[FFmpeg-devel] [PATCH 3/6] avcodec/pictordec: Optimize picmemset() for single plane full lines
Michael Niedermayer
michael at niedermayer.cc
Fri Aug 23 16:47:59 EEST 2019
On Sat, Aug 03, 2019 at 01:09:49PM +0200, Michael Niedermayer wrote:
> On Sat, Aug 03, 2019 at 12:43:32PM +1000, Peter Ross wrote:
> > On Sat, Aug 03, 2019 at 01:49:54AM +0200, Michael Niedermayer wrote:
> > > Fixes: Timeout (72sec -> 1sec)
> > > Fixes: 15512/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PICTOR_fuzzer-5663942342344704
> > >
> > > Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
> > > Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> > > ---
> > > libavcodec/pictordec.c | 16 +++++++++++++++-
> > > 1 file changed, 15 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/libavcodec/pictordec.c b/libavcodec/pictordec.c
> > > index 2e6fcdca52..5beb03cd5d 100644
> > > --- a/libavcodec/pictordec.c
> > > +++ b/libavcodec/pictordec.c
> > > @@ -66,6 +66,7 @@ static void picmemset(PicContext *s, AVFrame *frame, unsigned value, int run,
> > > int xl = *x;
> > > int yl = *y;
> > > int planel = *plane;
> > > + int pixels_per_value = 8/bits_per_plane;
> > > value <<= shift;
> > >
> > > d = frame->data[0] + yl * frame->linesize[0];
> > > @@ -74,7 +75,7 @@ static void picmemset(PicContext *s, AVFrame *frame, unsigned value, int run,
> > > for (j = 8-bits_per_plane; j >= 0; j -= bits_per_plane) {
> > > d[xl] |= (value >> j) & mask;
> > > xl += 1;
> > > - if (xl == s->width) {
> > > + while (xl == s->width) {
> > > yl -= 1;
> > > xl = 0;
> > > if (yl < 0) {
> > > @@ -86,6 +87,19 @@ static void picmemset(PicContext *s, AVFrame *frame, unsigned value, int run,
> > > mask <<= bits_per_plane;
> > > }
> > > d = frame->data[0] + yl * frame->linesize[0];
> > > + if (s->nb_planes == 1 &&
> > > + run*pixels_per_value >= s->width &&
> > > + pixels_per_value < s->width) {
> > > + int j;
> > > + for (j = 8-bits_per_plane; j >= 0; j -= bits_per_plane) {
> >
> > suggest naming it 'k' to avoid confusion with earlier for loop.
>
> actually, looking at this again, i think we should use the same j,
> This also now excludes s->width % pixels_per_value != 0 for which i suspect there
> is no testcase. Ill add support for this in case the fuzzer finds a case
> that way we then also have a testcase for implementing that corner case.
>
> heres the new code:
>
> --- a/libavcodec/pictordec.c
> +++ b/libavcodec/pictordec.c
> @@ -66,6 +66,7 @@ static void picmemset(PicContext *s, AVFrame *frame, unsigned value, int run,
> int xl = *x;
> int yl = *y;
> int planel = *plane;
> + int pixels_per_value = 8/bits_per_plane;
> value <<= shift;
>
> d = frame->data[0] + yl * frame->linesize[0];
> @@ -74,7 +75,7 @@ static void picmemset(PicContext *s, AVFrame *frame, unsigned value, int run,
> for (j = 8-bits_per_plane; j >= 0; j -= bits_per_plane) {
> d[xl] |= (value >> j) & mask;
> xl += 1;
> - if (xl == s->width) {
> + while (xl == s->width) {
> yl -= 1;
> xl = 0;
> if (yl < 0) {
> @@ -86,6 +87,19 @@ static void picmemset(PicContext *s, AVFrame *frame, unsigned value, int run,
> mask <<= bits_per_plane;
> }
> d = frame->data[0] + yl * frame->linesize[0];
> + if (s->nb_planes == 1 &&
> + run*pixels_per_value >= s->width &&
> + pixels_per_value < s->width &&
> + s->width % pixels_per_value == 0
> + ) {
> + for (; xl < pixels_per_value; xl ++) {
> + j = (j < bits_per_plane ? 8 : j) - bits_per_plane;
> + d[xl] |= (value >> j) & mask;
> + }
> + av_memcpy_backptr(d+xl, pixels_per_value, s->width - xl);
> + run -= s->width / pixels_per_value;
> + xl = s->width;
> + }
> }
> }
> run--;
will apply
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Whats the most studid thing your enemy could do ? Blow himself up
Whats the most studid thing you could do ? Give up your rights and
freedom because your enemy blew himself up.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20190823/90cbe337/attachment.sig>
More information about the ffmpeg-devel
mailing list