Re: [libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-05 Thread Luca Barbato
On 10/05/2012 09:25 AM, Diego Biurrun wrote: On Fri, Oct 05, 2012 at 02:48:20AM +0200, Luca Barbato wrote: On 10/05/2012 01:29 AM, Ronald S. Bultje wrote: On Thu, Oct 4, 2012 at 4:27 PM, Ronald S. Bultje rsbul...@gmail.com wrote: Oh right you mean for performance?

Re: [libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-05 Thread Ronald S. Bultje
Hi, On Fri, Oct 5, 2012 at 12:25 AM, Diego Biurrun di...@biurrun.de wrote: On Fri, Oct 05, 2012 at 02:48:20AM +0200, Luca Barbato wrote: On 10/05/2012 01:29 AM, Ronald S. Bultje wrote: On Thu, Oct 4, 2012 at 4:27 PM, Ronald S. Bultje rsbul...@gmail.com wrote: Oh right you mean for

Re: [libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-05 Thread Derek Buitenhuis
On 05/10/2012 11:47 AM, Ronald S. Bultje wrote: One of the mailinglists recently had an email about our slice threading impl relying on wakeups-only-by-signal, i.e. spurious wakeups could lead to races, can you look at that and see if it fixes it? Perhaps this is related:

Re: [libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-04 Thread Luca Barbato
On 10/04/2012 01:46 AM, Luca Barbato wrote: From: Ronald S. Bultje rsbul...@gmail.com The variable is copied to subsequent threads at the same time, so this may cause wrong ref_count[] values to be copied to subsequent threads. This bug was found using TSAN. Signed-off-by: Luca Barbato

Re: [libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-04 Thread Ronald S. Bultje
Hi, On Thu, Oct 4, 2012 at 12:59 PM, Luca Barbato lu_z...@gentoo.org wrote: On 10/04/2012 01:46 AM, Luca Barbato wrote: From: Ronald S. Bultje rsbul...@gmail.com The variable is copied to subsequent threads at the same time, so this may cause wrong ref_count[] values to be copied to

Re: [libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-04 Thread Ronald S. Bultje
On Thu, Oct 4, 2012 at 4:27 PM, Ronald S. Bultje rsbul...@gmail.com wrote: Hi, On Thu, Oct 4, 2012 at 12:59 PM, Luca Barbato lu_z...@gentoo.org wrote: On 10/04/2012 01:46 AM, Luca Barbato wrote: From: Ronald S. Bultje rsbul...@gmail.com The variable is copied to subsequent threads at the

Re: [libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-04 Thread Luca Barbato
On 10/05/2012 01:29 AM, Ronald S. Bultje wrote: On Thu, Oct 4, 2012 at 4:27 PM, Ronald S. Bultje rsbul...@gmail.com wrote: Oh right you mean for performance? fate-h264-conformance-cama2_vtc_b fate-h264-conformance-cama3_vtc_b should be ok. I'd push your patch + my style nits soon, I'll

[libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-03 Thread Ronald S. Bultje
From: Ronald S. Bultje rsbul...@gmail.com The variable is copied to subsequent threads at the same time, so this may cause wrong ref_count[] values to be copied to subsequent threads. --- libavcodec/h264_cabac.c | 41 - libavcodec/h264_cavlc.c | 33

[libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-03 Thread Ronald S. Bultje
From: Ronald S. Bultje rsbul...@gmail.com The variable is copied to subsequent threads at the same time, so this may cause wrong ref_count[] values to be copied to subsequent threads. --- libavcodec/h264_cabac.c | 41 - libavcodec/h264_cavlc.c | 33

[libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-03 Thread Ronald S. Bultje
From: Ronald S. Bultje rsbul...@gmail.com The variable is copied to subsequent threads at the same time, so this may cause wrong ref_count[] values to be copied to subsequent threads. This bug was found using TSAN. --- libavcodec/h264_cabac.c | 41 -

[libav-devel] [PATCH] h264: don't touch H264Context-ref_count[] during MB decoding.

2012-10-03 Thread Luca Barbato
From: Ronald S. Bultje rsbul...@gmail.com The variable is copied to subsequent threads at the same time, so this may cause wrong ref_count[] values to be copied to subsequent threads. This bug was found using TSAN. Signed-off-by: Luca Barbato lu_z...@gentoo.org --- Few cosmetic nits.