[FFmpeg-cvslog] YUV->BGR32 MMX support

[Kevin Coyle]

ffmpeg | branch: master | Kevin Coyle <kcoyle at netflix.com> | Fri Jul  3 11:16:42 2015 -0700| [126271138803fdb1eb6059f4f71c88254ae2a20b] | committer: Michael Niedermayer

YUV->BGR32 MMX support

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

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=126271138803fdb1eb6059f4f71c88254ae2a20b
---

 libswscale/x86/swscale_template.c |   30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c
index 36a606c..bbda6d0 100644
--- a/libswscale/x86/swscale_template.c
+++ b/libswscale/x86/swscale_template.c
@@ -399,6 +399,35 @@ static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
     }
 }
 
+static void RENAME(yuv2bgr32_X)(SwsContext *c, const int16_t *lumFilter,
+                                const int16_t **lumSrc, int lumFilterSize,
+                                const int16_t *chrFilter, const int16_t **chrUSrc,
+                                const int16_t **chrVSrc,
+                                int chrFilterSize, const int16_t **alpSrc,
+                                uint8_t *dest, int dstW, int dstY)
+{
+    x86_reg dummy=0;
+    x86_reg dstW_reg = dstW;
+    x86_reg uv_off = c->uv_offx2;
+
+    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
+        YSCALEYUV2PACKEDX
+        YSCALEYUV2RGBX
+        YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
+        "psraw                        $3, %%mm1         \n\t"
+        "psraw                        $3, %%mm7         \n\t"
+        "packuswb                  %%mm7, %%mm1         \n\t"
+        WRITEBGR32(%4, "%5", %%REGa, %%mm5, %%mm4, %%mm2, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
+        YSCALEYUV2PACKEDX_END
+    } else {
+        YSCALEYUV2PACKEDX
+        YSCALEYUV2RGBX
+        "pcmpeqd %%mm7, %%mm7 \n\t"
+        WRITEBGR32(%4, "%5", %%REGa, %%mm5, %%mm4, %%mm2, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
+        YSCALEYUV2PACKEDX_END
+    }
+}
+
 #define REAL_WRITERGB16(dst, dstw, index) \
     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
     "pand "MANGLE(bFC)", %%mm4  \n\t" /* G */\
@@ -1491,6 +1520,7 @@ static av_cold void RENAME(sws_init_swscale)(SwsContext *c)
                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
                     switch (c->dstFormat) {
                     case AV_PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;
+                    case AV_PIX_FMT_BGR32:   c->yuv2packedX = RENAME(yuv2bgr32_X);   break;
 #if HAVE_6REGS
                     case AV_PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;
 #endif