Skip to content

F
ffmpeg.wasm-core
Commit cf7d1c1a authored by Michael Niedermayer's avatar Michael Niedermayer
Browse files
Options

use unified yuv2rgb init 

{RGB,BGR}{1,4,8,15,16,24,32} output supported

Originally committed as revision 6579 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc
parent 7c307827
Show whitespace changes
Inline Side-by-side
Showing with 449 additions and 615 deletions
postproc/swscale.c
View file @ cf7d1c1a
...@@ -18,7 +18,7 @@ ...@@ -18,7 +18,7 @@
/* /*
supported Input formats: YV12, I420/IYUV, YUY2, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09 supported Input formats: YV12, I420/IYUV, YUY2, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09
supported output formats: YV12, I420/IYUV, BGR15, BGR16, BGR24, BGR32, Y8/Y800, YVU9/IF09 supported output formats: YV12, I420/IYUV, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
BGR15/16 support dithering BGR15/16 support dithering
unscaled special converters unscaled special converters
...@@ -103,15 +103,15 @@ untested special converters ...@@ -103,15 +103,15 @@ untested special converters
#define isPlanarYUV(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 || (x)==IMGFMT_YVU9) #define isPlanarYUV(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 || (x)==IMGFMT_YVU9)
#define isYUV(x) ((x)==IMGFMT_YUY2 || isPlanarYUV(x)) #define isYUV(x) ((x)==IMGFMT_YUY2 || isPlanarYUV(x))
#define isGray(x) ((x)==IMGFMT_Y800) #define isGray(x) ((x)==IMGFMT_Y800)
#define isRGB(x) (((x)&IMGFMT_RGB_MASK)==IMGFMT_RGB)
#define isBGR(x) (((x)&IMGFMT_BGR_MASK)==IMGFMT_BGR)
#define isSupportedIn(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 || (x)==IMGFMT_YUY2 \ #define isSupportedIn(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 || (x)==IMGFMT_YUY2 \
|| (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\ || (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\
|| (x)==IMGFMT_RGB32|| (x)==IMGFMT_RGB24\ || (x)==IMGFMT_RGB32|| (x)==IMGFMT_RGB24\
|| (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9) || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9)
#define isSupportedOut(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 \ #define isSupportedOut(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 \
|| (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\ || isRGB(x) || isBGR(x)\
|| (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9) || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9)
#define isRGB(x) (((x)&IMGFMT_RGB_MASK)==IMGFMT_RGB)
#define isBGR(x) (((x)&IMGFMT_BGR_MASK)==IMGFMT_BGR)
#define isPacked(x) ((x)==IMGFMT_YUY2 || isRGB(x) || isBGR(x)) #define isPacked(x) ((x)==IMGFMT_YUY2 || isRGB(x) || isBGR(x))
#define RGB2YUV_SHIFT 16 #define RGB2YUV_SHIFT 16
...@@ -208,26 +208,6 @@ static const uint64_t w1111 __attribute__((aligned(8))) = 0x00010001000100 ...@@ -208,26 +208,6 @@ static const uint64_t w1111 __attribute__((aligned(8))) = 0x00010001000100
// clipping helper table for C implementations: // clipping helper table for C implementations:
static unsigned char clip_table[768]; static unsigned char clip_table[768];
static unsigned short clip_table16b[768];
static unsigned short clip_table16g[768];
static unsigned short clip_table16r[768];
static unsigned short clip_table15b[768];
static unsigned short clip_table15g[768];
static unsigned short clip_table15r[768];
// yuv->rgb conversion tables:
static int yuvtab_2568[256];
static int yuvtab_3343[256];
static int yuvtab_0c92[256];
static int yuvtab_1a1e[256];
static int yuvtab_40cf[256];
// Needed for cubic scaler to catch overflows
static int clip_yuvtab_2568[768];
static int clip_yuvtab_3343[768];
static int clip_yuvtab_0c92[768];
static int clip_yuvtab_1a1e[768];
static int clip_yuvtab_40cf[768];
//global sws_flags from the command line //global sws_flags from the command line
int sws_flags=2; int sws_flags=2;
...@@ -250,7 +230,13 @@ void (*swScale)(SwsContext *context, uint8_t* src[], int srcStride[], int srcSli ...@@ -250,7 +230,13 @@ void (*swScale)(SwsContext *context, uint8_t* src[], int srcStride[], int srcSli
static SwsVector *getConvVec(SwsVector *a, SwsVector *b); static SwsVector *getConvVec(SwsVector *a, SwsVector *b);
static inline void orderYUV(int format, uint8_t * sortedP[], int sortedStride[], uint8_t * p[], int stride[]); static inline void orderYUV(int format, uint8_t * sortedP[], int sortedStride[], uint8_t * p[], int stride[]);
void *yuv2rgb_c_init (unsigned bpp, int mode, void *table_rV[256], void *table_gU[256], int table_gV[256], void *table_bU[256]);
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];
#ifdef CAN_COMPILE_X86_ASM #ifdef CAN_COMPILE_X86_ASM
void in_asm_used_var_warning_killer() void in_asm_used_var_warning_killer()
...@@ -414,214 +400,356 @@ static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilt ...@@ -414,214 +400,356 @@ static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilt
} }
} }
static inline void yuv2rgbXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, #define YSCALE_YUV_2_RGBX_C(type) \
for(i=0; i<(dstW>>1); i++){\
int j;\
int Y1=0;\
int Y2=0;\
int U=0;\
int V=0;\
type *r, *b, *g;\
const int i2= 2*i;\
\
for(j=0; j<lumFilterSize; j++)\
{\
Y1 += lumSrc[j][i2] * lumFilter[j];\
Y2 += lumSrc[j][i2+1] * lumFilter[j];\
}\
for(j=0; j<chrFilterSize; j++)\
{\
U += chrSrc[j][i] * chrFilter[j];\
V += chrSrc[j][i+2048] * chrFilter[j];\
}\
Y1>>=19;\
Y2>>=19;\
U >>=19;\
V >>=19;\
if((Y1|Y2|U|V)&256)\
{\
if(Y1>255) Y1=255;\
else if(Y1<0)Y1=0;\
if(Y2>255) Y2=255;\
else if(Y2<0)Y2=0;\
if(U>255) U=255;\
else if(U<0) U=0;\
if(V>255) V=255;\
else if(V<0) V=0;\
}\
r = c->table_rV[V];\
g = c->table_gU[U] + c->table_gV[V];\
b = c->table_bU[U];\
#define YSCALE_YUV_2_RGB2_C(type) \
for(i=0; i<(dstW>>1); i++){\
const int i2= 2*i;\
int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19;\
int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;\
int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19;\
int V= (uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19;\
type *r, *b, *g;\
r = c->table_rV[V];\
g = c->table_gU[U] + c->table_gV[V];\
b = c->table_bU[U];\
#define YSCALE_YUV_2_RGB1_C(type) \
for(i=0; i<(dstW>>1); i++){\
const int i2= 2*i;\
int Y1= buf0[i2 ]>>7;\
int Y2= buf0[i2+1]>>7;\
int U= (uvbuf1[i ])>>7;\
int V= (uvbuf1[i+2048])>>7;\
type *r, *b, *g;\
r = c->table_rV[V];\
g = c->table_gU[U] + c->table_gV[V];\
b = c->table_bU[U];\
#define YSCALE_YUV_2_RGB1B_C(type) \
for(i=0; i<(dstW>>1); i++){\
const int i2= 2*i;\
int Y1= buf0[i2 ]>>7;\
int Y2= buf0[i2+1]>>7;\
int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
int V= (uvbuf0[i+2048] + uvbuf1[i+2048])>>8;\
type *r, *b, *g;\
r = c->table_rV[V];\
g = c->table_gU[U] + c->table_gV[V];\
b = c->table_bU[U];\
#define YSCALE_YUV_2_ANYRGB_C(func)\
switch(c->dstFormat)\
{\
case IMGFMT_BGR32:\
case IMGFMT_RGB32:\
func(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;\
case IMGFMT_RGB24:\
func(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];\
((uint8_t*)dest)+=6;\
}\
break;\
case IMGFMT_BGR24:\
func(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];\
((uint8_t*)dest)+=6;\
}\
break;\
case IMGFMT_RGB16:\
case IMGFMT_BGR16:\
{\
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];\
func(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 IMGFMT_RGB15:\
case IMGFMT_BGR15:\
{\
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];\
func(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 IMGFMT_RGB8:\
case IMGFMT_BGR8:\
{\
const uint8_t * const d64= dither_8x8_73[y&7];\
const uint8_t * const d32= dither_8x8_32[y&7];\
func(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 IMGFMT_RGB4:\
case IMGFMT_BGR4:\
{\
const uint8_t * const d64= dither_8x8_73 [y&7];\
const uint8_t * const d128=dither_8x8_220[y&7];\
func(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 IMGFMT_RGB1:\
case IMGFMT_BGR1:\
{\
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;\
((uint8_t*)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+= ABS(last_new[y][i] - new);\
error_in3+= ABS(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)++;\
}\
}\
}\
*/\
}\
break;\
}\
static inline void yuv2rgbXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
uint8_t *dest, int dstW, int dstFormat) uint8_t *dest, int dstW, int y)
{ {
if(dstFormat==IMGFMT_BGR32)
{
int i; int i;
#ifdef WORDS_BIGENDIAN switch(c->dstFormat)
dest++;
#endif
for(i=0; i<(dstW>>1); i++){
int j;
int Y1=0;
int Y2=0;
int U=0;
int V=0;
int Cb, Cr, Cg;
for(j=0; j<lumFilterSize; j++)
{ {
Y1 += lumSrc[j][2*i] * lumFilter[j]; case IMGFMT_RGB32:
Y2 += lumSrc[j][2*i+1] * lumFilter[j]; case IMGFMT_BGR32:
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];
} }
for(j=0; j<chrFilterSize; j++) break;
{ case IMGFMT_RGB24:
U += chrSrc[j][i] * chrFilter[j]; YSCALE_YUV_2_RGBX_C(uint8_t)
V += chrSrc[j][i+2048] * chrFilter[j]; ((uint8_t*)dest)[0]= r[Y1];
} ((uint8_t*)dest)[1]= g[Y1];
Y1= clip_yuvtab_2568[ (Y1>>19) + 256 ]; ((uint8_t*)dest)[2]= b[Y1];
Y2= clip_yuvtab_2568[ (Y2>>19) + 256 ]; ((uint8_t*)dest)[3]= r[Y2];
U >>= 19; ((uint8_t*)dest)[4]= g[Y2];
V >>= 19; ((uint8_t*)dest)[5]= b[Y2];
((uint8_t*)dest)+=6;
Cb= clip_yuvtab_40cf[U+ 256];
Cg= clip_yuvtab_1a1e[V+ 256] + yuvtab_0c92[U+ 256];
Cr= clip_yuvtab_3343[V+ 256];
dest[8*i+0]=clip_table[((Y1 + Cb) >>13)];
dest[8*i+1]=clip_table[((Y1 + Cg) >>13)];
dest[8*i+2]=clip_table[((Y1 + Cr) >>13)];
dest[8*i+4]=clip_table[((Y2 + Cb) >>13)];
dest[8*i+5]=clip_table[((Y2 + Cg) >>13)];
dest[8*i+6]=clip_table[((Y2 + Cr) >>13)];
} }
break;
case IMGFMT_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];
((uint8_t*)dest)+=6;
} }
else if(dstFormat==IMGFMT_BGR24) break;
{ case IMGFMT_RGB16:
int i; case IMGFMT_BGR16:
for(i=0; i<(dstW>>1); i++){
int j;
int Y1=0;
int Y2=0;
int U=0;
int V=0;
int Cb, Cr, Cg;
for(j=0; j<lumFilterSize; j++)
{ {
Y1 += lumSrc[j][2*i] * lumFilter[j]; const int dr1= dither_2x2_8[y&1 ][0];
Y2 += lumSrc[j][2*i+1] * lumFilter[j]; 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];
} }
for(j=0; j<chrFilterSize; j++)
{
U += chrSrc[j][i] * chrFilter[j];
V += chrSrc[j][i+2048] * chrFilter[j];
} }
Y1= clip_yuvtab_2568[ (Y1>>19) + 256 ]; break;
Y2= clip_yuvtab_2568[ (Y2>>19) + 256 ]; case IMGFMT_RGB15:
U >>= 19; case IMGFMT_BGR15:
V >>= 19; {
const int dr1= dither_2x2_8[y&1 ][0];
Cb= clip_yuvtab_40cf[U+ 256]; const int dg1= dither_2x2_8[y&1 ][1];
Cg= clip_yuvtab_1a1e[V+ 256] + yuvtab_0c92[U+ 256]; const int db1= dither_2x2_8[(y&1)^1][0];
Cr= clip_yuvtab_3343[V+ 256]; const int dr2= dither_2x2_8[y&1 ][1];
const int dg2= dither_2x2_8[y&1 ][0];
dest[0]=clip_table[((Y1 + Cb) >>13)]; const int db2= dither_2x2_8[(y&1)^1][1];
dest[1]=clip_table[((Y1 + Cg) >>13)]; YSCALE_YUV_2_RGBX_C(uint16_t)
dest[2]=clip_table[((Y1 + Cr) >>13)]; ((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];
dest[3]=clip_table[((Y2 + Cb) >>13)];
dest[4]=clip_table[((Y2 + Cg) >>13)];
dest[5]=clip_table[((Y2 + Cr) >>13)];
dest+=6;
} }
} }
else if(dstFormat==IMGFMT_BGR16) break;
{ case IMGFMT_RGB8:
int i; case IMGFMT_BGR8:
#ifdef DITHER1XBPP
static int ditherb1=1<<14;
static int ditherg1=1<<13;
static int ditherr1=2<<14;
static int ditherb2=3<<14;
static int ditherg2=3<<13;
static int ditherr2=0<<14;
ditherb1 ^= (1^2)<<14;
ditherg1 ^= (1^2)<<13;
ditherr1 ^= (1^2)<<14;
ditherb2 ^= (3^0)<<14;
ditherg2 ^= (3^0)<<13;
ditherr2 ^= (3^0)<<14;
#else
const int ditherb1=0;
const int ditherg1=0;
const int ditherr1=0;
const int ditherb2=0;
const int ditherg2=0;
const int ditherr2=0;
#endif
for(i=0; i<(dstW>>1); i++){
int j;
int Y1=0;
int Y2=0;
int U=0;
int V=0;
int Cb, Cr, Cg;
for(j=0; j<lumFilterSize; j++)
{ {
Y1 += lumSrc[j][2*i] * lumFilter[j]; const uint8_t * const d64= dither_8x8_73[y&7];
Y2 += lumSrc[j][2*i+1] * lumFilter[j]; 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]];
} }
for(j=0; j<chrFilterSize; j++)
{
U += chrSrc[j][i] * chrFilter[j];
V += chrSrc[j][i+2048] * chrFilter[j];
} }
Y1= clip_yuvtab_2568[ (Y1>>19) + 256 ]; break;
Y2= clip_yuvtab_2568[ (Y2>>19) + 256 ]; case IMGFMT_RGB4:
U >>= 19; case IMGFMT_BGR4:
V >>= 19; {
const uint8_t * const d64= dither_8x8_73 [y&7];
Cb= clip_yuvtab_40cf[U+ 256]; const uint8_t * const d128=dither_8x8_220[y&7];
Cg= clip_yuvtab_1a1e[V+ 256] + yuvtab_0c92[U+ 256]; YSCALE_YUV_2_RGBX_C(uint8_t)
Cr= clip_yuvtab_3343[V+ 256]; ((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]];
((uint16_t*)dest)[2*i] =
clip_table16b[(Y1 + Cb + ditherb1) >>13] |
clip_table16g[(Y1 + Cg + ditherg1) >>13] |
clip_table16r[(Y1 + Cr + ditherr1) >>13];
((uint16_t*)dest)[2*i+1] =
clip_table16b[(Y2 + Cb + ditherb2) >>13] |
clip_table16g[(Y2 + Cg + ditherg2) >>13] |
clip_table16r[(Y2 + Cr + ditherr2) >>13];
} }
} }
else if(dstFormat==IMGFMT_BGR15) break;
case IMGFMT_RGB1:
case IMGFMT_BGR1:
{ {
int i; const uint8_t * const d128=dither_8x8_220[y&7];
#ifdef DITHER1XBPP uint8_t *g= c->table_gU[128] + c->table_gV[128];
static int ditherb1=1<<14; int acc=0;
static int ditherg1=1<<14; for(i=0; i<dstW-1; i+=2){
static int ditherr1=2<<14;
static int ditherb2=3<<14;
static int ditherg2=3<<14;
static int ditherr2=0<<14;
ditherb1 ^= (1^2)<<14;
ditherg1 ^= (1^2)<<14;
ditherr1 ^= (1^2)<<14;
ditherb2 ^= (3^0)<<14;
ditherg2 ^= (3^0)<<14;
ditherr2 ^= (3^0)<<14;
#else
const int ditherb1=0;
const int ditherg1=0;
const int ditherr1=0;
const int ditherb2=0;
const int ditherg2=0;
const int ditherr2=0;
#endif
for(i=0; i<(dstW>>1); i++){
int j; int j;
int Y1=0; int Y1=0;
int Y2=0; int Y2=0;
int U=0;
int V=0;
int Cb, Cr, Cg;
for(j=0; j<lumFilterSize; j++) for(j=0; j<lumFilterSize; j++)
{ {
Y1 += lumSrc[j][2*i] * lumFilter[j]; Y1 += lumSrc[j][i] * lumFilter[j];
Y2 += lumSrc[j][2*i+1] * lumFilter[j]; Y2 += lumSrc[j][i+1] * lumFilter[j];
} }
for(j=0; j<chrFilterSize; j++) Y1>>=19;
Y2>>=19;
if((Y1|Y2)&256)
{ {
U += chrSrc[j][i] * chrFilter[j]; if(Y1>255) Y1=255;
V += chrSrc[j][i+2048] * chrFilter[j]; 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;
((uint8_t*)dest)++;
} }
Y1= clip_yuvtab_2568[ (Y1>>19) + 256 ];
Y2= clip_yuvtab_2568[ (Y2>>19) + 256 ];
U >>= 19;
V >>= 19;
Cb= clip_yuvtab_40cf[U+ 256];
Cg= clip_yuvtab_1a1e[V+ 256] + yuvtab_0c92[U+ 256];
Cr= clip_yuvtab_3343[V+ 256];
((uint16_t*)dest)[2*i] =
clip_table15b[(Y1 + Cb + ditherb1) >>13] |
clip_table15g[(Y1 + Cg + ditherg1) >>13] |
clip_table15r[(Y1 + Cr + ditherr1) >>13];
((uint16_t*)dest)[2*i+1] =
clip_table15b[(Y2 + Cb + ditherb2) >>13] |
clip_table15g[(Y2 + Cg + ditherg2) >>13] |
clip_table15r[(Y2 + Cr + ditherr2) >>13];
} }
} }
break;
}
} }
...@@ -1330,22 +1458,6 @@ static void globalInit(){ ...@@ -1330,22 +1458,6 @@ static void globalInit(){
for(i=0; i<768; i++){ for(i=0; i<768; i++){
int c= MIN(MAX(i-256, 0), 255); int c= MIN(MAX(i-256, 0), 255);
clip_table[i]=c; clip_table[i]=c;
yuvtab_2568[c]= clip_yuvtab_2568[i]=(0x2568*(c-16))+(256<<13);
yuvtab_3343[c]= clip_yuvtab_3343[i]=0x3343*(c-128);
yuvtab_0c92[c]= clip_yuvtab_0c92[i]=-0x0c92*(c-128);
yuvtab_1a1e[c]= clip_yuvtab_1a1e[i]=-0x1a1e*(c-128);
yuvtab_40cf[c]= clip_yuvtab_40cf[i]=0x40cf*(c-128);
}
for(i=0; i<768; i++)
{
int v= clip_table[i];
clip_table16b[i]= v>>3;
clip_table16g[i]= (v<<3)&0x07E0;
clip_table16r[i]= (v<<8)&0xF800;
clip_table15b[i]= v>>3;
clip_table15g[i]= (v<<2)&0x03E0;
clip_table15r[i]= (v<<7)&0x7C00;
} }
cpuCaps= gCpuCaps; cpuCaps= gCpuCaps;
...@@ -1887,15 +1999,11 @@ SwsContext *getSwsContext(int srcW, int srcH, int srcFormat, int dstW, int dstH, ...@@ -1887,15 +1999,11 @@ SwsContext *getSwsContext(int srcW, int srcH, int srcFormat, int dstW, int dstH,
c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample); c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
c->chrDstW= -((-dstW) >> c->chrDstHSubSample); c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
c->chrDstH= -((-dstH) >> c->chrDstVSubSample); c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
/* printf("%d %d %d %d / %d %d %d %d //\n",
c->chrSrcW, if(isBGR(dstFormat))
c->chrSrcH, c->yuvTable= yuv2rgb_c_init(dstFormat & 0xFF, MODE_RGB, c->table_rV, c->table_gU, c->table_gV, c->table_bU);
c->chrDstW, if(isRGB(dstFormat))
c->chrDstH, c->yuvTable= yuv2rgb_c_init(dstFormat & 0xFF, MODE_BGR, c->table_rV, c->table_gU, c->table_gV, c->table_bU);
srcW,
srcH,
dstW,
dstH);*/
/* unscaled special Cases */ /* unscaled special Cases */
if(unscaled && !usesFilter) if(unscaled && !usesFilter)
...@@ -1924,6 +2032,7 @@ dstH);*/ ...@@ -1924,6 +2032,7 @@ dstH);*/
if((srcFormat==IMGFMT_YV12 || srcFormat==IMGFMT_I420) && isBGR(dstFormat)) if((srcFormat==IMGFMT_YV12 || srcFormat==IMGFMT_I420) && isBGR(dstFormat))
{ {
// FIXME multiple yuv2rgb converters wont work that way cuz that thing is full of globals&statics // FIXME multiple yuv2rgb converters wont work that way cuz that thing is full of globals&statics
//FIXME rgb vs. bgr ?
#ifdef WORDS_BIGENDIAN #ifdef WORDS_BIGENDIAN
if(dstFormat==IMGFMT_BGR32) if(dstFormat==IMGFMT_BGR32)
yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_BGR); yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_BGR);
...@@ -2613,6 +2722,8 @@ void freeSwsContext(SwsContext *c){ ...@@ -2613,6 +2722,8 @@ void freeSwsContext(SwsContext *c){
c->lumMmx2FilterPos=NULL; c->lumMmx2FilterPos=NULL;
if(c->chrMmx2FilterPos) free(c->chrMmx2FilterPos); if(c->chrMmx2FilterPos) free(c->chrMmx2FilterPos);
c->chrMmx2FilterPos=NULL; c->chrMmx2FilterPos=NULL;
if(c->yuvTable) free(c->yuvTable);
c->yuvTable=NULL;
free(c); free(c);
} }
......
postproc/swscale.h
View file @ cf7d1c1a
...@@ -88,6 +88,11 @@ typedef struct SwsContext{ ...@@ -88,6 +88,11 @@ typedef struct SwsContext{
int chrBufIndex; int chrBufIndex;
int dstY; int dstY;
int flags; int flags;
void * yuvTable;
void * table_rV[256];
void * table_gU[256];
int table_gV[256];
void * table_bU[256];
void (*swScale)(struct SwsContext *context, uint8_t* src[], int srcStride[], int srcSliceY, void (*swScale)(struct SwsContext *context, uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dst[], int dstStride[]); int srcSliceH, uint8_t* dst[], int dstStride[]);
......
postproc/swscale_template.c
View file @ cf7d1c1a
...@@ -752,18 +752,14 @@ static inline void RENAME(yuv2yuv1)(int16_t *lumSrc, int16_t *chrSrc, ...@@ -752,18 +752,14 @@ static inline void RENAME(yuv2yuv1)(int16_t *lumSrc, int16_t *chrSrc,
/** /**
* vertical scale YV12 to RGB * vertical scale YV12 to RGB
*/ */
static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, static inline void RENAME(yuv2rgbX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
uint8_t *dest, int dstW, int dstFormat, int16_t * lumMmxFilter, int16_t * chrMmxFilter) uint8_t *dest, int dstW, int16_t * lumMmxFilter, int16_t * chrMmxFilter, int dstY)
{ {
/* if(flags&SWS_FULL_UV_IPOL) switch(c->dstFormat)
{
//FIXME
}//FULL_UV_IPOL
else*/
{ {
#ifdef HAVE_MMX #ifdef HAVE_MMX
if(dstFormat == IMGFMT_BGR32) //FIXME untested case IMGFMT_BGR32:
{ {
asm volatile( asm volatile(
YSCALEYUV2RGBX YSCALEYUV2RGBX
...@@ -776,7 +772,8 @@ static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lu ...@@ -776,7 +772,8 @@ static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lu
: "%eax", "%ebx", "%ecx", "%edx", "%esi" : "%eax", "%ebx", "%ecx", "%edx", "%esi"
); );
} }
else if(dstFormat == IMGFMT_BGR24) //FIXME untested break;
case IMGFMT_BGR24:
{ {
asm volatile( asm volatile(
YSCALEYUV2RGBX YSCALEYUV2RGBX
...@@ -791,7 +788,8 @@ static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lu ...@@ -791,7 +788,8 @@ static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lu
: "%eax", "%ebx", "%ecx", "%edx", "%esi" : "%eax", "%ebx", "%ecx", "%edx", "%esi"
); );
} }
else if(dstFormat==IMGFMT_BGR15) break;
case IMGFMT_BGR15:
{ {
asm volatile( asm volatile(
YSCALEYUV2RGBX YSCALEYUV2RGBX
...@@ -811,7 +809,8 @@ static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lu ...@@ -811,7 +809,8 @@ static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lu
: "%eax", "%ebx", "%ecx", "%edx", "%esi" : "%eax", "%ebx", "%ecx", "%edx", "%esi"
); );
} }
else if(dstFormat==IMGFMT_BGR16) break;
case IMGFMT_BGR16:
{ {
asm volatile( asm volatile(
YSCALEYUV2RGBX YSCALEYUV2RGBX
...@@ -831,31 +830,33 @@ static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lu ...@@ -831,31 +830,33 @@ static inline void RENAME(yuv2rgbX)(int16_t *lumFilter, int16_t **lumSrc, int lu
: "%eax", "%ebx", "%ecx", "%edx", "%esi" : "%eax", "%ebx", "%ecx", "%edx", "%esi"
); );
} }
#else break;
yuv2rgbXinC(lumFilter, lumSrc, lumFilterSize,
chrFilter, chrSrc, chrFilterSize,
dest, dstW, dstFormat);
#endif #endif
} //!FULL_UV_IPOL default:
yuv2rgbXinC(c, lumFilter, lumSrc, lumFilterSize,
chrFilter, chrSrc, chrFilterSize,
dest, dstW, dstY);
break;
}
} }
/** /**
* vertical bilinear scale YV12 to RGB * vertical bilinear scale YV12 to RGB
*/ */
static inline void RENAME(yuv2rgb2)(uint16_t *buf0, uint16_t *buf1, uint16_t *uvbuf0, uint16_t *uvbuf1, static inline void RENAME(yuv2rgb2)(SwsContext *c, uint16_t *buf0, uint16_t *buf1, uint16_t *uvbuf0, uint16_t *uvbuf1,
uint8_t *dest, int dstW, int yalpha, int uvalpha, int dstFormat, int flags) uint8_t *dest, int dstW, int yalpha, int uvalpha, int y)
{ {
int yalpha1=yalpha^4095; int yalpha1=yalpha^4095;
int uvalpha1=uvalpha^4095; int uvalpha1=uvalpha^4095;
int i;
#if 0 //isnt used
if(flags&SWS_FULL_CHR_H_INT) if(flags&SWS_FULL_CHR_H_INT)
{ {
switch(dstFormat)
#ifdef HAVE_MMX
if(dstFormat==IMGFMT_BGR32)
{ {
#ifdef HAVE_MMX
case IMGFMT_BGR32:
asm volatile( asm volatile(
...@@ -879,9 +880,8 @@ FULL_YSCALEYUV2RGB ...@@ -879,9 +880,8 @@ FULL_YSCALEYUV2RGB
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} break;
else if(dstFormat==IMGFMT_BGR24) case IMGFMT_BGR24:
{
asm volatile( asm volatile(
FULL_YSCALEYUV2RGB FULL_YSCALEYUV2RGB
...@@ -929,9 +929,8 @@ FULL_YSCALEYUV2RGB ...@@ -929,9 +929,8 @@ FULL_YSCALEYUV2RGB
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax", "%ebx" : "%eax", "%ebx"
); );
} break;
else if(dstFormat==IMGFMT_BGR15) case IMGFMT_BGR15:
{
asm volatile( asm volatile(
FULL_YSCALEYUV2RGB FULL_YSCALEYUV2RGB
...@@ -963,9 +962,8 @@ FULL_YSCALEYUV2RGB ...@@ -963,9 +962,8 @@ FULL_YSCALEYUV2RGB
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} break;
else if(dstFormat==IMGFMT_BGR16) case IMGFMT_BGR16:
{
asm volatile( asm volatile(
FULL_YSCALEYUV2RGB FULL_YSCALEYUV2RGB
...@@ -997,8 +995,12 @@ FULL_YSCALEYUV2RGB ...@@ -997,8 +995,12 @@ FULL_YSCALEYUV2RGB
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} break;
#else #endif
case IMGFMT_RGB32:
#ifndef HAVE_MMX
case IMGFMT_BGR32:
#endif
if(dstFormat==IMGFMT_BGR32) if(dstFormat==IMGFMT_BGR32)
{ {
int i; int i;
...@@ -1060,13 +1062,14 @@ FULL_YSCALEYUV2RGB ...@@ -1060,13 +1062,14 @@ FULL_YSCALEYUV2RGB
clip_table15r[(Y + yuvtab_3343[V]) >>13]; clip_table15r[(Y + yuvtab_3343[V]) >>13];
} }
} }
#endif
}//FULL_UV_IPOL }//FULL_UV_IPOL
else else
{ {
#endif // if 0
#ifdef HAVE_MMX #ifdef HAVE_MMX
if(dstFormat==IMGFMT_BGR32) switch(c->dstFormat)
{ {
case IMGFMT_BGR32:
asm volatile( asm volatile(
YSCALEYUV2RGB YSCALEYUV2RGB
WRITEBGR32 WRITEBGR32
...@@ -1075,9 +1078,8 @@ FULL_YSCALEYUV2RGB ...@@ -1075,9 +1078,8 @@ FULL_YSCALEYUV2RGB
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} return;
else if(dstFormat==IMGFMT_BGR24) case IMGFMT_BGR24:
{
asm volatile( asm volatile(
"movl %4, %%ebx \n\t" "movl %4, %%ebx \n\t"
YSCALEYUV2RGB YSCALEYUV2RGB
...@@ -1087,9 +1089,8 @@ FULL_YSCALEYUV2RGB ...@@ -1087,9 +1089,8 @@ FULL_YSCALEYUV2RGB
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax", "%ebx" : "%eax", "%ebx"
); );
} return;
else if(dstFormat==IMGFMT_BGR15) case IMGFMT_BGR15:
{
asm volatile( asm volatile(
YSCALEYUV2RGB YSCALEYUV2RGB
/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
...@@ -1105,9 +1106,8 @@ FULL_YSCALEYUV2RGB ...@@ -1105,9 +1106,8 @@ FULL_YSCALEYUV2RGB
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} return;
else if(dstFormat==IMGFMT_BGR16) case IMGFMT_BGR16:
{
asm volatile( asm volatile(
YSCALEYUV2RGB YSCALEYUV2RGB
/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
...@@ -1123,176 +1123,38 @@ FULL_YSCALEYUV2RGB ...@@ -1123,176 +1123,38 @@ FULL_YSCALEYUV2RGB
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
return;
default: break;
} }
#else #endif //HAVE_MMX
if(dstFormat==IMGFMT_BGR32) YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB2_C)
{
int i;
#ifdef WORDS_BIGENDIAN
dest++;
#endif
for(i=0; i<dstW-1; i+=2){
// vertical linear interpolation && yuv2rgb in a single step:
int Y1=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];
int Y2=yuvtab_2568[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19)];
int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
int Cb= yuvtab_40cf[U];
int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
int Cr= yuvtab_3343[V];
dest[4*i+0]=clip_table[((Y1 + Cb) >>13)];
dest[4*i+1]=clip_table[((Y1 + Cg) >>13)];
dest[4*i+2]=clip_table[((Y1 + Cr) >>13)];
dest[4*i+4]=clip_table[((Y2 + Cb) >>13)];
dest[4*i+5]=clip_table[((Y2 + Cg) >>13)];
dest[4*i+6]=clip_table[((Y2 + Cr) >>13)];
}
}
else if(dstFormat==IMGFMT_BGR24)
{
int i;
for(i=0; i<dstW-1; i+=2){
// vertical linear interpolation && yuv2rgb in a single step:
int Y1=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];
int Y2=yuvtab_2568[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19)];
int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
int Cb= yuvtab_40cf[U];
int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
int Cr= yuvtab_3343[V];
dest[0]=clip_table[((Y1 + Cb) >>13)];
dest[1]=clip_table[((Y1 + Cg) >>13)];
dest[2]=clip_table[((Y1 + Cr) >>13)];
dest[3]=clip_table[((Y2 + Cb) >>13)];
dest[4]=clip_table[((Y2 + Cg) >>13)];
dest[5]=clip_table[((Y2 + Cr) >>13)];
dest+=6;
}
}
else if(dstFormat==IMGFMT_BGR16)
{
int i;
#ifdef DITHER1XBPP
static int ditherb1=1<<14;
static int ditherg1=1<<13;
static int ditherr1=2<<14;
static int ditherb2=3<<14;
static int ditherg2=3<<13;
static int ditherr2=0<<14;
ditherb1 ^= (1^2)<<14;
ditherg1 ^= (1^2)<<13;
ditherr1 ^= (1^2)<<14;
ditherb2 ^= (3^0)<<14;
ditherg2 ^= (3^0)<<13;
ditherr2 ^= (3^0)<<14;
#else
const int ditherb1=0;
const int ditherg1=0;
const int ditherr1=0;
const int ditherb2=0;
const int ditherg2=0;
const int ditherr2=0;
#endif
for(i=0; i<dstW-1; i+=2){
// vertical linear interpolation && yuv2rgb in a single step:
int Y1=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];
int Y2=yuvtab_2568[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19)];
int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
int Cb= yuvtab_40cf[U];
int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
int Cr= yuvtab_3343[V];
((uint16_t*)dest)[i] =
clip_table16b[(Y1 + Cb + ditherb1) >>13] |
clip_table16g[(Y1 + Cg + ditherg1) >>13] |
clip_table16r[(Y1 + Cr + ditherr1) >>13];
((uint16_t*)dest)[i+1] =
clip_table16b[(Y2 + Cb + ditherb2) >>13] |
clip_table16g[(Y2 + Cg + ditherg2) >>13] |
clip_table16r[(Y2 + Cr + ditherr2) >>13];
}
}
else if(dstFormat==IMGFMT_BGR15)
{
int i;
#ifdef DITHER1XBPP
static int ditherb1=1<<14;
static int ditherg1=1<<14;
static int ditherr1=2<<14;
static int ditherb2=3<<14;
static int ditherg2=3<<14;
static int ditherr2=0<<14;
ditherb1 ^= (1^2)<<14;
ditherg1 ^= (1^2)<<14;
ditherr1 ^= (1^2)<<14;
ditherb2 ^= (3^0)<<14;
ditherg2 ^= (3^0)<<14;
ditherr2 ^= (3^0)<<14;
#else
const int ditherb1=0;
const int ditherg1=0;
const int ditherr1=0;
const int ditherb2=0;
const int ditherg2=0;
const int ditherr2=0;
#endif
for(i=0; i<dstW-1; i+=2){
// vertical linear interpolation && yuv2rgb in a single step:
int Y1=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];
int Y2=yuvtab_2568[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19)];
int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
int Cb= yuvtab_40cf[U];
int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
int Cr= yuvtab_3343[V];
((uint16_t*)dest)[i] =
clip_table15b[(Y1 + Cb + ditherb1) >>13] |
clip_table15g[(Y1 + Cg + ditherg1) >>13] |
clip_table15r[(Y1 + Cr + ditherr1) >>13];
((uint16_t*)dest)[i+1] =
clip_table15b[(Y2 + Cb + ditherb2) >>13] |
clip_table15g[(Y2 + Cg + ditherg2) >>13] |
clip_table15r[(Y2 + Cr + ditherr2) >>13];
}
}
#endif
} //!FULL_UV_IPOL
} }
/** /**
* YV12 to RGB without scaling or interpolating * YV12 to RGB without scaling or interpolating
*/ */
static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *uvbuf1, static inline void RENAME(yuv2rgb1)(SwsContext *c, uint16_t *buf0, uint16_t *uvbuf0, uint16_t *uvbuf1,
uint8_t *dest, int dstW, int uvalpha, int dstFormat, int flags) uint8_t *dest, int dstW, int uvalpha, int dstFormat, int flags, int y)
{ {
int uvalpha1=uvalpha^4095; int uvalpha1=uvalpha^4095;
const int yalpha1=0; const int yalpha1=0;
int i;
uint16_t *buf1= buf0; //FIXME needed for the rgb1/bgr1
const int yalpha= 4096; //FIXME ...
if(flags&SWS_FULL_CHR_H_INT) if(flags&SWS_FULL_CHR_H_INT)
{ {
RENAME(yuv2rgb2)(buf0, buf0, uvbuf0, uvbuf1, dest, dstW, 0, uvalpha, dstFormat, flags); RENAME(yuv2rgb2)(c, buf0, buf0, uvbuf0, uvbuf1, dest, dstW, 0, uvalpha, y);
return; return;
} }
#ifdef HAVE_MMX #ifdef HAVE_MMX
if( uvalpha < 2048 ) // note this is not correct (shifts chrominance by 0.5 pixels) but its a bit faster if( uvalpha < 2048 ) // note this is not correct (shifts chrominance by 0.5 pixels) but its a bit faster
{ {
if(dstFormat==IMGFMT_BGR32) switch(dstFormat)
{ {
case IMGFMT_BGR32:
asm volatile( asm volatile(
YSCALEYUV2RGB1 YSCALEYUV2RGB1
WRITEBGR32 WRITEBGR32
...@@ -1300,9 +1162,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t * ...@@ -1300,9 +1162,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} return;
else if(dstFormat==IMGFMT_BGR24) case IMGFMT_BGR24:
{
asm volatile( asm volatile(
"movl %4, %%ebx \n\t" "movl %4, %%ebx \n\t"
YSCALEYUV2RGB1 YSCALEYUV2RGB1
...@@ -1311,9 +1172,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t * ...@@ -1311,9 +1172,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax", "%ebx" : "%eax", "%ebx"
); );
} return;
else if(dstFormat==IMGFMT_BGR15) case IMGFMT_BGR15:
{
asm volatile( asm volatile(
YSCALEYUV2RGB1 YSCALEYUV2RGB1
/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
...@@ -1327,9 +1187,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t * ...@@ -1327,9 +1187,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} return;
else if(dstFormat==IMGFMT_BGR16) case IMGFMT_BGR16:
{
asm volatile( asm volatile(
YSCALEYUV2RGB1 YSCALEYUV2RGB1
/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
...@@ -1344,12 +1203,14 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t * ...@@ -1344,12 +1203,14 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
return;
} }
} }
else else
{ {
if(dstFormat==IMGFMT_BGR32) switch(dstFormat)
{ {
case IMGFMT_BGR32:
asm volatile( asm volatile(
YSCALEYUV2RGB1b YSCALEYUV2RGB1b
WRITEBGR32 WRITEBGR32
...@@ -1357,9 +1218,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t * ...@@ -1357,9 +1218,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} return;
else if(dstFormat==IMGFMT_BGR24) case IMGFMT_BGR24:
{
asm volatile( asm volatile(
"movl %4, %%ebx \n\t" "movl %4, %%ebx \n\t"
YSCALEYUV2RGB1b YSCALEYUV2RGB1b
...@@ -1368,9 +1228,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t * ...@@ -1368,9 +1228,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax", "%ebx" : "%eax", "%ebx"
); );
} return;
else if(dstFormat==IMGFMT_BGR15) case IMGFMT_BGR15:
{
asm volatile( asm volatile(
YSCALEYUV2RGB1b YSCALEYUV2RGB1b
/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
...@@ -1384,9 +1243,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t * ...@@ -1384,9 +1243,8 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
} return;
else if(dstFormat==IMGFMT_BGR16) case IMGFMT_BGR16:
{
asm volatile( asm volatile(
YSCALEYUV2RGB1b YSCALEYUV2RGB1b
/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */ /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
...@@ -1401,156 +1259,16 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t * ...@@ -1401,156 +1259,16 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
"m" (yalpha1), "m" (uvalpha1) "m" (yalpha1), "m" (uvalpha1)
: "%eax" : "%eax"
); );
return;
} }
} }
#else
//FIXME write 2 versions (for even & odd lines)
if(dstFormat==IMGFMT_BGR32)
{
int i;
#ifdef WORDS_BIGENDIAN
dest++;
#endif
for(i=0; i<dstW-1; i+=2){
// vertical linear interpolation && yuv2rgb in a single step:
int Y1=yuvtab_2568[buf0[i]>>7];
int Y2=yuvtab_2568[buf0[i+1]>>7];
int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
int Cb= yuvtab_40cf[U];
int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
int Cr= yuvtab_3343[V];
dest[4*i+0]=clip_table[((Y1 + Cb) >>13)];
dest[4*i+1]=clip_table[((Y1 + Cg) >>13)];
dest[4*i+2]=clip_table[((Y1 + Cr) >>13)];
dest[4*i+4]=clip_table[((Y2 + Cb) >>13)];
dest[4*i+5]=clip_table[((Y2 + Cg) >>13)];
dest[4*i+6]=clip_table[((Y2 + Cr) >>13)];
}
}
else if(dstFormat==IMGFMT_BGR24)
{
int i;
for(i=0; i<dstW-1; i+=2){
// vertical linear interpolation && yuv2rgb in a single step:
int Y1=yuvtab_2568[buf0[i]>>7];
int Y2=yuvtab_2568[buf0[i+1]>>7];
int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
int Cb= yuvtab_40cf[U];
int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
int Cr= yuvtab_3343[V];
dest[0]=clip_table[((Y1 + Cb) >>13)];
dest[1]=clip_table[((Y1 + Cg) >>13)];
dest[2]=clip_table[((Y1 + Cr) >>13)];
dest[3]=clip_table[((Y2 + Cb) >>13)];
dest[4]=clip_table[((Y2 + Cg) >>13)];
dest[5]=clip_table[((Y2 + Cr) >>13)];
dest+=6;
}
}
else if(dstFormat==IMGFMT_BGR16)
{
int i;
#ifdef DITHER1XBPP
static int ditherb1=1<<14;
static int ditherg1=1<<13;
static int ditherr1=2<<14;
static int ditherb2=3<<14;
static int ditherg2=3<<13;
static int ditherr2=0<<14;
ditherb1 ^= (1^2)<<14;
ditherg1 ^= (1^2)<<13;
ditherr1 ^= (1^2)<<14;
ditherb2 ^= (3^0)<<14;
ditherg2 ^= (3^0)<<13;
ditherr2 ^= (3^0)<<14;
#else
const int ditherb1=0;
const int ditherg1=0;
const int ditherr1=0;
const int ditherb2=0;
const int ditherg2=0;
const int ditherr2=0;
#endif #endif
for(i=0; i<dstW-1; i+=2){ if( uvalpha < 2048 )
// vertical linear interpolation && yuv2rgb in a single step:
int Y1=yuvtab_2568[buf0[i]>>7];
int Y2=yuvtab_2568[buf0[i+1]>>7];
int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
int Cb= yuvtab_40cf[U];
int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
int Cr= yuvtab_3343[V];
((uint16_t*)dest)[i] =
clip_table16b[(Y1 + Cb + ditherb1) >>13] |
clip_table16g[(Y1 + Cg + ditherg1) >>13] |
clip_table16r[(Y1 + Cr + ditherr1) >>13];
((uint16_t*)dest)[i+1] =
clip_table16b[(Y2 + Cb + ditherb2) >>13] |
clip_table16g[(Y2 + Cg + ditherg2) >>13] |
clip_table16r[(Y2 + Cr + ditherr2) >>13];
}
}
else if(dstFormat==IMGFMT_BGR15)
{ {
int i; YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1_C)
#ifdef DITHER1XBPP }else{
static int ditherb1=1<<14; YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1B_C)
static int ditherg1=1<<14;
static int ditherr1=2<<14;
static int ditherb2=3<<14;
static int ditherg2=3<<14;
static int ditherr2=0<<14;
ditherb1 ^= (1^2)<<14;
ditherg1 ^= (1^2)<<14;
ditherr1 ^= (1^2)<<14;
ditherb2 ^= (3^0)<<14;
ditherg2 ^= (3^0)<<14;
ditherr2 ^= (3^0)<<14;
#else
const int ditherb1=0;
const int ditherg1=0;
const int ditherr1=0;
const int ditherb2=0;
const int ditherg2=0;
const int ditherr2=0;
#endif
for(i=0; i<dstW-1; i+=2){
// vertical linear interpolation && yuv2rgb in a single step:
int Y1=yuvtab_2568[buf0[i]>>7];
int Y2=yuvtab_2568[buf0[i+1]>>7];
int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
int Cb= yuvtab_40cf[U];
int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
int Cr= yuvtab_3343[V];
((uint16_t*)dest)[i] =
clip_table15b[(Y1 + Cb + ditherb1) >>13] |
clip_table15g[(Y1 + Cg + ditherg1) >>13] |
clip_table15r[(Y1 + Cr + ditherr1) >>13];
((uint16_t*)dest)[i+1] =
clip_table15b[(Y2 + Cb + ditherb2) >>13] |
clip_table15g[(Y2 + Cg + ditherg2) >>13] |
clip_table15r[(Y2 + Cr + ditherr2) >>13];
} }
}
#endif
} }
//FIXME yuy2* can read upto 7 samples to much //FIXME yuy2* can read upto 7 samples to much
...@@ -2814,24 +2532,24 @@ i--; ...@@ -2814,24 +2532,24 @@ i--;
{ {
int chrAlpha= vChrFilter[2*dstY+1]; int chrAlpha= vChrFilter[2*dstY+1];
RENAME(yuv2rgb1)(*lumSrcPtr, *chrSrcPtr, *(chrSrcPtr+1), RENAME(yuv2rgb1)(c, *lumSrcPtr, *chrSrcPtr, *(chrSrcPtr+1),
dest, dstW, chrAlpha, dstFormat, flags); dest, dstW, chrAlpha, dstFormat, flags, dstY);
} }
else if(vLumFilterSize == 2 && vChrFilterSize == 2) //BiLinear Upscale RGB else if(vLumFilterSize == 2 && vChrFilterSize == 2) //BiLinear Upscale RGB
{ {
int lumAlpha= vLumFilter[2*dstY+1]; int lumAlpha= vLumFilter[2*dstY+1];
int chrAlpha= vChrFilter[2*dstY+1]; int chrAlpha= vChrFilter[2*dstY+1];
RENAME(yuv2rgb2)(*lumSrcPtr, *(lumSrcPtr+1), *chrSrcPtr, *(chrSrcPtr+1), RENAME(yuv2rgb2)(c, *lumSrcPtr, *(lumSrcPtr+1), *chrSrcPtr, *(chrSrcPtr+1),
dest, dstW, lumAlpha, chrAlpha, dstFormat, flags); dest, dstW, lumAlpha, chrAlpha, dstY);
} }
else //General RGB else //General RGB
{ {
RENAME(yuv2rgbX)( RENAME(yuv2rgbX)(c,
vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
dest, dstW, dstFormat, dest, dstW,
lumMmxFilter+dstY*vLumFilterSize*4, chrMmxFilter+dstY*vChrFilterSize*4); lumMmxFilter+dstY*vLumFilterSize*4, chrMmxFilter+dstY*vChrFilterSize*4, dstY);
} }
} }
} }
...@@ -2851,10 +2569,10 @@ i--; ...@@ -2851,10 +2569,10 @@ i--;
{ {
ASSERT(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2); ASSERT(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2);
ASSERT(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2); ASSERT(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2);
yuv2rgbXinC( yuv2rgbXinC(c,
vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize, vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize, vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
dest, dstW, dstFormat); dest, dstW, dstY);
} }
} }
} }
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment