RepoMirrors/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2025-04-11 03:32:39 +00:00
Code Issues Projects Releases Wiki Activity

avfilter/vf_super2xsai: add slice threading support

Browse Source
This commit is contained in:
Paul B Mahol 2021-01-21 13:26:17 +01:00
parent c2ee42e032
commit f07a8b13f3
1 changed files with 38 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

60
libavfilter/vf_super2xsai.c
View File

@ -46,6 +46,10 @@ typedef struct Super2xSaIContext {
int is_be; int is_be;
} Super2xSaIContext; } Super2xSaIContext;
typedef struct ThreadData {
AVFrame *in, *out;
} ThreadData;
#define GET_RESULT(A, B, C, D) ((A != C || A != D) - (B != C || B != D)) #define GET_RESULT(A, B, C, D) ((A != C || A != D) - (B != C || B != D))
#define INTERPOLATE(A, B) (((A & hi_pixel_mask) >> 1) + ((B & hi_pixel_mask) >> 1) + (A & B & lo_pixel_mask)) #define INTERPOLATE(A, B) (((A & hi_pixel_mask) >> 1) + ((B & hi_pixel_mask) >> 1) + (A & B & lo_pixel_mask))
@ -53,12 +57,18 @@ typedef struct Super2xSaIContext {
#define Q_INTERPOLATE(A, B, C, D) ((A & q_hi_pixel_mask) >> 2) + ((B & q_hi_pixel_mask) >> 2) + ((C & q_hi_pixel_mask) >> 2) + ((D & q_hi_pixel_mask) >> 2) \ #define Q_INTERPOLATE(A, B, C, D) ((A & q_hi_pixel_mask) >> 2) + ((B & q_hi_pixel_mask) >> 2) + ((C & q_hi_pixel_mask) >> 2) + ((D & q_hi_pixel_mask) >> 2) \
+ ((((A & q_lo_pixel_mask) + (B & q_lo_pixel_mask) + (C & q_lo_pixel_mask) + (D & q_lo_pixel_mask)) >> 2) & q_lo_pixel_mask) + ((((A & q_lo_pixel_mask) + (B & q_lo_pixel_mask) + (C & q_lo_pixel_mask) + (D & q_lo_pixel_mask)) >> 2) & q_lo_pixel_mask)
static void super2xsai(AVFilterContext *ctx, static int super2xsai(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
const uint8_t *src, int src_linesize,
uint8_t *dst, int dst_linesize,
int width, int height)
{ {
Super2xSaIContext *s = ctx->priv; Super2xSaIContext *s = ctx->priv;
ThreadData *td = arg;
AVFrame *in = td->in;
AVFrame *out = td->out;
const uint8_t *src = in->data[0];
uint8_t *dst = out->data[0];
const int src_linesize = in->linesize[0];
const int dst_linesize = out->linesize[0];
const int width = in->width;
const int height = in->height;
unsigned int x, y; unsigned int x, y;
uint32_t color[4][4]; uint32_t color[4][4];
const uint8_t *src_line[4]; const uint8_t *src_line[4];
@ -67,18 +77,20 @@ static void super2xsai(AVFilterContext *ctx,
const uint32_t lo_pixel_mask = s->lo_pixel_mask; const uint32_t lo_pixel_mask = s->lo_pixel_mask;
const uint32_t q_hi_pixel_mask = s->q_hi_pixel_mask; const uint32_t q_hi_pixel_mask = s->q_hi_pixel_mask;
const uint32_t q_lo_pixel_mask = s->q_lo_pixel_mask; const uint32_t q_lo_pixel_mask = s->q_lo_pixel_mask;
const int slice_start = (height * jobnr) / nb_jobs;
const int slice_end = (height * (jobnr+1)) / nb_jobs;
/* Point to the first 4 lines, first line is duplicated */ /* Point to the first 4 lines, first line is duplicated */
src_line[0] = src; src_line[0] = src + src_linesize*FFMAX(slice_start - 1, 0);
src_line[1] = src; src_line[1] = src + src_linesize*slice_start;
src_line[2] = src + src_linesize*FFMIN(1, height-1); src_line[2] = src + src_linesize*FFMIN(slice_start + 1, height-1);
src_line[3] = src + src_linesize*FFMIN(2, height-1); src_line[3] = src + src_linesize*FFMIN(slice_start + 2, height-1);
#define READ_COLOR4(dst, src_line, off) dst = *((const uint32_t *)src_line + off) #define READ_COLOR4(dst, src_line, off) dst = *((const uint32_t *)src_line + off)
#define READ_COLOR3(dst, src_line, off) dst = AV_RL24 (src_line + 3*off) #define READ_COLOR3(dst, src_line, off) dst = AV_RL24 (src_line + 3*off)
#define READ_COLOR2(dst, src_line, off) dst = s->is_be ? AV_RB16(src_line + 2 * off) : AV_RL16(src_line + 2 * off) #define READ_COLOR2(dst, src_line, off) dst = s->is_be ? AV_RB16(src_line + 2 * off) : AV_RL16(src_line + 2 * off)
for (y = 0; y < height; y++) { for (y = slice_start; y < slice_end; y++) {
uint8_t *dst_line[2]; uint8_t *dst_line[2];
dst_line[0] = dst + dst_linesize*2*y; dst_line[0] = dst + dst_linesize*2*y;
@ -229,6 +241,8 @@ static void super2xsai(AVFilterContext *ctx,
if (y < height - 3) if (y < height - 3)
src_line[3] += src_linesize; src_line[3] += src_linesize;
} // y loop } // y loop
return 0;
} }
static int query_formats(AVFilterContext *ctx) static int query_formats(AVFilterContext *ctx)
@ -305,24 +319,25 @@ static int config_output(AVFilterLink *outlink)
return 0; return 0;
} }
static int filter_frame(AVFilterLink *inlink, AVFrame *inpicref) static int filter_frame(AVFilterLink *inlink, AVFrame *in)
{ {
AVFilterLink *outlink = inlink->dst->outputs[0]; AVFilterContext *ctx = inlink->dst;
AVFrame *outpicref = ff_get_video_buffer(outlink, outlink->w, outlink->h); AVFilterLink *outlink = ctx->outputs[0];
if (!outpicref) { ThreadData td;
av_frame_free(&inpicref); AVFrame *out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
if (!out) {
av_frame_free(&in);
return AVERROR(ENOMEM); return AVERROR(ENOMEM);
} }
av_frame_copy_props(outpicref, inpicref); av_frame_copy_props(out, in);
outpicref->width = outlink->w; out->width = outlink->w;
outpicref->height = outlink->h; out->height = outlink->h;
super2xsai(inlink->dst, inpicref->data[0], inpicref->linesize[0], td.in = in, td.out = out;
outpicref->data[0], outpicref->linesize[0], ctx->internal->execute(ctx, super2xsai, &td, NULL, FFMIN(in->height, ff_filter_get_nb_threads(ctx)));
inlink->w, inlink->h);
av_frame_free(&inpicref); av_frame_free(&in);
return ff_filter_frame(outlink, outpicref); return ff_filter_frame(outlink, out);
} }
static const AVFilterPad super2xsai_inputs[] = { static const AVFilterPad super2xsai_inputs[] = {
@ -351,4 +366,5 @@ AVFilter ff_vf_super2xsai = {
.query_formats = query_formats, .query_formats = query_formats,
.inputs = super2xsai_inputs, .inputs = super2xsai_inputs,
.outputs = super2xsai_outputs, .outputs = super2xsai_outputs,
.flags = AVFILTER_FLAG_SLICE_THREADS,
}; };