/*
- Copyright (C) 2001-2002 Michael Niedermayer <michaelni@gmx.at>
+ Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
*/
/*
- supported Input formats: YV12, I420, IYUV, YUY2, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8, Y800
- supported output formats: YV12, I420, IYUV, BGR15, BGR16, BGR24, BGR32, Y8, Y800
- BGR15/16 support dithering
+ supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09
+ supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
+ {BGR,RGB}{1,4,8,15,16} support dithering
- unscaled special converters
- YV12/I420/IYUV -> BGR15/BGR16/BGR24/BGR32
- YV12/I420/IYUV -> YV12/I420/IYUV
- YUY2/BGR15/BGR16/BGR24/BGR32/RGB24/RGB32 -> same format
+ unscaled special converters (YV12=I420=IYUV, Y800=Y8)
+ YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
+ x -> x
+ YUV9 -> YV12
+ YUV9/YV12 -> Y800
+ Y800 -> YUV9/YV12
BGR24 -> BGR32 & RGB24 -> RGB32
BGR32 -> BGR24 & RGB32 -> RGB24
BGR15 -> BGR16
*/
/*
-tested special converters
- YV12/I420 -> BGR16
+tested special converters (most are tested actually but i didnt write it down ...)
+ YV12 -> BGR16
YV12 -> YV12
BGR15 -> BGR16
BGR16 -> BGR16
+ YVU9 -> YV12
untested special converters
YV12/I420 -> BGR15/BGR24/BGR32 (its the yuv2rgb stuff, so it should be ok)
#else
#include <stdlib.h>
#endif
+#ifdef HAVE_ALTIVEC_H
+#include <altivec.h>
+#endif
#include "swscale.h"
+#include "swscale_internal.h"
#include "../cpudetect.h"
#include "../bswap.h"
#include "../libvo/img_format.h"
#include "rgb2rgb.h"
#include "../libvo/fastmemcpy.h"
-#include "../mp_msg.h"
-
-#define MSG_WARN(args...) mp_msg(MSGT_SWS,MSGL_WARN, ##args )
-#define MSG_FATAL(args...) mp_msg(MSGT_SWS,MSGL_FATAL, ##args )
-#define MSG_ERR(args...) mp_msg(MSGT_SWS,MSGL_ERR, ##args )
-#define MSG_V(args...) mp_msg(MSGT_SWS,MSGL_V, ##args )
-#define MSG_DBG2(args...) mp_msg(MSGT_SWS,MSGL_DBG2, ##args )
-#define MSG_INFO(args...) mp_msg(MSGT_SWS,MSGL_INFO, ##args )
#undef MOVNTQ
#undef PAVGB
#endif
//FIXME replace this with something faster
-#define isPlanarYUV(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 || (x)==IMGFMT_YVU9)
-#define isYUV(x) ((x)==IMGFMT_YUY2 || isPlanarYUV(x))
-#define isHalfChrV(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420)
-#define isHalfChrH(x) ((x)==IMGFMT_YUY2 || (x)==IMGFMT_YV12 || (x)==IMGFMT_I420)
+#define isPlanarYUV(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_YVU9 \
+ || (x)==IMGFMT_444P || (x)==IMGFMT_422P || (x)==IMGFMT_411P)
+#define isYUV(x) ((x)==IMGFMT_UYVY || (x)==IMGFMT_YUY2 || isPlanarYUV(x))
#define isGray(x) ((x)==IMGFMT_Y800)
-#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_RGB32|| (x)==IMGFMT_RGB24\
- || (x)==IMGFMT_Y800)
-#define isSupportedOut(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 \
- || (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\
- || (x)==IMGFMT_Y800)
#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 isSupportedIn(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_YUY2 || (x)==IMGFMT_UYVY\
+ || (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\
+ || (x)==IMGFMT_RGB32|| (x)==IMGFMT_RGB24\
+ || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9\
+ || (x)==IMGFMT_444P || (x)==IMGFMT_422P || (x)==IMGFMT_411P)
+#define isSupportedOut(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_YUY2 || (x)==IMGFMT_UYVY\
+ || (x)==IMGFMT_444P || (x)==IMGFMT_422P || (x)==IMGFMT_411P\
+ || isRGB(x) || isBGR(x)\
+ || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9)
+#define isPacked(x) ((x)==IMGFMT_YUY2 || (x)==IMGFMT_UYVY ||isRGB(x) || isBGR(x))
#define RGB2YUV_SHIFT 16
#define BY ((int)( 0.098*(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))
-extern int verbose; // defined in mplayer.c
+extern const int32_t Inverse_Table_6_9[8][4];
+
/*
NOTES
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
optimize bgr24 & bgr32
add BGR4 output support
write special BGR->BGR scaler
-deglobalize yuv2rgb*.c
*/
#define ABS(a) ((a) > 0 ? (a) : (-(a)))
#define MAX(a,b) ((a) < (b) ? (b) : (a))
#ifdef ARCH_X86
-#define CAN_COMPILE_X86_ASM
-#endif
-
-#ifdef CAN_COMPILE_X86_ASM
-static uint64_t __attribute__((aligned(8))) yCoeff= 0x2568256825682568LL;
-static uint64_t __attribute__((aligned(8))) vrCoeff= 0x3343334333433343LL;
-static uint64_t __attribute__((aligned(8))) ubCoeff= 0x40cf40cf40cf40cfLL;
-static uint64_t __attribute__((aligned(8))) vgCoeff= 0xE5E2E5E2E5E2E5E2LL;
-static uint64_t __attribute__((aligned(8))) ugCoeff= 0xF36EF36EF36EF36ELL;
static uint64_t __attribute__((aligned(8))) bF8= 0xF8F8F8F8F8F8F8F8LL;
static uint64_t __attribute__((aligned(8))) bFC= 0xFCFCFCFCFCFCFCFCLL;
-static uint64_t __attribute__((aligned(8))) w400= 0x0400040004000400LL;
-static uint64_t __attribute__((aligned(8))) w80= 0x0080008000800080LL;
static uint64_t __attribute__((aligned(8))) w10= 0x0010001000100010LL;
static uint64_t __attribute__((aligned(8))) w02= 0x0002000200020002LL;
static uint64_t __attribute__((aligned(8))) bm00001111=0x00000000FFFFFFFFLL;
static const uint64_t bgr2YOffset __attribute__((aligned(8))) = 0x1010101010101010ULL;
static const uint64_t bgr2UVOffset __attribute__((aligned(8)))= 0x8080808080808080ULL;
static const uint64_t w1111 __attribute__((aligned(8))) = 0x0001000100010001ULL;
-
-// FIXME remove
-static uint64_t __attribute__((aligned(8))) asm_yalpha1;
-static uint64_t __attribute__((aligned(8))) asm_uvalpha1;
#endif
// clipping helper table for C implementations:
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
-int sws_flags=2;
-
-//global srcFilter
-SwsFilter src_filter= {NULL, NULL, NULL, NULL};
-
-float sws_lum_gblur= 0.0;
-float sws_chr_gblur= 0.0;
-int sws_chr_vshift= 0;
-int sws_chr_hshift= 0;
-float sws_chr_sharpen= 0.0;
-float sws_lum_sharpen= 0.0;
-
-/* cpuCaps combined from cpudetect and whats actually compiled in
- (if there is no support for something compiled in it wont appear here) */
-static CpuCaps cpuCaps;
-
-void (*swScale)(SwsContext *context, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[])=NULL;
-
-static SwsVector *getConvVec(SwsVector *a, SwsVector *b);
-
-#ifdef CAN_COMPILE_X86_ASM
+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];
+
+#ifdef ARCH_X86
void in_asm_used_var_warning_killer()
{
- volatile int i= yCoeff+vrCoeff+ubCoeff+vgCoeff+ugCoeff+bF8+bFC+w400+w80+w10+
- bm00001111+bm00000111+bm11111000+b16Mask+g16Mask+r16Mask+b15Mask+g15Mask+r15Mask+asm_yalpha1+ asm_uvalpha1+
+ volatile int i= bF8+bFC+w10+
+ bm00001111+bm00000111+bm11111000+b16Mask+g16Mask+r16Mask+b15Mask+g15Mask+r15Mask+
M24A+M24B+M24C+w02 + b5Dither+g5Dither+r5Dither+g6Dither+dither4[0]+dither8[0]+bm01010101;
if(i) i=0;
}
#endif
-static int testFormat[]={
-IMGFMT_YV12,
-//IMGFMT_IYUV,
-IMGFMT_I420,
-IMGFMT_BGR15,
-IMGFMT_BGR16,
-IMGFMT_BGR24,
-IMGFMT_BGR32,
-//IMGFMT_Y8,
-IMGFMT_Y800,
-//IMGFMT_YUY2,
-0
-};
-
-static uint64_t getSSD(uint8_t *src1, uint8_t *src2, int stride1, int stride2, int w, int h){
- int x,y;
- uint64_t ssd=0;
-
- for(y=0; y<h; y++){
- for(x=0; x<w; x++){
- int d= src1[x + y*stride1] - src2[x + y*stride2];
- ssd+= d*d;
- }
- }
- return ssd;
-}
-
-// test by ref -> src -> dst -> out & compare out against ref
-// ref & out are YV12
-static void doTest(uint8_t *ref[3], int refStride[3], int w, int h, int srcFormat, int dstFormat,
- int srcW, int srcH, int dstW, int dstH, int flags){
- uint8_t *src[3];
- uint8_t *dst[3];
- uint8_t *out[3];
- int srcStride[3], dstStride[3];
- int i;
- uint64_t ssdY, ssdU, ssdV;
- SwsContext *srcContext, *dstContext, *outContext;
-
- for(i=0; i<3; i++){
- srcStride[i]= srcW*4;
- dstStride[i]= dstW*4;
- src[i]= malloc(srcStride[i]*srcH);
- dst[i]= malloc(dstStride[i]*dstH);
- out[i]= malloc(refStride[i]*h);
- }
-
- srcContext= getSwsContext(w, h, IMGFMT_YV12, srcW, srcH, srcFormat, flags, NULL, NULL);
- dstContext= getSwsContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, NULL, NULL);
- outContext= getSwsContext(dstW, dstH, dstFormat, w, h, IMGFMT_YV12, flags, NULL, NULL);
- if(srcContext==NULL ||dstContext==NULL ||outContext==NULL){
- printf("Failed allocating swsContext\n");
- goto end;
- }
-// printf("test %X %X %X -> %X %X %X\n", (int)ref[0], (int)ref[1], (int)ref[2],
-// (int)src[0], (int)src[1], (int)src[2]);
-
- srcContext->swScale(srcContext, ref, refStride, 0, h , src, srcStride);
- dstContext->swScale(dstContext, src, srcStride, 0, srcH, dst, dstStride);
- outContext->swScale(outContext, dst, dstStride, 0, dstH, out, refStride);
-
- ssdY= getSSD(ref[0], out[0], refStride[0], refStride[0], w, h);
- ssdU= getSSD(ref[1], out[1], refStride[1], refStride[1], (w+1)>>1, (h+1)>>1);
- ssdV= getSSD(ref[2], out[2], refStride[2], refStride[2], (w+1)>>1, (h+1)>>1);
-
- if(isGray(srcFormat) || isGray(dstFormat)) ssdU=ssdV=0; //FIXME check that output is really gray
-
- ssdY/= w*h;
- ssdU/= w*h/4;
- ssdV/= w*h/4;
-
- if(ssdY>10 || ssdU>10 || ssdV>10){
- printf(" %s %dx%d -> %s %4dx%4d flags=%2d SSD=%5lld,%5lld,%5lld\n",
- vo_format_name(srcFormat), srcW, srcH,
- vo_format_name(dstFormat), dstW, dstH,
- flags,
- ssdY, ssdU, ssdV);
- }
-
- end:
-
- freeSwsContext(srcContext);
- freeSwsContext(dstContext);
- freeSwsContext(outContext);
-
- for(i=0; i<3; i++){
- free(src[i]);
- free(dst[i]);
- free(out[i]);
- }
-}
-
-static void selfTest(uint8_t *src[3], int stride[3], int w, int h){
- int srcFormat, dstFormat, srcFormatIndex, dstFormatIndex;
- int srcW, srcH, dstW, dstH;
- int flags;
-
- for(srcFormatIndex=0; ;srcFormatIndex++){
- srcFormat= testFormat[srcFormatIndex];
- if(!srcFormat) break;
- for(dstFormatIndex=0; ;dstFormatIndex++){
- dstFormat= testFormat[dstFormatIndex];
- if(!dstFormat) break;
- if(!isSupportedOut(dstFormat)) continue;
-
- srcW= w+w/3;
- srcH= h+h/3;
- for(dstW=w; dstW<w*2; dstW+= dstW/3){
- for(dstH=h; dstH<h*2; dstH+= dstH/3){
- for(flags=1; flags<33; flags*=2)
- doTest(src, stride, w, h, srcFormat, dstFormat,
- srcW, srcH, dstW, dstH, flags);
- }
- }
- }
- }
-}
-
static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
- uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW)
+ uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW)
{
//FIXME Optimize (just quickly writen not opti..)
int i;
for(i=0; i<dstW; i++)
{
- int val=0;
+ int val=1<<18;
int j;
for(j=0; j<lumFilterSize; j++)
val += lumSrc[j][i] * lumFilter[j];
}
if(uDest != NULL)
- for(i=0; i<(dstW>>1); i++)
+ for(i=0; i<chrDstW; i++)
{
- int u=0;
- int v=0;
+ int u=1<<18;
+ int v=1<<18;
int j;
for(j=0; j<chrFilterSize; j++)
{
}
}
-static inline void yuv2rgbXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
+
+#define YSCALE_YUV_2_PACKEDX_C(type) \
+ 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 *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;\
+ }
+
+#define YSCALE_YUV_2_RGBX_C(type) \
+ YSCALE_YUV_2_PACKEDX_C(type)\
+ r = c->table_rV[V];\
+ g = c->table_gU[U] + c->table_gV[V];\
+ b = c->table_bU[U];\
+
+#define YSCALE_YUV_2_PACKED2_C \
+ 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;\
+
+#define YSCALE_YUV_2_RGB2_C(type) \
+ YSCALE_YUV_2_PACKED2_C\
+ 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_PACKED1_C \
+ 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;\
+
+#define YSCALE_YUV_2_RGB1_C(type) \
+ YSCALE_YUV_2_PACKED1_C\
+ 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_PACKED1B_C \
+ 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;\
+
+#define YSCALE_YUV_2_RGB1B_C(type) \
+ YSCALE_YUV_2_PACKED1B_C\
+ 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, func2)\
+ 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)[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 IMGFMT_RG4B:\
+ case IMGFMT_BG4B:\
+ {\
+ 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;\
+ case IMGFMT_YUY2:\
+ func2\
+ ((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 IMGFMT_UYVY:\
+ func2\
+ ((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;\
+ }\
+
+
+static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
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;
+ switch(c->dstFormat)
{
- int i;
-#ifdef WORDS_BIGENDIAN
- 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];
- Y2 += lumSrc[j][2*i+1] * lumFilter[j];
- }
- 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 ];
- 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];
-
- 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)];
+ case IMGFMT_RGB32:
+ 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];
}
- }
- else if(dstFormat==IMGFMT_BGR24)
- {
- int i;
- 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];
- Y2 += lumSrc[j][2*i+1] * lumFilter[j];
+ break;
+ case IMGFMT_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];
+ ((uint8_t*)dest)+=6;
+ }
+ 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;
+ }
+ 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];
+ 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];
+ }
+ 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];
+ 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];
}
- 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];
-
- 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++){
- 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];
- Y2 += lumSrc[j][2*i+1] * lumFilter[j];
+ 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];
+ 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];
+ }
+ 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];
+ 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);
}
- 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_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)
- {
- 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++){
- 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];
- Y2 += lumSrc[j][2*i+1] * lumFilter[j];
+ break;
+ case IMGFMT_RG4B:
+ case IMGFMT_BG4B:
+ {
+ 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]];
}
- for(j=0; j<chrFilterSize; j++)
- {
- U += chrSrc[j][i] * chrFilter[j];
- V += chrSrc[j][i+2048] * chrFilter[j];
+ }
+ 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];
+ 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;
+ ((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;
+ case IMGFMT_YUY2:
+ 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 IMGFMT_UYVY:
+ 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;
}
}
#define COMPILE_C
#endif
-#ifdef CAN_COMPILE_X86_ASM
+#ifdef ARCH_POWERPC
+#ifdef HAVE_ALTIVEC
+#define COMPILE_ALTIVEC
+#endif //HAVE_ALTIVEC
+#endif //ARCH_POWERPC
+
+#ifdef ARCH_X86
#if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
#define COMPILE_MMX
#if (defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
#define COMPILE_3DNOW
#endif
-#endif //CAN_COMPILE_X86_ASM
+#endif //ARCH_X86
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
+#undef HAVE_ALTIVEC
#define RENAME(a) a ## _C
#include "swscale_template.c"
#endif
-#ifdef CAN_COMPILE_X86_ASM
+#ifdef ARCH_POWERPC
+#ifdef COMPILE_ALTIVEC
+#undef RENAME
+#define HAVE_ALTIVEC
+#define RENAME(a) a ## _altivec
+#include "swscale_template.c"
+#endif
+#endif //ARCH_POWERPC
+
+#ifdef ARCH_X86
//X86 versions
/*
#include "swscale_template.c"
#endif
-#endif //CAN_COMPILE_X86_ASM
+#endif //ARCH_X86
-// minor note: the HAVE_xyz is messed up after that line so dont use it
+// minor note: the HAVE_xyz is messed up after that line so don't use it
-
-// old global scaler, dont use for new code
-// will use sws_flags from the command line
-void SwScale_YV12slice(unsigned char* src[], int srcStride[], int srcSliceY ,
- int srcSliceH, uint8_t* dst[], int dstStride, int dstbpp,
- int srcW, int srcH, int dstW, int dstH){
-
- static SwsContext *context=NULL;
- int dstFormat;
- int dstStride3[3]= {dstStride, dstStride>>1, dstStride>>1};
-
- switch(dstbpp)
- {
- case 8 : dstFormat= IMGFMT_Y8; break;
- case 12: dstFormat= IMGFMT_YV12; break;
- case 15: dstFormat= IMGFMT_BGR15; break;
- case 16: dstFormat= IMGFMT_BGR16; break;
- case 24: dstFormat= IMGFMT_BGR24; break;
- case 32: dstFormat= IMGFMT_BGR32; break;
- default: return;
- }
-
- if(!context) context=getSwsContextFromCmdLine(srcW, srcH, IMGFMT_YV12, dstW, dstH, dstFormat);
-
- context->swScale(context, src, srcStride, srcSliceY, srcSliceH, dst, dstStride3);
-}
-
-// will use sws_flags & src_filter (from cmd line)
-SwsContext *getSwsContextFromCmdLine(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat)
+static double getSplineCoeff(double a, double b, double c, double d, double dist)
{
- int flags=0;
- static int firstTime=1;
-
-#ifdef ARCH_X86
- if(gCpuCaps.hasMMX)
- asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions)
-#endif
- if(firstTime)
- {
- firstTime=0;
- flags= SWS_PRINT_INFO;
- }
- else if(verbose>1) flags= SWS_PRINT_INFO;
-
- if(src_filter.lumH) freeVec(src_filter.lumH);
- if(src_filter.lumV) freeVec(src_filter.lumV);
- if(src_filter.chrH) freeVec(src_filter.chrH);
- if(src_filter.chrV) freeVec(src_filter.chrV);
-
- if(sws_lum_gblur!=0.0){
- src_filter.lumH= getGaussianVec(sws_lum_gblur, 3.0);
- src_filter.lumV= getGaussianVec(sws_lum_gblur, 3.0);
- }else{
- src_filter.lumH= getIdentityVec();
- src_filter.lumV= getIdentityVec();
- }
-
- if(sws_chr_gblur!=0.0){
- src_filter.chrH= getGaussianVec(sws_chr_gblur, 3.0);
- src_filter.chrV= getGaussianVec(sws_chr_gblur, 3.0);
- }else{
- src_filter.chrH= getIdentityVec();
- src_filter.chrV= getIdentityVec();
- }
-
- if(sws_chr_sharpen!=0.0){
- SwsVector *g= getConstVec(-1.0, 3);
- SwsVector *id= getConstVec(10.0/sws_chr_sharpen, 1);
- g->coeff[1]=2.0;
- addVec(id, g);
- convVec(src_filter.chrH, id);
- convVec(src_filter.chrV, id);
- freeVec(g);
- freeVec(id);
- }
-
- if(sws_lum_sharpen!=0.0){
- SwsVector *g= getConstVec(-1.0, 3);
- SwsVector *id= getConstVec(10.0/sws_lum_sharpen, 1);
- g->coeff[1]=2.0;
- addVec(id, g);
- convVec(src_filter.lumH, id);
- convVec(src_filter.lumV, id);
- freeVec(g);
- freeVec(id);
- }
-
- if(sws_chr_hshift)
- shiftVec(src_filter.chrH, sws_chr_hshift);
-
- if(sws_chr_vshift)
- shiftVec(src_filter.chrV, sws_chr_vshift);
-
- normalizeVec(src_filter.chrH, 1.0);
- normalizeVec(src_filter.chrV, 1.0);
- normalizeVec(src_filter.lumH, 1.0);
- normalizeVec(src_filter.lumV, 1.0);
-
- if(verbose > 1) printVec(src_filter.chrH);
- if(verbose > 1) printVec(src_filter.lumH);
-
- switch(sws_flags)
- {
- case 0: flags|= SWS_FAST_BILINEAR; break;
- case 1: flags|= SWS_BILINEAR; break;
- case 2: flags|= SWS_BICUBIC; break;
- case 3: flags|= SWS_X; break;
- case 4: flags|= SWS_POINT; break;
- case 5: flags|= SWS_AREA; break;
- default:flags|= SWS_BILINEAR; break;
- }
-
- return getSwsContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, &src_filter, NULL);
+// printf("%f %f %f %f %f\n", a,b,c,d,dist);
+ if(dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
+ else return getSplineCoeff( 0.0,
+ b+ 2.0*c + 3.0*d,
+ c + 3.0*d,
+ -b- 3.0*c - 6.0*d,
+ dist-1.0);
}
-
static inline void initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
int srcW, int dstW, int filterAlign, int one, int flags,
SwsVector *srcFilter, SwsVector *dstFilter)
double *filter=NULL;
double *filter2=NULL;
#ifdef ARCH_X86
- if(gCpuCaps.hasMMX)
+ if(flags & SWS_CPU_CAPS_MMX)
asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions)
#endif
xDstInSrc+= xInc;
}
}
- else if(xInc <= (1<<16) || (flags&SWS_FAST_BILINEAR)) // upscale
+ else if((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
{
int i;
int xDstInSrc;
int j;
(*filterPos)[i]= xx;
- if((flags & SWS_BICUBIC) || (flags & SWS_X))
- {
- double d= ABS(((xx+1)<<16) - xDstInSrc)/(double)(1<<16);
- double y1,y2,y3,y4;
- double A= -0.6;
- if(flags & SWS_BICUBIC){
- // Equation is from VirtualDub
- y1 = ( + A*d - 2.0*A*d*d + A*d*d*d);
- y2 = (+ 1.0 - (A+3.0)*d*d + (A+2.0)*d*d*d);
- y3 = ( - A*d + (2.0*A+3.0)*d*d - (A+2.0)*d*d*d);
- y4 = ( + A*d*d - A*d*d*d);
- }else{
- // cubic interpolation (derived it myself)
- y1 = ( -2.0*d + 3.0*d*d - 1.0*d*d*d)/6.0;
- y2 = (6.0 -3.0*d - 6.0*d*d + 3.0*d*d*d)/6.0;
- y3 = ( +6.0*d + 3.0*d*d - 3.0*d*d*d)/6.0;
- y4 = ( -1.0*d + 1.0*d*d*d)/6.0;
- }
-
- filter[i*filterSize + 0]= y1;
- filter[i*filterSize + 1]= y2;
- filter[i*filterSize + 2]= y3;
- filter[i*filterSize + 3]= y4;
- }
- else
- {
//Bilinear upscale / linear interpolate / Area averaging
for(j=0; j<filterSize; j++)
{
filter[i*filterSize + j]= coeff;
xx++;
}
- }
xDstInSrc+= xInc;
}
}
- else // downscale
+ else
{
- int xDstInSrc;
- ASSERT(dstW <= srcW)
+ double xDstInSrc;
+ double sizeFactor, filterSizeInSrc;
+ const double xInc1= (double)xInc / (double)(1<<16);
+ int param= (flags&SWS_PARAM_MASK)>>SWS_PARAM_SHIFT;
+
+ 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 ? 2.0*param : 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;
+ else {
+ sizeFactor= 0.0; //GCC warning killer
+ ASSERT(0)
+ }
+
+ if(xInc1 <= 1.0) filterSizeInSrc= sizeFactor; // upscale
+ else filterSizeInSrc= sizeFactor*srcW / (double)dstW;
- if(flags&SWS_BICUBIC) filterSize= (int)ceil(1 + 4.0*srcW / (double)dstW);
- else if(flags&SWS_X) filterSize= (int)ceil(1 + 4.0*srcW / (double)dstW);
- else if(flags&SWS_AREA) filterSize= (int)ceil(1 + 1.0*srcW / (double)dstW);
- else /* BILINEAR */ filterSize= (int)ceil(1 + 2.0*srcW / (double)dstW);
- filter= (double*)memalign(8, dstW*sizeof(double)*filterSize);
+ filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible
+ if(filterSize > srcW-2) filterSize=srcW-2;
- xDstInSrc= xInc/2 - 0x8000;
+ filter= (double*)memalign(16, dstW*sizeof(double)*filterSize);
+
+ xDstInSrc= xInc1 / 2.0 - 0.5;
for(i=0; i<dstW; i++)
{
- int xx= (int)((double)xDstInSrc/(double)(1<<16) - (filterSize-1)*0.5 + 0.5);
+ int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);
int j;
(*filterPos)[i]= xx;
for(j=0; j<filterSize; j++)
{
- double d= ABS((xx<<16) - xDstInSrc)/(double)xInc;
+ double d= ABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;
double coeff;
- if((flags & SWS_BICUBIC) || (flags & SWS_X))
+ if(flags & SWS_BICUBIC)
{
- double A= -0.75;
-// d*=2;
+ double A= param ? -param*0.01 : -0.60;
+
// Equation is from VirtualDub
if(d<1.0)
coeff = (1.0 - (A+3.0)*d*d + (A+2.0)*d*d*d);
else
coeff=0.0;
}
+/* else if(flags & SWS_X)
+ {
+ double p= param ? param*0.01 : 0.3;
+ coeff = d ? sin(d*PI)/(d*PI) : 1.0;
+ coeff*= pow(2.0, - p*d*d);
+ }*/
+ else if(flags & SWS_X)
+ {
+ double A= param ? param*0.1 : 1.0;
+
+ if(d<1.0)
+ coeff = cos(d*PI);
+ else
+ coeff=-1.0;
+ if(coeff<0.0) coeff= -pow(-coeff, A);
+ else coeff= pow( coeff, A);
+ coeff= coeff*0.5 + 0.5;
+ }
else if(flags & SWS_AREA)
{
- double srcPixelSize= (1<<16)/(double)xInc;
+ 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;
else coeff=0.0;
}
- else
+ else if(flags & SWS_GAUSS)
+ {
+ double p= param ? param*0.1 : 3.0;
+ coeff = pow(2.0, - p*d*d);
+ }
+ else if(flags & SWS_SINC)
+ {
+ coeff = d ? sin(d*PI)/(d*PI) : 1.0;
+ }
+ else if(flags & SWS_LANCZOS)
+ {
+ double p= param ? param : 3.0;
+ coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;
+ if(d>p) coeff=0;
+ }
+ else if(flags & SWS_BILINEAR)
{
coeff= 1.0 - d;
if(coeff<0) coeff=0;
}
+ else if(flags & SWS_SPLINE)
+ {
+ double p=-2.196152422706632;
+ coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);
+ }
+ else {
+ coeff= 0.0; //GCC warning killer
+ ASSERT(0)
+ }
+
filter[i*filterSize + j]= coeff;
xx++;
}
- xDstInSrc+= xInc;
+ xDstInSrc+= xInc1;
}
}
scaleFilter.coeff= filter + i*filterSize;
scaleFilter.length= filterSize;
- if(srcFilter) outVec= getConvVec(srcFilter, &scaleFilter);
+ if(srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);
else outVec= &scaleFilter;
ASSERT(outVec->length == filter2Size)
(*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
- if(outVec != &scaleFilter) freeVec(outVec);
+ if(outVec != &scaleFilter) sws_freeVec(outVec);
}
free(filter); filter=NULL;
if(cutOff > SWS_MAX_REDUCE_CUTOFF) break;
- /* preserve Monotonicity because the core cant handle the filter otherwise */
+ /* preserve Monotonicity because the core can't handle the filter otherwise */
if(i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
// Move filter coeffs left
if(min>minFilterSize) minFilterSize= min;
}
+ if (flags & SWS_CPU_CAPS_ALTIVEC) {
+ // we can handle the special case 4,
+ // so we don't want to go to the full 8
+ if (minFilterSize < 5)
+ filterAlign = 4;
+
+ // we really don't want to waste our time
+ // doing useless computation, so fall-back on
+ // the scalar C code for very small filter.
+ // vectorizing is worth it only if you have
+ // decent-sized vector.
+ if (minFilterSize < 3)
+ filterAlign = 1;
+ }
+
ASSERT(minFilterSize > 0)
filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
ASSERT(filterSize > 0)
for(i=0; i<dstW; i++)
{
int j;
+ double error=0;
double sum=0;
double scale= one;
+
for(j=0; j<filterSize; j++)
{
sum+= filter[i*filterSize + j];
}
scale/= sum;
- for(j=0; j<filterSize; j++)
+ for(j=0; j<*outFilterSize; j++)
{
- (*outFilter)[i*(*outFilterSize) + j]= (int)(filter[i*filterSize + j]*scale);
+ double v= filter[i*filterSize + j]*scale + error;
+ int intV= floor(v + 0.5);
+ (*outFilter)[i*(*outFilterSize) + j]= intV;
+ error = v - intV;
}
}
}
#endif // ARCH_X86
-//FIXME remove
-void SwScale_Init(){
-}
-
static void globalInit(){
// generating tables:
int i;
for(i=0; i<768; i++){
int c= MIN(MAX(i-256, 0), 255);
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;
-
+static SwsFunc getSwsFunc(int flags){
+
#ifdef RUNTIME_CPUDETECT
-#ifdef CAN_COMPILE_X86_ASM
+#ifdef ARCH_X86
// ordered per speed fasterst first
- if(gCpuCaps.hasMMX2)
- swScale= swScale_MMX2;
- else if(gCpuCaps.has3DNow)
- swScale= swScale_3DNow;
- else if(gCpuCaps.hasMMX)
- swScale= swScale_MMX;
+ if(flags & SWS_CPU_CAPS_MMX2)
+ return swScale_MMX2;
+ else if(flags & SWS_CPU_CAPS_3DNOW)
+ return swScale_3DNow;
+ else if(flags & SWS_CPU_CAPS_MMX)
+ return swScale_MMX;
else
- swScale= swScale_C;
+ return swScale_C;
#else
- swScale= swScale_C;
- cpuCaps.hasMMX2 = cpuCaps.hasMMX = cpuCaps.has3DNow = 0;
+#ifdef ARCH_POWERPC
+ if(flags & SWS_CPU_CAPS_ALTIVEC)
+ return swScale_altivec;
+ else
+ return swScale_C;
+#endif
+ return swScale_C;
#endif
#else //RUNTIME_CPUDETECT
#ifdef HAVE_MMX2
- swScale= swScale_MMX2;
- cpuCaps.has3DNow = 0;
+ return swScale_MMX2;
#elif defined (HAVE_3DNOW)
- swScale= swScale_3DNow;
- cpuCaps.hasMMX2 = 0;
+ return swScale_3DNow;
#elif defined (HAVE_MMX)
- swScale= swScale_MMX;
- cpuCaps.hasMMX2 = cpuCaps.has3DNow = 0;
+ return swScale_MMX;
+#elif defined (HAVE_ALTIVEC)
+ return swScale_altivec;
#else
- swScale= swScale_C;
- cpuCaps.hasMMX2 = cpuCaps.hasMMX = cpuCaps.has3DNow = 0;
+ return swScale_C;
#endif
#endif //!RUNTIME_CPUDETECT
}
-static void PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
+static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dstParam[], int dstStride[]){
uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
/* Copy Y plane */
}
}
dst = dstParam[1] + dstStride[1]*srcSliceY;
- if(c->srcFormat==IMGFMT_YV12)
- interleaveBytes( src[1],src[2],dst,c->srcW,srcSliceH,srcStride[1],srcStride[2],dstStride[0] );
- else /* I420 & IYUV */
- interleaveBytes( src[2],src[1],dst,c->srcW,srcSliceH,srcStride[2],srcStride[1],dstStride[0] );
-}
+ interleaveBytes( src[1],src[2],dst,c->srcW,srcSliceH,srcStride[1],srcStride[2],dstStride[0] );
+ return srcSliceH;
+}
-/* Warper functions for yuv2bgr */
-static void planarYuvToBgr(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
+static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dstParam[], int dstStride[]){
uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
- if(c->srcFormat==IMGFMT_YV12)
- yuv2rgb( dst,src[0],src[1],src[2],c->srcW,srcSliceH,dstStride[0],srcStride[0],srcStride[1] );
- else /* I420 & IYUV */
- yuv2rgb( dst,src[0],src[2],src[1],c->srcW,srcSliceH,dstStride[0],srcStride[0],srcStride[1] );
+ yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
+
+ return srcSliceH;
}
-static void PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
+static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dstParam[], int dstStride[]){
uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
- if(c->srcFormat==IMGFMT_YV12)
- yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
- else /* I420 & IYUV */
- yv12toyuy2( src[0],src[2],src[1],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
-}
+ yv12touyvy( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
-static void bgr24to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*3==srcStride[0]*4)
- rgb24to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
+ return srcSliceH;
+}
- for(i=0; i<srcSliceH; i++)
- {
- rgb24to32(srcPtr, dstPtr, c->srcW*3);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
+/* {RGB,BGR}{15,16,24,32} -> {RGB,BGR}{15,16,24,32} */
+static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
+ int srcSliceH, uint8_t* dst[], int dstStride[]){
+ const int srcFormat= c->srcFormat;
+ const int dstFormat= c->dstFormat;
+ const int srcBpp= ((srcFormat&0xFF) + 7)>>3;
+ const int dstBpp= ((dstFormat&0xFF) + 7)>>3;
+ const int srcId= (srcFormat&0xFF)>>2; // 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8
+ const int dstId= (dstFormat&0xFF)>>2;
+ void (*conv)(const uint8_t *src, uint8_t *dst, unsigned src_size)=NULL;
+
+ /* BGR -> BGR */
+ if( (isBGR(srcFormat) && isBGR(dstFormat))
+ || (isRGB(srcFormat) && isRGB(dstFormat))){
+ switch(srcId | (dstId<<4)){
+ case 0x34: conv= rgb16to15; break;
+ case 0x36: conv= rgb24to15; break;
+ case 0x38: conv= rgb32to15; break;
+ case 0x43: conv= rgb15to16; break;
+ case 0x46: conv= rgb24to16; break;
+ case 0x48: conv= rgb32to16; break;
+ case 0x63: conv= rgb15to24; break;
+ case 0x64: conv= rgb16to24; break;
+ case 0x68: conv= rgb32to24; break;
+ case 0x83: conv= rgb15to32; break;
+ case 0x84: conv= rgb16to32; break;
+ case 0x86: conv= rgb24to32; break;
+ default: MSG_ERR("swScaler: internal error %s -> %s converter\n",
+ vo_format_name(srcFormat), vo_format_name(dstFormat)); break;
}
- }
-}
+ }else if( (isBGR(srcFormat) && isRGB(dstFormat))
+ || (isRGB(srcFormat) && isBGR(dstFormat))){
+ switch(srcId | (dstId<<4)){
+ case 0x33: conv= rgb15tobgr15; break;
+ case 0x34: conv= rgb16tobgr15; break;
+ case 0x36: conv= rgb24tobgr15; break;
+ case 0x38: conv= rgb32tobgr15; break;
+ case 0x43: conv= rgb15tobgr16; break;
+ case 0x44: conv= rgb16tobgr16; break;
+ case 0x46: conv= rgb24tobgr16; break;
+ case 0x48: conv= rgb32tobgr16; break;
+ case 0x63: conv= rgb15tobgr24; break;
+ case 0x64: conv= rgb16tobgr24; break;
+ case 0x66: conv= rgb24tobgr24; break;
+ case 0x68: conv= rgb32tobgr24; break;
+ case 0x83: conv= rgb15tobgr32; break;
+ case 0x84: conv= rgb16tobgr32; break;
+ case 0x86: conv= rgb24tobgr32; break;
+ case 0x88: conv= rgb32tobgr32; break;
+ default: MSG_ERR("swScaler: internal error %s -> %s converter\n",
+ vo_format_name(srcFormat), vo_format_name(dstFormat)); break;
+ }
+ }else{
+ MSG_ERR("swScaler: internal error %s -> %s converter\n",
+ vo_format_name(srcFormat), vo_format_name(dstFormat));
+ }
-static void bgr24to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*3==srcStride[0]*2)
- rgb24to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
+ if(dstStride[0]*srcBpp == srcStride[0]*dstBpp)
+ conv(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
else
{
int i;
for(i=0; i<srcSliceH; i++)
{
- rgb24to16(srcPtr, dstPtr, c->srcW*3);
+ conv(srcPtr, dstPtr, c->srcW*srcBpp);
srcPtr+= srcStride[0];
dstPtr+= dstStride[0];
}
}
+ return srcSliceH;
}
-static void bgr24to15Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
+static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*3==srcStride[0]*2)
- rgb24to15(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
- for(i=0; i<srcSliceH; i++)
- {
- rgb24to15(srcPtr, dstPtr, c->srcW*3);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
- }
- }
+ rgb24toyv12(
+ src[0],
+ dst[0]+ srcSliceY *dstStride[0],
+ dst[1]+(srcSliceY>>1)*dstStride[1],
+ dst[2]+(srcSliceY>>1)*dstStride[2],
+ c->srcW, srcSliceH,
+ dstStride[0], dstStride[1], srcStride[0]);
+ return srcSliceH;
}
-static void bgr32to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
+static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*4==srcStride[0]*3)
- rgb32to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
+ int i;
+
+ /* copy Y */
+ if(srcStride[0]==dstStride[0])
+ memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
+ else{
uint8_t *srcPtr= src[0];
uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
for(i=0; i<srcSliceH; i++)
{
- rgb32to24(srcPtr, dstPtr, c->srcW<<2);
+ memcpy(dstPtr, srcPtr, c->srcW);
srcPtr+= srcStride[0];
dstPtr+= dstStride[0];
}
- }
-}
+ }
-static void bgr32to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*4==srcStride[0]*2)
- rgb32to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
-
- for(i=0; i<srcSliceH; i++)
- {
- rgb32to16(srcPtr, dstPtr, c->srcW<<2);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
- }
- }
-}
-
-static void bgr32to15Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*4==srcStride[0]*2)
- rgb32to15(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
-
- for(i=0; i<srcSliceH; i++)
- {
- rgb32to15(srcPtr, dstPtr, c->srcW<<2);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
- }
- }
-}
-
-static void bgr15to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]==srcStride[0])
- rgb15to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
-
- for(i=0; i<srcSliceH; i++)
- {
- rgb15to16(srcPtr, dstPtr, c->srcW<<1);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
- }
- }
-}
-
-static void bgr15to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*2==srcStride[0]*3)
- rgb15to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
-
- for(i=0; i<srcSliceH; i++)
- {
- rgb15to24(srcPtr, dstPtr, c->srcW<<1);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
- }
- }
-}
-
-static void bgr15to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*2==srcStride[0]*4)
- rgb15to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
-
- for(i=0; i<srcSliceH; i++)
- {
- rgb15to32(srcPtr, dstPtr, c->srcW<<1);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
- }
- }
-}
-
-static void bgr16to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*2==srcStride[0]*3)
- rgb16to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
-
- for(i=0; i<srcSliceH; i++)
- {
- rgb16to24(srcPtr, dstPtr, c->srcW<<1);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
- }
- }
-}
-
-static void bgr16to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- if(dstStride[0]*2==srcStride[0]*4)
- rgb16to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
- else
- {
- int i;
- uint8_t *srcPtr= src[0];
- uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
-
- for(i=0; i<srcSliceH; i++)
- {
- rgb16to32(srcPtr, dstPtr, c->srcW<<1);
- srcPtr+= srcStride[0];
- dstPtr+= dstStride[0];
- }
- }
-}
-
-static void bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
- int srcSliceH, uint8_t* dst[], int dstStride[]){
-
- rgb24toyv12(
- src[0],
- dst[0]+ srcSliceY *dstStride[0],
- dst[1]+(srcSliceY>>1)*dstStride[1],
- dst[2]+(srcSliceY>>1)*dstStride[2],
- c->srcW, srcSliceH,
- dstStride[0], dstStride[1], srcStride[0]);
-}
+ if(c->dstFormat==IMGFMT_YV12){
+ planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);
+ planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);
+ }else{
+ planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);
+ planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);
+ }
+ return srcSliceH;
+}
/**
* bring pointers in YUV order instead of YVU
*/
-static inline void orderYUV(int format, uint8_t * sortedP[], int sortedStride[], uint8_t * p[], int stride[]){
- if(format == IMGFMT_YV12){
+static inline void sws_orderYUV(int format, uint8_t * sortedP[], int sortedStride[], uint8_t * p[], int stride[]){
+ if(format == IMGFMT_YV12 || format == IMGFMT_YVU9
+ || format == IMGFMT_444P || format == IMGFMT_422P || format == IMGFMT_411P){
sortedP[0]= p[0];
- sortedP[1]= p[1];
- sortedP[2]= p[2];
+ sortedP[1]= p[2];
+ sortedP[2]= p[1];
sortedStride[0]= stride[0];
- sortedStride[1]= stride[1];
- sortedStride[2]= stride[2];
+ sortedStride[1]= stride[2];
+ sortedStride[2]= stride[1];
}
- else if(isPacked(format) || isGray(format))
+ else if(isPacked(format) || isGray(format) || format == IMGFMT_Y8)
{
sortedP[0]= p[0];
sortedP[1]=
sortedStride[1]=
sortedStride[2]= 0;
}
- else /* I420 */
+ else if(format == IMGFMT_I420 || format == IMGFMT_IYUV)
{
sortedP[0]= p[0];
- sortedP[1]= p[2];
- sortedP[2]= p[1];
+ sortedP[1]= p[1];
+ sortedP[2]= p[2];
sortedStride[0]= stride[0];
- sortedStride[1]= stride[2];
- sortedStride[2]= stride[1];
+ sortedStride[1]= stride[1];
+ sortedStride[2]= stride[2];
+ }else{
+ MSG_ERR("internal error in orderYUV\n");
}
}
/* unscaled copy like stuff (assumes nearly identical formats) */
-static void simpleCopy(SwsContext *c, uint8_t* srcParam[], int srcStrideParam[], int srcSliceY,
- int srcSliceH, uint8_t* dstParam[], int dstStrideParam[]){
-
- int srcStride[3];
- int dstStride[3];
- uint8_t *src[3];
- uint8_t *dst[3];
-
- orderYUV(c->srcFormat, src, srcStride, srcParam, srcStrideParam);
- orderYUV(c->dstFormat, dst, dstStride, dstParam, dstStrideParam);
+static int simpleCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
+ int srcSliceH, uint8_t* dst[], int dstStride[]){
if(isPacked(c->srcFormat))
{
int plane;
for(plane=0; plane<3; plane++)
{
- int length= plane==0 ? c->srcW : ((c->srcW+1)>>1);
- int y= plane==0 ? srcSliceY: ((srcSliceY+1)>>1);
- int height= plane==0 ? srcSliceH: ((srcSliceH+1)>>1);
+ int length= plane==0 ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
+ int y= plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
+ int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
if((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)
{
if(!isGray(c->dstFormat))
- memset(dst[plane], 0, dstStride[plane]*height);
+ memset(dst[plane], 128, dstStride[plane]*height);
}
else
{
}
}
}
+ return srcSliceH;
}
static int remove_dup_fourcc(int fourcc)
{
switch(fourcc)
{
- case IMGFMT_IYUV: return IMGFMT_I420;
+ case IMGFMT_I420:
+ case IMGFMT_IYUV: return IMGFMT_YV12;
case IMGFMT_Y8 : return IMGFMT_Y800;
+ case IMGFMT_IF09: return IMGFMT_YVU9;
default: return fourcc;
}
}
static void getSubSampleFactors(int *h, int *v, int format){
switch(format){
+ case IMGFMT_UYVY:
case IMGFMT_YUY2:
*h=1;
*v=0;
break;
case IMGFMT_YV12:
- case IMGFMT_I420:
+ case IMGFMT_Y800: //FIXME remove after different subsamplings are fully implemented
*h=1;
*v=1;
break;
*h=2;
*v=2;
break;
+ case IMGFMT_444P:
+ *h=0;
+ *v=0;
+ break;
+ case IMGFMT_422P:
+ *h=1;
+ *v=0;
+ break;
+ case IMGFMT_411P:
+ *h=2;
+ *v=0;
+ break;
default:
*h=0;
*v=0;
}
}
-SwsContext *getSwsContext(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat, int flags,
+static uint16_t roundToInt16(int64_t f){
+ int r= (f + (1<<15))>>16;
+ if(r<-0x7FFF) return 0x8000;
+ else if(r> 0x7FFF) return 0x7FFF;
+ else return r;
+}
+
+/**
+ * @param inv_table the yuv2rgb coeffs, normally Inverse_Table_6_9[x]
+ * @param fullRange if 1 then the luma range is 0..255 if 0 its 16..235
+ * @return -1 if not supported
+ */
+int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
+ 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;
+
+ if(isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
+ memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
+ memcpy(c->dstColorspaceTable, table, sizeof(int)*4);
+
+ c->brightness= brightness;
+ c->contrast = contrast;
+ c->saturation= saturation;
+ c->srcRange = srcRange;
+ c->dstRange = dstRange;
+
+ c->uOffset= 0x0400040004000400LL;
+ c->vOffset= 0x0400040004000400LL;
+
+ if(!srcRange){
+ cy= (cy*255) / 219;
+ oy= 16<<16;
+ }
+
+ 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;
+
+ c->yCoeff= roundToInt16(cy *8192) * 0x0001000100010001ULL;
+ c->vrCoeff= roundToInt16(crv*8192) * 0x0001000100010001ULL;
+ c->ubCoeff= roundToInt16(cbu*8192) * 0x0001000100010001ULL;
+ c->vgCoeff= roundToInt16(cgv*8192) * 0x0001000100010001ULL;
+ c->ugCoeff= roundToInt16(cgu*8192) * 0x0001000100010001ULL;
+ c->yOffset= roundToInt16(oy * 8) * 0x0001000100010001ULL;
+
+ yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
+ //FIXME factorize
+
+ return 0;
+}
+
+/**
+ * @return -1 if not supported
+ */
+int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
+ if(isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
+
+ *inv_table = c->srcColorspaceTable;
+ *table = c->dstColorspaceTable;
+ *srcRange = c->srcRange;
+ *dstRange = c->dstRange;
+ *brightness= c->brightness;
+ *contrast = c->contrast;
+ *saturation= c->saturation;
+
+ return 0;
+}
+
+SwsContext *sws_getContext(int srcW, int srcH, int origSrcFormat, int dstW, int dstH, int origDstFormat, int flags,
SwsFilter *srcFilter, SwsFilter *dstFilter){
SwsContext *c;
int i;
- int usesFilter;
- int unscaled;
+ int usesVFilter, usesHFilter;
+ int unscaled, needsDither;
+ int srcFormat, dstFormat;
SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
#ifdef ARCH_X86
- if(gCpuCaps.hasMMX)
+ if(flags & SWS_CPU_CAPS_MMX)
asm volatile("emms\n\t"::: "memory");
#endif
- if(swScale==NULL) globalInit();
- /* avoid dupplicate Formats, so we dont need to check to much */
- srcFormat = remove_dup_fourcc(srcFormat);
- dstFormat = remove_dup_fourcc(dstFormat);
+#ifndef RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
+ flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC);
+#ifdef HAVE_MMX2
+ flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
+#elif defined (HAVE_3DNOW)
+ flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
+#elif defined (HAVE_MMX)
+ flags |= SWS_CPU_CAPS_MMX;
+#elif defined (HAVE_ALTIVEC)
+ flags |= SWS_CPU_CAPS_ALTIVEC;
+#endif
+#endif
+ if(clip_table[512] != 255) globalInit();
+ if(rgb15to16 == NULL) sws_rgb2rgb_init(flags);
+
+ /* avoid duplicate Formats, so we don't need to check to much */
+ srcFormat = remove_dup_fourcc(origSrcFormat);
+ dstFormat = remove_dup_fourcc(origDstFormat);
unscaled = (srcW == dstW && srcH == dstH);
+ needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
+ && (dstFormat&0xFF)<24
+ && ((dstFormat&0xFF)<(srcFormat&0xFF) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
if(!isSupportedIn(srcFormat))
{
c->flags= flags;
c->dstFormat= dstFormat;
c->srcFormat= srcFormat;
+ c->origDstFormat= origDstFormat;
+ c->origSrcFormat= origSrcFormat;
+ c->vRounder= 4* 0x0001000100010001ULL;
+
+ usesHFilter= usesVFilter= 0;
+ if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesVFilter=1;
+ if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesHFilter=1;
+ if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesVFilter=1;
+ if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesHFilter=1;
+ if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesVFilter=1;
+ if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesHFilter=1;
+ if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesVFilter=1;
+ if(srcFilter->chrH!=NULL && srcFilter->chrH->length>1) usesHFilter=1;
- usesFilter=0;
- if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesFilter=1;
- if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesFilter=1;
- if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesFilter=1;
- if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesFilter=1;
- if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesFilter=1;
- if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesFilter=1;
- if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesFilter=1;
- if(srcFilter->chrH!=NULL && srcFilter->chrH->length>1) usesFilter=1;
-
- /* unscaled special Cases */
- if(unscaled && !usesFilter)
- {
- /* yv12_to_nv12 */
- if((srcFormat == IMGFMT_YV12||srcFormat==IMGFMT_I420)&&dstFormat == IMGFMT_NV12)
- {
- c->swScale= PlanarToNV12Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
- /* yv12_to_yuy2 */
- if((srcFormat == IMGFMT_YV12||srcFormat==IMGFMT_I420)&&dstFormat == IMGFMT_YUY2)
- {
- c->swScale= PlanarToYuy2Wrapper;
+ getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
+ getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
- /* yuv2bgr */
- if(isPlanarYUV(srcFormat) && isBGR(dstFormat))
- {
- // FIXME multiple yuv2rgb converters wont work that way cuz that thing is full of globals&statics
-#ifdef WORDS_BIGENDIAN
- if(dstFormat==IMGFMT_BGR32)
- yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_BGR);
- else
- yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_RGB);
-#else
- yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_RGB);
-#endif
- c->swScale= planarYuvToBgr;
+ // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation
+ if((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
+ // drop some chroma lines if the user wants it
+ c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
+ c->chrSrcVSubSample+= c->vChrDrop;
- /* simple copy */
- if(srcFormat == dstFormat
- || ((isPlanarYUV(srcFormat)||isGray(srcFormat)) && (isPlanarYUV(dstFormat)||isGray(dstFormat))))
- {
- c->swScale= simpleCopy;
+ // drop every 2. pixel for chroma calculation unless user wants full chroma
+ if((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP))
+ c->chrSrcHSubSample=1;
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
+ c->chrIntHSubSample= c->chrDstHSubSample;
+ c->chrIntVSubSample= c->chrSrcVSubSample;
- /* bgr32to24 & rgb32to24*/
- if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR24)
- ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB24))
- {
- c->swScale= bgr32to24Wrapper;
+ // note the -((-x)>>y) is so that we allways round toward +inf
+ c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
+ c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
+ c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
+ c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
+ sws_setColorspaceDetails(c, Inverse_Table_6_9[SWS_CS_DEFAULT], 0, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, 0, 0, 1<<16, 1<<16);
- /* bgr32to16 & rgb32to16*/
- if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR16)
- ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB16))
+ /* unscaled special Cases */
+ if(unscaled && !usesHFilter && !usesVFilter)
+ {
+ /* yv12_to_nv12 */
+ if(srcFormat == IMGFMT_YV12 && dstFormat == IMGFMT_NV12)
{
- c->swScale= bgr32to16Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
+ c->swScale= PlanarToNV12Wrapper;
}
-
- /* bgr32to15 & rgb32to15*/
- if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR15)
- ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB15))
+ /* yuv2bgr */
+ if((srcFormat==IMGFMT_YV12 || srcFormat==IMGFMT_422P) && (isBGR(dstFormat) || isRGB(dstFormat)))
{
- c->swScale= bgr32to15Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
+ c->swScale= yuv2rgb_get_func_ptr(c);
}
- /* bgr24to32 & rgb24to32*/
- if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR32)
- ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB32))
- {
- c->swScale= bgr24to32Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
-
- /* bgr24to16 & rgb24to16*/
- if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR16)
- ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB16))
- {
- c->swScale= bgr24to16Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
-
- /* bgr24to15 & rgb24to15*/
- if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR15)
- ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB15))
+ if( srcFormat==IMGFMT_YVU9 && dstFormat==IMGFMT_YV12 )
{
- c->swScale= bgr24to15Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
+ c->swScale= yvu9toyv12Wrapper;
}
- /* bgr15to16 */
- if(srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR16)
- {
- c->swScale= bgr15to16Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
-
- /* bgr15to24 */
- if((srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR24)
- ||(srcFormat==IMGFMT_RGB15 && dstFormat==IMGFMT_RGB24))
- {
- c->swScale= bgr15to24Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
-
- /* bgr15to32 */
- if((srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR32)
- ||(srcFormat==IMGFMT_RGB15 && dstFormat==IMGFMT_RGB32))
- {
- c->swScale= bgr15to32Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
- }
-
- /* bgr16to24 */
- if((srcFormat==IMGFMT_BGR16 && dstFormat==IMGFMT_BGR24)
- ||(srcFormat==IMGFMT_RGB16 && dstFormat==IMGFMT_RGB24))
- {
- c->swScale= bgr16to24Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
+ /* bgr24toYV12 */
+ if(srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_YV12)
+ c->swScale= bgr24toyv12Wrapper;
+
+ /* rgb/bgr -> rgb/bgr (no dither needed forms) */
+ if( (isBGR(srcFormat) || isRGB(srcFormat))
+ && (isBGR(dstFormat) || isRGB(dstFormat))
+ && !needsDither)
+ c->swScale= rgb2rgbWrapper;
+
+ /* LQ converters if -sws 0 or -sws 4*/
+ if(c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
+ /* rgb/bgr -> rgb/bgr (dither needed forms) */
+ if( (isBGR(srcFormat) || isRGB(srcFormat))
+ && (isBGR(dstFormat) || isRGB(dstFormat))
+ && needsDither)
+ c->swScale= rgb2rgbWrapper;
+
+ /* yv12_to_yuy2 */
+ if(srcFormat == IMGFMT_YV12 &&
+ (dstFormat == IMGFMT_YUY2 || dstFormat == IMGFMT_UYVY))
+ {
+ if (dstFormat == IMGFMT_YUY2)
+ c->swScale= PlanarToYuy2Wrapper;
+ else
+ c->swScale= PlanarToUyvyWrapper;
+ }
}
- /* bgr16to32 */
- if((srcFormat==IMGFMT_BGR16 && dstFormat==IMGFMT_BGR32)
- ||(srcFormat==IMGFMT_RGB16 && dstFormat==IMGFMT_RGB32))
+ /* simple copy */
+ if( srcFormat == dstFormat
+ || (isPlanarYUV(srcFormat) && isGray(dstFormat))
+ || (isPlanarYUV(dstFormat) && isGray(srcFormat))
+ )
{
- c->swScale= bgr16to32Wrapper;
-
- if(flags&SWS_PRINT_INFO)
- MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
- vo_format_name(srcFormat), vo_format_name(dstFormat));
- return c;
+ c->swScale= simpleCopy;
}
- /* bgr24toYV12 */
- if(srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_YV12)
- {
- c->swScale= bgr24toyv12Wrapper;
-
+ if(c->swScale){
if(flags&SWS_PRINT_INFO)
MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n",
vo_format_name(srcFormat), vo_format_name(dstFormat));
}
}
- if(cpuCaps.hasMMX2)
+ if(flags & SWS_CPU_CAPS_MMX2)
{
c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
if(!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
if(flags&SWS_PRINT_INFO)
MSG_INFO("SwScaler: output Width is not a multiple of 32 -> no MMX2 scaler\n");
}
+ if(usesHFilter) c->canMMX2BeUsed=0;
}
else
c->canMMX2BeUsed=0;
- getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
- getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
-
- // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation
- if((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
-
- // drop eery 2. pixel for chroma calculation unless user wants full chroma
- if((isBGR(srcFormat) || isRGB(srcFormat) || srcFormat==IMGFMT_YUY2) && !(flags&SWS_FULL_CHR_V))
- c->chrSrcVSubSample=1;
-
- // drop eery 2. pixel for chroma calculation unless user wants full chroma
- if((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP))
- c->chrSrcHSubSample=1;
-
- c->chrIntHSubSample= c->chrDstHSubSample;
- c->chrIntVSubSample= c->chrSrcVSubSample;
-
- // note the -((-x)>>y) is so that we allways round toward +inf
- c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
- c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
- c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
- c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
-/* printf("%d %d %d %d / %d %d %d %d //\n",
- c->chrSrcW,
-c->chrSrcH,
-c->chrDstW,
-c->chrDstH,
-srcW,
-srcH,
-dstW,
-dstH);*/
c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
c->lumXInc+= 20;
c->chrXInc+= 20;
}
- //we dont use the x86asm scaler if mmx is available
- else if(cpuCaps.hasMMX)
+ //we don't use the x86asm scaler if mmx is available
+ else if(flags & SWS_CPU_CAPS_MMX)
{
c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
/* precalculate horizontal scaler filter coefficients */
{
- const int filterAlign= cpuCaps.hasMMX ? 4 : 1;
+ const int filterAlign=
+ (flags & SWS_CPU_CAPS_MMX) ? 4 :
+ (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
+ 1;
initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
- srcW , dstW, filterAlign, 1<<14, flags,
+ srcW , dstW, filterAlign, 1<<14,
+ (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags,
srcFilter->lumH, dstFilter->lumH);
initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
- (srcW+1)>>1, c->chrDstW, filterAlign, 1<<14, flags,
+ c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
+ (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
srcFilter->chrH, dstFilter->chrH);
#ifdef ARCH_X86
-// cant downscale !!!
+// can't downscale !!!
if(c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
{
c->lumMmx2Filter = (int16_t*)memalign(8, (dstW /8+8)*sizeof(int16_t));
/* precalculate vertical scaler filter coefficients */
- initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
- srcH , dstH, 1, (1<<12)-4, flags,
- srcFilter->lumV, dstFilter->lumV);
- initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
- (srcH+1)>>1, c->chrDstH, 1, (1<<12)-4, flags,
- srcFilter->chrV, dstFilter->chrV);
-
- // Calculate Buffer Sizes so that they wont run out while handling these damn slices
+ {
+ const int filterAlign=
+ (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
+ 1;
+
+ initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
+ srcH , dstH, filterAlign, (1<<12)-4,
+ (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags,
+ srcFilter->lumV, dstFilter->lumV);
+ initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
+ c->chrSrcH, c->chrDstH, filterAlign, (1<<12)-4,
+ (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
+ srcFilter->chrV, dstFilter->chrV);
+ }
+
+ // Calculate Buffer Sizes so that they won't run out while handling these damn slices
c->vLumBufSize= c->vLumFilterSize;
c->vChrBufSize= c->vChrFilterSize;
for(i=0; i<dstH; i++)
{
int chrI= i*c->chrDstH / dstH;
int nextSlice= MAX(c->vLumFilterPos[i ] + c->vLumFilterSize - 1,
- ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<1));
- nextSlice&= ~1; // Slices start at even boundaries
+ ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
+
+ nextSlice>>= c->chrSrcVSubSample;
+ nextSlice<<= c->chrSrcVSubSample;
if(c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice)
c->vLumBufSize= nextSlice - c->vLumFilterPos[i ];
- if(c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>1))
- c->vChrBufSize= (nextSlice>>1) - c->vChrFilterPos[chrI];
+ if(c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
+ c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
}
// allocate pixbufs (we use dynamic allocation because otherwise we would need to
ASSERT(c->chrDstH <= dstH)
- // pack filter data for mmx code
- if(cpuCaps.hasMMX)
- {
- c->lumMmxFilter= (int16_t*)memalign(8, c->vLumFilterSize* dstH*4*sizeof(int16_t));
- c->chrMmxFilter= (int16_t*)memalign(8, c->vChrFilterSize*c->chrDstH*4*sizeof(int16_t));
- for(i=0; i<c->vLumFilterSize*dstH; i++)
- c->lumMmxFilter[4*i]=c->lumMmxFilter[4*i+1]=c->lumMmxFilter[4*i+2]=c->lumMmxFilter[4*i+3]=
- c->vLumFilter[i];
- for(i=0; i<c->vChrFilterSize*c->chrDstH; i++)
- c->chrMmxFilter[4*i]=c->chrMmxFilter[4*i+1]=c->chrMmxFilter[4*i+2]=c->chrMmxFilter[4*i+3]=
- c->vChrFilter[i];
- }
-
if(flags&SWS_PRINT_INFO)
{
#ifdef DITHER1XBPP
MSG_INFO("\nSwScaler: Nearest Neighbor / POINT scaler, ");
else if(flags&SWS_AREA)
MSG_INFO("\nSwScaler: Area Averageing scaler, ");
+ else if(flags&SWS_BICUBLIN)
+ MSG_INFO("\nSwScaler: luma BICUBIC / chroma BILINEAR scaler, ");
+ else if(flags&SWS_GAUSS)
+ MSG_INFO("\nSwScaler: Gaussian scaler, ");
+ else if(flags&SWS_SINC)
+ MSG_INFO("\nSwScaler: Sinc scaler, ");
+ else if(flags&SWS_LANCZOS)
+ MSG_INFO("\nSwScaler: Lanczos scaler, ");
+ else if(flags&SWS_SPLINE)
+ MSG_INFO("\nSwScaler: Bicubic spline scaler, ");
else
MSG_INFO("\nSwScaler: ehh flags invalid?! ");
MSG_INFO("from %s to %s ",
vo_format_name(srcFormat), vo_format_name(dstFormat));
- if(cpuCaps.hasMMX2)
+ if(flags & SWS_CPU_CAPS_MMX2)
MSG_INFO("using MMX2\n");
- else if(cpuCaps.has3DNow)
+ else if(flags & SWS_CPU_CAPS_3DNOW)
MSG_INFO("using 3DNOW\n");
- else if(cpuCaps.hasMMX)
+ else if(flags & SWS_CPU_CAPS_MMX)
MSG_INFO("using MMX\n");
- else
+ else if(flags & SWS_CPU_CAPS_ALTIVEC)
+ MSG_INFO("using AltiVec\n");
+ else
MSG_INFO("using C\n");
}
- if((flags & SWS_PRINT_INFO) && verbose)
+ if(flags & SWS_PRINT_INFO)
{
- if(cpuCaps.hasMMX)
+ if(flags & SWS_CPU_CAPS_MMX)
{
if(c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
MSG_V("SwScaler: using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
if(isPlanarYUV(dstFormat))
{
if(c->vLumFilterSize==1)
- MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", cpuCaps.hasMMX ? "MMX" : "C");
+ MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
else
- MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (YV12 like)\n", cpuCaps.hasMMX ? "MMX" : "C");
+ MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
}
else
{
if(c->vLumFilterSize==1 && c->vChrFilterSize==2)
MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
- "SwScaler: 2-tap scaler for vertical chrominance scaling (BGR)\n",cpuCaps.hasMMX ? "MMX" : "C");
+ "SwScaler: 2-tap scaler for vertical chrominance scaling (BGR)\n",(flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
else if(c->vLumFilterSize==2 && c->vChrFilterSize==2)
- MSG_V("SwScaler: using 2-tap linear %s scaler for vertical scaling (BGR)\n", cpuCaps.hasMMX ? "MMX" : "C");
+ MSG_V("SwScaler: using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
else
- MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (BGR)\n", cpuCaps.hasMMX ? "MMX" : "C");
+ MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
}
if(dstFormat==IMGFMT_BGR24)
MSG_V("SwScaler: using %s YV12->BGR24 Converter\n",
- cpuCaps.hasMMX2 ? "MMX2" : (cpuCaps.hasMMX ? "MMX" : "C"));
+ (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
else if(dstFormat==IMGFMT_BGR32)
- MSG_V("SwScaler: using %s YV12->BGR32 Converter\n", cpuCaps.hasMMX ? "MMX" : "C");
+ MSG_V("SwScaler: using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
else if(dstFormat==IMGFMT_BGR16)
- MSG_V("SwScaler: using %s YV12->BGR16 Converter\n", cpuCaps.hasMMX ? "MMX" : "C");
+ MSG_V("SwScaler: using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
else if(dstFormat==IMGFMT_BGR15)
- MSG_V("SwScaler: using %s YV12->BGR15 Converter\n", cpuCaps.hasMMX ? "MMX" : "C");
+ MSG_V("SwScaler: using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
MSG_V("SwScaler: %dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
}
- if((flags & SWS_PRINT_INFO) && verbose>1)
+ if(flags & SWS_PRINT_INFO)
{
MSG_DBG2("SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
}
- c->swScale= swScale;
+ c->swScale= getSwsFunc(flags);
return c;
}
+/**
+ * swscale warper, so we don't need to export the SwsContext.
+ * assumes planar YUV to be in YUV order instead of YVU
+ */
+int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
+ int srcSliceH, uint8_t* dst[], int dstStride[]){
+ //copy strides, so they can safely be modified
+ int srcStride2[3]= {srcStride[0], srcStride[1], srcStride[2]};
+ int dstStride2[3]= {dstStride[0], dstStride[1], dstStride[2]};
+ return c->swScale(c, src, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);
+}
+
+/**
+ * swscale warper, so we don't need to export the SwsContext
+ */
+int sws_scale(SwsContext *c, uint8_t* srcParam[], int srcStrideParam[], int srcSliceY,
+ int srcSliceH, uint8_t* dstParam[], int dstStrideParam[]){
+ int srcStride[3];
+ int dstStride[3];
+ uint8_t *src[3];
+ uint8_t *dst[3];
+ sws_orderYUV(c->origSrcFormat, src, srcStride, srcParam, srcStrideParam);
+ sws_orderYUV(c->origDstFormat, dst, dstStride, dstParam, dstStrideParam);
+//printf("sws: slice %d %d\n", srcSliceY, srcSliceH);
+
+ return c->swScale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
+}
+
+SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
+ float lumaSharpen, float chromaSharpen,
+ float chromaHShift, float chromaVShift,
+ int verbose)
+{
+ SwsFilter *filter= malloc(sizeof(SwsFilter));
+
+ if(lumaGBlur!=0.0){
+ filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
+ filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
+ }else{
+ filter->lumH= sws_getIdentityVec();
+ filter->lumV= sws_getIdentityVec();
+ }
+
+ if(chromaGBlur!=0.0){
+ filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
+ filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
+ }else{
+ filter->chrH= sws_getIdentityVec();
+ filter->chrV= sws_getIdentityVec();
+ }
+
+ if(chromaSharpen!=0.0){
+ SwsVector *g= sws_getConstVec(-1.0, 3);
+ SwsVector *id= sws_getConstVec(10.0/chromaSharpen, 1);
+ g->coeff[1]=2.0;
+ sws_addVec(id, g);
+ sws_convVec(filter->chrH, id);
+ sws_convVec(filter->chrV, id);
+ sws_freeVec(g);
+ sws_freeVec(id);
+ }
+
+ if(lumaSharpen!=0.0){
+ SwsVector *g= sws_getConstVec(-1.0, 3);
+ SwsVector *id= sws_getConstVec(10.0/lumaSharpen, 1);
+ g->coeff[1]=2.0;
+ sws_addVec(id, g);
+ sws_convVec(filter->lumH, id);
+ sws_convVec(filter->lumV, id);
+ sws_freeVec(g);
+ sws_freeVec(id);
+ }
+
+ if(chromaHShift != 0.0)
+ sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
+
+ if(chromaVShift != 0.0)
+ sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
+
+ sws_normalizeVec(filter->chrH, 1.0);
+ sws_normalizeVec(filter->chrV, 1.0);
+ sws_normalizeVec(filter->lumH, 1.0);
+ sws_normalizeVec(filter->lumV, 1.0);
+
+ if(verbose) sws_printVec(filter->chrH);
+ if(verbose) sws_printVec(filter->lumH);
+
+ return filter;
+}
+
/**
* returns a normalized gaussian curve used to filter stuff
* quality=3 is high quality, lowwer is lowwer quality
*/
-
-SwsVector *getGaussianVec(double variance, double quality){
+SwsVector *sws_getGaussianVec(double variance, double quality){
const int length= (int)(variance*quality + 0.5) | 1;
int i;
double *coeff= memalign(sizeof(double), length*sizeof(double));
coeff[i]= exp( -dist*dist/(2*variance*variance) ) / sqrt(2*variance*PI);
}
- normalizeVec(vec, 1.0);
+ sws_normalizeVec(vec, 1.0);
return vec;
}
-SwsVector *getConstVec(double c, int length){
+SwsVector *sws_getConstVec(double c, int length){
int i;
double *coeff= memalign(sizeof(double), length*sizeof(double));
SwsVector *vec= malloc(sizeof(SwsVector));
}
-SwsVector *getIdentityVec(void){
+SwsVector *sws_getIdentityVec(void){
double *coeff= memalign(sizeof(double), sizeof(double));
SwsVector *vec= malloc(sizeof(SwsVector));
coeff[0]= 1.0;
return vec;
}
-void normalizeVec(SwsVector *a, double height){
+void sws_normalizeVec(SwsVector *a, double height){
int i;
double sum=0;
double inv;
inv= height/sum;
for(i=0; i<a->length; i++)
- a->coeff[i]*= height;
+ a->coeff[i]*= inv;
}
-void scaleVec(SwsVector *a, double scalar){
+void sws_scaleVec(SwsVector *a, double scalar){
int i;
for(i=0; i<a->length; i++)
a->coeff[i]*= scalar;
}
-static SwsVector *getConvVec(SwsVector *a, SwsVector *b){
+static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
int length= a->length + b->length - 1;
double *coeff= memalign(sizeof(double), length*sizeof(double));
int i, j;
return vec;
}
-static SwsVector *sumVec(SwsVector *a, SwsVector *b){
+static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
int length= MAX(a->length, b->length);
double *coeff= memalign(sizeof(double), length*sizeof(double));
int i;
return vec;
}
-static SwsVector *diffVec(SwsVector *a, SwsVector *b){
+static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
int length= MAX(a->length, b->length);
double *coeff= memalign(sizeof(double), length*sizeof(double));
int i;
}
/* shift left / or right if "shift" is negative */
-static SwsVector *getShiftedVec(SwsVector *a, int shift){
+static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
int length= a->length + ABS(shift)*2;
double *coeff= memalign(sizeof(double), length*sizeof(double));
int i;
return vec;
}
-void shiftVec(SwsVector *a, int shift){
- SwsVector *shifted= getShiftedVec(a, shift);
+void sws_shiftVec(SwsVector *a, int shift){
+ SwsVector *shifted= sws_getShiftedVec(a, shift);
free(a->coeff);
a->coeff= shifted->coeff;
a->length= shifted->length;
free(shifted);
}
-void addVec(SwsVector *a, SwsVector *b){
- SwsVector *sum= sumVec(a, b);
+void sws_addVec(SwsVector *a, SwsVector *b){
+ SwsVector *sum= sws_sumVec(a, b);
free(a->coeff);
a->coeff= sum->coeff;
a->length= sum->length;
free(sum);
}
-void subVec(SwsVector *a, SwsVector *b){
- SwsVector *diff= diffVec(a, b);
+void sws_subVec(SwsVector *a, SwsVector *b){
+ SwsVector *diff= sws_diffVec(a, b);
free(a->coeff);
a->coeff= diff->coeff;
a->length= diff->length;
free(diff);
}
-void convVec(SwsVector *a, SwsVector *b){
- SwsVector *conv= getConvVec(a, b);
- free(a->coeff);
+void sws_convVec(SwsVector *a, SwsVector *b){
+ SwsVector *conv= sws_getConvVec(a, b);
+ free(a->coeff);
a->coeff= conv->coeff;
a->length= conv->length;
free(conv);
}
-SwsVector *cloneVec(SwsVector *a){
+SwsVector *sws_cloneVec(SwsVector *a){
double *coeff= memalign(sizeof(double), a->length*sizeof(double));
int i;
SwsVector *vec= malloc(sizeof(SwsVector));
return vec;
}
-void printVec(SwsVector *a){
+void sws_printVec(SwsVector *a){
int i;
double max=0;
double min=0;
}
}
-void freeVec(SwsVector *a){
+void sws_freeVec(SwsVector *a){
if(!a) return;
if(a->coeff) free(a->coeff);
a->coeff=NULL;
free(a);
}
-void freeSwsContext(SwsContext *c){
+void sws_freeFilter(SwsFilter *filter){
+ if(!filter) return;
+
+ if(filter->lumH) sws_freeVec(filter->lumH);
+ if(filter->lumV) sws_freeVec(filter->lumV);
+ if(filter->chrH) sws_freeVec(filter->chrH);
+ if(filter->chrV) sws_freeVec(filter->chrV);
+ free(filter);
+}
+
+
+void sws_freeContext(SwsContext *c){
int i;
if(!c) return;
if(c->hChrFilterPos) free(c->hChrFilterPos);
c->hChrFilterPos = NULL;
- if(c->lumMmxFilter) free(c->lumMmxFilter);
- c->lumMmxFilter = NULL;
- if(c->chrMmxFilter) free(c->chrMmxFilter);
- c->chrMmxFilter = NULL;
-
if(c->lumMmx2Filter) free(c->lumMmx2Filter);
c->lumMmx2Filter=NULL;
if(c->chrMmx2Filter) free(c->chrMmx2Filter);
c->lumMmx2FilterPos=NULL;
if(c->chrMmx2FilterPos) free(c->chrMmx2FilterPos);
c->chrMmx2FilterPos=NULL;
+ if(c->yuvTable) free(c->yuvTable);
+ c->yuvTable=NULL;
free(c);
}
-