date:20181217

[FFmpeg-devel] [PATCH] libavcodec/fft_template: improve performance of the ff_fft_init in fft_template

2018-12-17 Thread Steven Liu

After patch:
init nbits = 17, get 1 samples, duration: 15221
Before patch:
init nbits = 17, get 1 samples, duration: 16105

test script:
DURATION=0
for((i=0;i<1;i++)) do
./libavcodec/tests/fft -n 17 &>output
T_DURATION=`grep "duration" output | awk -F"=" '{ print $2}'`
DURATION=`expr $DURATION + $T_DURATION`
done
TOTAL=`expr $DURATION / 1`
echo $TOTAL

Signed-off-by: Steven Liu 
---
 libavcodec/fft_template.c | 27 +--
 1 file changed, 17 insertions(+), 10 deletions(-)

diff --git a/libavcodec/fft_template.c b/libavcodec/fft_template.c
index 762c014bc8..5f6f52275d 100644
--- a/libavcodec/fft_template.c
+++ b/libavcodec/fft_template.c
@@ -257,21 +257,28 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int 
inverse)
 }
 #endif /* FFT_FIXED_32 */
 
+#define SPLIT_RADIX_PERMUTATION(num) \
+for(i=0; ifft_permutation == FF_FFT_PERM_SWAP_LSBS) \
+j = (j&~3) | ((j>>1)&1) | ((j<<1)&2); \
+k = -split_radix_permutation(i, n, s->inverse) & (n-1); \
+j = (j&~3) | ((j>>1)&1) | ((j<<1)&2); \
+k = -split_radix_permutation(i, n, s->inverse) & (n-1); \
+s->revtab##num[k] = j; \
+}
 
 if (s->fft_permutation == FF_FFT_PERM_AVX) {
 fft_perm_avx(s);
 } else {
-for(i=0; ifft_permutation == FF_FFT_PERM_SWAP_LSBS)
-j = (j&~3) | ((j>>1)&1) | ((j<<1)&2);
-k = -split_radix_permutation(i, n, s->inverse) & (n-1);
-if (s->revtab)
-s->revtab[k] = j;
-if (s->revtab32)
-s->revtab32[k] = j;
+if (s->revtab) {
+SPLIT_RADIX_PERMUTATION()
+}
+if (s->revtab32) {
+SPLIT_RADIX_PERMUTATION(32)
 }
+#undef SPLIT_RADIX_PERMUTATION
 }
 
 return 0;
-- 
2.15.2 (Apple Git-101.1)



___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH] avcodec/mips: [loongson] enable MSA optimization for loongson platform.

2018-12-17 Thread Shiyou Yin

Set initialization order of MSA after MMI to make it work on loongson 
platform(msa is supported by loongson2k、3a4000 etc.).
---
 libavcodec/mips/blockdsp_init_mips.c| 6 +++---
 libavcodec/mips/h264chroma_init_mips.c  | 6 +++---
 libavcodec/mips/h264dsp_init_mips.c | 6 +++---
 libavcodec/mips/h264pred_init_mips.c| 6 +++---
 libavcodec/mips/h264qpel_init_mips.c| 6 +++---
 libavcodec/mips/hpeldsp_init_mips.c | 6 +++---
 libavcodec/mips/idctdsp_init_mips.c | 6 +++---
 libavcodec/mips/mpegvideo_init_mips.c   | 6 +++---
 libavcodec/mips/pixblockdsp_init_mips.c | 6 +++---
 libavcodec/mips/vp8dsp_init_mips.c  | 6 +++---
 10 files changed, 30 insertions(+), 30 deletions(-)

diff --git a/libavcodec/mips/blockdsp_init_mips.c 
b/libavcodec/mips/blockdsp_init_mips.c
index 30ae95f..55ac1c3 100644
--- a/libavcodec/mips/blockdsp_init_mips.c
+++ b/libavcodec/mips/blockdsp_init_mips.c
@@ -45,10 +45,10 @@ static av_cold void blockdsp_init_mmi(BlockDSPContext *c)
 
 void ff_blockdsp_init_mips(BlockDSPContext *c)
 {
-#if HAVE_MSA
-blockdsp_init_msa(c);
-#endif  // #if HAVE_MSA
 #if HAVE_MMI
 blockdsp_init_mmi(c);
 #endif /* HAVE_MMI */
+#if HAVE_MSA
+blockdsp_init_msa(c);
+#endif  // #if HAVE_MSA
 }
diff --git a/libavcodec/mips/h264chroma_init_mips.c 
b/libavcodec/mips/h264chroma_init_mips.c
index 122148d..ae817e4 100644
--- a/libavcodec/mips/h264chroma_init_mips.c
+++ b/libavcodec/mips/h264chroma_init_mips.c
@@ -54,10 +54,10 @@ static av_cold void h264chroma_init_mmi(H264ChromaContext 
*c, int bit_depth)
 
 av_cold void ff_h264chroma_init_mips(H264ChromaContext *c, int bit_depth)
 {
-#if HAVE_MSA
-h264chroma_init_msa(c, bit_depth);
-#endif  // #if HAVE_MSA
 #if HAVE_MMI
 h264chroma_init_mmi(c, bit_depth);
 #endif /* HAVE_MMI */
+#if HAVE_MSA
+h264chroma_init_msa(c, bit_depth);
+#endif  // #if HAVE_MSA
 }
diff --git a/libavcodec/mips/h264dsp_init_mips.c 
b/libavcodec/mips/h264dsp_init_mips.c
index 1fe7f84..dc08a25 100644
--- a/libavcodec/mips/h264dsp_init_mips.c
+++ b/libavcodec/mips/h264dsp_init_mips.c
@@ -138,10 +138,10 @@ static av_cold void h264dsp_init_mmi(H264DSPContext * c, 
const int bit_depth,
 av_cold void ff_h264dsp_init_mips(H264DSPContext *c, const int bit_depth,
   const int chroma_format_idc)
 {
-#if HAVE_MSA
-h264dsp_init_msa(c, bit_depth, chroma_format_idc);
-#endif  // #if HAVE_MSA
 #if HAVE_MMI
 h264dsp_init_mmi(c, bit_depth, chroma_format_idc);
 #endif /* HAVE_MMI */
+#if HAVE_MSA
+h264dsp_init_msa(c, bit_depth, chroma_format_idc);
+#endif  // #if HAVE_MSA
 }
diff --git a/libavcodec/mips/h264pred_init_mips.c 
b/libavcodec/mips/h264pred_init_mips.c
index c33d8f7..63637b8 100644
--- a/libavcodec/mips/h264pred_init_mips.c
+++ b/libavcodec/mips/h264pred_init_mips.c
@@ -146,10 +146,10 @@ av_cold void ff_h264_pred_init_mips(H264PredContext *h, 
int codec_id,
 int bit_depth,
 const int chroma_format_idc)
 {
-#if HAVE_MSA
-h264_pred_init_msa(h, codec_id, bit_depth, chroma_format_idc);
-#endif  // #if HAVE_MSA
 #if HAVE_MMI
 h264_pred_init_mmi(h, codec_id, bit_depth, chroma_format_idc);
 #endif /* HAVE_MMI */
+#if HAVE_MSA
+h264_pred_init_msa(h, codec_id, bit_depth, chroma_format_idc);
+#endif  // #if HAVE_MSA
 }
diff --git a/libavcodec/mips/h264qpel_init_mips.c 
b/libavcodec/mips/h264qpel_init_mips.c
index 92219f8..33bae30 100644
--- a/libavcodec/mips/h264qpel_init_mips.c
+++ b/libavcodec/mips/h264qpel_init_mips.c
@@ -240,10 +240,10 @@ static av_cold void h264qpel_init_mmi(H264QpelContext *c, 
int bit_depth)
 
 av_cold void ff_h264qpel_init_mips(H264QpelContext *c, int bit_depth)
 {
-#if HAVE_MSA
-h264qpel_init_msa(c, bit_depth);
-#endif  // #if HAVE_MSA
 #if HAVE_MMI
 h264qpel_init_mmi(c, bit_depth);
 #endif /* HAVE_MMI */
+#if HAVE_MSA
+h264qpel_init_msa(c, bit_depth);
+#endif  // #if HAVE_MSA
 }
diff --git a/libavcodec/mips/hpeldsp_init_mips.c 
b/libavcodec/mips/hpeldsp_init_mips.c
index 363a045..d6f7a97 100644
--- a/libavcodec/mips/hpeldsp_init_mips.c
+++ b/libavcodec/mips/hpeldsp_init_mips.c
@@ -113,10 +113,10 @@ static void ff_hpeldsp_init_mmi(HpelDSPContext *c, int 
flags)
 
 void ff_hpeldsp_init_mips(HpelDSPContext *c, int flags)
 {
-#if HAVE_MSA
-ff_hpeldsp_init_msa(c, flags);
-#endif  // #if HAVE_MSA
 #if HAVE_MMI
 ff_hpeldsp_init_mmi(c, flags);
 #endif  // #if HAVE_MMI
+#if HAVE_MSA
+ff_hpeldsp_init_msa(c, flags);
+#endif  // #if HAVE_MSA
 }
diff --git a/libavcodec/mips/idctdsp_init_mips.c 
b/libavcodec/mips/idctdsp_init_mips.c
index bb33b55..85b76ca 100644
--- a/libavcodec/mips/idctdsp_init_mips.c
+++ b/libavcodec/mips/idctdsp_init_mips.c
@@ -65,10 +65,10 @@ static av_cold void idctdsp_init_mmi(IDCTDSPContext *c, 
AVCodecContext *avctx,
 av_cold void ff_idctdsp_init_mips(IDCTDSPContext *c, AVCodecContext *avctx,
   unsigned high_bit_depth)
 {
-#if HAVE_MSA
-idctdsp_init_msa(

Re: [FFmpeg-devel] NUT end timestamp

2018-12-17 Thread Paul B Mahol

On 12/14/18, Michael Niedermayer  wrote:
> Hi
>
> replying to a comment from IRC here because "durandal_1707 has quit" so it
> wouldnt be seen, and might be interresting for others so not mailing
> private
>
>  michaelni: nut does not take into consideration duration of
> last frame to return real duration
>  michaelni: so duration is always smaller by small amount
>  michaelni: looks like design failure
>
> Its a bit ago that i worked on nut but the last frame duration was
> considered,
> i do remember there was discussion about it.
>
> looking at nut.txt, it says this:
> All streams SHOULD end with EOR, where the pts of the EOR indicates the
> end presentation time of the final frame.
>
> That stores and gives the last duration. Our muxer seems to not implement
> this
> so its a implementation issue not a design one

If one need to parse all packets to get last duration that is flawed design.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH] avformat/mpegts: unset DTS/PTS for subtitle PES packets if PCR not available

2018-12-17 Thread Jan Ekström

On Sat, Dec 15, 2018, 20:50 Jan Ekström  Fixes issues when a subtitle packet is received before PCR for the
> program has been received, leading to wildly jumping timestamps
> on the lavf client side as well as in the re-ordering logic.
>
> This usually happens in case of multiplexes where the PCR of a
> program is not taken into account with subtitle tracks' DTS/PTS.
> --


Friendly ping. Sample where this improves the result is linked in the first
attempt @ https://patchwork.ffmpeg.org/patch/11419/

Jan
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] avformat/hls.c: Fix memory leak

2018-12-17 Thread Valery Kot

I just verified that patch is working. Build FFmpeg master with and
without patch, run HLS live stream from Youtube (extracted by
youtube_dl for Euronews live).

ffmpeg -i
https://manifest.googlevideo.com/api/manifest/hls_playlist/id/V2E-jOUVsd4.1/itag/95/source/yt_live_broadcast/requiressl/yes/ratebypass/yes/live/1/cmbypass/yes/goi/160/sgoap/gir%3Dyes%3Bitag%3D140/sgovp/gir%3Dyes%3Bitag%3D136/hls_chunk_host/r4---sn-5hnedn7e.googlevideo.com/ei/E2sXXOrRGdjGgAe_6abgBA/gcr/nl/playlist_type/DVR/initcwndbps/12170/mm/32/mn/sn-5hnedn7e/ms/lv/mv/m/pl/24/dover/11/keepalive/yes/mt/1545038545/disable_polymer/true/ip/80.255.245.45/ipbits/0/expire/1545060211/sparams/ip,ipbits,expire,id,itag,source,requiressl,ratebypass,live,cmbypass,goi,sgoap,sgovp,hls_chunk_host,ei,gcr,playlist_type,initcwndbps,mm,mn,ms,mv,pl/signature/517F67575565239FC2D7117212AB722FB68BA3A6.4A8F847C14BF8C702103AEC348C8AC553E812ECE/key/dg_yt0/playlist/index.m3u8
-c copy -t 3 -f mpegts -y NUL 2> _out.txt

Checking process memory (private working set) in Task Manager
Both instances start at 27MB RAM footprint.
Patched version within a minute grows to 31 MB and then stable.
Unmodified master RAM keeps growing, and after one hour reaches 90MB.
Thus 59MB leak in an hour! And keeps growing...

Youtube updates variant playlists app. every 2 seconds, with 3600
fragments each. Means leak of 8(bytes)*3600(pointers in
prev_segments)*30*60(playlists per hour) = 54M - correlates with
experimental results.
Of course, youtube is an extreme case, but leak also degrades over
time any live HLS input.

Please review.
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH]lsws/utils: Split "emms_c(); " call in two lines

2018-12-17 Thread Michael Niedermayer

On Mon, Dec 17, 2018 at 02:14:26AM +0100, Carl Eugen Hoyos wrote:
> Hi!
> 
> Attached patch fixes a useless warning when compiling with clang on arm.
> 
> Please comment, Carl Eugen

>  utils.c |3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> fb6799bbc59d79d9dfba9f8661aaecbed568f3a5  
> 0001-lsws-utils-Split-emms_c-in-two-lines.patch
> From b2773d1ca9473380b99400c2bbd5a6729622f51a Mon Sep 17 00:00:00 2001
> From: Carl Eugen Hoyos 
> Date: Mon, 17 Dec 2018 02:12:13 +0100
> Subject: [PATCH] lsws/utils: Split "emms_c();" in two lines.
> 
> Silences a clang warning when not compiling for x86:
> libswscale/utils.c:345:13: warning: while loop has empty body
> ---
>  libswscale/utils.c |3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/libswscale/utils.c b/libswscale/utils.c
> index df68bcc..de64694 100644
> --- a/libswscale/utils.c
> +++ b/libswscale/utils.c
> @@ -342,7 +342,8 @@ static av_cold int initFilter(int16_t **outFilter, 
> int32_t **filterPos,
>  const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8));
>  int ret= -1;
>  
> -emms_c(); // FIXME should not be required but IS (even for non-MMX 
> versions)
> +emms_c() // FIXME should not be required but IS (even for non-MMX 
> versions)
> +;

this looks like a bad change that could confuse people
i think a different solution should be found or the warning left if theres none

thx

[...]

-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

I have never wished to cater to the crowd; for what I know they do not
approve, and what they approve I do not know. -- Epicurus


signature.asc
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH v4] avformat/movenc: Added an option to disable SIDX atom

2018-12-17 Thread Ronak Patel


> On Dec 13, 2018, at 9:15 PM, Ronak Patel  
> wrote:
> 
> 
> 
>> On Dec 12, 2018, at 2:46 AM, Liu Steven  wrote:
>> 
>> 
>> 
 在 2018年12月12日，上午5:08，Ronak  写道：
 
 
 On Dec 11, 2018, at 3:28 PM, Gyan  wrote:
 
 
> On 12-12-2018 01:13 AM, Ronak wrote:
> 
> Looks like I found out why: 
> https://github.com/FFmpeg/FFmpeg/blob/master/libavformat/hlsenc.c#L790. 
> The hlsenc.c file overwrites whatever I pass in the command line...
 
 That is a bug but there's still an error in your command. It should be,
 
 
  -hls_ts_options movflags=+skip_sidx
 
 This is because mp4 is a child muxer in this context. That hls option 
 should be renamed.
 
 Line 790 should be patched by adding + before frag_custom
>>> 
>>> That doesn't work.
>>> 
>>> ffmpeg -i input.mp4 -hls_ts_options movflags=+skip_sidx -codec copy 
>>> -hls_time 0.97523809523809 -hls_segment_type fmp4 -hls_flags single_file 
>>> -hls_playlist_type vod output.m3u8
>>> ffmpeg version N-92677-gdd7d6034f1 Copyright (c) 2000-2018 the FFmpeg 
>>> developers
>>> built with gcc 4.4.6 (GCC) 20110731 (Red Hat 4.4.6-3)
>>> configuration: --prefix=/home/ronakp/ffmpeg_build 
>>> --pkg-config-flags=--static 
>>> --extra-cflags=-I/home/ronakp/ffmpeg_build/include 
>>> --extra-ldflags=-L/home/ronakp/ffmpeg_build/lib --extra-libs=-lpthread 
>>> --extra-libs=-lm --bindir=/home/ronakp/bin --enable-gpl --enable-libfdk_aac 
>>> --enable-libmp3lame --enable-libopus --enable-nonfree
>>> libavutil  56. 24.101 / 56. 24.101
>>> libavcodec 58. 42.100 / 58. 42.100
>>> libavformat58. 24.100 / 58. 24.100
>>> libavdevice58.  6.101 / 58.  6.101
>>> libavfilter 7. 46.101 /  7. 46.101
>>> libswscale  5.  4.100 /  5.  4.100
>>> libswresample   3.  4.100 /  3.  4.100
>>> libpostproc55.  4.100 / 55.  4.100
>>> Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'input.mp4':
>>> Metadata:
>>>  major_brand : isom
>>>  minor_version   : 1
>>>  compatible_brands: isom
>>>  creation_time   : 2013-11-14T18:23:26.00Z
>>> Duration: 02:39:09.39, start: 0.00, bitrate: 31 kb/s
>>>  Stream #0:0(und): Audio: aac (LC) (mp4a / 0x6134706D), 22050 Hz, mono, 
>>> fltp, 30 kb/s (default)
>>>  Metadata:
>>>creation_time   : 2013-11-14T18:23:26.00Z
>>>handler_name: GPAC ISO Audio Handler
>>> [hls @ 0x23e1b00] Opening 'output.m4s' for writing
>>> [hls @ 0x23e1b00] Some of provided format options in 'movflags=+skip_sidx' 
>>> are not recognized
>>> Could not write header for output file #0 (incorrect codec parameters ?): 
>>> Invalid argument
>>> Stream mapping:
>>> Stream #0:0 -> #0:0 (copy)
>>>  Last message repeated 1 times
>>> 
>>> Also, I think whatever options I set on the command line are going to be 
>>> overwritten by line 790 in hlsenc.c anyway no?
>> https://patchwork.ffmpeg.org/patch/11378/
>> 
>> Try this patch please :D
> 
> Thanks that works.
> 
> Let’s merge it!

Were you able to merge this patch?

> 
> 
>>> 
 
 Gyan
 
 ___
 ffmpeg-devel mailing list
 ffmpeg-devel@ffmpeg.org
 http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>> 
>>> ___
>>> ffmpeg-devel mailing list
>>> ffmpeg-devel@ffmpeg.org
>>> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>> 
>> 
>> 
>> ___
>> ffmpeg-devel mailing list
>> ffmpeg-devel@ffmpeg.org
>> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> ___
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

93 matches

Mail list logo