Skip to content

F
ffmpeg.wasm-core
Commit 367d9b29 authored by Michael Niedermayer's avatar Michael Niedermayer
Browse files
Options

Merge remote-tracking branch 'qatar/master' 

* qatar/master:
  swscale: K&R formatting cosmetics (part II)
  tiffdec: Add a malloc check and refactor another.
  faxcompr: Check malloc results and unify return path
  configure: escape colons in values written to config.fate
  ac3dsp: call femms/emms at the end of float_to_fixed24() for 3DNow and SSE
  matroska: Fix leaking memory allocated for laces.
  pthread: Fix crash due to fctx->delaying not being cleared.
  vp3: Assert on invalid filter_limit values.
  h264: fix 10bit biweight functions after recent x86inc.asm fixes.
  ffv1: Fix size mismatch in encode_line.
  movenc: Remove a dead initialization
  git-howto: Explain how to avoid Windows line endings in git checkouts.
  build: Move all arch OBJS declarations into arch subdirectory Makefiles.

Conflicts:
	configure
	libavcodec/vp3.c
	libavformat/matroskadec.c
	libavutil/Makefile
	libswscale/Makefile
	libswscale/swscale.c
	libswscale/swscale_internal.h
	libswscale/utils.c
Merged-by: 's avatarMichael Niedermayer <michaelni@gmx.at>
parents 62e5ef95 ef0ee7f6
Hide whitespace changes
Inline Side-by-side
Showing with 1187 additions and 962 deletions
configure
View file @ 367d9b29
......@@ -2803,7 +2803,11 @@ case $target_os in
;;
esac
echo "config:$arch:$subarch:$cpu:$target_os:$cc_ident:$FFMPEG_CONFIGURATION" >config.fate
esc(){
echo "$*" | sed 's/%/%25/g;s/:/%3a/g'
}
echo "config:$arch:$subarch:$cpu:$target_os:$(esc $cc_ident):$(esc $FFMPEG_CONFIGURATION)" >config.fate
check_cpp_condition stdlib.h "defined(__PIC__) || defined(__pic__) || defined(PIC)" && enable pic
......
doc/git-howto.texi
View file @ 367d9b29
......@@ -65,6 +65,14 @@ git clone git@@source.ffmpeg.org:ffmpeg <target>
This will put the FFmpeg sources into the directory @var{<target>} and let
you push back your changes to the remote repository.
Make sure that you do not have Windows line endings in your checkouts,
otherwise you may experience spurious compilation failures. One way to
achieve this is to run
@example
git config --global core.autocrlf false
@end example
@section Updating the source tree to the latest revision
......
libavcodec/faxcompr.c
View file @ 367d9b29
......@@ -275,12 +275,17 @@ int ff_ccitt_unpack(AVCodecContext *avctx,
{
int j;
GetBitContext gb;
int *runs, *ref, *runend;
int *runs, *ref = NULL, *runend;
int ret;
int runsize= avctx->width + 2;
int err = 0;
runs = av_malloc(runsize * sizeof(runs[0]));
ref = av_malloc(runsize * sizeof(ref[0]));
if (!runs || ! ref) {
err = AVERROR(ENOMEM);
goto fail;
}
ref[0] = avctx->width;
ref[1] = 0;
ref[2] = 0;
......@@ -290,9 +295,8 @@ int ff_ccitt_unpack(AVCodecContext *avctx,
if(compr == TIFF_G4){
ret = decode_group3_2d_line(avctx, &gb, avctx->width, runs, runend, ref);
if(ret < 0){
av_free(runs);
av_free(ref);
return -1;
err = -1;
goto fail;
}
}else{
int g3d1 = (compr == TIFF_G3) && !(opts & 1);
......@@ -313,7 +317,8 @@ int ff_ccitt_unpack(AVCodecContext *avctx,
}
dst += stride;
}
fail:
av_free(runs);
av_free(ref);
return 0;
return err;
}
libavcodec/ffv1.c
View file @ 367d9b29
......@@ -445,7 +445,7 @@ static inline int get_vlc_symbol(GetBitContext *gb, VlcState * const state, int
#if CONFIG_FFV1_ENCODER
static av_always_inline int encode_line(FFV1Context *s, int w,
int16_t *sample[2],
int16_t *sample[3],
int plane_index, int bits)
{
PlaneContext * const p= &s->plane[plane_index];
......
libavcodec/tiff.c
View file @ 367d9b29
......@@ -145,13 +145,18 @@ static int tiff_unpack_strip(TiffContext *s, uint8_t* dst, int stride, const uin
int c, line, pixels, code;
const uint8_t *ssrc = src;
int width = ((s->width * s->bpp) + 7) >> 3;
#if CONFIG_ZLIB
uint8_t *zbuf; unsigned long outlen;
if (size <= 0)
return AVERROR_INVALIDDATA;
#if CONFIG_ZLIB
if(s->compr == TIFF_DEFLATE || s->compr == TIFF_ADOBE_DEFLATE){
uint8_t *zbuf; unsigned long outlen;
int ret;
outlen = width * lines;
zbuf = av_malloc(outlen);
if (!zbuf)
return AVERROR(ENOMEM);
ret = tiff_uncompress(zbuf, &outlen, src, size);
if(ret != Z_OK){
av_log(s->avctx, AV_LOG_ERROR, "Uncompressing failed (%lu of %lu) with error %d\n", outlen, (unsigned long)width * lines, ret);
......@@ -180,11 +185,11 @@ static int tiff_unpack_strip(TiffContext *s, uint8_t* dst, int stride, const uin
}
if(s->compr == TIFF_CCITT_RLE || s->compr == TIFF_G3 || s->compr == TIFF_G4){
int i, ret = 0;
uint8_t *src2 = av_malloc(size + FF_INPUT_BUFFER_PADDING_SIZE);
uint8_t *src2 = av_malloc((unsigned)size + FF_INPUT_BUFFER_PADDING_SIZE);
if(!src2 || (unsigned)size + FF_INPUT_BUFFER_PADDING_SIZE < (unsigned)size){
if (!src2) {
av_log(s->avctx, AV_LOG_ERROR, "Error allocating temporary buffer\n");
return -1;
return AVERROR(ENOMEM);
}
if(s->fax_opts & 2){
av_log(s->avctx, AV_LOG_ERROR, "Uncompressed fax mode is not supported (yet)\n");
......
libavcodec/x86/ac3dsp.asm
View file @ 367d9b29
......@@ -223,7 +223,8 @@ cglobal float_to_fixed24_3dnow, 3,3,0, dst, src, len
add dstq, 32
sub lend, 8
ja .loop
REP_RET
femms
RET
INIT_XMM
cglobal float_to_fixed24_sse, 3,3,3, dst, src, len
......@@ -247,7 +248,8 @@ cglobal float_to_fixed24_sse, 3,3,3, dst, src, len
add dstq, 32
sub lend, 8
ja .loop
REP_RET
emms
RET
INIT_XMM
cglobal float_to_fixed24_sse2, 3,3,9, dst, src, len
......
libavcodec/x86/h264_weight_10bit.asm
View file @ 367d9b29
......@@ -155,7 +155,7 @@ WEIGHT_FUNC_HALF_MM sse4
%if ARCH_X86_32
DECLARE_REG_TMP 3
%else
DECLARE_REG_TMP 10
DECLARE_REG_TMP 7
%endif
%macro BIWEIGHT_PROLOGUE 0
......@@ -218,7 +218,7 @@ DECLARE_REG_TMP 10
%endmacro
%macro BIWEIGHT_FUNC_DBL 1
cglobal h264_biweight_16_10_%1
cglobal h264_biweight_16_10_%1, 0, 8, 8
BIWEIGHT_PROLOGUE
BIWEIGHT_SETUP %1
.nextrow
......@@ -238,7 +238,7 @@ BIWEIGHT_FUNC_DBL sse2
BIWEIGHT_FUNC_DBL sse4
%macro BIWEIGHT_FUNC 1
cglobal h264_biweight_8_10_%1
cglobal h264_biweight_8_10_%1, 0, 8, 8
BIWEIGHT_PROLOGUE
BIWEIGHT_SETUP %1
.nextrow
......@@ -256,7 +256,7 @@ BIWEIGHT_FUNC sse2
BIWEIGHT_FUNC sse4
%macro BIWEIGHT_FUNC_HALF 1
cglobal h264_biweight_4_10_%1
cglobal h264_biweight_4_10_%1, 0, 8, 8
BIWEIGHT_PROLOGUE
BIWEIGHT_SETUP %1
sar r3d, 1
......
libavformat/matroskadec.c
View file @ 367d9b29
......@@ -1940,8 +1940,8 @@ static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
if (size < cfs * h / 2) {
av_log(matroska->ctx, AV_LOG_ERROR,
"Corrupt int4 RM-style audio packet size\n");
av_free(lace_size);
return AVERROR_INVALIDDATA;
res = AVERROR_INVALIDDATA;
goto end;
}
for (x=0; x<h/2; x++)
memcpy(track->audio.buf+x*2*w+y*cfs,
......@@ -1950,16 +1950,16 @@ static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
if (size < w) {
av_log(matroska->ctx, AV_LOG_ERROR,
"Corrupt sipr RM-style audio packet size\n");
av_free(lace_size);
return AVERROR_INVALIDDATA;
res = AVERROR_INVALIDDATA;
goto end;
}
memcpy(track->audio.buf + y*w, data, w);
} else {
if (size < sps * w / sps) {
av_log(matroska->ctx, AV_LOG_ERROR,
"Corrupt generic RM-style audio packet size\n");
av_free(lace_size);
return AVERROR_INVALIDDATA;
res = AVERROR_INVALIDDATA;
goto end;
}
for (x=0; x<w/sps; x++)
memcpy(track->audio.buf+sps*(h*x+((h+1)/2)*(y&1)+(y>>1)), data+x*sps, sps);
......@@ -2049,6 +2049,7 @@ static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
}
}
end:
av_free(lace_size);
return res;
}
......
libavformat/movenc.c
View file @ 367d9b29
......@@ -939,7 +939,7 @@ static const AVCodecTag codec_3gp_tags[] = {
static int mov_find_codec_tag(AVFormatContext *s, MOVTrack *track)
{
int tag = track->enc->codec_tag;
int tag;
if (track->mode == MODE_MP4 || track->mode == MODE_PSP)
tag = mp4_get_codec_tag(s, track);
......
libavutil/Makefile
View file @ 367d9b29
......@@ -79,9 +79,6 @@ OBJS = adler32.o \
tree.o \
utils.o \
OBJS-$(ARCH_PPC) += ppc/cpu.o
OBJS-$(ARCH_X86) += x86/cpu.o
TESTPROGS = adler32 aes avstring base64 bprint cpu crc des eval file fifo \
lfg lls md5 opt pca parseutils random_seed rational sha tree
TESTPROGS-$(HAVE_LZO1X_999_COMPRESS) += lzo
......
libavutil/ppc/Makefile 0 → 100644
View file @ 367d9b29
OBJS += ppc/cpu.o \
libavutil/x86/Makefile 0 → 100644
View file @ 367d9b29
OBJS += x86/cpu.o \
libswscale/Makefile
View file @ 367d9b29
......@@ -14,22 +14,4 @@ OBJS = input.o \
utils.o \
yuv2rgb.o \
OBJS-$(ARCH_BFIN) += bfin/internal_bfin.o \
bfin/swscale_bfin.o \
bfin/yuv2rgb_bfin.o
ALTIVEC-OBJS += ppc/swscale_altivec.o \
ppc/yuv2rgb_altivec.o \
ppc/yuv2yuv_altivec.o
MMX-OBJS += x86/rgb2rgb.o \
x86/swscale_mmx.o \
x86/yuv2rgb_mmx.o
VIS-OBJS += sparc/yuv2rgb_vis.o
YASM-OBJS += x86/input.o \
x86/output.o \
x86/scale.o
$(SUBDIR)x86/swscale_mmx.o: CFLAGS += $(NOREDZONE_FLAGS)
OBJS-$(CONFIG_XMM_CLOBBER_TEST) += x86/w64xmmtest.o
TESTPROGS = colorspace swscale
libswscale/bfin/Makefile 0 → 100644
View file @ 367d9b29
OBJS += bfin/internal_bfin.o \
bfin/swscale_bfin.o \
bfin/yuv2rgb_bfin.o \
libswscale/ppc/Makefile 0 → 100644
View file @ 367d9b29
ALTIVEC-OBJS += ppc/swscale_altivec.o \
ppc/yuv2rgb_altivec.o \
ppc/yuv2yuv_altivec.o \
libswscale/sparc/Makefile 0 → 100644
View file @ 367d9b29
VIS-OBJS += sparc/yuv2rgb_vis.o \
libswscale/swscale.c
View file @ 367d9b29
......@@ -18,58 +18,59 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <assert.h>
#include <inttypes.h>
#include <string.h>
#include <math.h>
#include <stdio.h>
#include "config.h"
#include <assert.h>
#include "swscale.h"
#include "swscale_internal.h"
#include "rgb2rgb.h"
#include <string.h>
#include "libavutil/avassert.h"
#include "libavutil/intreadwrite.h"
#include "libavutil/cpu.h"
#include "libavutil/avutil.h"
#include "libavutil/mathematics.h"
#include "libavutil/bswap.h"
#include "libavutil/cpu.h"
#include "libavutil/intreadwrite.h"
#include "libavutil/mathematics.h"
#include "libavutil/pixdesc.h"
#include "config.h"
#include "rgb2rgb.h"
#include "swscale_internal.h"
#include "swscale.h"
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_128)[8][8] = {
{ 36, 68, 60, 92, 34, 66, 58, 90,},
{ 100, 4,124, 28, 98, 2,122, 26,},
{ 52, 84, 44, 76, 50, 82, 42, 74,},
{ 116, 20,108, 12,114, 18,106, 10,},
{ 32, 64, 56, 88, 38, 70, 62, 94,},
{ 96, 0,120, 24,102, 6,126, 30,},
{ 48, 80, 40, 72, 54, 86, 46, 78,},
{ 112, 16,104, 8,118, 22,110, 14,},
{ 36, 68, 60, 92, 34, 66, 58, 90, },
{ 100, 4, 124, 28, 98, 2, 122, 26, },
{ 52, 84, 44, 76, 50, 82, 42, 74, },
{ 116, 20, 108, 12, 114, 18, 106, 10, },
{ 32, 64, 56, 88, 38, 70, 62, 94, },
{ 96, 0, 120, 24, 102, 6, 126, 30, },
{ 48, 80, 40, 72, 54, 86, 46, 78, },
{ 112, 16, 104, 8, 118, 22, 110, 14, },
};
DECLARE_ALIGNED(8, const uint8_t, ff_sws_pb_64)[8] =
{ 64, 64, 64, 64, 64, 64, 64, 64 };
DECLARE_ALIGNED(8, const uint8_t, ff_sws_pb_64)[8] = {
64, 64, 64, 64, 64, 64, 64, 64
};
static av_always_inline void fillPlane(uint8_t* plane, int stride,
int width, int height,
int y, uint8_t val)
static av_always_inline void fillPlane(uint8_t *plane, int stride, int width,
int height, int y, uint8_t val)
{
int i;
uint8_t *ptr = plane + stride*y;
for (i=0; i<height; i++) {
uint8_t *ptr = plane + stride * y;
for (i = 0; i < height; i++) {
memset(ptr, val, width);
ptr += stride;
}
}
static void hScale16To19_c(SwsContext *c, int16_t *_dst, int dstW, const uint8_t *_src,
const int16_t *filter,
static void hScale16To19_c(SwsContext *c, int16_t *_dst, int dstW,
const uint8_t *_src, const int16_t *filter,
const int32_t *filterPos, int filterSize)
{
int i;
int32_t *dst = (int32_t *) _dst;
int32_t *dst = (int32_t *) _dst;
const uint16_t *src = (const uint16_t *) _src;
int bits = av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1;
int sh = bits - 4;
int bits = av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1;
int sh = bits - 4;
if((isAnyRGB(c->srcFormat) || c->srcFormat==PIX_FMT_PAL8) && av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1<15)
sh= 9;
......@@ -77,7 +78,7 @@ static void hScale16To19_c(SwsContext *c, int16_t *_dst, int dstW, const uint8_t
for (i = 0; i < dstW; i++) {
int j;
int srcPos = filterPos[i];
int val = 0;
int val = 0;
for (j = 0; j < filterSize; j++) {
val += src[srcPos + j] * filter[filterSize * i + j];
......@@ -87,13 +88,13 @@ static void hScale16To19_c(SwsContext *c, int16_t *_dst, int dstW, const uint8_t
}
}
static void hScale16To15_c(SwsContext *c, int16_t *dst, int dstW, const uint8_t *_src,
const int16_t *filter,
static void hScale16To15_c(SwsContext *c, int16_t *dst, int dstW,
const uint8_t *_src, const int16_t *filter,
const int32_t *filterPos, int filterSize)
{
int i;
const uint16_t *src = (const uint16_t *) _src;
int sh = av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1;
int sh = av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1;
if(sh<15)
sh= isAnyRGB(c->srcFormat) || c->srcFormat==PIX_FMT_PAL8 ? 13 : av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1;
......@@ -101,7 +102,7 @@ static void hScale16To15_c(SwsContext *c, int16_t *dst, int dstW, const uint8_t
for (i = 0; i < dstW; i++) {
int j;
int srcPos = filterPos[i];
int val = 0;
int val = 0;
for (j = 0; j < filterSize; j++) {
val += src[srcPos + j] * filter[filterSize * i + j];
......@@ -112,72 +113,71 @@ static void hScale16To15_c(SwsContext *c, int16_t *dst, int dstW, const uint8_t
}
// bilinear / bicubic scaling
static void hScale8To15_c(SwsContext *c, int16_t *dst, int dstW, const uint8_t *src,
const int16_t *filter, const int32_t *filterPos,
int filterSize)
static void hScale8To15_c(SwsContext *c, int16_t *dst, int dstW,
const uint8_t *src, const int16_t *filter,
const int32_t *filterPos, int filterSize)
{
int i;
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
int j;
int srcPos= filterPos[i];
int val=0;
for (j=0; j<filterSize; j++) {
val += ((int)src[srcPos + j])*filter[filterSize*i + j];
int srcPos = filterPos[i];
int val = 0;
for (j = 0; j < filterSize; j++) {
val += ((int)src[srcPos + j]) * filter[filterSize * i + j];
}
//filter += hFilterSize;
dst[i] = FFMIN(val>>7, (1<<15)-1); // the cubic equation does overflow ...
//dst[i] = val>>7;
dst[i] = FFMIN(val >> 7, (1 << 15) - 1); // the cubic equation does overflow ...
}
}
static void hScale8To19_c(SwsContext *c, int16_t *_dst, int dstW, const uint8_t *src,
const int16_t *filter, const int32_t *filterPos,
int filterSize)
static void hScale8To19_c(SwsContext *c, int16_t *_dst, int dstW,
const uint8_t *src, const int16_t *filter,
const int32_t *filterPos, int filterSize)
{
int i;
int32_t *dst = (int32_t *) _dst;
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
int j;
int srcPos= filterPos[i];
int val=0;
for (j=0; j<filterSize; j++) {
val += ((int)src[srcPos + j])*filter[filterSize*i + j];
int srcPos = filterPos[i];
int val = 0;
for (j = 0; j < filterSize; j++) {
val += ((int)src[srcPos + j]) * filter[filterSize * i + j];
}
//filter += hFilterSize;
dst[i] = FFMIN(val>>3, (1<<19)-1); // the cubic equation does overflow ...
//dst[i] = val>>7;
dst[i] = FFMIN(val >> 3, (1 << 19) - 1); // the cubic equation does overflow ...
}
}
//FIXME all pal and rgb srcFormats could do this convertion as well
//FIXME all scalers more complex than bilinear could do half of this transform
// FIXME all pal and rgb srcFormats could do this convertion as well
// FIXME all scalers more complex than bilinear could do half of this transform
static void chrRangeToJpeg_c(int16_t *dstU, int16_t *dstV, int width)
{
int i;
for (i = 0; i < width; i++) {
dstU[i] = (FFMIN(dstU[i],30775)*4663 - 9289992)>>12; //-264
dstV[i] = (FFMIN(dstV[i],30775)*4663 - 9289992)>>12; //-264
dstU[i] = (FFMIN(dstU[i], 30775) * 4663 - 9289992) >> 12; // -264
dstV[i] = (FFMIN(dstV[i], 30775) * 4663 - 9289992) >> 12; // -264
}
}
static void chrRangeFromJpeg_c(int16_t *dstU, int16_t *dstV, int width)
{
int i;
for (i = 0; i < width; i++) {
dstU[i] = (dstU[i]*1799 + 4081085)>>11; //1469
dstV[i] = (dstV[i]*1799 + 4081085)>>11; //1469
dstU[i] = (dstU[i] * 1799 + 4081085) >> 11; // 1469
dstV[i] = (dstV[i] * 1799 + 4081085) >> 11; // 1469
}
}
static void lumRangeToJpeg_c(int16_t *dst, int width)
{
int i;
for (i = 0; i < width; i++)
dst[i] = (FFMIN(dst[i],30189)*19077 - 39057361)>>14;
dst[i] = (FFMIN(dst[i], 30189) * 19077 - 39057361) >> 14;
}
static void lumRangeFromJpeg_c(int16_t *dst, int width)
{
int i;
for (i = 0; i < width; i++)
dst[i] = (dst[i]*14071 + 33561947)>>14;
dst[i] = (dst[i] * 14071 + 33561947) >> 14;
}
static void chrRangeToJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)
......@@ -186,27 +186,30 @@ static void chrRangeToJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)
int32_t *dstU = (int32_t *) _dstU;
int32_t *dstV = (int32_t *) _dstV;
for (i = 0; i < width; i++) {
dstU[i] = (FFMIN(dstU[i],30775<<4)*4663 - (9289992<<4))>>12; //-264
dstV[i] = (FFMIN(dstV[i],30775<<4)*4663 - (9289992<<4))>>12; //-264
dstU[i] = (FFMIN(dstU[i], 30775 << 4) * 4663 - (9289992 << 4)) >> 12; // -264
dstV[i] = (FFMIN(dstV[i], 30775 << 4) * 4663 - (9289992 << 4)) >> 12; // -264
}
}
static void chrRangeFromJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)
{
int i;
int32_t *dstU = (int32_t *) _dstU;
int32_t *dstV = (int32_t *) _dstV;
for (i = 0; i < width; i++) {
dstU[i] = (dstU[i]*1799 + (4081085<<4))>>11; //1469
dstV[i] = (dstV[i]*1799 + (4081085<<4))>>11; //1469
dstU[i] = (dstU[i] * 1799 + (4081085 << 4)) >> 11; // 1469
dstV[i] = (dstV[i] * 1799 + (4081085 << 4)) >> 11; // 1469
}
}
static void lumRangeToJpeg16_c(int16_t *_dst, int width)
{
int i;
int32_t *dst = (int32_t *) _dst;
for (i = 0; i < width; i++)
dst[i] = (FFMIN(dst[i],30189<<4)*4769 - (39057361<<2))>>12;
dst[i] = (FFMIN(dst[i], 30189 << 4) * 4769 - (39057361 << 2)) >> 12;
}
static void lumRangeFromJpeg16_c(int16_t *_dst, int width)
{
int i;
......@@ -219,12 +222,12 @@ static void hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
const uint8_t *src, int srcW, int xInc)
{
int i;
unsigned int xpos=0;
for (i=0;i<dstWidth;i++) {
register unsigned int xx=xpos>>16;
register unsigned int xalpha=(xpos&0xFFFF)>>9;
dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha;
xpos+=xInc;
unsigned int xpos = 0;
for (i = 0; i < dstWidth; i++) {
register unsigned int xx = xpos >> 16;
register unsigned int xalpha = (xpos & 0xFFFF) >> 9;
dst[i] = (src[xx] << 7) + (src[xx + 1] - src[xx]) * xalpha;
xpos += xInc;
}
for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
dst[i] = src[srcW-1]*128;
......@@ -232,26 +235,30 @@ static void hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
// *** horizontal scale Y line to temp buffer
static av_always_inline void hyscale(SwsContext *c, int16_t *dst, int dstWidth,
const uint8_t *src_in[4], int srcW, int xInc,
const uint8_t *src_in[4],
int srcW, int xInc,
const int16_t *hLumFilter,
const int32_t *hLumFilterPos, int hLumFilterSize,
const int32_t *hLumFilterPos,
int hLumFilterSize,
uint8_t *formatConvBuffer,
uint32_t *pal, int isAlpha)
{
void (*toYV12)(uint8_t *, const uint8_t *, const uint8_t *, const uint8_t *, int, uint32_t *) = isAlpha ? c->alpToYV12 : c->lumToYV12;
void (*toYV12)(uint8_t *, const uint8_t *, const uint8_t *, const uint8_t *, int, uint32_t *) =
isAlpha ? c->alpToYV12 : c->lumToYV12;
void (*convertRange)(int16_t *, int) = isAlpha ? NULL : c->lumConvertRange;
const uint8_t *src = src_in[isAlpha ? 3 : 0];
if (toYV12) {
toYV12(formatConvBuffer, src, src_in[1], src_in[2], srcW, pal);
src= formatConvBuffer;
src = formatConvBuffer;
} else if (c->readLumPlanar && !isAlpha) {
c->readLumPlanar(formatConvBuffer, src_in, srcW);
src = formatConvBuffer;
}
if (!c->hyscale_fast) {
c->hyScale(c, dst, dstWidth, src, hLumFilter, hLumFilterPos, hLumFilterSize);
c->hyScale(c, dst, dstWidth, src, hLumFilter,
hLumFilterPos, hLumFilterSize);
} else { // fast bilinear upscale / crap downscale
c->hyscale_fast(c, dst, dstWidth, src, srcW, xInc);
}
......@@ -265,13 +272,13 @@ static void hcscale_fast_c(SwsContext *c, int16_t *dst1, int16_t *dst2,
const uint8_t *src2, int srcW, int xInc)
{
int i;
unsigned int xpos=0;
for (i=0;i<dstWidth;i++) {
register unsigned int xx=xpos>>16;
register unsigned int xalpha=(xpos&0xFFFF)>>9;
dst1[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha);
dst2[i]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha);
xpos+=xInc;
unsigned int xpos = 0;
for (i = 0; i < dstWidth; i++) {
register unsigned int xx = xpos >> 16;
register unsigned int xalpha = (xpos & 0xFFFF) >> 9;
dst1[i] = (src1[xx] * (xalpha ^ 127) + src1[xx + 1] * xalpha);
dst2[i] = (src2[xx] * (xalpha ^ 127) + src2[xx + 1] * xalpha);
xpos += xInc;
}
for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
dst1[i] = src1[srcW-1]*128;
......@@ -279,23 +286,28 @@ static void hcscale_fast_c(SwsContext *c, int16_t *dst1, int16_t *dst2,
}
}
static av_always_inline void hcscale(SwsContext *c, int16_t *dst1, int16_t *dst2, int dstWidth,
static av_always_inline void hcscale(SwsContext *c, int16_t *dst1,
int16_t *dst2, int dstWidth,
const uint8_t *src_in[4],
int srcW, int xInc, const int16_t *hChrFilter,
const int32_t *hChrFilterPos, int hChrFilterSize,
int srcW, int xInc,
const int16_t *hChrFilter,
const int32_t *hChrFilterPos,
int hChrFilterSize,
uint8_t *formatConvBuffer, uint32_t *pal)
{
const uint8_t *src1 = src_in[1], *src2 = src_in[2];
if (c->chrToYV12) {
uint8_t *buf2 = formatConvBuffer + FFALIGN(srcW*2+78, 16);
uint8_t *buf2 = formatConvBuffer +
FFALIGN(srcW*2+78, 16);
c->chrToYV12(formatConvBuffer, buf2, src_in[0], src1, src2, srcW, pal);
src1= formatConvBuffer;
src2= buf2;
} else if (c->readChrPlanar) {
uint8_t *buf2 = formatConvBuffer + FFALIGN(srcW*2+78, 16);
uint8_t *buf2 = formatConvBuffer +
FFALIGN(srcW*2+78, 16);
c->readChrPlanar(formatConvBuffer, buf2, src_in, srcW);
src1= formatConvBuffer;
src2= buf2;
src1 = formatConvBuffer;
src2 = buf2;
}
if (!c->hcscale_fast) {
......@@ -310,88 +322,97 @@ static av_always_inline void hcscale(SwsContext *c, int16_t *dst1, int16_t *dst2
}
#define DEBUG_SWSCALE_BUFFERS 0
#define DEBUG_BUFFERS(...) if (DEBUG_SWSCALE_BUFFERS) av_log(c, AV_LOG_DEBUG, __VA_ARGS__)
#define DEBUG_BUFFERS(...) \
if (DEBUG_SWSCALE_BUFFERS) \
av_log(c, AV_LOG_DEBUG, __VA_ARGS__)
static int swScale(SwsContext *c, const uint8_t* src[],
static int swScale(SwsContext *c, const uint8_t *src[],
int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dst[], int dstStride[])
int srcSliceH, uint8_t *dst[], int dstStride[])
{
/* load a few things into local vars to make the code more readable? and faster */
const int srcW= c->srcW;
const int dstW= c->dstW;
const int dstH= c->dstH;
const int chrDstW= c->chrDstW;
const int chrSrcW= c->chrSrcW;
const int lumXInc= c->lumXInc;
const int chrXInc= c->chrXInc;
const enum PixelFormat dstFormat= c->dstFormat;
const int flags= c->flags;
int32_t *vLumFilterPos= c->vLumFilterPos;
int32_t *vChrFilterPos= c->vChrFilterPos;
int32_t *hLumFilterPos= c->hLumFilterPos;
int32_t *hChrFilterPos= c->hChrFilterPos;
int16_t *hLumFilter= c->hLumFilter;
int16_t *hChrFilter= c->hChrFilter;
int32_t *lumMmxFilter= c->lumMmxFilter;
int32_t *chrMmxFilter= c->chrMmxFilter;
const int vLumFilterSize= c->vLumFilterSize;
const int vChrFilterSize= c->vChrFilterSize;
const int hLumFilterSize= c->hLumFilterSize;
const int hChrFilterSize= c->hChrFilterSize;
int16_t **lumPixBuf= c->lumPixBuf;
int16_t **chrUPixBuf= c->chrUPixBuf;
int16_t **chrVPixBuf= c->chrVPixBuf;
int16_t **alpPixBuf= c->alpPixBuf;
const int vLumBufSize= c->vLumBufSize;
const int vChrBufSize= c->vChrBufSize;
uint8_t *formatConvBuffer= c->formatConvBuffer;
const int chrSrcSliceY= srcSliceY >> c->chrSrcVSubSample;
const int chrSrcSliceH= -((-srcSliceH) >> c->chrSrcVSubSample);
/* load a few things into local vars to make the code more readable?
* and faster */
const int srcW = c->srcW;
const int dstW = c->dstW;
const int dstH = c->dstH;
const int chrDstW = c->chrDstW;
const int chrSrcW = c->chrSrcW;
const int lumXInc = c->lumXInc;
const int chrXInc = c->chrXInc;
const enum PixelFormat dstFormat = c->dstFormat;
const int flags = c->flags;
int32_t *vLumFilterPos = c->vLumFilterPos;
int32_t *vChrFilterPos = c->vChrFilterPos;
int32_t *hLumFilterPos = c->hLumFilterPos;
int32_t *hChrFilterPos = c->hChrFilterPos;
int16_t *vLumFilter = c->vLumFilter;
int16_t *vChrFilter = c->vChrFilter;
int16_t *hLumFilter = c->hLumFilter;
int16_t *hChrFilter = c->hChrFilter;
int32_t *lumMmxFilter = c->lumMmxFilter;
int32_t *chrMmxFilter = c->chrMmxFilter;
const int vLumFilterSize = c->vLumFilterSize;
const int vChrFilterSize = c->vChrFilterSize;
const int hLumFilterSize = c->hLumFilterSize;
const int hChrFilterSize = c->hChrFilterSize;
int16_t **lumPixBuf = c->lumPixBuf;
int16_t **chrUPixBuf = c->chrUPixBuf;
int16_t **chrVPixBuf = c->chrVPixBuf;
int16_t **alpPixBuf = c->alpPixBuf;
const int vLumBufSize = c->vLumBufSize;
const int vChrBufSize = c->vChrBufSize;
uint8_t *formatConvBuffer = c->formatConvBuffer;
uint32_t *pal = c->pal_yuv;
yuv2planar1_fn yuv2plane1 = c->yuv2plane1;
yuv2planarX_fn yuv2planeX = c->yuv2planeX;
yuv2interleavedX_fn yuv2nv12cX = c->yuv2nv12cX;
yuv2packed1_fn yuv2packed1 = c->yuv2packed1;
yuv2packed2_fn yuv2packed2 = c->yuv2packed2;
yuv2packedX_fn yuv2packedX = c->yuv2packedX;
const int chrSrcSliceY = srcSliceY >> c->chrSrcVSubSample;
const int chrSrcSliceH = -((-srcSliceH) >> c->chrSrcVSubSample);
int should_dither = is9_OR_10BPS(c->srcFormat) ||
is16BPS(c->srcFormat);
int lastDstY;
uint32_t *pal=c->pal_yuv;
int should_dither= isNBPS(c->srcFormat) || is16BPS(c->srcFormat);
yuv2planar1_fn yuv2plane1 = c->yuv2plane1;
yuv2planarX_fn yuv2planeX = c->yuv2planeX;
yuv2interleavedX_fn yuv2nv12cX = c->yuv2nv12cX;
yuv2packed1_fn yuv2packed1 = c->yuv2packed1;
yuv2packed2_fn yuv2packed2 = c->yuv2packed2;
yuv2packedX_fn yuv2packedX = c->yuv2packedX;
/* vars which will change and which we need to store back in the context */
int dstY= c->dstY;
int lumBufIndex= c->lumBufIndex;
int chrBufIndex= c->chrBufIndex;
int lastInLumBuf= c->lastInLumBuf;
int lastInChrBuf= c->lastInChrBuf;
int dstY = c->dstY;
int lumBufIndex = c->lumBufIndex;
int chrBufIndex = c->chrBufIndex;
int lastInLumBuf = c->lastInLumBuf;
int lastInChrBuf = c->lastInChrBuf;
if (isPacked(c->srcFormat)) {
src[0]=
src[1]=
src[2]=
src[3]= src[0];
srcStride[0]=
srcStride[1]=
srcStride[2]=
srcStride[3]= srcStride[0];
src[0] =
src[1] =
src[2] =
src[3] = src[0];
srcStride[0] =
srcStride[1] =
srcStride[2] =
srcStride[3] = srcStride[0];
}
srcStride[1]<<= c->vChrDrop;
srcStride[2]<<= c->vChrDrop;
srcStride[1] <<= c->vChrDrop;
srcStride[2] <<= c->vChrDrop;
DEBUG_BUFFERS("swScale() %p[%d] %p[%d] %p[%d] %p[%d] -> %p[%d] %p[%d] %p[%d] %p[%d]\n",
src[0], srcStride[0], src[1], srcStride[1], src[2], srcStride[2], src[3], srcStride[3],
dst[0], dstStride[0], dst[1], dstStride[1], dst[2], dstStride[2], dst[3], dstStride[3]);
src[0], srcStride[0], src[1], srcStride[1],
src[2], srcStride[2], src[3], srcStride[3],
dst[0], dstStride[0], dst[1], dstStride[1],
dst[2], dstStride[2], dst[3], dstStride[3]);
DEBUG_BUFFERS("srcSliceY: %d srcSliceH: %d dstY: %d dstH: %d\n",
srcSliceY, srcSliceH, dstY, dstH);
srcSliceY, srcSliceH, dstY, dstH);
DEBUG_BUFFERS("vLumFilterSize: %d vLumBufSize: %d vChrFilterSize: %d vChrBufSize: %d\n",
vLumFilterSize, vLumBufSize, vChrFilterSize, vChrBufSize);
vLumFilterSize, vLumBufSize, vChrFilterSize, vChrBufSize);
if (dstStride[0]%16 !=0 || dstStride[1]%16 !=0 || dstStride[2]%16 !=0 || dstStride[3]%16 != 0) {
static int warnedAlready=0; //FIXME move this into the context perhaps
if (dstStride[0]%16 !=0 || dstStride[1]%16 !=0 ||
dstStride[2]%16 !=0 || dstStride[3]%16 != 0) {
static int warnedAlready = 0; // FIXME maybe move this into the context
if (flags & SWS_PRINT_INFO && !warnedAlready) {
av_log(c, AV_LOG_WARNING, "Warning: dstStride is not aligned!\n"
av_log(c, AV_LOG_WARNING,
"Warning: dstStride is not aligned!\n"
" ->cannot do aligned memory accesses anymore\n");
warnedAlready=1;
warnedAlready = 1;
}
}
......@@ -408,24 +429,24 @@ static int swScale(SwsContext *c, const uint8_t* src[],
}
/* Note the user might start scaling the picture in the middle so this
will not get executed. This is not really intended but works
currently, so people might do it. */
if (srcSliceY ==0) {
lumBufIndex=-1;
chrBufIndex=-1;
dstY=0;
lastInLumBuf= -1;
lastInChrBuf= -1;
* will not get executed. This is not really intended but works
* currently, so people might do it. */
if (srcSliceY == 0) {
lumBufIndex = -1;
chrBufIndex = -1;
dstY = 0;
lastInLumBuf = -1;
lastInChrBuf = -1;
}
if (!should_dither) {
c->chrDither8 = c->lumDither8 = ff_sws_pb_64;
}
lastDstY= dstY;
lastDstY = dstY;
for (;dstY < dstH; dstY++) {
const int chrDstY= dstY>>c->chrDstVSubSample;
uint8_t *dest[4] = {
for (; dstY < dstH; dstY++) {
const int chrDstY = dstY >> c->chrDstVSubSample;
uint8_t *dest[4] = {
dst[0] + dstStride[0] * dstY,
dst[1] + dstStride[1] * chrDstY,
dst[2] + dstStride[2] * chrDstY,
......@@ -433,9 +454,11 @@ static int swScale(SwsContext *c, const uint8_t* src[],
};
int use_mmx_vfilter= c->use_mmx_vfilter;
const int firstLumSrcY= FFMAX(1 - vLumFilterSize, vLumFilterPos[dstY]); //First line needed as input
const int firstLumSrcY2= FFMAX(1 - vLumFilterSize, vLumFilterPos[FFMIN(dstY | ((1<<c->chrDstVSubSample) - 1), dstH-1)]);
const int firstChrSrcY= FFMAX(1 - vChrFilterSize, vChrFilterPos[chrDstY]); //First line needed as input
// First line needed as input
const int firstLumSrcY = FFMAX(1 - vLumFilterSize, vLumFilterPos[dstY]);
const int firstLumSrcY2 = FFMAX(1 - vLumFilterSize, vLumFilterPos[FFMIN(dstY | ((1 << c->chrDstVSubSample) - 1), dstH - 1)]);
// First line needed as input
const int firstChrSrcY = FFMAX(1 - vChrFilterSize, vChrFilterPos[chrDstY]);
// Last line needed as input
int lastLumSrcY = FFMIN(c->srcH, firstLumSrcY + vLumFilterSize) - 1;
......@@ -443,30 +466,33 @@ static int swScale(SwsContext *c, const uint8_t* src[],
int lastChrSrcY = FFMIN(c->chrSrcH, firstChrSrcY + vChrFilterSize) - 1;
int enough_lines;
//handle holes (FAST_BILINEAR & weird filters)
if (firstLumSrcY > lastInLumBuf) lastInLumBuf= firstLumSrcY-1;
if (firstChrSrcY > lastInChrBuf) lastInChrBuf= firstChrSrcY-1;
// handle holes (FAST_BILINEAR & weird filters)
if (firstLumSrcY > lastInLumBuf)
lastInLumBuf = firstLumSrcY - 1;
if (firstChrSrcY > lastInChrBuf)
lastInChrBuf = firstChrSrcY - 1;
assert(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1);
assert(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1);
DEBUG_BUFFERS("dstY: %d\n", dstY);
DEBUG_BUFFERS("\tfirstLumSrcY: %d lastLumSrcY: %d lastInLumBuf: %d\n",
firstLumSrcY, lastLumSrcY, lastInLumBuf);
firstLumSrcY, lastLumSrcY, lastInLumBuf);
DEBUG_BUFFERS("\tfirstChrSrcY: %d lastChrSrcY: %d lastInChrBuf: %d\n",
firstChrSrcY, lastChrSrcY, lastInChrBuf);
firstChrSrcY, lastChrSrcY, lastInChrBuf);
// Do we have enough lines in this slice to output the dstY line
enough_lines = lastLumSrcY2 < srcSliceY + srcSliceH && lastChrSrcY < -((-srcSliceY - srcSliceH)>>c->chrSrcVSubSample);
enough_lines = lastLumSrcY2 < srcSliceY + srcSliceH &&
lastChrSrcY < -((-srcSliceY - srcSliceH) >> c->chrSrcVSubSample);
if (!enough_lines) {
lastLumSrcY = srcSliceY + srcSliceH - 1;
lastChrSrcY = chrSrcSliceY + chrSrcSliceH - 1;
DEBUG_BUFFERS("buffering slice: lastLumSrcY %d lastChrSrcY %d\n",
lastLumSrcY, lastChrSrcY);
lastLumSrcY, lastChrSrcY);
}
//Do horizontal scaling
while(lastInLumBuf < lastLumSrcY) {
// Do horizontal scaling
while (lastInLumBuf < lastLumSrcY) {
const uint8_t *src1[4] = {
src[0] + (lastInLumBuf + 1 - srcSliceY) * srcStride[0],
src[1] + (lastInLumBuf + 1 - srcSliceY) * srcStride[1],
......@@ -474,23 +500,21 @@ static int swScale(SwsContext *c, const uint8_t* src[],
src[3] + (lastInLumBuf + 1 - srcSliceY) * srcStride[3],
};
lumBufIndex++;
assert(lumBufIndex < 2*vLumBufSize);
assert(lumBufIndex < 2 * vLumBufSize);
assert(lastInLumBuf + 1 - srcSliceY < srcSliceH);
assert(lastInLumBuf + 1 - srcSliceY >= 0);
hyscale(c, lumPixBuf[ lumBufIndex ], dstW, src1, srcW, lumXInc,
hyscale(c, lumPixBuf[lumBufIndex], dstW, src1, srcW, lumXInc,
hLumFilter, hLumFilterPos, hLumFilterSize,
formatConvBuffer,
pal, 0);
formatConvBuffer, pal, 0);
if (CONFIG_SWSCALE_ALPHA && alpPixBuf)
hyscale(c, alpPixBuf[ lumBufIndex ], dstW, src1, srcW,
hyscale(c, alpPixBuf[lumBufIndex], dstW, src1, srcW,
lumXInc, hLumFilter, hLumFilterPos, hLumFilterSize,
formatConvBuffer,
pal, 1);
formatConvBuffer, pal, 1);
lastInLumBuf++;
DEBUG_BUFFERS("\t\tlumBufIndex %d: lastInLumBuf: %d\n",
lumBufIndex, lastInLumBuf);
lumBufIndex, lastInLumBuf);
}
while(lastInChrBuf < lastChrSrcY) {
while (lastInChrBuf < lastChrSrcY) {
const uint8_t *src1[4] = {
src[0] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[0],
src[1] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[1],
......@@ -498,50 +522,56 @@ static int swScale(SwsContext *c, const uint8_t* src[],
src[3] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[3],
};
chrBufIndex++;
assert(chrBufIndex < 2*vChrBufSize);
assert(chrBufIndex < 2 * vChrBufSize);
assert(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH));
assert(lastInChrBuf + 1 - chrSrcSliceY >= 0);
//FIXME replace parameters through context struct (some at least)
// FIXME replace parameters through context struct (some at least)
if (c->needs_hcscale)
hcscale(c, chrUPixBuf[chrBufIndex], chrVPixBuf[chrBufIndex],
chrDstW, src1, chrSrcW, chrXInc,
hChrFilter, hChrFilterPos, hChrFilterSize,
formatConvBuffer, pal);
chrDstW, src1, chrSrcW, chrXInc,
hChrFilter, hChrFilterPos, hChrFilterSize,
formatConvBuffer, pal);
lastInChrBuf++;
DEBUG_BUFFERS("\t\tchrBufIndex %d: lastInChrBuf: %d\n",
chrBufIndex, lastInChrBuf);
chrBufIndex, lastInChrBuf);
}
//wrap buf index around to stay inside the ring buffer
if (lumBufIndex >= vLumBufSize) lumBufIndex-= vLumBufSize;
if (chrBufIndex >= vChrBufSize) chrBufIndex-= vChrBufSize;
// wrap buf index around to stay inside the ring buffer
if (lumBufIndex >= vLumBufSize)
lumBufIndex -= vLumBufSize;
if (chrBufIndex >= vChrBufSize)
chrBufIndex -= vChrBufSize;
if (!enough_lines)
break; //we can't output a dstY line so let's try with the next slice
break; // we can't output a dstY line so let's try with the next slice
#if HAVE_MMX
updateMMXDitherTables(c, dstY, lumBufIndex, chrBufIndex, lastInLumBuf, lastInChrBuf);
updateMMXDitherTables(c, dstY, lumBufIndex, chrBufIndex,
lastInLumBuf, lastInChrBuf);
#endif
if (should_dither) {
c->chrDither8 = dither_8x8_128[chrDstY & 7];
c->lumDither8 = dither_8x8_128[dstY & 7];
c->lumDither8 = dither_8x8_128[dstY & 7];
}
if (dstY >= dstH-2) {
// hmm looks like we can't use MMX here without overwriting this array's tail
ff_sws_init_output_funcs(c, &yuv2plane1, &yuv2planeX, &yuv2nv12cX,
if (dstY >= dstH - 2) {
/* hmm looks like we can't use MMX here without overwriting
* this array's tail */
ff_sws_init_output_funcs(c, &yuv2plane1, &yuv2planeX, &yuv2nv12cX,
&yuv2packed1, &yuv2packed2, &yuv2packedX);
use_mmx_vfilter= 0;
}
{
const int16_t **lumSrcPtr= (const int16_t **)(void*) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
const int16_t **chrUSrcPtr= (const int16_t **)(void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
const int16_t **chrVSrcPtr= (const int16_t **)(void*) chrVPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **)(void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
int16_t *vLumFilter= c->vLumFilter;
int16_t *vChrFilter= c->vChrFilter;
if (isPlanarYUV(dstFormat) || (isGray(dstFormat) && !isALPHA(dstFormat))) { //YV12 like
const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
const int16_t **lumSrcPtr = (const int16_t **)(void*) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
const int16_t **chrUSrcPtr = (const int16_t **)(void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
const int16_t **chrVSrcPtr = (const int16_t **)(void*) chrVPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
const int16_t **alpSrcPtr = (CONFIG_SWSCALE_ALPHA && alpPixBuf) ?
(const int16_t **)(void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
int16_t *vLumFilter = c->vLumFilter;
int16_t *vChrFilter = c->vChrFilter;
if (isPlanarYUV(dstFormat) ||
(isGray(dstFormat) && !isALPHA(dstFormat))) { // YV12 like
const int chrSkipMask = (1 << c->chrDstVSubSample) - 1;
vLumFilter += dstY * vLumFilterSize;
vChrFilter += chrDstY * vChrFilterSize;
......@@ -564,53 +594,62 @@ static int swScale(SwsContext *c, const uint8_t* src[],
yuv2plane1(lumSrcPtr[0], dest[0], dstW, c->lumDither8, 0);
} else {
yuv2planeX(vLumFilter, vLumFilterSize,
lumSrcPtr, dest[0], dstW, c->lumDither8, 0);
lumSrcPtr, dest[0],
dstW, c->lumDither8, 0);
}
if (!((dstY&chrSkipMask) || isGray(dstFormat))) {
if (!((dstY & chrSkipMask) || isGray(dstFormat))) {
if (yuv2nv12cX) {
yuv2nv12cX(c, vChrFilter, vChrFilterSize, chrUSrcPtr, chrVSrcPtr, dest[1], chrDstW);
yuv2nv12cX(c, vChrFilter,
vChrFilterSize, chrUSrcPtr, chrVSrcPtr,
dest[1], chrDstW);
} else if (vChrFilterSize == 1) {
yuv2plane1(chrUSrcPtr[0], dest[1], chrDstW, c->chrDither8, 0);
yuv2plane1(chrVSrcPtr[0], dest[2], chrDstW, c->chrDither8, 3);
} else {
yuv2planeX(vChrFilter, vChrFilterSize,
chrUSrcPtr, dest[1], chrDstW, c->chrDither8, 0);
yuv2planeX(vChrFilter, vChrFilterSize,
chrVSrcPtr, dest[2], chrDstW, c->chrDither8, use_mmx_vfilter ? (c->uv_offx2 >> 1) : 3);
yuv2planeX(vChrFilter,
vChrFilterSize, chrUSrcPtr, dest[1],
chrDstW, c->chrDither8, 0);
yuv2planeX(vChrFilter,
vChrFilterSize, chrVSrcPtr, dest[2],
chrDstW, c->chrDither8, use_mmx_vfilter ? (c->uv_offx2 >> 1) : 3);
}
}
if (CONFIG_SWSCALE_ALPHA && alpPixBuf){
if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
if(use_mmx_vfilter){
vLumFilter= c->alpMmxFilter;
}
if (vLumFilterSize == 1) {
yuv2plane1(alpSrcPtr[0], dest[3], dstW, c->lumDither8, 0);
yuv2plane1(alpSrcPtr[0], dest[3], dstW,
c->lumDither8, 0);
} else {
yuv2planeX(vLumFilter, vLumFilterSize,
alpSrcPtr, dest[3], dstW, c->lumDither8, 0);
yuv2planeX(vLumFilter,
vLumFilterSize, alpSrcPtr, dest[3],
dstW, c->lumDither8, 0);
}
}
} else {
assert(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2);
assert(chrUSrcPtr + vChrFilterSize - 1 < chrUPixBuf + vChrBufSize*2);
if (c->yuv2packed1 && vLumFilterSize == 1 && vChrFilterSize <= 2) { //unscaled RGB
assert(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize * 2);
assert(chrUSrcPtr + vChrFilterSize - 1 < chrUPixBuf + vChrBufSize * 2);
if (c->yuv2packed1 && vLumFilterSize == 1 &&
vChrFilterSize <= 2) { // unscaled RGB
int chrAlpha = vChrFilterSize == 1 ? 0 : vChrFilter[2 * dstY + 1];
yuv2packed1(c, *lumSrcPtr, chrUSrcPtr, chrVSrcPtr,
alpPixBuf ? *alpSrcPtr : NULL,
dest[0], dstW, chrAlpha, dstY);
} else if (c->yuv2packed2 && vLumFilterSize == 2 && vChrFilterSize == 2) { //bilinear upscale RGB
} else if (c->yuv2packed2 && vLumFilterSize == 2 &&
vChrFilterSize == 2) { // bilinear upscale RGB
int lumAlpha = vLumFilter[2 * dstY + 1];
int chrAlpha = vChrFilter[2 * dstY + 1];
lumMmxFilter[2] =
lumMmxFilter[3] = vLumFilter[2 * dstY ] * 0x10001;
lumMmxFilter[3] = vLumFilter[2 * dstY] * 0x10001;
chrMmxFilter[2] =
chrMmxFilter[3] = vChrFilter[2 * chrDstY] * 0x10001;
yuv2packed2(c, lumSrcPtr, chrUSrcPtr, chrVSrcPtr,
alpPixBuf ? alpSrcPtr : NULL,
dest[0], dstW, lumAlpha, chrAlpha, dstY);
} else { //general RGB
} else { // general RGB
yuv2packedX(c, vLumFilter + dstY * vLumFilterSize,
lumSrcPtr, vLumFilterSize,
vChrFilter + dstY * vChrFilterSize,
......@@ -622,20 +661,20 @@ static int swScale(SwsContext *c, const uint8_t* src[],
}
if (isPlanar(dstFormat) && isALPHA(dstFormat) && !alpPixBuf)
fillPlane(dst[3], dstStride[3], dstW, dstY-lastDstY, lastDstY, 255);
fillPlane(dst[3], dstStride[3], dstW, dstY - lastDstY, lastDstY, 255);
#if HAVE_MMX2
if (av_get_cpu_flags() & AV_CPU_FLAG_MMX2)
__asm__ volatile("sfence":::"memory");
__asm__ volatile ("sfence" ::: "memory");
#endif
emms_c();
/* store changed local vars back in the context */
c->dstY= dstY;
c->lumBufIndex= lumBufIndex;
c->chrBufIndex= chrBufIndex;
c->lastInLumBuf= lastInLumBuf;
c->lastInChrBuf= lastInChrBuf;
c->dstY = dstY;
c->lumBufIndex = lumBufIndex;
c->chrBufIndex = chrBufIndex;
c->lastInLumBuf = lastInLumBuf;
c->lastInChrBuf = lastInChrBuf;
return dstY - lastDstY;
}
......@@ -662,7 +701,8 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
c->hyScale = c->hcScale = hScale8To19_c;
}
} else {
c->hyScale = c->hcScale = c->dstBpc > 10 ? hScale16To19_c : hScale16To15_c;
c->hyScale = c->hcScale = c->dstBpc > 10 ? hScale16To19_c
: hScale16To15_c;
}
if (c->srcRange != c->dstRange && !isAnyRGB(c->dstFormat)) {
......
libswscale/swscale_internal.h
View file @ 367d9b29
......@@ -657,8 +657,8 @@ const char *sws_format_name(enum PixelFormat format);
(x)==PIX_FMT_GBR24P \
)
#define isALPHA(x) \
(av_pix_fmt_descriptors[x].nb_components == 2 || \
#define isALPHA(x) \
(av_pix_fmt_descriptors[x].nb_components == 2 || \
av_pix_fmt_descriptors[x].nb_components == 4)
#if 1
......@@ -677,7 +677,7 @@ const char *sws_format_name(enum PixelFormat format);
(x) == PIX_FMT_PAL8)
#endif
#define isPlanar(x) \
#define isPlanar(x) \
(av_pix_fmt_descriptors[x].nb_components >= 2 && \
(av_pix_fmt_descriptors[x].flags & PIX_FMT_PLANAR))
......@@ -689,7 +689,7 @@ const char *sws_format_name(enum PixelFormat format);
((av_pix_fmt_descriptors[x].flags & \
(PIX_FMT_PLANAR | PIX_FMT_RGB)) == (PIX_FMT_PLANAR | PIX_FMT_RGB))
#define usePal(x) ((av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL) || \
#define usePal(x) ((av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL) || \
(av_pix_fmt_descriptors[x].flags & PIX_FMT_PSEUDOPAL) || \
(x) == PIX_FMT_Y400A)
......
libswscale/utils.c
View file @ 367d9b29
......@@ -18,14 +18,15 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#define _SVID_SOURCE //needed for MAP_ANONYMOUS
#include "config.h"
#define _SVID_SOURCE // needed for MAP_ANONYMOUS
#define _DARWIN_C_SOURCE // needed for MAP_ANON
#include <assert.h>
#include <inttypes.h>
#include <string.h>
#include <math.h>
#include <stdio.h>
#include "config.h"
#include <assert.h>
#include <string.h>
#if HAVE_SYS_MMAN_H
#include <sys/mman.h>
#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
......@@ -36,18 +37,19 @@
#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#endif
#include "swscale.h"
#include "swscale_internal.h"
#include "rgb2rgb.h"
#include "libavutil/intreadwrite.h"
#include "libavutil/x86_cpu.h"
#include "libavutil/cpu.h"
#include "libavutil/avassert.h"
#include "libavutil/avutil.h"
#include "libavutil/bswap.h"
#include "libavutil/cpu.h"
#include "libavutil/intreadwrite.h"
#include "libavutil/mathematics.h"
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
#include "libavutil/avassert.h"
#include "libavutil/x86_cpu.h"
#include "rgb2rgb.h"
#include "swscale.h"
#include "swscale_internal.h"
unsigned swscale_version(void)
{
......@@ -66,110 +68,110 @@ const char *swscale_license(void)
return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
}
#define RET 0xC3 //near return opcode for x86
#define RET 0xC3 // near return opcode for x86
typedef struct FormatEntry {
int is_supported_in, is_supported_out;
} FormatEntry;
static const FormatEntry format_entries[PIX_FMT_NB] = {
[PIX_FMT_YUV420P] = { 1 , 1 },
[PIX_FMT_YUYV422] = { 1 , 1 },
[PIX_FMT_RGB24] = { 1 , 1 },
[PIX_FMT_BGR24] = { 1 , 1 },
[PIX_FMT_YUV422P] = { 1 , 1 },
[PIX_FMT_YUV444P] = { 1 , 1 },
[PIX_FMT_YUV410P] = { 1 , 1 },
[PIX_FMT_YUV411P] = { 1 , 1 },
[PIX_FMT_GRAY8] = { 1 , 1 },
[PIX_FMT_MONOWHITE] = { 1 , 1 },
[PIX_FMT_MONOBLACK] = { 1 , 1 },
[PIX_FMT_PAL8] = { 1 , 0 },
[PIX_FMT_YUVJ420P] = { 1 , 1 },
[PIX_FMT_YUVJ422P] = { 1 , 1 },
[PIX_FMT_YUVJ444P] = { 1 , 1 },
[PIX_FMT_UYVY422] = { 1 , 1 },
[PIX_FMT_UYYVYY411] = { 0 , 0 },
[PIX_FMT_BGR8] = { 1 , 1 },
[PIX_FMT_BGR4] = { 0 , 1 },
[PIX_FMT_BGR4_BYTE] = { 1 , 1 },
[PIX_FMT_RGB8] = { 1 , 1 },
[PIX_FMT_RGB4] = { 0 , 1 },
[PIX_FMT_RGB4_BYTE] = { 1 , 1 },
[PIX_FMT_NV12] = { 1 , 1 },
[PIX_FMT_NV21] = { 1 , 1 },
[PIX_FMT_ARGB] = { 1 , 1 },
[PIX_FMT_RGBA] = { 1 , 1 },
[PIX_FMT_ABGR] = { 1 , 1 },
[PIX_FMT_BGRA] = { 1 , 1 },
[PIX_FMT_0RGB] = { 1 , 1 },
[PIX_FMT_RGB0] = { 1 , 1 },
[PIX_FMT_0BGR] = { 1 , 1 },
[PIX_FMT_BGR0] = { 1 , 1 },
[PIX_FMT_GRAY16BE] = { 1 , 1 },
[PIX_FMT_GRAY16LE] = { 1 , 1 },
[PIX_FMT_YUV440P] = { 1 , 1 },
[PIX_FMT_YUVJ440P] = { 1 , 1 },
[PIX_FMT_YUVA420P] = { 1 , 1 },
[PIX_FMT_YUVA444P] = { 1 , 1 },
[PIX_FMT_RGB48BE] = { 1 , 1 },
[PIX_FMT_RGB48LE] = { 1 , 1 },
[PIX_FMT_RGBA64BE] = { 1 , 0 },
[PIX_FMT_RGBA64LE] = { 1 , 0 },
[PIX_FMT_RGB565BE] = { 1 , 1 },
[PIX_FMT_RGB565LE] = { 1 , 1 },
[PIX_FMT_RGB555BE] = { 1 , 1 },
[PIX_FMT_RGB555LE] = { 1 , 1 },
[PIX_FMT_BGR565BE] = { 1 , 1 },
[PIX_FMT_BGR565LE] = { 1 , 1 },
[PIX_FMT_BGR555BE] = { 1 , 1 },
[PIX_FMT_BGR555LE] = { 1 , 1 },
[PIX_FMT_YUV420P16LE] = { 1 , 1 },
[PIX_FMT_YUV420P16BE] = { 1 , 1 },
[PIX_FMT_YUV422P16LE] = { 1 , 1 },
[PIX_FMT_YUV422P16BE] = { 1 , 1 },
[PIX_FMT_YUV444P16LE] = { 1 , 1 },
[PIX_FMT_YUV444P16BE] = { 1 , 1 },
[PIX_FMT_RGB444LE] = { 1 , 1 },
[PIX_FMT_RGB444BE] = { 1 , 1 },
[PIX_FMT_BGR444LE] = { 1 , 1 },
[PIX_FMT_BGR444BE] = { 1 , 1 },
[PIX_FMT_Y400A] = { 1 , 0 },
[PIX_FMT_BGR48BE] = { 1 , 1 },
[PIX_FMT_BGR48LE] = { 1 , 1 },
[PIX_FMT_BGRA64BE] = { 0 , 0 },
[PIX_FMT_BGRA64LE] = { 0 , 0 },
[PIX_FMT_YUV420P9BE] = { 1 , 1 },
[PIX_FMT_YUV420P9LE] = { 1 , 1 },
[PIX_FMT_YUV420P10BE] = { 1 , 1 },
[PIX_FMT_YUV420P10LE] = { 1 , 1 },
[PIX_FMT_YUV422P9BE] = { 1 , 1 },
[PIX_FMT_YUV422P9LE] = { 1 , 1 },
[PIX_FMT_YUV422P10BE] = { 1 , 1 },
[PIX_FMT_YUV422P10LE] = { 1 , 1 },
[PIX_FMT_YUV444P9BE] = { 1 , 1 },
[PIX_FMT_YUV444P9LE] = { 1 , 1 },
[PIX_FMT_YUV444P10BE] = { 1 , 1 },
[PIX_FMT_YUV444P10LE] = { 1 , 1 },
[PIX_FMT_GBRP] = { 1 , 0 },
[PIX_FMT_GBRP9LE] = { 1 , 0 },
[PIX_FMT_GBRP9BE] = { 1 , 0 },
[PIX_FMT_GBRP10LE] = { 1 , 0 },
[PIX_FMT_GBRP10BE] = { 1 , 0 },
[PIX_FMT_GBRP16LE] = { 1 , 0 },
[PIX_FMT_GBRP16BE] = { 1 , 0 },
[PIX_FMT_YUV420P] = { 1, 1 },
[PIX_FMT_YUYV422] = { 1, 1 },
[PIX_FMT_RGB24] = { 1, 1 },
[PIX_FMT_BGR24] = { 1, 1 },
[PIX_FMT_YUV422P] = { 1, 1 },
[PIX_FMT_YUV444P] = { 1, 1 },
[PIX_FMT_YUV410P] = { 1, 1 },
[PIX_FMT_YUV411P] = { 1, 1 },
[PIX_FMT_GRAY8] = { 1, 1 },
[PIX_FMT_MONOWHITE] = { 1, 1 },
[PIX_FMT_MONOBLACK] = { 1, 1 },
[PIX_FMT_PAL8] = { 1, 0 },
[PIX_FMT_YUVJ420P] = { 1, 1 },
[PIX_FMT_YUVJ422P] = { 1, 1 },
[PIX_FMT_YUVJ444P] = { 1, 1 },
[PIX_FMT_UYVY422] = { 1, 1 },
[PIX_FMT_UYYVYY411] = { 0, 0 },
[PIX_FMT_BGR8] = { 1, 1 },
[PIX_FMT_BGR4] = { 0, 1 },
[PIX_FMT_BGR4_BYTE] = { 1, 1 },
[PIX_FMT_RGB8] = { 1, 1 },
[PIX_FMT_RGB4] = { 0, 1 },
[PIX_FMT_RGB4_BYTE] = { 1, 1 },
[PIX_FMT_NV12] = { 1, 1 },
[PIX_FMT_NV21] = { 1, 1 },
[PIX_FMT_ARGB] = { 1, 1 },
[PIX_FMT_RGBA] = { 1, 1 },
[PIX_FMT_ABGR] = { 1, 1 },
[PIX_FMT_BGRA] = { 1, 1 },
[PIX_FMT_0RGB] = { 1, 1 },
[PIX_FMT_RGB0] = { 1, 1 },
[PIX_FMT_0BGR] = { 1, 1 },
[PIX_FMT_BGR0] = { 1, 1 },
[PIX_FMT_GRAY16BE] = { 1, 1 },
[PIX_FMT_GRAY16LE] = { 1, 1 },
[PIX_FMT_YUV440P] = { 1, 1 },
[PIX_FMT_YUVJ440P] = { 1, 1 },
[PIX_FMT_YUVA420P] = { 1, 1 },
[PIX_FMT_YUVA444P] = { 1, 1 },
[PIX_FMT_RGB48BE] = { 1, 1 },
[PIX_FMT_RGB48LE] = { 1, 1 },
[PIX_FMT_RGBA64BE] = { 1, 0 },
[PIX_FMT_RGBA64LE] = { 1, 0 },
[PIX_FMT_RGB565BE] = { 1, 1 },
[PIX_FMT_RGB565LE] = { 1, 1 },
[PIX_FMT_RGB555BE] = { 1, 1 },
[PIX_FMT_RGB555LE] = { 1, 1 },
[PIX_FMT_BGR565BE] = { 1, 1 },
[PIX_FMT_BGR565LE] = { 1, 1 },
[PIX_FMT_BGR555BE] = { 1, 1 },
[PIX_FMT_BGR555LE] = { 1, 1 },
[PIX_FMT_YUV420P16LE] = { 1, 1 },
[PIX_FMT_YUV420P16BE] = { 1, 1 },
[PIX_FMT_YUV422P16LE] = { 1, 1 },
[PIX_FMT_YUV422P16BE] = { 1, 1 },
[PIX_FMT_YUV444P16LE] = { 1, 1 },
[PIX_FMT_YUV444P16BE] = { 1, 1 },
[PIX_FMT_RGB444LE] = { 1, 1 },
[PIX_FMT_RGB444BE] = { 1, 1 },
[PIX_FMT_BGR444LE] = { 1, 1 },
[PIX_FMT_BGR444BE] = { 1, 1 },
[PIX_FMT_Y400A] = { 1, 0 },
[PIX_FMT_BGR48BE] = { 1, 1 },
[PIX_FMT_BGR48LE] = { 1, 1 },
[PIX_FMT_BGRA64BE] = { 0, 0 },
[PIX_FMT_BGRA64LE] = { 0, 0 },
[PIX_FMT_YUV420P9BE] = { 1, 1 },
[PIX_FMT_YUV420P9LE] = { 1, 1 },
[PIX_FMT_YUV420P10BE] = { 1, 1 },
[PIX_FMT_YUV420P10LE] = { 1, 1 },
[PIX_FMT_YUV422P9BE] = { 1, 1 },
[PIX_FMT_YUV422P9LE] = { 1, 1 },
[PIX_FMT_YUV422P10BE] = { 1, 1 },
[PIX_FMT_YUV422P10LE] = { 1, 1 },
[PIX_FMT_YUV444P9BE] = { 1, 1 },
[PIX_FMT_YUV444P9LE] = { 1, 1 },
[PIX_FMT_YUV444P10BE] = { 1, 1 },
[PIX_FMT_YUV444P10LE] = { 1, 1 },
[PIX_FMT_GBRP] = { 1, 0 },
[PIX_FMT_GBRP9LE] = { 1, 0 },
[PIX_FMT_GBRP9BE] = { 1, 0 },
[PIX_FMT_GBRP10LE] = { 1, 0 },
[PIX_FMT_GBRP10BE] = { 1, 0 },
[PIX_FMT_GBRP16LE] = { 1, 0 },
[PIX_FMT_GBRP16BE] = { 1, 0 },
};
int sws_isSupportedInput(enum PixelFormat pix_fmt)
{
return (unsigned)pix_fmt < PIX_FMT_NB ?
format_entries[pix_fmt].is_supported_in : 0;
format_entries[pix_fmt].is_supported_in : 0;
}
int sws_isSupportedOutput(enum PixelFormat pix_fmt)
{
return (unsigned)pix_fmt < PIX_FMT_NB ?
format_entries[pix_fmt].is_supported_out : 0;
format_entries[pix_fmt].is_supported_out : 0;
}
extern const int32_t ff_yuv2rgb_coeffs[8][4];
......@@ -181,261 +183,301 @@ const char *sws_format_name(enum PixelFormat format)
}
#endif
static double getSplineCoeff(double a, double b, double c, double d, double dist)
static double getSplineCoeff(double a, double b, double c, double d,
double dist)
{
if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
else return getSplineCoeff( 0.0,
b+ 2.0*c + 3.0*d,
c + 3.0*d,
-b- 3.0*c - 6.0*d,
dist-1.0);
if (dist <= 1.0)
return ((d * dist + c) * dist + b) * dist + a;
else
return getSplineCoeff(0.0,
b + 2.0 * c + 3.0 * d,
c + 3.0 * d,
-b - 3.0 * c - 6.0 * d,
dist - 1.0);
}
static 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, double param[2])
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,
double param[2])
{
int i;
int filterSize;
int filter2Size;
int minFilterSize;
int64_t *filter=NULL;
int64_t *filter2=NULL;
const int64_t fone= 1LL<<54;
int ret= -1;
int64_t *filter = NULL;
int64_t *filter2 = NULL;
const int64_t fone = 1LL << 54;
int ret = -1;
emms_c(); //FIXME this should not be required but it IS (even for non-MMX versions)
emms_c(); // FIXME should not be required but IS (even for non-MMX versions)
// 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);
// 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);
if (FFABS(xInc - 0x10000) <10) { // unscaled
if (FFABS(xInc - 0x10000) < 10) { // unscaled
int i;
filterSize= 1;
FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
filterSize = 1;
FF_ALLOCZ_OR_GOTO(NULL, filter,
dstW * sizeof(*filter) * filterSize, fail);
for (i=0; i<dstW; i++) {
filter[i*filterSize]= fone;
(*filterPos)[i]=i;
for (i = 0; i < dstW; i++) {
filter[i * filterSize] = fone;
(*filterPos)[i] = i;
}
} else if (flags&SWS_POINT) { // lame looking point sampling mode
} else if (flags & SWS_POINT) { // lame looking point sampling mode
int i;
int64_t xDstInSrc;
filterSize= 1;
FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
filterSize = 1;
FF_ALLOC_OR_GOTO(NULL, filter,
dstW * sizeof(*filter) * filterSize, fail);
xDstInSrc= xInc/2 - 0x8000;
for (i=0; i<dstW; i++) {
int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
xDstInSrc = xInc / 2 - 0x8000;
for (i = 0; i < dstW; i++) {
int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
(*filterPos)[i]= xx;
filter[i]= fone;
xDstInSrc+= xInc;
(*filterPos)[i] = xx;
filter[i] = fone;
xDstInSrc += xInc;
}
} else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale
} else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) ||
(flags & SWS_FAST_BILINEAR)) { // bilinear upscale
int i;
int64_t xDstInSrc;
filterSize= 2;
FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
filterSize = 2;
FF_ALLOC_OR_GOTO(NULL, filter,
dstW * sizeof(*filter) * filterSize, fail);
xDstInSrc= xInc/2 - 0x8000;
for (i=0; i<dstW; i++) {
int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
xDstInSrc = xInc / 2 - 0x8000;
for (i = 0; i < dstW; i++) {
int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
int j;
(*filterPos)[i]= xx;
//bilinear upscale / linear interpolate / area averaging
for (j=0; j<filterSize; j++) {
(*filterPos)[i] = xx;
// bilinear upscale / linear interpolate / area averaging
for (j = 0; j < filterSize; j++) {
int64_t coeff= fone - FFABS(((int64_t)xx<<16) - xDstInSrc)*(fone>>16);
if (coeff<0) coeff=0;
filter[i*filterSize + j]= coeff;
if (coeff < 0)
coeff = 0;
filter[i * filterSize + j] = coeff;
xx++;
}
xDstInSrc+= xInc;
xDstInSrc += xInc;
}
} else {
int64_t xDstInSrc;
int sizeFactor;
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;
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;
else {
sizeFactor= 0; //GCC warning killer
sizeFactor = 0; // GCC warning killer
assert(0);
}
if (xInc <= 1<<16) filterSize= 1 + sizeFactor; // upscale
else filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
if (xInc <= 1 << 16)
filterSize = 1 + sizeFactor; // upscale
else
filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW;
filterSize = FFMIN(filterSize, srcW - 2);
filterSize = FFMAX(filterSize, 1);
FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
FF_ALLOC_OR_GOTO(NULL, filter,
dstW * sizeof(*filter) * filterSize, fail);
xDstInSrc= xInc - 0x10000;
for (i=0; i<dstW; i++) {
int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
xDstInSrc = xInc - 0x10000;
for (i = 0; i < dstW; i++) {
int xx = (xDstInSrc - ((filterSize - 2) << 16)) / (1 << 17);
int j;
(*filterPos)[i]= xx;
for (j=0; j<filterSize; j++) {
int64_t d= (FFABS(((int64_t)xx<<17) - xDstInSrc))<<13;
(*filterPos)[i] = xx;
for (j = 0; j < filterSize; j++) {
int64_t d = (FFABS(((int64_t)xx << 17) - xDstInSrc)) << 13;
double floatd;
int64_t coeff;
if (xInc > 1<<16)
d= d*dstW/srcW;
floatd= d * (1.0/(1<<30));
if (xInc > 1 << 16)
d = d * dstW / srcW;
floatd = d * (1.0 / (1 << 30));
if (flags & SWS_BICUBIC) {
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);
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);
if (d >= 1LL<<31) {
if (d >= 1LL << 31) {
coeff = 0.0;
} else {
int64_t dd = (d * d) >> 30;
int64_t ddd = (dd * d) >> 30;
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);
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);
else
coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
coeff = (-B - 6 * C) * ddd +
(6 * B + 30 * C) * dd +
(-12 * B - 48 * C) * d +
(8 * B + 24 * C) * (1 << 30);
}
coeff *= fone>>(30+24);
coeff *= fone >> (30 + 24);
}
#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);
}
/* 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);
}*/
#endif
else if (flags & SWS_X) {
double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
double c;
if (floatd<1.0)
c = cos(floatd*M_PI);
if (floatd < 1.0)
c = cos(floatd * M_PI);
else
c = -1.0;
if (c < 0.0)
c = -pow(-c, A);
else
c=-1.0;
if (c<0.0) c= -pow(-c, A);
else c= pow( c, A);
coeff= (c*0.5 + 0.5)*fone;
c = pow(c, A);
coeff = (c * 0.5 + 0.5) * fone;
} else if (flags & SWS_AREA) {
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);
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);
} else if (flags & SWS_GAUSS) {
double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
coeff = (pow(2.0, - p*floatd*floatd))*fone;
double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
coeff = (pow(2.0, -p * floatd * floatd)) * fone;
} else if (flags & SWS_SINC) {
coeff = (d ? sin(floatd*M_PI)/(floatd*M_PI) : 1.0)*fone;
coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone;
} else if (flags & SWS_LANCZOS) {
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;
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;
} else if (flags & SWS_BILINEAR) {
coeff= (1<<30) - d;
if (coeff<0) coeff=0;
coeff = (1 << 30) - d;
if (coeff < 0)
coeff = 0;
coeff *= fone >> 30;
} else if (flags & SWS_SPLINE) {
double p=-2.196152422706632;
coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
double p = -2.196152422706632;
coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone;
} else {
coeff= 0.0; //GCC warning killer
coeff = 0.0; // GCC warning killer
assert(0);
}
filter[i*filterSize + j]= coeff;
filter[i * filterSize + j] = coeff;
xx++;
}
xDstInSrc+= 2*xInc;
xDstInSrc += 2 * xInc;
}
}
/* apply src & dst Filter to filter -> filter2
av_free(filter);
*/
assert(filterSize>0);
filter2Size= filterSize;
if (srcFilter) filter2Size+= srcFilter->length - 1;
if (dstFilter) filter2Size+= dstFilter->length - 1;
assert(filter2Size>0);
FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail);
for (i=0; i<dstW; i++) {
* av_free(filter);
*/
assert(filterSize > 0);
filter2Size = filterSize;
if (srcFilter)
filter2Size += srcFilter->length - 1;
if (dstFilter)
filter2Size += dstFilter->length - 1;
assert(filter2Size > 0);
FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size * dstW * sizeof(*filter2), fail);
for (i = 0; i < dstW; i++) {
int j, k;
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];
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];
}
} else {
for (j=0; j<filterSize; j++)
filter2[i*filter2Size + j]= filter[i*filterSize + j];
for (j = 0; j < filterSize; j++)
filter2[i * filter2Size + j] = filter[i * filterSize + j];
}
//FIXME dstFilter
// FIXME dstFilter
(*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
(*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2;
}
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).
minFilterSize= 0;
for (i=dstW-1; i>=0; i--) {
int min= filter2Size;
minFilterSize = 0;
for (i = dstW - 1; i >= 0; i--) {
int min = filter2Size;
int j;
int64_t cutOff=0.0;
int64_t cutOff = 0.0;
/* get rid of near zero elements on the left by shifting left */
for (j=0; j<filter2Size; j++) {
for (j = 0; j < filter2Size; j++) {
int k;
cutOff += FFABS(filter2[i*filter2Size]);
cutOff += FFABS(filter2[i * filter2Size]);
if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
break;
/* preserve monotonicity because the core can't handle the filter otherwise */
if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
/* preserve monotonicity because the core can't handle the
* filter otherwise */
if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1])
break;
// move filter coefficients left
for (k=1; k<filter2Size; k++)
filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
filter2[i*filter2Size + k - 1]= 0;
for (k = 1; k < filter2Size; k++)
filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k];
filter2[i * filter2Size + k - 1] = 0;
(*filterPos)[i]++;
}
cutOff=0;
cutOff = 0;
/* count near zeros on the right */
for (j=filter2Size-1; j>0; j--) {
cutOff += FFABS(filter2[i*filter2Size + j]);
for (j = filter2Size - 1; j > 0; j--) {
cutOff += FFABS(filter2[i * filter2Size + j]);
if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
break;
min--;
}
if (min>minFilterSize) minFilterSize= min;
if (min > minFilterSize)
minFilterSize = min;
}
if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) {
// we can handle the special case 4,
// so we don't want to go to the full 8
// we can handle the special case 4, so we don't want to go the full 8
if (minFilterSize < 5)
filterAlign = 4;
// We really don't want to waste our time
// doing useless computation, so fall back on
// the scalar C code for very small filters.
// Vectorizing is worth it only if you have a
// decent-sized vector.
/* 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. */
if (minFilterSize < 3)
filterAlign = 1;
}
......@@ -443,53 +485,58 @@ static int initFilter(int16_t **outFilter, int32_t **filterPos, int *outFilterSi
if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
// special case for unscaled vertical filtering
if (minFilterSize == 1 && filterAlign == 2)
filterAlign= 1;
filterAlign = 1;
}
assert(minFilterSize > 0);
filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1));
assert(filterSize > 0);
filter= av_malloc(filterSize*dstW*sizeof(*filter));
if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
filter = av_malloc(filterSize * dstW * sizeof(*filter));
if (filterSize >= MAX_FILTER_SIZE * 16 /
((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
goto fail;
*outFilterSize= filterSize;
*outFilterSize = filterSize;
if (flags&SWS_PRINT_INFO)
av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
if (flags & SWS_PRINT_INFO)
av_log(NULL, AV_LOG_VERBOSE,
"SwScaler: reducing / aligning filtersize %d -> %d\n",
filter2Size, filterSize);
/* try to reduce the filter-size (step2 reduce it) */
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
int j;
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;
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;
}
}
//FIXME try to align filterPos if possible
// FIXME try to align filterPos if possible
//fix borders
for (i=0; i<dstW; i++) {
// fix borders
for (i = 0; i < dstW; i++) {
int j;
if ((*filterPos)[i] < 0) {
// move filter coefficients left to compensate for filterPos
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;
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;
}
(*filterPos)[i]= 0;
}
if ((*filterPos)[i] + filterSize > srcW) {
int shift= (*filterPos)[i] + filterSize - srcW;
int shift = (*filterPos)[i] + filterSize - srcW;
// move filter coefficients right to compensate for filterPos
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;
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;
}
(*filterPos)[i]= srcW - filterSize;
}
......@@ -497,37 +544,40 @@ static int initFilter(int16_t **outFilter, int32_t **filterPos, int *outFilterSi
// Note the +1 is for the MMX scaler which reads over the end
/* align at 16 for AltiVec (needed by hScale_altivec_real) */
FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+3)*sizeof(int16_t), fail);
FF_ALLOCZ_OR_GOTO(NULL, *outFilter,
*outFilterSize * (dstW + 3) * sizeof(int16_t), fail);
/* normalize & store in outFilter */
for (i=0; i<dstW; i++) {
for (i = 0; i < dstW; i++) {
int j;
int64_t error=0;
int64_t sum=0;
int64_t error = 0;
int64_t sum = 0;
for (j=0; j<filterSize; j++) {
sum+= filter[i*filterSize + j];
for (j = 0; j < filterSize; j++) {
sum += filter[i * filterSize + j];
}
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;
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;
}
}
(*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;
(*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;
(*outFilter)[k + 1 * (*outFilterSize)] =
(*outFilter)[k + 2 * (*outFilterSize)] =
(*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k];
}
ret=0;
ret = 0;
fail:
av_free(filter);
av_free(filter2);
......@@ -535,7 +585,8 @@ fail:
}
#if HAVE_MMX2
static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)
static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode,
int16_t *filter, int32_t *filterPos, int numSplits)
{
uint8_t *fragmentA;
x86_reg imm8OfPShufW1A;
......@@ -550,18 +601,17 @@ static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *fil
int xpos, i;
// create an optimized horizontal scaling routine
/* This scaler is made of runtime-generated MMX2 code using specially
* tuned 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.
/* This scaler is made of runtime-generated MMX2 code using specially tuned
* 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.
*/
//code fragment
// code fragment
__asm__ volatile(
__asm__ volatile (
"jmp 9f \n\t"
// Begin
// Begin
"0: \n\t"
"movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
"movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
......@@ -581,27 +631,27 @@ static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *fil
"movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
"add $8, %%"REG_a" \n\t"
// End
// End
"9: \n\t"
// "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"
// "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"
"dec %1 \n\t"
"dec %2 \n\t"
"sub %0, %1 \n\t"
"sub %0, %2 \n\t"
"lea " LOCAL_MANGLE(9b) ", %3 \n\t"
"lea " LOCAL_MANGLE(9b) ", %3 \n\t"
"sub %0, %3 \n\t"
:"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
"=r" (fragmentLengthA)
);
: "=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
"=r" (fragmentLengthA)
);
__asm__ volatile(
__asm__ volatile (
"jmp 9f \n\t"
// Begin
// Begin
"0: \n\t"
"movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
"movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
......@@ -619,76 +669,81 @@ static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *fil
"movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
"add $8, %%"REG_a" \n\t"
// End
// End
"9: \n\t"
// "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"
// "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"
"dec %1 \n\t"
"dec %2 \n\t"
"sub %0, %1 \n\t"
"sub %0, %2 \n\t"
"lea " LOCAL_MANGLE(9b) ", %3 \n\t"
"lea " LOCAL_MANGLE(9b) ", %3 \n\t"
"sub %0, %3 \n\t"
:"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
"=r" (fragmentLengthB)
);
: "=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
"=r" (fragmentLengthB)
);
xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
fragmentPos=0;
xpos = 0; // lumXInc/2 - 0x8000; // difference between pixel centers
fragmentPos = 0;
for (i=0; i<dstW/numSplits; i++) {
int xx=xpos>>16;
for (i = 0; i < dstW / numSplits; i++) {
int xx = xpos >> 16;
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;
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;
if (filterCode) {
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;
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;
memcpy(filterCode + fragmentPos, fragment, fragmentLength);
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;
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;
}
}
fragmentPos+= fragmentLength;
fragmentPos += fragmentLength;
if (filterCode)
filterCode[fragmentPos]= RET;
filterCode[fragmentPos] = RET;
}
xpos+=xInc;
xpos += xInc;
}
if (filterCode)
filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
filterPos[((i / 2) + 1) & (~1)] = xpos >> 16; // needed to jump to the next part
return fragmentPos + 1;
}
......@@ -704,24 +759,27 @@ 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)
{
memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
memcpy(c->dstColorspaceTable, table, sizeof(int)*4);
memcpy(c->srcColorspaceTable, inv_table, sizeof(int) * 4);
memcpy(c->dstColorspaceTable, table, sizeof(int) * 4);
c->brightness= brightness;
c->contrast = contrast;
c->saturation= saturation;
c->srcRange = srcRange;
c->dstRange = dstRange;
if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
c->brightness = brightness;
c->contrast = contrast;
c->saturation = saturation;
c->srcRange = srcRange;
c->dstRange = dstRange;
if (isYUV(c->dstFormat) || isGray(c->dstFormat))
return -1;
c->dstFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[c->dstFormat]);
c->srcFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[c->srcFormat]);
ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
//FIXME factorize
ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
contrast, saturation);
// FIXME factorize
if (HAVE_ALTIVEC && av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC)
ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness,
contrast, saturation);
return 0;
}
......@@ -729,15 +787,16 @@ int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
int *srcRange, int **table, int *dstRange,
int *brightness, int *contrast, int *saturation)
{
if (!c || isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
if (!c || isYUV(c->dstFormat) || isGray(c->dstFormat))
return -1;
*inv_table = c->srcColorspaceTable;
*table = c->dstColorspaceTable;
*srcRange = c->srcRange;
*dstRange = c->dstRange;
*brightness= c->brightness;
*contrast = c->contrast;
*saturation= c->saturation;
*inv_table = c->srcColorspaceTable;
*table = c->dstColorspaceTable;
*srcRange = c->srcRange;
*dstRange = c->dstRange;
*brightness = c->brightness;
*contrast = c->contrast;
*saturation = c->saturation;
return 0;
}
......@@ -745,11 +804,20 @@ int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
static int handle_jpeg(enum PixelFormat *format)
{
switch (*format) {
case PIX_FMT_YUVJ420P: *format = PIX_FMT_YUV420P; return 1;
case PIX_FMT_YUVJ422P: *format = PIX_FMT_YUV422P; return 1;
case PIX_FMT_YUVJ444P: *format = PIX_FMT_YUV444P; return 1;
case PIX_FMT_YUVJ440P: *format = PIX_FMT_YUV440P; return 1;
default: return 0;
case PIX_FMT_YUVJ420P:
*format = PIX_FMT_YUV420P;
return 1;
case PIX_FMT_YUVJ422P:
*format = PIX_FMT_YUV422P;
return 1;
case PIX_FMT_YUVJ444P:
*format = PIX_FMT_YUV444P;
return 1;
case PIX_FMT_YUVJ440P:
*format = PIX_FMT_YUV440P;
return 1;
default:
return 0;
}
}
......@@ -766,7 +834,7 @@ static int handle_0alpha(enum PixelFormat *format)
SwsContext *sws_alloc_context(void)
{
SwsContext *c= av_mallocz(sizeof(SwsContext));
SwsContext *c = av_mallocz(sizeof(SwsContext));
c->av_class = &sws_context_class;
av_opt_set_defaults(c);
......@@ -779,20 +847,21 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
int i, j;
int usesVFilter, usesHFilter;
int unscaled;
SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
int srcW= c->srcW;
int srcH= c->srcH;
int dstW= c->dstW;
int dstH= c->dstH;
int dst_stride = FFALIGN(dstW * sizeof(int16_t)+66, 16);
SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
int srcW = c->srcW;
int srcH = c->srcH;
int dstW = c->dstW;
int dstH = c->dstH;
int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16);
int flags, cpu_flags;
enum PixelFormat srcFormat= c->srcFormat;
enum PixelFormat dstFormat= c->dstFormat;
enum PixelFormat srcFormat = c->srcFormat;
enum PixelFormat dstFormat = c->dstFormat;
cpu_flags = av_get_cpu_flags();
flags = c->flags;
emms_c();
if (!rgb15to16) sws_rgb2rgb_init();
if (!rgb15to16)
sws_rgb2rgb_init();
unscaled = (srcW == dstW && srcH == dstH);
......@@ -808,53 +877,59 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
}
if (!sws_isSupportedInput(srcFormat)) {
av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n", av_get_pix_fmt_name(srcFormat));
av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
av_get_pix_fmt_name(srcFormat));
return AVERROR(EINVAL);
}
if (!sws_isSupportedOutput(dstFormat)) {
av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n", av_get_pix_fmt_name(dstFormat));
av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n",
av_get_pix_fmt_name(dstFormat));
return AVERROR(EINVAL);
}
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))) {
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))) {
av_log(c, AV_LOG_ERROR, "Exactly one scaler algorithm must be chosen, got %X\n", i);
return AVERROR(EINVAL);
}
/* sanity check */
if (srcW<4 || srcH<1 || dstW<8 || dstH<1) { //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code
if (srcW < 4 || srcH < 1 || dstW < 8 || dstH < 1) {
/* FIXME check if these are enough and try to lower them after
* fixing the relevant parts of the code */
av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
srcW, srcH, dstW, dstH);
return AVERROR(EINVAL);
}
if (!dstFilter) dstFilter= &dummyFilter;
if (!srcFilter) srcFilter= &dummyFilter;
if (!dstFilter)
dstFilter = &dummyFilter;
if (!srcFilter)
srcFilter = &dummyFilter;
c->lumXInc= (((int64_t)srcW<<16) + (dstW>>1))/dstW;
c->lumYInc= (((int64_t)srcH<<16) + (dstH>>1))/dstH;
c->lumXInc = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
c->lumYInc = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH;
c->dstFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[dstFormat]);
c->srcFormatBpp = av_get_bits_per_pixel(&av_pix_fmt_descriptors[srcFormat]);
c->vRounder= 4* 0x0001000100010001ULL;
c->vRounder = 4 * 0x0001000100010001ULL;
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);
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);
getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
......@@ -867,48 +942,55 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
c->flags = flags;
}
}
// reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
/* reuse chroma for 2 pixels RGB/BGR unless user wants full
* chroma interpolation */
if (flags & SWS_FULL_CHR_H_INT &&
isAnyRGB(dstFormat) &&
dstFormat != PIX_FMT_RGBA &&
dstFormat != PIX_FMT_ARGB &&
dstFormat != PIX_FMT_BGRA &&
dstFormat != PIX_FMT_ABGR &&
isAnyRGB(dstFormat) &&
dstFormat != PIX_FMT_RGBA &&
dstFormat != PIX_FMT_ARGB &&
dstFormat != PIX_FMT_BGRA &&
dstFormat != PIX_FMT_ABGR &&
dstFormat != PIX_FMT_RGB24 &&
dstFormat != PIX_FMT_BGR24) {
av_log(c, AV_LOG_WARNING,
"full chroma interpolation for destination format '%s' not yet implemented\n",
sws_format_name(dstFormat));
flags &= ~SWS_FULL_CHR_H_INT;
av_get_pix_fmt_name(dstFormat));
flags &= ~SWS_FULL_CHR_H_INT;
c->flags = flags;
}
if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
c->chrDstHSubSample = 1;
// drop some chroma lines if the user wants it
c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
c->chrSrcVSubSample+= c->vChrDrop;
// drop every other pixel for chroma calculation unless user wants full chroma
if (isAnyRGB(srcFormat) && !(flags&SWS_FULL_CHR_H_INP)
&& srcFormat!=PIX_FMT_RGB8 && srcFormat!=PIX_FMT_BGR8
&& srcFormat!=PIX_FMT_RGB4 && srcFormat!=PIX_FMT_BGR4
&& srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
&& ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&SWS_FAST_BILINEAR)))
c->chrSrcHSubSample=1;
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) &&
srcFormat != PIX_FMT_RGB8 && srcFormat != PIX_FMT_BGR8 &&
srcFormat != PIX_FMT_RGB4 && srcFormat != PIX_FMT_BGR4 &&
srcFormat != PIX_FMT_RGB4_BYTE && srcFormat != PIX_FMT_BGR4_BYTE &&
((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
(flags & SWS_FAST_BILINEAR)))
c->chrSrcHSubSample = 1;
// Note the -((-x)>>y) is so that we always round toward +inf.
c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
c->chrSrcW = -((-srcW) >> c->chrSrcHSubSample);
c->chrSrcH = -((-srcH) >> c->chrSrcVSubSample);
c->chrDstW = -((-dstW) >> c->chrDstHSubSample);
c->chrDstH = -((-dstH) >> c->chrDstVSubSample);
/* unscaled special cases */
if (unscaled && !usesHFilter && !usesVFilter && (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
if (unscaled && !usesHFilter && !usesVFilter &&
(c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
ff_get_unscaled_swscale(c);
if (c->swScale) {
if (flags&SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
if (flags & SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO,
"using unscaled %s -> %s special converter\n",
av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
return 0;
}
......@@ -925,35 +1007,40 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
if (c->dstBpc == 16)
dst_stride <<= 1;
FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2 && c->srcBpc == 8 && c->dstBpc <= 10) {
c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {
if (flags&SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2 &&
c->srcBpc == 8 && c->dstBpc <= 10) {
c->canMMX2BeUsed = (dstW >= srcW && (dstW & 31) == 0 &&
(srcW & 15) == 0) ? 1 : 0;
if (!c->canMMX2BeUsed && dstW >= srcW && (srcW & 15) == 0
&& (flags & SWS_FAST_BILINEAR)) {
if (flags & SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO,
"output width is not a multiple of 32 -> no MMX2 scaler\n");
}
if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat)) c->canMMX2BeUsed=0;
}
else
c->canMMX2BeUsed=0;
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) {
if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat))
c->canMMX2BeUsed=0;
} else
c->canMMX2BeUsed = 0;
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) {
if (c->canMMX2BeUsed) {
c->lumXInc+= 20;
c->chrXInc+= 20;
c->lumXInc += 20;
c->chrXInc += 20;
}
//we don't use the x86 asm scaler if MMX is available
// we don't use the x86 asm scaler if MMX is available
else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX && c->dstBpc <= 10) {
c->lumXInc = ((int64_t)(srcW-2)<<16)/(dstW-2) - 20;
c->chrXInc = ((int64_t)(c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20;
c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
}
}
......@@ -962,8 +1049,10 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
#if HAVE_MMX2
// can't downscale !!!
if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {
c->lumMmx2FilterCodeSize = initMMX2HScaler( dstW, c->lumXInc, NULL, NULL, NULL, 8);
c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);
c->lumMmx2FilterCodeSize = initMMX2HScaler(dstW, c->lumXInc, NULL,
NULL, NULL, 8);
c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc,
NULL, NULL, NULL, 4);
#ifdef MAP_ANONYMOUS
c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
......@@ -985,13 +1074,16 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n");
return AVERROR(ENOMEM);
}
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);
initMMX2HScaler( dstW, c->lumXInc, c->lumMmx2FilterCode, c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
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);
initMMX2HScaler( dstW, c->lumXInc, c->lumMmx2FilterCode,
c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode,
c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
#ifdef MAP_ANONYMOUS
mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
......@@ -1000,107 +1092,122 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
} else
#endif /* HAVE_MMX2 */
{
const int filterAlign=
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 :
const int filterAlign =
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 :
(HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
1;
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, c->param) < 0)
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,
c->param) < 0)
goto fail;
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, c->param) < 0)
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,
c->param) < 0)
goto fail;
}
} // initialize horizontal stuff
/* precalculate vertical scaler filter coefficients */
{
const int filterAlign=
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 2 :
const int filterAlign =
(HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 2 :
(HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
1;
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, c->param) < 0)
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,
c->param) < 0)
goto fail;
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, c->param) < 0)
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,
c->param) < 0)
goto fail;
#if HAVE_ALTIVEC
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);
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);
for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
for (i = 0; i < c->vLumFilterSize * c->dstH; i++) {
int j;
short *p = (short *)&c->vYCoeffsBank[i];
for (j=0;j<8;j++)
for (j = 0; j < 8; j++)
p[j] = c->vLumFilter[i];
}
for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) {
int j;
short *p = (short *)&c->vCCoeffsBank[i];
for (j=0;j<8;j++)
for (j = 0; j < 8; j++)
p[j] = c->vChrFilter[i];
}
#endif
}
// calculate buffer sizes so that they won't run out while handling these damn slices
c->vLumBufSize= c->vLumFilterSize;
c->vChrBufSize= c->vChrFilterSize;
for (i=0; i<dstH; i++) {
int chrI = (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];
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];
}
// 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*2*sizeof(int16_t*), fail);
FF_ALLOC_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail);
FF_ALLOC_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail);
/* 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);
if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);
//Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna replace the 4000/8000)
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). */
/* align at 16 bytes for AltiVec */
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];
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];
}
// 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
c->uv_offx2 = dst_stride + 16;
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;
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;
}
if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
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];
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];
}
//try to avoid drawing green stuff between the right end and the stride end
for (i=0; i<c->vChrBufSize; i++)
// try to avoid drawing green stuff between the right end and the stride end
for (i = 0; i < c->vChrBufSize; i++)
if(av_pix_fmt_descriptors[c->dstFormat].comp[0].depth_minus1 == 15){
av_assert0(c->dstBpc > 10);
for(j=0; j<dst_stride/2+1; j++)
......@@ -1111,79 +1218,103 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
assert(c->chrDstH <= dstH);
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?! ");
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?! ");
av_log(c, AV_LOG_INFO, "from %s to %s%s ",
av_get_pix_fmt_name(srcFormat),
#ifdef DITHER1XBPP
dstFormat == PIX_FMT_BGR555 || dstFormat == PIX_FMT_BGR565 ||
dstFormat == PIX_FMT_BGR555 || dstFormat == PIX_FMT_BGR565 ||
dstFormat == PIX_FMT_RGB444BE || dstFormat == PIX_FMT_RGB444LE ||
dstFormat == PIX_FMT_BGR444BE || dstFormat == PIX_FMT_BGR444LE ? "dithered " : "",
dstFormat == PIX_FMT_BGR444BE || dstFormat == PIX_FMT_BGR444LE ?
"dithered " : "",
#else
"",
#endif
av_get_pix_fmt_name(dstFormat));
if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2) av_log(c, AV_LOG_INFO, "using MMX2\n");
else if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW) av_log(c, AV_LOG_INFO, "using 3DNOW\n");
else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) 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");
if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2)
av_log(c, AV_LOG_INFO, "using MMX2\n");
else if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW)
av_log(c, AV_LOG_INFO, "using 3DNOW\n");
else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX)
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");
av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
av_log(c, AV_LOG_DEBUG,
"lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
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);
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);
}
c->swScale= ff_getSwsFunc(c);
c->swScale = ff_getSwsFunc(c);
return 0;
fail: //FIXME replace things by appropriate error codes
fail: // FIXME replace things by appropriate error codes
return -1;
}
#if FF_API_SWS_GETCONTEXT
SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat,
int dstW, int dstH, enum PixelFormat dstFormat, int flags,
SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
int dstW, int dstH, enum PixelFormat dstFormat,
int flags, SwsFilter *srcFilter,
SwsFilter *dstFilter, const double *param)
{
SwsContext *c;
if(!(c=sws_alloc_context()))
if (!(c = sws_alloc_context()))
return NULL;
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);
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);
c->src0Alpha = handle_0alpha(&srcFormat);
c->dst0Alpha = handle_0alpha(&dstFormat);
c->srcFormat= srcFormat;
c->dstFormat= dstFormat;
c->srcFormat = srcFormat;
c->dstFormat = dstFormat;
if (param) {
c->param[0] = param[0];
c->param[1] = param[1];
}
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);
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);
if(sws_init_context(c, srcFilter, dstFilter) < 0){
if (sws_init_context(c, srcFilter, dstFilter) < 0) {
sws_freeContext(c);
return NULL;
}
......@@ -1197,28 +1328,28 @@ SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
float chromaHShift, float chromaVShift,
int verbose)
{
SwsFilter *filter= av_malloc(sizeof(SwsFilter));
SwsFilter *filter = av_malloc(sizeof(SwsFilter));
if (!filter)
return NULL;
if (lumaGBlur!=0.0) {
filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
if (lumaGBlur != 0.0) {
filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0);
} else {
filter->lumH= sws_getIdentityVec();
filter->lumV= sws_getIdentityVec();
filter->lumH = sws_getIdentityVec();
filter->lumV = sws_getIdentityVec();
}
if (chromaGBlur!=0.0) {
filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
if (chromaGBlur != 0.0) {
filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0);
} else {
filter->chrH= sws_getIdentityVec();
filter->chrV= sws_getIdentityVec();
filter->chrH = sws_getIdentityVec();
filter->chrV = sws_getIdentityVec();
}
if (chromaSharpen!=0.0) {
SwsVector *id= sws_getIdentityVec();
if (chromaSharpen != 0.0) {
SwsVector *id = sws_getIdentityVec();
sws_scaleVec(filter->chrH, -chromaSharpen);
sws_scaleVec(filter->chrV, -chromaSharpen);
sws_addVec(filter->chrH, id);
......@@ -1226,8 +1357,8 @@ SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
sws_freeVec(id);
}
if (lumaSharpen!=0.0) {
SwsVector *id= sws_getIdentityVec();
if (lumaSharpen != 0.0) {
SwsVector *id = sws_getIdentityVec();
sws_scaleVec(filter->lumH, -lumaSharpen);
sws_scaleVec(filter->lumV, -lumaSharpen);
sws_addVec(filter->lumH, id);
......@@ -1236,18 +1367,20 @@ SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
}
if (chromaHShift != 0.0)
sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5));
if (chromaVShift != 0.0)
sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5));
sws_normalizeVec(filter->chrH, 1.0);
sws_normalizeVec(filter->chrV, 1.0);
sws_normalizeVec(filter->lumH, 1.0);
sws_normalizeVec(filter->lumV, 1.0);
if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
if (verbose)
sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
if (verbose)
sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
return filter;
}
......@@ -1266,17 +1399,18 @@ SwsVector *sws_allocVec(int length)
SwsVector *sws_getGaussianVec(double variance, double quality)
{
const int length= (int)(variance*quality + 0.5) | 1;
const int length = (int)(variance * quality + 0.5) | 1;
int i;
double middle= (length-1)*0.5;
SwsVector *vec= sws_allocVec(length);
double middle = (length - 1) * 0.5;
SwsVector *vec = sws_allocVec(length);
if (!vec)
return NULL;
for (i=0; i<length; i++) {
double dist= i-middle;
vec->coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*M_PI);
for (i = 0; i < length; i++) {
double dist = i - middle;
vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
sqrt(2 * variance * M_PI);
}
sws_normalizeVec(vec, 1.0);
......@@ -1287,13 +1421,13 @@ SwsVector *sws_getGaussianVec(double variance, double quality)
SwsVector *sws_getConstVec(double c, int length)
{
int i;
SwsVector *vec= sws_allocVec(length);
SwsVector *vec = sws_allocVec(length);
if (!vec)
return NULL;
for (i=0; i<length; i++)
vec->coeff[i]= c;
for (i = 0; i < length; i++)
vec->coeff[i] = c;
return vec;
}
......@@ -1306,10 +1440,10 @@ SwsVector *sws_getIdentityVec(void)
static double sws_dcVec(SwsVector *a)
{
int i;
double sum=0;
double sum = 0;
for (i=0; i<a->length; i++)
sum+= a->coeff[i];
for (i = 0; i < a->length; i++)
sum += a->coeff[i];
return sum;
}
......@@ -1318,27 +1452,27 @@ void sws_scaleVec(SwsVector *a, double scalar)
{
int i;
for (i=0; i<a->length; i++)
a->coeff[i]*= scalar;
for (i = 0; i < a->length; i++)
a->coeff[i] *= scalar;
}
void sws_normalizeVec(SwsVector *a, double height)
{
sws_scaleVec(a, height/sws_dcVec(a));
sws_scaleVec(a, height / sws_dcVec(a));
}
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
{
int length= a->length + b->length - 1;
int length = a->length + b->length - 1;
int i, j;
SwsVector *vec= sws_getConstVec(0.0, length);
SwsVector *vec = sws_getConstVec(0.0, length);
if (!vec)
return NULL;
for (i=0; i<a->length; i++) {
for (j=0; j<b->length; j++) {
vec->coeff[i+j]+= a->coeff[i]*b->coeff[j];
for (i = 0; i < a->length; i++) {
for (j = 0; j < b->length; j++) {
vec->coeff[i + j] += a->coeff[i] * b->coeff[j];
}
}
......@@ -1347,30 +1481,34 @@ static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
{
int length= FFMAX(a->length, b->length);
int length = FFMAX(a->length, b->length);
int i;
SwsVector *vec= sws_getConstVec(0.0, length);
SwsVector *vec = sws_getConstVec(0.0, length);
if (!vec)
return NULL;
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];
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];
return vec;
}
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
{
int length= FFMAX(a->length, b->length);
int length = FFMAX(a->length, b->length);
int i;
SwsVector *vec= sws_getConstVec(0.0, length);
SwsVector *vec = sws_getConstVec(0.0, length);
if (!vec)
return NULL;
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];
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];
return vec;
}
......@@ -1378,15 +1516,16 @@ static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
/* shift left / or right if "shift" is negative */
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
{
int length= a->length + FFABS(shift)*2;
int length = a->length + FFABS(shift) * 2;
int i;
SwsVector *vec= sws_getConstVec(0.0, length);
SwsVector *vec = sws_getConstVec(0.0, length);
if (!vec)
return NULL;
for (i=0; i<a->length; i++) {
vec->coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
for (i = 0; i < a->length; i++) {
vec->coeff[i + (length - 1) / 2 -
(a->length - 1) / 2 - shift] = a->coeff[i];
}
return vec;
......@@ -1394,49 +1533,50 @@ static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
void sws_shiftVec(SwsVector *a, int shift)
{
SwsVector *shifted= sws_getShiftedVec(a, shift);
SwsVector *shifted = sws_getShiftedVec(a, shift);
av_free(a->coeff);
a->coeff= shifted->coeff;
a->length= shifted->length;
a->coeff = shifted->coeff;
a->length = shifted->length;
av_free(shifted);
}
void sws_addVec(SwsVector *a, SwsVector *b)
{
SwsVector *sum= sws_sumVec(a, b);
SwsVector *sum = sws_sumVec(a, b);
av_free(a->coeff);
a->coeff= sum->coeff;
a->length= sum->length;
a->coeff = sum->coeff;
a->length = sum->length;
av_free(sum);
}
void sws_subVec(SwsVector *a, SwsVector *b)
{
SwsVector *diff= sws_diffVec(a, b);
SwsVector *diff = sws_diffVec(a, b);
av_free(a->coeff);
a->coeff= diff->coeff;
a->length= diff->length;
a->coeff = diff->coeff;
a->length = diff->length;
av_free(diff);
}
void sws_convVec(SwsVector *a, SwsVector *b)
{
SwsVector *conv= sws_getConvVec(a, b);
SwsVector *conv = sws_getConvVec(a, b);
av_free(a->coeff);
a->coeff= conv->coeff;
a->length= conv->length;
a->coeff = conv->coeff;
a->length = conv->length;
av_free(conv);
}
SwsVector *sws_cloneVec(SwsVector *a)
{
int i;
SwsVector *vec= sws_allocVec(a->length);
SwsVector *vec = sws_allocVec(a->length);
if (!vec)
return NULL;
for (i=0; i<a->length; i++) vec->coeff[i]= a->coeff[i];
for (i = 0; i < a->length; i++)
vec->coeff[i] = a->coeff[i];
return vec;
}
......@@ -1444,65 +1584,75 @@ SwsVector *sws_cloneVec(SwsVector *a)
void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
{
int i;
double max=0;
double min=0;
double max = 0;
double min = 0;
double range;
for (i=0; i<a->length; i++)
if (a->coeff[i]>max) max= a->coeff[i];
for (i = 0; i < a->length; i++)
if (a->coeff[i] > max)
max = a->coeff[i];
for (i=0; i<a->length; i++)
if (a->coeff[i]<min) min= a->coeff[i];
for (i = 0; i < a->length; i++)
if (a->coeff[i] < min)
min = a->coeff[i];
range= max - min;
range = max - min;
for (i=0; i<a->length; i++) {
int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
for (i = 0; i < a->length; i++) {
int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5);
av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
for (;x>0; x--) av_log(log_ctx, log_level, " ");
for (; x > 0; x--)
av_log(log_ctx, log_level, " ");
av_log(log_ctx, log_level, "|\n");
}
}
void sws_freeVec(SwsVector *a)
{
if (!a) return;
if (!a)
return;
av_freep(&a->coeff);
a->length=0;
a->length = 0;
av_free(a);
}
void sws_freeFilter(SwsFilter *filter)
{
if (!filter) return;
if (filter->lumH) sws_freeVec(filter->lumH);
if (filter->lumV) sws_freeVec(filter->lumV);
if (filter->chrH) sws_freeVec(filter->chrH);
if (filter->chrV) sws_freeVec(filter->chrV);
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);
av_free(filter);
}
void sws_freeContext(SwsContext *c)
{
int i;
if (!c) return;
if (!c)
return;
if (c->lumPixBuf) {
for (i=0; i<c->vLumBufSize; i++)
for (i = 0; i < c->vLumBufSize; i++)
av_freep(&c->lumPixBuf[i]);
av_freep(&c->lumPixBuf);
}
if (c->chrUPixBuf) {
for (i=0; i<c->vChrBufSize; i++)
for (i = 0; i < c->vChrBufSize; i++)
av_freep(&c->chrUPixBuf[i]);
av_freep(&c->chrUPixBuf);
av_freep(&c->chrVPixBuf);
}
if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
for (i=0; i<c->vLumBufSize; i++)
for (i = 0; i < c->vLumBufSize; i++)
av_freep(&c->alpPixBuf[i]);
av_freep(&c->alpPixBuf);
}
......@@ -1523,17 +1673,21 @@ void sws_freeContext(SwsContext *c)
#if HAVE_MMX
#ifdef MAP_ANONYMOUS
if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);
if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);
if (c->lumMmx2FilterCode)
munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);
if (c->chrMmx2FilterCode)
munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);
#elif HAVE_VIRTUALALLOC
if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, 0, MEM_RELEASE);
if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, 0, MEM_RELEASE);
if (c->lumMmx2FilterCode)
VirtualFree(c->lumMmx2FilterCode, 0, MEM_RELEASE);
if (c->chrMmx2FilterCode)
VirtualFree(c->chrMmx2FilterCode, 0, MEM_RELEASE);
#else
av_free(c->lumMmx2FilterCode);
av_free(c->chrMmx2FilterCode);
#endif
c->lumMmx2FilterCode=NULL;
c->chrMmx2FilterCode=NULL;
c->lumMmx2FilterCode = NULL;
c->chrMmx2FilterCode = NULL;
#endif /* HAVE_MMX */
av_freep(&c->yuvTable);
......@@ -1542,12 +1696,16 @@ void sws_freeContext(SwsContext *c)
av_free(c);
}
struct SwsContext *sws_getCachedContext(struct SwsContext *context,
int srcW, int srcH, enum PixelFormat srcFormat,
int dstW, int dstH, enum PixelFormat dstFormat, int flags,
SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW,
int srcH, enum PixelFormat srcFormat,
int dstW, int dstH,
enum PixelFormat dstFormat, int flags,
SwsFilter *srcFilter,
SwsFilter *dstFilter,
const double *param)
{
static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
static const double default_param[2] = { SWS_PARAM_DEFAULT,
SWS_PARAM_DEFAULT };
if (!param)
param = default_param;
......@@ -1582,7 +1740,10 @@ struct SwsContext *sws_getCachedContext(struct SwsContext *context,
context->flags = flags;
context->param[0] = param[0];
context->param[1] = param[1];
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);
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);
if (sws_init_context(context, srcFilter, dstFilter) < 0) {
sws_freeContext(context);
return NULL;
......
libswscale/x86/Makefile 0 → 100644
View file @ 367d9b29
$(SUBDIR)x86/swscale_mmx.o: CFLAGS += $(NOREDZONE_FLAGS)
OBJS-$(CONFIG_XMM_CLOBBER_TEST) += x86/w64xmmtest.o
MMX-OBJS += x86/rgb2rgb.o \
x86/swscale_mmx.o \
x86/yuv2rgb_mmx.o \
YASM-OBJS += x86/input.o \
x86/output.o \
x86/scale.o \
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment