yading@11: /* yading@11: * Copyright (C) 2001-2003 Michael Niedermayer yading@11: * yading@11: * This file is part of FFmpeg. yading@11: * yading@11: * FFmpeg is free software; you can redistribute it and/or yading@11: * modify it under the terms of the GNU Lesser General Public yading@11: * License as published by the Free Software Foundation; either yading@11: * version 2.1 of the License, or (at your option) any later version. yading@11: * yading@11: * FFmpeg is distributed in the hope that it will be useful, yading@11: * but WITHOUT ANY WARRANTY; without even the implied warranty of yading@11: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU yading@11: * Lesser General Public License for more details. yading@11: * yading@11: * You should have received a copy of the GNU Lesser General Public yading@11: * License along with FFmpeg; if not, write to the Free Software yading@11: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA yading@11: */ yading@11: yading@11: #include "config.h" yading@11: yading@11: #define _SVID_SOURCE // needed for MAP_ANONYMOUS yading@11: #define _DARWIN_C_SOURCE // needed for MAP_ANON yading@11: #include yading@11: #include yading@11: #include yading@11: #include yading@11: #if HAVE_SYS_MMAN_H yading@11: #include yading@11: #if defined(MAP_ANON) && !defined(MAP_ANONYMOUS) yading@11: #define MAP_ANONYMOUS MAP_ANON yading@11: #endif yading@11: #endif yading@11: #if HAVE_VIRTUALALLOC yading@11: #define WIN32_LEAN_AND_MEAN yading@11: #include yading@11: #endif yading@11: yading@11: #include "libavutil/attributes.h" yading@11: #include "libavutil/avassert.h" yading@11: #include "libavutil/avutil.h" yading@11: #include "libavutil/bswap.h" yading@11: #include "libavutil/cpu.h" yading@11: #include "libavutil/intreadwrite.h" yading@11: #include "libavutil/mathematics.h" yading@11: #include "libavutil/opt.h" yading@11: #include "libavutil/pixdesc.h" yading@11: #include "libavutil/x86/asm.h" yading@11: #include "libavutil/x86/cpu.h" yading@11: #include "rgb2rgb.h" yading@11: #include "swscale.h" yading@11: #include "swscale_internal.h" yading@11: yading@11: unsigned swscale_version(void) yading@11: { yading@11: av_assert0(LIBSWSCALE_VERSION_MICRO >= 100); yading@11: return LIBSWSCALE_VERSION_INT; yading@11: } yading@11: yading@11: const char *swscale_configuration(void) yading@11: { yading@11: return FFMPEG_CONFIGURATION; yading@11: } yading@11: yading@11: const char *swscale_license(void) yading@11: { yading@11: #define LICENSE_PREFIX "libswscale license: " yading@11: return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1; yading@11: } yading@11: yading@11: #define RET 0xC3 // near return opcode for x86 yading@11: yading@11: typedef struct FormatEntry { yading@11: int is_supported_in, is_supported_out; yading@11: } FormatEntry; yading@11: yading@11: static const FormatEntry format_entries[AV_PIX_FMT_NB] = { yading@11: [AV_PIX_FMT_YUV420P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUYV422] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB24] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR24] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV410P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV411P] = { 1, 1 }, yading@11: [AV_PIX_FMT_GRAY8] = { 1, 1 }, yading@11: [AV_PIX_FMT_MONOWHITE] = { 1, 1 }, yading@11: [AV_PIX_FMT_MONOBLACK] = { 1, 1 }, yading@11: [AV_PIX_FMT_PAL8] = { 1, 0 }, yading@11: [AV_PIX_FMT_YUVJ420P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVJ422P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVJ444P] = { 1, 1 }, yading@11: [AV_PIX_FMT_UYVY422] = { 1, 1 }, yading@11: [AV_PIX_FMT_UYYVYY411] = { 0, 0 }, yading@11: [AV_PIX_FMT_BGR8] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR4] = { 0, 1 }, yading@11: [AV_PIX_FMT_BGR4_BYTE] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB8] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB4] = { 0, 1 }, yading@11: [AV_PIX_FMT_RGB4_BYTE] = { 1, 1 }, yading@11: [AV_PIX_FMT_NV12] = { 1, 1 }, yading@11: [AV_PIX_FMT_NV21] = { 1, 1 }, yading@11: [AV_PIX_FMT_ARGB] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGBA] = { 1, 1 }, yading@11: [AV_PIX_FMT_ABGR] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGRA] = { 1, 1 }, yading@11: [AV_PIX_FMT_0RGB] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB0] = { 1, 1 }, yading@11: [AV_PIX_FMT_0BGR] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR0] = { 1, 1 }, yading@11: [AV_PIX_FMT_GRAY16BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GRAY16LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV440P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVJ440P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA420P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA422P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA444P] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA420P9BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA420P9LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA422P9BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA422P9LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA444P9BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA444P9LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA420P10BE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA420P10LE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA422P10BE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA422P10LE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA444P10BE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA444P10LE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA420P16BE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA420P16LE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA422P16BE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA422P16LE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA444P16BE]= { 1, 1 }, yading@11: [AV_PIX_FMT_YUVA444P16LE]= { 1, 1 }, yading@11: [AV_PIX_FMT_RGB48BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB48LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGBA64BE] = { 1, 0 }, yading@11: [AV_PIX_FMT_RGBA64LE] = { 1, 0 }, yading@11: [AV_PIX_FMT_RGB565BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB565LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB555BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB555LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR565BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR565LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR555BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR555LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P16LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P16BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P16LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P16BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P16LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P16BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB444LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_RGB444BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR444LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR444BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_Y400A] = { 1, 0 }, yading@11: [AV_PIX_FMT_BGR48BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGR48LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_BGRA64BE] = { 0, 0 }, yading@11: [AV_PIX_FMT_BGRA64LE] = { 0, 0 }, yading@11: [AV_PIX_FMT_YUV420P9BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P9LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P10BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P10LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P12BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P12LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P14BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV420P14LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P9BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P9LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P10BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P10LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P12BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P12LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P14BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV422P14LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P9BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P9LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P10BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P10LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P12BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P12LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P14BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_YUV444P14LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP9LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP9BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP10LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP10BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP12LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP12BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP14LE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP14BE] = { 1, 1 }, yading@11: [AV_PIX_FMT_GBRP16LE] = { 1, 0 }, yading@11: [AV_PIX_FMT_GBRP16BE] = { 1, 0 }, yading@11: }; yading@11: yading@11: int sws_isSupportedInput(enum AVPixelFormat pix_fmt) yading@11: { yading@11: return (unsigned)pix_fmt < AV_PIX_FMT_NB ? yading@11: format_entries[pix_fmt].is_supported_in : 0; yading@11: } yading@11: yading@11: int sws_isSupportedOutput(enum AVPixelFormat pix_fmt) yading@11: { yading@11: return (unsigned)pix_fmt < AV_PIX_FMT_NB ? yading@11: format_entries[pix_fmt].is_supported_out : 0; yading@11: } yading@11: yading@11: extern const int32_t ff_yuv2rgb_coeffs[8][4]; yading@11: yading@11: #if FF_API_SWS_FORMAT_NAME yading@11: const char *sws_format_name(enum AVPixelFormat format) yading@11: { yading@11: const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(format); yading@11: if (desc) yading@11: return desc->name; yading@11: else yading@11: return "Unknown format"; yading@11: } yading@11: #endif yading@11: yading@11: static double getSplineCoeff(double a, double b, double c, double d, yading@11: double dist) yading@11: { yading@11: if (dist <= 1.0) yading@11: return ((d * dist + c) * dist + b) * dist + a; yading@11: else yading@11: return getSplineCoeff(0.0, yading@11: b + 2.0 * c + 3.0 * d, yading@11: c + 3.0 * d, yading@11: -b - 3.0 * c - 6.0 * d, yading@11: dist - 1.0); yading@11: } yading@11: yading@11: static int initFilter(int16_t **outFilter, int32_t **filterPos, yading@11: int *outFilterSize, int xInc, int srcW, int dstW, yading@11: int filterAlign, int one, int flags, int cpu_flags, yading@11: SwsVector *srcFilter, SwsVector *dstFilter, yading@11: double param[2]) yading@11: { yading@11: int i; yading@11: int filterSize; yading@11: int filter2Size; yading@11: int minFilterSize; yading@11: int64_t *filter = NULL; yading@11: int64_t *filter2 = NULL; yading@11: const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8)); yading@11: int ret = -1; yading@11: yading@11: emms_c(); // FIXME should not be required but IS (even for non-MMX versions) yading@11: yading@11: // NOTE: the +3 is for the MMX(+1) / SSE(+3) scaler which reads over the end yading@11: FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW + 3) * sizeof(**filterPos), fail); yading@11: yading@11: if (FFABS(xInc - 0x10000) < 10) { // unscaled yading@11: int i; yading@11: filterSize = 1; yading@11: FF_ALLOCZ_OR_GOTO(NULL, filter, yading@11: dstW * sizeof(*filter) * filterSize, fail); yading@11: yading@11: for (i = 0; i < dstW; i++) { yading@11: filter[i * filterSize] = fone; yading@11: (*filterPos)[i] = i; yading@11: } yading@11: } else if (flags & SWS_POINT) { // lame looking point sampling mode yading@11: int i; yading@11: int64_t xDstInSrc; yading@11: filterSize = 1; yading@11: FF_ALLOC_OR_GOTO(NULL, filter, yading@11: dstW * sizeof(*filter) * filterSize, fail); yading@11: yading@11: xDstInSrc = xInc / 2 - 0x8000; yading@11: for (i = 0; i < dstW; i++) { yading@11: int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16; yading@11: yading@11: (*filterPos)[i] = xx; yading@11: filter[i] = fone; yading@11: xDstInSrc += xInc; yading@11: } yading@11: } else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) || yading@11: (flags & SWS_FAST_BILINEAR)) { // bilinear upscale yading@11: int i; yading@11: int64_t xDstInSrc; yading@11: filterSize = 2; yading@11: FF_ALLOC_OR_GOTO(NULL, filter, yading@11: dstW * sizeof(*filter) * filterSize, fail); yading@11: yading@11: xDstInSrc = xInc / 2 - 0x8000; yading@11: for (i = 0; i < dstW; i++) { yading@11: int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16; yading@11: int j; yading@11: yading@11: (*filterPos)[i] = xx; yading@11: // bilinear upscale / linear interpolate / area averaging yading@11: for (j = 0; j < filterSize; j++) { yading@11: int64_t coeff= fone - FFABS(((int64_t)xx<<16) - xDstInSrc)*(fone>>16); yading@11: if (coeff < 0) yading@11: coeff = 0; yading@11: filter[i * filterSize + j] = coeff; yading@11: xx++; yading@11: } yading@11: xDstInSrc += xInc; yading@11: } yading@11: } else { yading@11: int64_t xDstInSrc; yading@11: int sizeFactor; yading@11: yading@11: if (flags & SWS_BICUBIC) yading@11: sizeFactor = 4; yading@11: else if (flags & SWS_X) yading@11: sizeFactor = 8; yading@11: else if (flags & SWS_AREA) yading@11: sizeFactor = 1; // downscale only, for upscale it is bilinear yading@11: else if (flags & SWS_GAUSS) yading@11: sizeFactor = 8; // infinite ;) yading@11: else if (flags & SWS_LANCZOS) yading@11: sizeFactor = param[0] != SWS_PARAM_DEFAULT ? ceil(2 * param[0]) : 6; yading@11: else if (flags & SWS_SINC) yading@11: sizeFactor = 20; // infinite ;) yading@11: else if (flags & SWS_SPLINE) yading@11: sizeFactor = 20; // infinite ;) yading@11: else if (flags & SWS_BILINEAR) yading@11: sizeFactor = 2; yading@11: else { yading@11: av_assert0(0); yading@11: } yading@11: yading@11: if (xInc <= 1 << 16) yading@11: filterSize = 1 + sizeFactor; // upscale yading@11: else yading@11: filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW; yading@11: yading@11: filterSize = FFMIN(filterSize, srcW - 2); yading@11: filterSize = FFMAX(filterSize, 1); yading@11: yading@11: FF_ALLOC_OR_GOTO(NULL, filter, yading@11: dstW * sizeof(*filter) * filterSize, fail); yading@11: yading@11: xDstInSrc = xInc - 0x10000; yading@11: for (i = 0; i < dstW; i++) { yading@11: int xx = (xDstInSrc - ((filterSize - 2) << 16)) / (1 << 17); yading@11: int j; yading@11: (*filterPos)[i] = xx; yading@11: for (j = 0; j < filterSize; j++) { yading@11: int64_t d = (FFABS(((int64_t)xx << 17) - xDstInSrc)) << 13; yading@11: double floatd; yading@11: int64_t coeff; yading@11: yading@11: if (xInc > 1 << 16) yading@11: d = d * dstW / srcW; yading@11: floatd = d * (1.0 / (1 << 30)); yading@11: yading@11: if (flags & SWS_BICUBIC) { yading@11: int64_t B = (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1 << 24); yading@11: int64_t C = (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1 << 24); yading@11: yading@11: if (d >= 1LL << 31) { yading@11: coeff = 0.0; yading@11: } else { yading@11: int64_t dd = (d * d) >> 30; yading@11: int64_t ddd = (dd * d) >> 30; yading@11: yading@11: if (d < 1LL << 30) yading@11: coeff = (12 * (1 << 24) - 9 * B - 6 * C) * ddd + yading@11: (-18 * (1 << 24) + 12 * B + 6 * C) * dd + yading@11: (6 * (1 << 24) - 2 * B) * (1 << 30); yading@11: else yading@11: coeff = (-B - 6 * C) * ddd + yading@11: (6 * B + 30 * C) * dd + yading@11: (-12 * B - 48 * C) * d + yading@11: (8 * B + 24 * C) * (1 << 30); yading@11: } yading@11: coeff /= (1LL<<54)/fone; yading@11: } yading@11: #if 0 yading@11: else if (flags & SWS_X) { yading@11: double p = param ? param * 0.01 : 0.3; yading@11: coeff = d ? sin(d * M_PI) / (d * M_PI) : 1.0; yading@11: coeff *= pow(2.0, -p * d * d); yading@11: } yading@11: #endif yading@11: else if (flags & SWS_X) { yading@11: double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0; yading@11: double c; yading@11: yading@11: if (floatd < 1.0) yading@11: c = cos(floatd * M_PI); yading@11: else yading@11: c = -1.0; yading@11: if (c < 0.0) yading@11: c = -pow(-c, A); yading@11: else yading@11: c = pow(c, A); yading@11: coeff = (c * 0.5 + 0.5) * fone; yading@11: } else if (flags & SWS_AREA) { yading@11: int64_t d2 = d - (1 << 29); yading@11: if (d2 * xInc < -(1LL << (29 + 16))) yading@11: coeff = 1.0 * (1LL << (30 + 16)); yading@11: else if (d2 * xInc < (1LL << (29 + 16))) yading@11: coeff = -d2 * xInc + (1LL << (29 + 16)); yading@11: else yading@11: coeff = 0.0; yading@11: coeff *= fone >> (30 + 16); yading@11: } else if (flags & SWS_GAUSS) { yading@11: double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0; yading@11: coeff = (pow(2.0, -p * floatd * floatd)) * fone; yading@11: } else if (flags & SWS_SINC) { yading@11: coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone; yading@11: } else if (flags & SWS_LANCZOS) { yading@11: double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0; yading@11: coeff = (d ? sin(floatd * M_PI) * sin(floatd * M_PI / p) / yading@11: (floatd * floatd * M_PI * M_PI / p) : 1.0) * fone; yading@11: if (floatd > p) yading@11: coeff = 0; yading@11: } else if (flags & SWS_BILINEAR) { yading@11: coeff = (1 << 30) - d; yading@11: if (coeff < 0) yading@11: coeff = 0; yading@11: coeff *= fone >> 30; yading@11: } else if (flags & SWS_SPLINE) { yading@11: double p = -2.196152422706632; yading@11: coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone; yading@11: } else { yading@11: av_assert0(0); yading@11: } yading@11: yading@11: filter[i * filterSize + j] = coeff; yading@11: xx++; yading@11: } yading@11: xDstInSrc += 2 * xInc; yading@11: } yading@11: } yading@11: yading@11: /* apply src & dst Filter to filter -> filter2 yading@11: * av_free(filter); yading@11: */ yading@11: av_assert0(filterSize > 0); yading@11: filter2Size = filterSize; yading@11: if (srcFilter) yading@11: filter2Size += srcFilter->length - 1; yading@11: if (dstFilter) yading@11: filter2Size += dstFilter->length - 1; yading@11: av_assert0(filter2Size > 0); yading@11: FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size * dstW * sizeof(*filter2), fail); yading@11: yading@11: for (i = 0; i < dstW; i++) { yading@11: int j, k; yading@11: yading@11: if (srcFilter) { yading@11: for (k = 0; k < srcFilter->length; k++) { yading@11: for (j = 0; j < filterSize; j++) yading@11: filter2[i * filter2Size + k + j] += yading@11: srcFilter->coeff[k] * filter[i * filterSize + j]; yading@11: } yading@11: } else { yading@11: for (j = 0; j < filterSize; j++) yading@11: filter2[i * filter2Size + j] = filter[i * filterSize + j]; yading@11: } yading@11: // FIXME dstFilter yading@11: yading@11: (*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2; yading@11: } yading@11: av_freep(&filter); yading@11: yading@11: /* try to reduce the filter-size (step1 find size and shift left) */ yading@11: // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not). yading@11: minFilterSize = 0; yading@11: for (i = dstW - 1; i >= 0; i--) { yading@11: int min = filter2Size; yading@11: int j; yading@11: int64_t cutOff = 0.0; yading@11: yading@11: /* get rid of near zero elements on the left by shifting left */ yading@11: for (j = 0; j < filter2Size; j++) { yading@11: int k; yading@11: cutOff += FFABS(filter2[i * filter2Size]); yading@11: yading@11: if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone) yading@11: break; yading@11: yading@11: /* preserve monotonicity because the core can't handle the yading@11: * filter otherwise */ yading@11: if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1]) yading@11: break; yading@11: yading@11: // move filter coefficients left yading@11: for (k = 1; k < filter2Size; k++) yading@11: filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k]; yading@11: filter2[i * filter2Size + k - 1] = 0; yading@11: (*filterPos)[i]++; yading@11: } yading@11: yading@11: cutOff = 0; yading@11: /* count near zeros on the right */ yading@11: for (j = filter2Size - 1; j > 0; j--) { yading@11: cutOff += FFABS(filter2[i * filter2Size + j]); yading@11: yading@11: if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone) yading@11: break; yading@11: min--; yading@11: } yading@11: yading@11: if (min > minFilterSize) yading@11: minFilterSize = min; yading@11: } yading@11: yading@11: if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) { yading@11: // we can handle the special case 4, so we don't want to go the full 8 yading@11: if (minFilterSize < 5) yading@11: filterAlign = 4; yading@11: yading@11: /* We really don't want to waste our time doing useless computation, so yading@11: * fall back on the scalar C code for very small filters. yading@11: * Vectorizing is worth it only if you have a decent-sized vector. */ yading@11: if (minFilterSize < 3) yading@11: filterAlign = 1; yading@11: } yading@11: yading@11: if (INLINE_MMX(cpu_flags)) { yading@11: // special case for unscaled vertical filtering yading@11: if (minFilterSize == 1 && filterAlign == 2) yading@11: filterAlign = 1; yading@11: } yading@11: yading@11: av_assert0(minFilterSize > 0); yading@11: filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1)); yading@11: av_assert0(filterSize > 0); yading@11: filter = av_malloc(filterSize * dstW * sizeof(*filter)); yading@11: if (filterSize >= MAX_FILTER_SIZE * 16 / yading@11: ((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter) { yading@11: av_log(NULL, AV_LOG_ERROR, "sws: filterSize %d is too large, try less extreem scaling or increase MAX_FILTER_SIZE and recompile\n", filterSize); yading@11: goto fail; yading@11: } yading@11: *outFilterSize = filterSize; yading@11: yading@11: if (flags & SWS_PRINT_INFO) yading@11: av_log(NULL, AV_LOG_VERBOSE, yading@11: "SwScaler: reducing / aligning filtersize %d -> %d\n", yading@11: filter2Size, filterSize); yading@11: /* try to reduce the filter-size (step2 reduce it) */ yading@11: for (i = 0; i < dstW; i++) { yading@11: int j; yading@11: yading@11: for (j = 0; j < filterSize; j++) { yading@11: if (j >= filter2Size) yading@11: filter[i * filterSize + j] = 0; yading@11: else yading@11: filter[i * filterSize + j] = filter2[i * filter2Size + j]; yading@11: if ((flags & SWS_BITEXACT) && j >= minFilterSize) yading@11: filter[i * filterSize + j] = 0; yading@11: } yading@11: } yading@11: yading@11: // FIXME try to align filterPos if possible yading@11: yading@11: // fix borders yading@11: for (i = 0; i < dstW; i++) { yading@11: int j; yading@11: if ((*filterPos)[i] < 0) { yading@11: // move filter coefficients left to compensate for filterPos yading@11: for (j = 1; j < filterSize; j++) { yading@11: int left = FFMAX(j + (*filterPos)[i], 0); yading@11: filter[i * filterSize + left] += filter[i * filterSize + j]; yading@11: filter[i * filterSize + j] = 0; yading@11: } yading@11: (*filterPos)[i]= 0; yading@11: } yading@11: yading@11: if ((*filterPos)[i] + filterSize > srcW) { yading@11: int shift = (*filterPos)[i] + filterSize - srcW; yading@11: // move filter coefficients right to compensate for filterPos yading@11: for (j = filterSize - 2; j >= 0; j--) { yading@11: int right = FFMIN(j + shift, filterSize - 1); yading@11: filter[i * filterSize + right] += filter[i * filterSize + j]; yading@11: filter[i * filterSize + j] = 0; yading@11: } yading@11: (*filterPos)[i]= srcW - filterSize; yading@11: } yading@11: } yading@11: yading@11: // Note the +1 is for the MMX scaler which reads over the end yading@11: /* align at 16 for AltiVec (needed by hScale_altivec_real) */ yading@11: FF_ALLOCZ_OR_GOTO(NULL, *outFilter, yading@11: *outFilterSize * (dstW + 3) * sizeof(int16_t), fail); yading@11: yading@11: /* normalize & store in outFilter */ yading@11: for (i = 0; i < dstW; i++) { yading@11: int j; yading@11: int64_t error = 0; yading@11: int64_t sum = 0; yading@11: yading@11: for (j = 0; j < filterSize; j++) { yading@11: sum += filter[i * filterSize + j]; yading@11: } yading@11: sum = (sum + one / 2) / one; yading@11: for (j = 0; j < *outFilterSize; j++) { yading@11: int64_t v = filter[i * filterSize + j] + error; yading@11: int intV = ROUNDED_DIV(v, sum); yading@11: (*outFilter)[i * (*outFilterSize) + j] = intV; yading@11: error = v - intV * sum; yading@11: } yading@11: } yading@11: yading@11: (*filterPos)[dstW + 0] = yading@11: (*filterPos)[dstW + 1] = yading@11: (*filterPos)[dstW + 2] = (*filterPos)[dstW - 1]; /* the MMX/SSE scaler will yading@11: * read over the end */ yading@11: for (i = 0; i < *outFilterSize; i++) { yading@11: int k = (dstW - 1) * (*outFilterSize) + i; yading@11: (*outFilter)[k + 1 * (*outFilterSize)] = yading@11: (*outFilter)[k + 2 * (*outFilterSize)] = yading@11: (*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k]; yading@11: } yading@11: yading@11: ret = 0; yading@11: yading@11: fail: yading@11: if(ret < 0) yading@11: av_log(NULL, AV_LOG_ERROR, "sws: initFilter failed\n"); yading@11: av_free(filter); yading@11: av_free(filter2); yading@11: return ret; yading@11: } yading@11: yading@11: #if HAVE_MMXEXT_INLINE yading@11: static int init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode, yading@11: int16_t *filter, int32_t *filterPos, yading@11: int numSplits) yading@11: { yading@11: uint8_t *fragmentA; yading@11: x86_reg imm8OfPShufW1A; yading@11: x86_reg imm8OfPShufW2A; yading@11: x86_reg fragmentLengthA; yading@11: uint8_t *fragmentB; yading@11: x86_reg imm8OfPShufW1B; yading@11: x86_reg imm8OfPShufW2B; yading@11: x86_reg fragmentLengthB; yading@11: int fragmentPos; yading@11: yading@11: int xpos, i; yading@11: yading@11: // create an optimized horizontal scaling routine yading@11: /* This scaler is made of runtime-generated MMXEXT code using specially tuned yading@11: * pshufw instructions. For every four output pixels, if four input pixels yading@11: * are enough for the fast bilinear scaling, then a chunk of fragmentB is yading@11: * used. If five input pixels are needed, then a chunk of fragmentA is used. yading@11: */ yading@11: yading@11: // code fragment yading@11: yading@11: __asm__ volatile ( yading@11: "jmp 9f \n\t" yading@11: // Begin yading@11: "0: \n\t" yading@11: "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t" yading@11: "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t" yading@11: "movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t" yading@11: "punpcklbw %%mm7, %%mm1 \n\t" yading@11: "punpcklbw %%mm7, %%mm0 \n\t" yading@11: "pshufw $0xFF, %%mm1, %%mm1 \n\t" yading@11: "1: \n\t" yading@11: "pshufw $0xFF, %%mm0, %%mm0 \n\t" yading@11: "2: \n\t" yading@11: "psubw %%mm1, %%mm0 \n\t" yading@11: "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t" yading@11: "pmullw %%mm3, %%mm0 \n\t" yading@11: "psllw $7, %%mm1 \n\t" yading@11: "paddw %%mm1, %%mm0 \n\t" yading@11: yading@11: "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t" yading@11: yading@11: "add $8, %%"REG_a" \n\t" yading@11: // End yading@11: "9: \n\t" yading@11: // "int $3 \n\t" yading@11: "lea " LOCAL_MANGLE(0b) ", %0 \n\t" yading@11: "lea " LOCAL_MANGLE(1b) ", %1 \n\t" yading@11: "lea " LOCAL_MANGLE(2b) ", %2 \n\t" yading@11: "dec %1 \n\t" yading@11: "dec %2 \n\t" yading@11: "sub %0, %1 \n\t" yading@11: "sub %0, %2 \n\t" yading@11: "lea " LOCAL_MANGLE(9b) ", %3 \n\t" yading@11: "sub %0, %3 \n\t" yading@11: yading@11: yading@11: : "=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A), yading@11: "=r" (fragmentLengthA) yading@11: ); yading@11: yading@11: __asm__ volatile ( yading@11: "jmp 9f \n\t" yading@11: // Begin yading@11: "0: \n\t" yading@11: "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t" yading@11: "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t" yading@11: "punpcklbw %%mm7, %%mm0 \n\t" yading@11: "pshufw $0xFF, %%mm0, %%mm1 \n\t" yading@11: "1: \n\t" yading@11: "pshufw $0xFF, %%mm0, %%mm0 \n\t" yading@11: "2: \n\t" yading@11: "psubw %%mm1, %%mm0 \n\t" yading@11: "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t" yading@11: "pmullw %%mm3, %%mm0 \n\t" yading@11: "psllw $7, %%mm1 \n\t" yading@11: "paddw %%mm1, %%mm0 \n\t" yading@11: yading@11: "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t" yading@11: yading@11: "add $8, %%"REG_a" \n\t" yading@11: // End yading@11: "9: \n\t" yading@11: // "int $3 \n\t" yading@11: "lea " LOCAL_MANGLE(0b) ", %0 \n\t" yading@11: "lea " LOCAL_MANGLE(1b) ", %1 \n\t" yading@11: "lea " LOCAL_MANGLE(2b) ", %2 \n\t" yading@11: "dec %1 \n\t" yading@11: "dec %2 \n\t" yading@11: "sub %0, %1 \n\t" yading@11: "sub %0, %2 \n\t" yading@11: "lea " LOCAL_MANGLE(9b) ", %3 \n\t" yading@11: "sub %0, %3 \n\t" yading@11: yading@11: yading@11: : "=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B), yading@11: "=r" (fragmentLengthB) yading@11: ); yading@11: yading@11: xpos = 0; // lumXInc/2 - 0x8000; // difference between pixel centers yading@11: fragmentPos = 0; yading@11: yading@11: for (i = 0; i < dstW / numSplits; i++) { yading@11: int xx = xpos >> 16; yading@11: yading@11: if ((i & 3) == 0) { yading@11: int a = 0; yading@11: int b = ((xpos + xInc) >> 16) - xx; yading@11: int c = ((xpos + xInc * 2) >> 16) - xx; yading@11: int d = ((xpos + xInc * 3) >> 16) - xx; yading@11: int inc = (d + 1 < 4); yading@11: uint8_t *fragment = (d + 1 < 4) ? fragmentB : fragmentA; yading@11: x86_reg imm8OfPShufW1 = (d + 1 < 4) ? imm8OfPShufW1B : imm8OfPShufW1A; yading@11: x86_reg imm8OfPShufW2 = (d + 1 < 4) ? imm8OfPShufW2B : imm8OfPShufW2A; yading@11: x86_reg fragmentLength = (d + 1 < 4) ? fragmentLengthB : fragmentLengthA; yading@11: int maxShift = 3 - (d + inc); yading@11: int shift = 0; yading@11: yading@11: if (filterCode) { yading@11: filter[i] = ((xpos & 0xFFFF) ^ 0xFFFF) >> 9; yading@11: filter[i + 1] = (((xpos + xInc) & 0xFFFF) ^ 0xFFFF) >> 9; yading@11: filter[i + 2] = (((xpos + xInc * 2) & 0xFFFF) ^ 0xFFFF) >> 9; yading@11: filter[i + 3] = (((xpos + xInc * 3) & 0xFFFF) ^ 0xFFFF) >> 9; yading@11: filterPos[i / 2] = xx; yading@11: yading@11: memcpy(filterCode + fragmentPos, fragment, fragmentLength); yading@11: yading@11: filterCode[fragmentPos + imm8OfPShufW1] = (a + inc) | yading@11: ((b + inc) << 2) | yading@11: ((c + inc) << 4) | yading@11: ((d + inc) << 6); yading@11: filterCode[fragmentPos + imm8OfPShufW2] = a | (b << 2) | yading@11: (c << 4) | yading@11: (d << 6); yading@11: yading@11: if (i + 4 - inc >= dstW) yading@11: shift = maxShift; // avoid overread yading@11: else if ((filterPos[i / 2] & 3) <= maxShift) yading@11: shift = filterPos[i / 2] & 3; // align yading@11: yading@11: if (shift && i >= shift) { yading@11: filterCode[fragmentPos + imm8OfPShufW1] += 0x55 * shift; yading@11: filterCode[fragmentPos + imm8OfPShufW2] += 0x55 * shift; yading@11: filterPos[i / 2] -= shift; yading@11: } yading@11: } yading@11: yading@11: fragmentPos += fragmentLength; yading@11: yading@11: if (filterCode) yading@11: filterCode[fragmentPos] = RET; yading@11: } yading@11: xpos += xInc; yading@11: } yading@11: if (filterCode) yading@11: filterPos[((i / 2) + 1) & (~1)] = xpos >> 16; // needed to jump to the next part yading@11: yading@11: return fragmentPos + 1; yading@11: } yading@11: #endif /* HAVE_MMXEXT_INLINE */ yading@11: yading@11: static void getSubSampleFactors(int *h, int *v, enum AVPixelFormat format) yading@11: { yading@11: const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(format); yading@11: *h = desc->log2_chroma_w; yading@11: *v = desc->log2_chroma_h; yading@11: } yading@11: yading@11: static void fill_rgb2yuv_table(SwsContext *c, const int table[4], int dstRange) yading@11: { yading@11: int64_t W, V, Z, Cy, Cu, Cv; yading@11: int64_t vr = table[0]; yading@11: int64_t ub = table[1]; yading@11: int64_t ug = -table[2]; yading@11: int64_t vg = -table[3]; yading@11: int64_t ONE = 65536; yading@11: int64_t cy = ONE; yading@11: uint8_t *p = (uint8_t*)c->input_rgb2yuv_table; yading@11: int i; yading@11: static const int8_t map[] = { yading@11: BY_IDX, GY_IDX, -1 , BY_IDX, BY_IDX, GY_IDX, -1 , BY_IDX, yading@11: RY_IDX, -1 , GY_IDX, RY_IDX, RY_IDX, -1 , GY_IDX, RY_IDX, yading@11: RY_IDX, GY_IDX, -1 , RY_IDX, RY_IDX, GY_IDX, -1 , RY_IDX, yading@11: BY_IDX, -1 , GY_IDX, BY_IDX, BY_IDX, -1 , GY_IDX, BY_IDX, yading@11: BU_IDX, GU_IDX, -1 , BU_IDX, BU_IDX, GU_IDX, -1 , BU_IDX, yading@11: RU_IDX, -1 , GU_IDX, RU_IDX, RU_IDX, -1 , GU_IDX, RU_IDX, yading@11: RU_IDX, GU_IDX, -1 , RU_IDX, RU_IDX, GU_IDX, -1 , RU_IDX, yading@11: BU_IDX, -1 , GU_IDX, BU_IDX, BU_IDX, -1 , GU_IDX, BU_IDX, yading@11: BV_IDX, GV_IDX, -1 , BV_IDX, BV_IDX, GV_IDX, -1 , BV_IDX, yading@11: RV_IDX, -1 , GV_IDX, RV_IDX, RV_IDX, -1 , GV_IDX, RV_IDX, yading@11: RV_IDX, GV_IDX, -1 , RV_IDX, RV_IDX, GV_IDX, -1 , RV_IDX, yading@11: BV_IDX, -1 , GV_IDX, BV_IDX, BV_IDX, -1 , GV_IDX, BV_IDX, yading@11: RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, yading@11: BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, yading@11: GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , yading@11: -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, yading@11: RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, yading@11: BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, yading@11: GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , yading@11: -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, yading@11: RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, yading@11: BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, yading@11: GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , yading@11: -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, //23 yading@11: -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //24 yading@11: -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //25 yading@11: -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //26 yading@11: -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //27 yading@11: -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //28 yading@11: -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //29 yading@11: -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //30 yading@11: -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //31 yading@11: BY_IDX, GY_IDX, RY_IDX, -1 , -1 , -1 , -1 , -1 , //32 yading@11: BU_IDX, GU_IDX, RU_IDX, -1 , -1 , -1 , -1 , -1 , //33 yading@11: BV_IDX, GV_IDX, RV_IDX, -1 , -1 , -1 , -1 , -1 , //34 yading@11: }; yading@11: yading@11: dstRange = 0; //FIXME range = 1 is handled elsewhere yading@11: yading@11: if (!dstRange) { yading@11: cy = cy * 255 / 219; yading@11: } else { yading@11: vr = vr * 224 / 255; yading@11: ub = ub * 224 / 255; yading@11: ug = ug * 224 / 255; yading@11: vg = vg * 224 / 255; yading@11: } yading@11: W = ROUNDED_DIV(ONE*ONE*ug, ub); yading@11: V = ROUNDED_DIV(ONE*ONE*vg, vr); yading@11: Z = ONE*ONE-W-V; yading@11: yading@11: Cy = ROUNDED_DIV(cy*Z, ONE); yading@11: Cu = ROUNDED_DIV(ub*Z, ONE); yading@11: Cv = ROUNDED_DIV(vr*Z, ONE); yading@11: yading@11: c->input_rgb2yuv_table[RY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cy); yading@11: c->input_rgb2yuv_table[GY_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cy); yading@11: c->input_rgb2yuv_table[BY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cy); yading@11: yading@11: c->input_rgb2yuv_table[RU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cu); yading@11: c->input_rgb2yuv_table[GU_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cu); yading@11: c->input_rgb2yuv_table[BU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(Z+W) , Cu); yading@11: yading@11: c->input_rgb2yuv_table[RV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(V+Z) , Cv); yading@11: c->input_rgb2yuv_table[GV_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cv); yading@11: c->input_rgb2yuv_table[BV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cv); yading@11: yading@11: if(/*!dstRange && */table == ff_yuv2rgb_coeffs[SWS_CS_DEFAULT]) { yading@11: c->input_rgb2yuv_table[BY_IDX] = ((int)(0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: c->input_rgb2yuv_table[BV_IDX] = (-(int)(0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: c->input_rgb2yuv_table[BU_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: c->input_rgb2yuv_table[GY_IDX] = ((int)(0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: c->input_rgb2yuv_table[GV_IDX] = (-(int)(0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: c->input_rgb2yuv_table[GU_IDX] = (-(int)(0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: c->input_rgb2yuv_table[RY_IDX] = ((int)(0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: c->input_rgb2yuv_table[RV_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: c->input_rgb2yuv_table[RU_IDX] = (-(int)(0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5)); yading@11: } yading@11: for(i=0; i= 0 ? c->input_rgb2yuv_table[map[i]] : 0); yading@11: } yading@11: yading@11: int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4], yading@11: int srcRange, const int table[4], int dstRange, yading@11: int brightness, int contrast, int saturation) yading@11: { yading@11: const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(c->dstFormat); yading@11: const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(c->srcFormat); yading@11: memcpy(c->srcColorspaceTable, inv_table, sizeof(int) * 4); yading@11: memcpy(c->dstColorspaceTable, table, sizeof(int) * 4); yading@11: yading@11: if(!isYUV(c->dstFormat) && !isGray(c->dstFormat)) yading@11: dstRange = 0; yading@11: if(!isYUV(c->srcFormat) && !isGray(c->srcFormat)) yading@11: srcRange = 0; yading@11: yading@11: c->brightness = brightness; yading@11: c->contrast = contrast; yading@11: c->saturation = saturation; yading@11: c->srcRange = srcRange; yading@11: c->dstRange = dstRange; yading@11: yading@11: if ((isYUV(c->dstFormat) || isGray(c->dstFormat)) && (isYUV(c->srcFormat) || isGray(c->srcFormat))) yading@11: return -1; yading@11: yading@11: c->dstFormatBpp = av_get_bits_per_pixel(desc_dst); yading@11: c->srcFormatBpp = av_get_bits_per_pixel(desc_src); yading@11: yading@11: if (!isYUV(c->dstFormat) && !isGray(c->dstFormat)) { yading@11: ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, yading@11: contrast, saturation); yading@11: // FIXME factorize yading@11: yading@11: if (HAVE_ALTIVEC && av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) yading@11: ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, yading@11: contrast, saturation); yading@11: } yading@11: yading@11: fill_rgb2yuv_table(c, table, dstRange); yading@11: yading@11: return 0; yading@11: } yading@11: yading@11: int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table, yading@11: int *srcRange, int **table, int *dstRange, yading@11: int *brightness, int *contrast, int *saturation) yading@11: { yading@11: if (!c || isYUV(c->dstFormat) || isGray(c->dstFormat)) yading@11: return -1; yading@11: yading@11: *inv_table = c->srcColorspaceTable; yading@11: *table = c->dstColorspaceTable; yading@11: *srcRange = c->srcRange; yading@11: *dstRange = c->dstRange; yading@11: *brightness = c->brightness; yading@11: *contrast = c->contrast; yading@11: *saturation = c->saturation; yading@11: yading@11: return 0; yading@11: } yading@11: yading@11: static int handle_jpeg(enum AVPixelFormat *format) yading@11: { yading@11: switch (*format) { yading@11: case AV_PIX_FMT_YUVJ420P: yading@11: *format = AV_PIX_FMT_YUV420P; yading@11: return 1; yading@11: case AV_PIX_FMT_YUVJ422P: yading@11: *format = AV_PIX_FMT_YUV422P; yading@11: return 1; yading@11: case AV_PIX_FMT_YUVJ444P: yading@11: *format = AV_PIX_FMT_YUV444P; yading@11: return 1; yading@11: case AV_PIX_FMT_YUVJ440P: yading@11: *format = AV_PIX_FMT_YUV440P; yading@11: return 1; yading@11: default: yading@11: return 0; yading@11: } yading@11: } yading@11: yading@11: static int handle_0alpha(enum AVPixelFormat *format) yading@11: { yading@11: switch (*format) { yading@11: case AV_PIX_FMT_0BGR : *format = AV_PIX_FMT_ABGR ; return 1; yading@11: case AV_PIX_FMT_BGR0 : *format = AV_PIX_FMT_BGRA ; return 4; yading@11: case AV_PIX_FMT_0RGB : *format = AV_PIX_FMT_ARGB ; return 1; yading@11: case AV_PIX_FMT_RGB0 : *format = AV_PIX_FMT_RGBA ; return 4; yading@11: default: return 0; yading@11: } yading@11: } yading@11: yading@11: SwsContext *sws_alloc_context(void) yading@11: { yading@11: SwsContext *c = av_mallocz(sizeof(SwsContext)); yading@11: yading@11: if (c) { yading@11: c->av_class = &sws_context_class; yading@11: av_opt_set_defaults(c); yading@11: } yading@11: yading@11: return c; yading@11: } yading@11: yading@11: av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter, yading@11: SwsFilter *dstFilter) yading@11: { yading@11: int i, j; yading@11: int usesVFilter, usesHFilter; yading@11: int unscaled; yading@11: SwsFilter dummyFilter = { NULL, NULL, NULL, NULL }; yading@11: int srcW = c->srcW; yading@11: int srcH = c->srcH; yading@11: int dstW = c->dstW; yading@11: int dstH = c->dstH; yading@11: int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16); yading@11: int flags, cpu_flags; yading@11: enum AVPixelFormat srcFormat = c->srcFormat; yading@11: enum AVPixelFormat dstFormat = c->dstFormat; yading@11: const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(srcFormat); yading@11: const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(dstFormat); yading@11: yading@11: cpu_flags = av_get_cpu_flags(); yading@11: flags = c->flags; yading@11: emms_c(); yading@11: if (!rgb15to16) yading@11: sws_rgb2rgb_init(); yading@11: yading@11: unscaled = (srcW == dstW && srcH == dstH); yading@11: yading@11: handle_jpeg(&srcFormat); yading@11: handle_jpeg(&dstFormat); yading@11: handle_0alpha(&srcFormat); yading@11: handle_0alpha(&dstFormat); yading@11: yading@11: if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat){ yading@11: av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n"); yading@11: c->srcFormat= srcFormat; yading@11: c->dstFormat= dstFormat; yading@11: } yading@11: yading@11: if (!sws_isSupportedInput(srcFormat)) { yading@11: av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n", yading@11: av_get_pix_fmt_name(srcFormat)); yading@11: return AVERROR(EINVAL); yading@11: } yading@11: if (!sws_isSupportedOutput(dstFormat)) { yading@11: av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n", yading@11: av_get_pix_fmt_name(dstFormat)); yading@11: return AVERROR(EINVAL); yading@11: } yading@11: yading@11: i = flags & (SWS_POINT | yading@11: SWS_AREA | yading@11: SWS_BILINEAR | yading@11: SWS_FAST_BILINEAR | yading@11: SWS_BICUBIC | yading@11: SWS_X | yading@11: SWS_GAUSS | yading@11: SWS_LANCZOS | yading@11: SWS_SINC | yading@11: SWS_SPLINE | yading@11: SWS_BICUBLIN); yading@11: if (!i || (i & (i - 1))) { yading@11: av_log(c, AV_LOG_ERROR, "Exactly one scaler algorithm must be chosen, got %X\n", i); yading@11: return AVERROR(EINVAL); yading@11: } yading@11: /* sanity check */ yading@11: if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) { yading@11: /* FIXME check if these are enough and try to lower them after yading@11: * fixing the relevant parts of the code */ yading@11: av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n", yading@11: srcW, srcH, dstW, dstH); yading@11: return AVERROR(EINVAL); yading@11: } yading@11: yading@11: if (!dstFilter) yading@11: dstFilter = &dummyFilter; yading@11: if (!srcFilter) yading@11: srcFilter = &dummyFilter; yading@11: yading@11: c->lumXInc = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW; yading@11: c->lumYInc = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH; yading@11: c->dstFormatBpp = av_get_bits_per_pixel(desc_dst); yading@11: c->srcFormatBpp = av_get_bits_per_pixel(desc_src); yading@11: c->vRounder = 4 * 0x0001000100010001ULL; yading@11: yading@11: usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) || yading@11: (srcFilter->chrV && srcFilter->chrV->length > 1) || yading@11: (dstFilter->lumV && dstFilter->lumV->length > 1) || yading@11: (dstFilter->chrV && dstFilter->chrV->length > 1); yading@11: usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) || yading@11: (srcFilter->chrH && srcFilter->chrH->length > 1) || yading@11: (dstFilter->lumH && dstFilter->lumH->length > 1) || yading@11: (dstFilter->chrH && dstFilter->chrH->length > 1); yading@11: yading@11: getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat); yading@11: getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat); yading@11: yading@11: if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) { yading@11: if (dstW&1) { yading@11: av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n"); yading@11: flags |= SWS_FULL_CHR_H_INT; yading@11: c->flags = flags; yading@11: } yading@11: } yading@11: yading@11: if(dstFormat == AV_PIX_FMT_BGR4_BYTE || yading@11: dstFormat == AV_PIX_FMT_RGB4_BYTE || yading@11: dstFormat == AV_PIX_FMT_BGR8 || yading@11: dstFormat == AV_PIX_FMT_RGB8) { yading@11: if (flags & SWS_ERROR_DIFFUSION && !(flags & SWS_FULL_CHR_H_INT)) { yading@11: av_log(c, AV_LOG_DEBUG, yading@11: "Error diffusion dither is only supported in full chroma interpolation for destination format '%s'\n", yading@11: av_get_pix_fmt_name(dstFormat)); yading@11: flags |= SWS_FULL_CHR_H_INT; yading@11: c->flags = flags; yading@11: } yading@11: if (!(flags & SWS_ERROR_DIFFUSION) && (flags & SWS_FULL_CHR_H_INT)) { yading@11: av_log(c, AV_LOG_DEBUG, yading@11: "Ordered dither is not supported in full chroma interpolation for destination format '%s'\n", yading@11: av_get_pix_fmt_name(dstFormat)); yading@11: flags |= SWS_ERROR_DIFFUSION; yading@11: c->flags = flags; yading@11: } yading@11: } yading@11: if (isPlanarRGB(dstFormat)) { yading@11: if (!(flags & SWS_FULL_CHR_H_INT)) { yading@11: av_log(c, AV_LOG_DEBUG, yading@11: "%s output is not supported with half chroma resolution, switching to full\n", yading@11: av_get_pix_fmt_name(dstFormat)); yading@11: flags |= SWS_FULL_CHR_H_INT; yading@11: c->flags = flags; yading@11: } yading@11: } yading@11: yading@11: /* reuse chroma for 2 pixels RGB/BGR unless user wants full yading@11: * chroma interpolation */ yading@11: if (flags & SWS_FULL_CHR_H_INT && yading@11: isAnyRGB(dstFormat) && yading@11: !isPlanarRGB(dstFormat) && yading@11: dstFormat != AV_PIX_FMT_RGBA && yading@11: dstFormat != AV_PIX_FMT_ARGB && yading@11: dstFormat != AV_PIX_FMT_BGRA && yading@11: dstFormat != AV_PIX_FMT_ABGR && yading@11: dstFormat != AV_PIX_FMT_RGB24 && yading@11: dstFormat != AV_PIX_FMT_BGR24 && yading@11: dstFormat != AV_PIX_FMT_BGR4_BYTE && yading@11: dstFormat != AV_PIX_FMT_RGB4_BYTE && yading@11: dstFormat != AV_PIX_FMT_BGR8 && yading@11: dstFormat != AV_PIX_FMT_RGB8 yading@11: ) { yading@11: av_log(c, AV_LOG_WARNING, yading@11: "full chroma interpolation for destination format '%s' not yet implemented\n", yading@11: av_get_pix_fmt_name(dstFormat)); yading@11: flags &= ~SWS_FULL_CHR_H_INT; yading@11: c->flags = flags; yading@11: } yading@11: if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT)) yading@11: c->chrDstHSubSample = 1; yading@11: yading@11: // drop some chroma lines if the user wants it yading@11: c->vChrDrop = (flags & SWS_SRC_V_CHR_DROP_MASK) >> yading@11: SWS_SRC_V_CHR_DROP_SHIFT; yading@11: c->chrSrcVSubSample += c->vChrDrop; yading@11: yading@11: /* drop every other pixel for chroma calculation unless user yading@11: * wants full chroma */ yading@11: if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP) && yading@11: srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 && yading@11: srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 && yading@11: srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE && yading@11: srcFormat != AV_PIX_FMT_GBRP9BE && srcFormat != AV_PIX_FMT_GBRP9LE && yading@11: srcFormat != AV_PIX_FMT_GBRP10BE && srcFormat != AV_PIX_FMT_GBRP10LE && yading@11: srcFormat != AV_PIX_FMT_GBRP12BE && srcFormat != AV_PIX_FMT_GBRP12LE && yading@11: srcFormat != AV_PIX_FMT_GBRP14BE && srcFormat != AV_PIX_FMT_GBRP14LE && yading@11: srcFormat != AV_PIX_FMT_GBRP16BE && srcFormat != AV_PIX_FMT_GBRP16LE && yading@11: ((dstW >> c->chrDstHSubSample) <= (srcW >> 1) || yading@11: (flags & SWS_FAST_BILINEAR))) yading@11: c->chrSrcHSubSample = 1; yading@11: yading@11: // Note the -((-x)>>y) is so that we always round toward +inf. yading@11: c->chrSrcW = -((-srcW) >> c->chrSrcHSubSample); yading@11: c->chrSrcH = -((-srcH) >> c->chrSrcVSubSample); yading@11: c->chrDstW = -((-dstW) >> c->chrDstHSubSample); yading@11: c->chrDstH = -((-dstH) >> c->chrDstVSubSample); yading@11: yading@11: FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail); yading@11: yading@11: /* unscaled special cases */ yading@11: if (unscaled && !usesHFilter && !usesVFilter && yading@11: (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) { yading@11: ff_get_unscaled_swscale(c); yading@11: yading@11: if (c->swScale) { yading@11: if (flags & SWS_PRINT_INFO) yading@11: av_log(c, AV_LOG_INFO, yading@11: "using unscaled %s -> %s special converter\n", yading@11: av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat)); yading@11: return 0; yading@11: } yading@11: } yading@11: yading@11: c->srcBpc = 1 + desc_src->comp[0].depth_minus1; yading@11: if (c->srcBpc < 8) yading@11: c->srcBpc = 8; yading@11: c->dstBpc = 1 + desc_dst->comp[0].depth_minus1; yading@11: if (c->dstBpc < 8) yading@11: c->dstBpc = 8; yading@11: if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8) yading@11: c->srcBpc = 16; yading@11: if (c->dstBpc == 16) yading@11: dst_stride <<= 1; yading@11: yading@11: if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) { yading@11: c->canMMXEXTBeUsed = (dstW >= srcW && (dstW & 31) == 0 && yading@11: (srcW & 15) == 0) ? 1 : 0; yading@11: if (!c->canMMXEXTBeUsed && dstW >= srcW && (srcW & 15) == 0 yading@11: yading@11: && (flags & SWS_FAST_BILINEAR)) { yading@11: if (flags & SWS_PRINT_INFO) yading@11: av_log(c, AV_LOG_INFO, yading@11: "output width is not a multiple of 32 -> no MMXEXT scaler\n"); yading@11: } yading@11: if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat)) yading@11: c->canMMXEXTBeUsed = 0; yading@11: } else yading@11: c->canMMXEXTBeUsed = 0; yading@11: yading@11: c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW; yading@11: c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH; yading@11: yading@11: /* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src yading@11: * to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do yading@11: * correct scaling. yading@11: * n-2 is the last chrominance sample available. yading@11: * This is not perfect, but no one should notice the difference, the more yading@11: * correct variant would be like the vertical one, but that would require yading@11: * some special code for the first and last pixel */ yading@11: if (flags & SWS_FAST_BILINEAR) { yading@11: if (c->canMMXEXTBeUsed) { yading@11: c->lumXInc += 20; yading@11: c->chrXInc += 20; yading@11: } yading@11: // we don't use the x86 asm scaler if MMX is available yading@11: else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) { yading@11: c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20; yading@11: c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20; yading@11: } yading@11: } yading@11: yading@11: #define USE_MMAP (HAVE_MMAP && HAVE_MPROTECT && defined MAP_ANONYMOUS) yading@11: yading@11: /* precalculate horizontal scaler filter coefficients */ yading@11: { yading@11: #if HAVE_MMXEXT_INLINE yading@11: // can't downscale !!! yading@11: if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) { yading@11: c->lumMmxextFilterCodeSize = init_hscaler_mmxext(dstW, c->lumXInc, NULL, yading@11: NULL, NULL, 8); yading@11: c->chrMmxextFilterCodeSize = init_hscaler_mmxext(c->chrDstW, c->chrXInc, yading@11: NULL, NULL, NULL, 4); yading@11: yading@11: #if USE_MMAP yading@11: c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize, yading@11: PROT_READ | PROT_WRITE, yading@11: MAP_PRIVATE | MAP_ANONYMOUS, yading@11: -1, 0); yading@11: c->chrMmxextFilterCode = mmap(NULL, c->chrMmxextFilterCodeSize, yading@11: PROT_READ | PROT_WRITE, yading@11: MAP_PRIVATE | MAP_ANONYMOUS, yading@11: -1, 0); yading@11: #elif HAVE_VIRTUALALLOC yading@11: c->lumMmxextFilterCode = VirtualAlloc(NULL, yading@11: c->lumMmxextFilterCodeSize, yading@11: MEM_COMMIT, yading@11: PAGE_EXECUTE_READWRITE); yading@11: c->chrMmxextFilterCode = VirtualAlloc(NULL, yading@11: c->chrMmxextFilterCodeSize, yading@11: MEM_COMMIT, yading@11: PAGE_EXECUTE_READWRITE); yading@11: #else yading@11: c->lumMmxextFilterCode = av_malloc(c->lumMmxextFilterCodeSize); yading@11: c->chrMmxextFilterCode = av_malloc(c->chrMmxextFilterCodeSize); yading@11: #endif yading@11: yading@11: #ifdef MAP_ANONYMOUS yading@11: if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED) yading@11: #else yading@11: if (!c->lumMmxextFilterCode || !c->chrMmxextFilterCode) yading@11: #endif yading@11: { yading@11: av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n"); yading@11: return AVERROR(ENOMEM); yading@11: } yading@11: yading@11: FF_ALLOCZ_OR_GOTO(c, c->hLumFilter, (dstW / 8 + 8) * sizeof(int16_t), fail); yading@11: FF_ALLOCZ_OR_GOTO(c, c->hChrFilter, (c->chrDstW / 4 + 8) * sizeof(int16_t), fail); yading@11: FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW / 2 / 8 + 8) * sizeof(int32_t), fail); yading@11: FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail); yading@11: yading@11: init_hscaler_mmxext( dstW, c->lumXInc, c->lumMmxextFilterCode, yading@11: c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8); yading@11: init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode, yading@11: c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4); yading@11: yading@11: #if USE_MMAP yading@11: mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ); yading@11: mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ); yading@11: #endif yading@11: } else yading@11: #endif /* HAVE_MMXEXT_INLINE */ yading@11: { yading@11: const int filterAlign = yading@11: (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 : yading@11: (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 : yading@11: 1; yading@11: yading@11: if (initFilter(&c->hLumFilter, &c->hLumFilterPos, yading@11: &c->hLumFilterSize, c->lumXInc, yading@11: srcW, dstW, filterAlign, 1 << 14, yading@11: (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags, yading@11: cpu_flags, srcFilter->lumH, dstFilter->lumH, yading@11: c->param) < 0) yading@11: goto fail; yading@11: if (initFilter(&c->hChrFilter, &c->hChrFilterPos, yading@11: &c->hChrFilterSize, c->chrXInc, yading@11: c->chrSrcW, c->chrDstW, filterAlign, 1 << 14, yading@11: (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags, yading@11: cpu_flags, srcFilter->chrH, dstFilter->chrH, yading@11: c->param) < 0) yading@11: goto fail; yading@11: } yading@11: } // initialize horizontal stuff yading@11: yading@11: /* precalculate vertical scaler filter coefficients */ yading@11: { yading@11: const int filterAlign = yading@11: (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 2 : yading@11: (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 : yading@11: 1; yading@11: yading@11: if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, yading@11: c->lumYInc, srcH, dstH, filterAlign, (1 << 12), yading@11: (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags, yading@11: cpu_flags, srcFilter->lumV, dstFilter->lumV, yading@11: c->param) < 0) yading@11: goto fail; yading@11: if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, yading@11: c->chrYInc, c->chrSrcH, c->chrDstH, yading@11: filterAlign, (1 << 12), yading@11: (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags, yading@11: cpu_flags, srcFilter->chrV, dstFilter->chrV, yading@11: c->param) < 0) yading@11: goto fail; yading@11: yading@11: #if HAVE_ALTIVEC yading@11: FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof(vector signed short) * c->vLumFilterSize * c->dstH, fail); yading@11: FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof(vector signed short) * c->vChrFilterSize * c->chrDstH, fail); yading@11: yading@11: for (i = 0; i < c->vLumFilterSize * c->dstH; i++) { yading@11: int j; yading@11: short *p = (short *)&c->vYCoeffsBank[i]; yading@11: for (j = 0; j < 8; j++) yading@11: p[j] = c->vLumFilter[i]; yading@11: } yading@11: yading@11: for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) { yading@11: int j; yading@11: short *p = (short *)&c->vCCoeffsBank[i]; yading@11: for (j = 0; j < 8; j++) yading@11: p[j] = c->vChrFilter[i]; yading@11: } yading@11: #endif yading@11: } yading@11: yading@11: // calculate buffer sizes so that they won't run out while handling these damn slices yading@11: c->vLumBufSize = c->vLumFilterSize; yading@11: c->vChrBufSize = c->vChrFilterSize; yading@11: for (i = 0; i < dstH; i++) { yading@11: int chrI = (int64_t)i * c->chrDstH / dstH; yading@11: int nextSlice = FFMAX(c->vLumFilterPos[i] + c->vLumFilterSize - 1, yading@11: ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1) yading@11: << c->chrSrcVSubSample)); yading@11: yading@11: nextSlice >>= c->chrSrcVSubSample; yading@11: nextSlice <<= c->chrSrcVSubSample; yading@11: if (c->vLumFilterPos[i] + c->vLumBufSize < nextSlice) yading@11: c->vLumBufSize = nextSlice - c->vLumFilterPos[i]; yading@11: if (c->vChrFilterPos[chrI] + c->vChrBufSize < yading@11: (nextSlice >> c->chrSrcVSubSample)) yading@11: c->vChrBufSize = (nextSlice >> c->chrSrcVSubSample) - yading@11: c->vChrFilterPos[chrI]; yading@11: } yading@11: yading@11: for (i = 0; i < 4; i++) yading@11: FF_ALLOCZ_OR_GOTO(c, c->dither_error[i], (c->dstW+2) * sizeof(int), fail); yading@11: yading@11: /* Allocate pixbufs (we use dynamic allocation because otherwise we would yading@11: * need to allocate several megabytes to handle all possible cases) */ yading@11: FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail); yading@11: FF_ALLOC_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail); yading@11: FF_ALLOC_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail); yading@11: if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) yading@11: FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail); yading@11: /* Note we need at least one pixel more at the end because of the MMX code yading@11: * (just in case someone wants to replace the 4000/8000). */ yading@11: /* align at 16 bytes for AltiVec */ yading@11: for (i = 0; i < c->vLumBufSize; i++) { yading@11: FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i + c->vLumBufSize], yading@11: dst_stride + 16, fail); yading@11: c->lumPixBuf[i] = c->lumPixBuf[i + c->vLumBufSize]; yading@11: } yading@11: // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate) yading@11: c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7); yading@11: c->uv_offx2 = dst_stride + 16; yading@11: for (i = 0; i < c->vChrBufSize; i++) { yading@11: FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i + c->vChrBufSize], yading@11: dst_stride * 2 + 32, fail); yading@11: c->chrUPixBuf[i] = c->chrUPixBuf[i + c->vChrBufSize]; yading@11: c->chrVPixBuf[i] = c->chrVPixBuf[i + c->vChrBufSize] yading@11: = c->chrUPixBuf[i] + (dst_stride >> 1) + 8; yading@11: } yading@11: if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) yading@11: for (i = 0; i < c->vLumBufSize; i++) { yading@11: FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i + c->vLumBufSize], yading@11: dst_stride + 16, fail); yading@11: c->alpPixBuf[i] = c->alpPixBuf[i + c->vLumBufSize]; yading@11: } yading@11: yading@11: // try to avoid drawing green stuff between the right end and the stride end yading@11: for (i = 0; i < c->vChrBufSize; i++) yading@11: if(desc_dst->comp[0].depth_minus1 == 15){ yading@11: av_assert0(c->dstBpc > 14); yading@11: for(j=0; jchrUPixBuf[i]))[j] = 1<<18; yading@11: } else yading@11: for(j=0; jchrUPixBuf[i]))[j] = 1<<14; yading@11: yading@11: av_assert0(c->chrDstH <= dstH); yading@11: yading@11: if (flags & SWS_PRINT_INFO) { yading@11: if (flags & SWS_FAST_BILINEAR) yading@11: av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, "); yading@11: else if (flags & SWS_BILINEAR) yading@11: av_log(c, AV_LOG_INFO, "BILINEAR scaler, "); yading@11: else if (flags & SWS_BICUBIC) yading@11: av_log(c, AV_LOG_INFO, "BICUBIC scaler, "); yading@11: else if (flags & SWS_X) yading@11: av_log(c, AV_LOG_INFO, "Experimental scaler, "); yading@11: else if (flags & SWS_POINT) yading@11: av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, "); yading@11: else if (flags & SWS_AREA) yading@11: av_log(c, AV_LOG_INFO, "Area Averaging scaler, "); yading@11: else if (flags & SWS_BICUBLIN) yading@11: av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, "); yading@11: else if (flags & SWS_GAUSS) yading@11: av_log(c, AV_LOG_INFO, "Gaussian scaler, "); yading@11: else if (flags & SWS_SINC) yading@11: av_log(c, AV_LOG_INFO, "Sinc scaler, "); yading@11: else if (flags & SWS_LANCZOS) yading@11: av_log(c, AV_LOG_INFO, "Lanczos scaler, "); yading@11: else if (flags & SWS_SPLINE) yading@11: av_log(c, AV_LOG_INFO, "Bicubic spline scaler, "); yading@11: else yading@11: av_log(c, AV_LOG_INFO, "ehh flags invalid?! "); yading@11: yading@11: av_log(c, AV_LOG_INFO, "from %s to %s%s ", yading@11: av_get_pix_fmt_name(srcFormat), yading@11: #ifdef DITHER1XBPP yading@11: dstFormat == AV_PIX_FMT_BGR555 || dstFormat == AV_PIX_FMT_BGR565 || yading@11: dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE || yading@11: dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ? yading@11: "dithered " : "", yading@11: #else yading@11: "", yading@11: #endif yading@11: av_get_pix_fmt_name(dstFormat)); yading@11: yading@11: if (INLINE_MMXEXT(cpu_flags)) yading@11: av_log(c, AV_LOG_INFO, "using MMXEXT\n"); yading@11: else if (INLINE_AMD3DNOW(cpu_flags)) yading@11: av_log(c, AV_LOG_INFO, "using 3DNOW\n"); yading@11: else if (INLINE_MMX(cpu_flags)) yading@11: av_log(c, AV_LOG_INFO, "using MMX\n"); yading@11: else if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) yading@11: av_log(c, AV_LOG_INFO, "using AltiVec\n"); yading@11: else yading@11: av_log(c, AV_LOG_INFO, "using C\n"); yading@11: yading@11: av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH); yading@11: av_log(c, AV_LOG_DEBUG, yading@11: "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", yading@11: c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc); yading@11: av_log(c, AV_LOG_DEBUG, yading@11: "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", yading@11: c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, yading@11: c->chrXInc, c->chrYInc); yading@11: } yading@11: yading@11: c->swScale = ff_getSwsFunc(c); yading@11: return 0; yading@11: fail: // FIXME replace things by appropriate error codes yading@11: return -1; yading@11: } yading@11: yading@11: #if FF_API_SWS_GETCONTEXT yading@11: SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat, yading@11: int dstW, int dstH, enum AVPixelFormat dstFormat, yading@11: int flags, SwsFilter *srcFilter, yading@11: SwsFilter *dstFilter, const double *param) yading@11: { yading@11: SwsContext *c; yading@11: yading@11: if (!(c = sws_alloc_context())) yading@11: return NULL; yading@11: yading@11: c->flags = flags; yading@11: c->srcW = srcW; yading@11: c->srcH = srcH; yading@11: c->dstW = dstW; yading@11: c->dstH = dstH; yading@11: c->srcRange = handle_jpeg(&srcFormat); yading@11: c->dstRange = handle_jpeg(&dstFormat); yading@11: c->src0Alpha = handle_0alpha(&srcFormat); yading@11: c->dst0Alpha = handle_0alpha(&dstFormat); yading@11: c->srcFormat = srcFormat; yading@11: c->dstFormat = dstFormat; yading@11: yading@11: if (param) { yading@11: c->param[0] = param[0]; yading@11: c->param[1] = param[1]; yading@11: } yading@11: sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange, yading@11: ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, yading@11: c->dstRange, 0, 1 << 16, 1 << 16); yading@11: yading@11: if (sws_init_context(c, srcFilter, dstFilter) < 0) { yading@11: sws_freeContext(c); yading@11: return NULL; yading@11: } yading@11: yading@11: return c; yading@11: } yading@11: #endif yading@11: yading@11: SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur, yading@11: float lumaSharpen, float chromaSharpen, yading@11: float chromaHShift, float chromaVShift, yading@11: int verbose) yading@11: { yading@11: SwsFilter *filter = av_malloc(sizeof(SwsFilter)); yading@11: if (!filter) yading@11: return NULL; yading@11: yading@11: if (lumaGBlur != 0.0) { yading@11: filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0); yading@11: filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0); yading@11: } else { yading@11: filter->lumH = sws_getIdentityVec(); yading@11: filter->lumV = sws_getIdentityVec(); yading@11: } yading@11: yading@11: if (chromaGBlur != 0.0) { yading@11: filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0); yading@11: filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0); yading@11: } else { yading@11: filter->chrH = sws_getIdentityVec(); yading@11: filter->chrV = sws_getIdentityVec(); yading@11: } yading@11: yading@11: if (chromaSharpen != 0.0) { yading@11: SwsVector *id = sws_getIdentityVec(); yading@11: sws_scaleVec(filter->chrH, -chromaSharpen); yading@11: sws_scaleVec(filter->chrV, -chromaSharpen); yading@11: sws_addVec(filter->chrH, id); yading@11: sws_addVec(filter->chrV, id); yading@11: sws_freeVec(id); yading@11: } yading@11: yading@11: if (lumaSharpen != 0.0) { yading@11: SwsVector *id = sws_getIdentityVec(); yading@11: sws_scaleVec(filter->lumH, -lumaSharpen); yading@11: sws_scaleVec(filter->lumV, -lumaSharpen); yading@11: sws_addVec(filter->lumH, id); yading@11: sws_addVec(filter->lumV, id); yading@11: sws_freeVec(id); yading@11: } yading@11: yading@11: if (chromaHShift != 0.0) yading@11: sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5)); yading@11: yading@11: if (chromaVShift != 0.0) yading@11: sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5)); yading@11: yading@11: sws_normalizeVec(filter->chrH, 1.0); yading@11: sws_normalizeVec(filter->chrV, 1.0); yading@11: sws_normalizeVec(filter->lumH, 1.0); yading@11: sws_normalizeVec(filter->lumV, 1.0); yading@11: yading@11: if (verbose) yading@11: sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG); yading@11: if (verbose) yading@11: sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG); yading@11: yading@11: return filter; yading@11: } yading@11: yading@11: SwsVector *sws_allocVec(int length) yading@11: { yading@11: SwsVector *vec; yading@11: yading@11: if(length <= 0 || length > INT_MAX/ sizeof(double)) yading@11: return NULL; yading@11: yading@11: vec = av_malloc(sizeof(SwsVector)); yading@11: if (!vec) yading@11: return NULL; yading@11: vec->length = length; yading@11: vec->coeff = av_malloc(sizeof(double) * length); yading@11: if (!vec->coeff) yading@11: av_freep(&vec); yading@11: return vec; yading@11: } yading@11: yading@11: SwsVector *sws_getGaussianVec(double variance, double quality) yading@11: { yading@11: const int length = (int)(variance * quality + 0.5) | 1; yading@11: int i; yading@11: double middle = (length - 1) * 0.5; yading@11: SwsVector *vec; yading@11: yading@11: if(variance < 0 || quality < 0) yading@11: return NULL; yading@11: yading@11: vec = sws_allocVec(length); yading@11: yading@11: if (!vec) yading@11: return NULL; yading@11: yading@11: for (i = 0; i < length; i++) { yading@11: double dist = i - middle; yading@11: vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) / yading@11: sqrt(2 * variance * M_PI); yading@11: } yading@11: yading@11: sws_normalizeVec(vec, 1.0); yading@11: yading@11: return vec; yading@11: } yading@11: yading@11: SwsVector *sws_getConstVec(double c, int length) yading@11: { yading@11: int i; yading@11: SwsVector *vec = sws_allocVec(length); yading@11: yading@11: if (!vec) yading@11: return NULL; yading@11: yading@11: for (i = 0; i < length; i++) yading@11: vec->coeff[i] = c; yading@11: yading@11: return vec; yading@11: } yading@11: yading@11: SwsVector *sws_getIdentityVec(void) yading@11: { yading@11: return sws_getConstVec(1.0, 1); yading@11: } yading@11: yading@11: static double sws_dcVec(SwsVector *a) yading@11: { yading@11: int i; yading@11: double sum = 0; yading@11: yading@11: for (i = 0; i < a->length; i++) yading@11: sum += a->coeff[i]; yading@11: yading@11: return sum; yading@11: } yading@11: yading@11: void sws_scaleVec(SwsVector *a, double scalar) yading@11: { yading@11: int i; yading@11: yading@11: for (i = 0; i < a->length; i++) yading@11: a->coeff[i] *= scalar; yading@11: } yading@11: yading@11: void sws_normalizeVec(SwsVector *a, double height) yading@11: { yading@11: sws_scaleVec(a, height / sws_dcVec(a)); yading@11: } yading@11: yading@11: static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b) yading@11: { yading@11: int length = a->length + b->length - 1; yading@11: int i, j; yading@11: SwsVector *vec = sws_getConstVec(0.0, length); yading@11: yading@11: if (!vec) yading@11: return NULL; yading@11: yading@11: for (i = 0; i < a->length; i++) { yading@11: for (j = 0; j < b->length; j++) { yading@11: vec->coeff[i + j] += a->coeff[i] * b->coeff[j]; yading@11: } yading@11: } yading@11: yading@11: return vec; yading@11: } yading@11: yading@11: static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b) yading@11: { yading@11: int length = FFMAX(a->length, b->length); yading@11: int i; yading@11: SwsVector *vec = sws_getConstVec(0.0, length); yading@11: yading@11: if (!vec) yading@11: return NULL; yading@11: yading@11: for (i = 0; i < a->length; i++) yading@11: vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i]; yading@11: for (i = 0; i < b->length; i++) yading@11: vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i]; yading@11: yading@11: return vec; yading@11: } yading@11: yading@11: static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b) yading@11: { yading@11: int length = FFMAX(a->length, b->length); yading@11: int i; yading@11: SwsVector *vec = sws_getConstVec(0.0, length); yading@11: yading@11: if (!vec) yading@11: return NULL; yading@11: yading@11: for (i = 0; i < a->length; i++) yading@11: vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i]; yading@11: for (i = 0; i < b->length; i++) yading@11: vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] -= b->coeff[i]; yading@11: yading@11: return vec; yading@11: } yading@11: yading@11: /* shift left / or right if "shift" is negative */ yading@11: static SwsVector *sws_getShiftedVec(SwsVector *a, int shift) yading@11: { yading@11: int length = a->length + FFABS(shift) * 2; yading@11: int i; yading@11: SwsVector *vec = sws_getConstVec(0.0, length); yading@11: yading@11: if (!vec) yading@11: return NULL; yading@11: yading@11: for (i = 0; i < a->length; i++) { yading@11: vec->coeff[i + (length - 1) / 2 - yading@11: (a->length - 1) / 2 - shift] = a->coeff[i]; yading@11: } yading@11: yading@11: return vec; yading@11: } yading@11: yading@11: void sws_shiftVec(SwsVector *a, int shift) yading@11: { yading@11: SwsVector *shifted = sws_getShiftedVec(a, shift); yading@11: av_free(a->coeff); yading@11: a->coeff = shifted->coeff; yading@11: a->length = shifted->length; yading@11: av_free(shifted); yading@11: } yading@11: yading@11: void sws_addVec(SwsVector *a, SwsVector *b) yading@11: { yading@11: SwsVector *sum = sws_sumVec(a, b); yading@11: av_free(a->coeff); yading@11: a->coeff = sum->coeff; yading@11: a->length = sum->length; yading@11: av_free(sum); yading@11: } yading@11: yading@11: void sws_subVec(SwsVector *a, SwsVector *b) yading@11: { yading@11: SwsVector *diff = sws_diffVec(a, b); yading@11: av_free(a->coeff); yading@11: a->coeff = diff->coeff; yading@11: a->length = diff->length; yading@11: av_free(diff); yading@11: } yading@11: yading@11: void sws_convVec(SwsVector *a, SwsVector *b) yading@11: { yading@11: SwsVector *conv = sws_getConvVec(a, b); yading@11: av_free(a->coeff); yading@11: a->coeff = conv->coeff; yading@11: a->length = conv->length; yading@11: av_free(conv); yading@11: } yading@11: yading@11: SwsVector *sws_cloneVec(SwsVector *a) yading@11: { yading@11: int i; yading@11: SwsVector *vec = sws_allocVec(a->length); yading@11: yading@11: if (!vec) yading@11: return NULL; yading@11: yading@11: for (i = 0; i < a->length; i++) yading@11: vec->coeff[i] = a->coeff[i]; yading@11: yading@11: return vec; yading@11: } yading@11: yading@11: void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level) yading@11: { yading@11: int i; yading@11: double max = 0; yading@11: double min = 0; yading@11: double range; yading@11: yading@11: for (i = 0; i < a->length; i++) yading@11: if (a->coeff[i] > max) yading@11: max = a->coeff[i]; yading@11: yading@11: for (i = 0; i < a->length; i++) yading@11: if (a->coeff[i] < min) yading@11: min = a->coeff[i]; yading@11: yading@11: range = max - min; yading@11: yading@11: for (i = 0; i < a->length; i++) { yading@11: int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5); yading@11: av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]); yading@11: for (; x > 0; x--) yading@11: av_log(log_ctx, log_level, " "); yading@11: av_log(log_ctx, log_level, "|\n"); yading@11: } yading@11: } yading@11: yading@11: void sws_freeVec(SwsVector *a) yading@11: { yading@11: if (!a) yading@11: return; yading@11: av_freep(&a->coeff); yading@11: a->length = 0; yading@11: av_free(a); yading@11: } yading@11: yading@11: void sws_freeFilter(SwsFilter *filter) yading@11: { yading@11: if (!filter) yading@11: return; yading@11: yading@11: if (filter->lumH) yading@11: sws_freeVec(filter->lumH); yading@11: if (filter->lumV) yading@11: sws_freeVec(filter->lumV); yading@11: if (filter->chrH) yading@11: sws_freeVec(filter->chrH); yading@11: if (filter->chrV) yading@11: sws_freeVec(filter->chrV); yading@11: av_free(filter); yading@11: } yading@11: yading@11: void sws_freeContext(SwsContext *c) yading@11: { yading@11: int i; yading@11: if (!c) yading@11: return; yading@11: yading@11: if (c->lumPixBuf) { yading@11: for (i = 0; i < c->vLumBufSize; i++) yading@11: av_freep(&c->lumPixBuf[i]); yading@11: av_freep(&c->lumPixBuf); yading@11: } yading@11: yading@11: if (c->chrUPixBuf) { yading@11: for (i = 0; i < c->vChrBufSize; i++) yading@11: av_freep(&c->chrUPixBuf[i]); yading@11: av_freep(&c->chrUPixBuf); yading@11: av_freep(&c->chrVPixBuf); yading@11: } yading@11: yading@11: if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) { yading@11: for (i = 0; i < c->vLumBufSize; i++) yading@11: av_freep(&c->alpPixBuf[i]); yading@11: av_freep(&c->alpPixBuf); yading@11: } yading@11: yading@11: for (i = 0; i < 4; i++) yading@11: av_freep(&c->dither_error[i]); yading@11: yading@11: av_freep(&c->vLumFilter); yading@11: av_freep(&c->vChrFilter); yading@11: av_freep(&c->hLumFilter); yading@11: av_freep(&c->hChrFilter); yading@11: #if HAVE_ALTIVEC yading@11: av_freep(&c->vYCoeffsBank); yading@11: av_freep(&c->vCCoeffsBank); yading@11: #endif yading@11: yading@11: av_freep(&c->vLumFilterPos); yading@11: av_freep(&c->vChrFilterPos); yading@11: av_freep(&c->hLumFilterPos); yading@11: av_freep(&c->hChrFilterPos); yading@11: yading@11: #if HAVE_MMX_INLINE yading@11: #if USE_MMAP yading@11: if (c->lumMmxextFilterCode) yading@11: munmap(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize); yading@11: if (c->chrMmxextFilterCode) yading@11: munmap(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize); yading@11: #elif HAVE_VIRTUALALLOC yading@11: if (c->lumMmxextFilterCode) yading@11: VirtualFree(c->lumMmxextFilterCode, 0, MEM_RELEASE); yading@11: if (c->chrMmxextFilterCode) yading@11: VirtualFree(c->chrMmxextFilterCode, 0, MEM_RELEASE); yading@11: #else yading@11: av_free(c->lumMmxextFilterCode); yading@11: av_free(c->chrMmxextFilterCode); yading@11: #endif yading@11: c->lumMmxextFilterCode = NULL; yading@11: c->chrMmxextFilterCode = NULL; yading@11: #endif /* HAVE_MMX_INLINE */ yading@11: yading@11: av_freep(&c->yuvTable); yading@11: av_freep(&c->formatConvBuffer); yading@11: yading@11: av_free(c); yading@11: } yading@11: yading@11: struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW, yading@11: int srcH, enum AVPixelFormat srcFormat, yading@11: int dstW, int dstH, yading@11: enum AVPixelFormat dstFormat, int flags, yading@11: SwsFilter *srcFilter, yading@11: SwsFilter *dstFilter, yading@11: const double *param) yading@11: { yading@11: static const double default_param[2] = { SWS_PARAM_DEFAULT, yading@11: SWS_PARAM_DEFAULT }; yading@11: yading@11: if (!param) yading@11: param = default_param; yading@11: yading@11: if (context && yading@11: (context->srcW != srcW || yading@11: context->srcH != srcH || yading@11: context->srcFormat != srcFormat || yading@11: context->dstW != dstW || yading@11: context->dstH != dstH || yading@11: context->dstFormat != dstFormat || yading@11: context->flags != flags || yading@11: context->param[0] != param[0] || yading@11: context->param[1] != param[1])) { yading@11: sws_freeContext(context); yading@11: context = NULL; yading@11: } yading@11: yading@11: if (!context) { yading@11: if (!(context = sws_alloc_context())) yading@11: return NULL; yading@11: context->srcW = srcW; yading@11: context->srcH = srcH; yading@11: context->srcRange = handle_jpeg(&srcFormat); yading@11: context->src0Alpha = handle_0alpha(&srcFormat); yading@11: context->srcFormat = srcFormat; yading@11: context->dstW = dstW; yading@11: context->dstH = dstH; yading@11: context->dstRange = handle_jpeg(&dstFormat); yading@11: context->dst0Alpha = handle_0alpha(&dstFormat); yading@11: context->dstFormat = dstFormat; yading@11: context->flags = flags; yading@11: context->param[0] = param[0]; yading@11: context->param[1] = param[1]; yading@11: sws_setColorspaceDetails(context, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], yading@11: context->srcRange, yading@11: ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, yading@11: context->dstRange, 0, 1 << 16, 1 << 16); yading@11: if (sws_init_context(context, srcFilter, dstFilter) < 0) { yading@11: sws_freeContext(context); yading@11: return NULL; yading@11: } yading@11: } yading@11: return context; yading@11: }