a4388ebd |
/*
* Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of FFmpeg.
* |
819ee683 |
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version. |
a4388ebd |
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
819ee683 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details. |
a4388ebd |
* |
819ee683 |
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software |
a4388ebd |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
|
ef0ee7f6 |
#include "config.h"
#define _SVID_SOURCE // needed for MAP_ANONYMOUS |
262b6dba |
#define _DARWIN_C_SOURCE // needed for MAP_ANON |
a4388ebd |
#include <inttypes.h>
#include <math.h>
#include <stdio.h> |
ef0ee7f6 |
#include <string.h> |
a4388ebd |
#if HAVE_SYS_MMAN_H
#include <sys/mman.h>
#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
#define MAP_ANONYMOUS MAP_ANON
#endif
#endif
#if HAVE_VIRTUALALLOC
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#endif |
ef0ee7f6 |
|
5a6e3c03 |
#include "libavutil/attributes.h" |
367d9b29 |
#include "libavutil/avassert.h" |
a4388ebd |
#include "libavutil/avutil.h"
#include "libavutil/bswap.h" |
ef0ee7f6 |
#include "libavutil/cpu.h"
#include "libavutil/intreadwrite.h" |
5089ce1b |
#include "libavutil/mathematics.h" |
f34fcdc8 |
#include "libavutil/opt.h" |
a4388ebd |
#include "libavutil/pixdesc.h" |
c318626c |
#include "libavutil/x86/asm.h" |
e0c6cce4 |
#include "libavutil/x86/cpu.h" |
ef0ee7f6 |
#include "rgb2rgb.h"
#include "swscale.h"
#include "swscale_internal.h" |
a4388ebd |
unsigned swscale_version(void)
{ |
0f91a577 |
av_assert0(LIBSWSCALE_VERSION_MICRO >= 100); |
a4388ebd |
return LIBSWSCALE_VERSION_INT;
}
const char *swscale_configuration(void)
{ |
e528cdac |
return FFMPEG_CONFIGURATION; |
a4388ebd |
}
const char *swscale_license(void)
{
#define LICENSE_PREFIX "libswscale license: " |
0cb88628 |
return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1; |
a4388ebd |
}
|
ef0ee7f6 |
#define RET 0xC3 // near return opcode for x86 |
a4388ebd |
|
c899bedb |
typedef struct FormatEntry { |
e8794f1d |
int is_supported_in, is_supported_out; |
c899bedb |
} FormatEntry;
|
716d413c |
static const FormatEntry format_entries[AV_PIX_FMT_NB] = {
[AV_PIX_FMT_YUV420P] = { 1, 1 },
[AV_PIX_FMT_YUYV422] = { 1, 1 },
[AV_PIX_FMT_RGB24] = { 1, 1 },
[AV_PIX_FMT_BGR24] = { 1, 1 },
[AV_PIX_FMT_YUV422P] = { 1, 1 },
[AV_PIX_FMT_YUV444P] = { 1, 1 },
[AV_PIX_FMT_YUV410P] = { 1, 1 },
[AV_PIX_FMT_YUV411P] = { 1, 1 },
[AV_PIX_FMT_GRAY8] = { 1, 1 },
[AV_PIX_FMT_MONOWHITE] = { 1, 1 },
[AV_PIX_FMT_MONOBLACK] = { 1, 1 },
[AV_PIX_FMT_PAL8] = { 1, 0 },
[AV_PIX_FMT_YUVJ420P] = { 1, 1 },
[AV_PIX_FMT_YUVJ422P] = { 1, 1 },
[AV_PIX_FMT_YUVJ444P] = { 1, 1 },
[AV_PIX_FMT_UYVY422] = { 1, 1 },
[AV_PIX_FMT_UYYVYY411] = { 0, 0 },
[AV_PIX_FMT_BGR8] = { 1, 1 },
[AV_PIX_FMT_BGR4] = { 0, 1 },
[AV_PIX_FMT_BGR4_BYTE] = { 1, 1 },
[AV_PIX_FMT_RGB8] = { 1, 1 },
[AV_PIX_FMT_RGB4] = { 0, 1 },
[AV_PIX_FMT_RGB4_BYTE] = { 1, 1 },
[AV_PIX_FMT_NV12] = { 1, 1 },
[AV_PIX_FMT_NV21] = { 1, 1 },
[AV_PIX_FMT_ARGB] = { 1, 1 },
[AV_PIX_FMT_RGBA] = { 1, 1 },
[AV_PIX_FMT_ABGR] = { 1, 1 },
[AV_PIX_FMT_BGRA] = { 1, 1 }, |
ac627b3d |
[AV_PIX_FMT_0RGB] = { 1, 1 },
[AV_PIX_FMT_RGB0] = { 1, 1 },
[AV_PIX_FMT_0BGR] = { 1, 1 },
[AV_PIX_FMT_BGR0] = { 1, 1 }, |
716d413c |
[AV_PIX_FMT_GRAY16BE] = { 1, 1 },
[AV_PIX_FMT_GRAY16LE] = { 1, 1 },
[AV_PIX_FMT_YUV440P] = { 1, 1 },
[AV_PIX_FMT_YUVJ440P] = { 1, 1 },
[AV_PIX_FMT_YUVA420P] = { 1, 1 }, |
ac627b3d |
[AV_PIX_FMT_YUVA422P] = { 1, 1 },
[AV_PIX_FMT_YUVA444P] = { 1, 1 }, |
7658295b |
[AV_PIX_FMT_YUVA420P9BE] = { 1, 1 },
[AV_PIX_FMT_YUVA420P9LE] = { 1, 1 },
[AV_PIX_FMT_YUVA422P9BE] = { 1, 1 },
[AV_PIX_FMT_YUVA422P9LE] = { 1, 1 },
[AV_PIX_FMT_YUVA444P9BE] = { 1, 1 },
[AV_PIX_FMT_YUVA444P9LE] = { 1, 1 },
[AV_PIX_FMT_YUVA420P10BE]= { 1, 1 },
[AV_PIX_FMT_YUVA420P10LE]= { 1, 1 },
[AV_PIX_FMT_YUVA422P10BE]= { 1, 1 },
[AV_PIX_FMT_YUVA422P10LE]= { 1, 1 },
[AV_PIX_FMT_YUVA444P10BE]= { 1, 1 },
[AV_PIX_FMT_YUVA444P10LE]= { 1, 1 },
[AV_PIX_FMT_YUVA420P16BE]= { 1, 1 },
[AV_PIX_FMT_YUVA420P16LE]= { 1, 1 },
[AV_PIX_FMT_YUVA422P16BE]= { 1, 1 },
[AV_PIX_FMT_YUVA422P16LE]= { 1, 1 },
[AV_PIX_FMT_YUVA444P16BE]= { 1, 1 },
[AV_PIX_FMT_YUVA444P16LE]= { 1, 1 }, |
716d413c |
[AV_PIX_FMT_RGB48BE] = { 1, 1 },
[AV_PIX_FMT_RGB48LE] = { 1, 1 }, |
ac627b3d |
[AV_PIX_FMT_RGBA64BE] = { 1, 0 },
[AV_PIX_FMT_RGBA64LE] = { 1, 0 }, |
716d413c |
[AV_PIX_FMT_RGB565BE] = { 1, 1 },
[AV_PIX_FMT_RGB565LE] = { 1, 1 },
[AV_PIX_FMT_RGB555BE] = { 1, 1 },
[AV_PIX_FMT_RGB555LE] = { 1, 1 },
[AV_PIX_FMT_BGR565BE] = { 1, 1 },
[AV_PIX_FMT_BGR565LE] = { 1, 1 },
[AV_PIX_FMT_BGR555BE] = { 1, 1 },
[AV_PIX_FMT_BGR555LE] = { 1, 1 },
[AV_PIX_FMT_YUV420P16LE] = { 1, 1 },
[AV_PIX_FMT_YUV420P16BE] = { 1, 1 },
[AV_PIX_FMT_YUV422P16LE] = { 1, 1 },
[AV_PIX_FMT_YUV422P16BE] = { 1, 1 },
[AV_PIX_FMT_YUV444P16LE] = { 1, 1 },
[AV_PIX_FMT_YUV444P16BE] = { 1, 1 },
[AV_PIX_FMT_RGB444LE] = { 1, 1 },
[AV_PIX_FMT_RGB444BE] = { 1, 1 },
[AV_PIX_FMT_BGR444LE] = { 1, 1 },
[AV_PIX_FMT_BGR444BE] = { 1, 1 },
[AV_PIX_FMT_Y400A] = { 1, 0 },
[AV_PIX_FMT_BGR48BE] = { 1, 1 },
[AV_PIX_FMT_BGR48LE] = { 1, 1 }, |
ac627b3d |
[AV_PIX_FMT_BGRA64BE] = { 0, 0 },
[AV_PIX_FMT_BGRA64LE] = { 0, 0 }, |
716d413c |
[AV_PIX_FMT_YUV420P9BE] = { 1, 1 },
[AV_PIX_FMT_YUV420P9LE] = { 1, 1 },
[AV_PIX_FMT_YUV420P10BE] = { 1, 1 },
[AV_PIX_FMT_YUV420P10LE] = { 1, 1 }, |
ac627b3d |
[AV_PIX_FMT_YUV420P12BE] = { 1, 1 },
[AV_PIX_FMT_YUV420P12LE] = { 1, 1 },
[AV_PIX_FMT_YUV420P14BE] = { 1, 1 },
[AV_PIX_FMT_YUV420P14LE] = { 1, 1 }, |
716d413c |
[AV_PIX_FMT_YUV422P9BE] = { 1, 1 },
[AV_PIX_FMT_YUV422P9LE] = { 1, 1 },
[AV_PIX_FMT_YUV422P10BE] = { 1, 1 },
[AV_PIX_FMT_YUV422P10LE] = { 1, 1 }, |
ac627b3d |
[AV_PIX_FMT_YUV422P12BE] = { 1, 1 },
[AV_PIX_FMT_YUV422P12LE] = { 1, 1 },
[AV_PIX_FMT_YUV422P14BE] = { 1, 1 },
[AV_PIX_FMT_YUV422P14LE] = { 1, 1 }, |
716d413c |
[AV_PIX_FMT_YUV444P9BE] = { 1, 1 },
[AV_PIX_FMT_YUV444P9LE] = { 1, 1 },
[AV_PIX_FMT_YUV444P10BE] = { 1, 1 },
[AV_PIX_FMT_YUV444P10LE] = { 1, 1 }, |
ac627b3d |
[AV_PIX_FMT_YUV444P12BE] = { 1, 1 },
[AV_PIX_FMT_YUV444P12LE] = { 1, 1 },
[AV_PIX_FMT_YUV444P14BE] = { 1, 1 },
[AV_PIX_FMT_YUV444P14LE] = { 1, 1 }, |
716d413c |
[AV_PIX_FMT_GBRP] = { 1, 0 },
[AV_PIX_FMT_GBRP9LE] = { 1, 0 },
[AV_PIX_FMT_GBRP9BE] = { 1, 0 },
[AV_PIX_FMT_GBRP10LE] = { 1, 0 },
[AV_PIX_FMT_GBRP10BE] = { 1, 0 }, |
ac627b3d |
[AV_PIX_FMT_GBRP12LE] = { 1, 0 },
[AV_PIX_FMT_GBRP12BE] = { 1, 0 },
[AV_PIX_FMT_GBRP14LE] = { 1, 0 },
[AV_PIX_FMT_GBRP14BE] = { 1, 0 }, |
716d413c |
[AV_PIX_FMT_GBRP16LE] = { 1, 0 },
[AV_PIX_FMT_GBRP16BE] = { 1, 0 }, |
c899bedb |
};
|
716d413c |
int sws_isSupportedInput(enum AVPixelFormat pix_fmt) |
a4388ebd |
{ |
716d413c |
return (unsigned)pix_fmt < AV_PIX_FMT_NB ? |
ef0ee7f6 |
format_entries[pix_fmt].is_supported_in : 0; |
a4388ebd |
}
|
716d413c |
int sws_isSupportedOutput(enum AVPixelFormat pix_fmt) |
a4388ebd |
{ |
716d413c |
return (unsigned)pix_fmt < AV_PIX_FMT_NB ? |
ef0ee7f6 |
format_entries[pix_fmt].is_supported_out : 0; |
a4388ebd |
}
extern const int32_t ff_yuv2rgb_coeffs[8][4];
|
b932eb1b |
#if FF_API_SWS_FORMAT_NAME |
716d413c |
const char *sws_format_name(enum AVPixelFormat format) |
b932eb1b |
{ |
0a7068fa |
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(format);
if (desc)
return desc->name; |
a4388ebd |
else
return "Unknown format"; |
b932eb1b |
}
#endif
|
ef0ee7f6 |
static double getSplineCoeff(double a, double b, double c, double d,
double dist) |
a4388ebd |
{ |
ef0ee7f6 |
if (dist <= 1.0)
return ((d * dist + c) * dist + b) * dist + a;
else
return getSplineCoeff(0.0,
b + 2.0 * c + 3.0 * d,
c + 3.0 * d,
-b - 3.0 * c - 6.0 * d,
dist - 1.0); |
a4388ebd |
}
|
ef0ee7f6 |
static int initFilter(int16_t **outFilter, int32_t **filterPos,
int *outFilterSize, int xInc, int srcW, int dstW,
int filterAlign, int one, int flags, int cpu_flags,
SwsVector *srcFilter, SwsVector *dstFilter, |
367d9b29 |
double param[2]) |
a4388ebd |
{
int i;
int filterSize;
int filter2Size;
int minFilterSize; |
ef0ee7f6 |
int64_t *filter = NULL;
int64_t *filter2 = NULL; |
7de21960 |
const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8)); |
ef0ee7f6 |
int ret = -1; |
e66149e7 |
|
ef0ee7f6 |
emms_c(); // FIXME should not be required but IS (even for non-MMX versions) |
a4388ebd |
|
ef0ee7f6 |
// NOTE: the +3 is for the MMX(+1) / SSE(+3) scaler which reads over the end
FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW + 3) * sizeof(**filterPos), fail); |
a4388ebd |
|
ef0ee7f6 |
if (FFABS(xInc - 0x10000) < 10) { // unscaled |
a4388ebd |
int i; |
ef0ee7f6 |
filterSize = 1;
FF_ALLOCZ_OR_GOTO(NULL, filter,
dstW * sizeof(*filter) * filterSize, fail); |
a4388ebd |
|
ef0ee7f6 |
for (i = 0; i < dstW; i++) {
filter[i * filterSize] = fone;
(*filterPos)[i] = i; |
a4388ebd |
} |
ef0ee7f6 |
} else if (flags & SWS_POINT) { // lame looking point sampling mode |
a4388ebd |
int i; |
90cad5ed |
int64_t xDstInSrc; |
ef0ee7f6 |
filterSize = 1;
FF_ALLOC_OR_GOTO(NULL, filter,
dstW * sizeof(*filter) * filterSize, fail); |
a4388ebd |
|
ef0ee7f6 |
xDstInSrc = xInc / 2 - 0x8000;
for (i = 0; i < dstW; i++) {
int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16; |
a4388ebd |
|
ef0ee7f6 |
(*filterPos)[i] = xx;
filter[i] = fone;
xDstInSrc += xInc; |
a4388ebd |
} |
ef0ee7f6 |
} else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) ||
(flags & SWS_FAST_BILINEAR)) { // bilinear upscale |
a4388ebd |
int i; |
90cad5ed |
int64_t xDstInSrc; |
ef0ee7f6 |
filterSize = 2;
FF_ALLOC_OR_GOTO(NULL, filter,
dstW * sizeof(*filter) * filterSize, fail); |
a4388ebd |
|
ef0ee7f6 |
xDstInSrc = xInc / 2 - 0x8000;
for (i = 0; i < dstW; i++) {
int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16; |
a4388ebd |
int j;
|
ef0ee7f6 |
(*filterPos)[i] = xx;
// bilinear upscale / linear interpolate / area averaging
for (j = 0; j < filterSize; j++) { |
90cad5ed |
int64_t coeff= fone - FFABS(((int64_t)xx<<16) - xDstInSrc)*(fone>>16); |
ef0ee7f6 |
if (coeff < 0)
coeff = 0;
filter[i * filterSize + j] = coeff; |
a4388ebd |
xx++;
} |
ef0ee7f6 |
xDstInSrc += xInc; |
a4388ebd |
}
} else { |
90cad5ed |
int64_t xDstInSrc; |
a4388ebd |
int sizeFactor;
|
ef0ee7f6 |
if (flags & SWS_BICUBIC)
sizeFactor = 4;
else if (flags & SWS_X)
sizeFactor = 8;
else if (flags & SWS_AREA)
sizeFactor = 1; // downscale only, for upscale it is bilinear
else if (flags & SWS_GAUSS)
sizeFactor = 8; // infinite ;)
else if (flags & SWS_LANCZOS)
sizeFactor = param[0] != SWS_PARAM_DEFAULT ? ceil(2 * param[0]) : 6;
else if (flags & SWS_SINC)
sizeFactor = 20; // infinite ;)
else if (flags & SWS_SPLINE)
sizeFactor = 20; // infinite ;)
else if (flags & SWS_BILINEAR)
sizeFactor = 2; |
a4388ebd |
else { |
e003da71 |
av_assert0(0); |
a4388ebd |
}
|
ef0ee7f6 |
if (xInc <= 1 << 16)
filterSize = 1 + sizeFactor; // upscale
else
filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW; |
a4388ebd |
|
1254022e |
filterSize = FFMIN(filterSize, srcW - 2);
filterSize = FFMAX(filterSize, 1); |
a4388ebd |
|
ef0ee7f6 |
FF_ALLOC_OR_GOTO(NULL, filter,
dstW * sizeof(*filter) * filterSize, fail); |
a4388ebd |
|
ef0ee7f6 |
xDstInSrc = xInc - 0x10000;
for (i = 0; i < dstW; i++) { |
9a2387cf |
int xx = (xDstInSrc - ((int64_t)(filterSize - 2) << 16)) / (1 << 17); |
a4388ebd |
int j; |
ef0ee7f6 |
(*filterPos)[i] = xx;
for (j = 0; j < filterSize; j++) {
int64_t d = (FFABS(((int64_t)xx << 17) - xDstInSrc)) << 13; |
a4388ebd |
double floatd;
int64_t coeff;
|
ef0ee7f6 |
if (xInc > 1 << 16)
d = d * dstW / srcW;
floatd = d * (1.0 / (1 << 30)); |
a4388ebd |
if (flags & SWS_BICUBIC) { |
ef0ee7f6 |
int64_t B = (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1 << 24);
int64_t C = (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1 << 24); |
a4388ebd |
|
ef0ee7f6 |
if (d >= 1LL << 31) { |
8305041e |
coeff = 0.0;
} else {
int64_t dd = (d * d) >> 30;
int64_t ddd = (dd * d) >> 30;
|
ef0ee7f6 |
if (d < 1LL << 30)
coeff = (12 * (1 << 24) - 9 * B - 6 * C) * ddd +
(-18 * (1 << 24) + 12 * B + 6 * C) * dd +
(6 * (1 << 24) - 2 * B) * (1 << 30); |
8305041e |
else |
ef0ee7f6 |
coeff = (-B - 6 * C) * ddd +
(6 * B + 30 * C) * dd +
(-12 * B - 48 * C) * d +
(8 * B + 24 * C) * (1 << 30); |
8305041e |
} |
7de21960 |
coeff /= (1LL<<54)/fone; |
a4388ebd |
} |
ef0ee7f6 |
#if 0
else if (flags & SWS_X) {
double p = param ? param * 0.01 : 0.3;
coeff = d ? sin(d * M_PI) / (d * M_PI) : 1.0;
coeff *= pow(2.0, -p * d * d); |
a4388ebd |
} |
ef0ee7f6 |
#endif |
a4388ebd |
else if (flags & SWS_X) { |
ef0ee7f6 |
double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0; |
a4388ebd |
double c;
|
ef0ee7f6 |
if (floatd < 1.0)
c = cos(floatd * M_PI); |
a4388ebd |
else |
ef0ee7f6 |
c = -1.0;
if (c < 0.0)
c = -pow(-c, A); |
a4388ebd |
else |
ef0ee7f6 |
c = pow(c, A);
coeff = (c * 0.5 + 0.5) * fone; |
a4388ebd |
} else if (flags & SWS_AREA) { |
ef0ee7f6 |
int64_t d2 = d - (1 << 29);
if (d2 * xInc < -(1LL << (29 + 16)))
coeff = 1.0 * (1LL << (30 + 16));
else if (d2 * xInc < (1LL << (29 + 16)))
coeff = -d2 * xInc + (1LL << (29 + 16));
else
coeff = 0.0;
coeff *= fone >> (30 + 16); |
a4388ebd |
} else if (flags & SWS_GAUSS) { |
ef0ee7f6 |
double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
coeff = (pow(2.0, -p * floatd * floatd)) * fone; |
a4388ebd |
} else if (flags & SWS_SINC) { |
ef0ee7f6 |
coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone; |
a4388ebd |
} else if (flags & SWS_LANCZOS) { |
ef0ee7f6 |
double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
coeff = (d ? sin(floatd * M_PI) * sin(floatd * M_PI / p) /
(floatd * floatd * M_PI * M_PI / p) : 1.0) * fone;
if (floatd > p)
coeff = 0; |
a4388ebd |
} else if (flags & SWS_BILINEAR) { |
ef0ee7f6 |
coeff = (1 << 30) - d;
if (coeff < 0)
coeff = 0; |
a4388ebd |
coeff *= fone >> 30;
} else if (flags & SWS_SPLINE) { |
ef0ee7f6 |
double p = -2.196152422706632;
coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone; |
a4388ebd |
} else { |
e003da71 |
av_assert0(0); |
a4388ebd |
}
|
ef0ee7f6 |
filter[i * filterSize + j] = coeff; |
a4388ebd |
xx++;
} |
ef0ee7f6 |
xDstInSrc += 2 * xInc; |
a4388ebd |
}
}
/* apply src & dst Filter to filter -> filter2 |
ef0ee7f6 |
* av_free(filter);
*/ |
e003da71 |
av_assert0(filterSize > 0); |
ef0ee7f6 |
filter2Size = filterSize;
if (srcFilter)
filter2Size += srcFilter->length - 1;
if (dstFilter)
filter2Size += dstFilter->length - 1; |
e003da71 |
av_assert0(filter2Size > 0); |
ef0ee7f6 |
FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size * dstW * sizeof(*filter2), fail);
for (i = 0; i < dstW; i++) { |
a4388ebd |
int j, k;
|
ef0ee7f6 |
if (srcFilter) {
for (k = 0; k < srcFilter->length; k++) {
for (j = 0; j < filterSize; j++)
filter2[i * filter2Size + k + j] +=
srcFilter->coeff[k] * filter[i * filterSize + j]; |
a4388ebd |
}
} else { |
ef0ee7f6 |
for (j = 0; j < filterSize; j++)
filter2[i * filter2Size + j] = filter[i * filterSize + j]; |
a4388ebd |
} |
ef0ee7f6 |
// FIXME dstFilter |
a4388ebd |
|
ef0ee7f6 |
(*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2; |
a4388ebd |
}
av_freep(&filter);
/* try to reduce the filter-size (step1 find size and shift left) */
// Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not). |
ef0ee7f6 |
minFilterSize = 0;
for (i = dstW - 1; i >= 0; i--) {
int min = filter2Size; |
a4388ebd |
int j; |
ef0ee7f6 |
int64_t cutOff = 0.0; |
a4388ebd |
/* get rid of near zero elements on the left by shifting left */ |
ef0ee7f6 |
for (j = 0; j < filter2Size; j++) { |
a4388ebd |
int k; |
ef0ee7f6 |
cutOff += FFABS(filter2[i * filter2Size]); |
a4388ebd |
|
ef0ee7f6 |
if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
break; |
a4388ebd |
|
ef0ee7f6 |
/* preserve monotonicity because the core can't handle the
* filter otherwise */
if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1])
break; |
a4388ebd |
// move filter coefficients left |
ef0ee7f6 |
for (k = 1; k < filter2Size; k++)
filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k];
filter2[i * filter2Size + k - 1] = 0; |
a4388ebd |
(*filterPos)[i]++;
}
|
ef0ee7f6 |
cutOff = 0; |
a4388ebd |
/* count near zeros on the right */ |
ef0ee7f6 |
for (j = filter2Size - 1; j > 0; j--) {
cutOff += FFABS(filter2[i * filter2Size + j]); |
a4388ebd |
|
ef0ee7f6 |
if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
break; |
a4388ebd |
min--;
}
|
ef0ee7f6 |
if (min > minFilterSize)
minFilterSize = min; |
a4388ebd |
}
|
e66149e7 |
if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) { |
ef0ee7f6 |
// we can handle the special case 4, so we don't want to go the full 8 |
a4388ebd |
if (minFilterSize < 5)
filterAlign = 4;
|
ef0ee7f6 |
/* We really don't want to waste our time doing useless computation, so
* fall back on the scalar C code for very small filters.
* Vectorizing is worth it only if you have a decent-sized vector. */ |
a4388ebd |
if (minFilterSize < 3)
filterAlign = 1;
}
|
e0c6cce4 |
if (INLINE_MMX(cpu_flags)) { |
a4388ebd |
// special case for unscaled vertical filtering
if (minFilterSize == 1 && filterAlign == 2) |
ef0ee7f6 |
filterAlign = 1; |
a4388ebd |
}
|
e003da71 |
av_assert0(minFilterSize > 0); |
ef0ee7f6 |
filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1)); |
e003da71 |
av_assert0(filterSize > 0); |
ef0ee7f6 |
filter = av_malloc(filterSize * dstW * sizeof(*filter));
if (filterSize >= MAX_FILTER_SIZE * 16 / |
733f85b7 |
((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter) {
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); |
a4388ebd |
goto fail; |
733f85b7 |
} |
ef0ee7f6 |
*outFilterSize = filterSize; |
a4388ebd |
|
ef0ee7f6 |
if (flags & SWS_PRINT_INFO)
av_log(NULL, AV_LOG_VERBOSE,
"SwScaler: reducing / aligning filtersize %d -> %d\n",
filter2Size, filterSize); |
a4388ebd |
/* try to reduce the filter-size (step2 reduce it) */ |
ef0ee7f6 |
for (i = 0; i < dstW; i++) { |
a4388ebd |
int j;
|
ef0ee7f6 |
for (j = 0; j < filterSize; j++) {
if (j >= filter2Size)
filter[i * filterSize + j] = 0;
else
filter[i * filterSize + j] = filter2[i * filter2Size + j];
if ((flags & SWS_BITEXACT) && j >= minFilterSize)
filter[i * filterSize + j] = 0; |
a4388ebd |
}
}
|
ef0ee7f6 |
// FIXME try to align filterPos if possible |
a4388ebd |
|
ef0ee7f6 |
// fix borders |
367d9b29 |
for (i = 0; i < dstW; i++) { |
a4388ebd |
int j;
if ((*filterPos)[i] < 0) {
// move filter coefficients left to compensate for filterPos |
367d9b29 |
for (j = 1; j < filterSize; j++) {
int left = FFMAX(j + (*filterPos)[i], 0);
filter[i * filterSize + left] += filter[i * filterSize + j];
filter[i * filterSize + j] = 0; |
a4388ebd |
}
(*filterPos)[i]= 0;
}
if ((*filterPos)[i] + filterSize > srcW) { |
367d9b29 |
int shift = (*filterPos)[i] + filterSize - srcW; |
a4388ebd |
// move filter coefficients right to compensate for filterPos |
367d9b29 |
for (j = filterSize - 2; j >= 0; j--) {
int right = FFMIN(j + shift, filterSize - 1);
filter[i * filterSize + right] += filter[i * filterSize + j];
filter[i * filterSize + j] = 0; |
a4388ebd |
}
(*filterPos)[i]= srcW - filterSize;
}
}
// Note the +1 is for the MMX scaler which reads over the end
/* align at 16 for AltiVec (needed by hScale_altivec_real) */ |
ef0ee7f6 |
FF_ALLOCZ_OR_GOTO(NULL, *outFilter,
*outFilterSize * (dstW + 3) * sizeof(int16_t), fail); |
a4388ebd |
/* normalize & store in outFilter */ |
ef0ee7f6 |
for (i = 0; i < dstW; i++) { |
a4388ebd |
int j; |
ef0ee7f6 |
int64_t error = 0;
int64_t sum = 0; |
a4388ebd |
|
ef0ee7f6 |
for (j = 0; j < filterSize; j++) {
sum += filter[i * filterSize + j]; |
a4388ebd |
} |
ef0ee7f6 |
sum = (sum + one / 2) / one;
for (j = 0; j < *outFilterSize; j++) {
int64_t v = filter[i * filterSize + j] + error;
int intV = ROUNDED_DIV(v, sum);
(*outFilter)[i * (*outFilterSize) + j] = intV;
error = v - intV * sum; |
a4388ebd |
}
}
|
ef0ee7f6 |
(*filterPos)[dstW + 0] =
(*filterPos)[dstW + 1] =
(*filterPos)[dstW + 2] = (*filterPos)[dstW - 1]; /* the MMX/SSE scaler will
* read over the end */
for (i = 0; i < *outFilterSize; i++) {
int k = (dstW - 1) * (*outFilterSize) + i; |
ea540401 |
(*outFilter)[k + 1 * (*outFilterSize)] =
(*outFilter)[k + 2 * (*outFilterSize)] =
(*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k]; |
a4388ebd |
}
|
ef0ee7f6 |
ret = 0;
|
a4388ebd |
fail: |
733f85b7 |
if(ret < 0)
av_log(NULL, AV_LOG_ERROR, "sws: initFilter failed\n"); |
a4388ebd |
av_free(filter);
av_free(filter2);
return ret;
}
|
17337f54 |
#if HAVE_MMXEXT_INLINE |
d8eda370 |
static int init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode,
int16_t *filter, int32_t *filterPos,
int numSplits) |
a4388ebd |
{
uint8_t *fragmentA;
x86_reg imm8OfPShufW1A;
x86_reg imm8OfPShufW2A;
x86_reg fragmentLengthA;
uint8_t *fragmentB;
x86_reg imm8OfPShufW1B;
x86_reg imm8OfPShufW2B;
x86_reg fragmentLengthB;
int fragmentPos;
int xpos, i;
// create an optimized horizontal scaling routine |
652f5185 |
/* This scaler is made of runtime-generated MMXEXT code using specially tuned |
ef0ee7f6 |
* pshufw instructions. For every four output pixels, if four input pixels
* are enough for the fast bilinear scaling, then a chunk of fragmentB is
* used. If five input pixels are needed, then a chunk of fragmentA is used. |
a4388ebd |
*/
|
ef0ee7f6 |
// code fragment |
a4388ebd |
|
ef0ee7f6 |
__asm__ volatile ( |
a4388ebd |
"jmp 9f \n\t" |
ef0ee7f6 |
// Begin |
a4388ebd |
"0: \n\t"
"movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
"movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
"movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t"
"punpcklbw %%mm7, %%mm1 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
"pshufw $0xFF, %%mm1, %%mm1 \n\t"
"1: \n\t"
"pshufw $0xFF, %%mm0, %%mm0 \n\t"
"2: \n\t"
"psubw %%mm1, %%mm0 \n\t"
"movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
"pmullw %%mm3, %%mm0 \n\t"
"psllw $7, %%mm1 \n\t"
"paddw %%mm1, %%mm0 \n\t"
"movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
"add $8, %%"REG_a" \n\t" |
ef0ee7f6 |
// End |
a4388ebd |
"9: \n\t" |
ef0ee7f6 |
// "int $3 \n\t"
"lea " LOCAL_MANGLE(0b) ", %0 \n\t"
"lea " LOCAL_MANGLE(1b) ", %1 \n\t"
"lea " LOCAL_MANGLE(2b) ", %2 \n\t" |
a4388ebd |
"dec %1 \n\t"
"dec %2 \n\t"
"sub %0, %1 \n\t"
"sub %0, %2 \n\t" |
ef0ee7f6 |
"lea " LOCAL_MANGLE(9b) ", %3 \n\t" |
a4388ebd |
"sub %0, %3 \n\t"
|
ef0ee7f6 |
: "=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
"=r" (fragmentLengthA)
); |
a4388ebd |
|
ef0ee7f6 |
__asm__ volatile ( |
a4388ebd |
"jmp 9f \n\t" |
ef0ee7f6 |
// Begin |
a4388ebd |
"0: \n\t"
"movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
"movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
"pshufw $0xFF, %%mm0, %%mm1 \n\t"
"1: \n\t"
"pshufw $0xFF, %%mm0, %%mm0 \n\t"
"2: \n\t"
"psubw %%mm1, %%mm0 \n\t"
"movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
"pmullw %%mm3, %%mm0 \n\t"
"psllw $7, %%mm1 \n\t"
"paddw %%mm1, %%mm0 \n\t"
"movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
"add $8, %%"REG_a" \n\t" |
ef0ee7f6 |
// End |
a4388ebd |
"9: \n\t" |
ef0ee7f6 |
// "int $3 \n\t"
"lea " LOCAL_MANGLE(0b) ", %0 \n\t"
"lea " LOCAL_MANGLE(1b) ", %1 \n\t"
"lea " LOCAL_MANGLE(2b) ", %2 \n\t" |
a4388ebd |
"dec %1 \n\t"
"dec %2 \n\t"
"sub %0, %1 \n\t"
"sub %0, %2 \n\t" |
ef0ee7f6 |
"lea " LOCAL_MANGLE(9b) ", %3 \n\t" |
a4388ebd |
"sub %0, %3 \n\t"
|
ef0ee7f6 |
: "=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
"=r" (fragmentLengthB)
); |
a4388ebd |
|
ef0ee7f6 |
xpos = 0; // lumXInc/2 - 0x8000; // difference between pixel centers
fragmentPos = 0; |
a4388ebd |
|
ef0ee7f6 |
for (i = 0; i < dstW / numSplits; i++) {
int xx = xpos >> 16; |
a4388ebd |
|
ef0ee7f6 |
if ((i & 3) == 0) {
int a = 0;
int b = ((xpos + xInc) >> 16) - xx;
int c = ((xpos + xInc * 2) >> 16) - xx;
int d = ((xpos + xInc * 3) >> 16) - xx;
int inc = (d + 1 < 4);
uint8_t *fragment = (d + 1 < 4) ? fragmentB : fragmentA;
x86_reg imm8OfPShufW1 = (d + 1 < 4) ? imm8OfPShufW1B : imm8OfPShufW1A;
x86_reg imm8OfPShufW2 = (d + 1 < 4) ? imm8OfPShufW2B : imm8OfPShufW2A;
x86_reg fragmentLength = (d + 1 < 4) ? fragmentLengthB : fragmentLengthA;
int maxShift = 3 - (d + inc);
int shift = 0; |
a4388ebd |
if (filterCode) { |
ef0ee7f6 |
filter[i] = ((xpos & 0xFFFF) ^ 0xFFFF) >> 9;
filter[i + 1] = (((xpos + xInc) & 0xFFFF) ^ 0xFFFF) >> 9;
filter[i + 2] = (((xpos + xInc * 2) & 0xFFFF) ^ 0xFFFF) >> 9;
filter[i + 3] = (((xpos + xInc * 3) & 0xFFFF) ^ 0xFFFF) >> 9;
filterPos[i / 2] = xx; |
a4388ebd |
memcpy(filterCode + fragmentPos, fragment, fragmentLength);
|
ef0ee7f6 |
filterCode[fragmentPos + imm8OfPShufW1] = (a + inc) |
((b + inc) << 2) |
((c + inc) << 4) |
((d + inc) << 6);
filterCode[fragmentPos + imm8OfPShufW2] = a | (b << 2) |
(c << 4) |
(d << 6);
if (i + 4 - inc >= dstW)
shift = maxShift; // avoid overread
else if ((filterPos[i / 2] & 3) <= maxShift)
shift = filterPos[i / 2] & 3; // align
if (shift && i >= shift) {
filterCode[fragmentPos + imm8OfPShufW1] += 0x55 * shift;
filterCode[fragmentPos + imm8OfPShufW2] += 0x55 * shift;
filterPos[i / 2] -= shift; |
a4388ebd |
}
}
|
ef0ee7f6 |
fragmentPos += fragmentLength; |
a4388ebd |
if (filterCode) |
ef0ee7f6 |
filterCode[fragmentPos] = RET; |
a4388ebd |
} |
ef0ee7f6 |
xpos += xInc; |
a4388ebd |
}
if (filterCode) |
ef0ee7f6 |
filterPos[((i / 2) + 1) & (~1)] = xpos >> 16; // needed to jump to the next part |
a4388ebd |
return fragmentPos + 1;
} |
17337f54 |
#endif /* HAVE_MMXEXT_INLINE */ |
a4388ebd |
|
716d413c |
static void getSubSampleFactors(int *h, int *v, enum AVPixelFormat format) |
a4388ebd |
{ |
0a7068fa |
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(format);
*h = desc->log2_chroma_w;
*v = desc->log2_chroma_h; |
a4388ebd |
}
|
96c1e6d4 |
int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
int srcRange, const int table[4], int dstRange,
int brightness, int contrast, int saturation) |
a4388ebd |
{ |
0a7068fa |
const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(c->dstFormat);
const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(c->srcFormat); |
ef0ee7f6 |
memcpy(c->srcColorspaceTable, inv_table, sizeof(int) * 4);
memcpy(c->dstColorspaceTable, table, sizeof(int) * 4); |
a4388ebd |
|
8edf9b1f |
if(!isYUV(c->dstFormat) && !isGray(c->dstFormat))
dstRange = 0;
if(!isYUV(c->srcFormat) && !isGray(c->srcFormat))
srcRange = 0;
|
ef0ee7f6 |
c->brightness = brightness;
c->contrast = contrast;
c->saturation = saturation;
c->srcRange = srcRange;
c->dstRange = dstRange; |
8edf9b1f |
|
ef0ee7f6 |
if (isYUV(c->dstFormat) || isGray(c->dstFormat))
return -1; |
a4388ebd |
|
0a7068fa |
c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
c->srcFormatBpp = av_get_bits_per_pixel(desc_src); |
635d4aed |
|
ef0ee7f6 |
ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
contrast, saturation);
// FIXME factorize |
a4388ebd |
|
e66149e7 |
if (HAVE_ALTIVEC && av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) |
ef0ee7f6 |
ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness,
contrast, saturation); |
a4388ebd |
return 0;
}
|
96c1e6d4 |
int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
int *srcRange, int **table, int *dstRange,
int *brightness, int *contrast, int *saturation) |
a4388ebd |
{ |
367d9b29 |
if (!c || isYUV(c->dstFormat) || isGray(c->dstFormat)) |
ef0ee7f6 |
return -1; |
a4388ebd |
|
ef0ee7f6 |
*inv_table = c->srcColorspaceTable;
*table = c->dstColorspaceTable;
*srcRange = c->srcRange;
*dstRange = c->dstRange;
*brightness = c->brightness;
*contrast = c->contrast;
*saturation = c->saturation; |
a4388ebd |
return 0;
}
|
716d413c |
static int handle_jpeg(enum AVPixelFormat *format) |
a4388ebd |
{
switch (*format) { |
716d413c |
case AV_PIX_FMT_YUVJ420P:
*format = AV_PIX_FMT_YUV420P; |
ef0ee7f6 |
return 1; |
716d413c |
case AV_PIX_FMT_YUVJ422P:
*format = AV_PIX_FMT_YUV422P; |
ef0ee7f6 |
return 1; |
716d413c |
case AV_PIX_FMT_YUVJ444P:
*format = AV_PIX_FMT_YUV444P; |
ef0ee7f6 |
return 1; |
716d413c |
case AV_PIX_FMT_YUVJ440P:
*format = AV_PIX_FMT_YUV440P; |
ef0ee7f6 |
return 1;
default:
return 0; |
1c910d2f |
}
}
|
ac627b3d |
static int handle_0alpha(enum AVPixelFormat *format) |
1c910d2f |
{
switch (*format) { |
ac627b3d |
case AV_PIX_FMT_0BGR : *format = AV_PIX_FMT_ABGR ; return 1;
case AV_PIX_FMT_BGR0 : *format = AV_PIX_FMT_BGRA ; return 4;
case AV_PIX_FMT_0RGB : *format = AV_PIX_FMT_ARGB ; return 1;
case AV_PIX_FMT_RGB0 : *format = AV_PIX_FMT_RGBA ; return 4; |
b8dbff33 |
default: return 0; |
a4388ebd |
}
}
|
c24b404b |
SwsContext *sws_alloc_context(void)
{ |
ef0ee7f6 |
SwsContext *c = av_mallocz(sizeof(SwsContext)); |
635d4aed |
|
f73f76fd |
if (c) {
c->av_class = &sws_context_class;
av_opt_set_defaults(c);
} |
635d4aed |
return c;
}
|
5a6e3c03 |
av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
SwsFilter *dstFilter) |
c24b404b |
{ |
29fb1b5a |
int i, j; |
a4388ebd |
int usesVFilter, usesHFilter;
int unscaled; |
ef0ee7f6 |
SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
int srcW = c->srcW;
int srcH = c->srcH;
int dstW = c->dstW;
int dstH = c->dstH; |
367d9b29 |
int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16); |
e66149e7 |
int flags, cpu_flags; |
716d413c |
enum AVPixelFormat srcFormat = c->srcFormat;
enum AVPixelFormat dstFormat = c->dstFormat; |
0a7068fa |
const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(srcFormat);
const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(dstFormat); |
635d4aed |
|
e66149e7 |
cpu_flags = av_get_cpu_flags();
flags = c->flags; |
c4fd283a |
emms_c(); |
ef0ee7f6 |
if (!rgb15to16)
sws_rgb2rgb_init(); |
a4388ebd |
unscaled = (srcW == dstW && srcH == dstH);
|
5bccb2d0 |
handle_jpeg(&srcFormat);
handle_jpeg(&dstFormat); |
1c910d2f |
handle_0alpha(&srcFormat);
handle_0alpha(&dstFormat); |
5bccb2d0 |
if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat){
av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
c->srcFormat= srcFormat;
c->dstFormat= dstFormat;
}
|
07f49ca1 |
if (!sws_isSupportedInput(srcFormat)) { |
ef0ee7f6 |
av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n", |
367d9b29 |
av_get_pix_fmt_name(srcFormat)); |
635d4aed |
return AVERROR(EINVAL); |
a4388ebd |
} |
07f49ca1 |
if (!sws_isSupportedOutput(dstFormat)) { |
ef0ee7f6 |
av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n", |
367d9b29 |
av_get_pix_fmt_name(dstFormat)); |
635d4aed |
return AVERROR(EINVAL); |
a4388ebd |
}
|
ef0ee7f6 |
i = flags & (SWS_POINT |
SWS_AREA |
SWS_BILINEAR |
SWS_FAST_BILINEAR |
SWS_BICUBIC |
SWS_X |
SWS_GAUSS |
SWS_LANCZOS |
SWS_SINC |
SWS_SPLINE |
SWS_BICUBLIN);
if (!i || (i & (i - 1))) { |
0421b6dc |
av_log(c, AV_LOG_ERROR, "Exactly one scaler algorithm must be chosen, got %X\n", i); |
635d4aed |
return AVERROR(EINVAL); |
a4388ebd |
}
/* sanity check */ |
425c30dd |
if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) { |
ef0ee7f6 |
/* FIXME check if these are enough and try to lower them after
* fixing the relevant parts of the code */ |
3636e791 |
av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n", |
a4388ebd |
srcW, srcH, dstW, dstH); |
635d4aed |
return AVERROR(EINVAL); |
a4388ebd |
}
|
ef0ee7f6 |
if (!dstFilter)
dstFilter = &dummyFilter;
if (!srcFilter)
srcFilter = &dummyFilter; |
a4388ebd |
|
ef0ee7f6 |
c->lumXInc = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
c->lumYInc = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH; |
0a7068fa |
c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
c->srcFormatBpp = av_get_bits_per_pixel(desc_src); |
ef0ee7f6 |
c->vRounder = 4 * 0x0001000100010001ULL; |
a4388ebd |
|
ef0ee7f6 |
usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) ||
(srcFilter->chrV && srcFilter->chrV->length > 1) ||
(dstFilter->lumV && dstFilter->lumV->length > 1) ||
(dstFilter->chrV && dstFilter->chrV->length > 1);
usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) ||
(srcFilter->chrH && srcFilter->chrH->length > 1) ||
(dstFilter->lumH && dstFilter->lumH->length > 1) ||
(dstFilter->chrH && dstFilter->chrH->length > 1); |
a4388ebd |
getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
|
87a72b91 |
|
ba6e36f9 |
if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) {
if (dstW&1) {
av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
flags |= SWS_FULL_CHR_H_INT;
c->flags = flags; |
87a72b91 |
}
} |
ef0ee7f6 |
/* reuse chroma for 2 pixels RGB/BGR unless user wants full
* chroma interpolation */ |
87a72b91 |
if (flags & SWS_FULL_CHR_H_INT && |
ef0ee7f6 |
isAnyRGB(dstFormat) && |
716d413c |
dstFormat != AV_PIX_FMT_RGBA &&
dstFormat != AV_PIX_FMT_ARGB &&
dstFormat != AV_PIX_FMT_BGRA &&
dstFormat != AV_PIX_FMT_ABGR &&
dstFormat != AV_PIX_FMT_RGB24 &&
dstFormat != AV_PIX_FMT_BGR24) { |
87a72b91 |
av_log(c, AV_LOG_WARNING,
"full chroma interpolation for destination format '%s' not yet implemented\n", |
367d9b29 |
av_get_pix_fmt_name(dstFormat)); |
ef0ee7f6 |
flags &= ~SWS_FULL_CHR_H_INT; |
87a72b91 |
c->flags = flags; |
ba6e36f9 |
} |
ef0ee7f6 |
if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
c->chrDstHSubSample = 1; |
a4388ebd |
// drop some chroma lines if the user wants it |
ef0ee7f6 |
c->vChrDrop = (flags & SWS_SRC_V_CHR_DROP_MASK) >>
SWS_SRC_V_CHR_DROP_SHIFT;
c->chrSrcVSubSample += c->vChrDrop;
/* drop every other pixel for chroma calculation unless user
* wants full chroma */
if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP) && |
716d413c |
srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 &&
srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 &&
srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE && |
ef0ee7f6 |
((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
(flags & SWS_FAST_BILINEAR)))
c->chrSrcHSubSample = 1; |
a4388ebd |
// Note the -((-x)>>y) is so that we always round toward +inf. |
ef0ee7f6 |
c->chrSrcW = -((-srcW) >> c->chrSrcHSubSample);
c->chrSrcH = -((-srcH) >> c->chrSrcVSubSample);
c->chrDstW = -((-dstW) >> c->chrDstHSubSample);
c->chrDstH = -((-dstH) >> c->chrDstVSubSample); |
a4388ebd |
|
21d82554 |
FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
|
a4388ebd |
/* unscaled special cases */ |
ef0ee7f6 |
if (unscaled && !usesHFilter && !usesVFilter &&
(c->srcRange == c->dstRange || isAnyRGB(dstFormat))) { |
a4388ebd |
ff_get_unscaled_swscale(c);
if (c->swScale) { |
ef0ee7f6 |
if (flags & SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO,
"using unscaled %s -> %s special converter\n", |
e1197b9e |
av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat)); |
635d4aed |
return 0; |
a4388ebd |
}
}
|
0a7068fa |
c->srcBpc = 1 + desc_src->comp[0].depth_minus1; |
28c1115a |
if (c->srcBpc < 8)
c->srcBpc = 8; |
0a7068fa |
c->dstBpc = 1 + desc_dst->comp[0].depth_minus1; |
28c1115a |
if (c->dstBpc < 8)
c->dstBpc = 8; |
ac627b3d |
if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8) |
ca1dfea1 |
c->srcBpc = 16; |
28c1115a |
if (c->dstBpc == 16) |
45f6ffe5 |
dst_stride <<= 1; |
4b874fc4 |
|
77aedc77 |
if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) { |
a65bdceb |
c->canMMXEXTBeUsed = (dstW >= srcW && (dstW & 31) == 0 &&
(srcW & 15) == 0) ? 1 : 0;
if (!c->canMMXEXTBeUsed && dstW >= srcW && (srcW & 15) == 0 |
4b874fc4 |
|
ef0ee7f6 |
&& (flags & SWS_FAST_BILINEAR)) {
if (flags & SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO, |
652f5185 |
"output width is not a multiple of 32 -> no MMXEXT scaler\n"); |
a4388ebd |
} |
367d9b29 |
if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat)) |
a65bdceb |
c->canMMXEXTBeUsed = 0; |
ef0ee7f6 |
} else |
a65bdceb |
c->canMMXEXTBeUsed = 0; |
ef0ee7f6 |
c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW;
c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH;
/* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src
* to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do
* correct scaling.
* n-2 is the last chrominance sample available.
* This is not perfect, but no one should notice the difference, the more
* correct variant would be like the vertical one, but that would require
* some special code for the first and last pixel */
if (flags & SWS_FAST_BILINEAR) { |
a65bdceb |
if (c->canMMXEXTBeUsed) { |
ef0ee7f6 |
c->lumXInc += 20;
c->chrXInc += 20; |
a4388ebd |
} |
ef0ee7f6 |
// we don't use the x86 asm scaler if MMX is available |
77aedc77 |
else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) { |
ef0ee7f6 |
c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20;
c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20; |
a4388ebd |
}
}
|
95cd815c |
#define USE_MMAP (HAVE_MMAP && HAVE_MPROTECT && defined MAP_ANONYMOUS)
|
a4388ebd |
/* precalculate horizontal scaler filter coefficients */
{ |
17337f54 |
#if HAVE_MMXEXT_INLINE |
a4388ebd |
// can't downscale !!! |
a65bdceb |
if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) { |
d8eda370 |
c->lumMmxextFilterCodeSize = init_hscaler_mmxext(dstW, c->lumXInc, NULL,
NULL, NULL, 8);
c->chrMmxextFilterCodeSize = init_hscaler_mmxext(c->chrDstW, c->chrXInc,
NULL, NULL, NULL, 4); |
a4388ebd |
|
95cd815c |
#if USE_MMAP |
a65bdceb |
c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize,
PROT_READ | PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS,
-1, 0);
c->chrMmxextFilterCode = mmap(NULL, c->chrMmxextFilterCodeSize,
PROT_READ | PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS,
-1, 0); |
a4388ebd |
#elif HAVE_VIRTUALALLOC |
a65bdceb |
c->lumMmxextFilterCode = VirtualAlloc(NULL,
c->lumMmxextFilterCodeSize,
MEM_COMMIT,
PAGE_EXECUTE_READWRITE);
c->chrMmxextFilterCode = VirtualAlloc(NULL,
c->chrMmxextFilterCodeSize,
MEM_COMMIT,
PAGE_EXECUTE_READWRITE); |
a4388ebd |
#else |
a65bdceb |
c->lumMmxextFilterCode = av_malloc(c->lumMmxextFilterCodeSize);
c->chrMmxextFilterCode = av_malloc(c->chrMmxextFilterCodeSize); |
a4388ebd |
#endif
|
2671ab39 |
#ifdef MAP_ANONYMOUS |
4b874fc4 |
if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED) |
2671ab39 |
#else |
a65bdceb |
if (!c->lumMmxextFilterCode || !c->chrMmxextFilterCode) |
2671ab39 |
#endif |
32a92d32 |
{
av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n"); |
635d4aed |
return AVERROR(ENOMEM); |
32a92d32 |
} |
a4388ebd |
|
ef0ee7f6 |
FF_ALLOCZ_OR_GOTO(c, c->hLumFilter, (dstW / 8 + 8) * sizeof(int16_t), fail);
FF_ALLOCZ_OR_GOTO(c, c->hChrFilter, (c->chrDstW / 4 + 8) * sizeof(int16_t), fail);
FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW / 2 / 8 + 8) * sizeof(int32_t), fail);
FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail); |
a4388ebd |
|
add7513e |
init_hscaler_mmxext( dstW, c->lumXInc, c->lumMmxextFilterCode,
c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8); |
d8eda370 |
init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode, |
add7513e |
c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4); |
a4388ebd |
|
95cd815c |
#if USE_MMAP |
a65bdceb |
mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ);
mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ); |
a4388ebd |
#endif
} else |
17337f54 |
#endif /* HAVE_MMXEXT_INLINE */ |
a4388ebd |
{ |
ef0ee7f6 |
const int filterAlign =
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 : |
e66149e7 |
(HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 : |
a4388ebd |
1;
|
ef0ee7f6 |
if (initFilter(&c->hLumFilter, &c->hLumFilterPos,
&c->hLumFilterSize, c->lumXInc,
srcW, dstW, filterAlign, 1 << 14,
(flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
cpu_flags, srcFilter->lumH, dstFilter->lumH, |
367d9b29 |
c->param) < 0) |
a4388ebd |
goto fail; |
ef0ee7f6 |
if (initFilter(&c->hChrFilter, &c->hChrFilterPos,
&c->hChrFilterSize, c->chrXInc,
c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
(flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
cpu_flags, srcFilter->chrH, dstFilter->chrH, |
367d9b29 |
c->param) < 0) |
a4388ebd |
goto fail;
}
} // initialize horizontal stuff
/* precalculate vertical scaler filter coefficients */
{ |
ef0ee7f6 |
const int filterAlign =
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 2 : |
e66149e7 |
(HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 : |
a4388ebd |
1;
|
ef0ee7f6 |
if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize,
c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
(flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
cpu_flags, srcFilter->lumV, dstFilter->lumV, |
367d9b29 |
c->param) < 0) |
a4388ebd |
goto fail; |
ef0ee7f6 |
if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize,
c->chrYInc, c->chrSrcH, c->chrDstH,
filterAlign, (1 << 12),
(flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
cpu_flags, srcFilter->chrV, dstFilter->chrV, |
367d9b29 |
c->param) < 0) |
a4388ebd |
goto fail;
|
f684f3c5 |
#if HAVE_ALTIVEC |
ef0ee7f6 |
FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof(vector signed short) * c->vLumFilterSize * c->dstH, fail);
FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof(vector signed short) * c->vChrFilterSize * c->chrDstH, fail); |
a4388ebd |
|
ef0ee7f6 |
for (i = 0; i < c->vLumFilterSize * c->dstH; i++) { |
a4388ebd |
int j;
short *p = (short *)&c->vYCoeffsBank[i]; |
ef0ee7f6 |
for (j = 0; j < 8; j++) |
a4388ebd |
p[j] = c->vLumFilter[i];
}
|
ef0ee7f6 |
for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) { |
a4388ebd |
int j;
short *p = (short *)&c->vCCoeffsBank[i]; |
ef0ee7f6 |
for (j = 0; j < 8; j++) |
a4388ebd |
p[j] = c->vChrFilter[i];
}
#endif
}
// calculate buffer sizes so that they won't run out while handling these damn slices |
ef0ee7f6 |
c->vLumBufSize = c->vLumFilterSize;
c->vChrBufSize = c->vChrFilterSize;
for (i = 0; i < dstH; i++) {
int chrI = (int64_t)i * c->chrDstH / dstH;
int nextSlice = FFMAX(c->vLumFilterPos[i] + c->vLumFilterSize - 1,
((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)
<< c->chrSrcVSubSample));
nextSlice >>= c->chrSrcVSubSample;
nextSlice <<= c->chrSrcVSubSample;
if (c->vLumFilterPos[i] + c->vLumBufSize < nextSlice)
c->vLumBufSize = nextSlice - c->vLumFilterPos[i];
if (c->vChrFilterPos[chrI] + c->vChrBufSize <
(nextSlice >> c->chrSrcVSubSample))
c->vChrBufSize = (nextSlice >> c->chrSrcVSubSample) -
c->vChrFilterPos[chrI]; |
a4388ebd |
}
|
ef0ee7f6 |
/* Allocate pixbufs (we use dynamic allocation because otherwise we would
* need to allocate several megabytes to handle all possible cases) */
FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
FF_ALLOC_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
FF_ALLOC_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail); |
a4388ebd |
if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) |
ef0ee7f6 |
FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
/* Note we need at least one pixel more at the end because of the MMX code
* (just in case someone wants to replace the 4000/8000). */ |
a4388ebd |
/* align at 16 bytes for AltiVec */ |
ef0ee7f6 |
for (i = 0; i < c->vLumBufSize; i++) {
FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i + c->vLumBufSize],
dst_stride + 16, fail);
c->lumPixBuf[i] = c->lumPixBuf[i + c->vLumBufSize]; |
a4388ebd |
} |
baba2eed |
// 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate) |
ca1dfea1 |
c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7); |
042f9d62 |
c->uv_offx2 = dst_stride + 16; |
ef0ee7f6 |
for (i = 0; i < c->vChrBufSize; i++) {
FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i + c->vChrBufSize],
dst_stride * 2 + 32, fail);
c->chrUPixBuf[i] = c->chrUPixBuf[i + c->vChrBufSize];
c->chrVPixBuf[i] = c->chrVPixBuf[i + c->vChrBufSize]
= c->chrUPixBuf[i] + (dst_stride >> 1) + 8; |
a4388ebd |
}
if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) |
ef0ee7f6 |
for (i = 0; i < c->vLumBufSize; i++) {
FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i + c->vLumBufSize],
dst_stride + 16, fail);
c->alpPixBuf[i] = c->alpPixBuf[i + c->vLumBufSize]; |
a4388ebd |
}
|
ef0ee7f6 |
// try to avoid drawing green stuff between the right end and the stride end
for (i = 0; i < c->vChrBufSize; i++) |
a291345b |
if(desc_dst->comp[0].depth_minus1 == 15){ |
fa36f334 |
av_assert0(c->dstBpc > 14); |
29fb1b5a |
for(j=0; j<dst_stride/2+1; j++)
((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
} else
for(j=0; j<dst_stride+1; j++)
((int16_t*)(c->chrUPixBuf[i]))[j] = 1<<14; |
a4388ebd |
|
e003da71 |
av_assert0(c->chrDstH <= dstH); |
a4388ebd |
|
ef0ee7f6 |
if (flags & SWS_PRINT_INFO) {
if (flags & SWS_FAST_BILINEAR)
av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
else if (flags & SWS_BILINEAR)
av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
else if (flags & SWS_BICUBIC)
av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
else if (flags & SWS_X)
av_log(c, AV_LOG_INFO, "Experimental scaler, ");
else if (flags & SWS_POINT)
av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
else if (flags & SWS_AREA)
av_log(c, AV_LOG_INFO, "Area Averaging scaler, ");
else if (flags & SWS_BICUBLIN)
av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
else if (flags & SWS_GAUSS)
av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
else if (flags & SWS_SINC)
av_log(c, AV_LOG_INFO, "Sinc scaler, ");
else if (flags & SWS_LANCZOS)
av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
else if (flags & SWS_SPLINE)
av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
else
av_log(c, AV_LOG_INFO, "ehh flags invalid?! "); |
a4388ebd |
av_log(c, AV_LOG_INFO, "from %s to %s%s ", |
e1197b9e |
av_get_pix_fmt_name(srcFormat), |
a4388ebd |
#ifdef DITHER1XBPP |
716d413c |
dstFormat == AV_PIX_FMT_BGR555 || dstFormat == AV_PIX_FMT_BGR565 ||
dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE ||
dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ? |
ef0ee7f6 |
"dithered " : "", |
a4388ebd |
#else
"",
#endif |
e1197b9e |
av_get_pix_fmt_name(dstFormat)); |
a4388ebd |
|
e0c6cce4 |
if (INLINE_MMXEXT(cpu_flags)) |
652f5185 |
av_log(c, AV_LOG_INFO, "using MMXEXT\n"); |
e0c6cce4 |
else if (INLINE_AMD3DNOW(cpu_flags)) |
ef0ee7f6 |
av_log(c, AV_LOG_INFO, "using 3DNOW\n"); |
e0c6cce4 |
else if (INLINE_MMX(cpu_flags)) |
ef0ee7f6 |
av_log(c, AV_LOG_INFO, "using MMX\n");
else if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC)
av_log(c, AV_LOG_INFO, "using AltiVec\n");
else
av_log(c, AV_LOG_INFO, "using C\n"); |
a4388ebd |
av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH); |
ef0ee7f6 |
av_log(c, AV_LOG_DEBUG,
"lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", |
a4388ebd |
c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc); |
ef0ee7f6 |
av_log(c, AV_LOG_DEBUG,
"chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH,
c->chrXInc, c->chrYInc); |
a4388ebd |
}
|
ef0ee7f6 |
c->swScale = ff_getSwsFunc(c); |
635d4aed |
return 0; |
ef0ee7f6 |
fail: // FIXME replace things by appropriate error codes |
635d4aed |
return -1;
} |
a4388ebd |
|
0810a584 |
#if FF_API_SWS_GETCONTEXT |
716d413c |
SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
int dstW, int dstH, enum AVPixelFormat dstFormat, |
ef0ee7f6 |
int flags, SwsFilter *srcFilter,
SwsFilter *dstFilter, const double *param) |
635d4aed |
{
SwsContext *c;
|
ef0ee7f6 |
if (!(c = sws_alloc_context())) |
635d4aed |
return NULL;
|
ef0ee7f6 |
c->flags = flags;
c->srcW = srcW;
c->srcH = srcH;
c->dstW = dstW;
c->dstH = dstH;
c->srcRange = handle_jpeg(&srcFormat);
c->dstRange = handle_jpeg(&dstFormat); |
1c910d2f |
c->src0Alpha = handle_0alpha(&srcFormat);
c->dst0Alpha = handle_0alpha(&dstFormat); |
ef0ee7f6 |
c->srcFormat = srcFormat;
c->dstFormat = dstFormat; |
635d4aed |
if (param) {
c->param[0] = param[0];
c->param[1] = param[1];
} |
ef0ee7f6 |
sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange,
ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/,
c->dstRange, 0, 1 << 16, 1 << 16); |
635d4aed |
|
ef0ee7f6 |
if (sws_init_context(c, srcFilter, dstFilter) < 0) { |
635d4aed |
sws_freeContext(c);
return NULL;
}
return c; |
a4388ebd |
} |
0810a584 |
#endif |
a4388ebd |
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
float lumaSharpen, float chromaSharpen,
float chromaHShift, float chromaVShift,
int verbose)
{ |
ef0ee7f6 |
SwsFilter *filter = av_malloc(sizeof(SwsFilter)); |
a4388ebd |
if (!filter)
return NULL;
|
ef0ee7f6 |
if (lumaGBlur != 0.0) {
filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0); |
a4388ebd |
} else { |
ef0ee7f6 |
filter->lumH = sws_getIdentityVec();
filter->lumV = sws_getIdentityVec(); |
a4388ebd |
}
|
ef0ee7f6 |
if (chromaGBlur != 0.0) {
filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0); |
a4388ebd |
} else { |
ef0ee7f6 |
filter->chrH = sws_getIdentityVec();
filter->chrV = sws_getIdentityVec(); |
a4388ebd |
}
|
ef0ee7f6 |
if (chromaSharpen != 0.0) {
SwsVector *id = sws_getIdentityVec(); |
a4388ebd |
sws_scaleVec(filter->chrH, -chromaSharpen);
sws_scaleVec(filter->chrV, -chromaSharpen);
sws_addVec(filter->chrH, id);
sws_addVec(filter->chrV, id);
sws_freeVec(id);
}
|
ef0ee7f6 |
if (lumaSharpen != 0.0) {
SwsVector *id = sws_getIdentityVec(); |
a4388ebd |
sws_scaleVec(filter->lumH, -lumaSharpen);
sws_scaleVec(filter->lumV, -lumaSharpen);
sws_addVec(filter->lumH, id);
sws_addVec(filter->lumV, id);
sws_freeVec(id);
}
if (chromaHShift != 0.0) |
ef0ee7f6 |
sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5)); |
a4388ebd |
if (chromaVShift != 0.0) |
ef0ee7f6 |
sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5)); |
a4388ebd |
sws_normalizeVec(filter->chrH, 1.0);
sws_normalizeVec(filter->chrV, 1.0);
sws_normalizeVec(filter->lumH, 1.0);
sws_normalizeVec(filter->lumV, 1.0);
|
ef0ee7f6 |
if (verbose)
sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
if (verbose)
sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG); |
a4388ebd |
return filter;
}
SwsVector *sws_allocVec(int length)
{ |
fe573d1a |
SwsVector *vec;
if(length <= 0 || length > INT_MAX/ sizeof(double))
return NULL;
vec = av_malloc(sizeof(SwsVector)); |
a4388ebd |
if (!vec)
return NULL;
vec->length = length;
vec->coeff = av_malloc(sizeof(double) * length);
if (!vec->coeff)
av_freep(&vec);
return vec;
}
SwsVector *sws_getGaussianVec(double variance, double quality)
{ |
ef0ee7f6 |
const int length = (int)(variance * quality + 0.5) | 1; |
a4388ebd |
int i; |
ef0ee7f6 |
double middle = (length - 1) * 0.5; |
e823e736 |
SwsVector *vec;
if(variance < 0 || quality < 0)
return NULL;
vec = sws_allocVec(length); |
a4388ebd |
if (!vec)
return NULL;
|
ef0ee7f6 |
for (i = 0; i < length; i++) {
double dist = i - middle;
vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
sqrt(2 * variance * M_PI); |
a4388ebd |
}
sws_normalizeVec(vec, 1.0);
return vec;
}
SwsVector *sws_getConstVec(double c, int length)
{
int i; |
ef0ee7f6 |
SwsVector *vec = sws_allocVec(length); |
a4388ebd |
if (!vec)
return NULL;
|
ef0ee7f6 |
for (i = 0; i < length; i++)
vec->coeff[i] = c; |
a4388ebd |
return vec;
}
SwsVector *sws_getIdentityVec(void)
{
return sws_getConstVec(1.0, 1);
}
|
627686e6 |
static double sws_dcVec(SwsVector *a) |
a4388ebd |
{
int i; |
ef0ee7f6 |
double sum = 0; |
a4388ebd |
|
ef0ee7f6 |
for (i = 0; i < a->length; i++)
sum += a->coeff[i]; |
a4388ebd |
return sum;
}
void sws_scaleVec(SwsVector *a, double scalar)
{
int i;
|
ef0ee7f6 |
for (i = 0; i < a->length; i++)
a->coeff[i] *= scalar; |
a4388ebd |
}
void sws_normalizeVec(SwsVector *a, double height)
{ |
ef0ee7f6 |
sws_scaleVec(a, height / sws_dcVec(a)); |
a4388ebd |
}
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
{ |
ef0ee7f6 |
int length = a->length + b->length - 1; |
a4388ebd |
int i, j; |
ef0ee7f6 |
SwsVector *vec = sws_getConstVec(0.0, length); |
a4388ebd |
if (!vec)
return NULL;
|
ef0ee7f6 |
for (i = 0; i < a->length; i++) {
for (j = 0; j < b->length; j++) {
vec->coeff[i + j] += a->coeff[i] * b->coeff[j]; |
a4388ebd |
}
}
return vec;
}
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
{ |
ef0ee7f6 |
int length = FFMAX(a->length, b->length); |
a4388ebd |
int i; |
ef0ee7f6 |
SwsVector *vec = sws_getConstVec(0.0, length); |
a4388ebd |
if (!vec)
return NULL;
|
ef0ee7f6 |
for (i = 0; i < a->length; i++)
vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
for (i = 0; i < b->length; i++)
vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i]; |
a4388ebd |
return vec;
}
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
{ |
ef0ee7f6 |
int length = FFMAX(a->length, b->length); |
a4388ebd |
int i; |
ef0ee7f6 |
SwsVector *vec = sws_getConstVec(0.0, length); |
a4388ebd |
if (!vec)
return NULL;
|
ef0ee7f6 |
for (i = 0; i < a->length; i++)
vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
for (i = 0; i < b->length; i++)
vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] -= b->coeff[i]; |
a4388ebd |
return vec;
}
/* shift left / or right if "shift" is negative */
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
{ |
ef0ee7f6 |
int length = a->length + FFABS(shift) * 2; |
a4388ebd |
int i; |
ef0ee7f6 |
SwsVector *vec = sws_getConstVec(0.0, length); |
a4388ebd |
if (!vec)
return NULL;
|
ef0ee7f6 |
for (i = 0; i < a->length; i++) {
vec->coeff[i + (length - 1) / 2 -
(a->length - 1) / 2 - shift] = a->coeff[i]; |
a4388ebd |
}
return vec;
}
void sws_shiftVec(SwsVector *a, int shift)
{ |
ef0ee7f6 |
SwsVector *shifted = sws_getShiftedVec(a, shift); |
a4388ebd |
av_free(a->coeff); |
ef0ee7f6 |
a->coeff = shifted->coeff;
a->length = shifted->length; |
a4388ebd |
av_free(shifted);
}
void sws_addVec(SwsVector *a, SwsVector *b)
{ |
ef0ee7f6 |
SwsVector *sum = sws_sumVec(a, b); |
a4388ebd |
av_free(a->coeff); |
ef0ee7f6 |
a->coeff = sum->coeff;
a->length = sum->length; |
a4388ebd |
av_free(sum);
}
void sws_subVec(SwsVector *a, SwsVector *b)
{ |
ef0ee7f6 |
SwsVector *diff = sws_diffVec(a, b); |
a4388ebd |
av_free(a->coeff); |
ef0ee7f6 |
a->coeff = diff->coeff;
a->length = diff->length; |
a4388ebd |
av_free(diff);
}
void sws_convVec(SwsVector *a, SwsVector *b)
{ |
ef0ee7f6 |
SwsVector *conv = sws_getConvVec(a, b); |
a4388ebd |
av_free(a->coeff); |
ef0ee7f6 |
a->coeff = conv->coeff;
a->length = conv->length; |
a4388ebd |
av_free(conv);
}
SwsVector *sws_cloneVec(SwsVector *a)
{
int i; |
ef0ee7f6 |
SwsVector *vec = sws_allocVec(a->length); |
a4388ebd |
if (!vec)
return NULL;
|
ef0ee7f6 |
for (i = 0; i < a->length; i++)
vec->coeff[i] = a->coeff[i]; |
a4388ebd |
return vec;
}
void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
{
int i; |
ef0ee7f6 |
double max = 0;
double min = 0; |
a4388ebd |
double range;
|
ef0ee7f6 |
for (i = 0; i < a->length; i++)
if (a->coeff[i] > max)
max = a->coeff[i]; |
a4388ebd |
|
ef0ee7f6 |
for (i = 0; i < a->length; i++)
if (a->coeff[i] < min)
min = a->coeff[i]; |
a4388ebd |
|
ef0ee7f6 |
range = max - min; |
a4388ebd |
|
ef0ee7f6 |
for (i = 0; i < a->length; i++) {
int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5); |
a4388ebd |
av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]); |
ef0ee7f6 |
for (; x > 0; x--)
av_log(log_ctx, log_level, " "); |
a4388ebd |
av_log(log_ctx, log_level, "|\n");
}
}
void sws_freeVec(SwsVector *a)
{ |
ef0ee7f6 |
if (!a)
return; |
a4388ebd |
av_freep(&a->coeff); |
ef0ee7f6 |
a->length = 0; |
a4388ebd |
av_free(a);
}
void sws_freeFilter(SwsFilter *filter)
{ |
ef0ee7f6 |
if (!filter)
return;
if (filter->lumH)
sws_freeVec(filter->lumH);
if (filter->lumV)
sws_freeVec(filter->lumV);
if (filter->chrH)
sws_freeVec(filter->chrH);
if (filter->chrV)
sws_freeVec(filter->chrV); |
a4388ebd |
av_free(filter);
}
void sws_freeContext(SwsContext *c)
{
int i; |
ef0ee7f6 |
if (!c)
return; |
a4388ebd |
if (c->lumPixBuf) { |
ef0ee7f6 |
for (i = 0; i < c->vLumBufSize; i++) |
a4388ebd |
av_freep(&c->lumPixBuf[i]);
av_freep(&c->lumPixBuf);
}
|
986f0d86 |
if (c->chrUPixBuf) { |
ef0ee7f6 |
for (i = 0; i < c->vChrBufSize; i++) |
986f0d86 |
av_freep(&c->chrUPixBuf[i]);
av_freep(&c->chrUPixBuf);
av_freep(&c->chrVPixBuf); |
a4388ebd |
}
if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) { |
ef0ee7f6 |
for (i = 0; i < c->vLumBufSize; i++) |
a4388ebd |
av_freep(&c->alpPixBuf[i]);
av_freep(&c->alpPixBuf);
}
av_freep(&c->vLumFilter);
av_freep(&c->vChrFilter);
av_freep(&c->hLumFilter);
av_freep(&c->hChrFilter); |
f684f3c5 |
#if HAVE_ALTIVEC |
a4388ebd |
av_freep(&c->vYCoeffsBank);
av_freep(&c->vCCoeffsBank);
#endif
av_freep(&c->vLumFilterPos);
av_freep(&c->vChrFilterPos);
av_freep(&c->hLumFilterPos);
av_freep(&c->hChrFilterPos);
|
17337f54 |
#if HAVE_MMX_INLINE |
95cd815c |
#if USE_MMAP |
a65bdceb |
if (c->lumMmxextFilterCode)
munmap(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize);
if (c->chrMmxextFilterCode)
munmap(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize); |
a4388ebd |
#elif HAVE_VIRTUALALLOC |
a65bdceb |
if (c->lumMmxextFilterCode)
VirtualFree(c->lumMmxextFilterCode, 0, MEM_RELEASE);
if (c->chrMmxextFilterCode)
VirtualFree(c->chrMmxextFilterCode, 0, MEM_RELEASE); |
a4388ebd |
#else |
a65bdceb |
av_free(c->lumMmxextFilterCode);
av_free(c->chrMmxextFilterCode); |
a4388ebd |
#endif |
a65bdceb |
c->lumMmxextFilterCode = NULL;
c->chrMmxextFilterCode = NULL; |
17337f54 |
#endif /* HAVE_MMX_INLINE */ |
a4388ebd |
av_freep(&c->yuvTable); |
485d73ef |
av_freep(&c->formatConvBuffer); |
a4388ebd |
av_free(c);
}
|
ef0ee7f6 |
struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW, |
716d413c |
int srcH, enum AVPixelFormat srcFormat, |
ef0ee7f6 |
int dstW, int dstH, |
716d413c |
enum AVPixelFormat dstFormat, int flags, |
ef0ee7f6 |
SwsFilter *srcFilter,
SwsFilter *dstFilter,
const double *param) |
a4388ebd |
{ |
ef0ee7f6 |
static const double default_param[2] = { SWS_PARAM_DEFAULT,
SWS_PARAM_DEFAULT }; |
a4388ebd |
if (!param)
param = default_param;
|
5a011d8b |
if (context && |
22b6a24c |
(context->srcW != srcW ||
context->srcH != srcH ||
context->srcFormat != srcFormat ||
context->dstW != dstW ||
context->dstH != dstH ||
context->dstFormat != dstFormat ||
context->flags != flags ||
context->param[0] != param[0] || |
b623d0cb |
context->param[1] != param[1])) {
sws_freeContext(context);
context = NULL;
} |
5a011d8b |
|
a4388ebd |
if (!context) { |
0810a584 |
if (!(context = sws_alloc_context()))
return NULL;
context->srcW = srcW;
context->srcH = srcH; |
9c158e49 |
context->srcRange = handle_jpeg(&srcFormat); |
1c910d2f |
context->src0Alpha = handle_0alpha(&srcFormat); |
0810a584 |
context->srcFormat = srcFormat; |
12eef0d1 |
context->dstW = dstW;
context->dstH = dstH; |
9c158e49 |
context->dstRange = handle_jpeg(&dstFormat); |
1c910d2f |
context->dst0Alpha = handle_0alpha(&dstFormat); |
0810a584 |
context->dstFormat = dstFormat;
context->flags = flags;
context->param[0] = param[0];
context->param[1] = param[1]; |
ef0ee7f6 |
sws_setColorspaceDetails(context, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT],
context->srcRange,
ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/,
context->dstRange, 0, 1 << 16, 1 << 16); |
0810a584 |
if (sws_init_context(context, srcFilter, dstFilter) < 0) {
sws_freeContext(context);
return NULL;
} |
a4388ebd |
}
return context;
} |