generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

x86/swr: use lavu helper macros to check CPU extensions

Browse Source 
Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
James Almer 2014-07-02 22:16:50 -03:00 committed by Michael Niedermayer
parent 4c145b692c
commit 9937362c54
3 changed files with 25 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
libswresample/x86/audio_convert_init.c
View File

@ -18,6 +18,7 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/x86/cpu.h"
#include "libswresample/swresample_internal.h"
#include "libswresample/audioconvert.h"
@ -41,24 +42,24 @@ av_cold void swri_audio_convert_init_x86(struct AudioConvert *ac,
//FIXME add memcpy case
#define MULTI_CAPS_FUNC(flag, cap) \
if (mm_flags & flag) {\
if (EXTERNAL_##flag(mm_flags)) {\
if( out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S16 || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_S16P)\
ac->simd_f = ff_int16_to_int32_a_ ## cap;\
if( out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S32P)\
ac->simd_f = ff_int32_to_int16_a_ ## cap;\
}
MULTI_CAPS_FUNC(AV_CPU_FLAG_MMX, mmx)
MULTI_CAPS_FUNC(AV_CPU_FLAG_SSE2, sse2)
MULTI_CAPS_FUNC(MMX, mmx)
MULTI_CAPS_FUNC(SSE2, sse2)
if(mm_flags & AV_CPU_FLAG_MMX) {
if(EXTERNAL_MMX(mm_flags)) {
if(channels == 6) {
if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P)
ac->simd_f = ff_pack_6ch_float_to_float_a_mmx;
}
}
if(mm_flags & AV_CPU_FLAG_SSE2) {
if(EXTERNAL_SSE2(mm_flags)) {
if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32P)
ac->simd_f = ff_int32_to_float_a_sse2;
if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S16 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S16P)
@ -105,7 +106,7 @@ MULTI_CAPS_FUNC(AV_CPU_FLAG_SSE2, sse2)
ac->simd_f = ff_unpack_2ch_float_to_int16_a_sse2;
}
}
if(mm_flags & AV_CPU_FLAG_SSSE3) {
if(EXTERNAL_SSSE3(mm_flags)) {
if(channels == 2) {
if( out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_S16)
ac->simd_f = ff_unpack_2ch_int16_to_int16_a_ssse3;
@ -115,7 +116,7 @@ MULTI_CAPS_FUNC(AV_CPU_FLAG_SSE2, sse2)
ac->simd_f = ff_unpack_2ch_int16_to_float_a_ssse3;
}
}
if(mm_flags & AV_CPU_FLAG_SSE4) {
if(EXTERNAL_SSE4(mm_flags)) {
if(channels == 6) {
if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_FLTP || out_fmt == AV_SAMPLE_FMT_S32 && in_fmt == AV_SAMPLE_FMT_S32P)
ac->simd_f = ff_pack_6ch_float_to_float_a_sse4;
@ -125,7 +126,7 @@ MULTI_CAPS_FUNC(AV_CPU_FLAG_SSE2, sse2)
ac->simd_f = ff_pack_6ch_float_to_int32_a_sse4;
}
}
if(HAVE_AVX_EXTERNAL && mm_flags & AV_CPU_FLAG_AVX) {
if(EXTERNAL_AVX(mm_flags)) {
if( out_fmt == AV_SAMPLE_FMT_FLT && in_fmt == AV_SAMPLE_FMT_S32 || out_fmt == AV_SAMPLE_FMT_FLTP && in_fmt == AV_SAMPLE_FMT_S32P)
ac->simd_f = ff_int32_to_float_a_avx;
if(channels == 6) {

11
libswresample/x86/rematrix_init.c
View File

@ -18,6 +18,7 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/x86/cpu.h"
#include "libswresample/swresample_internal.h"
#define D(type, simd) \
@ -30,6 +31,7 @@ D(int16, mmx)
D(int16, sse2)
av_cold void swri_rematrix_init_x86(struct SwrContext *s){
#if HAVE_YASM
int mm_flags = av_get_cpu_flags();
int nb_in = av_get_channel_layout_nb_channels(s->in_ch_layout);
int nb_out = av_get_channel_layout_nb_channels(s->out_ch_layout);
@ -40,11 +42,11 @@ av_cold void swri_rematrix_init_x86(struct SwrContext *s){
s->mix_2_1_simd = NULL;
if (s->midbuf.fmt == AV_SAMPLE_FMT_S16P){
if(mm_flags & AV_CPU_FLAG_MMX) {
if(EXTERNAL_MMX(mm_flags)) {
s->mix_1_1_simd = ff_mix_1_1_a_int16_mmx;
s->mix_2_1_simd = ff_mix_2_1_a_int16_mmx;
}
if(mm_flags & AV_CPU_FLAG_SSE2) {
if(EXTERNAL_SSE2(mm_flags)) {
s->mix_1_1_simd = ff_mix_1_1_a_int16_sse2;
s->mix_2_1_simd = ff_mix_2_1_a_int16_sse2;
}
@ -64,11 +66,11 @@ av_cold void swri_rematrix_init_x86(struct SwrContext *s){
((int16_t*)s->native_simd_one)[1] = 14;
((int16_t*)s->native_simd_one)[0] = 16384;
} else if(s->midbuf.fmt == AV_SAMPLE_FMT_FLTP){
if(mm_flags & AV_CPU_FLAG_SSE) {
if(EXTERNAL_SSE(mm_flags)) {
s->mix_1_1_simd = ff_mix_1_1_a_float_sse;
s->mix_2_1_simd = ff_mix_2_1_a_float_sse;
}
if(HAVE_AVX_EXTERNAL && mm_flags & AV_CPU_FLAG_AVX) {
if(EXTERNAL_AVX(mm_flags)) {
s->mix_1_1_simd = ff_mix_1_1_a_float_avx;
s->mix_2_1_simd = ff_mix_2_1_a_float_avx;
}
@ -77,4 +79,5 @@ av_cold void swri_rematrix_init_x86(struct SwrContext *s){
s->native_simd_one = av_mallocz(sizeof(float));
memcpy(s->native_simd_one, s->native_one, sizeof(float));
}
#endif
}

17
libswresample/x86/resample_init.c
View File

@ -25,6 +25,7 @@
* @author Michael Niedermayer <michaelni@gmx.at>
*/
#include "libavutil/x86/cpu.h"
#include "libswresample/resample.h"
#define RESAMPLE_FUNCS(type, opt) \
@ -48,39 +49,39 @@ void swri_resample_dsp_x86_init(ResampleContext *c)
switch(c->format){
case AV_SAMPLE_FMT_S16P:
if (ARCH_X86_32 && HAVE_MMXEXT_EXTERNAL && mm_flags & AV_CPU_FLAG_MMX2) {
if (ARCH_X86_32 && EXTERNAL_MMXEXT(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_int16_mmxext
: ff_resample_common_int16_mmxext;
}
if (HAVE_SSE2_EXTERNAL && mm_flags & AV_CPU_FLAG_SSE2) {
if (EXTERNAL_SSE2(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_int16_sse2
: ff_resample_common_int16_sse2;
}
if (HAVE_XOP_EXTERNAL && mm_flags & AV_CPU_FLAG_XOP) {
if (EXTERNAL_XOP(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_int16_xop
: ff_resample_common_int16_xop;
}
break;
case AV_SAMPLE_FMT_FLTP:
if (HAVE_SSE_EXTERNAL && mm_flags & AV_CPU_FLAG_SSE) {
if (EXTERNAL_SSE(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_float_sse
: ff_resample_common_float_sse;
}
if (HAVE_AVX_EXTERNAL && mm_flags & AV_CPU_FLAG_AVX) {
if (EXTERNAL_AVX(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_float_avx
: ff_resample_common_float_avx;
}
if (HAVE_FMA3_EXTERNAL && mm_flags & AV_CPU_FLAG_FMA3) {
if (EXTERNAL_FMA3(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_float_fma3
: ff_resample_common_float_fma3;
}
if (HAVE_FMA4_EXTERNAL && mm_flags & AV_CPU_FLAG_FMA4) {
if (EXTERNAL_FMA4(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_float_fma4
: ff_resample_common_float_fma4;
}
break;
case AV_SAMPLE_FMT_DBLP:
if (HAVE_SSE2_EXTERNAL && mm_flags & AV_CPU_FLAG_SSE2) {
if (EXTERNAL_SSE2(mm_flags)) {
c->dsp.resample = c->linear ? ff_resample_linear_double_sse2
: ff_resample_common_double_sse2;
}