[FFmpeg-cvslog] sws/yuv2rgb: clarify precision of coeff and offset for mmx code

[Clément Bœsch]

ffmpeg | branch: master | Clément Bœsch <u at pkh.me> | Fri Feb 26 22:34:10 2016 +0100| [1e7a37f0a07f3f2077352d43602e36024f6c9f2e] | committer: Clément Bœsch

sws/yuv2rgb: clarify precision of coeff and offset for mmx code

It makes easier looking at the difference with the generic code just
below.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1e7a37f0a07f3f2077352d43602e36024f6c9f2e
---

 libswscale/yuv2rgb.c |   12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/libswscale/yuv2rgb.c b/libswscale/yuv2rgb.c
index 9d79d79..62abb7d 100644
--- a/libswscale/yuv2rgb.c
+++ b/libswscale/yuv2rgb.c
@@ -817,12 +817,12 @@ av_cold int ff_yuv2rgb_c_init_tables(SwsContext *c, const int inv_table[4],
 
     c->uOffset = 0x0400040004000400LL;
     c->vOffset = 0x0400040004000400LL;
-    c->yCoeff  = roundToInt16(cy  * 8192) * 0x0001000100010001ULL;
-    c->vrCoeff = roundToInt16(crv * 8192) * 0x0001000100010001ULL;
-    c->ubCoeff = roundToInt16(cbu * 8192) * 0x0001000100010001ULL;
-    c->vgCoeff = roundToInt16(cgv * 8192) * 0x0001000100010001ULL;
-    c->ugCoeff = roundToInt16(cgu * 8192) * 0x0001000100010001ULL;
-    c->yOffset = roundToInt16(oy  *    8) * 0x0001000100010001ULL;
+    c->yCoeff  = roundToInt16(cy  * (1 << 13)) * 0x0001000100010001ULL;
+    c->vrCoeff = roundToInt16(crv * (1 << 13)) * 0x0001000100010001ULL;
+    c->ubCoeff = roundToInt16(cbu * (1 << 13)) * 0x0001000100010001ULL;
+    c->vgCoeff = roundToInt16(cgv * (1 << 13)) * 0x0001000100010001ULL;
+    c->ugCoeff = roundToInt16(cgu * (1 << 13)) * 0x0001000100010001ULL;
+    c->yOffset = roundToInt16(oy  * (1 <<  3)) * 0x0001000100010001ULL;
 
     c->yuv2rgb_y_coeff   = (int16_t)roundToInt16(cy  * (1 << 13));
     c->yuv2rgb_y_offset  = (int16_t)roundToInt16(oy  * (1 <<  9));