00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
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/cpu.h"
00044 #include "libavutil/avutil.h"
00045 #include "libavutil/bswap.h"
00046 #include "libavutil/mathematics.h"
00047 #include "libavutil/opt.h"
00048 #include "libavutil/pixdesc.h"
00049
00050 unsigned swscale_version(void)
00051 {
00052 return LIBSWSCALE_VERSION_INT;
00053 }
00054
00055 const char *swscale_configuration(void)
00056 {
00057 return LIBAV_CONFIGURATION;
00058 }
00059
00060 const char *swscale_license(void)
00061 {
00062 #define LICENSE_PREFIX "libswscale license: "
00063 return LICENSE_PREFIX LIBAV_LICENSE + sizeof(LICENSE_PREFIX) - 1;
00064 }
00065
00066 #define RET 0xC3 //near return opcode for x86
00067
00068 typedef struct FormatEntry {
00069 int is_supported_in, is_supported_out;
00070 } FormatEntry;
00071
00072 static const FormatEntry format_entries[PIX_FMT_NB] = {
00073 [PIX_FMT_YUV420P] = { 1 , 1 },
00074 [PIX_FMT_YUYV422] = { 1 , 1 },
00075 [PIX_FMT_RGB24] = { 1 , 1 },
00076 [PIX_FMT_BGR24] = { 1 , 1 },
00077 [PIX_FMT_YUV422P] = { 1 , 1 },
00078 [PIX_FMT_YUV444P] = { 1 , 1 },
00079 [PIX_FMT_YUV410P] = { 1 , 1 },
00080 [PIX_FMT_YUV411P] = { 1 , 1 },
00081 [PIX_FMT_GRAY8] = { 1 , 1 },
00082 [PIX_FMT_MONOWHITE] = { 1 , 1 },
00083 [PIX_FMT_MONOBLACK] = { 1 , 1 },
00084 [PIX_FMT_PAL8] = { 1 , 0 },
00085 [PIX_FMT_YUVJ420P] = { 1 , 1 },
00086 [PIX_FMT_YUVJ422P] = { 1 , 1 },
00087 [PIX_FMT_YUVJ444P] = { 1 , 1 },
00088 [PIX_FMT_UYVY422] = { 1 , 1 },
00089 [PIX_FMT_UYYVYY411] = { 0 , 0 },
00090 [PIX_FMT_BGR8] = { 1 , 1 },
00091 [PIX_FMT_BGR4] = { 0 , 1 },
00092 [PIX_FMT_BGR4_BYTE] = { 1 , 1 },
00093 [PIX_FMT_RGB8] = { 1 , 1 },
00094 [PIX_FMT_RGB4] = { 0 , 1 },
00095 [PIX_FMT_RGB4_BYTE] = { 1 , 1 },
00096 [PIX_FMT_NV12] = { 1 , 1 },
00097 [PIX_FMT_NV21] = { 1 , 1 },
00098 [PIX_FMT_ARGB] = { 1 , 1 },
00099 [PIX_FMT_RGBA] = { 1 , 1 },
00100 [PIX_FMT_ABGR] = { 1 , 1 },
00101 [PIX_FMT_BGRA] = { 1 , 1 },
00102 [PIX_FMT_GRAY16BE] = { 1 , 1 },
00103 [PIX_FMT_GRAY16LE] = { 1 , 1 },
00104 [PIX_FMT_YUV440P] = { 1 , 1 },
00105 [PIX_FMT_YUVJ440P] = { 1 , 1 },
00106 [PIX_FMT_YUVA420P] = { 1 , 1 },
00107 [PIX_FMT_RGB48BE] = { 1 , 1 },
00108 [PIX_FMT_RGB48LE] = { 1 , 1 },
00109 [PIX_FMT_RGB565BE] = { 1 , 1 },
00110 [PIX_FMT_RGB565LE] = { 1 , 1 },
00111 [PIX_FMT_RGB555BE] = { 1 , 1 },
00112 [PIX_FMT_RGB555LE] = { 1 , 1 },
00113 [PIX_FMT_BGR565BE] = { 1 , 1 },
00114 [PIX_FMT_BGR565LE] = { 1 , 1 },
00115 [PIX_FMT_BGR555BE] = { 1 , 1 },
00116 [PIX_FMT_BGR555LE] = { 1 , 1 },
00117 [PIX_FMT_YUV420P16LE] = { 1 , 1 },
00118 [PIX_FMT_YUV420P16BE] = { 1 , 1 },
00119 [PIX_FMT_YUV422P16LE] = { 1 , 1 },
00120 [PIX_FMT_YUV422P16BE] = { 1 , 1 },
00121 [PIX_FMT_YUV444P16LE] = { 1 , 1 },
00122 [PIX_FMT_YUV444P16BE] = { 1 , 1 },
00123 [PIX_FMT_RGB444LE] = { 1 , 1 },
00124 [PIX_FMT_RGB444BE] = { 1 , 1 },
00125 [PIX_FMT_BGR444LE] = { 1 , 1 },
00126 [PIX_FMT_BGR444BE] = { 1 , 1 },
00127 [PIX_FMT_Y400A] = { 1 , 0 },
00128 [PIX_FMT_BGR48BE] = { 1 , 1 },
00129 [PIX_FMT_BGR48LE] = { 1 , 1 },
00130 [PIX_FMT_YUV420P9BE] = { 1 , 1 },
00131 [PIX_FMT_YUV420P9LE] = { 1 , 1 },
00132 [PIX_FMT_YUV420P10BE] = { 1 , 1 },
00133 [PIX_FMT_YUV420P10LE] = { 1 , 1 },
00134 [PIX_FMT_YUV422P9BE] = { 1 , 1 },
00135 [PIX_FMT_YUV422P9LE] = { 1 , 1 },
00136 [PIX_FMT_YUV422P10BE] = { 1 , 1 },
00137 [PIX_FMT_YUV422P10LE] = { 1 , 1 },
00138 [PIX_FMT_YUV444P9BE] = { 1 , 1 },
00139 [PIX_FMT_YUV444P9LE] = { 1 , 1 },
00140 [PIX_FMT_YUV444P10BE] = { 1 , 1 },
00141 [PIX_FMT_YUV444P10LE] = { 1 , 1 },
00142 [PIX_FMT_GBRP] = { 1 , 0 },
00143 [PIX_FMT_GBRP9LE] = { 1 , 0 },
00144 [PIX_FMT_GBRP9BE] = { 1 , 0 },
00145 [PIX_FMT_GBRP10LE] = { 1 , 0 },
00146 [PIX_FMT_GBRP10BE] = { 1 , 0 },
00147 [PIX_FMT_GBRP16LE] = { 1 , 0 },
00148 [PIX_FMT_GBRP16BE] = { 1 , 0 },
00149 };
00150
00151 int sws_isSupportedInput(enum PixelFormat pix_fmt)
00152 {
00153 return (unsigned)pix_fmt < PIX_FMT_NB ?
00154 format_entries[pix_fmt].is_supported_in : 0;
00155 }
00156
00157 int sws_isSupportedOutput(enum PixelFormat pix_fmt)
00158 {
00159 return (unsigned)pix_fmt < PIX_FMT_NB ?
00160 format_entries[pix_fmt].is_supported_out : 0;
00161 }
00162
00163 extern const int32_t ff_yuv2rgb_coeffs[8][4];
00164
00165 const char *sws_format_name(enum PixelFormat format)
00166 {
00167 if ((unsigned)format < PIX_FMT_NB && av_pix_fmt_descriptors[format].name)
00168 return av_pix_fmt_descriptors[format].name;
00169 else
00170 return "Unknown format";
00171 }
00172
00173 static double getSplineCoeff(double a, double b, double c, double d, double dist)
00174 {
00175 if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
00176 else return getSplineCoeff( 0.0,
00177 b+ 2.0*c + 3.0*d,
00178 c + 3.0*d,
00179 -b- 3.0*c - 6.0*d,
00180 dist-1.0);
00181 }
00182
00183 static int initFilter(int16_t **outFilter, int32_t **filterPos, int *outFilterSize, int xInc,
00184 int srcW, int dstW, int filterAlign, int one, int flags, int cpu_flags,
00185 SwsVector *srcFilter, SwsVector *dstFilter, double param[2], int is_horizontal)
00186 {
00187 int i;
00188 int filterSize;
00189 int filter2Size;
00190 int minFilterSize;
00191 int64_t *filter=NULL;
00192 int64_t *filter2=NULL;
00193 const int64_t fone= 1LL<<54;
00194 int ret= -1;
00195
00196 emms_c();
00197
00198
00199 FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW+3)*sizeof(**filterPos), fail);
00200
00201 if (FFABS(xInc - 0x10000) <10) {
00202 int i;
00203 filterSize= 1;
00204 FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
00205
00206 for (i=0; i<dstW; i++) {
00207 filter[i*filterSize]= fone;
00208 (*filterPos)[i]=i;
00209 }
00210
00211 } else if (flags&SWS_POINT) {
00212 int i;
00213 int xDstInSrc;
00214 filterSize= 1;
00215 FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
00216
00217 xDstInSrc= xInc/2 - 0x8000;
00218 for (i=0; i<dstW; i++) {
00219 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
00220
00221 (*filterPos)[i]= xx;
00222 filter[i]= fone;
00223 xDstInSrc+= xInc;
00224 }
00225 } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) {
00226 int i;
00227 int xDstInSrc;
00228 filterSize= 2;
00229 FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
00230
00231 xDstInSrc= xInc/2 - 0x8000;
00232 for (i=0; i<dstW; i++) {
00233 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
00234 int j;
00235
00236 (*filterPos)[i]= xx;
00237
00238 for (j=0; j<filterSize; j++) {
00239 int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
00240 if (coeff<0) coeff=0;
00241 filter[i*filterSize + j]= coeff;
00242 xx++;
00243 }
00244 xDstInSrc+= xInc;
00245 }
00246 } else {
00247 int64_t xDstInSrc;
00248 int sizeFactor;
00249
00250 if (flags&SWS_BICUBIC) sizeFactor= 4;
00251 else if (flags&SWS_X) sizeFactor= 8;
00252 else if (flags&SWS_AREA) sizeFactor= 1;
00253 else if (flags&SWS_GAUSS) sizeFactor= 8;
00254 else if (flags&SWS_LANCZOS) sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
00255 else if (flags&SWS_SINC) sizeFactor= 20;
00256 else if (flags&SWS_SPLINE) sizeFactor= 20;
00257 else if (flags&SWS_BILINEAR) sizeFactor= 2;
00258 else {
00259 sizeFactor= 0;
00260 assert(0);
00261 }
00262
00263 if (xInc <= 1<<16) filterSize= 1 + sizeFactor;
00264 else filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
00265
00266 filterSize = av_clip(filterSize, 1, srcW - 2);
00267
00268 FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
00269
00270 xDstInSrc= xInc - 0x10000;
00271 for (i=0; i<dstW; i++) {
00272 int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
00273 int j;
00274 (*filterPos)[i]= xx;
00275 for (j=0; j<filterSize; j++) {
00276 int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
00277 double floatd;
00278 int64_t coeff;
00279
00280 if (xInc > 1<<16)
00281 d= d*dstW/srcW;
00282 floatd= d * (1.0/(1<<30));
00283
00284 if (flags & SWS_BICUBIC) {
00285 int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1<<24);
00286 int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
00287
00288 if (d >= 1LL<<31) {
00289 coeff = 0.0;
00290 } else {
00291 int64_t dd = (d * d) >> 30;
00292 int64_t ddd = (dd * d) >> 30;
00293
00294 if (d < 1LL<<30)
00295 coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
00296 else
00297 coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
00298 }
00299 coeff *= fone>>(30+24);
00300 }
00301
00302
00303
00304
00305
00306 else if (flags & SWS_X) {
00307 double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
00308 double c;
00309
00310 if (floatd<1.0)
00311 c = cos(floatd*M_PI);
00312 else
00313 c=-1.0;
00314 if (c<0.0) c= -pow(-c, A);
00315 else c= pow( c, A);
00316 coeff= (c*0.5 + 0.5)*fone;
00317 } else if (flags & SWS_AREA) {
00318 int64_t d2= d - (1<<29);
00319 if (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
00320 else if (d2*xInc < (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
00321 else coeff=0.0;
00322 coeff *= fone>>(30+16);
00323 } else if (flags & SWS_GAUSS) {
00324 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
00325 coeff = (pow(2.0, - p*floatd*floatd))*fone;
00326 } else if (flags & SWS_SINC) {
00327 coeff = (d ? sin(floatd*M_PI)/(floatd*M_PI) : 1.0)*fone;
00328 } else if (flags & SWS_LANCZOS) {
00329 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
00330 coeff = (d ? sin(floatd*M_PI)*sin(floatd*M_PI/p)/(floatd*floatd*M_PI*M_PI/p) : 1.0)*fone;
00331 if (floatd>p) coeff=0;
00332 } else if (flags & SWS_BILINEAR) {
00333 coeff= (1<<30) - d;
00334 if (coeff<0) coeff=0;
00335 coeff *= fone >> 30;
00336 } else if (flags & SWS_SPLINE) {
00337 double p=-2.196152422706632;
00338 coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
00339 } else {
00340 coeff= 0.0;
00341 assert(0);
00342 }
00343
00344 filter[i*filterSize + j]= coeff;
00345 xx++;
00346 }
00347 xDstInSrc+= 2*xInc;
00348 }
00349 }
00350
00351
00352
00353
00354 assert(filterSize>0);
00355 filter2Size= filterSize;
00356 if (srcFilter) filter2Size+= srcFilter->length - 1;
00357 if (dstFilter) filter2Size+= dstFilter->length - 1;
00358 assert(filter2Size>0);
00359 FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail);
00360
00361 for (i=0; i<dstW; i++) {
00362 int j, k;
00363
00364 if(srcFilter) {
00365 for (k=0; k<srcFilter->length; k++) {
00366 for (j=0; j<filterSize; j++)
00367 filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
00368 }
00369 } else {
00370 for (j=0; j<filterSize; j++)
00371 filter2[i*filter2Size + j]= filter[i*filterSize + j];
00372 }
00373
00374
00375 (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
00376 }
00377 av_freep(&filter);
00378
00379
00380
00381 minFilterSize= 0;
00382 for (i=dstW-1; i>=0; i--) {
00383 int min= filter2Size;
00384 int j;
00385 int64_t cutOff=0.0;
00386
00387
00388 for (j=0; j<filter2Size; j++) {
00389 int k;
00390 cutOff += FFABS(filter2[i*filter2Size]);
00391
00392 if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
00393
00394
00395 if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
00396
00397
00398 for (k=1; k<filter2Size; k++)
00399 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
00400 filter2[i*filter2Size + k - 1]= 0;
00401 (*filterPos)[i]++;
00402 }
00403
00404 cutOff=0;
00405
00406 for (j=filter2Size-1; j>0; j--) {
00407 cutOff += FFABS(filter2[i*filter2Size + j]);
00408
00409 if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
00410 min--;
00411 }
00412
00413 if (min>minFilterSize) minFilterSize= min;
00414 }
00415
00416 if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) {
00417
00418
00419 if (minFilterSize < 5)
00420 filterAlign = 4;
00421
00422
00423
00424
00425
00426
00427 if (minFilterSize < 3)
00428 filterAlign = 1;
00429 }
00430
00431 if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
00432
00433 if (minFilterSize == 1 && filterAlign == 2)
00434 filterAlign= 1;
00435 }
00436
00437 assert(minFilterSize > 0);
00438 filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
00439 assert(filterSize > 0);
00440 filter= av_malloc(filterSize*dstW*sizeof(*filter));
00441 if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
00442 goto fail;
00443 *outFilterSize= filterSize;
00444
00445 if (flags&SWS_PRINT_INFO)
00446 av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
00447
00448 for (i=0; i<dstW; i++) {
00449 int j;
00450
00451 for (j=0; j<filterSize; j++) {
00452 if (j>=filter2Size) filter[i*filterSize + j]= 0;
00453 else filter[i*filterSize + j]= filter2[i*filter2Size + j];
00454 if((flags & SWS_BITEXACT) && j>=minFilterSize)
00455 filter[i*filterSize + j]= 0;
00456 }
00457 }
00458
00459
00460
00461
00462 if (is_horizontal) {
00463 for (i = 0; i < dstW; i++) {
00464 int j;
00465 if ((*filterPos)[i] < 0) {
00466
00467 for (j = 1; j < filterSize; j++) {
00468 int left = FFMAX(j + (*filterPos)[i], 0);
00469 filter[i * filterSize + left] += filter[i * filterSize + j];
00470 filter[i * filterSize + j ] = 0;
00471 }
00472 (*filterPos)[i] = 0;
00473 }
00474
00475 if ((*filterPos)[i] + filterSize > srcW) {
00476 int shift = (*filterPos)[i] + filterSize - srcW;
00477
00478 for (j = filterSize - 2; j >= 0; j--) {
00479 int right = FFMIN(j + shift, filterSize - 1);
00480 filter[i * filterSize + right] += filter[i * filterSize + j];
00481 filter[i * filterSize + j ] = 0;
00482 }
00483 (*filterPos)[i] = srcW - filterSize;
00484 }
00485 }
00486 }
00487
00488
00489
00490 FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+3)*sizeof(int16_t), fail);
00491
00492
00493 for (i=0; i<dstW; i++) {
00494 int j;
00495 int64_t error=0;
00496 int64_t sum=0;
00497
00498 for (j=0; j<filterSize; j++) {
00499 sum+= filter[i*filterSize + j];
00500 }
00501 sum= (sum + one/2)/ one;
00502 for (j=0; j<*outFilterSize; j++) {
00503 int64_t v= filter[i*filterSize + j] + error;
00504 int intV= ROUNDED_DIV(v, sum);
00505 (*outFilter)[i*(*outFilterSize) + j]= intV;
00506 error= v - intV*sum;
00507 }
00508 }
00509
00510 (*filterPos)[dstW+0] =
00511 (*filterPos)[dstW+1] =
00512 (*filterPos)[dstW+2] = (*filterPos)[dstW-1];
00513 for (i=0; i<*outFilterSize; i++) {
00514 int k= (dstW - 1) * (*outFilterSize) + i;
00515 (*outFilter)[k + 1 * (*outFilterSize)] =
00516 (*outFilter)[k + 2 * (*outFilterSize)] =
00517 (*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k];
00518 }
00519
00520 ret=0;
00521 fail:
00522 av_free(filter);
00523 av_free(filter2);
00524 return ret;
00525 }
00526
00527 #if HAVE_MMX2
00528 static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)
00529 {
00530 uint8_t *fragmentA;
00531 x86_reg imm8OfPShufW1A;
00532 x86_reg imm8OfPShufW2A;
00533 x86_reg fragmentLengthA;
00534 uint8_t *fragmentB;
00535 x86_reg imm8OfPShufW1B;
00536 x86_reg imm8OfPShufW2B;
00537 x86_reg fragmentLengthB;
00538 int fragmentPos;
00539
00540 int xpos, i;
00541
00542
00543
00544
00545
00546
00547
00548
00549
00550
00551
00552 __asm__ volatile(
00553 "jmp 9f \n\t"
00554
00555 "0: \n\t"
00556 "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
00557 "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
00558 "movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t"
00559 "punpcklbw %%mm7, %%mm1 \n\t"
00560 "punpcklbw %%mm7, %%mm0 \n\t"
00561 "pshufw $0xFF, %%mm1, %%mm1 \n\t"
00562 "1: \n\t"
00563 "pshufw $0xFF, %%mm0, %%mm0 \n\t"
00564 "2: \n\t"
00565 "psubw %%mm1, %%mm0 \n\t"
00566 "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
00567 "pmullw %%mm3, %%mm0 \n\t"
00568 "psllw $7, %%mm1 \n\t"
00569 "paddw %%mm1, %%mm0 \n\t"
00570
00571 "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
00572
00573 "add $8, %%"REG_a" \n\t"
00574
00575 "9: \n\t"
00576
00577 "lea " LOCAL_MANGLE(0b) ", %0 \n\t"
00578 "lea " LOCAL_MANGLE(1b) ", %1 \n\t"
00579 "lea " LOCAL_MANGLE(2b) ", %2 \n\t"
00580 "dec %1 \n\t"
00581 "dec %2 \n\t"
00582 "sub %0, %1 \n\t"
00583 "sub %0, %2 \n\t"
00584 "lea " LOCAL_MANGLE(9b) ", %3 \n\t"
00585 "sub %0, %3 \n\t"
00586
00587
00588 :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
00589 "=r" (fragmentLengthA)
00590 );
00591
00592 __asm__ volatile(
00593 "jmp 9f \n\t"
00594
00595 "0: \n\t"
00596 "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
00597 "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
00598 "punpcklbw %%mm7, %%mm0 \n\t"
00599 "pshufw $0xFF, %%mm0, %%mm1 \n\t"
00600 "1: \n\t"
00601 "pshufw $0xFF, %%mm0, %%mm0 \n\t"
00602 "2: \n\t"
00603 "psubw %%mm1, %%mm0 \n\t"
00604 "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
00605 "pmullw %%mm3, %%mm0 \n\t"
00606 "psllw $7, %%mm1 \n\t"
00607 "paddw %%mm1, %%mm0 \n\t"
00608
00609 "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
00610
00611 "add $8, %%"REG_a" \n\t"
00612
00613 "9: \n\t"
00614
00615 "lea " LOCAL_MANGLE(0b) ", %0 \n\t"
00616 "lea " LOCAL_MANGLE(1b) ", %1 \n\t"
00617 "lea " LOCAL_MANGLE(2b) ", %2 \n\t"
00618 "dec %1 \n\t"
00619 "dec %2 \n\t"
00620 "sub %0, %1 \n\t"
00621 "sub %0, %2 \n\t"
00622 "lea " LOCAL_MANGLE(9b) ", %3 \n\t"
00623 "sub %0, %3 \n\t"
00624
00625
00626 :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
00627 "=r" (fragmentLengthB)
00628 );
00629
00630 xpos= 0;
00631 fragmentPos=0;
00632
00633 for (i=0; i<dstW/numSplits; i++) {
00634 int xx=xpos>>16;
00635
00636 if ((i&3) == 0) {
00637 int a=0;
00638 int b=((xpos+xInc)>>16) - xx;
00639 int c=((xpos+xInc*2)>>16) - xx;
00640 int d=((xpos+xInc*3)>>16) - xx;
00641 int inc = (d+1<4);
00642 uint8_t *fragment = (d+1<4) ? fragmentB : fragmentA;
00643 x86_reg imm8OfPShufW1 = (d+1<4) ? imm8OfPShufW1B : imm8OfPShufW1A;
00644 x86_reg imm8OfPShufW2 = (d+1<4) ? imm8OfPShufW2B : imm8OfPShufW2A;
00645 x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;
00646 int maxShift= 3-(d+inc);
00647 int shift=0;
00648
00649 if (filterCode) {
00650 filter[i ] = (( xpos & 0xFFFF) ^ 0xFFFF)>>9;
00651 filter[i+1] = (((xpos+xInc ) & 0xFFFF) ^ 0xFFFF)>>9;
00652 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
00653 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
00654 filterPos[i/2]= xx;
00655
00656 memcpy(filterCode + fragmentPos, fragment, fragmentLength);
00657
00658 filterCode[fragmentPos + imm8OfPShufW1]=
00659 (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);
00660 filterCode[fragmentPos + imm8OfPShufW2]=
00661 a | (b<<2) | (c<<4) | (d<<6);
00662
00663 if (i+4-inc>=dstW) shift=maxShift;
00664 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3;
00665
00666 if (shift && i>=shift) {
00667 filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;
00668 filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;
00669 filterPos[i/2]-=shift;
00670 }
00671 }
00672
00673 fragmentPos+= fragmentLength;
00674
00675 if (filterCode)
00676 filterCode[fragmentPos]= RET;
00677 }
00678 xpos+=xInc;
00679 }
00680 if (filterCode)
00681 filterPos[((i/2)+1)&(~1)]= xpos>>16;
00682
00683 return fragmentPos + 1;
00684 }
00685 #endif
00686
00687 static void getSubSampleFactors(int *h, int *v, enum PixelFormat format)
00688 {
00689 *h = av_pix_fmt_descriptors[format].log2_chroma_w;
00690 *v = av_pix_fmt_descriptors[format].log2_chroma_h;
00691 }
00692
00693 int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
00694 int srcRange, const int table[4], int dstRange,
00695 int brightness, int contrast, int saturation)
00696 {
00697 memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
00698 memcpy(c->dstColorspaceTable, table, sizeof(int)*4);
00699
00700 c->brightness= brightness;
00701 c->contrast = contrast;
00702 c->saturation= saturation;
00703 c->srcRange = srcRange;
00704 c->dstRange = dstRange;
00705 if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
00706
00707 c->dstFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[c->dstFormat]);
00708 c->srcFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[c->srcFormat]);
00709
00710 ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
00711
00712
00713 if (HAVE_ALTIVEC && av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC)
00714 ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
00715 return 0;
00716 }
00717
00718 int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
00719 int *srcRange, int **table, int *dstRange,
00720 int *brightness, int *contrast, int *saturation)
00721 {
00722 if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
00723
00724 *inv_table = c->srcColorspaceTable;
00725 *table = c->dstColorspaceTable;
00726 *srcRange = c->srcRange;
00727 *dstRange = c->dstRange;
00728 *brightness= c->brightness;
00729 *contrast = c->contrast;
00730 *saturation= c->saturation;
00731
00732 return 0;
00733 }
00734
00735 static int handle_jpeg(enum PixelFormat *format)
00736 {
00737 switch (*format) {
00738 case PIX_FMT_YUVJ420P: *format = PIX_FMT_YUV420P; return 1;
00739 case PIX_FMT_YUVJ422P: *format = PIX_FMT_YUV422P; return 1;
00740 case PIX_FMT_YUVJ444P: *format = PIX_FMT_YUV444P; return 1;
00741 case PIX_FMT_YUVJ440P: *format = PIX_FMT_YUV440P; return 1;
00742 default: return 0;
00743 }
00744 }
00745
00746 SwsContext *sws_alloc_context(void)
00747 {
00748 SwsContext *c= av_mallocz(sizeof(SwsContext));
00749
00750 c->av_class = &sws_context_class;
00751 av_opt_set_defaults(c);
00752
00753 return c;
00754 }
00755
00756 int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
00757 {
00758 int i;
00759 int usesVFilter, usesHFilter;
00760 int unscaled;
00761 SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
00762 int srcW= c->srcW;
00763 int srcH= c->srcH;
00764 int dstW= c->dstW;
00765 int dstH= c->dstH;
00766 int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 16, 16), dst_stride_px = dst_stride >> 1;
00767 int flags, cpu_flags;
00768 enum PixelFormat srcFormat= c->srcFormat;
00769 enum PixelFormat dstFormat= c->dstFormat;
00770
00771 cpu_flags = av_get_cpu_flags();
00772 flags = c->flags;
00773 emms_c();
00774 if (!rgb15to16) sws_rgb2rgb_init();
00775
00776 unscaled = (srcW == dstW && srcH == dstH);
00777
00778 if (!sws_isSupportedInput(srcFormat)) {
00779 av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n", sws_format_name(srcFormat));
00780 return AVERROR(EINVAL);
00781 }
00782 if (!sws_isSupportedOutput(dstFormat)) {
00783 av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n", sws_format_name(dstFormat));
00784 return AVERROR(EINVAL);
00785 }
00786
00787 i= flags & ( SWS_POINT
00788 |SWS_AREA
00789 |SWS_BILINEAR
00790 |SWS_FAST_BILINEAR
00791 |SWS_BICUBIC
00792 |SWS_X
00793 |SWS_GAUSS
00794 |SWS_LANCZOS
00795 |SWS_SINC
00796 |SWS_SPLINE
00797 |SWS_BICUBLIN);
00798 if(!i || (i & (i-1))) {
00799 av_log(c, AV_LOG_ERROR, "Exactly one scaler algorithm must be chosen\n");
00800 return AVERROR(EINVAL);
00801 }
00802
00803 if (srcW<4 || srcH<1 || dstW<8 || dstH<1) {
00804 av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
00805 srcW, srcH, dstW, dstH);
00806 return AVERROR(EINVAL);
00807 }
00808
00809 if (!dstFilter) dstFilter= &dummyFilter;
00810 if (!srcFilter) srcFilter= &dummyFilter;
00811
00812 c->lumXInc= (((int64_t)srcW<<16) + (dstW>>1))/dstW;
00813 c->lumYInc= (((int64_t)srcH<<16) + (dstH>>1))/dstH;
00814 c->dstFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[dstFormat]);
00815 c->srcFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[srcFormat]);
00816 c->vRounder= 4* 0x0001000100010001ULL;
00817
00818 usesVFilter = (srcFilter->lumV && srcFilter->lumV->length>1) ||
00819 (srcFilter->chrV && srcFilter->chrV->length>1) ||
00820 (dstFilter->lumV && dstFilter->lumV->length>1) ||
00821 (dstFilter->chrV && dstFilter->chrV->length>1);
00822 usesHFilter = (srcFilter->lumH && srcFilter->lumH->length>1) ||
00823 (srcFilter->chrH && srcFilter->chrH->length>1) ||
00824 (dstFilter->lumH && dstFilter->lumH->length>1) ||
00825 (dstFilter->chrH && dstFilter->chrH->length>1);
00826
00827 getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
00828 getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
00829
00830
00831 if (flags & SWS_FULL_CHR_H_INT &&
00832 isAnyRGB(dstFormat) &&
00833 dstFormat != PIX_FMT_RGBA &&
00834 dstFormat != PIX_FMT_ARGB &&
00835 dstFormat != PIX_FMT_BGRA &&
00836 dstFormat != PIX_FMT_ABGR &&
00837 dstFormat != PIX_FMT_RGB24 &&
00838 dstFormat != PIX_FMT_BGR24) {
00839 av_log(c, AV_LOG_ERROR,
00840 "full chroma interpolation for destination format '%s' not yet implemented\n",
00841 sws_format_name(dstFormat));
00842 flags &= ~SWS_FULL_CHR_H_INT;
00843 c->flags = flags;
00844 }
00845 if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
00846
00847
00848 c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
00849 c->chrSrcVSubSample+= c->vChrDrop;
00850
00851
00852 if (isAnyRGB(srcFormat) && !(flags&SWS_FULL_CHR_H_INP)
00853 && srcFormat!=PIX_FMT_RGB8 && srcFormat!=PIX_FMT_BGR8
00854 && srcFormat!=PIX_FMT_RGB4 && srcFormat!=PIX_FMT_BGR4
00855 && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
00856 && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&SWS_FAST_BILINEAR)))
00857 c->chrSrcHSubSample=1;
00858
00859
00860 c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
00861 c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
00862 c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
00863 c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
00864
00865
00866 if (unscaled && !usesHFilter && !usesVFilter && (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
00867 ff_get_unscaled_swscale(c);
00868
00869 if (c->swScale) {
00870 if (flags&SWS_PRINT_INFO)
00871 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
00872 sws_format_name(srcFormat), sws_format_name(dstFormat));
00873 return 0;
00874 }
00875 }
00876
00877 c->srcBpc = 1 + av_pix_fmt_descriptors[srcFormat].comp[0].depth_minus1;
00878 if (c->srcBpc < 8)
00879 c->srcBpc = 8;
00880 c->dstBpc = 1 + av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1;
00881 if (c->dstBpc < 8)
00882 c->dstBpc = 8;
00883 if (c->dstBpc == 16)
00884 dst_stride <<= 1;
00885 FF_ALLOC_OR_GOTO(c, c->formatConvBuffer,
00886 (FFALIGN(srcW, 16) * 2 * FFALIGN(c->srcBpc, 8) >> 3) + 16,
00887 fail);
00888 if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2 && c->srcBpc == 8 && c->dstBpc <= 10) {
00889 c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
00890 if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {
00891 if (flags&SWS_PRINT_INFO)
00892 av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
00893 }
00894 if (usesHFilter) c->canMMX2BeUsed=0;
00895 }
00896 else
00897 c->canMMX2BeUsed=0;
00898
00899 c->chrXInc= (((int64_t)c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
00900 c->chrYInc= (((int64_t)c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
00901
00902
00903
00904
00905
00906
00907
00908 if (flags&SWS_FAST_BILINEAR) {
00909 if (c->canMMX2BeUsed) {
00910 c->lumXInc+= 20;
00911 c->chrXInc+= 20;
00912 }
00913
00914 else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
00915 c->lumXInc = ((int64_t)(srcW-2)<<16)/(dstW-2) - 20;
00916 c->chrXInc = ((int64_t)(c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
00917 }
00918 }
00919
00920
00921 {
00922 #if HAVE_MMX2
00923
00924 if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {
00925 c->lumMmx2FilterCodeSize = initMMX2HScaler( dstW, c->lumXInc, NULL, NULL, NULL, 8);
00926 c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);
00927
00928 #ifdef MAP_ANONYMOUS
00929 c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
00930 c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
00931 #elif HAVE_VIRTUALALLOC
00932 c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
00933 c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
00934 #else
00935 c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize);
00936 c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);
00937 #endif
00938
00939 if (!c->lumMmx2FilterCode || !c->chrMmx2FilterCode)
00940 return AVERROR(ENOMEM);
00941 FF_ALLOCZ_OR_GOTO(c, c->hLumFilter , (dstW /8+8)*sizeof(int16_t), fail);
00942 FF_ALLOCZ_OR_GOTO(c, c->hChrFilter , (c->chrDstW /4+8)*sizeof(int16_t), fail);
00943 FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW /2/8+8)*sizeof(int32_t), fail);
00944 FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW/2/4+8)*sizeof(int32_t), fail);
00945
00946 initMMX2HScaler( dstW, c->lumXInc, c->lumMmx2FilterCode, c->hLumFilter, c->hLumFilterPos, 8);
00947 initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->hChrFilter, c->hChrFilterPos, 4);
00948
00949 #ifdef MAP_ANONYMOUS
00950 mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
00951 mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
00952 #endif
00953 } else
00954 #endif
00955 {
00956 const int filterAlign=
00957 (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 :
00958 (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
00959 1;
00960
00961 if (initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
00962 srcW , dstW, filterAlign, 1<<14,
00963 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags, cpu_flags,
00964 srcFilter->lumH, dstFilter->lumH, c->param, 1) < 0)
00965 goto fail;
00966 if (initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
00967 c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
00968 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags, cpu_flags,
00969 srcFilter->chrH, dstFilter->chrH, c->param, 1) < 0)
00970 goto fail;
00971 }
00972 }
00973
00974
00975 {
00976 const int filterAlign=
00977 (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 2 :
00978 (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
00979 1;
00980
00981 if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
00982 srcH , dstH, filterAlign, (1<<12),
00983 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags, cpu_flags,
00984 srcFilter->lumV, dstFilter->lumV, c->param, 0) < 0)
00985 goto fail;
00986 if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
00987 c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
00988 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags, cpu_flags,
00989 srcFilter->chrV, dstFilter->chrV, c->param, 0) < 0)
00990 goto fail;
00991
00992 #if HAVE_ALTIVEC
00993 FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof (vector signed short)*c->vLumFilterSize*c->dstH, fail);
00994 FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH, fail);
00995
00996 for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
00997 int j;
00998 short *p = (short *)&c->vYCoeffsBank[i];
00999 for (j=0;j<8;j++)
01000 p[j] = c->vLumFilter[i];
01001 }
01002
01003 for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
01004 int j;
01005 short *p = (short *)&c->vCCoeffsBank[i];
01006 for (j=0;j<8;j++)
01007 p[j] = c->vChrFilter[i];
01008 }
01009 #endif
01010 }
01011
01012
01013 c->vLumBufSize= c->vLumFilterSize;
01014 c->vChrBufSize= c->vChrFilterSize;
01015 for (i=0; i<dstH; i++) {
01016 int chrI = (int64_t) i * c->chrDstH / dstH;
01017 int nextSlice= FFMAX(c->vLumFilterPos[i ] + c->vLumFilterSize - 1,
01018 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
01019
01020 nextSlice>>= c->chrSrcVSubSample;
01021 nextSlice<<= c->chrSrcVSubSample;
01022 if (c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice)
01023 c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
01024 if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
01025 c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
01026 }
01027
01028
01029
01030 FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize*3*sizeof(int16_t*), fail);
01031 FF_ALLOC_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize*3*sizeof(int16_t*), fail);
01032 FF_ALLOC_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize*3*sizeof(int16_t*), fail);
01033 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
01034 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*3*sizeof(int16_t*), fail);
01035
01036
01037 for (i=0; i<c->vLumBufSize; i++) {
01038 FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i+c->vLumBufSize], dst_stride+16, fail);
01039 c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize];
01040 }
01041
01042 c->uv_off_px = dst_stride_px + 64 / (c->dstBpc &~ 7);
01043 c->uv_off_byte = dst_stride + 16;
01044 for (i=0; i<c->vChrBufSize; i++) {
01045 FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i+c->vChrBufSize], dst_stride*2+32, fail);
01046 c->chrUPixBuf[i] = c->chrUPixBuf[i+c->vChrBufSize];
01047 c->chrVPixBuf[i] = c->chrVPixBuf[i+c->vChrBufSize] = c->chrUPixBuf[i] + (dst_stride >> 1) + 8;
01048 }
01049 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
01050 for (i=0; i<c->vLumBufSize; i++) {
01051 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i+c->vLumBufSize], dst_stride+16, fail);
01052 c->alpPixBuf[i] = c->alpPixBuf[i+c->vLumBufSize];
01053 }
01054
01055
01056 for (i=0; i<c->vChrBufSize; i++)
01057 memset(c->chrUPixBuf[i], 64, dst_stride*2+1);
01058
01059 assert(c->chrDstH <= dstH);
01060
01061 if (flags&SWS_PRINT_INFO) {
01062 if (flags&SWS_FAST_BILINEAR) av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
01063 else if (flags&SWS_BILINEAR) av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
01064 else if (flags&SWS_BICUBIC) av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
01065 else if (flags&SWS_X) av_log(c, AV_LOG_INFO, "Experimental scaler, ");
01066 else if (flags&SWS_POINT) av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
01067 else if (flags&SWS_AREA) av_log(c, AV_LOG_INFO, "Area Averaging scaler, ");
01068 else if (flags&SWS_BICUBLIN) av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
01069 else if (flags&SWS_GAUSS) av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
01070 else if (flags&SWS_SINC) av_log(c, AV_LOG_INFO, "Sinc scaler, ");
01071 else if (flags&SWS_LANCZOS) av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
01072 else if (flags&SWS_SPLINE) av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
01073 else av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
01074
01075 av_log(c, AV_LOG_INFO, "from %s to %s%s ",
01076 sws_format_name(srcFormat),
01077 #ifdef DITHER1XBPP
01078 dstFormat == PIX_FMT_BGR555 || dstFormat == PIX_FMT_BGR565 ||
01079 dstFormat == PIX_FMT_RGB444BE || dstFormat == PIX_FMT_RGB444LE ||
01080 dstFormat == PIX_FMT_BGR444BE || dstFormat == PIX_FMT_BGR444LE ? "dithered " : "",
01081 #else
01082 "",
01083 #endif
01084 sws_format_name(dstFormat));
01085
01086 if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2) av_log(c, AV_LOG_INFO, "using MMX2\n");
01087 else if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW) av_log(c, AV_LOG_INFO, "using 3DNOW\n");
01088 else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) av_log(c, AV_LOG_INFO, "using MMX\n");
01089 else if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) av_log(c, AV_LOG_INFO, "using AltiVec\n");
01090 else av_log(c, AV_LOG_INFO, "using C\n");
01091
01092 av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
01093 av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
01094 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
01095 av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
01096 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
01097 }
01098
01099 c->swScale= ff_getSwsFunc(c);
01100 return 0;
01101 fail:
01102 return -1;
01103 }
01104
01105 #if FF_API_SWS_GETCONTEXT
01106 SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat,
01107 int dstW, int dstH, enum PixelFormat dstFormat, int flags,
01108 SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
01109 {
01110 SwsContext *c;
01111
01112 if(!(c=sws_alloc_context()))
01113 return NULL;
01114
01115 c->flags= flags;
01116 c->srcW= srcW;
01117 c->srcH= srcH;
01118 c->dstW= dstW;
01119 c->dstH= dstH;
01120 c->srcRange = handle_jpeg(&srcFormat);
01121 c->dstRange = handle_jpeg(&dstFormat);
01122 c->srcFormat= srcFormat;
01123 c->dstFormat= dstFormat;
01124
01125 if (param) {
01126 c->param[0] = param[0];
01127 c->param[1] = param[1];
01128 }
01129 sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] , c->dstRange, 0, 1<<16, 1<<16);
01130
01131 if(sws_init_context(c, srcFilter, dstFilter) < 0){
01132 sws_freeContext(c);
01133 return NULL;
01134 }
01135
01136 return c;
01137 }
01138 #endif
01139
01140 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
01141 float lumaSharpen, float chromaSharpen,
01142 float chromaHShift, float chromaVShift,
01143 int verbose)
01144 {
01145 SwsFilter *filter= av_malloc(sizeof(SwsFilter));
01146 if (!filter)
01147 return NULL;
01148
01149 if (lumaGBlur!=0.0) {
01150 filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
01151 filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
01152 } else {
01153 filter->lumH= sws_getIdentityVec();
01154 filter->lumV= sws_getIdentityVec();
01155 }
01156
01157 if (chromaGBlur!=0.0) {
01158 filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
01159 filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
01160 } else {
01161 filter->chrH= sws_getIdentityVec();
01162 filter->chrV= sws_getIdentityVec();
01163 }
01164
01165 if (chromaSharpen!=0.0) {
01166 SwsVector *id= sws_getIdentityVec();
01167 sws_scaleVec(filter->chrH, -chromaSharpen);
01168 sws_scaleVec(filter->chrV, -chromaSharpen);
01169 sws_addVec(filter->chrH, id);
01170 sws_addVec(filter->chrV, id);
01171 sws_freeVec(id);
01172 }
01173
01174 if (lumaSharpen!=0.0) {
01175 SwsVector *id= sws_getIdentityVec();
01176 sws_scaleVec(filter->lumH, -lumaSharpen);
01177 sws_scaleVec(filter->lumV, -lumaSharpen);
01178 sws_addVec(filter->lumH, id);
01179 sws_addVec(filter->lumV, id);
01180 sws_freeVec(id);
01181 }
01182
01183 if (chromaHShift != 0.0)
01184 sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
01185
01186 if (chromaVShift != 0.0)
01187 sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
01188
01189 sws_normalizeVec(filter->chrH, 1.0);
01190 sws_normalizeVec(filter->chrV, 1.0);
01191 sws_normalizeVec(filter->lumH, 1.0);
01192 sws_normalizeVec(filter->lumV, 1.0);
01193
01194 if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
01195 if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
01196
01197 return filter;
01198 }
01199
01200 SwsVector *sws_allocVec(int length)
01201 {
01202 SwsVector *vec = av_malloc(sizeof(SwsVector));
01203 if (!vec)
01204 return NULL;
01205 vec->length = length;
01206 vec->coeff = av_malloc(sizeof(double) * length);
01207 if (!vec->coeff)
01208 av_freep(&vec);
01209 return vec;
01210 }
01211
01212 SwsVector *sws_getGaussianVec(double variance, double quality)
01213 {
01214 const int length= (int)(variance*quality + 0.5) | 1;
01215 int i;
01216 double middle= (length-1)*0.5;
01217 SwsVector *vec= sws_allocVec(length);
01218
01219 if (!vec)
01220 return NULL;
01221
01222 for (i=0; i<length; i++) {
01223 double dist= i-middle;
01224 vec->coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*M_PI);
01225 }
01226
01227 sws_normalizeVec(vec, 1.0);
01228
01229 return vec;
01230 }
01231
01232 SwsVector *sws_getConstVec(double c, int length)
01233 {
01234 int i;
01235 SwsVector *vec= sws_allocVec(length);
01236
01237 if (!vec)
01238 return NULL;
01239
01240 for (i=0; i<length; i++)
01241 vec->coeff[i]= c;
01242
01243 return vec;
01244 }
01245
01246 SwsVector *sws_getIdentityVec(void)
01247 {
01248 return sws_getConstVec(1.0, 1);
01249 }
01250
01251 static double sws_dcVec(SwsVector *a)
01252 {
01253 int i;
01254 double sum=0;
01255
01256 for (i=0; i<a->length; i++)
01257 sum+= a->coeff[i];
01258
01259 return sum;
01260 }
01261
01262 void sws_scaleVec(SwsVector *a, double scalar)
01263 {
01264 int i;
01265
01266 for (i=0; i<a->length; i++)
01267 a->coeff[i]*= scalar;
01268 }
01269
01270 void sws_normalizeVec(SwsVector *a, double height)
01271 {
01272 sws_scaleVec(a, height/sws_dcVec(a));
01273 }
01274
01275 static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
01276 {
01277 int length= a->length + b->length - 1;
01278 int i, j;
01279 SwsVector *vec= sws_getConstVec(0.0, length);
01280
01281 if (!vec)
01282 return NULL;
01283
01284 for (i=0; i<a->length; i++) {
01285 for (j=0; j<b->length; j++) {
01286 vec->coeff[i+j]+= a->coeff[i]*b->coeff[j];
01287 }
01288 }
01289
01290 return vec;
01291 }
01292
01293 static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
01294 {
01295 int length= FFMAX(a->length, b->length);
01296 int i;
01297 SwsVector *vec= sws_getConstVec(0.0, length);
01298
01299 if (!vec)
01300 return NULL;
01301
01302 for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
01303 for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
01304
01305 return vec;
01306 }
01307
01308 static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
01309 {
01310 int length= FFMAX(a->length, b->length);
01311 int i;
01312 SwsVector *vec= sws_getConstVec(0.0, length);
01313
01314 if (!vec)
01315 return NULL;
01316
01317 for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
01318 for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
01319
01320 return vec;
01321 }
01322
01323
01324 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
01325 {
01326 int length= a->length + FFABS(shift)*2;
01327 int i;
01328 SwsVector *vec= sws_getConstVec(0.0, length);
01329
01330 if (!vec)
01331 return NULL;
01332
01333 for (i=0; i<a->length; i++) {
01334 vec->coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
01335 }
01336
01337 return vec;
01338 }
01339
01340 void sws_shiftVec(SwsVector *a, int shift)
01341 {
01342 SwsVector *shifted= sws_getShiftedVec(a, shift);
01343 av_free(a->coeff);
01344 a->coeff= shifted->coeff;
01345 a->length= shifted->length;
01346 av_free(shifted);
01347 }
01348
01349 void sws_addVec(SwsVector *a, SwsVector *b)
01350 {
01351 SwsVector *sum= sws_sumVec(a, b);
01352 av_free(a->coeff);
01353 a->coeff= sum->coeff;
01354 a->length= sum->length;
01355 av_free(sum);
01356 }
01357
01358 void sws_subVec(SwsVector *a, SwsVector *b)
01359 {
01360 SwsVector *diff= sws_diffVec(a, b);
01361 av_free(a->coeff);
01362 a->coeff= diff->coeff;
01363 a->length= diff->length;
01364 av_free(diff);
01365 }
01366
01367 void sws_convVec(SwsVector *a, SwsVector *b)
01368 {
01369 SwsVector *conv= sws_getConvVec(a, b);
01370 av_free(a->coeff);
01371 a->coeff= conv->coeff;
01372 a->length= conv->length;
01373 av_free(conv);
01374 }
01375
01376 SwsVector *sws_cloneVec(SwsVector *a)
01377 {
01378 int i;
01379 SwsVector *vec= sws_allocVec(a->length);
01380
01381 if (!vec)
01382 return NULL;
01383
01384 for (i=0; i<a->length; i++) vec->coeff[i]= a->coeff[i];
01385
01386 return vec;
01387 }
01388
01389 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
01390 {
01391 int i;
01392 double max=0;
01393 double min=0;
01394 double range;
01395
01396 for (i=0; i<a->length; i++)
01397 if (a->coeff[i]>max) max= a->coeff[i];
01398
01399 for (i=0; i<a->length; i++)
01400 if (a->coeff[i]<min) min= a->coeff[i];
01401
01402 range= max - min;
01403
01404 for (i=0; i<a->length; i++) {
01405 int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
01406 av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
01407 for (;x>0; x--) av_log(log_ctx, log_level, " ");
01408 av_log(log_ctx, log_level, "|\n");
01409 }
01410 }
01411
01412 void sws_freeVec(SwsVector *a)
01413 {
01414 if (!a) return;
01415 av_freep(&a->coeff);
01416 a->length=0;
01417 av_free(a);
01418 }
01419
01420 void sws_freeFilter(SwsFilter *filter)
01421 {
01422 if (!filter) return;
01423
01424 if (filter->lumH) sws_freeVec(filter->lumH);
01425 if (filter->lumV) sws_freeVec(filter->lumV);
01426 if (filter->chrH) sws_freeVec(filter->chrH);
01427 if (filter->chrV) sws_freeVec(filter->chrV);
01428 av_free(filter);
01429 }
01430
01431 void sws_freeContext(SwsContext *c)
01432 {
01433 int i;
01434 if (!c) return;
01435
01436 if (c->lumPixBuf) {
01437 for (i=0; i<c->vLumBufSize; i++)
01438 av_freep(&c->lumPixBuf[i]);
01439 av_freep(&c->lumPixBuf);
01440 }
01441
01442 if (c->chrUPixBuf) {
01443 for (i=0; i<c->vChrBufSize; i++)
01444 av_freep(&c->chrUPixBuf[i]);
01445 av_freep(&c->chrUPixBuf);
01446 av_freep(&c->chrVPixBuf);
01447 }
01448
01449 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
01450 for (i=0; i<c->vLumBufSize; i++)
01451 av_freep(&c->alpPixBuf[i]);
01452 av_freep(&c->alpPixBuf);
01453 }
01454
01455 av_freep(&c->vLumFilter);
01456 av_freep(&c->vChrFilter);
01457 av_freep(&c->hLumFilter);
01458 av_freep(&c->hChrFilter);
01459 #if HAVE_ALTIVEC
01460 av_freep(&c->vYCoeffsBank);
01461 av_freep(&c->vCCoeffsBank);
01462 #endif
01463
01464 av_freep(&c->vLumFilterPos);
01465 av_freep(&c->vChrFilterPos);
01466 av_freep(&c->hLumFilterPos);
01467 av_freep(&c->hChrFilterPos);
01468
01469 #if HAVE_MMX
01470 #ifdef MAP_ANONYMOUS
01471 if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);
01472 if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);
01473 #elif HAVE_VIRTUALALLOC
01474 if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, 0, MEM_RELEASE);
01475 if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, 0, MEM_RELEASE);
01476 #else
01477 av_free(c->lumMmx2FilterCode);
01478 av_free(c->chrMmx2FilterCode);
01479 #endif
01480 c->lumMmx2FilterCode=NULL;
01481 c->chrMmx2FilterCode=NULL;
01482 #endif
01483
01484 av_freep(&c->yuvTable);
01485 av_free(c->formatConvBuffer);
01486
01487 av_free(c);
01488 }
01489
01490 struct SwsContext *sws_getCachedContext(struct SwsContext *context,
01491 int srcW, int srcH, enum PixelFormat srcFormat,
01492 int dstW, int dstH, enum PixelFormat dstFormat, int flags,
01493 SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
01494 {
01495 static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
01496
01497 if (!param)
01498 param = default_param;
01499
01500 if (context &&
01501 (context->srcW != srcW ||
01502 context->srcH != srcH ||
01503 context->srcFormat != srcFormat ||
01504 context->dstW != dstW ||
01505 context->dstH != dstH ||
01506 context->dstFormat != dstFormat ||
01507 context->flags != flags ||
01508 context->param[0] != param[0] ||
01509 context->param[1] != param[1])) {
01510 sws_freeContext(context);
01511 context = NULL;
01512 }
01513
01514 if (!context) {
01515 if (!(context = sws_alloc_context()))
01516 return NULL;
01517 context->srcW = srcW;
01518 context->srcH = srcH;
01519 context->srcRange = handle_jpeg(&srcFormat);
01520 context->srcFormat = srcFormat;
01521 context->dstW = dstW;
01522 context->dstH = dstH;
01523 context->dstRange = handle_jpeg(&dstFormat);
01524 context->dstFormat = dstFormat;
01525 context->flags = flags;
01526 context->param[0] = param[0];
01527 context->param[1] = param[1];
01528 sws_setColorspaceDetails(context, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], context->srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] , context->dstRange, 0, 1<<16, 1<<16);
01529 if (sws_init_context(context, srcFilter, dstFilter) < 0) {
01530 sws_freeContext(context);
01531 return NULL;
01532 }
01533 }
01534 return context;
01535 }
01536