---
 libavresample/x86/audio_convert.asm    |   69 ++++++++++++++++++++++++++++++++
 libavresample/x86/audio_convert_init.c |    5 ++
 2 files changed, 74 insertions(+), 0 deletions(-)

diff --git a/libavresample/x86/audio_convert.asm 
b/libavresample/x86/audio_convert.asm
index 53091ef..9559bd5 100644
--- a/libavresample/x86/audio_convert.asm
+++ b/libavresample/x86/audio_convert.asm
@@ -1128,3 +1128,72 @@ CONV_FLT_TO_FLTP_2CH
 INIT_XMM avx
 CONV_FLT_TO_FLTP_2CH
 %endif
+
+;------------------------------------------------------------------------------
+; void ff_conv_flt_to_fltp_6ch(float *const *dst, float *src, int len,
+;                              int channels);
+;------------------------------------------------------------------------------
+
+INIT_XMM sse
+cglobal conv_flt_to_fltp_6ch, 2,8,7, dst, src, dst1, dst2, dst3, dst4, dst5, 
len
+%if ARCH_X86_64
+    mov      lend, r2d
+%else
+    %define lend dword r2m
+%endif
+    mov     dst1q, [dstq+  gprsize]
+    mov     dst2q, [dstq+2*gprsize]
+    mov     dst3q, [dstq+3*gprsize]
+    mov     dst4q, [dstq+4*gprsize]
+    mov     dst5q, [dstq+5*gprsize]
+    mov      dstq, [dstq          ]
+    sub     dst1q, dstq
+    sub     dst2q, dstq
+    sub     dst3q, dstq
+    sub     dst4q, dstq
+    sub     dst5q, dstq
+    ALIGN 16
+.loop:
+    mova       m0, [srcq         ]
+    mova       m1, [srcq+  mmsize]
+    mova       m2, [srcq+2*mmsize]
+    mova       m3, [srcq+3*mmsize]
+    mova       m4, [srcq+4*mmsize]
+    mova       m5, [srcq+5*mmsize]
+    movhlps    m6, m0
+    movlhps    m0, m0
+    unpckhps   m0, m1
+    unpcklps   m6, m2
+    movhlps    m2, m2
+    unpcklps   m1, m2
+    SWAP 2, 6
+    movhlps    m6, m3
+    movlhps    m3, m3
+    unpckhps   m3, m4
+    unpcklps   m6, m5
+    movhlps    m5, m5
+    unpcklps   m4, m5
+    SWAP 5, 6
+    mova       m6, m0
+    unpcklpd   m0, m3
+    unpckhpd   m6, m3
+    SWAP 3, 6
+    mova       m6, m2
+    unpcklpd   m2, m5
+    unpckhpd   m6, m5
+    SWAP 5, 6
+    mova       m6, m1
+    unpcklpd   m1, m4
+    unpckhpd   m6, m4
+    SWAP 4, 6
+    mova  [dstq      ], m0
+    mova  [dstq+dst1q], m3
+    mova  [dstq+dst2q], m2
+    mova  [dstq+dst3q], m5
+    mova  [dstq+dst4q], m1
+    mova  [dstq+dst5q], m4
+    add      srcq, mmsize*6
+    add      dstq, mmsize
+    sub      lend, mmsize/4
+    jg .loop
+    REP_RET
diff --git a/libavresample/x86/audio_convert_init.c 
b/libavresample/x86/audio_convert_init.c
index a06f9af..feb5987 100644
--- a/libavresample/x86/audio_convert_init.c
+++ b/libavresample/x86/audio_convert_init.c
@@ -115,6 +115,9 @@ extern void ff_conv_flt_to_fltp_2ch_sse(float *const *dst, 
float *src, int len,
 extern void ff_conv_flt_to_fltp_2ch_avx(float *const *dst, float *src, int len,
                                         int channels);
 
+extern void ff_conv_flt_to_fltp_6ch_sse(float *const *dst, float *src, int len,
+                                        int channels);
+
 av_cold void ff_audio_convert_init_x86(AudioConvert *ac)
 {
 #if HAVE_YASM
@@ -133,6 +136,8 @@ av_cold void ff_audio_convert_init_x86(AudioConvert *ac)
                                   2, 16, 8, "SSE", 
ff_conv_fltp_to_flt_2ch_sse);
         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);
+        ff_audio_convert_set_func(ac, AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_FLT,
+                                  6, 16, 4, "SSE", 
ff_conv_flt_to_fltp_6ch_sse);
     }
     if (mm_flags & AV_CPU_FLAG_SSE2 && HAVE_SSE) {
         if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
-- 
1.7.1

_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to