86/videodsp_init.c View File

@@ -100,20 +100,20 @@ void ff_prefetch_3dnow(uint8_t *buf, ptrdiff_t stride, int h);
av_cold void ff_videodsp_init_x86(VideoDSPContext *ctx, int bpc)
{
#if HAVE_YASM
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

#if ARCH_X86_32
if (bpc <= 8 && mm_flags & AV_CPU_FLAG_MMX) {
if (bpc <= 8 && cpu_flags & AV_CPU_FLAG_MMX) {
ctx->emulated_edge_mc = emulated_edge_mc_mmx;
}
if (mm_flags & AV_CPU_FLAG_3DNOW) {
if (cpu_flags & AV_CPU_FLAG_3DNOW) {
ctx->prefetch = ff_prefetch_3dnow;
}
#endif /* ARCH_X86_32 */
if (mm_flags & AV_CPU_FLAG_MMXEXT) {
if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
ctx->prefetch = ff_prefetch_mmxext;
}
if (bpc <= 8 && mm_flags & AV_CPU_FLAG_SSE) {
if (bpc <= 8 && cpu_flags & AV_CPU_FLAG_SSE) {
ctx->emulated_edge_mc = emulated_edge_mc_sse;
}
#endif /* HAVE_YASM */


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

@@ -31,13 +31,13 @@ void ff_vorbis_inverse_coupling_sse(float *mag, float *ang,
av_cold void ff_vorbisdsp_init_x86(VorbisDSPContext *dsp)
{
#if HAVE_YASM
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

#if ARCH_X86_32
if (mm_flags & AV_CPU_FLAG_3DNOW)
if (cpu_flags & AV_CPU_FLAG_3DNOW)
dsp->vorbis_inverse_coupling = ff_vorbis_inverse_coupling_3dnow;
#endif /* ARCH_X86_32 */
if (mm_flags & AV_CPU_FLAG_SSE)
if (cpu_flags & AV_CPU_FLAG_SSE)
dsp->vorbis_inverse_coupling = ff_vorbis_inverse_coupling_sse;
#endif /* HAVE_YASM */
}

+ 4
- 4
libavcodec/x86/vp3dsp_init.c View File

@@ -42,16 +42,16 @@ void ff_vp3_h_loop_filter_mmxext(uint8_t *src, int stride,

av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags)
{
int cpuflags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

#if ARCH_X86_32
if (EXTERNAL_MMX(cpuflags)) {
if (EXTERNAL_MMX(cpu_flags)) {
c->idct_put = ff_vp3_idct_put_mmx;
c->idct_add = ff_vp3_idct_add_mmx;
}
#endif

if (EXTERNAL_MMXEXT(cpuflags)) {
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->idct_dc_add = ff_vp3_idct_dc_add_mmxext;

if (!(flags & CODEC_FLAG_BITEXACT)) {
@@ -60,7 +60,7 @@ av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags)
}
}

if (EXTERNAL_SSE2(cpuflags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
c->idct_put = ff_vp3_idct_put_sse2;
c->idct_add = ff_vp3_idct_add_sse2;
}


+ 3
- 3
libavcodec/x86/vp56dsp_init.c View File

@@ -32,16 +32,16 @@ void ff_vp6_filter_diag4_sse2(uint8_t *dst, uint8_t *src, int stride,

av_cold void ff_vp56dsp_init_x86(VP56DSPContext* c, enum AVCodecID codec)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (CONFIG_VP6_DECODER && codec == AV_CODEC_ID_VP6) {
#if ARCH_X86_32
if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
c->vp6_filter_diag4 = ff_vp6_filter_diag4_mmx;
}
#endif

if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
c->vp6_filter_diag4 = ff_vp6_filter_diag4_sse2;
}
}


+ 8
- 8
libavcodec/x86/vp8dsp_init.c View File

@@ -316,9 +316,9 @@ DECLARE_LOOP_FILTER(sse4)
av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
{
#if HAVE_YASM
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (mm_flags & AV_CPU_FLAG_MMX) {
if (cpu_flags & AV_CPU_FLAG_MMX) {
c->vp8_idct_dc_add = ff_vp8_idct_dc_add_mmx;
c->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_mmx;
#if ARCH_X86_32
@@ -349,7 +349,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)

/* 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. */
if (mm_flags & AV_CPU_FLAG_MMXEXT) {
if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
VP8_MC_FUNC(2, 4, mmxext);
VP8_BILINEAR_MC_FUNC(2, 4, mmxext);
#if ARCH_X86_32
@@ -373,14 +373,14 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
#endif
}

if (mm_flags & AV_CPU_FLAG_SSE) {
if (cpu_flags & AV_CPU_FLAG_SSE) {
c->vp8_idct_add = ff_vp8_idct_add_sse;
c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse;
c->put_vp8_epel_pixels_tab[0][0][0] =
c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse;
}

if (mm_flags & (AV_CPU_FLAG_SSE2|AV_CPU_FLAG_SSE2SLOW)) {
if (cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) {
VP8_LUMA_MC_FUNC(0, 16, sse2);
VP8_MC_FUNC(1, 8, sse2);
VP8_BILINEAR_MC_FUNC(0, 16, sse2);
@@ -395,7 +395,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
c->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_mbedge_sse2;
}

if (mm_flags & AV_CPU_FLAG_SSE2) {
if (cpu_flags & AV_CPU_FLAG_SSE2) {
c->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_sse2;

c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse2;
@@ -407,7 +407,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_sse2;
}

if (mm_flags & AV_CPU_FLAG_SSSE3) {
if (cpu_flags & AV_CPU_FLAG_SSSE3) {
VP8_LUMA_MC_FUNC(0, 16, ssse3);
VP8_MC_FUNC(1, 8, ssse3);
VP8_MC_FUNC(2, 4, ssse3);
@@ -429,7 +429,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext* c)
c->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_mbedge_ssse3;
}

if (mm_flags & AV_CPU_FLAG_SSE4) {
if (cpu_flags & AV_CPU_FLAG_SSE4) {
c->vp8_idct_dc_add = ff_vp8_idct_dc_add_sse4;

c->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter_simple_sse4;


+ 5
- 5
libavfilter/x86/af_volume_init.c View File

@@ -35,24 +35,24 @@ void ff_scale_samples_s32_avx(uint8_t *dst, const uint8_t *src, int len,

av_cold void ff_volume_init_x86(VolumeContext *vol)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();
enum AVSampleFormat sample_fmt = av_get_packed_sample_fmt(vol->sample_fmt);

if (sample_fmt == AV_SAMPLE_FMT_S16) {
if (EXTERNAL_SSE2(mm_flags) && vol->volume_i < 32768) {
if (EXTERNAL_SSE2(cpu_flags) && vol->volume_i < 32768) {
vol->scale_samples = ff_scale_samples_s16_sse2;
vol->samples_align = 8;
}
} else if (sample_fmt == AV_SAMPLE_FMT_S32) {
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
vol->scale_samples = ff_scale_samples_s32_sse2;
vol->samples_align = 4;
}
if (EXTERNAL_SSSE3(mm_flags) && mm_flags & AV_CPU_FLAG_ATOM) {
if (EXTERNAL_SSSE3(cpu_flags) && cpu_flags & AV_CPU_FLAG_ATOM) {
vol->scale_samples = ff_scale_samples_s32_ssse3_atom;
vol->samples_align = 4;
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
vol->scale_samples = ff_scale_samples_s32_avx;
vol->samples_align = 8;
}


+ 8
- 8
libavresample/x86/audio_convert_init.c View File

@@ -145,15 +145,15 @@ void ff_conv_flt_to_fltp_6ch_avx (float *const *dst, float *src, int len,

av_cold void ff_audio_convert_init_x86(AudioConvert *ac)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_MMX(mm_flags)) {
if (EXTERNAL_MMX(cpu_flags)) {
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S32,
0, 1, 8, "MMX", ff_conv_s32_to_s16_mmx);
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
6, 1, 4, "MMX", ff_conv_fltp_to_flt_6ch_mmx);
}
if (EXTERNAL_SSE(mm_flags)) {
if (EXTERNAL_SSE(cpu_flags)) {
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
6, 1, 2, "SSE", ff_conv_fltp_to_s16_6ch_sse);
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
@@ -161,8 +161,8 @@ av_cold void ff_audio_convert_init_x86(AudioConvert *ac)
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_FLT,
2, 16, 4, "SSE", ff_conv_flt_to_fltp_2ch_sse);
}
if (EXTERNAL_SSE2(mm_flags)) {
if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
if (EXTERNAL_SSE2(cpu_flags)) {
if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) {
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S32,
0, 16, 16, "SSE2", ff_conv_s32_to_s16_sse2);
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P,
@@ -206,7 +206,7 @@ av_cold void ff_audio_convert_init_x86(AudioConvert *ac)
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_FLT,
6, 16, 4, "SSE2", ff_conv_flt_to_fltp_6ch_sse2);
}
if (EXTERNAL_SSSE3(mm_flags)) {
if (EXTERNAL_SSSE3(cpu_flags)) {
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16P,
6, 16, 4, "SSSE3", ff_conv_s16p_to_flt_6ch_ssse3);
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLTP,
@@ -220,13 +220,13 @@ av_cold void ff_audio_convert_init_x86(AudioConvert *ac)
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_FLT,
6, 16, 4, "SSSE3", ff_conv_flt_to_s16p_6ch_ssse3);
}
if (EXTERNAL_SSE4(mm_flags)) {
if (EXTERNAL_SSE4(cpu_flags)) {
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16,
0, 16, 8, "SSE4", ff_conv_s16_to_flt_sse4);
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLTP,
6, 16, 4, "SSE4", ff_conv_fltp_to_flt_6ch_sse4);
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S32,
0, 32, 16, "AVX", ff_conv_s32_to_flt_avx);
ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_FLT,


+ 10
- 10
libavresample/x86/audio_mix_init.c View File

@@ -106,7 +106,7 @@ DEFINE_MIX_3_8_TO_1_2(7)
DEFINE_MIX_3_8_TO_1_2(8)

#define SET_MIX_3_8_TO_1_2(chan) \
if (EXTERNAL_SSE(mm_flags)) { \
if (EXTERNAL_SSE(cpu_flags)) { \
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT,\
chan, 1, 16, 4, "SSE", \
ff_mix_ ## chan ## _to_1_fltp_flt_sse); \
@@ -114,7 +114,7 @@ DEFINE_MIX_3_8_TO_1_2(8)
chan, 2, 16, 4, "SSE", \
ff_mix_## chan ##_to_2_fltp_flt_sse); \
} \
if (EXTERNAL_SSE2(mm_flags)) { \
if (EXTERNAL_SSE2(cpu_flags)) { \
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_FLT,\
chan, 1, 16, 8, "SSE2", \
ff_mix_ ## chan ## _to_1_s16p_flt_sse2); \
@@ -122,7 +122,7 @@ DEFINE_MIX_3_8_TO_1_2(8)
chan, 2, 16, 8, "SSE2", \
ff_mix_ ## chan ## _to_2_s16p_flt_sse2); \
} \
if (EXTERNAL_SSE4(mm_flags)) { \
if (EXTERNAL_SSE4(cpu_flags)) { \
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_FLT,\
chan, 1, 16, 8, "SSE4", \
ff_mix_ ## chan ## _to_1_s16p_flt_sse4); \
@@ -130,7 +130,7 @@ DEFINE_MIX_3_8_TO_1_2(8)
chan, 2, 16, 8, "SSE4", \
ff_mix_ ## chan ## _to_2_s16p_flt_sse4); \
} \
if (EXTERNAL_AVX(mm_flags)) { \
if (EXTERNAL_AVX(cpu_flags)) { \
int ptr_align = 32; \
int smp_align = 8; \
if (ARCH_X86_32 || chan >= 6) { \
@@ -150,7 +150,7 @@ DEFINE_MIX_3_8_TO_1_2(8)
chan, 2, 16, 8, "AVX", \
ff_mix_ ## chan ## _to_2_s16p_flt_avx); \
} \
if (EXTERNAL_FMA4(mm_flags)) { \
if (EXTERNAL_FMA4(cpu_flags)) { \
int ptr_align = 32; \
int smp_align = 8; \
if (ARCH_X86_32 || chan >= 6) { \
@@ -174,15 +174,15 @@ DEFINE_MIX_3_8_TO_1_2(8)
av_cold void ff_audio_mix_init_x86(AudioMix *am)
{
#if HAVE_YASM
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_SSE(mm_flags)) {
if (EXTERNAL_SSE(cpu_flags)) {
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT,
2, 1, 16, 8, "SSE", ff_mix_2_to_1_fltp_flt_sse);
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT,
1, 2, 16, 4, "SSE", ff_mix_1_to_2_fltp_flt_sse);
}
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_FLT,
2, 1, 16, 8, "SSE2", ff_mix_2_to_1_s16p_flt_sse2);
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_Q8,
@@ -190,13 +190,13 @@ av_cold void ff_audio_mix_init_x86(AudioMix *am)
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_FLT,
1, 2, 16, 8, "SSE2", ff_mix_1_to_2_s16p_flt_sse2);
}
if (EXTERNAL_SSE4(mm_flags)) {
if (EXTERNAL_SSE4(cpu_flags)) {
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_FLT,
2, 1, 16, 8, "SSE4", ff_mix_2_to_1_s16p_flt_sse4);
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_S16P, AV_MIX_COEFF_TYPE_FLT,
1, 2, 16, 8, "SSE4", ff_mix_1_to_2_s16p_flt_sse4);
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT,
2, 1, 32, 16, "AVX", ff_mix_2_to_1_fltp_flt_avx);
ff_audio_mix_set_func(am, AV_SAMPLE_FMT_FLTP, AV_MIX_COEFF_TYPE_FLT,


+ 6
- 6
libavresample/x86/dither_init.c View File

@@ -34,26 +34,26 @@ void ff_dither_int_to_float_triangular_avx(float *dst, int *src0, int len);
av_cold void ff_dither_init_x86(DitherDSPContext *ddsp,
enum AVResampleDitherMethod method)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
ddsp->quantize = ff_quantize_sse2;
ddsp->ptr_align = 16;
ddsp->samples_align = 8;
}

if (method == AV_RESAMPLE_DITHER_RECTANGULAR) {
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
ddsp->dither_int_to_float = ff_dither_int_to_float_rectangular_sse2;
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
ddsp->dither_int_to_float = ff_dither_int_to_float_rectangular_avx;
}
} else {
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
ddsp->dither_int_to_float = ff_dither_int_to_float_triangular_sse2;
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
ddsp->dither_int_to_float = ff_dither_int_to_float_triangular_avx;
}
}


+ 1
- 3
libavutil/ppc/float_dsp_init.c View File

@@ -27,9 +27,7 @@
av_cold void ff_float_dsp_init_ppc(AVFloatDSPContext *fdsp, int bit_exact)
{
#if HAVE_ALTIVEC
int mm_flags = av_get_cpu_flags();

if (!(mm_flags & AV_CPU_FLAG_ALTIVEC))
if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
return;

fdsp->vector_fmul = ff_vector_fmul_altivec;


+ 6
- 6
libavutil/x86/float_dsp_init.c View File

@@ -124,17 +124,17 @@ static void vector_fmul_window_sse(float *dst, const float *src0,

av_cold void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp)
{
int mm_flags = av_get_cpu_flags();
int cpu_flags = av_get_cpu_flags();

#if HAVE_6REGS && HAVE_INLINE_ASM
if (INLINE_AMD3DNOWEXT(mm_flags)) {
if (INLINE_AMD3DNOWEXT(cpu_flags)) {
fdsp->vector_fmul_window = vector_fmul_window_3dnowext;
}
if (INLINE_SSE(mm_flags)) {
if (INLINE_SSE(cpu_flags)) {
fdsp->vector_fmul_window = vector_fmul_window_sse;
}
#endif
if (EXTERNAL_SSE(mm_flags)) {
if (EXTERNAL_SSE(cpu_flags)) {
fdsp->vector_fmul = ff_vector_fmul_sse;
fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_sse;
fdsp->vector_fmul_scalar = ff_vector_fmul_scalar_sse;
@@ -143,10 +143,10 @@ av_cold void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp)
fdsp->scalarproduct_float = ff_scalarproduct_float_sse;
fdsp->butterflies_float = ff_butterflies_float_sse;
}
if (EXTERNAL_SSE2(mm_flags)) {
if (EXTERNAL_SSE2(cpu_flags)) {
fdsp->vector_dmul_scalar = ff_vector_dmul_scalar_sse2;
}
if (EXTERNAL_AVX(mm_flags)) {
if (EXTERNAL_AVX(cpu_flags)) {
fdsp->vector_fmul = ff_vector_fmul_avx;
fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_avx;
fdsp->vector_dmul_scalar = ff_vector_dmul_scalar_avx;


Write Preview
Loading…
Cancel
Save