[FFmpeg-devel] [PATCH] Swscale YUV2RGB table generator

Kostya kostya.shishkov
Mon Feb 9 09:49:57 CET 2009


On Sun, Feb 08, 2009 at 09:08:01PM +0100, Michael Niedermayer wrote:
> On Sun, Feb 08, 2009 at 08:27:49PM +0200, Kostya wrote:
> > Here's LGPLed version of YUV2RGB tables generator.
> > While it may be not perfect, it seems to work.
> > 
> > Please test and/or give your comments.
[...]
> > +#define RGB(i)                                      \
> > +    U = pu[i];                                      \
> > +    V = pv[i];                                      \
> > +    r = (void *)c->table_rV[V];                     \
> > +    g = (void *)(c->table_gU[U] + c->table_gV[V]);  \
> > +    b = (void *)c->table_bU[U];
> > +
> > +#define DST1(i)                         \
> > +    Y = py_1[2*i];                      \
> > +    dst_1[2*i] = r[Y] + g[Y] + b[Y];    \
> > +    Y = py_1[2*i+1];                    \
> > +    dst_1[2*i+1] = r[Y] + g[Y] + b[Y];
> > +
> > +#define DST2(i)                         \
> > +    Y = py_2[2*i];                      \
> > +    dst_2[2*i] = r[Y] + g[Y] + b[Y];    \
> > +    Y = py_2[2*i+1];                    \
> > +    dst_2[2*i+1] = r[Y] + g[Y] + b[Y];
> > +
> > +#define DST1RGB(i)                                                \
> > +    Y = py_1[2*i];                                                \
> > +    dst_1[6*i] = r[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = b[Y];  \
> > +    Y = py_1[2*i+1];                                              \
> > +    dst_1[6*i+3] = r[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = b[Y];
> > +
> > +#define DST2RGB(i)                                                \
> > +    Y = py_2[2*i];                                                \
> > +    dst_2[6*i] = r[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = b[Y];  \
> > +    Y = py_2[2*i+1];                                              \
> > +    dst_2[6*i+3] = r[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = b[Y];
> > +
> > +#define DST1BGR(i)                                                \
> > +    Y = py_1[2*i];                                                \
> > +    dst_1[6*i] = b[Y]; dst_1[6*i+1] = g[Y]; dst_1[6*i+2] = r[Y];  \
> > +    Y = py_1[2*i+1];                                              \
> > +    dst_1[6*i+3] = b[Y]; dst_1[6*i+4] = g[Y]; dst_1[6*i+5] = r[Y];
> > +
> > +#define DST2BGR(i)                                                \
> > +    Y = py_2[2*i];                                                \
> > +    dst_2[6*i] = b[Y]; dst_2[6*i+1] = g[Y]; dst_2[6*i+2] = r[Y];  \
> > +    Y = py_2[2*i+1];                                              \
> > +    dst_2[6*i+3] = b[Y]; dst_2[6*i+4] = g[Y]; dst_2[6*i+5] = r[Y];
> > +
> 
> this doesnt look rewritten
> i think this has to be split out into a seperate GPL file or it has to
> be rewritten as well
 
Rewritten (our goal is to get rid of GPL there, isn't it?)
Personally I don't care about SwScale or GPL/LGPL issues, I just don't like
the fact FFmpeg has different behaviour in GPL mode. 
 
[...]
> -- 
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
-------------- next part --------------
Index: Makefile
===================================================================
--- Makefile	(revision 28473)
+++ Makefile	(working copy)
@@ -5,10 +5,9 @@
 
 HEADERS = swscale.h
 
-OBJS = rgb2rgb.o swscale.o swscale_avoption.o
+OBJS = rgb2rgb.o swscale.o swscale_avoption.o yuv2rgb2.o
 
 OBJS-$(ARCH_BFIN)          +=  internal_bfin.o swscale_bfin.o yuv2rgb_bfin.o
-OBJS-$(CONFIG_GPL)         +=  yuv2rgb.o
 OBJS-$(CONFIG_MLIB)        +=  yuv2rgb_mlib.o
 OBJS-$(HAVE_ALTIVEC)       +=  yuv2rgb_altivec.o
 OBJS-$(HAVE_VIS)           +=  yuv2rgb_vis.o
Index: swscale.c
===================================================================
--- swscale.c	(revision 28473)
+++ swscale.c	(working copy)
@@ -174,7 +174,7 @@
 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 
-extern const int32_t Inverse_Table_6_9[8][4];
+extern const int32_t ff_yuv2rgb_coeffs[8][4];
 
 static const double rgb2yuv_table[8][9]={
     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
@@ -2076,7 +2076,7 @@
 }
 
 /**
- * @param inv_table the yuv2rgb coefficients, normally Inverse_Table_6_9[x]
+ * @param inv_table the yuv2rgb coefficients, normally ff_yuv2rgb_coeffs[x]
  * @param fullRange if 1 then the luma range is 0..255 if 0 it is 16..235
  * @return -1 if not supported
  */
@@ -2321,7 +2321,7 @@
     c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
     c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
 
-    sws_setColorspaceDetails(c, Inverse_Table_6_9[SWS_CS_DEFAULT], srcRange, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
+    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
 
     /* unscaled special cases */
     if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))
Index: yuv2rgb2.c
===================================================================
--- yuv2rgb2.c	(revision 0)
+++ yuv2rgb2.c	(revision 0)
@@ -0,0 +1,689 @@
+/*
+ * yuv2rgb.c, Software YUV to RGB converter
+ *
+ *  Copyright (C) 2009 Konstantin Shishkov
+ *
+ *  MMX/MMX2 template stuff (needed for fast movntq support),
+ *  1,4,8bpp support and context / deglobalize stuff
+ *  by Michael Niedermayer (michaelni at gmx.at)
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <inttypes.h>
+#include <assert.h>
+
+#include "config.h"
+#include "rgb2rgb.h"
+#include "swscale.h"
+#include "swscale_internal.h"
+
+#define DITHER1XBPP // only for MMX
+
+extern const uint8_t dither_8x8_32[8][8];
+extern const uint8_t dither_8x8_73[8][8];
+extern const uint8_t dither_8x8_220[8][8];
+
+#if HAVE_MMX
+
+/* hope these constant values are cache line aligned */
+DECLARE_ASM_CONST(8, uint64_t, mmx_00ffw)   = 0x00ff00ff00ff00ffULL;
+DECLARE_ASM_CONST(8, uint64_t, mmx_redmask) = 0xf8f8f8f8f8f8f8f8ULL;
+DECLARE_ASM_CONST(8, uint64_t, mmx_grnmask) = 0xfcfcfcfcfcfcfcfcULL;
+
+//MMX versions
+#undef RENAME
+#undef HAVE_MMX2
+#undef HAVE_AMD3DNOW
+#define HAVE_MMX2 0
+#define HAVE_AMD3DNOW 0
+#define RENAME(a) a ## _MMX
+#include "yuv2rgb_template.c"
+
+//MMX2 versions
+#undef RENAME
+#undef HAVE_MMX2
+#define HAVE_MMX2 1
+#define RENAME(a) a ## _MMX2
+#include "yuv2rgb_template.c"
+
+#endif /* HAVE_MMX */
+
+const int32_t ff_yuv2rgb_coeffs[8][4] = {
+    {117504, 138453, 13954, 34903}, /* no sequence_display_extension */
+    {117504, 138453, 13954, 34903}, /* ITU-R Rec. 709 (1990) */
+    {104597, 132201, 25675, 53279}, /* unspecified */
+    {104597, 132201, 25675, 53279}, /* reserved */
+    {104448, 132798, 24759, 53109}, /* FCC */
+    {104597, 132201, 25675, 53279}, /* ITU-R Rec. 624-4 System B, G */
+    {104597, 132201, 25675, 53279}, /* SMPTE 170M */
+    {117579, 136230, 16907, 35559}  /* SMPTE 240M (1987) */
+};
+
+#define LOADCHROMA(i)                               \
+    U = pu[i];                                      \
+    V = pv[i];                                      \
+    r = (void *)c->table_rV[V];                     \
+    g = (void *)(c->table_gU[U] + c->table_gV[V]);  \
+    b = (void *)c->table_bU[U];
+
+#define PUTRGB(dst,src,i,o)          \
+    Y = src[2*i+o];                  \
+    dst[2*i  ] = r[Y] + g[Y] + b[Y]; \
+    Y = src[2*i+1-o];                \
+    dst[2*i+1] = r[Y] + g[Y] + b[Y];
+
+#define PUTRGB24(dst,src,i)                                  \
+    Y = src[2*i];                                            \
+    dst[6*i+0] = r[Y]; dst[6*i+1] = g[Y]; dst[6*i+2] = b[Y]; \
+    Y = src[2*i+1];                                          \
+    dst[6*i+3] = r[Y]; dst[6*i+4] = g[Y]; dst[6*i+5] = b[Y];
+
+#define PUTBGR24(dst,src,i)                                  \
+    Y = src[2*i];                                            \
+    dst[6*i+0] = b[Y]; dst[6*i+1] = g[Y]; dst[6*i+2] = r[Y]; \
+    Y = src[2*i+1];                                          \
+    dst[6*i+3] = b[Y]; dst[6*i+4] = g[Y]; dst[6*i+5] = r[Y];
+
+#define YUV2RGBFUNC(func_name, dst_type) \
+static int func_name(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, \
+                     int srcSliceH, uint8_t* dst[], int dstStride[]){\
+    int y;\
+\
+    if (c->srcFormat == PIX_FMT_YUV422P){\
+        srcStride[1] *= 2;\
+        srcStride[2] *= 2;\
+    }\
+    for (y=0; y<srcSliceH; y+=2){\
+        dst_type *dst_1= (dst_type*)(dst[0] + (y+srcSliceY  )*dstStride[0]);\
+        dst_type *dst_2= (dst_type*)(dst[0] + (y+srcSliceY+1)*dstStride[0]);\
+        dst_type av_unused *r, *b;\
+        dst_type *g;\
+        uint8_t *py_1= src[0] + y*srcStride[0];\
+        uint8_t *py_2= py_1 + srcStride[0];\
+        uint8_t *pu= src[1] + (y>>1)*srcStride[1];\
+        uint8_t *pv= src[2] + (y>>1)*srcStride[2];\
+        unsigned int h_size= c->dstW>>3;\
+        while (h_size--) {\
+            int av_unused U, V;\
+            int Y;\
+
+#define ENDYUV2RGBLINE(dst_delta)\
+            pu += 4;\
+            pv += 4;\
+            py_1 += 8;\
+            py_2 += 8;\
+            dst_1 += dst_delta;\
+            dst_2 += dst_delta;\
+        }\
+        if (c->dstW & 4) {\
+            int av_unused Y, U, V;\
+
+#define ENDYUV2RGBFUNC()\
+        }\
+    }\
+    return srcSliceH;\
+}
+
+#define CLOSEYUV2RGBFUNC(dst_delta)\
+    ENDYUV2RGBLINE(dst_delta)\
+    ENDYUV2RGBFUNC()
+
+YUV2RGBFUNC(yuv2rgb_c_32, uint32_t)
+    LOADCHROMA(0);
+    PUTRGB(dst_1,py_1,0,0);
+    PUTRGB(dst_2,py_2,0,1);
+
+    LOADCHROMA(1);
+    PUTRGB(dst_2,py_2,1,1);
+    PUTRGB(dst_1,py_1,1,0);
+    LOADCHROMA(1);
+    PUTRGB(dst_2,py_2,1,1);
+    PUTRGB(dst_1,py_1,1,0);
+
+    LOADCHROMA(2);
+    PUTRGB(dst_1,py_1,2,0);
+    PUTRGB(dst_2,py_2,2,1);
+
+    LOADCHROMA(3);
+    PUTRGB(dst_2,py_2,3,1);
+    PUTRGB(dst_1,py_1,3,0);
+ENDYUV2RGBLINE(8)
+    LOADCHROMA(0);
+    PUTRGB(dst_1,py_1,0,0);
+    PUTRGB(dst_2,py_2,0,1);
+
+    LOADCHROMA(1);
+    PUTRGB(dst_2,py_2,1,1);
+    PUTRGB(dst_1,py_1,1,0);
+ENDYUV2RGBFUNC()
+
+YUV2RGBFUNC(yuv2rgb_c_24_rgb, uint8_t)
+    LOADCHROMA(0);
+    PUTRGB24(dst_1,py_1,0);
+    PUTRGB24(dst_2,py_2,0);
+
+    LOADCHROMA(1);
+    PUTRGB24(dst_2,py_2,1);
+    PUTRGB24(dst_1,py_1,1);
+
+    LOADCHROMA(2);
+    PUTRGB24(dst_1,py_1,2);
+    PUTRGB24(dst_2,py_2,2);
+
+    LOADCHROMA(3);
+    PUTRGB24(dst_2,py_2,3);
+    PUTRGB24(dst_1,py_1,3);
+ENDYUV2RGBLINE(24)
+    LOADCHROMA(0);
+    PUTRGB24(dst_1,py_1,0);
+    PUTRGB24(dst_2,py_2,0);
+
+    LOADCHROMA(1);
+    PUTRGB24(dst_2,py_2,1);
+    PUTRGB24(dst_1,py_1,1);
+ENDYUV2RGBFUNC()
+
+// only trivial mods from yuv2rgb_c_24_rgb
+YUV2RGBFUNC(yuv2rgb_c_24_bgr, uint8_t)
+    LOADCHROMA(0);
+    PUTBGR24(dst_1,py_1,0);
+    PUTBGR24(dst_2,py_2,0);
+
+    LOADCHROMA(1);
+    PUTBGR24(dst_2,py_2,1);
+    PUTBGR24(dst_1,py_1,1);
+
+    LOADCHROMA(2);
+    PUTBGR24(dst_1,py_1,2);
+    PUTBGR24(dst_2,py_2,2);
+
+    LOADCHROMA(3);
+    PUTBGR24(dst_2,py_2,3);
+    PUTBGR24(dst_1,py_1,3);
+ENDYUV2RGBLINE(24)
+    LOADCHROMA(0);
+    PUTBGR24(dst_1,py_1,0);
+    PUTBGR24(dst_2,py_2,0);
+
+    LOADCHROMA(1);
+    PUTBGR24(dst_2,py_2,1);
+    PUTBGR24(dst_1,py_1,1);
+ENDYUV2RGBFUNC()
+
+// This is exactly the same code as yuv2rgb_c_32 except for the types of
+// r, g, b, dst_1, dst_2
+YUV2RGBFUNC(yuv2rgb_c_16, uint16_t)
+    LOADCHROMA(0);
+    PUTRGB(dst_1,py_1,0,0);
+    PUTRGB(dst_2,py_2,0,1);
+
+    LOADCHROMA(1);
+    PUTRGB(dst_2,py_2,1,1);
+    PUTRGB(dst_1,py_1,1,0);
+
+    LOADCHROMA(2);
+    PUTRGB(dst_1,py_1,2,0);
+    PUTRGB(dst_2,py_2,2,1);
+
+    LOADCHROMA(3);
+    PUTRGB(dst_2,py_2,3,1);
+    PUTRGB(dst_1,py_1,3,0);
+CLOSEYUV2RGBFUNC(8)
+
+// This is exactly the same code as yuv2rgb_c_32 except for the types of
+// r, g, b, dst_1, dst_2
+YUV2RGBFUNC(yuv2rgb_c_8, uint8_t)
+    LOADCHROMA(0);
+    PUTRGB(dst_1,py_1,0,0);
+    PUTRGB(dst_2,py_2,0,1);
+
+    LOADCHROMA(1);
+    PUTRGB(dst_2,py_2,1,1);
+    PUTRGB(dst_1,py_1,1,0);
+
+    LOADCHROMA(2);
+    PUTRGB(dst_1,py_1,2,0);
+    PUTRGB(dst_2,py_2,2,1);
+
+    LOADCHROMA(3);
+    PUTRGB(dst_2,py_2,3,1);
+    PUTRGB(dst_1,py_1,3,0);
+CLOSEYUV2RGBFUNC(8)
+
+// r, g, b, dst_1, dst_2
+YUV2RGBFUNC(yuv2rgb_c_8_ordered_dither, uint8_t)
+    const uint8_t *d32= dither_8x8_32[y&7];
+    const uint8_t *d64= dither_8x8_73[y&7];
+#define PUTRGB8(dst,src,i,o)                                    \
+    Y = src[2*i];                                               \
+    dst[2*i]   = r[Y+d32[0+o]] + g[Y+d32[0+o]] + b[Y+d64[0+o]]; \
+    Y = src[2*i+1];                                             \
+    dst[2*i+1] = r[Y+d32[1+o]] + g[Y+d32[1+o]] + b[Y+d64[1+o]];
+
+    LOADCHROMA(0);
+    PUTRGB8(dst_1,py_1,0,0);
+    PUTRGB8(dst_2,py_2,0,0+8);
+
+    LOADCHROMA(1);
+    PUTRGB8(dst_2,py_2,1,2+8);
+    PUTRGB8(dst_1,py_1,1,2);
+
+    LOADCHROMA(2);
+    PUTRGB8(dst_1,py_1,2,4);
+    PUTRGB8(dst_2,py_2,2,4+8);
+
+    LOADCHROMA(3);
+    PUTRGB8(dst_2,py_2,3,6+8);
+    PUTRGB8(dst_1,py_1,3,6);
+CLOSEYUV2RGBFUNC(8)
+
+
+// This is exactly the same code as yuv2rgb_c_32 except for the types of
+// r, g, b, dst_1, dst_2
+YUV2RGBFUNC(yuv2rgb_c_4, uint8_t)
+    int acc;
+#define PUTRGB4(dst,src,i)          \
+    Y = src[2*i];                   \
+    acc = r[Y] + g[Y] + b[Y];       \
+    Y = src[2*i+1];                 \
+    acc |= (r[Y] + g[Y] + b[Y])<<4; \
+    dst[i] = acc;
+
+    LOADCHROMA(0);
+    PUTRGB4(dst_1,py_1,0);
+    PUTRGB4(dst_2,py_2,0);
+
+    LOADCHROMA(1);
+    PUTRGB4(dst_2,py_2,1);
+    PUTRGB4(dst_1,py_1,1);
+
+    LOADCHROMA(2);
+    PUTRGB4(dst_1,py_1,2);
+    PUTRGB4(dst_2,py_2,2);
+
+    LOADCHROMA(3);
+    PUTRGB4(dst_2,py_2,3);
+    PUTRGB4(dst_1,py_1,3);
+CLOSEYUV2RGBFUNC(4)
+
+YUV2RGBFUNC(yuv2rgb_c_4_ordered_dither, uint8_t)
+    const uint8_t *d64= dither_8x8_73[y&7];
+    const uint8_t *d128=dither_8x8_220[y&7];
+    int acc;
+
+#define PUTRGB4D(dst,src,i,o)                                     \
+    Y = src[2*i];                                                 \
+    acc = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]];        \
+    Y = src[2*i+1];                                               \
+    acc |= (r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]])<<4;  \
+    dst[i]= acc;
+
+    LOADCHROMA(0);
+    PUTRGB4D(dst_1,py_1,0,0);
+    PUTRGB4D(dst_2,py_2,0,0+8);
+
+    LOADCHROMA(1);
+    PUTRGB4D(dst_2,py_2,1,2+8);
+    PUTRGB4D(dst_1,py_1,1,2);
+
+    LOADCHROMA(2);
+    PUTRGB4D(dst_1,py_1,2,4);
+    PUTRGB4D(dst_2,py_2,2,4+8);
+
+    LOADCHROMA(3);
+    PUTRGB4D(dst_2,py_2,3,6+8);
+    PUTRGB4D(dst_1,py_1,3,6);
+CLOSEYUV2RGBFUNC(4)
+
+// This is exactly the same code as yuv2rgb_c_32 except for the types of
+// r, g, b, dst_1, dst_2
+YUV2RGBFUNC(yuv2rgb_c_4b, uint8_t)
+    LOADCHROMA(0);
+    PUTRGB(dst_1,py_1,0,0);
+    PUTRGB(dst_2,py_2,0,1);
+
+    LOADCHROMA(1);
+    PUTRGB(dst_2,py_2,1,1);
+    PUTRGB(dst_1,py_1,1,0);
+
+    LOADCHROMA(2);
+    PUTRGB(dst_1,py_1,2,0);
+    PUTRGB(dst_2,py_2,2,1);
+
+    LOADCHROMA(3);
+    PUTRGB(dst_2,py_2,3,1);
+    PUTRGB(dst_1,py_1,3,0);
+CLOSEYUV2RGBFUNC(8)
+
+YUV2RGBFUNC(yuv2rgb_c_4b_ordered_dither, uint8_t)
+    const uint8_t *d64= dither_8x8_73[y&7];
+    const uint8_t *d128=dither_8x8_220[y&7];
+
+#define PUTRGB4DB(dst,src,i,o)                                    \
+    Y = src[2*i];                                                 \
+    dst[2*i]   = r[Y+d128[0+o]] + g[Y+d64[0+o]] + b[Y+d128[0+o]]; \
+    Y = src[2*i+1];                                               \
+    dst[2*i+1] = r[Y+d128[1+o]] + g[Y+d64[1+o]] + b[Y+d128[1+o]];
+
+    LOADCHROMA(0);
+    PUTRGB4DB(dst_1,py_1,0,0);
+    PUTRGB4DB(dst_2,py_2,0,0+8);
+
+    LOADCHROMA(1);
+    PUTRGB4DB(dst_2,py_2,1,2+8);
+    PUTRGB4DB(dst_1,py_1,1,2);
+
+    LOADCHROMA(2);
+    PUTRGB4DB(dst_1,py_1,2,4);
+    PUTRGB4DB(dst_2,py_2,2,4+8);
+
+    LOADCHROMA(3);
+    PUTRGB4DB(dst_2,py_2,3,6+8);
+    PUTRGB4DB(dst_1,py_1,3,6);
+CLOSEYUV2RGBFUNC(8)
+
+YUV2RGBFUNC(yuv2rgb_c_1_ordered_dither, uint8_t)
+        const uint8_t *d128=dither_8x8_220[y&7];
+        char out_1=0, out_2=0;
+        g= c->table_gU[128] + c->table_gV[128];
+
+#define PUTRGB1(out,src,i,o)    \
+    Y = src[2*i];               \
+    out+= out + g[Y+d128[0+o]]; \
+    Y = src[2*i+1];             \
+    out+= out + g[Y+d128[1+o]];
+
+    PUTRGB1(out_1,py_1,0,0);
+    PUTRGB1(out_2,py_2,0,0+8);
+
+    PUTRGB1(out_2,py_2,1,2+8);
+    PUTRGB1(out_1,py_1,1,2);
+
+    PUTRGB1(out_1,py_1,2,4);
+    PUTRGB1(out_2,py_2,2,4+8);
+
+    PUTRGB1(out_2,py_2,3,6+8);
+    PUTRGB1(out_1,py_1,3,6);
+
+    dst_1[0]= out_1;
+    dst_2[0]= out_2;
+CLOSEYUV2RGBFUNC(1)
+
+SwsFunc yuv2rgb_get_func_ptr (SwsContext *c)
+{
+#if HAVE_MMX2 || HAVE_MMX
+    if(c->flags & SWS_CPU_CAPS_MMX2){
+        switch(c->dstFormat){
+        case PIX_FMT_RGB32:  return yuv420_rgb32_MMX2;
+        case PIX_FMT_BGR24:  return yuv420_rgb24_MMX2;
+        case PIX_FMT_RGB565: return yuv420_rgb16_MMX2;
+        case PIX_FMT_RGB555: return yuv420_rgb15_MMX2;
+        }
+    }
+    if(c->flags & SWS_CPU_CAPS_MMX){
+        switch(c->dstFormat){
+        case PIX_FMT_RGB32:  return yuv420_rgb32_MMX;
+        case PIX_FMT_BGR24:  return yuv420_rgb24_MMX;
+        case PIX_FMT_RGB565: return yuv420_rgb16_MMX;
+        case PIX_FMT_RGB555: return yuv420_rgb15_MMX;
+        }
+    }
+#endif
+#if HAVE_VIS
+    {
+        SwsFunc t= yuv2rgb_init_vis(c);
+        if (t) return t;
+    }
+#endif
+#if CONFIG_MLIB
+    {
+        SwsFunc t= yuv2rgb_init_mlib(c);
+        if (t) return t;
+    }
+#endif
+#if HAVE_ALTIVEC
+    if(c->flags & SWS_CPU_CAPS_ALTIVEC){
+        SwsFunc t = yuv2rgb_init_altivec(c);
+        if (t) return t;
+    }
+#endif
+
+#if ARCH_BFIN
+    if(c->flags & SWS_CPU_CAPS_BFIN){
+        SwsFunc t = ff_bfin_yuv2rgb_get_func_ptr (c);
+        if (t) return t;
+    }
+#endif
+
+    av_log(c, AV_LOG_WARNING, "No accelerated colorspace conversion found.\n");
+
+    switch(c->dstFormat){
+    case PIX_FMT_BGR32_1:
+    case PIX_FMT_RGB32_1:
+    case PIX_FMT_BGR32:
+    case PIX_FMT_RGB32: return yuv2rgb_c_32;
+    case PIX_FMT_RGB24: return yuv2rgb_c_24_rgb;
+    case PIX_FMT_BGR24: return yuv2rgb_c_24_bgr;
+    case PIX_FMT_RGB565:
+    case PIX_FMT_BGR565:
+    case PIX_FMT_RGB555:
+    case PIX_FMT_BGR555: return yuv2rgb_c_16;
+    case PIX_FMT_RGB8:
+    case PIX_FMT_BGR8:  return yuv2rgb_c_8_ordered_dither;
+    case PIX_FMT_RGB4:
+    case PIX_FMT_BGR4:  return yuv2rgb_c_4_ordered_dither;
+    case PIX_FMT_RGB4_BYTE:
+    case PIX_FMT_BGR4_BYTE:  return yuv2rgb_c_4b_ordered_dither;
+    case PIX_FMT_MONOBLACK:  return yuv2rgb_c_1_ordered_dither;
+    default:
+        assert(0);
+    }
+    return NULL;
+}
+
+static void fill_table(uint8_t* table[256], int elemsize, int inc, uint8_t *y_table)
+{
+    int i;
+    int64_t cb = 0;
+
+    y_table -= elemsize * (inc >> 9);
+
+    for(i = 0; i < 256; i++){
+        table[i] = y_table + elemsize * ((cb + 0x8000) >> 16);
+        cb += inc;
+    }
+}
+
+static void fill_gv_table(int table[256], int elemsize, int inc, int off)
+{
+    int i;
+    int64_t cb = 0;
+
+    off -= inc >> 9;
+
+    for(i = 0; i < 256; i++){
+        table[i] = elemsize * (off + ((cb + 0x8000) >> 16));
+        cb += inc;
+    }
+}
+
+av_cold int yuv2rgb_c_init_tables (SwsContext *c, const int inv_table[4], int fullRange,
+                                   int brightness, int contrast, int saturation)
+{
+    const int isRgb =      c->dstFormat==PIX_FMT_RGB32
+                        || c->dstFormat==PIX_FMT_RGB32_1
+                        || c->dstFormat==PIX_FMT_BGR24
+                        || c->dstFormat==PIX_FMT_RGB565
+                        || c->dstFormat==PIX_FMT_RGB555
+                        || c->dstFormat==PIX_FMT_RGB8
+                        || c->dstFormat==PIX_FMT_RGB4
+                        || c->dstFormat==PIX_FMT_RGB4_BYTE
+                        || c->dstFormat==PIX_FMT_MONOBLACK;
+    const int bpp = fmt_depth(c->dstFormat);
+    uint8_t *y_table;
+    uint16_t *y_table16;
+    uint32_t *y_table32;
+    int i, base, rbase, gbase, bbase;
+
+    int64_t crv =  inv_table[0];
+    int64_t cbu =  inv_table[1];
+    int64_t cgu = -inv_table[2];
+    int64_t cgv = -inv_table[3];
+    int64_t cy  = 1<<16;
+    int64_t oy  = 0;
+
+    int64_t yb = 0;
+
+    if(!fullRange){
+        cy= (cy*255) / 219;
+        oy= 16<<16;
+    }else{
+        crv= (crv*224) / 255;
+        cbu= (cbu*224) / 255;
+        cgu= (cgu*224) / 255;
+        cgv= (cgv*224) / 255;
+    }
+
+    cy = (cy *contrast             )>>16;
+    crv= (crv*contrast * saturation)>>32;
+    cbu= (cbu*contrast * saturation)>>32;
+    cgu= (cgu*contrast * saturation)>>32;
+    cgv= (cgv*contrast * saturation)>>32;
+    oy -= 256*brightness;
+
+    //scale coefficients by cy
+    crv = (crv << 16) / cy;
+    cbu = (cbu << 16) / cy;
+    cgu = (cgu << 16) / cy;
+    cgv = (cgv << 16) / cy;
+
+    av_free(c->yuvTable);
+
+    switch(bpp){
+    case 1:
+        c->yuvTable = av_malloc(512);
+        y_table = c->yuvTable;
+        yb = -(384<<16) - oy;
+        for(i = 0; i < 512; i++){
+            y_table[i] = av_clip_uint8((yb + 0x8000) >> 16) >> 7;
+            yb += cy;
+        }
+        fill_table(c->table_gU, 1, cgu, y_table + 256 - (oy >> 16));
+        fill_gv_table(c->table_gV, 1, cgv, 0);
+        break;
+    case 4:
+    case 4|128:
+        rbase = isRgb ? 3 : 0;
+        gbase = 1;
+        bbase = isRgb ? 0 : 3;
+        c->yuvTable = av_malloc(1024*3);
+        y_table = c->yuvTable;
+        yb = -(384<<16) - oy;
+        for(i = 0; i < 1024; i++){
+            int yval = av_clip_uint8((yb + 0x8000) >> 16);
+            y_table[i     ] = (yval >> 7) << rbase;
+            y_table[i+1024] = ((yval + 43) / 85) << gbase;
+            y_table[i+2048] = (yval >> 7) << bbase;
+            yb += cy;
+        }
+        fill_table(c->table_rV, 1, crv, y_table + 232        - (oy >> 16));
+        fill_table(c->table_gU, 1, cgu, y_table + 232 + 1024 - (oy >> 16));
+        fill_table(c->table_bU, 1, cbu, y_table + 232 + 2048 - (oy >> 16));        
+        fill_gv_table(c->table_gV, 1, cgv, 0);
+        break;
+    case 8:
+        rbase = isRgb ? 5 : 0;
+        gbase = isRgb ? 2 : 3;
+        bbase = isRgb ? 0 : 6;
+        c->yuvTable = av_malloc(1024*3);
+        y_table = c->yuvTable;
+        yb = -(384<<16) - oy;
+        for(i = 0; i < 1024; i++){
+            int yval = av_clip_uint8((yb + 0x8000) >> 16);
+            y_table[i     ] = ((yval + 18) / 36) << rbase;
+            y_table[i+1024] = ((yval + 18) / 36) << gbase;
+            y_table[i+2048] = ((yval + 43) / 85) << bbase;
+            yb += cy;
+        }
+        fill_table(c->table_rV, 1, crv, y_table + 327        - (oy >> 16));
+        fill_table(c->table_gU, 1, cgu, y_table + 327 + 1024 - (oy >> 16));
+        fill_table(c->table_bU, 1, cbu, y_table + 327 + 2048 - (oy >> 16));        
+        fill_gv_table(c->table_gV, 1, cgv, 0);
+        break;
+    case 15:
+    case 16:
+        rbase = isRgb ? bpp - 5 : 0;
+        gbase = 5;
+        bbase = isRgb ? 0 : (bpp - 5);
+        c->yuvTable = av_malloc(1024*3*2);
+        y_table16 = c->yuvTable;
+        yb = -(384<<16) - oy;
+        for(i = 0; i < 1024; i++){
+            uint8_t yval = av_clip_uint8((yb + 0x8000) >> 16);
+            y_table16[i     ] = (yval >> 3)          << rbase;
+            y_table16[i+1024] = (yval >> (18 - bpp)) << gbase;
+            y_table16[i+2048] = (yval >> 3)          << bbase;
+            yb += cy;
+        }
+        fill_table(c->table_rV, 2, crv, y_table16 + 342        - (oy >> 16));
+        fill_table(c->table_gU, 2, cgu, y_table16 + 342 + 1024 - (oy >> 16));
+        fill_table(c->table_bU, 2, cbu, y_table16 + 342 + 2048 - (oy >> 16));        
+        fill_gv_table(c->table_gV, 2, cgv, 0);
+        break;
+    case 24:
+        c->yuvTable = av_malloc(1024);
+        y_table = c->yuvTable;
+        yb = -(384<<16) - oy;
+        for(i = 0; i < 1024; i++){
+            y_table[i] = av_clip_uint8((yb + 0x8000) >> 16);
+            yb += cy;
+        }
+        fill_table(c->table_rV, 1, crv, y_table + 384 - (oy >> 16));
+        fill_table(c->table_gU, 1, cgu, y_table + 384 - (oy >> 16));
+        fill_table(c->table_bU, 1, cbu, y_table + 384 - (oy >> 16));
+        fill_gv_table(c->table_gV, 1, cgv, 16);
+        break;
+    case 32:
+        base = (c->dstFormat == PIX_FMT_RGB32_1 || c->dstFormat == PIX_FMT_BGR32_1) ? 8 : 0;
+        rbase = base + (isRgb ? 16 : 0);
+        gbase = base + 8;
+        bbase = base + (isRgb ? 0 : 16);
+        c->yuvTable = av_malloc(1024*3*4);
+        y_table32 = c->yuvTable;
+        yb = -(384<<16) - oy;
+        for(i = 0; i < 1024; i++){
+            uint8_t yval = av_clip_uint8((yb + 0x8000) >> 16);
+            y_table32[i     ] = yval << rbase;
+            y_table32[i+1024] = yval << gbase;
+            y_table32[i+2048] = yval << bbase;
+            yb += cy;
+        }
+        fill_table(c->table_rV, 4, crv, y_table32 + 384        - (oy >> 16));
+        fill_table(c->table_gU, 4, cgu, y_table32 + 384 + 1024 - (oy >> 16));
+        fill_table(c->table_bU, 4, cbu, y_table32 + 384 + 2048 - (oy >> 16));        
+        fill_gv_table(c->table_gV, 4, cgv, 0);
+        break;
+    default:
+        c->yuvTable = NULL;
+        av_log(c, AV_LOG_ERROR, "%ibpp not supported by yuv2rgb\n", bpp);
+        return -1;
+    }
+    return 0;
+}



More information about the ffmpeg-devel mailing list