[FFmpeg-devel] avcodec/prores_ks reduce twice fdct calls

Submitted by Michael Niedermayer on Dec. 31, 2018, 2:46 a.m.

Details

Message ID 20181231024657.GP3501@michaelspb
State New
Headers show

Commit Message

Michael Niedermayer Dec. 31, 2018, 2:46 a.m.
On Sun, Dec 30, 2018 at 10:57:23PM +0200, Alex Mogurenko wrote:
> fdct done twice for each block. first time during quant calculation, second during slice encoding. so if we pre-save dct coefficients no need to do fdct second time.
> disadvantages: requires more memory
> advantages: improves performance ~4-5%
> ---
>  libavcodec/proresenc_kostya.c | 74 ++++++++++++++++++++++++-----------
>  1 file changed, 52 insertions(+), 22 deletions(-)

breaks fate

TEST    vsynth1-prores_ks
Test vsynth1-prores_ks failed. Look at tests/data/fate/vsynth1-prores_ks.err for details.
make: *** [fate-vsynth1-prores_ks] Error 1

[...]

Comments

Alex Mogurenko Dec. 31, 2018, 9:12 a.m.
I seems to be lame as failed to find how to run fate to check prores_ks :(

On Mon, Dec 31, 2018 at 4:46 AM Michael Niedermayer <michael@niedermayer.cc>
wrote:

> On Sun, Dec 30, 2018 at 10:57:23PM +0200, Alex Mogurenko wrote:
> > fdct done twice for each block. first time during quant calculation,
> second during slice encoding. so if we pre-save dct coefficients no need to
> do fdct second time.
> > disadvantages: requires more memory
> > advantages: improves performance ~4-5%
> > ---
> >  libavcodec/proresenc_kostya.c | 74 ++++++++++++++++++++++++-----------
> >  1 file changed, 52 insertions(+), 22 deletions(-)
>
> breaks fate
>
> TEST    vsynth1-prores_ks
> --- ./tests/ref/vsynth/vsynth1-prores_ks        2018-12-28
> 17:54:41.361383177 +0100
> +++ tests/data/fate/vsynth1-prores_ks   2018-12-31 03:45:17.207673799 +0100
> @@ -1,4 +1,4 @@
> -fe41a284da97ea5ec8866ca9a55b84da *tests/data/fate/vsynth1-prores_ks.mov
> -3858911 tests/data/fate/vsynth1-prores_ks.mov
> -100eb002413fe7a632d440dfbdf7e3ff
> *tests/data/fate/vsynth1-prores_ks.out.rawvideo
> -stddev:    3.17 PSNR: 38.09 MAXDIFF:   39 bytes:  7603200/  7603200
> +ba6294d95b96f032b90f804f112ab98a *tests/data/fate/vsynth1-prores_ks.mov
> +3867422 tests/data/fate/vsynth1-prores_ks.mov
> +e85510eadb1ff115e85c480d8e1011a4
> *tests/data/fate/vsynth1-prores_ks.out.rawvideo
> +stddev:   19.38 PSNR: 22.38 MAXDIFF:  210 bytes:  7603200/  7603200
> Test vsynth1-prores_ks failed. Look at
> tests/data/fate/vsynth1-prores_ks.err for details.
> make: *** [fate-vsynth1-prores_ks] Error 1
>
> [...]
> --
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> "You are 36 times more likely to die in a bathtub than at the hands of a
> terrorist. Also, you are 2.5 times more likely to become a president and
> 2 times more likely to become an astronaut, than to die in a terrorist
> attack." -- Thoughty2
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
Derek Buitenhuis Dec. 31, 2018, 3:56 p.m.
On 31/12/2018 09:12, Alex Mogurenko wrote:
> I seems to be lame as failed to find how to run fate to check prores_ks :(

General way to run all of FATE:

$ configure --samples=/home/path/for/samples/ --enable-gpl [...]
$ make fate-rsync
$ make fate

There are sub-targets, too, as listed in Michaels mail, like fate-vsynth1.

- Derek
Alex Mogurenko Dec. 31, 2018, 9:02 p.m.
thanks,
just sent new patch

On Mon, Dec 31, 2018 at 5:56 PM Derek Buitenhuis <derek.buitenhuis@gmail.com>
wrote:

> On 31/12/2018 09:12, Alex Mogurenko wrote:
> > I seems to be lame as failed to find how to run fate to check prores_ks
> :(
>
> General way to run all of FATE:
>
> $ configure --samples=/home/path/for/samples/ --enable-gpl [...]
> $ make fate-rsync
> $ make fate
>
> There are sub-targets, too, as listed in Michaels mail, like fate-vsynth1.
>
> - Derek
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
Alex Mogurenko Jan. 4, 2019, 9:24 a.m.
ping

On Mon, Dec 31, 2018 at 11:02 PM Alex Mogurenko <alex@mogurenko.com> wrote:

> thanks,
> just sent new patch
>
> On Mon, Dec 31, 2018 at 5:56 PM Derek Buitenhuis <
> derek.buitenhuis@gmail.com> wrote:
>
>> On 31/12/2018 09:12, Alex Mogurenko wrote:
>> > I seems to be lame as failed to find how to run fate to check prores_ks
>> :(
>>
>> General way to run all of FATE:
>>
>> $ configure --samples=/home/path/for/samples/ --enable-gpl [...]
>> $ make fate-rsync
>> $ make fate
>>
>> There are sub-targets, too, as listed in Michaels mail, like fate-vsynth1.
>>
>> - Derek
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel@ffmpeg.org
>> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>
>

Patch hide | download patch | download mbox

--- ./tests/ref/vsynth/vsynth1-prores_ks	2018-12-28 17:54:41.361383177 +0100
+++ tests/data/fate/vsynth1-prores_ks	2018-12-31 03:45:17.207673799 +0100
@@ -1,4 +1,4 @@ 
-fe41a284da97ea5ec8866ca9a55b84da *tests/data/fate/vsynth1-prores_ks.mov
-3858911 tests/data/fate/vsynth1-prores_ks.mov
-100eb002413fe7a632d440dfbdf7e3ff *tests/data/fate/vsynth1-prores_ks.out.rawvideo
-stddev:    3.17 PSNR: 38.09 MAXDIFF:   39 bytes:  7603200/  7603200
+ba6294d95b96f032b90f804f112ab98a *tests/data/fate/vsynth1-prores_ks.mov
+3867422 tests/data/fate/vsynth1-prores_ks.mov
+e85510eadb1ff115e85c480d8e1011a4 *tests/data/fate/vsynth1-prores_ks.out.rawvideo
+stddev:   19.38 PSNR: 22.38 MAXDIFF:  210 bytes:  7603200/  7603200