[FFmpeg-cvslog] swr: add ff_int32_to_float_a_avx

[Michael Niedermayer]

ffmpeg | branch: master | Michael Niedermayer <michaelni at gmx.at> | Thu May  3 15:52:36 2012 +0200| [c1fe2db3769d1a9e2e8d3ea718e4306648bafede] | committer: Michael Niedermayer

swr: add ff_int32_to_float_a_avx

Signed-off-by: Michael Niedermayer <michaelni at gmx.at>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c1fe2db3769d1a9e2e8d3ea718e4306648bafede
---

 libswresample/x86/audio_convert.asm |   12 +++++++++---
 libswresample/x86/swresample_x86.c  |    6 ++++++
 2 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/libswresample/x86/audio_convert.asm b/libswresample/x86/audio_convert.asm
index f220e2d..4973b5c 100644
--- a/libswresample/x86/audio_convert.asm
+++ b/libswresample/x86/audio_convert.asm
@@ -22,7 +22,7 @@
 %include "libavutil/x86/x86util.asm"
 
 SECTION_RODATA
-
+align 32
 flt2pm31: times 8 dd 4.6566129e-10
 flt2p31 : times 8 dd 2147483648.0
 flt2p15 : times 8 dd 32768.0
@@ -87,8 +87,8 @@ int32_to_float_u_int %+ SUFFIX
     cvtdq2ps  m0, m0
     cvtdq2ps  m1, m1
 %endif
-    mulps m0, m2
-    mulps m1, m2
+    mulps m0, m0, m2
+    mulps m1, m1, m2
     mov%1 [         dstq+lenq], m0
     mov%1 [mmsize + dstq+lenq], m1
     add lenq, 2*mmsize
@@ -243,3 +243,9 @@ FLOAT_TO_INT32 u
 FLOAT_TO_INT32 a
 FLOAT_TO_INT16 u
 FLOAT_TO_INT16 a
+
+%if HAVE_AVX
+INIT_YMM avx
+INT32_TO_FLOAT u
+INT32_TO_FLOAT a
+%endif
diff --git a/libswresample/x86/swresample_x86.c b/libswresample/x86/swresample_x86.c
index 9e5ac49..dc6ade5 100644
--- a/libswresample/x86/swresample_x86.c
+++ b/libswresample/x86/swresample_x86.c
@@ -33,6 +33,8 @@ void ff_int16_to_float_a_sse2(uint8_t **dst, const uint8_t **src, int len);
 void ff_float_to_int32_a_sse2(uint8_t **dst, const uint8_t **src, int len);
 void ff_float_to_int16_a_sse2(uint8_t **dst, const uint8_t **src, int len);
 
+void ff_int32_to_float_a_avx(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,
@@ -64,4 +66,8 @@ MULTI_CAPS_FUNC(AV_CPU_FLAG_SSE, sse)
         if(   out_fmt == AV_SAMPLE_FMT_S16  && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S16P && in_fmt == AV_SAMPLE_FMT_FLTP)
             ac->simd_f =  ff_float_to_int16_a_sse2;
     }
+    if(HAVE_AVX && mm_flags & AV_CPU_FLAG_AVX) {
+        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;
+    }
 }