Skip to content

F
ffmpeg
Commit 62d176de authored by Pedro Arthur's avatar Pedro Arthur
Browse files
Options

swscale: refactor vertical scaler

parent fffae8e6
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 93 additions and 42 deletions
libswscale/Makefile
View file @ 62d176de
......@@ -17,6 +17,7 @@ OBJS = alphablend.o \
yuv2rgb.o \
slice.o \
hscale.o \
vscale.o \
OBJS-$(CONFIG_SHARED) += log2_tab.o
......
libswscale/slice.c
View file @ 62d176de
......@@ -214,6 +214,7 @@ int ff_init_filters(SwsContext * c)
int index;
int num_ydesc;
int num_cdesc;
int num_vdesc = isPlanarYUV(c->dstFormat) && !isGray(c->dstFormat) ? 2 : 1;
int need_lum_conv = c->lumToYV12 || c->readLumPlanar || c->alpToYV12 || c->readAlpPlanar;
int need_chr_conv = c->chrToYV12 || c->readChrPlanar;
int srcIdx, dstIdx;
......@@ -228,8 +229,8 @@ int ff_init_filters(SwsContext * c)
num_ydesc = need_lum_conv ? 2 : 1;
num_cdesc = need_chr_conv ? 2 : 1;
c->numSlice = FFMAX(num_ydesc, num_cdesc) + 1;
c->numDesc = num_ydesc + num_cdesc;
c->numSlice = FFMAX(num_ydesc, num_cdesc) + 2;
c->numDesc = num_ydesc + num_cdesc + num_vdesc;
c->descIndex[0] = num_ydesc;
c->descIndex[1] = num_ydesc + num_cdesc;
......@@ -243,12 +244,13 @@ int ff_init_filters(SwsContext * c)
res = alloc_slice(&c->slice[0], c->srcFormat, c->srcH, c->chrSrcH, c->chrSrcHSubSample, c->chrSrcVSubSample, 0);
if (res < 0) goto cleanup;
for (i = 1; i < c->numSlice-1; ++i) {
for (i = 1; i < c->numSlice-2; ++i) {
res = alloc_slice(&c->slice[i], c->srcFormat, c->vLumFilterSize + MAX_LINES_AHEAD, c->vChrFilterSize + MAX_LINES_AHEAD, c->chrSrcHSubSample, c->chrSrcVSubSample, 0);
if (res < 0) goto cleanup;
res = alloc_lines(&c->slice[i], FFALIGN(c->srcW*2+78, 16), c->srcW);
if (res < 0) goto cleanup;
}
// horizontal scaler output
res = alloc_slice(&c->slice[i], c->srcFormat, c->vLumFilterSize + MAX_LINES_AHEAD, c->vChrFilterSize + MAX_LINES_AHEAD, c->chrDstHSubSample, c->chrDstVSubSample, 1);
if (res < 0) goto cleanup;
res = alloc_lines(&c->slice[i], dst_stride, c->dstW);
......@@ -256,6 +258,11 @@ int ff_init_filters(SwsContext * c)
fill_ones(&c->slice[i], dst_stride>>1, c->dstBpc == 16);
// vertical scaler output
++i;
res = alloc_slice(&c->slice[i], c->dstFormat, c->dstH, c->chrDstH, c->chrDstHSubSample, c->chrDstVSubSample, 0);
if (res < 0) goto cleanup;
index = 0;
srcIdx = 0;
dstIdx = 1;
......@@ -290,6 +297,13 @@ int ff_init_filters(SwsContext * c)
ff_init_desc_no_chr(&c->desc[index], &c->slice[srcIdx], &c->slice[dstIdx]);
}
++index;
{
srcIdx = c->numSlice - 2;
dstIdx = c->numSlice - 1;
ff_init_vscale(c, c->desc + index, c->slice + srcIdx, c->slice + dstIdx);
}
return 0;
cleanup:
......
libswscale/swscale.c
View file @ 62d176de
......@@ -326,8 +326,8 @@ static int swscale(SwsContext *c, const uint8_t *src[],
#endif
const int dstW = c->dstW;
const int dstH = c->dstH;
const int chrDstW = c->chrDstW;
#ifndef NEW_FILTER
const int chrDstW = c->chrDstW;
const int chrSrcW = c->chrSrcW;
const int lumXInc = c->lumXInc;
const int chrXInc = c->chrXInc;
......@@ -341,9 +341,9 @@ static int swscale(SwsContext *c, const uint8_t *src[],
int32_t *hChrFilterPos = c->hChrFilterPos;
int16_t *hLumFilter = c->hLumFilter;
int16_t *hChrFilter = c->hChrFilter;
#endif
int32_t *lumMmxFilter = c->lumMmxFilter;
int32_t *chrMmxFilter = c->chrMmxFilter;
#endif
const int vLumFilterSize = c->vLumFilterSize;
const int vChrFilterSize = c->vChrFilterSize;
#ifndef NEW_FILTER
......@@ -381,13 +381,18 @@ static int swscale(SwsContext *c, const uint8_t *src[],
int lastInChrBuf = c->lastInChrBuf;
// int perform_gamma = c->is_internal_gamma;
#ifdef NEW_FILTER
int lumStart = 0;
int lumEnd = c->descIndex[0];
int chrStart = lumEnd;
int chrEnd = c->descIndex[1];
int vStart = chrEnd;
int vEnd = c->numDesc;
SwsSlice *src_slice = &c->slice[lumStart];
SwsSlice *dst_slice = &c->slice[c->numSlice-1];
SwsSlice *hout_slice = &c->slice[c->numSlice-2];
SwsSlice *vout_slice = &c->slice[c->numSlice-1];
SwsFilterDescriptor *desc = c->desc;
#endif
int hasLumHoles = 1;
int hasChrHoles = 1;
......@@ -460,20 +465,28 @@ static int swscale(SwsContext *c, const uint8_t *src[],
}
lastDstY = dstY;
ff_init_slice_from_src(src_slice, (uint8_t**)src, srcStride, c->srcW,
srcSliceY, srcSliceH,
chrSrcSliceY, chrSrcSliceH);
dst_slice->plane[0].sliceY = lastInLumBuf + 1;
dst_slice->plane[1].sliceY = lastInChrBuf + 1;
dst_slice->plane[2].sliceY = lastInChrBuf + 1;
dst_slice->plane[3].sliceY = lastInLumBuf + 1;
#ifdef NEW_FILTER
ff_init_vscale_pfn(c, yuv2plane1, yuv2planeX, yuv2nv12cX,
yuv2packed1, yuv2packed2, yuv2packedX, yuv2anyX, c->use_mmx_vfilter);
dst_slice->plane[0].sliceH =
dst_slice->plane[1].sliceH =
dst_slice->plane[2].sliceH =
dst_slice->plane[3].sliceH = 0;
dst_slice->width = dstW;
ff_init_slice_from_src(src_slice, (uint8_t**)src, srcStride, c->srcW,
srcSliceY, srcSliceH, chrSrcSliceY, chrSrcSliceH);
ff_init_slice_from_src(vout_slice, (uint8_t**)dst, dstStride, c->dstW,
dstY, dstH, dstY >> c->chrDstVSubSample,
FF_CEIL_RSHIFT(dstH, c->chrDstVSubSample));
hout_slice->plane[0].sliceY = lastInLumBuf + 1;
hout_slice->plane[1].sliceY = lastInChrBuf + 1;
hout_slice->plane[2].sliceY = lastInChrBuf + 1;
hout_slice->plane[3].sliceY = lastInLumBuf + 1;
hout_slice->plane[0].sliceH =
hout_slice->plane[1].sliceH =
hout_slice->plane[2].sliceH =
hout_slice->plane[3].sliceH = 0;
hout_slice->width = dstW;
#endif
for (; dstY < dstH; dstY++) {
const int chrDstY = dstY >> c->chrDstVSubSample;
......@@ -496,29 +509,35 @@ static int swscale(SwsContext *c, const uint8_t *src[],
int lastLumSrcY2 = FFMIN(c->srcH, firstLumSrcY2 + vLumFilterSize) - 1;
int lastChrSrcY = FFMIN(c->chrSrcH, firstChrSrcY + vChrFilterSize) - 1;
int enough_lines;
#ifdef NEW_FILTER
int i;
int posY, cPosY, firstPosY, lastPosY, firstCPosY, lastCPosY;
#endif
// handle holes (FAST_BILINEAR & weird filters)
if (firstLumSrcY > lastInLumBuf) {
#ifdef NEW_FILTER
hasLumHoles = lastInLumBuf != firstLumSrcY - 1;
lastInLumBuf = firstLumSrcY - 1;
if (hasLumHoles) {
dst_slice->plane[0].sliceY = lastInLumBuf + 1;
dst_slice->plane[3].sliceY = lastInLumBuf + 1;
dst_slice->plane[0].sliceH =
dst_slice->plane[3].sliceH = 0;
hout_slice->plane[0].sliceY = lastInLumBuf + 1;
hout_slice->plane[3].sliceY = lastInLumBuf + 1;
hout_slice->plane[0].sliceH =
hout_slice->plane[3].sliceH = 0;
}
#endif
lastInLumBuf = firstLumSrcY - 1;
}
if (firstChrSrcY > lastInChrBuf) {
#ifdef NEW_FILTER
hasChrHoles = lastInChrBuf != firstChrSrcY - 1;
lastInChrBuf = firstChrSrcY - 1;
if (hasChrHoles) {
dst_slice->plane[1].sliceY = lastInChrBuf + 1;
dst_slice->plane[2].sliceY = lastInChrBuf + 1;
dst_slice->plane[1].sliceH =
dst_slice->plane[2].sliceH = 0;
hout_slice->plane[1].sliceY = lastInChrBuf + 1;
hout_slice->plane[2].sliceY = lastInChrBuf + 1;
hout_slice->plane[1].sliceH =
hout_slice->plane[2].sliceH = 0;
}
#endif
lastInChrBuf = firstChrSrcY - 1;
}
av_assert0(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1);
av_assert0(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1);
......@@ -541,7 +560,7 @@ static int swscale(SwsContext *c, const uint8_t *src[],
}
#ifdef NEW_FILTER
posY = dst_slice->plane[0].sliceY + dst_slice->plane[0].sliceH;
posY = hout_slice->plane[0].sliceY + hout_slice->plane[0].sliceH;
if (posY <= lastLumSrcY && !hasLumHoles) {
firstPosY = FFMAX(firstLumSrcY, posY);
lastPosY = FFMIN(lastLumSrcY + MAX_LINES_AHEAD, srcSliceY + srcSliceH - 1);
......@@ -550,7 +569,7 @@ static int swscale(SwsContext *c, const uint8_t *src[],
lastPosY = lastLumSrcY;
}
cPosY = dst_slice->plane[1].sliceY + dst_slice->plane[1].sliceH;
cPosY = hout_slice->plane[1].sliceY + hout_slice->plane[1].sliceH;
if (cPosY <= lastChrSrcY && !hasChrHoles) {
firstCPosY = FFMAX(firstChrSrcY, cPosY);
lastCPosY = FFMIN(lastChrSrcY + MAX_LINES_AHEAD, FF_CEIL_RSHIFT(srcSliceY + srcSliceH, c->chrSrcVSubSample) - 1);
......@@ -559,7 +578,7 @@ static int swscale(SwsContext *c, const uint8_t *src[],
lastCPosY = lastChrSrcY;
}
ff_rotate_slice(dst_slice, lastPosY, lastCPosY);
ff_rotate_slice(hout_slice, lastPosY, lastCPosY);
if (posY < lastLumSrcY + 1) {
for (i = lumStart; i < lumEnd; ++i)
......@@ -650,22 +669,20 @@ static int swscale(SwsContext *c, const uint8_t *src[],
ff_sws_init_output_funcs(c, &yuv2plane1, &yuv2planeX, &yuv2nv12cX,
&yuv2packed1, &yuv2packed2, &yuv2packedX, &yuv2anyX);
use_mmx_vfilter= 0;
ff_init_vscale_pfn(c, yuv2plane1, yuv2planeX, yuv2nv12cX,
yuv2packed1, yuv2packed2, yuv2packedX, yuv2anyX, use_mmx_vfilter);
}
{
#ifdef NEW_FILTER
const int16_t **lumSrcPtr = (const int16_t **)(void*) dst_slice->plane[0].line + firstLumSrcY - dst_slice->plane[0].sliceY;
const int16_t **chrUSrcPtr = (const int16_t **)(void*) dst_slice->plane[1].line + firstChrSrcY - dst_slice->plane[1].sliceY;
const int16_t **chrVSrcPtr = (const int16_t **)(void*) dst_slice->plane[2].line + firstChrSrcY - dst_slice->plane[2].sliceY;
const int16_t **alpSrcPtr = (CONFIG_SWSCALE_ALPHA && alpPixBuf) ?
(const int16_t **)(void*) dst_slice->plane[3].line + firstLumSrcY - dst_slice->plane[3].sliceY : NULL;
for (i = vStart; i < vEnd; ++i)
desc[i].process(c, &desc[i], dstY, 1);
#else
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;
#endif
int16_t *vLumFilter = c->vLumFilter;
int16_t *vChrFilter = c->vChrFilter;
......@@ -766,6 +783,7 @@ static int swscale(SwsContext *c, const uint8_t *src[],
chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
alpSrcPtr, dest, dstW, dstY);
}
#endif
//if (perform_gamma)
// gamma_convert(dest, dstW, c->gamma);
}
......
libswscale/swscale_internal.h
View file @ 62d176de
......@@ -1002,6 +1002,15 @@ typedef struct FilterContext
int xInc;
} FilterContext;
typedef struct VScalerContext
{
uint16_t *filter[2];
int32_t *filter_pos;
int filter_size;
int isMMX;
void *pfn;
} VScalerContext;
// warp input lines in the form (src + width*i + j) to slice format (line[i][j])
int ff_init_slice_from_src(SwsSlice * s, uint8_t *src[4], int stride[4], int srcW, int lumY, int lumH, int chrY, int chrH);
......@@ -1033,9 +1042,18 @@ int ff_init_desc_chscale(SwsFilterDescriptor *desc, SwsSlice *src, SwsSlice *dst
int ff_init_desc_no_chr(SwsFilterDescriptor *desc, SwsSlice * src, SwsSlice *dst);
/// initializes vertical scaling descriptors
int ff_init_vscale(SwsContext *c, SwsFilterDescriptor *desc, SwsSlice *src, SwsSlice *dst);
/// setup vertical scaler functions
void ff_init_vscale_pfn(SwsContext *c, yuv2planar1_fn yuv2plane1, yuv2planarX_fn yuv2planeX,
yuv2interleavedX_fn yuv2nv12cX, yuv2packed1_fn yuv2packed1, yuv2packed2_fn yuv2packed2,
yuv2packedX_fn yuv2packedX, yuv2anyX_fn yuv2anyX, int use_mmx);
//number of extra lines to process
#define MAX_LINES_AHEAD 4
// enable use of refactored scaler code
#define NEW_FILTER
#endif /* SWSCALE_SWSCALE_INTERNAL_H */
libswscale/vscale.c 0 → 100644
View file @ 62d176de
This diff is collapsed.
libswscale/x86/swscale.c
View file @ 62d176de
......@@ -86,9 +86,9 @@ void ff_updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrB
const int dstH= c->dstH;
const int flags= c->flags;
#ifdef NEW_FILTER
SwsPlane *lumPlane = &c->slice[c->numSlice-1].plane[0];
SwsPlane *chrUPlane = &c->slice[c->numSlice-1].plane[1];
SwsPlane *alpPlane = &c->slice[c->numSlice-1].plane[3];
SwsPlane *lumPlane = &c->slice[c->numSlice-2].plane[0];
SwsPlane *chrUPlane = &c->slice[c->numSlice-2].plane[1];
SwsPlane *alpPlane = &c->slice[c->numSlice-2].plane[3];
#else
int16_t **lumPixBuf= c->lumPixBuf;
int16_t **chrUPixBuf= c->chrUPixBuf;
......
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