2015-07-24 09:09:16 +02:00
|
|
|
/*
|
|
|
|
* Copyright (C) 2012 Mark Himsley
|
|
|
|
*
|
|
|
|
* get_scene_score() Copyright (c) 2011 Stefano Sabatini
|
|
|
|
* taken from libavfilter/vf_select.c
|
|
|
|
*
|
|
|
|
* This file is part of FFmpeg.
|
|
|
|
*
|
|
|
|
* FFmpeg is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
|
|
* License as published by the Free Software Foundation; either
|
|
|
|
* version 2.1 of the License, or (at your option) any later version.
|
|
|
|
*
|
|
|
|
* FFmpeg is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* Lesser General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
|
|
* License along with FFmpeg; if not, write to the Free Software
|
|
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
|
|
*/
|
|
|
|
|
|
|
|
/**
|
|
|
|
* @file
|
|
|
|
* filter for upsampling or downsampling a progressive source
|
|
|
|
*/
|
|
|
|
|
|
|
|
#define DEBUG
|
|
|
|
|
|
|
|
#include "libavutil/avassert.h"
|
|
|
|
#include "libavutil/imgutils.h"
|
|
|
|
#include "libavutil/internal.h"
|
|
|
|
#include "libavutil/opt.h"
|
|
|
|
#include "libavutil/pixdesc.h"
|
|
|
|
#include "libavutil/pixelutils.h"
|
|
|
|
|
|
|
|
#include "avfilter.h"
|
|
|
|
#include "internal.h"
|
|
|
|
#include "video.h"
|
avfilter/vf_framerate: add SIMD functions for frame blending
Blend function speedups on x86_64 Core i5 4460:
ffmpeg -f lavfi -i allyuv -vf framerate=60:threads=1 -f null none
C: 447548411 decicycles in Blend, 2048 runs, 0 skips
SSSE3: 130020087 decicycles in Blend, 2048 runs, 0 skips
AVX2: 128508221 decicycles in Blend, 2048 runs, 0 skips
ffmpeg -f lavfi -i allyuv -vf format=yuv420p12,framerate=60:threads=1 -f null none
C: 228932745 decicycles in Blend, 2048 runs, 0 skips
SSE4: 123357781 decicycles in Blend, 2048 runs, 0 skips
AVX2: 121215353 decicycles in Blend, 2048 runs, 0 skips
Signed-off-by: Marton Balint <cus@passwd.hu>
2018-01-08 02:05:45 +02:00
|
|
|
#include "framerate.h"
|
2015-07-24 09:09:16 +02:00
|
|
|
|
|
|
|
#define OFFSET(x) offsetof(FrameRateContext, x)
|
|
|
|
#define V AV_OPT_FLAG_VIDEO_PARAM
|
|
|
|
#define F AV_OPT_FLAG_FILTERING_PARAM
|
|
|
|
#define FRAMERATE_FLAG_SCD 01
|
|
|
|
|
|
|
|
static const AVOption framerate_options[] = {
|
|
|
|
{"fps", "required output frames per second rate", OFFSET(dest_frame_rate), AV_OPT_TYPE_VIDEO_RATE, {.str="50"}, 0, INT_MAX, V|F },
|
|
|
|
|
2018-01-15 23:19:46 +02:00
|
|
|
{"interp_start", "point to start linear interpolation", OFFSET(interp_start), AV_OPT_TYPE_INT, {.i64=15}, 0, 255, V|F },
|
|
|
|
{"interp_end", "point to end linear interpolation", OFFSET(interp_end), AV_OPT_TYPE_INT, {.i64=240}, 0, 255, V|F },
|
2017-12-10 01:29:12 +02:00
|
|
|
{"scene", "scene change level", OFFSET(scene_score), AV_OPT_TYPE_DOUBLE, {.dbl=8.2}, 0, INT_MAX, V|F },
|
2015-07-24 09:09:16 +02:00
|
|
|
|
|
|
|
{"flags", "set flags", OFFSET(flags), AV_OPT_TYPE_FLAGS, {.i64=1}, 0, INT_MAX, V|F, "flags" },
|
|
|
|
{"scene_change_detect", "enable scene change detection", 0, AV_OPT_TYPE_CONST, {.i64=FRAMERATE_FLAG_SCD}, INT_MIN, INT_MAX, V|F, "flags" },
|
|
|
|
{"scd", "enable scene change detection", 0, AV_OPT_TYPE_CONST, {.i64=FRAMERATE_FLAG_SCD}, INT_MIN, INT_MAX, V|F, "flags" },
|
|
|
|
|
|
|
|
{NULL}
|
|
|
|
};
|
|
|
|
|
|
|
|
AVFILTER_DEFINE_CLASS(framerate);
|
|
|
|
|
2015-09-07 20:30:22 +02:00
|
|
|
static av_always_inline int64_t sad_8x8_16(const uint16_t *src1, ptrdiff_t stride1,
|
|
|
|
const uint16_t *src2, ptrdiff_t stride2)
|
|
|
|
{
|
|
|
|
int sum = 0;
|
|
|
|
int x, y;
|
|
|
|
|
|
|
|
for (y = 0; y < 8; y++) {
|
|
|
|
for (x = 0; x < 8; x++)
|
|
|
|
sum += FFABS(src1[x] - src2[x]);
|
|
|
|
src1 += stride1;
|
|
|
|
src2 += stride2;
|
|
|
|
}
|
|
|
|
return sum;
|
|
|
|
}
|
|
|
|
|
2017-12-10 20:01:03 +02:00
|
|
|
static int64_t scene_sad16(FrameRateContext *s, const uint16_t *p1, int p1_linesize, const uint16_t* p2, int p2_linesize, const int width, const int height)
|
2015-09-07 20:30:22 +02:00
|
|
|
{
|
2017-12-10 01:00:52 +02:00
|
|
|
int64_t sad;
|
|
|
|
int x, y;
|
2017-12-10 20:01:03 +02:00
|
|
|
for (sad = y = 0; y < height - 7; y += 8) {
|
|
|
|
for (x = 0; x < width - 7; x += 8) {
|
2017-12-10 01:00:52 +02:00
|
|
|
sad += sad_8x8_16(p1 + y * p1_linesize + x,
|
|
|
|
p1_linesize,
|
|
|
|
p2 + y * p2_linesize + x,
|
|
|
|
p2_linesize);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return sad;
|
|
|
|
}
|
2015-09-07 20:30:22 +02:00
|
|
|
|
2017-12-10 20:01:03 +02:00
|
|
|
static int64_t scene_sad8(FrameRateContext *s, uint8_t *p1, int p1_linesize, uint8_t* p2, int p2_linesize, const int width, const int height)
|
2017-12-10 01:00:52 +02:00
|
|
|
{
|
|
|
|
int64_t sad;
|
|
|
|
int x, y;
|
2017-12-10 20:01:03 +02:00
|
|
|
for (sad = y = 0; y < height - 7; y += 8) {
|
|
|
|
for (x = 0; x < width - 7; x += 8) {
|
2017-12-10 01:00:52 +02:00
|
|
|
sad += s->sad(p1 + y * p1_linesize + x,
|
|
|
|
p1_linesize,
|
|
|
|
p2 + y * p2_linesize + x,
|
|
|
|
p2_linesize);
|
2015-09-07 20:30:22 +02:00
|
|
|
}
|
|
|
|
}
|
2017-12-10 01:00:52 +02:00
|
|
|
emms_c();
|
|
|
|
return sad;
|
2015-09-07 20:30:22 +02:00
|
|
|
}
|
|
|
|
|
2015-07-24 09:09:16 +02:00
|
|
|
static double get_scene_score(AVFilterContext *ctx, AVFrame *crnt, AVFrame *next)
|
|
|
|
{
|
|
|
|
FrameRateContext *s = ctx->priv;
|
|
|
|
double ret = 0;
|
|
|
|
|
|
|
|
ff_dlog(ctx, "get_scene_score()\n");
|
|
|
|
|
2017-12-10 04:18:49 +02:00
|
|
|
if (crnt->height == next->height &&
|
2015-07-24 09:09:16 +02:00
|
|
|
crnt->width == next->width) {
|
|
|
|
int64_t sad;
|
|
|
|
double mafd, diff;
|
|
|
|
|
|
|
|
ff_dlog(ctx, "get_scene_score() process\n");
|
2017-12-10 01:00:52 +02:00
|
|
|
if (s->bitdepth == 8)
|
2017-12-10 20:01:03 +02:00
|
|
|
sad = scene_sad8(s, crnt->data[0], crnt->linesize[0], next->data[0], next->linesize[0], crnt->width, crnt->height);
|
2017-12-10 01:00:52 +02:00
|
|
|
else
|
2017-12-10 20:01:03 +02:00
|
|
|
sad = scene_sad16(s, (const uint16_t*)crnt->data[0], crnt->linesize[0] / 2, (const uint16_t*)next->data[0], next->linesize[0] / 2, crnt->width, crnt->height);
|
2015-07-24 09:09:16 +02:00
|
|
|
|
2017-12-10 20:01:03 +02:00
|
|
|
mafd = (double)sad * 100.0 / FFMAX(1, (crnt->height & ~7) * (crnt->width & ~7)) / (1 << s->bitdepth);
|
2015-07-24 09:09:16 +02:00
|
|
|
diff = fabs(mafd - s->prev_mafd);
|
|
|
|
ret = av_clipf(FFMIN(mafd, diff), 0, 100.0);
|
|
|
|
s->prev_mafd = mafd;
|
|
|
|
}
|
2017-12-10 01:00:52 +02:00
|
|
|
ff_dlog(ctx, "get_scene_score() result is:%f\n", ret);
|
2015-07-24 09:09:16 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2017-12-09 23:46:27 +02:00
|
|
|
typedef struct ThreadData {
|
|
|
|
AVFrame *copy_src1, *copy_src2;
|
|
|
|
uint16_t src1_factor, src2_factor;
|
|
|
|
} ThreadData;
|
|
|
|
|
2018-01-08 22:44:42 +02:00
|
|
|
static int filter_slice(AVFilterContext *ctx, void *arg, int job, int nb_jobs)
|
2017-12-09 23:46:27 +02:00
|
|
|
{
|
|
|
|
FrameRateContext *s = ctx->priv;
|
|
|
|
ThreadData *td = arg;
|
|
|
|
uint16_t src1_factor = td->src1_factor;
|
|
|
|
uint16_t src2_factor = td->src2_factor;
|
2018-01-08 22:44:42 +02:00
|
|
|
int plane;
|
2017-12-09 23:46:27 +02:00
|
|
|
|
|
|
|
for (plane = 0; plane < 4 && td->copy_src1->data[plane] && td->copy_src2->data[plane]; plane++) {
|
|
|
|
int cpy_line_width = s->line_size[plane];
|
|
|
|
uint8_t *cpy_src1_data = td->copy_src1->data[plane];
|
|
|
|
int cpy_src1_line_size = td->copy_src1->linesize[plane];
|
|
|
|
uint8_t *cpy_src2_data = td->copy_src2->data[plane];
|
|
|
|
int cpy_src2_line_size = td->copy_src2->linesize[plane];
|
|
|
|
int cpy_src_h = (plane > 0 && plane < 3) ? (td->copy_src1->height >> s->vsub) : (td->copy_src1->height);
|
|
|
|
uint8_t *cpy_dst_data = s->work->data[plane];
|
|
|
|
int cpy_dst_line_size = s->work->linesize[plane];
|
|
|
|
const int start = (cpy_src_h * job ) / nb_jobs;
|
|
|
|
const int end = (cpy_src_h * (job+1)) / nb_jobs;
|
|
|
|
cpy_src1_data += start * cpy_src1_line_size;
|
|
|
|
cpy_src2_data += start * cpy_src2_line_size;
|
|
|
|
cpy_dst_data += start * cpy_dst_line_size;
|
|
|
|
|
2018-01-08 22:44:42 +02:00
|
|
|
s->blend(cpy_src1_data, cpy_src1_line_size,
|
|
|
|
cpy_src2_data, cpy_src2_line_size,
|
|
|
|
cpy_dst_data, cpy_dst_line_size,
|
|
|
|
cpy_line_width, end - start,
|
2018-01-15 23:19:46 +02:00
|
|
|
src1_factor, src2_factor, s->blend_factor_max >> 1);
|
2017-12-09 23:46:27 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
static int blend_frames(AVFilterContext *ctx, int interpolate)
|
2015-09-07 20:30:22 +02:00
|
|
|
{
|
|
|
|
FrameRateContext *s = ctx->priv;
|
|
|
|
AVFilterLink *outlink = ctx->outputs[0];
|
|
|
|
double interpolate_scene_score = 0;
|
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
if ((s->flags & FRAMERATE_FLAG_SCD)) {
|
|
|
|
if (s->score >= 0.0)
|
|
|
|
interpolate_scene_score = s->score;
|
2017-12-10 04:18:49 +02:00
|
|
|
else
|
2017-12-20 00:49:41 +02:00
|
|
|
interpolate_scene_score = s->score = get_scene_score(ctx, s->f0, s->f1);
|
2017-12-10 00:00:57 +02:00
|
|
|
ff_dlog(ctx, "blend_frames() interpolate scene score:%f\n", interpolate_scene_score);
|
2015-09-07 20:30:22 +02:00
|
|
|
}
|
|
|
|
// decide if the shot-change detection allows us to blend two frames
|
2017-12-20 00:49:41 +02:00
|
|
|
if (interpolate_scene_score < s->scene_score) {
|
2017-12-09 23:46:27 +02:00
|
|
|
ThreadData td;
|
2017-12-20 00:49:41 +02:00
|
|
|
td.copy_src1 = s->f0;
|
|
|
|
td.copy_src2 = s->f1;
|
|
|
|
td.src2_factor = interpolate;
|
2018-01-15 23:19:46 +02:00
|
|
|
td.src1_factor = s->blend_factor_max - td.src2_factor;
|
2015-09-07 20:30:22 +02:00
|
|
|
|
|
|
|
// get work-space for output frame
|
|
|
|
s->work = ff_get_video_buffer(outlink, outlink->w, outlink->h);
|
|
|
|
if (!s->work)
|
|
|
|
return AVERROR(ENOMEM);
|
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
av_frame_copy_props(s->work, s->f0);
|
2015-09-07 20:30:22 +02:00
|
|
|
|
2017-12-10 00:00:57 +02:00
|
|
|
ff_dlog(ctx, "blend_frames() INTERPOLATE to create work frame\n");
|
avfilter/vf_framerate: add SIMD functions for frame blending
Blend function speedups on x86_64 Core i5 4460:
ffmpeg -f lavfi -i allyuv -vf framerate=60:threads=1 -f null none
C: 447548411 decicycles in Blend, 2048 runs, 0 skips
SSSE3: 130020087 decicycles in Blend, 2048 runs, 0 skips
AVX2: 128508221 decicycles in Blend, 2048 runs, 0 skips
ffmpeg -f lavfi -i allyuv -vf format=yuv420p12,framerate=60:threads=1 -f null none
C: 228932745 decicycles in Blend, 2048 runs, 0 skips
SSE4: 123357781 decicycles in Blend, 2048 runs, 0 skips
AVX2: 121215353 decicycles in Blend, 2048 runs, 0 skips
Signed-off-by: Marton Balint <cus@passwd.hu>
2018-01-08 02:05:45 +02:00
|
|
|
ctx->internal->execute(ctx, filter_slice, &td, NULL, FFMIN(FFMAX(1, outlink->h >> 2), ff_filter_get_nb_threads(ctx)));
|
2015-09-07 20:30:22 +02:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
static int process_work_frame(AVFilterContext *ctx)
|
2015-09-07 20:30:22 +02:00
|
|
|
{
|
|
|
|
FrameRateContext *s = ctx->priv;
|
2017-12-20 00:49:41 +02:00
|
|
|
int64_t work_pts;
|
2018-01-15 23:19:46 +02:00
|
|
|
int64_t interpolate, interpolate8;
|
2017-12-20 00:49:41 +02:00
|
|
|
int ret;
|
2015-07-24 09:09:16 +02:00
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
if (!s->f1)
|
2015-07-24 09:09:16 +02:00
|
|
|
return 0;
|
2017-12-20 00:49:41 +02:00
|
|
|
if (!s->f0 && !s->flush)
|
2015-07-24 09:09:16 +02:00
|
|
|
return 0;
|
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
work_pts = s->start_pts + av_rescale_q(s->n, av_inv_q(s->dest_frame_rate), s->dest_time_base);
|
|
|
|
|
|
|
|
if (work_pts >= s->pts1 && !s->flush)
|
|
|
|
return 0;
|
2015-07-24 09:09:16 +02:00
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
if (!s->f0) {
|
|
|
|
s->work = av_frame_clone(s->f1);
|
|
|
|
} else {
|
|
|
|
if (work_pts >= s->pts1 + s->delta && s->flush)
|
|
|
|
return 0;
|
|
|
|
|
2018-01-15 23:19:46 +02:00
|
|
|
interpolate = av_rescale(work_pts - s->pts0, s->blend_factor_max, s->delta);
|
|
|
|
interpolate8 = av_rescale(work_pts - s->pts0, 256, s->delta);
|
|
|
|
ff_dlog(ctx, "process_work_frame() interpolate: %"PRId64"/256\n", interpolate8);
|
|
|
|
if (interpolate >= s->blend_factor_max || interpolate8 > s->interp_end) {
|
2017-12-20 00:49:41 +02:00
|
|
|
s->work = av_frame_clone(s->f1);
|
2018-01-15 23:19:46 +02:00
|
|
|
} else if (interpolate <= 0 || interpolate8 < s->interp_start) {
|
2017-12-20 00:49:41 +02:00
|
|
|
s->work = av_frame_clone(s->f0);
|
2015-07-24 09:09:16 +02:00
|
|
|
} else {
|
2017-12-20 00:49:41 +02:00
|
|
|
ret = blend_frames(ctx, interpolate);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
if (ret == 0)
|
2018-01-15 23:19:46 +02:00
|
|
|
s->work = av_frame_clone(interpolate > (s->blend_factor_max >> 1) ? s->f1 : s->f0);
|
2015-07-24 09:09:16 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-09-07 20:30:22 +02:00
|
|
|
if (!s->work)
|
2015-07-24 09:09:16 +02:00
|
|
|
return AVERROR(ENOMEM);
|
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
s->work->pts = work_pts;
|
|
|
|
s->n++;
|
2015-07-24 09:09:16 +02:00
|
|
|
|
2017-04-05 01:41:15 +02:00
|
|
|
return 1;
|
2015-07-24 09:09:16 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static av_cold int init(AVFilterContext *ctx)
|
|
|
|
{
|
|
|
|
FrameRateContext *s = ctx->priv;
|
2017-12-20 00:49:41 +02:00
|
|
|
s->start_pts = AV_NOPTS_VALUE;
|
2015-07-24 09:09:16 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static av_cold void uninit(AVFilterContext *ctx)
|
|
|
|
{
|
|
|
|
FrameRateContext *s = ctx->priv;
|
2017-12-20 00:49:41 +02:00
|
|
|
av_frame_free(&s->f0);
|
|
|
|
av_frame_free(&s->f1);
|
2015-07-24 09:09:16 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int query_formats(AVFilterContext *ctx)
|
|
|
|
{
|
|
|
|
static const enum AVPixelFormat pix_fmts[] = {
|
|
|
|
AV_PIX_FMT_YUV410P,
|
|
|
|
AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUVJ411P,
|
|
|
|
AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVJ420P,
|
|
|
|
AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUVJ422P,
|
|
|
|
AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUVJ440P,
|
|
|
|
AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUVJ444P,
|
2015-09-07 20:30:22 +02:00
|
|
|
AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV420P12,
|
|
|
|
AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV422P12,
|
|
|
|
AV_PIX_FMT_YUV444P9, AV_PIX_FMT_YUV444P10, AV_PIX_FMT_YUV444P12,
|
2015-07-24 09:09:16 +02:00
|
|
|
AV_PIX_FMT_NONE
|
|
|
|
};
|
|
|
|
|
|
|
|
AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
|
|
|
|
if (!fmts_list)
|
|
|
|
return AVERROR(ENOMEM);
|
|
|
|
return ff_set_common_formats(ctx, fmts_list);
|
|
|
|
}
|
|
|
|
|
2018-01-08 22:44:42 +02:00
|
|
|
static void blend_frames_c(BLEND_FUNC_PARAMS)
|
|
|
|
{
|
|
|
|
int line, pixel;
|
|
|
|
for (line = 0; line < height; line++) {
|
2018-01-15 23:19:46 +02:00
|
|
|
for (pixel = 0; pixel < width; pixel++)
|
|
|
|
dst[pixel] = ((src1[pixel] * factor1) + (src2[pixel] * factor2) + half) >> BLEND_FACTOR_DEPTH8;
|
2018-01-08 22:44:42 +02:00
|
|
|
src1 += src1_linesize;
|
|
|
|
src2 += src2_linesize;
|
|
|
|
dst += dst_linesize;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void blend_frames16_c(BLEND_FUNC_PARAMS)
|
|
|
|
{
|
|
|
|
int line, pixel;
|
|
|
|
uint16_t *dstw = (uint16_t *)dst;
|
|
|
|
uint16_t *src1w = (uint16_t *)src1;
|
|
|
|
uint16_t *src2w = (uint16_t *)src2;
|
|
|
|
width /= 2;
|
|
|
|
src1_linesize /= 2;
|
|
|
|
src2_linesize /= 2;
|
|
|
|
dst_linesize /= 2;
|
|
|
|
for (line = 0; line < height; line++) {
|
|
|
|
for (pixel = 0; pixel < width; pixel++)
|
2018-01-15 23:19:46 +02:00
|
|
|
dstw[pixel] = ((src1w[pixel] * factor1) + (src2w[pixel] * factor2) + half) >> BLEND_FACTOR_DEPTH16;
|
2018-01-08 22:44:42 +02:00
|
|
|
src1w += src1_linesize;
|
|
|
|
src2w += src2_linesize;
|
|
|
|
dstw += dst_linesize;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
avfilter/vf_framerate: add SIMD functions for frame blending
Blend function speedups on x86_64 Core i5 4460:
ffmpeg -f lavfi -i allyuv -vf framerate=60:threads=1 -f null none
C: 447548411 decicycles in Blend, 2048 runs, 0 skips
SSSE3: 130020087 decicycles in Blend, 2048 runs, 0 skips
AVX2: 128508221 decicycles in Blend, 2048 runs, 0 skips
ffmpeg -f lavfi -i allyuv -vf format=yuv420p12,framerate=60:threads=1 -f null none
C: 228932745 decicycles in Blend, 2048 runs, 0 skips
SSE4: 123357781 decicycles in Blend, 2048 runs, 0 skips
AVX2: 121215353 decicycles in Blend, 2048 runs, 0 skips
Signed-off-by: Marton Balint <cus@passwd.hu>
2018-01-08 02:05:45 +02:00
|
|
|
void ff_framerate_init(FrameRateContext *s)
|
|
|
|
{
|
|
|
|
if (s->bitdepth == 8) {
|
|
|
|
s->blend_factor_max = 1 << BLEND_FACTOR_DEPTH8;
|
|
|
|
s->blend = blend_frames_c;
|
|
|
|
} else {
|
|
|
|
s->blend_factor_max = 1 << BLEND_FACTOR_DEPTH16;
|
|
|
|
s->blend = blend_frames16_c;
|
|
|
|
}
|
|
|
|
if (ARCH_X86)
|
|
|
|
ff_framerate_init_x86(s);
|
|
|
|
}
|
|
|
|
|
2015-07-24 09:09:16 +02:00
|
|
|
static int config_input(AVFilterLink *inlink)
|
|
|
|
{
|
|
|
|
AVFilterContext *ctx = inlink->dst;
|
|
|
|
FrameRateContext *s = ctx->priv;
|
|
|
|
const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
|
|
|
|
int plane;
|
|
|
|
|
|
|
|
for (plane = 0; plane < 4; plane++) {
|
|
|
|
s->line_size[plane] = av_image_get_linesize(inlink->format, inlink->w,
|
|
|
|
plane);
|
|
|
|
}
|
|
|
|
|
2015-09-07 20:30:22 +02:00
|
|
|
s->bitdepth = pix_desc->comp[0].depth;
|
2015-07-24 09:09:16 +02:00
|
|
|
s->vsub = pix_desc->log2_chroma_h;
|
|
|
|
|
|
|
|
s->sad = av_pixelutils_get_sad_fn(3, 3, 2, s); // 8x8 both sources aligned
|
|
|
|
if (!s->sad)
|
|
|
|
return AVERROR(EINVAL);
|
|
|
|
|
|
|
|
s->srce_time_base = inlink->time_base;
|
|
|
|
|
avfilter/vf_framerate: add SIMD functions for frame blending
Blend function speedups on x86_64 Core i5 4460:
ffmpeg -f lavfi -i allyuv -vf framerate=60:threads=1 -f null none
C: 447548411 decicycles in Blend, 2048 runs, 0 skips
SSSE3: 130020087 decicycles in Blend, 2048 runs, 0 skips
AVX2: 128508221 decicycles in Blend, 2048 runs, 0 skips
ffmpeg -f lavfi -i allyuv -vf format=yuv420p12,framerate=60:threads=1 -f null none
C: 228932745 decicycles in Blend, 2048 runs, 0 skips
SSE4: 123357781 decicycles in Blend, 2048 runs, 0 skips
AVX2: 121215353 decicycles in Blend, 2048 runs, 0 skips
Signed-off-by: Marton Balint <cus@passwd.hu>
2018-01-08 02:05:45 +02:00
|
|
|
ff_framerate_init(s);
|
2015-09-07 20:30:22 +02:00
|
|
|
|
2015-07-24 09:09:16 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_frame(AVFilterLink *inlink, AVFrame *inpicref)
|
|
|
|
{
|
2017-04-05 01:41:15 +02:00
|
|
|
int ret;
|
2015-07-24 09:09:16 +02:00
|
|
|
AVFilterContext *ctx = inlink->dst;
|
|
|
|
FrameRateContext *s = ctx->priv;
|
2017-12-20 00:49:41 +02:00
|
|
|
int64_t pts;
|
2015-07-24 09:09:16 +02:00
|
|
|
|
|
|
|
if (inpicref->interlaced_frame)
|
|
|
|
av_log(ctx, AV_LOG_WARNING, "Interlaced frame found - the output will not be correct.\n");
|
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
if (inpicref->pts == AV_NOPTS_VALUE) {
|
|
|
|
av_log(ctx, AV_LOG_WARNING, "Ignoring frame without PTS.\n");
|
|
|
|
return 0;
|
|
|
|
}
|
2015-07-24 09:09:16 +02:00
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
pts = av_rescale_q(inpicref->pts, s->srce_time_base, s->dest_time_base);
|
|
|
|
if (s->f1 && pts == s->pts1) {
|
|
|
|
av_log(ctx, AV_LOG_WARNING, "Ignoring frame with same PTS.\n");
|
|
|
|
return 0;
|
2015-07-24 09:09:16 +02:00
|
|
|
}
|
|
|
|
|
2017-12-20 00:49:41 +02:00
|
|
|
av_frame_free(&s->f0);
|
|
|
|
s->f0 = s->f1;
|
|
|
|
s->pts0 = s->pts1;
|
|
|
|
s->f1 = inpicref;
|
|
|
|
s->pts1 = pts;
|
|
|
|
s->delta = s->pts1 - s->pts0;
|
|
|
|
s->score = -1.0;
|
|
|
|
|
|
|
|
if (s->delta < 0) {
|
|
|
|
av_log(ctx, AV_LOG_WARNING, "PTS discontinuity.\n");
|
|
|
|
s->start_pts = s->pts1;
|
|
|
|
s->n = 0;
|
|
|
|
av_frame_free(&s->f0);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (s->start_pts == AV_NOPTS_VALUE)
|
|
|
|
s->start_pts = s->pts1;
|
|
|
|
|
|
|
|
do {
|
|
|
|
ret = process_work_frame(ctx);
|
|
|
|
if (ret <= 0)
|
|
|
|
return ret;
|
|
|
|
ret = ff_filter_frame(ctx->outputs[0], s->work);
|
|
|
|
} while (ret >= 0);
|
|
|
|
|
|
|
|
return ret;
|
2015-07-24 09:09:16 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int config_output(AVFilterLink *outlink)
|
|
|
|
{
|
|
|
|
AVFilterContext *ctx = outlink->src;
|
|
|
|
FrameRateContext *s = ctx->priv;
|
|
|
|
int exact;
|
|
|
|
|
|
|
|
ff_dlog(ctx, "config_output()\n");
|
|
|
|
|
|
|
|
ff_dlog(ctx,
|
|
|
|
"config_output() input time base:%u/%u (%f)\n",
|
|
|
|
ctx->inputs[0]->time_base.num,ctx->inputs[0]->time_base.den,
|
|
|
|
av_q2d(ctx->inputs[0]->time_base));
|
|
|
|
|
|
|
|
// make sure timebase is small enough to hold the framerate
|
|
|
|
|
|
|
|
exact = av_reduce(&s->dest_time_base.num, &s->dest_time_base.den,
|
|
|
|
av_gcd((int64_t)s->srce_time_base.num * s->dest_frame_rate.num,
|
|
|
|
(int64_t)s->srce_time_base.den * s->dest_frame_rate.den ),
|
|
|
|
(int64_t)s->srce_time_base.den * s->dest_frame_rate.num, INT_MAX);
|
|
|
|
|
|
|
|
av_log(ctx, AV_LOG_INFO,
|
|
|
|
"time base:%u/%u -> %u/%u exact:%d\n",
|
|
|
|
s->srce_time_base.num, s->srce_time_base.den,
|
|
|
|
s->dest_time_base.num, s->dest_time_base.den, exact);
|
|
|
|
if (!exact) {
|
|
|
|
av_log(ctx, AV_LOG_WARNING, "Timebase conversion is not exact\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
outlink->frame_rate = s->dest_frame_rate;
|
|
|
|
outlink->time_base = s->dest_time_base;
|
|
|
|
|
|
|
|
ff_dlog(ctx,
|
|
|
|
"config_output() output time base:%u/%u (%f) w:%d h:%d\n",
|
|
|
|
outlink->time_base.num, outlink->time_base.den,
|
|
|
|
av_q2d(outlink->time_base),
|
|
|
|
outlink->w, outlink->h);
|
|
|
|
|
|
|
|
|
|
|
|
av_log(ctx, AV_LOG_INFO, "fps -> fps:%u/%u scene score:%f interpolate start:%d end:%d\n",
|
|
|
|
s->dest_frame_rate.num, s->dest_frame_rate.den,
|
|
|
|
s->scene_score, s->interp_start, s->interp_end);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int request_frame(AVFilterLink *outlink)
|
|
|
|
{
|
|
|
|
AVFilterContext *ctx = outlink->src;
|
|
|
|
FrameRateContext *s = ctx->priv;
|
2017-12-20 00:49:41 +02:00
|
|
|
int ret;
|
2015-07-24 09:09:16 +02:00
|
|
|
|
|
|
|
ff_dlog(ctx, "request_frame()\n");
|
|
|
|
|
2017-04-05 01:41:15 +02:00
|
|
|
ret = ff_request_frame(ctx->inputs[0]);
|
2017-12-20 00:49:41 +02:00
|
|
|
if (ret == AVERROR_EOF && s->f1 && !s->flush) {
|
2017-04-05 01:41:15 +02:00
|
|
|
s->flush = 1;
|
2017-12-20 00:49:41 +02:00
|
|
|
ret = process_work_frame(ctx);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
ret = ret ? ff_filter_frame(ctx->outputs[0], s->work) : AVERROR_EOF;
|
2017-04-05 01:41:15 +02:00
|
|
|
}
|
2017-12-20 00:49:41 +02:00
|
|
|
|
2017-04-05 01:41:15 +02:00
|
|
|
ff_dlog(ctx, "request_frame() source's request_frame() returned:%d\n", ret);
|
2017-12-20 00:49:41 +02:00
|
|
|
return ret;
|
2015-07-24 09:09:16 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static const AVFilterPad framerate_inputs[] = {
|
|
|
|
{
|
|
|
|
.name = "default",
|
|
|
|
.type = AVMEDIA_TYPE_VIDEO,
|
|
|
|
.config_props = config_input,
|
|
|
|
.filter_frame = filter_frame,
|
|
|
|
},
|
|
|
|
{ NULL }
|
|
|
|
};
|
|
|
|
|
|
|
|
static const AVFilterPad framerate_outputs[] = {
|
|
|
|
{
|
|
|
|
.name = "default",
|
|
|
|
.type = AVMEDIA_TYPE_VIDEO,
|
|
|
|
.request_frame = request_frame,
|
|
|
|
.config_props = config_output,
|
|
|
|
},
|
|
|
|
{ NULL }
|
|
|
|
};
|
|
|
|
|
|
|
|
AVFilter ff_vf_framerate = {
|
|
|
|
.name = "framerate",
|
|
|
|
.description = NULL_IF_CONFIG_SMALL("Upsamples or downsamples progressive source between specified frame rates."),
|
|
|
|
.priv_size = sizeof(FrameRateContext),
|
|
|
|
.priv_class = &framerate_class,
|
|
|
|
.init = init,
|
|
|
|
.uninit = uninit,
|
|
|
|
.query_formats = query_formats,
|
|
|
|
.inputs = framerate_inputs,
|
|
|
|
.outputs = framerate_outputs,
|
2017-12-09 23:46:27 +02:00
|
|
|
.flags = AVFILTER_FLAG_SLICE_THREADS,
|
2015-07-24 09:09:16 +02:00
|
|
|
};
|