Merge remote branch 'qatar/master'

[ffmpeg] / libswscale / swscale.c

diff --git a/libswscale/swscale.c b/libswscale/swscale.c
index b72ed79ae9794544a3bda2382c9fb549db31f7a8..7be3344b4f2204a85a374a6cc541ecc694d5b22c 100644 (file)
--- a/libswscale/swscale.c
+++ b/libswscale/swscale.c
@@ -266,6 +266,81 @@ DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 };
 #endif
 
+DECLARE_ALIGNED(8, const uint8_t, dithers)[8][8][8]={
+{
+  {   0,  1,  0,  1,  0,  1,  0,  1,},
+  {   1,  0,  1,  0,  1,  0,  1,  0,},
+  {   0,  1,  0,  1,  0,  1,  0,  1,},
+  {   1,  0,  1,  0,  1,  0,  1,  0,},
+  {   0,  1,  0,  1,  0,  1,  0,  1,},
+  {   1,  0,  1,  0,  1,  0,  1,  0,},
+  {   0,  1,  0,  1,  0,  1,  0,  1,},
+  {   1,  0,  1,  0,  1,  0,  1,  0,},
+},{
+  {   1,  2,  1,  2,  1,  2,  1,  2,},
+  {   3,  0,  3,  0,  3,  0,  3,  0,},
+  {   1,  2,  1,  2,  1,  2,  1,  2,},
+  {   3,  0,  3,  0,  3,  0,  3,  0,},
+  {   1,  2,  1,  2,  1,  2,  1,  2,},
+  {   3,  0,  3,  0,  3,  0,  3,  0,},
+  {   1,  2,  1,  2,  1,  2,  1,  2,},
+  {   3,  0,  3,  0,  3,  0,  3,  0,},
+},{
+  {   2,  4,  3,  5,  2,  4,  3,  5,},
+  {   6,  0,  7,  1,  6,  0,  7,  1,},
+  {   3,  5,  2,  4,  3,  5,  2,  4,},
+  {   7,  1,  6,  0,  7,  1,  6,  0,},
+  {   2,  4,  3,  5,  2,  4,  3,  5,},
+  {   6,  0,  7,  1,  6,  0,  7,  1,},
+  {   3,  5,  2,  4,  3,  5,  2,  4,},
+  {   7,  1,  6,  0,  7,  1,  6,  0,},
+},{
+  {   4,  8,  7, 11,  4,  8,  7, 11,},
+  {  12,  0, 15,  3, 12,  0, 15,  3,},
+  {   6, 10,  5,  9,  6, 10,  5,  9,},
+  {  14,  2, 13,  1, 14,  2, 13,  1,},
+  {   4,  8,  7, 11,  4,  8,  7, 11,},
+  {  12,  0, 15,  3, 12,  0, 15,  3,},
+  {   6, 10,  5,  9,  6, 10,  5,  9,},
+  {  14,  2, 13,  1, 14,  2, 13,  1,},
+},{
+  {   9, 17, 15, 23,  8, 16, 14, 22,},
+  {  25,  1, 31,  7, 24,  0, 30,  6,},
+  {  13, 21, 11, 19, 12, 20, 10, 18,},
+  {  29,  5, 27,  3, 28,  4, 26,  2,},
+  {   8, 16, 14, 22,  9, 17, 15, 23,},
+  {  24,  0, 30,  6, 25,  1, 31,  7,},
+  {  12, 20, 10, 18, 13, 21, 11, 19,},
+  {  28,  4, 26,  2, 29,  5, 27,  3,},
+},{
+  {  18, 34, 30, 46, 17, 33, 29, 45,},
+  {  50,  2, 62, 14, 49,  1, 61, 13,},
+  {  26, 42, 22, 38, 25, 41, 21, 37,},
+  {  58, 10, 54,  6, 57,  9, 53,  5,},
+  {  16, 32, 28, 44, 19, 35, 31, 47,},
+  {  48,  0, 60, 12, 51,  3, 63, 15,},
+  {  24, 40, 20, 36, 27, 43, 23, 39,},
+  {  56,  8, 52,  4, 59, 11, 55,  7,},
+},{
+  {  18, 34, 30, 46, 17, 33, 29, 45,},
+  {  50,  2, 62, 14, 49,  1, 61, 13,},
+  {  26, 42, 22, 38, 25, 41, 21, 37,},
+  {  58, 10, 54,  6, 57,  9, 53,  5,},
+  {  16, 32, 28, 44, 19, 35, 31, 47,},
+  {  48,  0, 60, 12, 51,  3, 63, 15,},
+  {  24, 40, 20, 36, 27, 43, 23, 39,},
+  {  56,  8, 52,  4, 59, 11, 55,  7,},
+},{
+  {  36, 68, 60, 92, 34, 66, 58, 90,},
+  { 100,  4,124, 28, 98,  2,122, 26,},
+  {  52, 84, 44, 76, 50, 82, 42, 74,},
+  { 116, 20,108, 12,114, 18,106, 10,},
+  {  32, 64, 56, 88, 38, 70, 62, 94,},
+  {  96,  0,120, 24,102,  6,126, 30,},
+  {  48, 80, 40, 72, 54, 86, 46, 78,},
+  { 112, 16,104,  8,118, 22,110, 14,},
+}};
+
 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
                                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
                                                     const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
@@ -1657,7 +1732,7 @@ static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[],
         if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
             dstPtr += ALT32_CORR;
 
-        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
+        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0 && !(srcStride[0]%srcBpp))
             conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
         else {
             int i;
@@ -1761,15 +1836,26 @@ static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[
                         srcPtr2 += srcStride[plane]/2;
                     }
                 } else {
-                    // FIXME Maybe dither instead.
                     for (i = 0; i < height; i++) {
-                        for (j = 0; j < length; j++)
-                            dstPtr[j] = srcPtr2[j]>>(depth-8);
+                        uint8_t *dither= dithers[depth-9][i&7];
+                        for (j = 0; j < length-7; j+=8){
+                            dstPtr[j+0] = (srcPtr2[j+0] + dither[0])>>(depth-8);
+                            dstPtr[j+1] = (srcPtr2[j+1] + dither[1])>>(depth-8);
+                            dstPtr[j+2] = (srcPtr2[j+2] + dither[2])>>(depth-8);
+                            dstPtr[j+3] = (srcPtr2[j+3] + dither[3])>>(depth-8);
+                            dstPtr[j+4] = (srcPtr2[j+4] + dither[4])>>(depth-8);
+                            dstPtr[j+5] = (srcPtr2[j+5] + dither[5])>>(depth-8);
+                            dstPtr[j+6] = (srcPtr2[j+6] + dither[6])>>(depth-8);
+                            dstPtr[j+7] = (srcPtr2[j+7] + dither[7])>>(depth-8);
+                        }
+                        for (; j < length; j++)
+                            dstPtr[j] = (srcPtr2[j] + dither[j&7])>>(depth-8);
                         dstPtr  += dstStride[plane];
                         srcPtr2 += srcStride[plane]/2;
                     }
                 }
             } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
+                //FIXME add dither
                 if (!isBE(c->srcFormat)) srcPtr++;
                 for (i=0; i<height; i++) {
                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];