[FFmpeg-cvslog] Merge commit '99434f4df81b6801b2b535d5b9143305595784f6'

Clément Bœsch git at videolan.org
Thu Mar 30 11:28:21 EEST 2017


ffmpeg | branch: master | Clément Bœsch <cboesch at gopro.com> | Thu Mar 30 10:23:25 2017 +0200| [f291a9a1adab8ba41e06073b8a9e51f528be6532] | committer: Clément Bœsch

Merge commit '99434f4df81b6801b2b535d5b9143305595784f6'

* commit '99434f4df81b6801b2b535d5b9143305595784f6':
  float_dsp: Have implementation match function pointer prototype

Merged-by: Clément Bœsch <cboesch at gopro.com>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f291a9a1adab8ba41e06073b8a9e51f528be6532
---

 libavutil/x86/float_dsp_init.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libavutil/x86/float_dsp_init.c b/libavutil/x86/float_dsp_init.c
index c836a78..736b9ad 100644
--- a/libavutil/x86/float_dsp_init.c
+++ b/libavutil/x86/float_dsp_init.c
@@ -63,7 +63,7 @@ void ff_vector_fmul_reverse_avx(float *dst, const float *src0,
 
 float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order);
 
-void ff_butterflies_float_sse(float *src0, float *src1, int len);
+void ff_butterflies_float_sse(float *av_restrict src0, float *av_restrict src1, int len);
 
 av_cold void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp)
 {


======================================================================

diff --cc libavutil/x86/float_dsp_init.c
index c836a78,dcbe3f1..736b9ad
--- a/libavutil/x86/float_dsp_init.c
+++ b/libavutil/x86/float_dsp_init.c
@@@ -63,7 -54,73 +63,7 @@@ void ff_vector_fmul_reverse_avx(float *
  
  float ff_scalarproduct_float_sse(const float *v1, const float *v2, int order);
  
- void ff_butterflies_float_sse(float *src0, float *src1, int len);
 -void ff_butterflies_float_sse(float *restrict src0, float *restrict src1, int len);
 -
 -#if HAVE_6REGS && HAVE_INLINE_ASM
 -static void vector_fmul_window_3dnowext(float *dst, const float *src0,
 -                                        const float *src1, const float *win,
 -                                        int len)
 -{
 -    x86_reg i = -len * 4;
 -    x86_reg j =  len * 4 - 8;
 -    __asm__ volatile (
 -        "1:                             \n"
 -        "pswapd (%5, %1), %%mm1         \n"
 -        "movq   (%5, %0), %%mm0         \n"
 -        "pswapd (%4, %1), %%mm5         \n"
 -        "movq   (%3, %0), %%mm4         \n"
 -        "movq      %%mm0, %%mm2         \n"
 -        "movq      %%mm1, %%mm3         \n"
 -        "pfmul     %%mm4, %%mm2         \n" // src0[len + i] * win[len + i]
 -        "pfmul     %%mm5, %%mm3         \n" // src1[j]       * win[len + j]
 -        "pfmul     %%mm4, %%mm1         \n" // src0[len + i] * win[len + j]
 -        "pfmul     %%mm5, %%mm0         \n" // src1[j]       * win[len + i]
 -        "pfadd     %%mm3, %%mm2         \n"
 -        "pfsub     %%mm0, %%mm1         \n"
 -        "pswapd    %%mm2, %%mm2         \n"
 -        "movq      %%mm1, (%2, %0)      \n"
 -        "movq      %%mm2, (%2, %1)      \n"
 -        "sub          $8, %1            \n"
 -        "add          $8, %0            \n"
 -        "jl           1b                \n"
 -        "femms                          \n"
 -        : "+r"(i), "+r"(j)
 -        : "r"(dst + len), "r"(src0 + len), "r"(src1), "r"(win + len)
 -    );
 -}
 -
 -static void vector_fmul_window_sse(float *dst, const float *src0,
 -                                   const float *src1, const float *win, int len)
 -{
 -    x86_reg i = -len * 4;
 -    x86_reg j =  len * 4 - 16;
 -    __asm__ volatile (
 -        "1:                             \n"
 -        "movaps      (%5, %1), %%xmm1   \n"
 -        "movaps      (%5, %0), %%xmm0   \n"
 -        "movaps      (%4, %1), %%xmm5   \n"
 -        "movaps      (%3, %0), %%xmm4   \n"
 -        "shufps $0x1b, %%xmm1, %%xmm1   \n"
 -        "shufps $0x1b, %%xmm5, %%xmm5   \n"
 -        "movaps        %%xmm0, %%xmm2   \n"
 -        "movaps        %%xmm1, %%xmm3   \n"
 -        "mulps         %%xmm4, %%xmm2   \n" // src0[len + i] * win[len + i]
 -        "mulps         %%xmm5, %%xmm3   \n" // src1[j]       * win[len + j]
 -        "mulps         %%xmm4, %%xmm1   \n" // src0[len + i] * win[len + j]
 -        "mulps         %%xmm5, %%xmm0   \n" // src1[j]       * win[len + i]
 -        "addps         %%xmm3, %%xmm2   \n"
 -        "subps         %%xmm0, %%xmm1   \n"
 -        "shufps $0x1b, %%xmm2, %%xmm2   \n"
 -        "movaps        %%xmm1, (%2, %0) \n"
 -        "movaps        %%xmm2, (%2, %1) \n"
 -        "sub              $16, %1       \n"
 -        "add              $16, %0       \n"
 -        "jl                1b           \n"
 -        : "+r"(i), "+r"(j)
 -        : "r"(dst + len), "r"(src0 + len), "r"(src1), "r"(win + len)
 -    );
 -}
 -#endif /* HAVE_6REGS && HAVE_INLINE_ASM */
++void ff_butterflies_float_sse(float *av_restrict src0, float *av_restrict src1, int len);
  
  av_cold void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp)
  {



More information about the ffmpeg-cvslog mailing list