FFmpeg/libavcodec/vp9lpf.c
Andreas Rheinhardt 7bd3b73716 avcodec/vp9: Switch to ProgressFrames
This already fixes a race in the vp9-encparams test. In this test,
side data is added to the current frame after having been decoded
(and therefore after ff_thread_finish_setup() has been called).
Yet the update_thread_context callback called ff_thread_ref_frame()
and therefore av_frame_ref() with this frame as source frame and
the ensuing read was unsynchronised with adding the side data,
i.e. there was a data race.

By switching to the ProgressFrame API the implicit av_frame_ref()
is removed and the race fixed except if this frame is later reused by
a show-existing-frame which uses an explicit av_frame_ref().
The vp9-encparams test does not cover this, so this commit
already fixes all the races in this test.

This decoder kept multiple references to the same ThreadFrames
in the same context and therefore had lots of implicit av_frame_ref()
even when decoding single-threaded. This incurred lots of small
allocations: When decoding an ordinary 10s video in single-threaded
mode the number of allocations reported by Valgrind went down
from 57,814 to 20,908; for 10 threads it went down from 84,223 to
21,901.

Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2024-04-19 13:18:04 +02:00

204 lines
8.8 KiB
C

/*
* VP9 compatible video decoder
*
* Copyright (C) 2013 Ronald S. Bultje <rsbultje gmail com>
* Copyright (C) 2013 Clément Bœsch <u pkh me>
*
* 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
*/
#include "avcodec.h"
#include "vp9dec.h"
static av_always_inline void filter_plane_cols(VP9Context *s, int col, int ss_h, int ss_v,
uint8_t *lvl, uint8_t (*mask)[4],
uint8_t *dst, ptrdiff_t ls)
{
int y, x, bytesperpixel = s->bytesperpixel;
// filter edges between columns (e.g. block1 | block2)
for (y = 0; y < 8; y += 2 << ss_v, dst += 16 * ls, lvl += 16 << ss_v) {
uint8_t *ptr = dst, *l = lvl, *hmask1 = mask[y], *hmask2 = mask[y + 1 + ss_v];
unsigned hm1 = hmask1[0] | hmask1[1] | hmask1[2], hm13 = hmask1[3];
unsigned hm2 = hmask2[1] | hmask2[2], hm23 = hmask2[3];
unsigned hm = hm1 | hm2 | hm13 | hm23;
for (x = 1; hm & ~(x - 1); x <<= 1, ptr += 8 * bytesperpixel >> ss_h) {
if (col || x > 1) {
if (hm1 & x) {
int L = *l, H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
if (hmask1[0] & x) {
if (hmask2[0] & x) {
av_assert2(l[8 << ss_v] == L);
s->dsp.loop_filter_16[0](ptr, ls, E, I, H);
} else {
s->dsp.loop_filter_8[2][0](ptr, ls, E, I, H);
}
} else if (hm2 & x) {
L = l[8 << ss_v];
H |= (L >> 4) << 8;
E |= s->filter_lut.mblim_lut[L] << 8;
I |= s->filter_lut.lim_lut[L] << 8;
s->dsp.loop_filter_mix2[!!(hmask1[1] & x)]
[!!(hmask2[1] & x)]
[0](ptr, ls, E, I, H);
} else {
s->dsp.loop_filter_8[!!(hmask1[1] & x)]
[0](ptr, ls, E, I, H);
}
} else if (hm2 & x) {
int L = l[8 << ss_v], H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
s->dsp.loop_filter_8[!!(hmask2[1] & x)]
[0](ptr + 8 * ls, ls, E, I, H);
}
}
if (ss_h) {
if (x & 0xAA)
l += 2;
} else {
if (hm13 & x) {
int L = *l, H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
if (hm23 & x) {
L = l[8 << ss_v];
H |= (L >> 4) << 8;
E |= s->filter_lut.mblim_lut[L] << 8;
I |= s->filter_lut.lim_lut[L] << 8;
s->dsp.loop_filter_mix2[0][0][0](ptr + 4 * bytesperpixel, ls, E, I, H);
} else {
s->dsp.loop_filter_8[0][0](ptr + 4 * bytesperpixel, ls, E, I, H);
}
} else if (hm23 & x) {
int L = l[8 << ss_v], H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
s->dsp.loop_filter_8[0][0](ptr + 8 * ls + 4 * bytesperpixel, ls, E, I, H);
}
l++;
}
}
}
}
static av_always_inline void filter_plane_rows(VP9Context *s, int row, int ss_h, int ss_v,
uint8_t *lvl, uint8_t (*mask)[4],
uint8_t *dst, ptrdiff_t ls)
{
int y, x, bytesperpixel = s->bytesperpixel;
// block1
// filter edges between rows (e.g. ------)
// block2
for (y = 0; y < 8; y++, dst += 8 * ls >> ss_v) {
uint8_t *ptr = dst, *l = lvl, *vmask = mask[y];
unsigned vm = vmask[0] | vmask[1] | vmask[2], vm3 = vmask[3];
for (x = 1; vm & ~(x - 1); x <<= (2 << ss_h), ptr += 16 * bytesperpixel, l += 2 << ss_h) {
if (row || y) {
if (vm & x) {
int L = *l, H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
if (vmask[0] & x) {
if (vmask[0] & (x << (1 + ss_h))) {
av_assert2(l[1 + ss_h] == L);
s->dsp.loop_filter_16[1](ptr, ls, E, I, H);
} else {
s->dsp.loop_filter_8[2][1](ptr, ls, E, I, H);
}
} else if (vm & (x << (1 + ss_h))) {
L = l[1 + ss_h];
H |= (L >> 4) << 8;
E |= s->filter_lut.mblim_lut[L] << 8;
I |= s->filter_lut.lim_lut[L] << 8;
s->dsp.loop_filter_mix2[!!(vmask[1] & x)]
[!!(vmask[1] & (x << (1 + ss_h)))]
[1](ptr, ls, E, I, H);
} else {
s->dsp.loop_filter_8[!!(vmask[1] & x)]
[1](ptr, ls, E, I, H);
}
} else if (vm & (x << (1 + ss_h))) {
int L = l[1 + ss_h], H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
s->dsp.loop_filter_8[!!(vmask[1] & (x << (1 + ss_h)))]
[1](ptr + 8 * bytesperpixel, ls, E, I, H);
}
}
if (!ss_v) {
if (vm3 & x) {
int L = *l, H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
if (vm3 & (x << (1 + ss_h))) {
L = l[1 + ss_h];
H |= (L >> 4) << 8;
E |= s->filter_lut.mblim_lut[L] << 8;
I |= s->filter_lut.lim_lut[L] << 8;
s->dsp.loop_filter_mix2[0][0][1](ptr + ls * 4, ls, E, I, H);
} else {
s->dsp.loop_filter_8[0][1](ptr + ls * 4, ls, E, I, H);
}
} else if (vm3 & (x << (1 + ss_h))) {
int L = l[1 + ss_h], H = L >> 4;
int E = s->filter_lut.mblim_lut[L], I = s->filter_lut.lim_lut[L];
s->dsp.loop_filter_8[0][1](ptr + ls * 4 + 8 * bytesperpixel, ls, E, I, H);
}
}
}
if (ss_v) {
if (y & 1)
lvl += 16;
} else {
lvl += 8;
}
}
}
void ff_vp9_loopfilter_sb(AVCodecContext *avctx, VP9Filter *lflvl,
int row, int col, ptrdiff_t yoff, ptrdiff_t uvoff)
{
VP9Context *s = avctx->priv_data;
AVFrame *f = s->s.frames[CUR_FRAME].tf.f;
uint8_t *dst = f->data[0] + yoff;
ptrdiff_t ls_y = f->linesize[0], ls_uv = f->linesize[1];
uint8_t (*uv_masks)[8][4] = lflvl->mask[s->ss_h | s->ss_v];
int p;
/* FIXME: In how far can we interleave the v/h loopfilter calls? E.g.
* if you think of them as acting on a 8x8 block max, we can interleave
* each v/h within the single x loop, but that only works if we work on
* 8 pixel blocks, and we won't always do that (we want at least 16px
* to use SSE2 optimizations, perhaps 32 for AVX2) */
filter_plane_cols(s, col, 0, 0, lflvl->level, lflvl->mask[0][0], dst, ls_y);
filter_plane_rows(s, row, 0, 0, lflvl->level, lflvl->mask[0][1], dst, ls_y);
for (p = 0; p < 2; p++) {
dst = f->data[1 + p] + uvoff;
filter_plane_cols(s, col, s->ss_h, s->ss_v, lflvl->level, uv_masks[0], dst, ls_uv);
filter_plane_rows(s, row, s->ss_h, s->ss_v, lflvl->level, uv_masks[1], dst, ls_uv);
}
}