[FFmpeg-cvslog] avfilter/vf_displace: add slice threading support
Paul B Mahol
git at videolan.org
Sun Jun 18 12:12:41 EEST 2023
ffmpeg | branch: master | Paul B Mahol <onemda at gmail.com> | Sun Jun 18 11:01:56 2023 +0200| [54dc06e09e1d06f9184ffb4327d8e75b5651d431] | committer: Paul B Mahol
avfilter/vf_displace: add slice threading support
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=54dc06e09e1d06f9184ffb4327d8e75b5651d431
---
libavfilter/vf_displace.c | 109 ++++++++++++++++++++++++++++------------------
1 file changed, 66 insertions(+), 43 deletions(-)
diff --git a/libavfilter/vf_displace.c b/libavfilter/vf_displace.c
index 11909949b1..4df98d4cc1 100644
--- a/libavfilter/vf_displace.c
+++ b/libavfilter/vf_displace.c
@@ -45,8 +45,7 @@ typedef struct DisplaceContext {
uint8_t blank[4];
FFFrameSync fs;
- void (*displace)(struct DisplaceContext *s, const AVFrame *in,
- const AVFrame *xpic, const AVFrame *ypic, AVFrame *out);
+ int (*displace_slice)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs);
} DisplaceContext;
#define OFFSET(x) offsetof(DisplaceContext, x)
@@ -76,29 +75,38 @@ static const enum AVPixelFormat pix_fmts[] = {
AV_PIX_FMT_GRAY8, AV_PIX_FMT_NONE
};
-static void displace_planar(DisplaceContext *s, const AVFrame *in,
- const AVFrame *xpic, const AVFrame *ypic,
- AVFrame *out)
+typedef struct ThreadData {
+ AVFrame *in, *xin, *yin, *out;
+} ThreadData;
+
+static int displace_planar(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
{
- int plane, x, y;
+ DisplaceContext *s = ctx->priv;
+ const ThreadData *td = arg;
+ const AVFrame *in = td->in;
+ const AVFrame *xin = td->xin;
+ const AVFrame *yin = td->yin;
+ const AVFrame *out = td->out;
- for (plane = 0; plane < s->nb_planes; plane++) {
+ for (int plane = 0; plane < s->nb_planes; plane++) {
const int h = s->height[plane];
const int w = s->width[plane];
+ const int slice_start = (h * jobnr ) / nb_jobs;
+ const int slice_end = (h * (jobnr+1)) / nb_jobs;
const int dlinesize = out->linesize[plane];
const int slinesize = in->linesize[plane];
- const int xlinesize = xpic->linesize[plane];
- const int ylinesize = ypic->linesize[plane];
+ const int xlinesize = xin->linesize[plane];
+ const int ylinesize = yin->linesize[plane];
const uint8_t *src = in->data[plane];
- const uint8_t *ysrc = ypic->data[plane];
- const uint8_t *xsrc = xpic->data[plane];
- uint8_t *dst = out->data[plane];
+ const uint8_t *ysrc = yin->data[plane] + slice_start * ylinesize;
+ const uint8_t *xsrc = xin->data[plane] + slice_start * xlinesize;
+ uint8_t *dst = out->data[plane] + slice_start * dlinesize;
const uint8_t blank = s->blank[plane];
- for (y = 0; y < h; y++) {
+ for (int y = slice_start; y < slice_end; y++) {
switch (s->edge) {
case EDGE_BLANK:
- for (x = 0; x < w; x++) {
+ for (int x = 0; x < w; x++) {
int Y = y + ysrc[x] - 128;
int X = x + xsrc[x] - 128;
@@ -109,14 +117,14 @@ static void displace_planar(DisplaceContext *s, const AVFrame *in,
}
break;
case EDGE_SMEAR:
- for (x = 0; x < w; x++) {
+ for (int x = 0; x < w; x++) {
int Y = av_clip(y + ysrc[x] - 128, 0, h - 1);
int X = av_clip(x + xsrc[x] - 128, 0, w - 1);
dst[x] = src[Y * slinesize + X];
}
break;
case EDGE_WRAP:
- for (x = 0; x < w; x++) {
+ for (int x = 0; x < w; x++) {
int Y = (y + ysrc[x] - 128) % h;
int X = (x + xsrc[x] - 128) % w;
@@ -128,7 +136,7 @@ static void displace_planar(DisplaceContext *s, const AVFrame *in,
}
break;
case EDGE_MIRROR:
- for (x = 0; x < w; x++) {
+ for (int x = 0; x < w; x++) {
int Y = y + ysrc[x] - 128;
int X = x + xsrc[x] - 128;
@@ -150,31 +158,37 @@ static void displace_planar(DisplaceContext *s, const AVFrame *in,
dst += dlinesize;
}
}
+ return 0;
}
-static void displace_packed(DisplaceContext *s, const AVFrame *in,
- const AVFrame *xpic, const AVFrame *ypic,
- AVFrame *out)
+static int displace_packed(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
{
+ DisplaceContext *s = ctx->priv;
+ const ThreadData *td = arg;
+ const AVFrame *in = td->in;
+ const AVFrame *xin = td->xin;
+ const AVFrame *yin = td->yin;
+ const AVFrame *out = td->out;
const int step = s->step;
const int h = s->height[0];
const int w = s->width[0];
+ const int slice_start = (h * jobnr ) / nb_jobs;
+ const int slice_end = (h * (jobnr+1)) / nb_jobs;
const int dlinesize = out->linesize[0];
const int slinesize = in->linesize[0];
- const int xlinesize = xpic->linesize[0];
- const int ylinesize = ypic->linesize[0];
+ const int xlinesize = xin->linesize[0];
+ const int ylinesize = yin->linesize[0];
const uint8_t *src = in->data[0];
- const uint8_t *ysrc = ypic->data[0];
- const uint8_t *xsrc = xpic->data[0];
+ const uint8_t *ysrc = yin->data[0] + slice_start * ylinesize;
+ const uint8_t *xsrc = xin->data[0] + slice_start * xlinesize;
+ uint8_t *dst = out->data[0] + slice_start * dlinesize;
const uint8_t *blank = s->blank;
- uint8_t *dst = out->data[0];
- int c, x, y;
- for (y = 0; y < h; y++) {
+ for (int y = slice_start; y < slice_end; y++) {
switch (s->edge) {
case EDGE_BLANK:
- for (x = 0; x < w; x++) {
- for (c = 0; c < s->nb_components; c++) {
+ for (int x = 0; x < w; x++) {
+ for (int c = 0; c < s->nb_components; c++) {
int Y = y + (ysrc[x * step + c] - 128);
int X = x + (xsrc[x * step + c] - 128);
@@ -186,8 +200,8 @@ static void displace_packed(DisplaceContext *s, const AVFrame *in,
}
break;
case EDGE_SMEAR:
- for (x = 0; x < w; x++) {
- for (c = 0; c < s->nb_components; c++) {
+ for (int x = 0; x < w; x++) {
+ for (int c = 0; c < s->nb_components; c++) {
int Y = av_clip(y + (ysrc[x * step + c] - 128), 0, h - 1);
int X = av_clip(x + (xsrc[x * step + c] - 128), 0, w - 1);
@@ -196,8 +210,8 @@ static void displace_packed(DisplaceContext *s, const AVFrame *in,
}
break;
case EDGE_WRAP:
- for (x = 0; x < w; x++) {
- for (c = 0; c < s->nb_components; c++) {
+ for (int x = 0; x < w; x++) {
+ for (int c = 0; c < s->nb_components; c++) {
int Y = (y + (ysrc[x * step + c] - 128)) % h;
int X = (x + (xsrc[x * step + c] - 128)) % w;
@@ -210,8 +224,8 @@ static void displace_packed(DisplaceContext *s, const AVFrame *in,
}
break;
case EDGE_MIRROR:
- for (x = 0; x < w; x++) {
- for (c = 0; c < s->nb_components; c++) {
+ for (int x = 0; x < w; x++) {
+ for (int c = 0; c < s->nb_components; c++) {
int Y = y + ysrc[x * step + c] - 128;
int X = x + xsrc[x * step + c] - 128;
@@ -233,6 +247,7 @@ static void displace_packed(DisplaceContext *s, const AVFrame *in,
xsrc += xlinesize;
dst += dlinesize;
}
+ return 0;
}
static int process_frame(FFFrameSync *fs)
@@ -240,12 +255,12 @@ static int process_frame(FFFrameSync *fs)
AVFilterContext *ctx = fs->parent;
DisplaceContext *s = fs->opaque;
AVFilterLink *outlink = ctx->outputs[0];
- AVFrame *out, *in, *xpic, *ypic;
+ AVFrame *out, *in, *xin, *yin;
int ret;
- if ((ret = ff_framesync_get_frame(&s->fs, 0, &in, 0)) < 0 ||
- (ret = ff_framesync_get_frame(&s->fs, 1, &xpic, 0)) < 0 ||
- (ret = ff_framesync_get_frame(&s->fs, 2, &ypic, 0)) < 0)
+ if ((ret = ff_framesync_get_frame(&s->fs, 0, &in, 0)) < 0 ||
+ (ret = ff_framesync_get_frame(&s->fs, 1, &xin, 0)) < 0 ||
+ (ret = ff_framesync_get_frame(&s->fs, 2, &yin, 0)) < 0)
return ret;
if (ctx->is_disabled) {
@@ -253,12 +268,19 @@ static int process_frame(FFFrameSync *fs)
if (!out)
return AVERROR(ENOMEM);
} else {
+ ThreadData td;
+
out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
if (!out)
return AVERROR(ENOMEM);
av_frame_copy_props(out, in);
- s->displace(s, in, xpic, ypic, out);
+ td.in = in;
+ td.xin = xin;
+ td.yin = yin;
+ td.out = out;
+ ff_filter_execute(ctx, s->displace_slice, &td, NULL,
+ FFMIN(outlink->h, ff_filter_get_nb_threads(ctx)));
}
out->pts = av_rescale_q(s->fs.pts, s->fs.time_base, outlink->time_base);
@@ -276,9 +298,9 @@ static int config_input(AVFilterLink *inlink)
s->nb_components = desc->nb_components;
if (s->nb_planes > 1 || s->nb_components == 1)
- s->displace = displace_planar;
+ s->displace_slice = displace_planar;
else
- s->displace = displace_packed;
+ s->displace_slice = displace_packed;
if (!(desc->flags & AV_PIX_FMT_FLAG_RGB)) {
s->blank[1] = s->blank[2] = 128;
@@ -398,5 +420,6 @@ const AVFilter ff_vf_displace = {
FILTER_OUTPUTS(displace_outputs),
FILTER_PIXFMTS_ARRAY(pix_fmts),
.priv_class = &displace_class,
- .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
+ .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL |
+ AVFILTER_FLAG_SLICE_THREADS,
};
More information about the ffmpeg-cvslog
mailing list