swscale: eliminate redundant SwsInternal accesses

This is a purely cosmetic commit aimed at replacing accesses to
SwsInternal.opts by direct access to SwsContext wherever convenient.

Sponsored-by: Sovereign Tech Fund
Signed-off-by: Niklas Haas <git@haasn.dev>
master
Niklas Haas 2 months ago
parent ed5dd67562
commit 6a91a165fd
  1. 2
      libswscale/options.c
  2. 93
      libswscale/swscale.c
  3. 233
      libswscale/utils.c
  4. 16
      tests/checkasm/sw_gbrp.c
  5. 16
      tests/checkasm/sw_range_convert.c
  6. 29
      tests/checkasm/sw_rgb.c
  7. 22
      tests/checkasm/sw_scale.c

@ -27,7 +27,7 @@ static const char *sws_context_to_name(void *ptr)
return "swscaler";
}
#define OFFSET(x) offsetof(SwsInternal, opts.x)
#define OFFSET(x) offsetof(SwsContext, x)
#define DEFAULT 0
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM

@ -895,7 +895,7 @@ static int scale_cascaded(SwsInternal *c,
uint8_t * const dstSlice[], const int dstStride[],
int dstSliceY, int dstSliceH)
{
const int dstH0 = sws_internal(c->cascaded_context[0])->opts.dst_h;
const int dstH0 = c->cascaded_context[0]->dst_h;
int ret = scale_internal(c->cascaded_context[0],
srcSlice, srcStride, srcSliceY, srcSliceH,
c->cascaded_tmp[0], c->cascaded_tmpStride[0],
@ -915,13 +915,13 @@ static int scale_internal(SwsContext *sws,
int dstSliceY, int dstSliceH)
{
SwsInternal *c = sws_internal(sws);
const int scale_dst = dstSliceY > 0 || dstSliceH < c->opts.dst_h;
const int scale_dst = dstSliceY > 0 || dstSliceH < sws->dst_h;
const int frame_start = scale_dst || !c->sliceDir;
int i, ret;
const uint8_t *src2[4];
uint8_t *dst2[4];
int macro_height_src = isBayer(c->opts.src_format) ? 2 : (1 << c->chrSrcVSubSample);
int macro_height_dst = isBayer(c->opts.dst_format) ? 2 : (1 << c->chrDstVSubSample);
int macro_height_src = isBayer(sws->src_format) ? 2 : (1 << c->chrSrcVSubSample);
int macro_height_dst = isBayer(sws->dst_format) ? 2 : (1 << c->chrDstVSubSample);
// copy strides, so they can safely be modified
int srcStride2[4];
int dstStride2[4];
@ -933,25 +933,25 @@ static int scale_internal(SwsContext *sws,
}
if ((srcSliceY & (macro_height_src - 1)) ||
((srcSliceH & (macro_height_src - 1)) && srcSliceY + srcSliceH != c->opts.src_h) ||
srcSliceY + srcSliceH > c->opts.src_h ||
(isBayer(c->opts.src_format) && srcSliceH <= 1)) {
((srcSliceH & (macro_height_src - 1)) && srcSliceY + srcSliceH != sws->src_h) ||
srcSliceY + srcSliceH > sws->src_h ||
(isBayer(sws->src_format) && srcSliceH <= 1)) {
av_log(c, AV_LOG_ERROR, "Slice parameters %d, %d are invalid\n", srcSliceY, srcSliceH);
return AVERROR(EINVAL);
}
if ((dstSliceY & (macro_height_dst - 1)) ||
((dstSliceH & (macro_height_dst - 1)) && dstSliceY + dstSliceH != c->opts.dst_h) ||
dstSliceY + dstSliceH > c->opts.dst_h) {
((dstSliceH & (macro_height_dst - 1)) && dstSliceY + dstSliceH != sws->dst_h) ||
dstSliceY + dstSliceH > sws->dst_h) {
av_log(c, AV_LOG_ERROR, "Slice parameters %d, %d are invalid\n", dstSliceY, dstSliceH);
return AVERROR(EINVAL);
}
if (!check_image_pointers(srcSlice, c->opts.src_format, srcStride)) {
if (!check_image_pointers(srcSlice, sws->src_format, srcStride)) {
av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
return AVERROR(EINVAL);
}
if (!check_image_pointers((const uint8_t* const*)dstSlice, c->opts.dst_format, dstStride)) {
if (!check_image_pointers((const uint8_t* const*)dstSlice, sws->dst_format, dstStride)) {
av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
return AVERROR(EINVAL);
}
@ -960,22 +960,19 @@ static int scale_internal(SwsContext *sws,
if (srcSliceH == 0)
return 0;
if (c->opts.gamma_flag && c->cascaded_context[0])
if (sws->gamma_flag && c->cascaded_context[0])
return scale_gamma(c, srcSlice, srcStride, srcSliceY, srcSliceH,
dstSlice, dstStride, dstSliceY, dstSliceH);
if (c->cascaded_context[0] && srcSliceY == 0 &&
srcSliceH == sws_internal(c->cascaded_context[0])->opts.src_h)
{
if (c->cascaded_context[0] && srcSliceY == 0 && srcSliceH == c->cascaded_context[0]->src_h)
return scale_cascaded(c, srcSlice, srcStride, srcSliceY, srcSliceH,
dstSlice, dstStride, dstSliceY, dstSliceH);
}
if (!srcSliceY && (c->opts.flags & SWS_BITEXACT) && c->opts.dither == SWS_DITHER_ED && c->dither_error[0])
if (!srcSliceY && (sws->flags & SWS_BITEXACT) && sws->dither == SWS_DITHER_ED && c->dither_error[0])
for (i = 0; i < 4; i++)
memset(c->dither_error[i], 0, sizeof(c->dither_error[0][0]) * (c->opts.dst_w+2));
memset(c->dither_error[i], 0, sizeof(c->dither_error[0][0]) * (sws->dst_w+2));
if (usePal(c->opts.src_format))
if (usePal(sws->src_format))
ff_update_palette(c, (const uint32_t *)srcSlice[1]);
memcpy(src2, srcSlice, sizeof(src2));
@ -984,7 +981,7 @@ static int scale_internal(SwsContext *sws,
memcpy(dstStride2, dstStride, sizeof(dstStride2));
if (frame_start && !scale_dst) {
if (srcSliceY != 0 && srcSliceY + srcSliceH != c->opts.src_h) {
if (srcSliceY != 0 && srcSliceY + srcSliceH != sws->src_h) {
av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
return AVERROR(EINVAL);
}
@ -993,7 +990,7 @@ static int scale_internal(SwsContext *sws,
} else if (scale_dst)
c->sliceDir = 1;
if (c->src0Alpha && !c->dst0Alpha && isALPHA(c->opts.dst_format)) {
if (c->src0Alpha && !c->dst0Alpha && isALPHA(sws->dst_format)) {
uint8_t *base;
int x,y;
@ -1005,15 +1002,15 @@ static int scale_internal(SwsContext *sws,
base = srcStride[0] < 0 ? c->rgb0_scratch - srcStride[0] * (srcSliceH-1) :
c->rgb0_scratch;
for (y=0; y<srcSliceH; y++){
memcpy(base + srcStride[0]*y, src2[0] + srcStride[0]*y, 4*c->opts.src_w);
for (x=c->src0Alpha-1; x<4*c->opts.src_w; x+=4) {
memcpy(base + srcStride[0]*y, src2[0] + srcStride[0]*y, 4*sws->src_w);
for (x=c->src0Alpha-1; x<4*sws->src_w; x+=4) {
base[ srcStride[0]*y + x] = 0xFF;
}
}
src2[0] = base;
}
if (c->srcXYZ && !(c->dstXYZ && c->opts.src_w==c->opts.dst_w && c->opts.src_h==c->opts.dst_h)) {
if (c->srcXYZ && !(c->dstXYZ && sws->src_w==sws->dst_w && sws->src_h==sws->dst_h)) {
uint8_t *base;
av_fast_malloc(&c->xyz_scratch, &c->xyz_scratch_allocated,
@ -1024,7 +1021,7 @@ static int scale_internal(SwsContext *sws,
base = srcStride[0] < 0 ? c->xyz_scratch - srcStride[0] * (srcSliceH-1) :
c->xyz_scratch;
ff_xyz12Torgb48(c, base, srcStride[0], src2[0], srcStride[0], c->opts.src_w, srcSliceH);
ff_xyz12Torgb48(c, base, srcStride[0], src2[0], srcStride[0], sws->src_w, srcSliceH);
src2[0] = base;
}
@ -1036,19 +1033,19 @@ static int scale_internal(SwsContext *sws,
}
src2[0] += (srcSliceH - 1) * srcStride[0];
if (!usePal(c->opts.src_format))
if (!usePal(sws->src_format))
src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];
src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];
src2[3] += (srcSliceH - 1) * srcStride[3];
dst2[0] += ( c->opts.dst_h - 1) * dstStride[0];
dst2[1] += ((c->opts.dst_h >> c->chrDstVSubSample) - 1) * dstStride[1];
dst2[2] += ((c->opts.dst_h >> c->chrDstVSubSample) - 1) * dstStride[2];
dst2[3] += ( c->opts.dst_h - 1) * dstStride[3];
dst2[0] += ( sws->dst_h - 1) * dstStride[0];
dst2[1] += ((sws->dst_h >> c->chrDstVSubSample) - 1) * dstStride[1];
dst2[2] += ((sws->dst_h >> c->chrDstVSubSample) - 1) * dstStride[2];
dst2[3] += ( sws->dst_h - 1) * dstStride[3];
srcSliceY_internal = c->opts.src_h-srcSliceY-srcSliceH;
srcSliceY_internal = sws->src_h-srcSliceY-srcSliceH;
}
reset_ptr(src2, c->opts.src_format);
reset_ptr((void*)dst2, c->opts.dst_format);
reset_ptr(src2, sws->src_format);
reset_ptr((void*)dst2, sws->dst_format);
if (c->convert_unscaled) {
int offset = srcSliceY_internal;
@ -1058,13 +1055,13 @@ static int scale_internal(SwsContext *sws,
if (scale_dst) {
av_assert0(offset == 0);
for (i = 0; i < 4 && src2[i]; i++) {
if (!src2[i] || (i > 0 && usePal(c->opts.src_format)))
if (!src2[i] || (i > 0 && usePal(sws->src_format)))
break;
src2[i] += (dstSliceY >> ((i == 1 || i == 2) ? c->chrSrcVSubSample : 0)) * srcStride2[i];
}
for (i = 0; i < 4 && dst2[i]; i++) {
if (!dst2[i] || (i > 0 && usePal(c->opts.dst_format)))
if (!dst2[i] || (i > 0 && usePal(sws->dst_format)))
break;
dst2[i] -= (dstSliceY >> ((i == 1 || i == 2) ? c->chrDstVSubSample : 0)) * dstStride2[i];
}
@ -1081,7 +1078,7 @@ static int scale_internal(SwsContext *sws,
dst2, dstStride2, dstSliceY, dstSliceH);
}
if (c->dstXYZ && !(c->srcXYZ && c->opts.src_w==c->opts.dst_w && c->opts.src_h==c->opts.dst_h)) {
if (c->dstXYZ && !(c->srcXYZ && sws->src_w==sws->dst_w && sws->src_h==sws->dst_h)) {
uint8_t *dst;
if (scale_dst) {
@ -1091,16 +1088,16 @@ static int scale_internal(SwsContext *sws,
av_assert0(dstY >= ret);
av_assert0(ret >= 0);
av_assert0(c->opts.dst_h >= dstY);
av_assert0(sws->dst_h >= dstY);
dst = dst2[0] + (dstY - ret) * dstStride2[0];
}
/* replace on the same data */
ff_rgb48Toxyz12(c, dst, dstStride2[0], dst, dstStride2[0], c->opts.dst_w, ret);
ff_rgb48Toxyz12(c, dst, dstStride2[0], dst, dstStride2[0], sws->dst_w, ret);
}
/* reset slice direction at end of frame */
if ((srcSliceY_internal + srcSliceH == c->opts.src_h) || scale_dst)
if ((srcSliceY_internal + srcSliceH == sws->src_h) || scale_dst)
c->sliceDir = 0;
return ret;
@ -1124,9 +1121,9 @@ int sws_frame_start(SwsContext *sws, AVFrame *dst, const AVFrame *src)
return ret;
if (!dst->buf[0]) {
dst->width = c->opts.dst_w;
dst->height = c->opts.dst_h;
dst->format = c->opts.dst_format;
dst->width = sws->dst_w;
dst->height = sws->dst_h;
dst->format = sws->dst_format;
ret = av_frame_get_buffer(dst, 0);
if (ret < 0)
@ -1177,10 +1174,10 @@ int sws_receive_slice(SwsContext *sws, unsigned int slice_start,
/* wait until complete input has been received */
if (!(c->src_ranges.nb_ranges == 1 &&
c->src_ranges.ranges[0].start == 0 &&
c->src_ranges.ranges[0].len == c->opts.src_h))
c->src_ranges.ranges[0].len == sws->src_h))
return AVERROR(EAGAIN);
if ((slice_start > 0 || slice_height < c->opts.dst_h) &&
if ((slice_start > 0 || slice_height < sws->dst_h) &&
(slice_start % align || slice_height % align)) {
av_log(c, AV_LOG_ERROR,
"Incorrectly aligned output: %u/%u not multiples of %u\n",
@ -1192,7 +1189,7 @@ int sws_receive_slice(SwsContext *sws, unsigned int slice_start,
int nb_jobs = c->nb_slice_ctx;
int ret = 0;
if (sws_internal(c->slice_ctx[0])->opts.dither == SWS_DITHER_ED)
if (c->slice_ctx[0]->dither == SWS_DITHER_ED)
nb_jobs = 1;
c->dst_slice_start = slice_start;
@ -1218,7 +1215,7 @@ int sws_receive_slice(SwsContext *sws, unsigned int slice_start,
}
return scale_internal(sws, (const uint8_t * const *)c->frame_src->data,
c->frame_src->linesize, 0, c->opts.src_h,
c->frame_src->linesize, 0, sws->src_h,
dst, c->frame_dst->linesize, slice_start, slice_height);
}
@ -1256,7 +1253,7 @@ int attribute_align_arg sws_scale(SwsContext *sws,
}
return scale_internal(sws, srcSlice, srcStride, srcSliceY, srcSliceH,
dst, dstStride, 0, c->opts.dst_h);
dst, dstStride, 0, sws->dst_h);
}
void ff_sws_slice_worker(void *priv, int jobnr, int threadnr,
@ -1284,7 +1281,7 @@ void ff_sws_slice_worker(void *priv, int jobnr, int threadnr,
}
err = scale_internal(sws, (const uint8_t * const *)parent->frame_src->data,
parent->frame_src->linesize, 0, c->opts.src_h,
parent->frame_src->linesize, 0, sws->src_h,
dst, parent->frame_dst->linesize,
parent->dst_slice_start + slice_start, slice_end - slice_start);
}

@ -1026,12 +1026,13 @@ static int handle_xyz(enum AVPixelFormat *format)
}
}
static void handle_formats(SwsInternal *c)
static void handle_formats(SwsContext *sws)
{
c->src0Alpha |= handle_0alpha(&c->opts.src_format);
c->dst0Alpha |= handle_0alpha(&c->opts.dst_format);
c->srcXYZ |= handle_xyz(&c->opts.src_format);
c->dstXYZ |= handle_xyz(&c->opts.dst_format);
SwsInternal *c = sws_internal(sws);
c->src0Alpha |= handle_0alpha(&sws->src_format);
c->dst0Alpha |= handle_0alpha(&sws->dst_format);
c->srcXYZ |= handle_xyz(&sws->src_format);
c->dstXYZ |= handle_xyz(&sws->dst_format);
if (c->srcXYZ || c->dstXYZ)
fill_xyztables(c);
}
@ -1063,17 +1064,17 @@ int sws_setColorspaceDetails(SwsContext *sws, const int inv_table[4],
return parent_ret;
}
handle_formats(c);
desc_dst = av_pix_fmt_desc_get(c->opts.dst_format);
desc_src = av_pix_fmt_desc_get(c->opts.src_format);
handle_formats(sws);
desc_dst = av_pix_fmt_desc_get(sws->dst_format);
desc_src = av_pix_fmt_desc_get(sws->src_format);
if(range_override_needed(c->opts.dst_format))
if(range_override_needed(sws->dst_format))
dstRange = 0;
if(range_override_needed(c->opts.src_format))
if(range_override_needed(sws->src_format))
srcRange = 0;
if (c->opts.src_range != srcRange ||
c->opts.dst_range != dstRange ||
if (sws->src_range != srcRange ||
sws->dst_range != dstRange ||
c->brightness != brightness ||
c->contrast != contrast ||
c->saturation != saturation ||
@ -1090,8 +1091,8 @@ int sws_setColorspaceDetails(SwsContext *sws, const int inv_table[4],
c->brightness = brightness;
c->contrast = contrast;
c->saturation = saturation;
c->opts.src_range = srcRange;
c->opts.dst_range = dstRange;
sws->src_range = srcRange;
sws->dst_range = dstRange;
if (need_reinit)
ff_sws_init_range_convert(c);
@ -1105,27 +1106,27 @@ int sws_setColorspaceDetails(SwsContext *sws, const int inv_table[4],
if (!need_reinit)
return 0;
if ((isYUV(c->opts.dst_format) || isGray(c->opts.dst_format)) && (isYUV(c->opts.src_format) || isGray(c->opts.src_format))) {
if ((isYUV(sws->dst_format) || isGray(sws->dst_format)) && (isYUV(sws->src_format) || isGray(sws->src_format))) {
if (!c->cascaded_context[0] &&
memcmp(c->dstColorspaceTable, c->srcColorspaceTable, sizeof(int) * 4) &&
c->opts.src_w && c->opts.src_h && c->opts.dst_w && c->opts.dst_h) {
sws->src_w && sws->src_h && sws->dst_w && sws->dst_h) {
enum AVPixelFormat tmp_format;
int tmp_width, tmp_height;
int srcW = c->opts.src_w;
int srcH = c->opts.src_h;
int dstW = c->opts.dst_w;
int dstH = c->opts.dst_h;
int srcW = sws->src_w;
int srcH = sws->src_h;
int dstW = sws->dst_w;
int dstH = sws->dst_h;
int ret;
av_log(c, AV_LOG_VERBOSE, "YUV color matrix differs for YUV->YUV, using intermediate RGB to convert\n");
if (isNBPS(c->opts.dst_format) || is16BPS(c->opts.dst_format)) {
if (isALPHA(c->opts.src_format) && isALPHA(c->opts.dst_format)) {
if (isNBPS(sws->dst_format) || is16BPS(sws->dst_format)) {
if (isALPHA(sws->src_format) && isALPHA(sws->dst_format)) {
tmp_format = AV_PIX_FMT_BGRA64;
} else {
tmp_format = AV_PIX_FMT_BGR48;
}
} else {
if (isALPHA(c->opts.src_format) && isALPHA(c->opts.dst_format)) {
if (isALPHA(sws->src_format) && isALPHA(sws->dst_format)) {
tmp_format = AV_PIX_FMT_BGRA;
} else {
tmp_format = AV_PIX_FMT_BGR24;
@ -1145,14 +1146,13 @@ int sws_setColorspaceDetails(SwsContext *sws, const int inv_table[4],
if (ret < 0)
return ret;
c->cascaded_context[0] = alloc_set_opts(srcW, srcH, c->opts.src_format,
c->cascaded_context[0] = alloc_set_opts(srcW, srcH, sws->src_format,
tmp_width, tmp_height, tmp_format,
c->opts.flags,
c->opts.scaler_params);
sws->flags, sws->scaler_params);
if (!c->cascaded_context[0])
return -1;
sws_internal(c->cascaded_context[0])->opts.alpha_blend = c->opts.alpha_blend;
c->cascaded_context[0]->alpha_blend = sws->alpha_blend;
ret = sws_init_context(c->cascaded_context[0], NULL , NULL);
if (ret < 0)
return ret;
@ -1162,13 +1162,12 @@ int sws_setColorspaceDetails(SwsContext *sws, const int inv_table[4],
brightness, contrast, saturation);
c->cascaded_context[1] = alloc_set_opts(tmp_width, tmp_height, tmp_format,
dstW, dstH, c->opts.dst_format,
c->opts.flags,
c->opts.scaler_params);
dstW, dstH, sws->dst_format,
sws->flags, sws->scaler_params);
if (!c->cascaded_context[1])
return -1;
sws_internal(c->cascaded_context[1])->opts.src_range = srcRange;
sws_internal(c->cascaded_context[1])->opts.dst_range = dstRange;
c->cascaded_context[1]->src_range = srcRange;
c->cascaded_context[1]->dst_range = dstRange;
ret = sws_init_context(c->cascaded_context[1], NULL , NULL);
if (ret < 0)
return ret;
@ -1183,7 +1182,7 @@ int sws_setColorspaceDetails(SwsContext *sws, const int inv_table[4],
return 0;
}
if (!isYUV(c->opts.dst_format) && !isGray(c->opts.dst_format)) {
if (!isYUV(sws->dst_format) && !isGray(sws->dst_format)) {
ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
contrast, saturation);
// FIXME factorize
@ -1215,8 +1214,8 @@ int sws_getColorspaceDetails(SwsContext *sws, int **inv_table,
*inv_table = c->srcColorspaceTable;
*table = c->dstColorspaceTable;
*srcRange = range_override_needed(c->opts.src_format) ? 1 : c->opts.src_range;
*dstRange = range_override_needed(c->opts.dst_format) ? 1 : c->opts.dst_range;
*srcRange = range_override_needed(sws->src_format) ? 1 : sws->src_range;
*dstRange = range_override_needed(sws->dst_format) ? 1 : sws->dst_range;
*brightness = c->brightness;
*contrast = c->contrast;
*saturation = c->saturation;
@ -1321,10 +1320,10 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
int unscaled;
SwsInternal *c = sws_internal(sws);
SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
int srcW = c->opts.src_w;
int srcH = c->opts.src_h;
int dstW = c->opts.dst_w;
int dstH = c->opts.dst_h;
int srcW = sws->src_w;
int srcH = sws->src_h;
int dstW = sws->dst_w;
int dstH = sws->dst_h;
int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16);
int flags, cpu_flags;
enum AVPixelFormat srcFormat, dstFormat;
@ -1335,25 +1334,25 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
static const float float_mult = 1.0f / 255.0f;
cpu_flags = av_get_cpu_flags();
flags = c->opts.flags;
flags = sws->flags;
emms_c();
unscaled = (srcW == dstW && srcH == dstH);
if (!c->contrast && !c->saturation && !c->dstFormatBpp)
sws_setColorspaceDetails(sws, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->opts.src_range,
sws_setColorspaceDetails(sws, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], sws->src_range,
ff_yuv2rgb_coeffs[SWS_CS_DEFAULT],
c->opts.dst_range, 0, 1 << 16, 1 << 16);
sws->dst_range, 0, 1 << 16, 1 << 16);
handle_formats(c);
srcFormat = c->opts.src_format;
dstFormat = c->opts.dst_format;
handle_formats(sws);
srcFormat = sws->src_format;
dstFormat = sws->dst_format;
desc_src = av_pix_fmt_desc_get(srcFormat);
desc_dst = av_pix_fmt_desc_get(dstFormat);
// If the source has no alpha then disable alpha blendaway
if (c->src0Alpha)
c->opts.alpha_blend = SWS_ALPHA_BLEND_NONE;
sws->alpha_blend = SWS_ALPHA_BLEND_NONE;
if (!(unscaled && sws_isSupportedEndiannessConversion(srcFormat) &&
av_pix_fmt_swap_endianness(srcFormat) == dstFormat)) {
@ -1390,7 +1389,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
flags |= SWS_BICUBIC;
else
flags |= SWS_BICUBIC;
c->opts.flags = flags;
sws->flags = flags;
} else if (i & (i - 1)) {
av_log(c, AV_LOG_ERROR,
"Exactly one scaler algorithm must be chosen, got %X\n", i);
@ -1407,7 +1406,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
if (flags & SWS_FAST_BILINEAR) {
if (srcW < 8 || dstW < 8) {
flags ^= SWS_FAST_BILINEAR | SWS_BILINEAR;
c->opts.flags = flags;
sws->flags = flags;
}
}
@ -1440,46 +1439,46 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
if (dstW&1) {
av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
flags |= SWS_FULL_CHR_H_INT;
c->opts.flags = flags;
sws->flags = flags;
}
if ( c->chrSrcHSubSample == 0
&& c->chrSrcVSubSample == 0
&& c->opts.dither != SWS_DITHER_BAYER //SWS_FULL_CHR_H_INT is currently not supported with SWS_DITHER_BAYER
&& !(c->opts.flags & SWS_FAST_BILINEAR)
&& sws->dither != SWS_DITHER_BAYER //SWS_FULL_CHR_H_INT is currently not supported with SWS_DITHER_BAYER
&& !(sws->flags & SWS_FAST_BILINEAR)
) {
av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to input having non subsampled chroma\n");
flags |= SWS_FULL_CHR_H_INT;
c->opts.flags = flags;
sws->flags = flags;
}
}
if (c->opts.dither == SWS_DITHER_AUTO) {
if (sws->dither == SWS_DITHER_AUTO) {
if (flags & SWS_ERROR_DIFFUSION)
c->opts.dither = SWS_DITHER_ED;
sws->dither = SWS_DITHER_ED;
}
if(dstFormat == AV_PIX_FMT_BGR4_BYTE ||
dstFormat == AV_PIX_FMT_RGB4_BYTE ||
dstFormat == AV_PIX_FMT_BGR8 ||
dstFormat == AV_PIX_FMT_RGB8) {
if (c->opts.dither == SWS_DITHER_AUTO)
c->opts.dither = (flags & SWS_FULL_CHR_H_INT) ? SWS_DITHER_ED : SWS_DITHER_BAYER;
if (sws->dither == SWS_DITHER_AUTO)
sws->dither = (flags & SWS_FULL_CHR_H_INT) ? SWS_DITHER_ED : SWS_DITHER_BAYER;
if (!(flags & SWS_FULL_CHR_H_INT)) {
if (c->opts.dither == SWS_DITHER_ED || c->opts.dither == SWS_DITHER_A_DITHER || c->opts.dither == SWS_DITHER_X_DITHER || c->opts.dither == SWS_DITHER_NONE) {
if (sws->dither == SWS_DITHER_ED || sws->dither == SWS_DITHER_A_DITHER || sws->dither == SWS_DITHER_X_DITHER || sws->dither == SWS_DITHER_NONE) {
av_log(c, AV_LOG_DEBUG,
"Desired dithering only supported in full chroma interpolation for destination format '%s'\n",
av_get_pix_fmt_name(dstFormat));
flags |= SWS_FULL_CHR_H_INT;
c->opts.flags = flags;
sws->flags = flags;
}
}
if (flags & SWS_FULL_CHR_H_INT) {
if (c->opts.dither == SWS_DITHER_BAYER) {
if (sws->dither == SWS_DITHER_BAYER) {
av_log(c, AV_LOG_DEBUG,
"Ordered dither is not supported in full chroma interpolation for destination format '%s'\n",
av_get_pix_fmt_name(dstFormat));
c->opts.dither = SWS_DITHER_ED;
sws->dither = SWS_DITHER_ED;
}
}
}
@ -1489,7 +1488,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
"%s output is not supported with half chroma resolution, switching to full\n",
av_get_pix_fmt_name(dstFormat));
flags |= SWS_FULL_CHR_H_INT;
c->opts.flags = flags;
sws->flags = flags;
}
}
@ -1523,7 +1522,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
"full chroma interpolation for destination format '%s' not yet implemented\n",
av_get_pix_fmt_name(dstFormat));
flags &= ~SWS_FULL_CHR_H_INT;
c->opts.flags = flags;
sws->flags = flags;
}
if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
c->chrDstHSubSample = 1;
@ -1585,7 +1584,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
av_log(c, AV_LOG_INFO,
"output width is not a multiple of 32 -> no MMXEXT scaler\n");
}
if (usesHFilter || isNBPS(c->opts.src_format) || is16BPS(c->opts.src_format) || isAnyRGB(c->opts.src_format))
if (usesHFilter || isNBPS(sws->src_format) || is16BPS(sws->src_format) || isAnyRGB(sws->src_format))
c->canMMXEXTBeUsed = 0;
} else
c->canMMXEXTBeUsed = 0;
@ -1616,8 +1615,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
c->gamma_value = 2.2;
tmpFmt = AV_PIX_FMT_RGBA64LE;
if (!unscaled && c->opts.gamma_flag && (srcFormat != tmpFmt || dstFormat != tmpFmt)) {
if (!unscaled && sws->gamma_flag && (srcFormat != tmpFmt || dstFormat != tmpFmt)) {
SwsInternal *c2;
c->cascaded_context[0] = NULL;
@ -1629,7 +1627,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
srcW, srcH, tmpFmt,
flags, NULL, NULL,
c->opts.scaler_params);
sws->scaler_params);
if (!c->cascaded_context[0]) {
return AVERROR(ENOMEM);
}
@ -1637,7 +1635,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFmt,
dstW, dstH, tmpFmt,
flags, srcFilter, dstFilter,
c->opts.scaler_params);
sws->scaler_params);
if (!c->cascaded_context[1])
return AVERROR(ENOMEM);
@ -1669,7 +1667,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
c->cascaded_context[2] = sws_getContext(dstW, dstH, tmpFmt,
dstW, dstH, dstFormat,
flags, NULL, NULL,
c->opts.scaler_params);
sws->scaler_params);
if (!c->cascaded_context[2])
return AVERROR(ENOMEM);
}
@ -1690,14 +1688,14 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
srcW, srcH, tmpFormat,
flags, srcFilter, NULL,
c->opts.scaler_params);
sws->scaler_params);
if (!c->cascaded_context[0])
return AVERROR(ENOMEM);
c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFormat,
dstW, dstH, dstFormat,
flags, NULL, dstFilter,
c->opts.scaler_params);
sws->scaler_params);
if (!c->cascaded_context[1])
return AVERROR(ENOMEM);
return 0;
@ -1720,11 +1718,11 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
if (CONFIG_SWSCALE_ALPHA && isALPHA(srcFormat) && !isALPHA(dstFormat)) {
enum AVPixelFormat tmpFormat = alphaless_fmt(srcFormat);
if (tmpFormat != AV_PIX_FMT_NONE && c->opts.alpha_blend != SWS_ALPHA_BLEND_NONE) {
if (tmpFormat != AV_PIX_FMT_NONE && sws->alpha_blend != SWS_ALPHA_BLEND_NONE) {
if (!unscaled ||
dstFormat != tmpFormat ||
usesHFilter || usesVFilter ||
c->opts.src_range != c->opts.dst_range
sws->src_range != sws->dst_range
) {
c->cascaded_mainindex = 1;
ret = av_image_alloc(c->cascaded_tmp[0], c->cascaded_tmpStride[0],
@ -1734,22 +1732,22 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
c->cascaded_context[0] = alloc_set_opts(srcW, srcH, srcFormat,
srcW, srcH, tmpFormat,
flags, c->opts.scaler_params);
flags, sws->scaler_params);
if (!c->cascaded_context[0])
return AVERROR(EINVAL);
sws_internal(c->cascaded_context[0])->opts.alpha_blend = c->opts.alpha_blend;
c->cascaded_context[0]->alpha_blend = sws->alpha_blend;
ret = sws_init_context(c->cascaded_context[0], NULL , NULL);
if (ret < 0)
return ret;
c->cascaded_context[1] = alloc_set_opts(srcW, srcH, tmpFormat,
dstW, dstH, dstFormat,
flags, c->opts.scaler_params);
flags, sws->scaler_params);
if (!c->cascaded_context[1])
return AVERROR(EINVAL);
sws_internal(c->cascaded_context[1])->opts.src_range = c->opts.src_range;
sws_internal(c->cascaded_context[1])->opts.dst_range = c->opts.dst_range;
c->cascaded_context[1]->src_range = sws->src_range;
c->cascaded_context[1]->dst_range = sws->dst_range;
ret = sws_init_context(c->cascaded_context[1], srcFilter , dstFilter);
if (ret < 0)
return ret;
@ -1761,9 +1759,9 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
/* alpha blend special case, note this has been split via cascaded contexts if its scaled */
if (unscaled && !usesHFilter && !usesVFilter &&
c->opts.alpha_blend != SWS_ALPHA_BLEND_NONE &&
sws->alpha_blend != SWS_ALPHA_BLEND_NONE &&
isALPHA(srcFormat) &&
(c->opts.src_range == c->opts.dst_range || isAnyRGB(dstFormat)) &&
(sws->src_range == sws->dst_range || isAnyRGB(dstFormat)) &&
alphaless_fmt(srcFormat) == dstFormat
) {
c->convert_unscaled = ff_sws_alphablendaway;
@ -1777,7 +1775,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
/* unscaled special cases */
if (unscaled && !usesHFilter && !usesVFilter &&
(c->opts.src_range == c->opts.dst_range || isAnyRGB(dstFormat) ||
(sws->src_range == sws->dst_range || isAnyRGB(dstFormat) ||
isFloat(srcFormat) || isFloat(dstFormat) || isBayer(srcFormat))){
ff_get_unscaled_swscale(c);
@ -1873,7 +1871,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
srcW, dstW, filterAlign, 1 << 14,
(flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
cpu_flags, srcFilter->lumH, dstFilter->lumH,
c->opts.scaler_params,
sws->scaler_params,
get_local_pos(c, 0, 0, 0),
get_local_pos(c, 0, 0, 0))) < 0)
goto fail;
@ -1884,9 +1882,9 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
(flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
cpu_flags, srcFilter->chrH, dstFilter->chrH,
c->opts.scaler_params,
get_local_pos(c, c->chrSrcHSubSample, c->opts.src_h_chr_pos, 0),
get_local_pos(c, c->chrDstHSubSample, c->opts.dst_h_chr_pos, 0))) < 0)
sws->scaler_params,
get_local_pos(c, c->chrSrcHSubSample, sws->src_h_chr_pos, 0),
get_local_pos(c, c->chrDstHSubSample, sws->dst_h_chr_pos, 0))) < 0)
goto fail;
if (ff_shuffle_filter_coefficients(c, c->hChrFilterPos, c->hChrFilterSize, c->hChrFilter, c->chrDstW) < 0)
goto nomem;
@ -1903,7 +1901,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
(flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
cpu_flags, srcFilter->lumV, dstFilter->lumV,
c->opts.scaler_params,
sws->scaler_params,
get_local_pos(c, 0, 0, 1),
get_local_pos(c, 0, 0, 1))) < 0)
goto fail;
@ -1912,18 +1910,18 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
filterAlign, (1 << 12),
(flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
cpu_flags, srcFilter->chrV, dstFilter->chrV,
c->opts.scaler_params,
get_local_pos(c, c->chrSrcVSubSample, c->opts.src_v_chr_pos, 1),
get_local_pos(c, c->chrDstVSubSample, c->opts.dst_v_chr_pos, 1))) < 0)
sws->scaler_params,
get_local_pos(c, c->chrSrcVSubSample, sws->src_v_chr_pos, 1),
get_local_pos(c, c->chrDstVSubSample, sws->dst_v_chr_pos, 1))) < 0)
goto fail;
#if HAVE_ALTIVEC
if (!FF_ALLOC_TYPED_ARRAY(c->vYCoeffsBank, c->vLumFilterSize * c->opts.dst_h) ||
if (!FF_ALLOC_TYPED_ARRAY(c->vYCoeffsBank, c->vLumFilterSize * sws->dst_h) ||
!FF_ALLOC_TYPED_ARRAY(c->vCCoeffsBank, c->vChrFilterSize * c->chrDstH))
goto nomem;
for (i = 0; i < c->vLumFilterSize * c->opts.dst_h; i++) {
for (i = 0; i < c->vLumFilterSize * sws->dst_h; i++) {
int j;
short *p = (short *)&c->vYCoeffsBank[i];
for (j = 0; j < 8; j++)
@ -1940,10 +1938,10 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
}
for (i = 0; i < 4; i++)
if (!FF_ALLOCZ_TYPED_ARRAY(c->dither_error[i], c->opts.dst_w + 3))
if (!FF_ALLOCZ_TYPED_ARRAY(c->dither_error[i], sws->dst_w + 3))
goto nomem;
c->needAlpha = (CONFIG_SWSCALE_ALPHA && isALPHA(c->opts.src_format) && isALPHA(c->opts.dst_format)) ? 1 : 0;
c->needAlpha = (CONFIG_SWSCALE_ALPHA && isALPHA(sws->src_format) && isALPHA(sws->dst_format)) ? 1 : 0;
// 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
@ -1985,7 +1983,7 @@ static av_cold int sws_init_single_context(SwsContext *sws, SwsFilter *srcFilter
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",
c->opts.src_w, c->opts.src_h, c->opts.dst_w, c->opts.dst_h, c->lumXInc, c->lumYInc);
sws->src_w, sws->src_h, sws->dst_w, sws->dst_h, 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,
@ -2017,14 +2015,14 @@ fail: // FIXME replace things by appropriate error codes
c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
tmpW, tmpH, tmpFormat,
flags, srcFilter, NULL,
c->opts.scaler_params);
sws->scaler_params);
if (!c->cascaded_context[0])
return AVERROR(ENOMEM);
c->cascaded_context[1] = sws_getContext(tmpW, tmpH, tmpFormat,
dstW, dstH, dstFormat,
flags, NULL, dstFilter,
c->opts.scaler_params);
sws->scaler_params);
if (!c->cascaded_context[1])
return AVERROR(ENOMEM);
return 0;
@ -2039,41 +2037,38 @@ static int context_init_threaded(SwsContext *sws,
int ret;
ret = avpriv_slicethread_create(&c->slicethread, (void*) sws,
ff_sws_slice_worker, NULL, c->opts.threads);
ff_sws_slice_worker, NULL, sws->threads);
if (ret == AVERROR(ENOSYS)) {
c->opts.threads = 1;
sws->threads = 1;
return 0;
} else if (ret < 0)
return ret;
c->opts.threads = ret;
sws->threads = ret;
c->slice_ctx = av_calloc(c->opts.threads, sizeof(*c->slice_ctx));
c->slice_err = av_calloc(c->opts.threads, sizeof(*c->slice_err));
c->slice_ctx = av_calloc(sws->threads, sizeof(*c->slice_ctx));
c->slice_err = av_calloc(sws->threads, sizeof(*c->slice_err));
if (!c->slice_ctx || !c->slice_err)
return AVERROR(ENOMEM);
for (int i = 0; i < c->opts.threads; i++) {
SwsInternal *c2;
c->slice_ctx[i] = sws_alloc_context();
if (!c->slice_ctx[i])
for (int i = 0; i < sws->threads; i++) {
SwsContext *slice;
slice = c->slice_ctx[i] = sws_alloc_context();
if (!slice)
return AVERROR(ENOMEM);
c2 = sws_internal(c->slice_ctx[i]);
sws_internal(slice)->parent = sws;
c->nb_slice_ctx++;
c2->parent = sws;
ret = av_opt_copy((void*)c->slice_ctx[i], (void*)c);
ret = av_opt_copy(slice, sws);
if (ret < 0)
return ret;
slice->threads = 1;
c2->opts.threads = 1;
ret = sws_init_single_context(c->slice_ctx[i], src_filter, dst_filter);
ret = sws_init_single_context(slice, src_filter, dst_filter);
if (ret < 0)
return ret;
if (c2->opts.dither == SWS_DITHER_ED) {
if (slice->dither == SWS_DITHER_ED) {
av_log(c, AV_LOG_VERBOSE,
"Error-diffusion dither is in use, scaling will be single-threaded.");
break;
@ -2099,17 +2094,17 @@ av_cold int sws_init_context(SwsContext *sws, SwsFilter *srcFilter,
if (ff_thread_once(&rgb2rgb_once, ff_sws_rgb2rgb_init) != 0)
return AVERROR_UNKNOWN;
src_format = c->opts.src_format;
dst_format = c->opts.dst_format;
c->opts.src_range |= handle_jpeg(&c->opts.src_format);
c->opts.dst_range |= handle_jpeg(&c->opts.dst_format);
src_format = sws->src_format;
dst_format = sws->dst_format;
sws->src_range |= handle_jpeg(&sws->src_format);
sws->dst_range |= handle_jpeg(&sws->dst_format);
if (src_format != c->opts.src_format || dst_format != c->opts.dst_format)
if (src_format != sws->src_format || dst_format != sws->dst_format)
av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
if (c->opts.threads != 1) {
if (sws->threads != 1) {
ret = context_init_threaded(sws, srcFilter, dstFilter);
if (ret < 0 || c->opts.threads > 1)
if (ret < 0 || sws->threads > 1)
return ret;
// threading disabled in this build, init as single-threaded
}

@ -135,13 +135,13 @@ static void check_output_yuv2gbrp(void)
fail();
c = sws_internal(sws);
c->opts.flags |= SWS_FULL_CHR_H_INT;
sws->flags |= SWS_FULL_CHR_H_INT;
for (fmi = 0; fmi < FF_ARRAY_ELEMS(planar_fmts); fmi++) {
for (fsi = 0; fsi < FILTER_SIZES; fsi++) {
for (isi = 0; isi < FF_ARRAY_ELEMS(input_sizes); isi++ ) {
desc = av_pix_fmt_desc_get(planar_fmts[fmi]);
c->opts.dst_format = planar_fmts[fmi];
sws->dst_format = planar_fmts[fmi];
dstW = input_sizes[isi];
luma_filter_size = filter_sizes[fsi];
@ -229,8 +229,8 @@ static void check_input_planar_rgb_to_y(void)
for (fmi = 0; fmi < FF_ARRAY_ELEMS(planar_fmts); fmi++) {
for (isi = 0; isi < FF_ARRAY_ELEMS(input_sizes); isi++ ) {
desc = av_pix_fmt_desc_get(planar_fmts[fmi]);
c->opts.src_format = planar_fmts[fmi];
c->opts.dst_format = AV_PIX_FMT_YUVA444P16;
sws->src_format = planar_fmts[fmi];
sws->dst_format = AV_PIX_FMT_YUVA444P16;
byte_size = 2;
dstW = input_sizes[isi];
@ -300,8 +300,8 @@ static void check_input_planar_rgb_to_uv(void)
for (fmi = 0; fmi < FF_ARRAY_ELEMS(planar_fmts); fmi++) {
for (isi = 0; isi < FF_ARRAY_ELEMS(input_sizes); isi++ ) {
desc = av_pix_fmt_desc_get(planar_fmts[fmi]);
c->opts.src_format = planar_fmts[fmi];
c->opts.dst_format = AV_PIX_FMT_YUVA444P16;
sws->src_format = planar_fmts[fmi];
sws->dst_format = AV_PIX_FMT_YUVA444P16;
byte_size = 2;
dstW = input_sizes[isi];
@ -373,8 +373,8 @@ static void check_input_planar_rgb_to_a(void)
if (!(desc->flags & AV_PIX_FMT_FLAG_ALPHA))
continue;
c->opts.src_format = planar_fmts[fmi];
c->opts.dst_format = AV_PIX_FMT_YUVA444P16;
sws->src_format = planar_fmts[fmi];
sws->dst_format = AV_PIX_FMT_YUVA444P16;
byte_size = 2;
dstW = input_sizes[isi];

@ -73,16 +73,16 @@ static void check_lumConvertRange(int from)
fail();
c = sws_internal(sws);
c->opts.src_range = from;
c->opts.dst_range = !from;
sws->src_range = from;
sws->dst_range = !from;
for (int pfi = 0; pfi < FF_ARRAY_ELEMS(pixel_formats); pfi++) {
enum AVPixelFormat pix_fmt = pixel_formats[pfi];
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
int bit_depth = desc->comp[0].depth;
int sample_size = bit_depth == 16 ? sizeof(int32_t) : sizeof(int16_t);
c->opts.src_format = pix_fmt;
c->opts.dst_format = pix_fmt;
sws->src_format = pix_fmt;
sws->dst_format = pix_fmt;
c->dstBpc = bit_depth;
ff_sws_init_scale(c);
for (int dstWi = 0; dstWi < FF_ARRAY_ELEMS(input_sizes); dstWi++) {
@ -123,16 +123,16 @@ static void check_chrConvertRange(int from)
fail();
c = sws_internal(sws);
c->opts.src_range = from;
c->opts.dst_range = !from;
sws->src_range = from;
sws->dst_range = !from;
for (int pfi = 0; pfi < FF_ARRAY_ELEMS(pixel_formats); pfi++) {
enum AVPixelFormat pix_fmt = pixel_formats[pfi];
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
int bit_depth = desc->comp[0].depth;
int sample_size = bit_depth == 16 ? sizeof(int32_t) : sizeof(int16_t);
c->opts.src_format = pix_fmt;
c->opts.dst_format = pix_fmt;
sws->src_format = pix_fmt;
sws->dst_format = pix_fmt;
c->dstBpc = bit_depth;
ff_sws_init_scale(c);
for (int dstWi = 0; dstWi < FF_ARRAY_ELEMS(input_sizes); dstWi++) {

@ -127,9 +127,10 @@ static int cmp_off_by_n(const uint8_t *ref, const uint8_t *test, size_t n, int a
return 0;
}
static void check_rgb24toyv12(SwsInternal *ctx)
static void check_rgb24toyv12(SwsContext *sws)
{
static const int input_sizes[] = {16, 128, 512, MAX_LINE_SIZE, -MAX_LINE_SIZE};
SwsInternal *ctx = sws_internal(sws);
LOCAL_ALIGNED_32(uint8_t, src, [BUFSIZE * 3]);
LOCAL_ALIGNED_32(uint8_t, buf_y_0, [BUFSIZE]);
@ -353,8 +354,10 @@ static const enum AVPixelFormat rgb_formats[] = {
AV_PIX_FMT_ARGB,
};
static void check_rgb_to_y(SwsInternal *ctx)
static void check_rgb_to_y(SwsContext *sws)
{
SwsInternal *ctx = sws_internal(sws);
LOCAL_ALIGNED_16(uint8_t, src24, [MAX_LINE_SIZE * 3]);
LOCAL_ALIGNED_16(uint8_t, src32, [MAX_LINE_SIZE * 4]);
LOCAL_ALIGNED_32(uint8_t, dst0_y, [MAX_LINE_SIZE * 2]);
@ -370,7 +373,7 @@ static void check_rgb_to_y(SwsInternal *ctx)
for (int i = 0; i < FF_ARRAY_ELEMS(rgb_formats); i++) {
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(rgb_formats[i]);
ctx->opts.src_format = rgb_formats[i];
sws->src_format = rgb_formats[i];
ff_sws_init_scale(ctx);
for (int j = 0; j < FF_ARRAY_ELEMS(input_sizes); j++) {
@ -389,15 +392,17 @@ static void check_rgb_to_y(SwsInternal *ctx)
if (desc->nb_components == 3 ||
// only bench native endian formats
(ctx->opts.src_format == AV_PIX_FMT_RGB32 || ctx->opts.src_format == AV_PIX_FMT_RGB32_1))
(sws->src_format == AV_PIX_FMT_RGB32 || sws->src_format == AV_PIX_FMT_RGB32_1))
bench_new(dst1_y, src, NULL, NULL, w, ctx->input_rgb2yuv_table, NULL);
}
}
}
}
static void check_rgb_to_uv(SwsInternal *ctx)
static void check_rgb_to_uv(SwsContext *sws)
{
SwsInternal *ctx = sws_internal(sws);
LOCAL_ALIGNED_16(uint8_t, src24, [MAX_LINE_SIZE * 3]);
LOCAL_ALIGNED_16(uint8_t, src32, [MAX_LINE_SIZE * 4]);
LOCAL_ALIGNED_16(uint8_t, dst0_u, [MAX_LINE_SIZE * 2]);
@ -417,8 +422,8 @@ static void check_rgb_to_uv(SwsInternal *ctx)
const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(src_fmt);
ctx->chrSrcHSubSample = (i % 2) ? 0 : 1;
ctx->opts.src_format = src_fmt;
ctx->opts.dst_format = ctx->chrSrcHSubSample ? AV_PIX_FMT_YUV420P : AV_PIX_FMT_YUV444P;
sws->src_format = src_fmt;
sws->dst_format = ctx->chrSrcHSubSample ? AV_PIX_FMT_YUV420P : AV_PIX_FMT_YUV444P;
ff_sws_init_scale(ctx);
for (int j = 0; j < FF_ARRAY_ELEMS(input_sizes); j++) {
@ -441,7 +446,7 @@ static void check_rgb_to_uv(SwsInternal *ctx)
if (desc->nb_components == 3 ||
// only bench native endian formats
(ctx->opts.src_format == AV_PIX_FMT_RGB32 || ctx->opts.src_format == AV_PIX_FMT_RGB32_1))
(sws->src_format == AV_PIX_FMT_RGB32 || sws->src_format == AV_PIX_FMT_RGB32_1))
bench_new(dst1_u, dst1_v, NULL, src, src, w, ctx->input_rgb2yuv_table, NULL);
}
}
@ -451,7 +456,6 @@ static void check_rgb_to_uv(SwsInternal *ctx)
void checkasm_check_sw_rgb(void)
{
SwsContext *sws;
SwsInternal *c;
ff_sws_rgb2rgb_init();
@ -485,14 +489,13 @@ void checkasm_check_sw_rgb(void)
if (!sws)
fail();
c = sws_internal(sws);
check_rgb_to_y(c);
check_rgb_to_y(sws);
report("rgb_to_y");
check_rgb_to_uv(c);
check_rgb_to_uv(sws);
report("rgb_to_uv");
check_rgb24toyv12(c);
check_rgb24toyv12(sws);
report("rgb24toyv12");
sws_freeContext(sws);

@ -123,12 +123,12 @@ static void check_yuv2yuv1(int accurate)
randomize_buffers((uint8_t*)dither, 8);
randomize_buffers((uint8_t*)src_pixels, LARGEST_INPUT_SIZE * sizeof(int16_t));
sws = sws_alloc_context();
c = sws_internal(sws);
if (accurate)
c->opts.flags |= SWS_ACCURATE_RND;
sws->flags |= SWS_ACCURATE_RND;
if (sws_init_context(sws, NULL, NULL) < 0)
fail();
c = sws_internal(sws);
ff_sws_init_scale(c);
for (isi = 0; isi < FF_ARRAY_ELEMS(input_sizes); ++isi) {
dstW = input_sizes[isi];
@ -190,12 +190,12 @@ static void check_yuv2yuvX(int accurate)
memset(dither, d_val, LARGEST_INPUT_SIZE);
randomize_buffers((uint8_t*)src_pixels, LARGEST_FILTER * LARGEST_INPUT_SIZE * sizeof(int16_t));
sws = sws_alloc_context();
c = sws_internal(sws);
if (accurate)
c->opts.flags |= SWS_ACCURATE_RND;
sws->flags |= SWS_ACCURATE_RND;
if (sws_init_context(sws, NULL, NULL) < 0)
fail();
c = sws_internal(sws);
ff_sws_init_scale(c);
for(isi = 0; isi < FF_ARRAY_ELEMS(input_sizes); ++isi){
dstW = input_sizes[isi];
@ -341,20 +341,20 @@ static void check_hscale(void)
filter[SRC_PIXELS * width + i] = rnd();
}
c->opts.dst_w = c->chrDstW = input_sizes[dstWi];
sws->dst_w = c->chrDstW = input_sizes[dstWi];
ff_sws_init_scale(c);
memcpy(filterAvx2, filter, sizeof(uint16_t) * (SRC_PIXELS * MAX_FILTER_WIDTH + MAX_FILTER_WIDTH));
ff_shuffle_filter_coefficients(c, filterPosAvx, width, filterAvx2, c->opts.dst_w);
ff_shuffle_filter_coefficients(c, filterPosAvx, width, filterAvx2, sws->dst_w);
if (check_func(c->hcScale, "hscale_%d_to_%d__fs_%d_dstW_%d", c->srcBpc, c->dstBpc + 1, width, c->opts.dst_w)) {
if (check_func(c->hcScale, "hscale_%d_to_%d__fs_%d_dstW_%d", c->srcBpc, c->dstBpc + 1, width, sws->dst_w)) {
memset(dst0, 0, SRC_PIXELS * sizeof(dst0[0]));
memset(dst1, 0, SRC_PIXELS * sizeof(dst1[0]));
call_ref(NULL, dst0, c->opts.dst_w, src, filter, filterPos, width);
call_new(NULL, dst1, c->opts.dst_w, src, filterAvx2, filterPosAvx, width);
if (memcmp(dst0, dst1, c->opts.dst_w * sizeof(dst0[0])))
call_ref(NULL, dst0, sws->dst_w, src, filter, filterPos, width);
call_new(NULL, dst1, sws->dst_w, src, filterAvx2, filterPosAvx, width);
if (memcmp(dst0, dst1, sws->dst_w * sizeof(dst0[0])))
fail();
bench_new(NULL, dst0, c->opts.dst_w, src, filter, filterPosAvx, width);
bench_new(NULL, dst0, sws->dst_w, src, filter, filterPosAvx, width);
}
}
}

Loading…
Cancel
Save