Libav
|
00001 /* 00002 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at> 00003 * 00004 * This file is part of FFmpeg. 00005 * 00006 * FFmpeg is free software; you can redistribute it and/or 00007 * modify it under the terms of the GNU Lesser General Public 00008 * License as published by the Free Software Foundation; either 00009 * version 2.1 of the License, or (at your option) any later version. 00010 * 00011 * FFmpeg is distributed in the hope that it will be useful, 00012 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00013 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00014 * Lesser General Public License for more details. 00015 * 00016 * You should have received a copy of the GNU Lesser General Public 00017 * License along with FFmpeg; if not, write to the Free Software 00018 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00019 */ 00020 00021 #define _SVID_SOURCE //needed for MAP_ANONYMOUS 00022 #include <inttypes.h> 00023 #include <string.h> 00024 #include <math.h> 00025 #include <stdio.h> 00026 #include "config.h" 00027 #include <assert.h> 00028 #if HAVE_SYS_MMAN_H 00029 #include <sys/mman.h> 00030 #if defined(MAP_ANON) && !defined(MAP_ANONYMOUS) 00031 #define MAP_ANONYMOUS MAP_ANON 00032 #endif 00033 #endif 00034 #if HAVE_VIRTUALALLOC 00035 #define WIN32_LEAN_AND_MEAN 00036 #include <windows.h> 00037 #endif 00038 #include "swscale.h" 00039 #include "swscale_internal.h" 00040 #include "rgb2rgb.h" 00041 #include "libavutil/intreadwrite.h" 00042 #include "libavutil/x86_cpu.h" 00043 #include "libavutil/avutil.h" 00044 #include "libavutil/bswap.h" 00045 #include "libavutil/mathematics.h" 00046 #include "libavutil/pixdesc.h" 00047 00048 unsigned swscale_version(void) 00049 { 00050 return LIBSWSCALE_VERSION_INT; 00051 } 00052 00053 const char *swscale_configuration(void) 00054 { 00055 return FFMPEG_CONFIGURATION; 00056 } 00057 00058 const char *swscale_license(void) 00059 { 00060 #define LICENSE_PREFIX "libswscale license: " 00061 return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1; 00062 } 00063 00064 #define RET 0xC3 //near return opcode for x86 00065 00066 #define isSupportedIn(x) ( \ 00067 (x)==PIX_FMT_YUV420P \ 00068 || (x)==PIX_FMT_YUVA420P \ 00069 || (x)==PIX_FMT_YUYV422 \ 00070 || (x)==PIX_FMT_UYVY422 \ 00071 || (x)==PIX_FMT_RGB48BE \ 00072 || (x)==PIX_FMT_RGB48LE \ 00073 || (x)==PIX_FMT_RGB32 \ 00074 || (x)==PIX_FMT_RGB32_1 \ 00075 || (x)==PIX_FMT_BGR24 \ 00076 || (x)==PIX_FMT_BGR565 \ 00077 || (x)==PIX_FMT_BGR555 \ 00078 || (x)==PIX_FMT_BGR32 \ 00079 || (x)==PIX_FMT_BGR32_1 \ 00080 || (x)==PIX_FMT_RGB24 \ 00081 || (x)==PIX_FMT_RGB565 \ 00082 || (x)==PIX_FMT_RGB555 \ 00083 || (x)==PIX_FMT_GRAY8 \ 00084 || (x)==PIX_FMT_YUV410P \ 00085 || (x)==PIX_FMT_YUV440P \ 00086 || (x)==PIX_FMT_NV12 \ 00087 || (x)==PIX_FMT_NV21 \ 00088 || (x)==PIX_FMT_GRAY16BE \ 00089 || (x)==PIX_FMT_GRAY16LE \ 00090 || (x)==PIX_FMT_YUV444P \ 00091 || (x)==PIX_FMT_YUV422P \ 00092 || (x)==PIX_FMT_YUV411P \ 00093 || (x)==PIX_FMT_YUVJ420P \ 00094 || (x)==PIX_FMT_YUVJ422P \ 00095 || (x)==PIX_FMT_YUVJ440P \ 00096 || (x)==PIX_FMT_YUVJ444P \ 00097 || (x)==PIX_FMT_PAL8 \ 00098 || (x)==PIX_FMT_BGR8 \ 00099 || (x)==PIX_FMT_RGB8 \ 00100 || (x)==PIX_FMT_BGR4_BYTE \ 00101 || (x)==PIX_FMT_RGB4_BYTE \ 00102 || (x)==PIX_FMT_YUV440P \ 00103 || (x)==PIX_FMT_MONOWHITE \ 00104 || (x)==PIX_FMT_MONOBLACK \ 00105 || (x)==PIX_FMT_YUV420P16LE \ 00106 || (x)==PIX_FMT_YUV422P16LE \ 00107 || (x)==PIX_FMT_YUV444P16LE \ 00108 || (x)==PIX_FMT_YUV420P16BE \ 00109 || (x)==PIX_FMT_YUV422P16BE \ 00110 || (x)==PIX_FMT_YUV444P16BE \ 00111 ) 00112 00113 int sws_isSupportedInput(enum PixelFormat pix_fmt) 00114 { 00115 return isSupportedIn(pix_fmt); 00116 } 00117 00118 #define isSupportedOut(x) ( \ 00119 (x)==PIX_FMT_YUV420P \ 00120 || (x)==PIX_FMT_YUVA420P \ 00121 || (x)==PIX_FMT_YUYV422 \ 00122 || (x)==PIX_FMT_UYVY422 \ 00123 || (x)==PIX_FMT_YUV444P \ 00124 || (x)==PIX_FMT_YUV422P \ 00125 || (x)==PIX_FMT_YUV411P \ 00126 || (x)==PIX_FMT_YUVJ420P \ 00127 || (x)==PIX_FMT_YUVJ422P \ 00128 || (x)==PIX_FMT_YUVJ440P \ 00129 || (x)==PIX_FMT_YUVJ444P \ 00130 || isAnyRGB(x) \ 00131 || (x)==PIX_FMT_NV12 \ 00132 || (x)==PIX_FMT_NV21 \ 00133 || (x)==PIX_FMT_GRAY16BE \ 00134 || (x)==PIX_FMT_GRAY16LE \ 00135 || (x)==PIX_FMT_GRAY8 \ 00136 || (x)==PIX_FMT_YUV410P \ 00137 || (x)==PIX_FMT_YUV440P \ 00138 || (x)==PIX_FMT_YUV420P16LE \ 00139 || (x)==PIX_FMT_YUV422P16LE \ 00140 || (x)==PIX_FMT_YUV444P16LE \ 00141 || (x)==PIX_FMT_YUV420P16BE \ 00142 || (x)==PIX_FMT_YUV422P16BE \ 00143 || (x)==PIX_FMT_YUV444P16BE \ 00144 ) 00145 00146 int sws_isSupportedOutput(enum PixelFormat pix_fmt) 00147 { 00148 return isSupportedOut(pix_fmt); 00149 } 00150 00151 extern const int32_t ff_yuv2rgb_coeffs[8][4]; 00152 00153 const char *sws_format_name(enum PixelFormat format) 00154 { 00155 if ((unsigned)format < PIX_FMT_NB && av_pix_fmt_descriptors[format].name) 00156 return av_pix_fmt_descriptors[format].name; 00157 else 00158 return "Unknown format"; 00159 } 00160 00161 static double getSplineCoeff(double a, double b, double c, double d, double dist) 00162 { 00163 // printf("%f %f %f %f %f\n", a,b,c,d,dist); 00164 if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a; 00165 else return getSplineCoeff( 0.0, 00166 b+ 2.0*c + 3.0*d, 00167 c + 3.0*d, 00168 -b- 3.0*c - 6.0*d, 00169 dist-1.0); 00170 } 00171 00172 static int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc, 00173 int srcW, int dstW, int filterAlign, int one, int flags, 00174 SwsVector *srcFilter, SwsVector *dstFilter, double param[2]) 00175 { 00176 int i; 00177 int filterSize; 00178 int filter2Size; 00179 int minFilterSize; 00180 int64_t *filter=NULL; 00181 int64_t *filter2=NULL; 00182 const int64_t fone= 1LL<<54; 00183 int ret= -1; 00184 #if ARCH_X86 00185 if (flags & SWS_CPU_CAPS_MMX) 00186 __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions) 00187 #endif 00188 00189 // NOTE: the +1 is for the MMX scaler which reads over the end 00190 FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW+1)*sizeof(int16_t), fail); 00191 00192 if (FFABS(xInc - 0x10000) <10) { // unscaled 00193 int i; 00194 filterSize= 1; 00195 FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail); 00196 00197 for (i=0; i<dstW; i++) { 00198 filter[i*filterSize]= fone; 00199 (*filterPos)[i]=i; 00200 } 00201 00202 } else if (flags&SWS_POINT) { // lame looking point sampling mode 00203 int i; 00204 int xDstInSrc; 00205 filterSize= 1; 00206 FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail); 00207 00208 xDstInSrc= xInc/2 - 0x8000; 00209 for (i=0; i<dstW; i++) { 00210 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16; 00211 00212 (*filterPos)[i]= xx; 00213 filter[i]= fone; 00214 xDstInSrc+= xInc; 00215 } 00216 } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale 00217 int i; 00218 int xDstInSrc; 00219 filterSize= 2; 00220 FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail); 00221 00222 xDstInSrc= xInc/2 - 0x8000; 00223 for (i=0; i<dstW; i++) { 00224 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16; 00225 int j; 00226 00227 (*filterPos)[i]= xx; 00228 //bilinear upscale / linear interpolate / area averaging 00229 for (j=0; j<filterSize; j++) { 00230 int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16); 00231 if (coeff<0) coeff=0; 00232 filter[i*filterSize + j]= coeff; 00233 xx++; 00234 } 00235 xDstInSrc+= xInc; 00236 } 00237 } else { 00238 int xDstInSrc; 00239 int sizeFactor; 00240 00241 if (flags&SWS_BICUBIC) sizeFactor= 4; 00242 else if (flags&SWS_X) sizeFactor= 8; 00243 else if (flags&SWS_AREA) sizeFactor= 1; //downscale only, for upscale it is bilinear 00244 else if (flags&SWS_GAUSS) sizeFactor= 8; // infinite ;) 00245 else if (flags&SWS_LANCZOS) sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6; 00246 else if (flags&SWS_SINC) sizeFactor= 20; // infinite ;) 00247 else if (flags&SWS_SPLINE) sizeFactor= 20; // infinite ;) 00248 else if (flags&SWS_BILINEAR) sizeFactor= 2; 00249 else { 00250 sizeFactor= 0; //GCC warning killer 00251 assert(0); 00252 } 00253 00254 if (xInc <= 1<<16) filterSize= 1 + sizeFactor; // upscale 00255 else filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW; 00256 00257 if (filterSize > srcW-2) filterSize=srcW-2; 00258 00259 FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail); 00260 00261 xDstInSrc= xInc - 0x10000; 00262 for (i=0; i<dstW; i++) { 00263 int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17); 00264 int j; 00265 (*filterPos)[i]= xx; 00266 for (j=0; j<filterSize; j++) { 00267 int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13; 00268 double floatd; 00269 int64_t coeff; 00270 00271 if (xInc > 1<<16) 00272 d= d*dstW/srcW; 00273 floatd= d * (1.0/(1<<30)); 00274 00275 if (flags & SWS_BICUBIC) { 00276 int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1<<24); 00277 int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24); 00278 int64_t dd = ( d*d)>>30; 00279 int64_t ddd= (dd*d)>>30; 00280 00281 if (d < 1LL<<30) 00282 coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30); 00283 else if (d < 1LL<<31) 00284 coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30); 00285 else 00286 coeff=0.0; 00287 coeff *= fone>>(30+24); 00288 } 00289 /* else if (flags & SWS_X) { 00290 double p= param ? param*0.01 : 0.3; 00291 coeff = d ? sin(d*PI)/(d*PI) : 1.0; 00292 coeff*= pow(2.0, - p*d*d); 00293 }*/ 00294 else if (flags & SWS_X) { 00295 double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0; 00296 double c; 00297 00298 if (floatd<1.0) 00299 c = cos(floatd*M_PI); 00300 else 00301 c=-1.0; 00302 if (c<0.0) c= -pow(-c, A); 00303 else c= pow( c, A); 00304 coeff= (c*0.5 + 0.5)*fone; 00305 } else if (flags & SWS_AREA) { 00306 int64_t d2= d - (1<<29); 00307 if (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16)); 00308 else if (d2*xInc < (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16)); 00309 else coeff=0.0; 00310 coeff *= fone>>(30+16); 00311 } else if (flags & SWS_GAUSS) { 00312 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0; 00313 coeff = (pow(2.0, - p*floatd*floatd))*fone; 00314 } else if (flags & SWS_SINC) { 00315 coeff = (d ? sin(floatd*M_PI)/(floatd*M_PI) : 1.0)*fone; 00316 } else if (flags & SWS_LANCZOS) { 00317 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0; 00318 coeff = (d ? sin(floatd*M_PI)*sin(floatd*M_PI/p)/(floatd*floatd*M_PI*M_PI/p) : 1.0)*fone; 00319 if (floatd>p) coeff=0; 00320 } else if (flags & SWS_BILINEAR) { 00321 coeff= (1<<30) - d; 00322 if (coeff<0) coeff=0; 00323 coeff *= fone >> 30; 00324 } else if (flags & SWS_SPLINE) { 00325 double p=-2.196152422706632; 00326 coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone; 00327 } else { 00328 coeff= 0.0; //GCC warning killer 00329 assert(0); 00330 } 00331 00332 filter[i*filterSize + j]= coeff; 00333 xx++; 00334 } 00335 xDstInSrc+= 2*xInc; 00336 } 00337 } 00338 00339 /* apply src & dst Filter to filter -> filter2 00340 av_free(filter); 00341 */ 00342 assert(filterSize>0); 00343 filter2Size= filterSize; 00344 if (srcFilter) filter2Size+= srcFilter->length - 1; 00345 if (dstFilter) filter2Size+= dstFilter->length - 1; 00346 assert(filter2Size>0); 00347 FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail); 00348 00349 for (i=0; i<dstW; i++) { 00350 int j, k; 00351 00352 if(srcFilter) { 00353 for (k=0; k<srcFilter->length; k++) { 00354 for (j=0; j<filterSize; j++) 00355 filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j]; 00356 } 00357 } else { 00358 for (j=0; j<filterSize; j++) 00359 filter2[i*filter2Size + j]= filter[i*filterSize + j]; 00360 } 00361 //FIXME dstFilter 00362 00363 (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2; 00364 } 00365 av_freep(&filter); 00366 00367 /* try to reduce the filter-size (step1 find size and shift left) */ 00368 // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not). 00369 minFilterSize= 0; 00370 for (i=dstW-1; i>=0; i--) { 00371 int min= filter2Size; 00372 int j; 00373 int64_t cutOff=0.0; 00374 00375 /* get rid of near zero elements on the left by shifting left */ 00376 for (j=0; j<filter2Size; j++) { 00377 int k; 00378 cutOff += FFABS(filter2[i*filter2Size]); 00379 00380 if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break; 00381 00382 /* preserve monotonicity because the core can't handle the filter otherwise */ 00383 if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break; 00384 00385 // move filter coefficients left 00386 for (k=1; k<filter2Size; k++) 00387 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k]; 00388 filter2[i*filter2Size + k - 1]= 0; 00389 (*filterPos)[i]++; 00390 } 00391 00392 cutOff=0; 00393 /* count near zeros on the right */ 00394 for (j=filter2Size-1; j>0; j--) { 00395 cutOff += FFABS(filter2[i*filter2Size + j]); 00396 00397 if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break; 00398 min--; 00399 } 00400 00401 if (min>minFilterSize) minFilterSize= min; 00402 } 00403 00404 if (flags & SWS_CPU_CAPS_ALTIVEC) { 00405 // we can handle the special case 4, 00406 // so we don't want to go to the full 8 00407 if (minFilterSize < 5) 00408 filterAlign = 4; 00409 00410 // We really don't want to waste our time 00411 // doing useless computation, so fall back on 00412 // the scalar C code for very small filters. 00413 // Vectorizing is worth it only if you have a 00414 // decent-sized vector. 00415 if (minFilterSize < 3) 00416 filterAlign = 1; 00417 } 00418 00419 if (flags & SWS_CPU_CAPS_MMX) { 00420 // special case for unscaled vertical filtering 00421 if (minFilterSize == 1 && filterAlign == 2) 00422 filterAlign= 1; 00423 } 00424 00425 assert(minFilterSize > 0); 00426 filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1)); 00427 assert(filterSize > 0); 00428 filter= av_malloc(filterSize*dstW*sizeof(*filter)); 00429 if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter) 00430 goto fail; 00431 *outFilterSize= filterSize; 00432 00433 if (flags&SWS_PRINT_INFO) 00434 av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize); 00435 /* try to reduce the filter-size (step2 reduce it) */ 00436 for (i=0; i<dstW; i++) { 00437 int j; 00438 00439 for (j=0; j<filterSize; j++) { 00440 if (j>=filter2Size) filter[i*filterSize + j]= 0; 00441 else filter[i*filterSize + j]= filter2[i*filter2Size + j]; 00442 if((flags & SWS_BITEXACT) && j>=minFilterSize) 00443 filter[i*filterSize + j]= 0; 00444 } 00445 } 00446 00447 //FIXME try to align filterPos if possible 00448 00449 //fix borders 00450 for (i=0; i<dstW; i++) { 00451 int j; 00452 if ((*filterPos)[i] < 0) { 00453 // move filter coefficients left to compensate for filterPos 00454 for (j=1; j<filterSize; j++) { 00455 int left= FFMAX(j + (*filterPos)[i], 0); 00456 filter[i*filterSize + left] += filter[i*filterSize + j]; 00457 filter[i*filterSize + j]=0; 00458 } 00459 (*filterPos)[i]= 0; 00460 } 00461 00462 if ((*filterPos)[i] + filterSize > srcW) { 00463 int shift= (*filterPos)[i] + filterSize - srcW; 00464 // move filter coefficients right to compensate for filterPos 00465 for (j=filterSize-2; j>=0; j--) { 00466 int right= FFMIN(j + shift, filterSize-1); 00467 filter[i*filterSize +right] += filter[i*filterSize +j]; 00468 filter[i*filterSize +j]=0; 00469 } 00470 (*filterPos)[i]= srcW - filterSize; 00471 } 00472 } 00473 00474 // Note the +1 is for the MMX scaler which reads over the end 00475 /* align at 16 for AltiVec (needed by hScale_altivec_real) */ 00476 FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+1)*sizeof(int16_t), fail); 00477 00478 /* normalize & store in outFilter */ 00479 for (i=0; i<dstW; i++) { 00480 int j; 00481 int64_t error=0; 00482 int64_t sum=0; 00483 00484 for (j=0; j<filterSize; j++) { 00485 sum+= filter[i*filterSize + j]; 00486 } 00487 sum= (sum + one/2)/ one; 00488 for (j=0; j<*outFilterSize; j++) { 00489 int64_t v= filter[i*filterSize + j] + error; 00490 int intV= ROUNDED_DIV(v, sum); 00491 (*outFilter)[i*(*outFilterSize) + j]= intV; 00492 error= v - intV*sum; 00493 } 00494 } 00495 00496 (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end 00497 for (i=0; i<*outFilterSize; i++) { 00498 int j= dstW*(*outFilterSize); 00499 (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)]; 00500 } 00501 00502 ret=0; 00503 fail: 00504 av_free(filter); 00505 av_free(filter2); 00506 return ret; 00507 } 00508 00509 #if ARCH_X86 && (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) 00510 static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits) 00511 { 00512 uint8_t *fragmentA; 00513 x86_reg imm8OfPShufW1A; 00514 x86_reg imm8OfPShufW2A; 00515 x86_reg fragmentLengthA; 00516 uint8_t *fragmentB; 00517 x86_reg imm8OfPShufW1B; 00518 x86_reg imm8OfPShufW2B; 00519 x86_reg fragmentLengthB; 00520 int fragmentPos; 00521 00522 int xpos, i; 00523 00524 // create an optimized horizontal scaling routine 00525 /* This scaler is made of runtime-generated MMX2 code using specially 00526 * tuned pshufw instructions. For every four output pixels, if four 00527 * input pixels are enough for the fast bilinear scaling, then a chunk 00528 * of fragmentB is used. If five input pixels are needed, then a chunk 00529 * of fragmentA is used. 00530 */ 00531 00532 //code fragment 00533 00534 __asm__ volatile( 00535 "jmp 9f \n\t" 00536 // Begin 00537 "0: \n\t" 00538 "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t" 00539 "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t" 00540 "movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t" 00541 "punpcklbw %%mm7, %%mm1 \n\t" 00542 "punpcklbw %%mm7, %%mm0 \n\t" 00543 "pshufw $0xFF, %%mm1, %%mm1 \n\t" 00544 "1: \n\t" 00545 "pshufw $0xFF, %%mm0, %%mm0 \n\t" 00546 "2: \n\t" 00547 "psubw %%mm1, %%mm0 \n\t" 00548 "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t" 00549 "pmullw %%mm3, %%mm0 \n\t" 00550 "psllw $7, %%mm1 \n\t" 00551 "paddw %%mm1, %%mm0 \n\t" 00552 00553 "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t" 00554 00555 "add $8, %%"REG_a" \n\t" 00556 // End 00557 "9: \n\t" 00558 // "int $3 \n\t" 00559 "lea " LOCAL_MANGLE(0b) ", %0 \n\t" 00560 "lea " LOCAL_MANGLE(1b) ", %1 \n\t" 00561 "lea " LOCAL_MANGLE(2b) ", %2 \n\t" 00562 "dec %1 \n\t" 00563 "dec %2 \n\t" 00564 "sub %0, %1 \n\t" 00565 "sub %0, %2 \n\t" 00566 "lea " LOCAL_MANGLE(9b) ", %3 \n\t" 00567 "sub %0, %3 \n\t" 00568 00569 00570 :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A), 00571 "=r" (fragmentLengthA) 00572 ); 00573 00574 __asm__ volatile( 00575 "jmp 9f \n\t" 00576 // Begin 00577 "0: \n\t" 00578 "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t" 00579 "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t" 00580 "punpcklbw %%mm7, %%mm0 \n\t" 00581 "pshufw $0xFF, %%mm0, %%mm1 \n\t" 00582 "1: \n\t" 00583 "pshufw $0xFF, %%mm0, %%mm0 \n\t" 00584 "2: \n\t" 00585 "psubw %%mm1, %%mm0 \n\t" 00586 "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t" 00587 "pmullw %%mm3, %%mm0 \n\t" 00588 "psllw $7, %%mm1 \n\t" 00589 "paddw %%mm1, %%mm0 \n\t" 00590 00591 "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t" 00592 00593 "add $8, %%"REG_a" \n\t" 00594 // End 00595 "9: \n\t" 00596 // "int $3 \n\t" 00597 "lea " LOCAL_MANGLE(0b) ", %0 \n\t" 00598 "lea " LOCAL_MANGLE(1b) ", %1 \n\t" 00599 "lea " LOCAL_MANGLE(2b) ", %2 \n\t" 00600 "dec %1 \n\t" 00601 "dec %2 \n\t" 00602 "sub %0, %1 \n\t" 00603 "sub %0, %2 \n\t" 00604 "lea " LOCAL_MANGLE(9b) ", %3 \n\t" 00605 "sub %0, %3 \n\t" 00606 00607 00608 :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B), 00609 "=r" (fragmentLengthB) 00610 ); 00611 00612 xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers 00613 fragmentPos=0; 00614 00615 for (i=0; i<dstW/numSplits; i++) { 00616 int xx=xpos>>16; 00617 00618 if ((i&3) == 0) { 00619 int a=0; 00620 int b=((xpos+xInc)>>16) - xx; 00621 int c=((xpos+xInc*2)>>16) - xx; 00622 int d=((xpos+xInc*3)>>16) - xx; 00623 int inc = (d+1<4); 00624 uint8_t *fragment = (d+1<4) ? fragmentB : fragmentA; 00625 x86_reg imm8OfPShufW1 = (d+1<4) ? imm8OfPShufW1B : imm8OfPShufW1A; 00626 x86_reg imm8OfPShufW2 = (d+1<4) ? imm8OfPShufW2B : imm8OfPShufW2A; 00627 x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA; 00628 int maxShift= 3-(d+inc); 00629 int shift=0; 00630 00631 if (filterCode) { 00632 filter[i ] = (( xpos & 0xFFFF) ^ 0xFFFF)>>9; 00633 filter[i+1] = (((xpos+xInc ) & 0xFFFF) ^ 0xFFFF)>>9; 00634 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9; 00635 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9; 00636 filterPos[i/2]= xx; 00637 00638 memcpy(filterCode + fragmentPos, fragment, fragmentLength); 00639 00640 filterCode[fragmentPos + imm8OfPShufW1]= 00641 (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6); 00642 filterCode[fragmentPos + imm8OfPShufW2]= 00643 a | (b<<2) | (c<<4) | (d<<6); 00644 00645 if (i+4-inc>=dstW) shift=maxShift; //avoid overread 00646 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align 00647 00648 if (shift && i>=shift) { 00649 filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift; 00650 filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift; 00651 filterPos[i/2]-=shift; 00652 } 00653 } 00654 00655 fragmentPos+= fragmentLength; 00656 00657 if (filterCode) 00658 filterCode[fragmentPos]= RET; 00659 } 00660 xpos+=xInc; 00661 } 00662 if (filterCode) 00663 filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part 00664 00665 return fragmentPos + 1; 00666 } 00667 #endif /* ARCH_X86 && (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) */ 00668 00669 static void getSubSampleFactors(int *h, int *v, enum PixelFormat format) 00670 { 00671 *h = av_pix_fmt_descriptors[format].log2_chroma_w; 00672 *v = av_pix_fmt_descriptors[format].log2_chroma_h; 00673 } 00674 00675 static uint16_t roundToInt16(int64_t f) 00676 { 00677 int r= (f + (1<<15))>>16; 00678 if (r<-0x7FFF) return 0x8000; 00679 else if (r> 0x7FFF) return 0x7FFF; 00680 else return r; 00681 } 00682 00683 int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation) 00684 { 00685 int64_t crv = inv_table[0]; 00686 int64_t cbu = inv_table[1]; 00687 int64_t cgu = -inv_table[2]; 00688 int64_t cgv = -inv_table[3]; 00689 int64_t cy = 1<<16; 00690 int64_t oy = 0; 00691 00692 memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4); 00693 memcpy(c->dstColorspaceTable, table, sizeof(int)*4); 00694 00695 c->brightness= brightness; 00696 c->contrast = contrast; 00697 c->saturation= saturation; 00698 c->srcRange = srcRange; 00699 c->dstRange = dstRange; 00700 if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1; 00701 00702 c->uOffset= 0x0400040004000400LL; 00703 c->vOffset= 0x0400040004000400LL; 00704 00705 if (!srcRange) { 00706 cy= (cy*255) / 219; 00707 oy= 16<<16; 00708 } else { 00709 crv= (crv*224) / 255; 00710 cbu= (cbu*224) / 255; 00711 cgu= (cgu*224) / 255; 00712 cgv= (cgv*224) / 255; 00713 } 00714 00715 cy = (cy *contrast )>>16; 00716 crv= (crv*contrast * saturation)>>32; 00717 cbu= (cbu*contrast * saturation)>>32; 00718 cgu= (cgu*contrast * saturation)>>32; 00719 cgv= (cgv*contrast * saturation)>>32; 00720 00721 oy -= 256*brightness; 00722 00723 c->yCoeff= roundToInt16(cy *8192) * 0x0001000100010001ULL; 00724 c->vrCoeff= roundToInt16(crv*8192) * 0x0001000100010001ULL; 00725 c->ubCoeff= roundToInt16(cbu*8192) * 0x0001000100010001ULL; 00726 c->vgCoeff= roundToInt16(cgv*8192) * 0x0001000100010001ULL; 00727 c->ugCoeff= roundToInt16(cgu*8192) * 0x0001000100010001ULL; 00728 c->yOffset= roundToInt16(oy * 8) * 0x0001000100010001ULL; 00729 00730 c->yuv2rgb_y_coeff = (int16_t)roundToInt16(cy <<13); 00731 c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9); 00732 c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13); 00733 c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13); 00734 c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13); 00735 c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13); 00736 00737 ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation); 00738 //FIXME factorize 00739 00740 #if HAVE_ALTIVEC 00741 if (c->flags & SWS_CPU_CAPS_ALTIVEC) 00742 ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation); 00743 #endif 00744 return 0; 00745 } 00746 00747 int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation) 00748 { 00749 if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1; 00750 00751 *inv_table = c->srcColorspaceTable; 00752 *table = c->dstColorspaceTable; 00753 *srcRange = c->srcRange; 00754 *dstRange = c->dstRange; 00755 *brightness= c->brightness; 00756 *contrast = c->contrast; 00757 *saturation= c->saturation; 00758 00759 return 0; 00760 } 00761 00762 static int handle_jpeg(enum PixelFormat *format) 00763 { 00764 switch (*format) { 00765 case PIX_FMT_YUVJ420P: 00766 *format = PIX_FMT_YUV420P; 00767 return 1; 00768 case PIX_FMT_YUVJ422P: 00769 *format = PIX_FMT_YUV422P; 00770 return 1; 00771 case PIX_FMT_YUVJ444P: 00772 *format = PIX_FMT_YUV444P; 00773 return 1; 00774 case PIX_FMT_YUVJ440P: 00775 *format = PIX_FMT_YUV440P; 00776 return 1; 00777 default: 00778 return 0; 00779 } 00780 } 00781 00782 SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, 00783 int dstW, int dstH, enum PixelFormat dstFormat, int flags, 00784 SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param) 00785 { 00786 SwsContext *c; 00787 int i; 00788 int usesVFilter, usesHFilter; 00789 int unscaled; 00790 int srcRange, dstRange; 00791 SwsFilter dummyFilter= {NULL, NULL, NULL, NULL}; 00792 #if ARCH_X86 00793 if (flags & SWS_CPU_CAPS_MMX) 00794 __asm__ volatile("emms\n\t"::: "memory"); 00795 #endif 00796 00797 #if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off 00798 flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN); 00799 flags |= ff_hardcodedcpuflags(); 00800 #endif /* CONFIG_RUNTIME_CPUDETECT */ 00801 if (!rgb15to16) sws_rgb2rgb_init(flags); 00802 00803 unscaled = (srcW == dstW && srcH == dstH); 00804 00805 srcRange = handle_jpeg(&srcFormat); 00806 dstRange = handle_jpeg(&dstFormat); 00807 00808 if (!isSupportedIn(srcFormat)) { 00809 av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat)); 00810 return NULL; 00811 } 00812 if (!isSupportedOut(dstFormat)) { 00813 av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat)); 00814 return NULL; 00815 } 00816 00817 i= flags & ( SWS_POINT 00818 |SWS_AREA 00819 |SWS_BILINEAR 00820 |SWS_FAST_BILINEAR 00821 |SWS_BICUBIC 00822 |SWS_X 00823 |SWS_GAUSS 00824 |SWS_LANCZOS 00825 |SWS_SINC 00826 |SWS_SPLINE 00827 |SWS_BICUBLIN); 00828 if(!i || (i & (i-1))) { 00829 av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n"); 00830 return NULL; 00831 } 00832 00833 /* sanity check */ 00834 if (srcW<4 || srcH<1 || dstW<8 || dstH<1) { //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code 00835 av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n", 00836 srcW, srcH, dstW, dstH); 00837 return NULL; 00838 } 00839 if(srcW > VOFW || dstW > VOFW) { 00840 av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n"); 00841 return NULL; 00842 } 00843 00844 if (!dstFilter) dstFilter= &dummyFilter; 00845 if (!srcFilter) srcFilter= &dummyFilter; 00846 00847 FF_ALLOCZ_OR_GOTO(NULL, c, sizeof(SwsContext), fail); 00848 00849 c->av_class = &sws_context_class; 00850 c->srcW= srcW; 00851 c->srcH= srcH; 00852 c->dstW= dstW; 00853 c->dstH= dstH; 00854 c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW; 00855 c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH; 00856 c->flags= flags; 00857 c->dstFormat= dstFormat; 00858 c->srcFormat= srcFormat; 00859 c->dstFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[dstFormat]); 00860 c->srcFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[srcFormat]); 00861 c->vRounder= 4* 0x0001000100010001ULL; 00862 00863 usesVFilter = (srcFilter->lumV && srcFilter->lumV->length>1) || 00864 (srcFilter->chrV && srcFilter->chrV->length>1) || 00865 (dstFilter->lumV && dstFilter->lumV->length>1) || 00866 (dstFilter->chrV && dstFilter->chrV->length>1); 00867 usesHFilter = (srcFilter->lumH && srcFilter->lumH->length>1) || 00868 (srcFilter->chrH && srcFilter->chrH->length>1) || 00869 (dstFilter->lumH && dstFilter->lumH->length>1) || 00870 (dstFilter->chrH && dstFilter->chrH->length>1); 00871 00872 getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat); 00873 getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat); 00874 00875 // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation 00876 if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1; 00877 00878 // drop some chroma lines if the user wants it 00879 c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT; 00880 c->chrSrcVSubSample+= c->vChrDrop; 00881 00882 // drop every other pixel for chroma calculation unless user wants full chroma 00883 if (isAnyRGB(srcFormat) && !(flags&SWS_FULL_CHR_H_INP) 00884 && srcFormat!=PIX_FMT_RGB8 && srcFormat!=PIX_FMT_BGR8 00885 && srcFormat!=PIX_FMT_RGB4 && srcFormat!=PIX_FMT_BGR4 00886 && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE 00887 && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT)))) 00888 c->chrSrcHSubSample=1; 00889 00890 if (param) { 00891 c->param[0] = param[0]; 00892 c->param[1] = param[1]; 00893 } else { 00894 c->param[0] = 00895 c->param[1] = SWS_PARAM_DEFAULT; 00896 } 00897 00898 // Note the -((-x)>>y) is so that we always round toward +inf. 00899 c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample); 00900 c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample); 00901 c->chrDstW= -((-dstW) >> c->chrDstHSubSample); 00902 c->chrDstH= -((-dstH) >> c->chrDstVSubSample); 00903 00904 sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16); 00905 00906 /* unscaled special cases */ 00907 if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isAnyRGB(dstFormat))) { 00908 ff_get_unscaled_swscale(c); 00909 00910 if (c->swScale) { 00911 if (flags&SWS_PRINT_INFO) 00912 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n", 00913 sws_format_name(srcFormat), sws_format_name(dstFormat)); 00914 return c; 00915 } 00916 } 00917 00918 if (flags & SWS_CPU_CAPS_MMX2) { 00919 c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0; 00920 if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) { 00921 if (flags&SWS_PRINT_INFO) 00922 av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n"); 00923 } 00924 if (usesHFilter) c->canMMX2BeUsed=0; 00925 } 00926 else 00927 c->canMMX2BeUsed=0; 00928 00929 c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW; 00930 c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH; 00931 00932 // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst 00933 // but only for the FAST_BILINEAR mode otherwise do correct scaling 00934 // n-2 is the last chrominance sample available 00935 // this is not perfect, but no one should notice the difference, the more correct variant 00936 // would be like the vertical one, but that would require some special code for the 00937 // first and last pixel 00938 if (flags&SWS_FAST_BILINEAR) { 00939 if (c->canMMX2BeUsed) { 00940 c->lumXInc+= 20; 00941 c->chrXInc+= 20; 00942 } 00943 //we don't use the x86 asm scaler if MMX is available 00944 else if (flags & SWS_CPU_CAPS_MMX) { 00945 c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20; 00946 c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20; 00947 } 00948 } 00949 00950 /* precalculate horizontal scaler filter coefficients */ 00951 { 00952 #if ARCH_X86 && (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) 00953 // can't downscale !!! 00954 if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) { 00955 c->lumMmx2FilterCodeSize = initMMX2HScaler( dstW, c->lumXInc, NULL, NULL, NULL, 8); 00956 c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4); 00957 00958 #ifdef MAP_ANONYMOUS 00959 c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0); 00960 c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0); 00961 #elif HAVE_VIRTUALALLOC 00962 c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE); 00963 c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE); 00964 #else 00965 c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize); 00966 c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize); 00967 #endif 00968 00969 if (!c->lumMmx2FilterCode || !c->chrMmx2FilterCode) 00970 goto fail; 00971 FF_ALLOCZ_OR_GOTO(c, c->hLumFilter , (dstW /8+8)*sizeof(int16_t), fail); 00972 FF_ALLOCZ_OR_GOTO(c, c->hChrFilter , (c->chrDstW /4+8)*sizeof(int16_t), fail); 00973 FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW /2/8+8)*sizeof(int32_t), fail); 00974 FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW/2/4+8)*sizeof(int32_t), fail); 00975 00976 initMMX2HScaler( dstW, c->lumXInc, c->lumMmx2FilterCode, c->hLumFilter, c->hLumFilterPos, 8); 00977 initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->hChrFilter, c->hChrFilterPos, 4); 00978 00979 #ifdef MAP_ANONYMOUS 00980 mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ); 00981 mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ); 00982 #endif 00983 } else 00984 #endif /* ARCH_X86 && (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) */ 00985 { 00986 const int filterAlign= 00987 (flags & SWS_CPU_CAPS_MMX) ? 4 : 00988 (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 : 00989 1; 00990 00991 if (initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc, 00992 srcW , dstW, filterAlign, 1<<14, 00993 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags, 00994 srcFilter->lumH, dstFilter->lumH, c->param) < 0) 00995 goto fail; 00996 if (initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc, 00997 c->chrSrcW, c->chrDstW, filterAlign, 1<<14, 00998 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags, 00999 srcFilter->chrH, dstFilter->chrH, c->param) < 0) 01000 goto fail; 01001 } 01002 } // initialize horizontal stuff 01003 01004 /* precalculate vertical scaler filter coefficients */ 01005 { 01006 const int filterAlign= 01007 (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 : 01008 (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 : 01009 1; 01010 01011 if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc, 01012 srcH , dstH, filterAlign, (1<<12), 01013 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags, 01014 srcFilter->lumV, dstFilter->lumV, c->param) < 0) 01015 goto fail; 01016 if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc, 01017 c->chrSrcH, c->chrDstH, filterAlign, (1<<12), 01018 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags, 01019 srcFilter->chrV, dstFilter->chrV, c->param) < 0) 01020 goto fail; 01021 01022 #if HAVE_ALTIVEC 01023 FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof (vector signed short)*c->vLumFilterSize*c->dstH, fail); 01024 FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH, fail); 01025 01026 for (i=0;i<c->vLumFilterSize*c->dstH;i++) { 01027 int j; 01028 short *p = (short *)&c->vYCoeffsBank[i]; 01029 for (j=0;j<8;j++) 01030 p[j] = c->vLumFilter[i]; 01031 } 01032 01033 for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) { 01034 int j; 01035 short *p = (short *)&c->vCCoeffsBank[i]; 01036 for (j=0;j<8;j++) 01037 p[j] = c->vChrFilter[i]; 01038 } 01039 #endif 01040 } 01041 01042 // calculate buffer sizes so that they won't run out while handling these damn slices 01043 c->vLumBufSize= c->vLumFilterSize; 01044 c->vChrBufSize= c->vChrFilterSize; 01045 for (i=0; i<dstH; i++) { 01046 int chrI= i*c->chrDstH / dstH; 01047 int nextSlice= FFMAX(c->vLumFilterPos[i ] + c->vLumFilterSize - 1, 01048 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample)); 01049 01050 nextSlice>>= c->chrSrcVSubSample; 01051 nextSlice<<= c->chrSrcVSubSample; 01052 if (c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice) 01053 c->vLumBufSize= nextSlice - c->vLumFilterPos[i]; 01054 if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample)) 01055 c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI]; 01056 } 01057 01058 // allocate pixbufs (we use dynamic allocation because otherwise we would need to 01059 // allocate several megabytes to handle all possible cases) 01060 FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail); 01061 FF_ALLOC_OR_GOTO(c, c->chrPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail); 01062 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) 01063 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail); 01064 //Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna replace the 4000/8000) 01065 /* align at 16 bytes for AltiVec */ 01066 for (i=0; i<c->vLumBufSize; i++) { 01067 FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i+c->vLumBufSize], VOF+1, fail); 01068 c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize]; 01069 } 01070 for (i=0; i<c->vChrBufSize; i++) { 01071 FF_ALLOC_OR_GOTO(c, c->chrPixBuf[i+c->vChrBufSize], (VOF+1)*2, fail); 01072 c->chrPixBuf[i] = c->chrPixBuf[i+c->vChrBufSize]; 01073 } 01074 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) 01075 for (i=0; i<c->vLumBufSize; i++) { 01076 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i+c->vLumBufSize], VOF+1, fail); 01077 c->alpPixBuf[i] = c->alpPixBuf[i+c->vLumBufSize]; 01078 } 01079 01080 //try to avoid drawing green stuff between the right end and the stride end 01081 for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2); 01082 01083 assert(2*VOFW == VOF); 01084 01085 assert(c->chrDstH <= dstH); 01086 01087 if (flags&SWS_PRINT_INFO) { 01088 if (flags&SWS_FAST_BILINEAR) 01089 av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, "); 01090 else if (flags&SWS_BILINEAR) 01091 av_log(c, AV_LOG_INFO, "BILINEAR scaler, "); 01092 else if (flags&SWS_BICUBIC) 01093 av_log(c, AV_LOG_INFO, "BICUBIC scaler, "); 01094 else if (flags&SWS_X) 01095 av_log(c, AV_LOG_INFO, "Experimental scaler, "); 01096 else if (flags&SWS_POINT) 01097 av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, "); 01098 else if (flags&SWS_AREA) 01099 av_log(c, AV_LOG_INFO, "Area Averaging scaler, "); 01100 else if (flags&SWS_BICUBLIN) 01101 av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, "); 01102 else if (flags&SWS_GAUSS) 01103 av_log(c, AV_LOG_INFO, "Gaussian scaler, "); 01104 else if (flags&SWS_SINC) 01105 av_log(c, AV_LOG_INFO, "Sinc scaler, "); 01106 else if (flags&SWS_LANCZOS) 01107 av_log(c, AV_LOG_INFO, "Lanczos scaler, "); 01108 else if (flags&SWS_SPLINE) 01109 av_log(c, AV_LOG_INFO, "Bicubic spline scaler, "); 01110 else 01111 av_log(c, AV_LOG_INFO, "ehh flags invalid?! "); 01112 01113 av_log(c, AV_LOG_INFO, "from %s to %s%s ", 01114 sws_format_name(srcFormat), 01115 #ifdef DITHER1XBPP 01116 dstFormat == PIX_FMT_BGR555 || dstFormat == PIX_FMT_BGR565 || 01117 dstFormat == PIX_FMT_RGB444BE || dstFormat == PIX_FMT_RGB444LE || 01118 dstFormat == PIX_FMT_BGR444BE || dstFormat == PIX_FMT_BGR444LE ? "dithered " : "", 01119 #else 01120 "", 01121 #endif 01122 sws_format_name(dstFormat)); 01123 01124 if (flags & SWS_CPU_CAPS_MMX2) 01125 av_log(c, AV_LOG_INFO, "using MMX2\n"); 01126 else if (flags & SWS_CPU_CAPS_3DNOW) 01127 av_log(c, AV_LOG_INFO, "using 3DNOW\n"); 01128 else if (flags & SWS_CPU_CAPS_MMX) 01129 av_log(c, AV_LOG_INFO, "using MMX\n"); 01130 else if (flags & SWS_CPU_CAPS_ALTIVEC) 01131 av_log(c, AV_LOG_INFO, "using AltiVec\n"); 01132 else 01133 av_log(c, AV_LOG_INFO, "using C\n"); 01134 01135 if (flags & SWS_CPU_CAPS_MMX) { 01136 if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR)) 01137 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n"); 01138 else { 01139 if (c->hLumFilterSize==4) 01140 av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n"); 01141 else if (c->hLumFilterSize==8) 01142 av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n"); 01143 else 01144 av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n"); 01145 01146 if (c->hChrFilterSize==4) 01147 av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n"); 01148 else if (c->hChrFilterSize==8) 01149 av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n"); 01150 else 01151 av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n"); 01152 } 01153 } else { 01154 #if ARCH_X86 01155 av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n"); 01156 #else 01157 if (flags & SWS_FAST_BILINEAR) 01158 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n"); 01159 else 01160 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n"); 01161 #endif 01162 } 01163 if (isPlanarYUV(dstFormat)) { 01164 if (c->vLumFilterSize==1) 01165 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01166 else 01167 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01168 } else { 01169 if (c->vLumFilterSize==1 && c->vChrFilterSize==2) 01170 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n" 01171 " 2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01172 else if (c->vLumFilterSize==2 && c->vChrFilterSize==2) 01173 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01174 else 01175 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01176 } 01177 01178 if (dstFormat==PIX_FMT_BGR24) 01179 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n", 01180 (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C")); 01181 else if (dstFormat==PIX_FMT_RGB32) 01182 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01183 else if (dstFormat==PIX_FMT_BGR565) 01184 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01185 else if (dstFormat==PIX_FMT_BGR555) 01186 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01187 else if (dstFormat == PIX_FMT_RGB444BE || dstFormat == PIX_FMT_RGB444LE || 01188 dstFormat == PIX_FMT_BGR444BE || dstFormat == PIX_FMT_BGR444LE) 01189 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR12 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"); 01190 01191 av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH); 01192 av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", 01193 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc); 01194 av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", 01195 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc); 01196 } 01197 01198 c->swScale= ff_getSwsFunc(c); 01199 return c; 01200 01201 fail: 01202 sws_freeContext(c); 01203 return NULL; 01204 } 01205 01206 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur, 01207 float lumaSharpen, float chromaSharpen, 01208 float chromaHShift, float chromaVShift, 01209 int verbose) 01210 { 01211 SwsFilter *filter= av_malloc(sizeof(SwsFilter)); 01212 if (!filter) 01213 return NULL; 01214 01215 if (lumaGBlur!=0.0) { 01216 filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0); 01217 filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0); 01218 } else { 01219 filter->lumH= sws_getIdentityVec(); 01220 filter->lumV= sws_getIdentityVec(); 01221 } 01222 01223 if (chromaGBlur!=0.0) { 01224 filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0); 01225 filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0); 01226 } else { 01227 filter->chrH= sws_getIdentityVec(); 01228 filter->chrV= sws_getIdentityVec(); 01229 } 01230 01231 if (chromaSharpen!=0.0) { 01232 SwsVector *id= sws_getIdentityVec(); 01233 sws_scaleVec(filter->chrH, -chromaSharpen); 01234 sws_scaleVec(filter->chrV, -chromaSharpen); 01235 sws_addVec(filter->chrH, id); 01236 sws_addVec(filter->chrV, id); 01237 sws_freeVec(id); 01238 } 01239 01240 if (lumaSharpen!=0.0) { 01241 SwsVector *id= sws_getIdentityVec(); 01242 sws_scaleVec(filter->lumH, -lumaSharpen); 01243 sws_scaleVec(filter->lumV, -lumaSharpen); 01244 sws_addVec(filter->lumH, id); 01245 sws_addVec(filter->lumV, id); 01246 sws_freeVec(id); 01247 } 01248 01249 if (chromaHShift != 0.0) 01250 sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5)); 01251 01252 if (chromaVShift != 0.0) 01253 sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5)); 01254 01255 sws_normalizeVec(filter->chrH, 1.0); 01256 sws_normalizeVec(filter->chrV, 1.0); 01257 sws_normalizeVec(filter->lumH, 1.0); 01258 sws_normalizeVec(filter->lumV, 1.0); 01259 01260 if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG); 01261 if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG); 01262 01263 return filter; 01264 } 01265 01266 SwsVector *sws_allocVec(int length) 01267 { 01268 SwsVector *vec = av_malloc(sizeof(SwsVector)); 01269 if (!vec) 01270 return NULL; 01271 vec->length = length; 01272 vec->coeff = av_malloc(sizeof(double) * length); 01273 if (!vec->coeff) 01274 av_freep(&vec); 01275 return vec; 01276 } 01277 01278 SwsVector *sws_getGaussianVec(double variance, double quality) 01279 { 01280 const int length= (int)(variance*quality + 0.5) | 1; 01281 int i; 01282 double middle= (length-1)*0.5; 01283 SwsVector *vec= sws_allocVec(length); 01284 01285 if (!vec) 01286 return NULL; 01287 01288 for (i=0; i<length; i++) { 01289 double dist= i-middle; 01290 vec->coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*M_PI); 01291 } 01292 01293 sws_normalizeVec(vec, 1.0); 01294 01295 return vec; 01296 } 01297 01298 SwsVector *sws_getConstVec(double c, int length) 01299 { 01300 int i; 01301 SwsVector *vec= sws_allocVec(length); 01302 01303 if (!vec) 01304 return NULL; 01305 01306 for (i=0; i<length; i++) 01307 vec->coeff[i]= c; 01308 01309 return vec; 01310 } 01311 01312 SwsVector *sws_getIdentityVec(void) 01313 { 01314 return sws_getConstVec(1.0, 1); 01315 } 01316 01317 static double sws_dcVec(SwsVector *a) 01318 { 01319 int i; 01320 double sum=0; 01321 01322 for (i=0; i<a->length; i++) 01323 sum+= a->coeff[i]; 01324 01325 return sum; 01326 } 01327 01328 void sws_scaleVec(SwsVector *a, double scalar) 01329 { 01330 int i; 01331 01332 for (i=0; i<a->length; i++) 01333 a->coeff[i]*= scalar; 01334 } 01335 01336 void sws_normalizeVec(SwsVector *a, double height) 01337 { 01338 sws_scaleVec(a, height/sws_dcVec(a)); 01339 } 01340 01341 static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b) 01342 { 01343 int length= a->length + b->length - 1; 01344 int i, j; 01345 SwsVector *vec= sws_getConstVec(0.0, length); 01346 01347 if (!vec) 01348 return NULL; 01349 01350 for (i=0; i<a->length; i++) { 01351 for (j=0; j<b->length; j++) { 01352 vec->coeff[i+j]+= a->coeff[i]*b->coeff[j]; 01353 } 01354 } 01355 01356 return vec; 01357 } 01358 01359 static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b) 01360 { 01361 int length= FFMAX(a->length, b->length); 01362 int i; 01363 SwsVector *vec= sws_getConstVec(0.0, length); 01364 01365 if (!vec) 01366 return NULL; 01367 01368 for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i]; 01369 for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i]; 01370 01371 return vec; 01372 } 01373 01374 static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b) 01375 { 01376 int length= FFMAX(a->length, b->length); 01377 int i; 01378 SwsVector *vec= sws_getConstVec(0.0, length); 01379 01380 if (!vec) 01381 return NULL; 01382 01383 for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i]; 01384 for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i]; 01385 01386 return vec; 01387 } 01388 01389 /* shift left / or right if "shift" is negative */ 01390 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift) 01391 { 01392 int length= a->length + FFABS(shift)*2; 01393 int i; 01394 SwsVector *vec= sws_getConstVec(0.0, length); 01395 01396 if (!vec) 01397 return NULL; 01398 01399 for (i=0; i<a->length; i++) { 01400 vec->coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i]; 01401 } 01402 01403 return vec; 01404 } 01405 01406 void sws_shiftVec(SwsVector *a, int shift) 01407 { 01408 SwsVector *shifted= sws_getShiftedVec(a, shift); 01409 av_free(a->coeff); 01410 a->coeff= shifted->coeff; 01411 a->length= shifted->length; 01412 av_free(shifted); 01413 } 01414 01415 void sws_addVec(SwsVector *a, SwsVector *b) 01416 { 01417 SwsVector *sum= sws_sumVec(a, b); 01418 av_free(a->coeff); 01419 a->coeff= sum->coeff; 01420 a->length= sum->length; 01421 av_free(sum); 01422 } 01423 01424 void sws_subVec(SwsVector *a, SwsVector *b) 01425 { 01426 SwsVector *diff= sws_diffVec(a, b); 01427 av_free(a->coeff); 01428 a->coeff= diff->coeff; 01429 a->length= diff->length; 01430 av_free(diff); 01431 } 01432 01433 void sws_convVec(SwsVector *a, SwsVector *b) 01434 { 01435 SwsVector *conv= sws_getConvVec(a, b); 01436 av_free(a->coeff); 01437 a->coeff= conv->coeff; 01438 a->length= conv->length; 01439 av_free(conv); 01440 } 01441 01442 SwsVector *sws_cloneVec(SwsVector *a) 01443 { 01444 int i; 01445 SwsVector *vec= sws_allocVec(a->length); 01446 01447 if (!vec) 01448 return NULL; 01449 01450 for (i=0; i<a->length; i++) vec->coeff[i]= a->coeff[i]; 01451 01452 return vec; 01453 } 01454 01455 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level) 01456 { 01457 int i; 01458 double max=0; 01459 double min=0; 01460 double range; 01461 01462 for (i=0; i<a->length; i++) 01463 if (a->coeff[i]>max) max= a->coeff[i]; 01464 01465 for (i=0; i<a->length; i++) 01466 if (a->coeff[i]<min) min= a->coeff[i]; 01467 01468 range= max - min; 01469 01470 for (i=0; i<a->length; i++) { 01471 int x= (int)((a->coeff[i]-min)*60.0/range +0.5); 01472 av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]); 01473 for (;x>0; x--) av_log(log_ctx, log_level, " "); 01474 av_log(log_ctx, log_level, "|\n"); 01475 } 01476 } 01477 01478 #if LIBSWSCALE_VERSION_MAJOR < 1 01479 void sws_printVec(SwsVector *a) 01480 { 01481 sws_printVec2(a, NULL, AV_LOG_DEBUG); 01482 } 01483 #endif 01484 01485 void sws_freeVec(SwsVector *a) 01486 { 01487 if (!a) return; 01488 av_freep(&a->coeff); 01489 a->length=0; 01490 av_free(a); 01491 } 01492 01493 void sws_freeFilter(SwsFilter *filter) 01494 { 01495 if (!filter) return; 01496 01497 if (filter->lumH) sws_freeVec(filter->lumH); 01498 if (filter->lumV) sws_freeVec(filter->lumV); 01499 if (filter->chrH) sws_freeVec(filter->chrH); 01500 if (filter->chrV) sws_freeVec(filter->chrV); 01501 av_free(filter); 01502 } 01503 01504 void sws_freeContext(SwsContext *c) 01505 { 01506 int i; 01507 if (!c) return; 01508 01509 if (c->lumPixBuf) { 01510 for (i=0; i<c->vLumBufSize; i++) 01511 av_freep(&c->lumPixBuf[i]); 01512 av_freep(&c->lumPixBuf); 01513 } 01514 01515 if (c->chrPixBuf) { 01516 for (i=0; i<c->vChrBufSize; i++) 01517 av_freep(&c->chrPixBuf[i]); 01518 av_freep(&c->chrPixBuf); 01519 } 01520 01521 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) { 01522 for (i=0; i<c->vLumBufSize; i++) 01523 av_freep(&c->alpPixBuf[i]); 01524 av_freep(&c->alpPixBuf); 01525 } 01526 01527 av_freep(&c->vLumFilter); 01528 av_freep(&c->vChrFilter); 01529 av_freep(&c->hLumFilter); 01530 av_freep(&c->hChrFilter); 01531 #if HAVE_ALTIVEC 01532 av_freep(&c->vYCoeffsBank); 01533 av_freep(&c->vCCoeffsBank); 01534 #endif 01535 01536 av_freep(&c->vLumFilterPos); 01537 av_freep(&c->vChrFilterPos); 01538 av_freep(&c->hLumFilterPos); 01539 av_freep(&c->hChrFilterPos); 01540 01541 #if ARCH_X86 01542 #ifdef MAP_ANONYMOUS 01543 if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize); 01544 if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize); 01545 #elif HAVE_VIRTUALALLOC 01546 if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, 0, MEM_RELEASE); 01547 if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, 0, MEM_RELEASE); 01548 #else 01549 av_free(c->lumMmx2FilterCode); 01550 av_free(c->chrMmx2FilterCode); 01551 #endif 01552 c->lumMmx2FilterCode=NULL; 01553 c->chrMmx2FilterCode=NULL; 01554 #endif /* ARCH_X86 */ 01555 01556 av_freep(&c->yuvTable); 01557 01558 av_free(c); 01559 } 01560 01561 struct SwsContext *sws_getCachedContext(struct SwsContext *context, 01562 int srcW, int srcH, enum PixelFormat srcFormat, 01563 int dstW, int dstH, enum PixelFormat dstFormat, int flags, 01564 SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param) 01565 { 01566 static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT}; 01567 01568 if (!param) 01569 param = default_param; 01570 01571 if (context && 01572 (context->srcW != srcW || 01573 context->srcH != srcH || 01574 context->srcFormat != srcFormat || 01575 context->dstW != dstW || 01576 context->dstH != dstH || 01577 context->dstFormat != dstFormat || 01578 context->flags != flags || 01579 context->param[0] != param[0] || 01580 context->param[1] != param[1])) { 01581 sws_freeContext(context); 01582 context = NULL; 01583 } 01584 01585 if (!context) { 01586 return sws_getContext(srcW, srcH, srcFormat, 01587 dstW, dstH, dstFormat, flags, 01588 srcFilter, dstFilter, param); 01589 } 01590 return context; 01591 } 01592