summaryrefslogtreecommitdiffstats
path: root/libswscale/swscale.c
diff options
context:
space:
mode:
Diffstat (limited to 'libswscale/swscale.c')
-rw-r--r--libswscale/swscale.c682
1 files changed, 355 insertions, 327 deletions
diff --git a/libswscale/swscale.c b/libswscale/swscale.c
index 52d235960b..a7c42c8371 100644
--- a/libswscale/swscale.c
+++ b/libswscale/swscale.c
@@ -116,6 +116,7 @@ unsigned swscale_version(void)
|| (x)==PIX_FMT_RGB555 \
|| (x)==PIX_FMT_GRAY8 \
|| (x)==PIX_FMT_YUV410P \
+ || (x)==PIX_FMT_YUV440P \
|| (x)==PIX_FMT_GRAY16BE \
|| (x)==PIX_FMT_GRAY16LE \
|| (x)==PIX_FMT_YUV444P \
@@ -127,6 +128,8 @@ unsigned swscale_version(void)
|| (x)==PIX_FMT_BGR4_BYTE \
|| (x)==PIX_FMT_RGB4_BYTE \
|| (x)==PIX_FMT_YUV440P \
+ || (x)==PIX_FMT_MONOWHITE \
+ || (x)==PIX_FMT_MONOBLACK \
)
#define isSupportedOut(x) ( \
(x)==PIX_FMT_YUV420P \
@@ -143,6 +146,7 @@ unsigned swscale_version(void)
|| (x)==PIX_FMT_GRAY16LE \
|| (x)==PIX_FMT_GRAY8 \
|| (x)==PIX_FMT_YUV410P \
+ || (x)==PIX_FMT_YUV440P \
)
#define isPacked(x) ( \
(x)==PIX_FMT_PAL8 \
@@ -153,15 +157,15 @@ unsigned swscale_version(void)
)
#define RGB2YUV_SHIFT 15
-#define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
-#define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
-#define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
-#define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
-#define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
-#define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
-#define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
-#define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
-#define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
+#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
+#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
+#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
+#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
+#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
+#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
+#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
+#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];
@@ -258,11 +262,112 @@ static unsigned char clip_table[768];
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
-extern const uint8_t dither_2x2_4[2][8];
-extern const uint8_t dither_2x2_8[2][8];
-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];
+const uint8_t __attribute__((aligned(8))) dither_2x2_4[2][8]={
+{ 1, 3, 1, 3, 1, 3, 1, 3, },
+{ 2, 0, 2, 0, 2, 0, 2, 0, },
+};
+
+const uint8_t __attribute__((aligned(8))) dither_2x2_8[2][8]={
+{ 6, 2, 6, 2, 6, 2, 6, 2, },
+{ 0, 4, 0, 4, 0, 4, 0, 4, },
+};
+
+const uint8_t __attribute__((aligned(8))) dither_8x8_32[8][8]={
+{ 17, 9, 23, 15, 16, 8, 22, 14, },
+{ 5, 29, 3, 27, 4, 28, 2, 26, },
+{ 21, 13, 19, 11, 20, 12, 18, 10, },
+{ 0, 24, 6, 30, 1, 25, 7, 31, },
+{ 16, 8, 22, 14, 17, 9, 23, 15, },
+{ 4, 28, 2, 26, 5, 29, 3, 27, },
+{ 20, 12, 18, 10, 21, 13, 19, 11, },
+{ 1, 25, 7, 31, 0, 24, 6, 30, },
+};
+
+#if 0
+const uint8_t __attribute__((aligned(8))) dither_8x8_64[8][8]={
+{ 0, 48, 12, 60, 3, 51, 15, 63, },
+{ 32, 16, 44, 28, 35, 19, 47, 31, },
+{ 8, 56, 4, 52, 11, 59, 7, 55, },
+{ 40, 24, 36, 20, 43, 27, 39, 23, },
+{ 2, 50, 14, 62, 1, 49, 13, 61, },
+{ 34, 18, 46, 30, 33, 17, 45, 29, },
+{ 10, 58, 6, 54, 9, 57, 5, 53, },
+{ 42, 26, 38, 22, 41, 25, 37, 21, },
+};
+#endif
+
+const uint8_t __attribute__((aligned(8))) dither_8x8_73[8][8]={
+{ 0, 55, 14, 68, 3, 58, 17, 72, },
+{ 37, 18, 50, 32, 40, 22, 54, 35, },
+{ 9, 64, 5, 59, 13, 67, 8, 63, },
+{ 46, 27, 41, 23, 49, 31, 44, 26, },
+{ 2, 57, 16, 71, 1, 56, 15, 70, },
+{ 39, 21, 52, 34, 38, 19, 51, 33, },
+{ 11, 66, 7, 62, 10, 65, 6, 60, },
+{ 48, 30, 43, 25, 47, 29, 42, 24, },
+};
+
+#if 0
+const uint8_t __attribute__((aligned(8))) dither_8x8_128[8][8]={
+{ 68, 36, 92, 60, 66, 34, 90, 58, },
+{ 20, 116, 12, 108, 18, 114, 10, 106, },
+{ 84, 52, 76, 44, 82, 50, 74, 42, },
+{ 0, 96, 24, 120, 6, 102, 30, 126, },
+{ 64, 32, 88, 56, 70, 38, 94, 62, },
+{ 16, 112, 8, 104, 22, 118, 14, 110, },
+{ 80, 48, 72, 40, 86, 54, 78, 46, },
+{ 4, 100, 28, 124, 2, 98, 26, 122, },
+};
+#endif
+
+#if 1
+const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
+{117, 62, 158, 103, 113, 58, 155, 100, },
+{ 34, 199, 21, 186, 31, 196, 17, 182, },
+{144, 89, 131, 76, 141, 86, 127, 72, },
+{ 0, 165, 41, 206, 10, 175, 52, 217, },
+{110, 55, 151, 96, 120, 65, 162, 107, },
+{ 28, 193, 14, 179, 38, 203, 24, 189, },
+{138, 83, 124, 69, 148, 93, 134, 79, },
+{ 7, 172, 48, 213, 3, 168, 45, 210, },
+};
+#elif 1
+// tries to correct a gamma of 1.5
+const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
+{ 0, 143, 18, 200, 2, 156, 25, 215, },
+{ 78, 28, 125, 64, 89, 36, 138, 74, },
+{ 10, 180, 3, 161, 16, 195, 8, 175, },
+{109, 51, 93, 38, 121, 60, 105, 47, },
+{ 1, 152, 23, 210, 0, 147, 20, 205, },
+{ 85, 33, 134, 71, 81, 30, 130, 67, },
+{ 14, 190, 6, 171, 12, 185, 5, 166, },
+{117, 57, 101, 44, 113, 54, 97, 41, },
+};
+#elif 1
+// tries to correct a gamma of 2.0
+const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
+{ 0, 124, 8, 193, 0, 140, 12, 213, },
+{ 55, 14, 104, 42, 66, 19, 119, 52, },
+{ 3, 168, 1, 145, 6, 187, 3, 162, },
+{ 86, 31, 70, 21, 99, 39, 82, 28, },
+{ 0, 134, 11, 206, 0, 129, 9, 200, },
+{ 62, 17, 114, 48, 58, 16, 109, 45, },
+{ 5, 181, 2, 157, 4, 175, 1, 151, },
+{ 95, 36, 78, 26, 90, 34, 74, 24, },
+};
+#else
+// tries to correct a gamma of 2.5
+const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
+{ 0, 107, 3, 187, 0, 125, 6, 212, },
+{ 39, 7, 86, 28, 49, 11, 102, 36, },
+{ 1, 158, 0, 131, 3, 180, 1, 151, },
+{ 68, 19, 52, 12, 81, 25, 64, 17, },
+{ 0, 119, 5, 203, 0, 113, 4, 195, },
+{ 45, 9, 96, 33, 42, 8, 91, 30, },
+{ 2, 172, 1, 144, 2, 165, 0, 137, },
+{ 77, 23, 60, 15, 72, 21, 56, 14, },
+};
+#endif
const char *sws_format_name(enum PixelFormat format)
{
@@ -434,7 +539,7 @@ static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFil
}
}
-#define YSCALE_YUV_2_PACKEDX_C(type) \
+#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type) \
for (i=0; i<(dstW>>1); i++){\
int j;\
int Y1 = 1<<18;\
@@ -458,6 +563,9 @@ static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFil
Y2>>=19;\
U >>=19;\
V >>=19;\
+
+#define YSCALE_YUV_2_PACKEDX_C(type) \
+ YSCALE_YUV_2_PACKEDX_NOCLIP_C(type)\
if ((Y1|Y2|U|V)&256)\
{\
if (Y1>255) Y1=255; \
@@ -470,14 +578,51 @@ static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFil
else if (V<0) V=0; \
}
-#define YSCALE_YUV_2_GRAY16_C(type) \
+#define YSCALE_YUV_2_PACKEDX_FULL_C \
+ for (i=0; i<dstW; i++){\
+ int j;\
+ int Y = 0;\
+ int U = -128<<19;\
+ int V = -128<<19;\
+ int R,G,B;\
+ \
+ for (j=0; j<lumFilterSize; j++){\
+ Y += lumSrc[j][i ] * lumFilter[j];\
+ }\
+ for (j=0; j<chrFilterSize; j++){\
+ U += chrSrc[j][i ] * chrFilter[j];\
+ V += chrSrc[j][i+VOFW] * chrFilter[j];\
+ }\
+ Y >>=10;\
+ U >>=10;\
+ V >>=10;\
+
+#define YSCALE_YUV_2_RGBX_FULL_C(rnd) \
+ YSCALE_YUV_2_PACKEDX_FULL_C\
+ Y-= c->yuv2rgb_y_offset;\
+ Y*= c->yuv2rgb_y_coeff;\
+ Y+= rnd;\
+ R= Y + V*c->yuv2rgb_v2r_coeff;\
+ G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
+ B= Y + U*c->yuv2rgb_u2b_coeff;\
+ if ((R|G|B)&(0xC0000000)){\
+ if (R>=(256<<22)) R=(256<<22)-1; \
+ else if (R<0)R=0; \
+ if (G>=(256<<22)) G=(256<<22)-1; \
+ else if (G<0)G=0; \
+ if (B>=(256<<22)) B=(256<<22)-1; \
+ else if (B<0)B=0; \
+ }\
+
+
+#define YSCALE_YUV_2_GRAY16_C \
for (i=0; i<(dstW>>1); i++){\
int j;\
int Y1 = 1<<18;\
int Y2 = 1<<18;\
int U = 1<<18;\
int V = 1<<18;\
- type av_unused *r, *b, *g;\
+ \
const int i2= 2*i;\
\
for (j=0; j<lumFilterSize; j++)\
@@ -496,7 +641,7 @@ static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFil
}
#define YSCALE_YUV_2_RGBX_C(type) \
- YSCALE_YUV_2_PACKEDX_C(type) \
+ YSCALE_YUV_2_PACKEDX_C(type) /* FIXME fix tables so that cliping is not needed and then use _NOCLIP*/\
r = (type *)c->table_rV[V]; \
g = (type *)(c->table_gU[U] + c->table_gV[V]); \
b = (type *)c->table_bU[U]; \
@@ -558,7 +703,57 @@ static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFil
g = (type *)(c->table_gU[U] + c->table_gV[V]);\
b = (type *)c->table_bU[U];\
-#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16)\
+#define YSCALE_YUV_2_MONO2_C \
+ const uint8_t * const d128=dither_8x8_220[y&7];\
+ uint8_t *g= c->table_gU[128] + c->table_gV[128];\
+ for (i=0; i<dstW-7; i+=8){\
+ int acc;\
+ acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
+ acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
+ acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
+ acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
+ acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
+ acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
+ acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
+ acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
+ ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
+ dest++;\
+ }\
+
+
+#define YSCALE_YUV_2_MONOX_C \
+ const uint8_t * const d128=dither_8x8_220[y&7];\
+ uint8_t *g= c->table_gU[128] + c->table_gV[128];\
+ int acc=0;\
+ for (i=0; i<dstW-1; i+=2){\
+ int j;\
+ int Y1=1<<18;\
+ int Y2=1<<18;\
+\
+ for (j=0; j<lumFilterSize; j++)\
+ {\
+ Y1 += lumSrc[j][i] * lumFilter[j];\
+ Y2 += lumSrc[j][i+1] * lumFilter[j];\
+ }\
+ Y1>>=19;\
+ Y2>>=19;\
+ if ((Y1|Y2)&256)\
+ {\
+ if (Y1>255) Y1=255;\
+ else if (Y1<0)Y1=0;\
+ if (Y2>255) Y2=255;\
+ else if (Y2<0)Y2=0;\
+ }\
+ acc+= acc + g[Y1+d128[(i+0)&7]];\
+ acc+= acc + g[Y2+d128[(i+1)&7]];\
+ if ((i&7)==6){\
+ ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
+ dest++;\
+ }\
+ }
+
+
+#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
switch(c->dstFormat)\
{\
case PIX_FMT_RGB32:\
@@ -656,67 +851,9 @@ static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFil
}\
break;\
case PIX_FMT_MONOBLACK:\
+ case PIX_FMT_MONOWHITE:\
{\
- const uint8_t * const d128=dither_8x8_220[y&7];\
- uint8_t *g= c->table_gU[128] + c->table_gV[128];\
- for (i=0; i<dstW-7; i+=8){\
- int acc;\
- acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
- acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
- acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
- acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
- acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
- acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
- acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
- acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
- ((uint8_t*)dest)[0]= acc;\
- dest++;\
- }\
-\
-/*\
-((uint8_t*)dest)-= dstW>>4;\
-{\
- int acc=0;\
- int left=0;\
- static int top[1024];\
- static int last_new[1024][1024];\
- static int last_in3[1024][1024];\
- static int drift[1024][1024];\
- int topLeft=0;\
- int shift=0;\
- int count=0;\
- const uint8_t * const d128=dither_8x8_220[y&7];\
- int error_new=0;\
- int error_in3=0;\
- int f=0;\
- \
- for (i=dstW>>1; i<dstW; i++){\
- int in= ((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19);\
- int in2 = (76309 * (in - 16) + 32768) >> 16;\
- int in3 = (in2 < 0) ? 0 : ((in2 > 255) ? 255 : in2);\
- int old= (left*7 + topLeft + top[i]*5 + top[i+1]*3)/20 + in3\
- + (last_new[y][i] - in3)*f/256;\
- int new= old> 128 ? 255 : 0;\
-\
- error_new+= FFABS(last_new[y][i] - new);\
- error_in3+= FFABS(last_in3[y][i] - in3);\
- f= error_new - error_in3*4;\
- if (f<0) f=0;\
- if (f>256) f=256;\
-\
- topLeft= top[i];\
- left= top[i]= old - new;\
- last_new[y][i]= new;\
- last_in3[y][i]= in3;\
-\
- acc+= acc + (new&1);\
- if ((i&7)==6){\
- ((uint8_t*)dest)[0]= acc;\
- ((uint8_t*)dest)++;\
- }\
- }\
-}\
-*/\
+ func_monoblack\
}\
break;\
case PIX_FMT_YUYV422:\
@@ -759,171 +896,51 @@ static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **l
uint8_t *dest, int dstW, int y)
{
int i;
- switch(c->dstFormat)
- {
- case PIX_FMT_BGR32:
- case PIX_FMT_RGB32:
- case PIX_FMT_BGR32_1:
- case PIX_FMT_RGB32_1:
- YSCALE_YUV_2_RGBX_C(uint32_t)
- ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];
- ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];
- }
- break;
+ YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)
+}
+
+static inline void yuv2rgbXinC_full(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
+ int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
+ uint8_t *dest, int dstW, int y)
+{
+ int i;
+ int step= fmt_depth(c->dstFormat)/8;
+ int aidx= 3;
+
+ switch(c->dstFormat){
+ case PIX_FMT_ARGB:
+ dest++;
+ aidx= 0;
case PIX_FMT_RGB24:
- YSCALE_YUV_2_RGBX_C(uint8_t)
- ((uint8_t*)dest)[0]= r[Y1];
- ((uint8_t*)dest)[1]= g[Y1];
- ((uint8_t*)dest)[2]= b[Y1];
- ((uint8_t*)dest)[3]= r[Y2];
- ((uint8_t*)dest)[4]= g[Y2];
- ((uint8_t*)dest)[5]= b[Y2];
- dest+=6;
+ aidx--;
+ case PIX_FMT_RGBA:
+ YSCALE_YUV_2_RGBX_FULL_C(1<<21)
+ dest[aidx]= 0;
+ dest[0]= R>>22;
+ dest[1]= G>>22;
+ dest[2]= B>>22;
+ dest+= step;
}
break;
+ case PIX_FMT_ABGR:
+ dest++;
+ aidx= 0;
case PIX_FMT_BGR24:
- YSCALE_YUV_2_RGBX_C(uint8_t)
- ((uint8_t*)dest)[0]= b[Y1];
- ((uint8_t*)dest)[1]= g[Y1];
- ((uint8_t*)dest)[2]= r[Y1];
- ((uint8_t*)dest)[3]= b[Y2];
- ((uint8_t*)dest)[4]= g[Y2];
- ((uint8_t*)dest)[5]= r[Y2];
- dest+=6;
- }
- break;
- case PIX_FMT_RGB565:
- case PIX_FMT_BGR565:
- {
- const int dr1= dither_2x2_8[y&1 ][0];
- const int dg1= dither_2x2_4[y&1 ][0];
- const int db1= dither_2x2_8[(y&1)^1][0];
- const int dr2= dither_2x2_8[y&1 ][1];
- const int dg2= dither_2x2_4[y&1 ][1];
- const int db2= dither_2x2_8[(y&1)^1][1];
- YSCALE_YUV_2_RGBX_C(uint16_t)
- ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];
- ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];
- }
- }
- break;
- case PIX_FMT_RGB555:
- case PIX_FMT_BGR555:
- {
- const int dr1= dither_2x2_8[y&1 ][0];
- const int dg1= dither_2x2_8[y&1 ][1];
- const int db1= dither_2x2_8[(y&1)^1][0];
- const int dr2= dither_2x2_8[y&1 ][1];
- const int dg2= dither_2x2_8[y&1 ][0];
- const int db2= dither_2x2_8[(y&1)^1][1];
- YSCALE_YUV_2_RGBX_C(uint16_t)
- ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];
- ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];
- }
- }
- break;
- case PIX_FMT_RGB8:
- case PIX_FMT_BGR8:
- {
- const uint8_t * const d64= dither_8x8_73[y&7];
- const uint8_t * const d32= dither_8x8_32[y&7];
- YSCALE_YUV_2_RGBX_C(uint8_t)
- ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];
- ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];
- }
- }
- break;
- case PIX_FMT_RGB4:
- case PIX_FMT_BGR4:
- {
- const uint8_t * const d64= dither_8x8_73 [y&7];
- const uint8_t * const d128=dither_8x8_220[y&7];
- YSCALE_YUV_2_RGBX_C(uint8_t)
- ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]
- +((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);
- }
- }
- break;
- case PIX_FMT_RGB4_BYTE:
- case PIX_FMT_BGR4_BYTE:
- {
- const uint8_t * const d64= dither_8x8_73 [y&7];
- const uint8_t * const d128=dither_8x8_220[y&7];
- YSCALE_YUV_2_RGBX_C(uint8_t)
- ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];
- ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];
- }
- }
- break;
- case PIX_FMT_MONOBLACK:
- {
- const uint8_t * const d128=dither_8x8_220[y&7];
- uint8_t *g= c->table_gU[128] + c->table_gV[128];
- int acc=0;
- for (i=0; i<dstW-1; i+=2){
- int j;
- int Y1=1<<18;
- int Y2=1<<18;
-
- for (j=0; j<lumFilterSize; j++)
- {
- Y1 += lumSrc[j][i] * lumFilter[j];
- Y2 += lumSrc[j][i+1] * lumFilter[j];
- }
- Y1>>=19;
- Y2>>=19;
- if ((Y1|Y2)&256)
- {
- if (Y1>255) Y1=255;
- else if (Y1<0)Y1=0;
- if (Y2>255) Y2=255;
- else if (Y2<0)Y2=0;
- }
- acc+= acc + g[Y1+d128[(i+0)&7]];
- acc+= acc + g[Y2+d128[(i+1)&7]];
- if ((i&7)==6){
- ((uint8_t*)dest)[0]= acc;
- dest++;
- }
- }
- }
- break;
- case PIX_FMT_YUYV422:
- YSCALE_YUV_2_PACKEDX_C(void)
- ((uint8_t*)dest)[2*i2+0]= Y1;
- ((uint8_t*)dest)[2*i2+1]= U;
- ((uint8_t*)dest)[2*i2+2]= Y2;
- ((uint8_t*)dest)[2*i2+3]= V;
- }
- break;
- case PIX_FMT_UYVY422:
- YSCALE_YUV_2_PACKEDX_C(void)
- ((uint8_t*)dest)[2*i2+0]= U;
- ((uint8_t*)dest)[2*i2+1]= Y1;
- ((uint8_t*)dest)[2*i2+2]= V;
- ((uint8_t*)dest)[2*i2+3]= Y2;
- }
- break;
- case PIX_FMT_GRAY16BE:
- YSCALE_YUV_2_GRAY16_C(void)
- ((uint8_t*)dest)[2*i2+0]= Y1>>8;
- ((uint8_t*)dest)[2*i2+1]= Y1;
- ((uint8_t*)dest)[2*i2+2]= Y2>>8;
- ((uint8_t*)dest)[2*i2+3]= Y2;
- }
- break;
- case PIX_FMT_GRAY16LE:
- YSCALE_YUV_2_GRAY16_C(void)
- ((uint8_t*)dest)[2*i2+0]= Y1;
- ((uint8_t*)dest)[2*i2+1]= Y1>>8;
- ((uint8_t*)dest)[2*i2+2]= Y2;
- ((uint8_t*)dest)[2*i2+3]= Y2>>8;
+ aidx--;
+ case PIX_FMT_BGRA:
+ YSCALE_YUV_2_RGBX_FULL_C(1<<21)
+ dest[aidx]= 0;
+ dest[0]= B>>22;
+ dest[1]= G>>22;
+ dest[2]= R>>22;
+ dest+= step;
}
break;
+ default:
+ assert(0);
}
}
-
//Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
//Plain C versions
#if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) || !defined(CONFIG_GPL)
@@ -1036,8 +1053,9 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
int filterSize;
int filter2Size;
int minFilterSize;
- double *filter=NULL;
- double *filter2=NULL;
+ int64_t *filter=NULL;
+ int64_t *filter2=NULL;
+ const int64_t fone= 1LL<<54;
int ret= -1;
#if defined(ARCH_X86)
if (flags & SWS_CPU_CAPS_MMX)
@@ -1051,12 +1069,11 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
{
int i;
filterSize= 1;
- filter= av_malloc(dstW*sizeof(double)*filterSize);
- for (i=0; i<dstW*filterSize; i++) filter[i]=0;
+ filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
for (i=0; i<dstW; i++)
{
- filter[i*filterSize]=1;
+ filter[i*filterSize]= fone;
(*filterPos)[i]=i;
}
@@ -1066,7 +1083,7 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
int i;
int xDstInSrc;
filterSize= 1;
- filter= av_malloc(dstW*sizeof(double)*filterSize);
+ filter= av_malloc(dstW*sizeof(*filter)*filterSize);
xDstInSrc= xInc/2 - 0x8000;
for (i=0; i<dstW; i++)
@@ -1074,7 +1091,7 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
(*filterPos)[i]= xx;
- filter[i]= 1.0;
+ filter[i]= fone;
xDstInSrc+= xInc;
}
}
@@ -1085,7 +1102,7 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
if (flags&SWS_BICUBIC) filterSize= 4;
else if (flags&SWS_X ) filterSize= 4;
else filterSize= 2; // SWS_BILINEAR / SWS_AREA
- filter= av_malloc(dstW*sizeof(double)*filterSize);
+ filter= av_malloc(dstW*sizeof(*filter)*filterSize);
xDstInSrc= xInc/2 - 0x8000;
for (i=0; i<dstW; i++)
@@ -1097,8 +1114,7 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
//Bilinear upscale / linear interpolate / Area averaging
for (j=0; j<filterSize; j++)
{
- double d= FFABS((xx<<16) - xDstInSrc)/(double)(1<<16);
- double coeff= 1.0 - d;
+ int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
if (coeff<0) coeff=0;
filter[i*filterSize + j]= coeff;
xx++;
@@ -1108,52 +1124,59 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
}
else
{
- double xDstInSrc;
- double sizeFactor, filterSizeInSrc;
- const double xInc1= (double)xInc / (double)(1<<16);
-
- if (flags&SWS_BICUBIC) sizeFactor= 4.0;
- else if (flags&SWS_X) sizeFactor= 8.0;
- else if (flags&SWS_AREA) sizeFactor= 1.0; //downscale only, for upscale it is bilinear
- else if (flags&SWS_GAUSS) sizeFactor= 8.0; // infinite ;)
- else if (flags&SWS_LANCZOS) sizeFactor= param[0] != SWS_PARAM_DEFAULT ? 2.0*param[0] : 6.0;
- else if (flags&SWS_SINC) sizeFactor= 20.0; // infinite ;)
- else if (flags&SWS_SPLINE) sizeFactor= 20.0; // infinite ;)
- else if (flags&SWS_BILINEAR) sizeFactor= 2.0;
+ int xDstInSrc;
+ int sizeFactor;
+
+ if (flags&SWS_BICUBIC) sizeFactor= 4;
+ else if (flags&SWS_X) sizeFactor= 8;
+ else if (flags&SWS_AREA) sizeFactor= 1; //downscale only, for upscale it is bilinear
+ else if (flags&SWS_GAUSS) sizeFactor= 8; // infinite ;)
+ else if (flags&SWS_LANCZOS) sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
+ else if (flags&SWS_SINC) sizeFactor= 20; // infinite ;)
+ else if (flags&SWS_SPLINE) sizeFactor= 20; // infinite ;)
+ else if (flags&SWS_BILINEAR) sizeFactor= 2;
else {
- sizeFactor= 0.0; //GCC warning killer
+ sizeFactor= 0; //GCC warning killer
assert(0);
}
- if (xInc1 <= 1.0) filterSizeInSrc= sizeFactor; // upscale
- else filterSizeInSrc= sizeFactor*srcW / (double)dstW;
+ if (xInc <= 1<<16) filterSize= 1 + sizeFactor; // upscale
+ else filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
- filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible
if (filterSize > srcW-2) filterSize=srcW-2;
- filter= av_malloc(dstW*sizeof(double)*filterSize);
+ filter= av_malloc(dstW*sizeof(*filter)*filterSize);
- xDstInSrc= xInc1 / 2.0 - 0.5;
+ xDstInSrc= xInc - 0x10000;
for (i=0; i<dstW; i++)
{
- int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);
+ int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
int j;
(*filterPos)[i]= xx;
for (j=0; j<filterSize; j++)
{
- double d= FFABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;
- double coeff;
+ int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
+ double floatd;
+ int64_t coeff;
+
+ if (xInc > 1<<16)
+ d= d*dstW/srcW;
+ floatd= d * (1.0/(1<<30));
+
if (flags & SWS_BICUBIC)
{
- double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;
- double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;
-
- if (d<1.0)
- coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;
- else if (d<2.0)
- coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;
+ int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1<<24);
+ int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
+ int64_t dd = ( d*d)>>30;
+ int64_t ddd= (dd*d)>>30;
+
+ if (d < 1LL<<30)
+ coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
+ else if (d < 1LL<<31)
+ coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
else
coeff=0.0;
+ coeff *= fone>>(30+24);
}
/* else if (flags & SWS_X)
{
@@ -1164,46 +1187,49 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
else if (flags & SWS_X)
{
double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
+ double c;
- if (d<1.0)
- coeff = cos(d*PI);
+ if (floatd<1.0)
+ c = cos(floatd*PI);
else
- coeff=-1.0;
- if (coeff<0.0) coeff= -pow(-coeff, A);
- else coeff= pow( coeff, A);
- coeff= coeff*0.5 + 0.5;
+ c=-1.0;
+ if (c<0.0) c= -pow(-c, A);
+ else c= pow( c, A);
+ coeff= (c*0.5 + 0.5)*fone;
}
else if (flags & SWS_AREA)
{
- double srcPixelSize= 1.0/xInc1;
- if (d + srcPixelSize/2 < 0.5) coeff= 1.0;
- else if (d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5;
+ int64_t d2= d - (1<<29);
+ if (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
+ else if (d2*xInc < (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
else coeff=0.0;
+ coeff *= fone>>(30+16);
}
else if (flags & SWS_GAUSS)
{
double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
- coeff = pow(2.0, - p*d*d);
+ coeff = (pow(2.0, - p*floatd*floatd))*fone;
}
else if (flags & SWS_SINC)
{
- coeff = d ? sin(d*PI)/(d*PI) : 1.0;
+ coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
}
else if (flags & SWS_LANCZOS)
{
double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
- coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;
- if (d>p) coeff=0;
+ coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
+ if (floatd>p) coeff=0;
}
else if (flags & SWS_BILINEAR)
{
- coeff= 1.0 - d;
+ coeff= (1<<30) - d;
if (coeff<0) coeff=0;
+ coeff *= fone >> 30;
}
else if (flags & SWS_SPLINE)
{
double p=-2.196152422706632;
- coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);
+ coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d) * fone;
}
else {
coeff= 0.0; //GCC warning killer
@@ -1213,7 +1239,7 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
filter[i*filterSize + j]= coeff;
xx++;
}
- xDstInSrc+= xInc1;
+ xDstInSrc+= 2*xInc;
}
}
@@ -1225,31 +1251,24 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
if (srcFilter) filter2Size+= srcFilter->length - 1;
if (dstFilter) filter2Size+= dstFilter->length - 1;
assert(filter2Size>0);
- filter2= av_malloc(filter2Size*dstW*sizeof(double));
+ filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
for (i=0; i<dstW; i++)
{
- int j;
- SwsVector scaleFilter;
- SwsVector *outVec;
-
- scaleFilter.coeff= filter + i*filterSize;
- scaleFilter.length= filterSize;
-
- if (srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);
- else outVec= &scaleFilter;
-
- assert(outVec->length == filter2Size);
- //FIXME dstFilter
+ int j, k;
- for (j=0; j<outVec->length; j++)
- {
- filter2[i*filter2Size + j]= outVec->coeff[j];
+ if(srcFilter){
+ for (k=0; k<srcFilter->length; k++){
+ for (j=0; j<filterSize; j++)
+ filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
+ }
+ }else{
+ for (j=0; j<filterSize; j++)
+ filter2[i*filter2Size + j]= filter[i*filterSize + j];
}
+ //FIXME dstFilter
(*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
-
- if (outVec != &scaleFilter) sws_freeVec(outVec);
}
av_freep(&filter);
@@ -1260,7 +1279,7 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
{
int min= filter2Size;
int j;
- double cutOff=0.0;
+ int64_t cutOff=0.0;
/* get rid off near zero elements on the left by shifting left */
for (j=0; j<filter2Size; j++)
@@ -1268,7 +1287,7 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
int k;
cutOff += FFABS(filter2[i*filter2Size]);
- if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;
+ if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
/* preserve monotonicity because the core can't handle the filter otherwise */
if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
@@ -1276,17 +1295,17 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
// Move filter coeffs left
for (k=1; k<filter2Size; k++)
filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
- filter2[i*filter2Size + k - 1]= 0.0;
+ filter2[i*filter2Size + k - 1]= 0;
(*filterPos)[i]++;
}
- cutOff=0.0;
+ cutOff=0;
/* count near zeros on the right */
for (j=filter2Size-1; j>0; j--)
{
cutOff += FFABS(filter2[i*filter2Size + j]);
- if (cutOff > SWS_MAX_REDUCE_CUTOFF) break;
+ if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
min--;
}
@@ -1317,7 +1336,7 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
assert(minFilterSize > 0);
filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
assert(filterSize > 0);
- filter= av_malloc(filterSize*dstW*sizeof(double));
+ filter= av_malloc(filterSize*dstW*sizeof(*filter));
if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
goto error;
*outFilterSize= filterSize;
@@ -1331,8 +1350,10 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
for (j=0; j<filterSize; j++)
{
- if (j>=filter2Size) filter[i*filterSize + j]= 0.0;
+ if (j>=filter2Size) filter[i*filterSize + j]= 0;
else filter[i*filterSize + j]= filter2[i*filter2Size + j];
+ if((flags & SWS_BITEXACT) && j>=minFilterSize)
+ filter[i*filterSize + j]= 0;
}
}
@@ -1377,21 +1398,20 @@ static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outF
for (i=0; i<dstW; i++)
{
int j;
- double error=0;
- double sum=0;
- double scale= one;
+ int64_t error=0;
+ int64_t sum=0;
for (j=0; j<filterSize; j++)
{
sum+= filter[i*filterSize + j];
}
- scale/= sum;
+ sum= (sum + one/2)/ one;
for (j=0; j<*outFilterSize; j++)
{
- double v= filter[i*filterSize + j]*scale + error;
- int intV= floor(v + 0.5);
+ int64_t v= filter[i*filterSiz