diff options
| author | Michael Niedermayer <michaelni@gmx.at> | 2012-04-28 17:04:42 +0200 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2012-04-28 17:06:11 +0200 |
| commit | 832c3b10d239582d332cc949ef17142cec86d843 (patch) | |
| tree | 8caa875c535edb9ddb82357465fdec3be2871f07 | |
| parent | 95057b197268eaa45a96923d9a13e4181b2af809 (diff) | |
| download | ffmpeg-832c3b10d239582d332cc949ef17142cec86d843.tar.gz | |
swr: add int32_to_float_sse2
could be done for sse/3dnow too if someone wants
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
| -rw-r--r-- | libswresample/x86/audio_convert.asm | 44 | ||||
| -rw-r--r-- | libswresample/x86/swresample_x86.c | 7 |
2 files changed, 51 insertions, 0 deletions
diff --git a/libswresample/x86/audio_convert.asm b/libswresample/x86/audio_convert.asm index 59de2d48fa..67f63092f4 100644 --- a/libswresample/x86/audio_convert.asm +++ b/libswresample/x86/audio_convert.asm @@ -21,6 +21,10 @@ %include "libavutil/x86/x86inc.asm" %include "libavutil/x86/x86util.asm" +SECTION_RODATA + +flt2pm31: times 8 dd 4.6566129e-10 + SECTION .text %macro INT16_TO_INT32 1 @@ -55,6 +59,42 @@ int16_to_int32_u_int %+ SUFFIX REP_RET %endmacro +%macro INT32_TO_FLOAT 1 +cglobal int32_to_float_%1, 3, 3, 3, dst, src, len + mov srcq, [srcq] + mov dstq, [dstq] +%ifidn %1, a + test dstq, mmsize-1 + jne int32_to_float_u_int %+ SUFFIX + test srcq, mmsize-1 + jne int32_to_float_u_int %+ SUFFIX +%else +int32_to_float_u_int %+ SUFFIX +%endif + add srcq, lenq + add dstq, lenq + neg lenq + mova m2, [flt2pm31] +.next: +%ifidn %1, a + cvtdq2ps m0, [ srcq+lenq] + cvtdq2ps m1, [mmsize + srcq+lenq] +%else + movu m0, [ srcq+lenq] + movu m1, [mmsize + srcq+lenq] + cvtdq2ps m0, m0 + cvtdq2ps m1, m1 +%endif + mulps m0, m2 + mulps m1, m2 + mov%1 [ dstq+lenq], m0 + mov%1 [mmsize + dstq+lenq], m1 + add lenq, 2*mmsize + jl .next + REP_RET +%endmacro + + INIT_MMX mmx INT16_TO_INT32 u INT16_TO_INT32 a @@ -62,3 +102,7 @@ INT16_TO_INT32 a INIT_XMM sse INT16_TO_INT32 u INT16_TO_INT32 a + +INIT_XMM sse2 +INT32_TO_FLOAT u +INT32_TO_FLOAT a diff --git a/libswresample/x86/swresample_x86.c b/libswresample/x86/swresample_x86.c index f2a14c1faf..996d72460f 100644 --- a/libswresample/x86/swresample_x86.c +++ b/libswresample/x86/swresample_x86.c @@ -26,6 +26,8 @@ MULTI_CAPS_FUNC_DECL(mmx) MULTI_CAPS_FUNC_DECL(sse) +void ff_int32_to_float_a_sse2(uint8_t **dst, const uint8_t **src, int len); + void swri_audio_convert_init_x86(struct AudioConvert *ac, enum AVSampleFormat out_fmt, enum AVSampleFormat in_fmt, @@ -44,4 +46,9 @@ void swri_audio_convert_init_x86(struct AudioConvert *ac, MULTI_CAPS_FUNC(AV_CPU_FLAG_MMX, mmx) MULTI_CAPS_FUNC(AV_CPU_FLAG_SSE, sse) + + if(mm_flags & AV_CPU_FLAG_SSE2) { + 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; + } } |