Re: [libav-devel] [PATCH] avformat: Free the internal codec context at the end

On Apr 12, 2017 8:00 AM, "Vittorio Giovara" wrote: On Tue, Apr 11, 2017 at 7:54 PM, Sean McGovern wrote: > On Apr 11, 2017 7:48 PM, "Luca Barbato" wrote: > > Avoid a use after free in avformat_find_stream_info. > > CC:

Re: [libav-devel] [PATCH] hevc: Add NEON 16x16 IDCT

On Fri, 7 Apr 2017, Martin Storsjö wrote: These 8 macro invocations adds 8 copies of the almost identical same code though - the size of this object file is inflated a great deal due to this. Before this, the code size ends up at 2280 bytes, after this patch it ends up at 15976 bytes. This

Re: [libav-devel] [PATCH] hevc: Add NEON 16x16 IDCT

On Wed, 12 Apr 2017, Alexandra Hájková wrote: The speedup vs C code is around 6-13x. --- Use irp to avoid the repetition. libavcodec/arm/hevc_idct.S| 196 ++ libavcodec/arm/hevcdsp_init_arm.c | 4 + 2 files changed, 200 insertions(+) diff --git

Re: [libav-devel] [PATCH] hevc: Optimize NEON 8x8 IDCT using col_limit

On Wed, 12 Apr 2017, Alexandra Hájková wrote: --- libavcodec/arm/hevc_idct.S | 26 +++--- 1 file changed, 15 insertions(+), 11 deletions(-) Given that the whole col_limit parameter seems to not be clearly defined, I'd skip this patch for now until it has been clarified.

Re: [libav-devel] [PATCH] checkasm: Test sub-idcts for hevc

On Wed, 12 Apr 2017, Martin Storsjö wrote: This allows testing and benchmarking use of the col_limit parameter in the idct implementations. --- I can't say I really understand the logic for setting the col_limit parameter based on the range on where non-zero coefficients are set, but this set

Re: [libav-devel] [PATCH] mem: uninline av_malloc(z)_array()

On 12/04/2017 15:27, Anton Khirnov wrote: > Inlining public functions hardcodes their implementation into the ABI, > so it should be avoided unless there is a very good reason for it. No > such reason exists in this case. > --- > libavutil/mem.c | 14 ++ > libavutil/mem.h | 14

Re: [libav-devel] [PATCH] golomb: Simplify get_ur_golomb_jpegls

On 11/04/2017 17:41, Diego Biurrun wrote: > Not necessary IMO. the line is extra long so with just the ; you get something quite weird looking at 80 cols. ___ libav-devel mailing list libav-devel@libav.org

[libav-devel] [PATCH] mem: uninline av_malloc(z)_array()

Inlining public functions hardcodes their implementation into the ABI, so it should be avoided unless there is a very good reason for it. No such reason exists in this case. --- libavutil/mem.c | 14 ++ libavutil/mem.h | 14 ++ 2 files changed, 16 insertions(+), 12

Re: [libav-devel] [PATCH] avutil: av_frame_ref: copy source frame flags

Le 12/04/2017 à 12:16, Hendrik Leppkes a écrit : >> >> dst->format = src->format; >> +dst->flags = src->flags; >> dst->width = src->width; >> dst->height = src->height; >> dst->channel_layout = src->channel_layout; > > You must be

Re: [libav-devel] [PATCH 2/7] utvideodec: Support UQY2

On 11/04/2017 10:11, Diego Biurrun wrote: > extradata is uint8_t*, so this should be PRIu8. Maybe it could be done > in a separate patch though. It should since just above there is the same thing. I'd push this one now. lu ___ libav-devel mailing list

Re: [libav-devel] [PATCH] checkasm: Test sub-idcts for hevc

On 12/04/2017 13:53, Martin Storsjö wrote: > This allows testing and benchmarking use of the col_limit parameter > in the idct implementations. > --- > I can't say I really understand the logic for setting the col_limit > parameter based on the range on where non-zero coefficients are set, > but

Re: [libav-devel] [PATCH] hevc: Optimize NEON 8x8 IDCT using col_limit

On Wed, 12 Apr 2017, Alexandra Hájková wrote: --- libavcodec/arm/hevc_idct.S | 26 +++--- 1 file changed, 15 insertions(+), 11 deletions(-) diff --git a/libavcodec/arm/hevc_idct.S b/libavcodec/arm/hevc_idct.S index 4124fc8..29135ad 100644 --- a/libavcodec/arm/hevc_idct.S +++

Re: [libav-devel] [PATCH] avformat: Free the internal codec context at the end

On Tue, Apr 11, 2017 at 7:54 PM, Sean McGovern wrote: > On Apr 11, 2017 7:48 PM, "Luca Barbato" wrote: > > Avoid a use after free in avformat_find_stream_info. > > CC: libav-sta...@libav.org > --- > libavformat/utils.c | 6 +- > 1 file changed, 1

[libav-devel] [PATCH] hevc: Optimize NEON 8x8 IDCT using col_limit

--- libavcodec/arm/hevc_idct.S | 26 +++--- 1 file changed, 15 insertions(+), 11 deletions(-) diff --git a/libavcodec/arm/hevc_idct.S b/libavcodec/arm/hevc_idct.S index 4124fc8..29135ad 100644 --- a/libavcodec/arm/hevc_idct.S +++ b/libavcodec/arm/hevc_idct.S @@ -58,7 +58,7 @@

[libav-devel] [PATCH] checkasm: Test sub-idcts for hevc

This allows testing and benchmarking use of the col_limit parameter in the idct implementations. --- I can't say I really understand the logic for setting the col_limit parameter based on the range on where non-zero coefficients are set, but this set up at least works with the reference C version

Re: [libav-devel] [PATCH] avutil: av_frame_ref: copy source frame flags

On Wed, Apr 12, 2017 at 12:06 PM, Francois Cartegnie wrote: > This happens at least with h264 (I have not tested other codecs). > > When corrupted frames output is set, any corrupted > frame is flagged but that information is lost when > av_frame_ref is used on output. > > The

[libav-devel] [PATCH] avutil: av_frame_ref: copy source frame flags

This happens at least with h264 (I have not tested other codecs). When corrupted frames output is set, any corrupted frame is flagged but that information is lost when av_frame_ref is used on output. The other local only h264 fix would be to copy flags in h264dec output_frame(). Francois ---

Re: [libav-devel] [PATCH] thread: Define ff_mutex_* macros as stub functions when threads are disabled

On 31/03/2017 19:35, Diego Biurrun wrote: > Silences a bunch of "statement with no effect" warnings with threads disabled. > --- > > Now with stub functions, should be more correct. > > libavutil/thread.h | 8 > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git

Re: [libav-devel] [PATCH] thread: Define ff_mutex_* macros as stub functions when threads are disabled

On Fri, Mar 31, 2017 at 07:35:35PM +0200, Diego Biurrun wrote: > Silences a bunch of "statement with no effect" warnings with threads disabled. > --- > > Now with stub functions, should be more correct. > > libavutil/thread.h | 8 > 1 file changed, 4 insertions(+), 4 deletions(-) ping

[libav-devel] [PATCH] hevc: Add NEON 16x16 IDCT

The speedup vs C code is around 6-13x. --- Use irp to avoid the repetition. libavcodec/arm/hevc_idct.S| 196 ++ libavcodec/arm/hevcdsp_init_arm.c | 4 + 2 files changed, 200 insertions(+) diff --git a/libavcodec/arm/hevc_idct.S

Re: [libav-devel] [PATCH 1/8] hevc: properly handle no_rasl_output_flag when removing pictures from the DPB

Yes, I probably messed up. Merging code from very different 2 year old code is not exactly easy. I don't know how you people do this... On Tue, Apr 11, 2017 at 3:33 PM, Hendrik Leppkes wrote: > On Tue, Apr 11, 2017 at 3:29 PM, Anton Khirnov wrote: >>