On Sat, Nov 28, 2015 at 8:48 AM, Ronald S. Bultje wrote:
> Hi,
>
> On Sat, Nov 28, 2015 at 12:46 AM, Ganesh Ajjanagadde
> wrote:
>>
>> I personally consider this net ~ 10x and overall perf numbers sufficient
>> for using dynamic initialization all the
Le septidi 7 frimaire, an CCXXIV, Clement Boesch a écrit :
> "returned internally by demuxers" maybe
That sounds to me like an oxymoron. IMHO, being in internal.h and named
FFERROR instead of AVERROR should be enough (even separately).
> nit: space issue
Locally fixed.
> when they data?
On Fri, 2015-11-27 at 19:35 -0500, Ganesh Ajjanagadde wrote:
>
> @@ -40,6 +39,57 @@
> * Tables in this file are shared by the AAC decoders and encoder
> */
>
> +#define POW_SF2_ZERO200///< ff_aac_pow2sf_tab index
> corresponding to pow(2, 0);
Move POW_SF2_ZERO to aac.h, which is
On Sat, Nov 28, 2015 at 7:27 AM, Ganesh Ajjanagadde wrote:
> On Sat, Nov 28, 2015 at 4:41 AM, Michael Niedermayer wrote:
>> On Sat, Nov 28, 2015 at 12:46:31AM -0500, Ganesh Ajjanagadde wrote:
>>> This further speeds up runtime initialization, with identical
Hi,
On Sat, Nov 28, 2015 at 12:46 AM, Ganesh Ajjanagadde wrote:
> I personally consider this net ~ 10x and overall perf numbers sufficient
> for using dynamic initialization all the time here, especially since the
> tables are large.
This is a logical fallacy. By this
On Sat, Nov 28, 2015 at 3:38 AM, Rostislav Pehlivanov
wrote:
> On Fri, 2015-11-27 at 19:35 -0500, Ganesh Ajjanagadde wrote:
>>
>> @@ -40,6 +39,57 @@
>> * Tables in this file are shared by the AAC decoders and encoder
>> */
>>
>> +#define POW_SF2_ZERO200///<
On Sat, Nov 28, 2015 at 12:46:31AM -0500, Ganesh Ajjanagadde wrote:
> This further speeds up runtime initialization, with identical generated
> tables.
>
> Sample benchmark (x86-64, Haswell, GNU/Linux):
>
> old:
> 34441423 decicycles in mpegaudio_tableinit,8192 runs, 0 skips
>
> new:
On 22 November 2015 at 04:20, Matt Oliver wrote:
> On 22 November 2015 at 03:48, Ricardo Constantino
> wrote:
>
>> >+ --enable-gcrypt enable gmp, needed for rtmp(t)e support
>> >+ if openssl, librtmp or gmp is not used
On Sat, Nov 28, 2015 at 4:41 AM, Michael Niedermayer wrote:
> On Sat, Nov 28, 2015 at 12:46:31AM -0500, Ganesh Ajjanagadde wrote:
>> This further speeds up runtime initialization, with identical generated
>> tables.
>>
>> Sample benchmark (x86-64, Haswell, GNU/Linux):
>>
>>
On 11/28/15, Rick Kern wrote:
> Enable with configure --enable-vtenc and encode using -codec:v vtenc_h264.
>
> Signed-off-by: Rick Kern
> ---
> MAINTAINERS|1 +
> configure |4 +
> libavcodec/Makefile|1 +
>
On Sat, Nov 28, 2015 at 10:21 AM, Ganesh Ajjanagadde wrote:
> On Sat, Nov 28, 2015 at 10:06 AM, Rostislav Pehlivanov
> wrote:
>> On Sat, 2015-11-28 at 09:45 -0500, Ganesh Ajjanagadde wrote:
>>>
>>> Removed the av_cold, did not find a way to cleanly get rid
On Sat, Nov 28, 2015 at 10:06 AM, Rostislav Pehlivanov
wrote:
> On Sat, 2015-11-28 at 09:45 -0500, Ganesh Ajjanagadde wrote:
>>
>> Removed the av_cold, did not find a way to cleanly get rid of the
>> inline. I do suggest giving it some thought; it is needless
>> executable
>>
This further speeds up runtime initialization, with identical generated tables.
Sample benchmark (x86-64, Haswell, GNU/Linux):
old:
34441423 decicycles in mpegaudio_tableinit,8192 runs, 0 skips
new:
10776291 decicycles in mpegaudio_tableinit,8192 runs, 0 skips
Most low
This does some miscellaneous stuff mainly avoiding the usage of pow to
achieve significant speedups. This is not speed critical, but is
unnecessary latency and cycles wasted for a user.
All tables tested and are identical to the old ones
(bit-exact even in floating point case).
Sample benchmark
This is useful for build-time table generation (--enable-hardcoded-tables),
by providing compat shims for hosts that have broken libms.
This file is deliberately kept minimal; functions can always be added on
an as-needed basis.
Reviewed-by: Clément Bœsch
Reviewed-by: Ronald S.
Enable with configure --enable-vtenc and encode using -codec:v vtenc_h264.
Signed-off-by: Rick Kern
---
MAINTAINERS|1 +
configure |4 +
libavcodec/Makefile|1 +
libavcodec/allcodecs.c |1 +
libavcodec/vtenc.c | 1212
Fixed code review issues.
Rick Kern (1):
Added VideoToolbox H.264 encoder.
MAINTAINERS|1 +
configure |4 +
libavcodec/Makefile|1 +
libavcodec/allcodecs.c |1 +
libavcodec/vtenc.c | 1212
5
On Sat, 2015-11-28 at 09:45 -0500, Ganesh Ajjanagadde wrote:
>
> Removed the av_cold, did not find a way to cleanly get rid of the
> inline. I do suggest giving it some thought; it is needless
> executable
> bloat since the inline does not help perf. Pushed, thanks.
>
Hardly a blot as it's
On 11/28/2015 5:03 PM, Ganesh Ajjanagadde wrote:
> +static inline double ff_cbrt(double x)
> +{
> +return x < 0 ? -pow(-x, 1.0 / 3.0) : pow(x, 1.0 / 3.0);
> +}
> +#define cbrt ff_cbrt
Didn't Clément say to not pollute the global namespace like this?
- Derek
On systems having cbrt, there is no reason to use the slow pow function.
Sample benchmark (x86-64, Haswell, GNU/Linux):
new:
5124920 decicycles in cbrt_tableinit, 1 runs, 0 skips
old:
12321680 decicycles in cbrt_tableinit, 1 runs, 0 skips
Reviewed-by: Ronald S. Bultje
On Sat, Nov 28, 2015 at 7:50 AM, Ganesh Ajjanagadde wrote:
> On Sat, Nov 28, 2015 at 3:38 AM, Rostislav Pehlivanov
> wrote:
>> On Fri, 2015-11-27 at 19:35 -0500, Ganesh Ajjanagadde wrote:
>>>
>>> @@ -40,6 +39,57 @@
>>> * Tables in this file are shared by
On Sat, Nov 28, 2015 at 09:08:52AM +0100, Nicolas George wrote:
> Le septidi 7 frimaire, an CCXXIV, Clement Boesch a écrit :
> > "returned internally by demuxers" maybe
>
> That sounds to me like an oxymoron. IMHO, being in internal.h and named
> FFERROR instead of AVERROR should be enough (even
This is a patch set that significantly speeds up dynamic table initialization
(the default) on all platforms. The degree of speedup may vary; best results
are obtained on a fully C99 compliant libm. In particular, benchmarks refer to
the GNU/Linux platform.
Patches tested with FATE with and
On Sat, Nov 28, 2015 at 1:31 PM, Derek Buitenhuis
wrote:
> On 11/28/2015 5:03 PM, Ganesh Ajjanagadde wrote:
>> +static inline double ff_cbrt(double x)
>> +{
>> +return x < 0 ? -pow(-x, 1.0 / 3.0) : pow(x, 1.0 / 3.0);
>> +}
>> +#define cbrt ff_cbrt
>
> Didn't
On Sat, Nov 28, 2015 at 2:45 PM, Derek Buitenhuis
wrote:
> On 11/28/2015 7:33 PM, Ganesh Ajjanagadde wrote:
>> 2. Clement's idea AFAIK does not work, since the names must be
>> identical to the standard C names for the build to work with/without
>> hardcoded tables.
>
On 11/28/2015 7:51 PM, Ganesh Ajjanagadde wrote:
> In principle of course, e.g with more ifdefry, configure, or something
> of that sort. I do not believe this is what he meant.
I will await his reply to clarify, then. Perhaps do not write so
matter-of-factly.
> In any case, the point is moot -
On Sat, Nov 28, 2015 at 3:29 PM, Derek Buitenhuis
wrote:
> On 11/28/2015 5:03 PM, Ganesh Ajjanagadde wrote:
>> +if (i % 4 == 0)
>
> I doubt it particularly matters in this case, but you can
> avoid a modulo operator here with i & 3, like elsewhere
> in the
K coding style is implied but not listed in 'Coding Rules'.
Signed-off-by: Rick Kern
---
doc/developer.texi | 3 +++
1 file changed, 3 insertions(+)
diff --git a/doc/developer.texi b/doc/developer.texi
index cad1c29..9a901d8 100644
--- a/doc/developer.texi
+++
Enable with configure --enable-vtenc and encode using -codec:v vtenc_h264.
Signed-off-by: Rick Kern
---
MAINTAINERS|1 +
configure |4 +
libavcodec/Makefile|1 +
libavcodec/allcodecs.c |1 +
libavcodec/vtenc.c | 1269
vtenc.c was updated to match the coding style of the project.
I used several other files to get a feel for the project's coding style. Is
there documentation for this?
Rick Kern (1):
Added VideoToolbox H.264 encoder.
MAINTAINERS|1 +
configure |4 +
On 11/28/2015 5:03 PM, Ganesh Ajjanagadde wrote:
> +if (i % 4 == 0)
I doubt it particularly matters in this case, but you can
avoid a modulo operator here with i & 3, like elsewhere
in the file. I doubt the compiler is that dumb, though.
Same for other occurrences.
- Derek
On 11/28/2015 9:19 PM, Ganesh Ajjanagadde wrote:
> I actually prefer keeping all at i % 4, since that conveys intent
> better IMHO, and would change uniformly to that. It takes a reader a
> couple of seconds extra to register what i & 3 is doing for no gain in
> C code. No compiler I know of in
On Sat, Nov 28, 2015 at 02:06:07PM -0500, Rick Kern wrote:
> vtenc.c was updated to match the coding style of the project.
>
> I used several other files to get a feel for the project's coding
> style. Is there documentation for this?
Yes: https://ffmpeg.org/developer.html#Coding-Rules-1
On Sat, Nov 28, 2015 at 4:35 PM, Derek Buitenhuis
wrote:
> On 11/28/2015 9:26 PM, Derek Buitenhuis wrote:
>> Changing all of them sounds fine to me.
>
> [21:31] <@ubitux> re: i%4 vs i&3; you should ask him to make sure the
> compiler is smart enough regarding the
Signed-off-by: Paul B Mahol
---
doc/filters.texi | 57 ++
libavfilter/Makefile | 1 +
libavfilter/af_apulsator.c | 270 +
libavfilter/allfilters.c | 1 +
4 files changed, 329 insertions(+)
create mode
On Sat, Nov 28, 2015 at 5:26 PM, Paul B Mahol wrote:
> Signed-off-by: Paul B Mahol
[...]
> +case SQUARE:
> +val = (phs < 0.5) ? -1 : +1;
Something I missed: consider using e.g FFSIGN(phs - 0.5) or FFSIGN(0.5
- phase), can't recall. This is
On 11/28/2015 7:33 PM, Ganesh Ajjanagadde wrote:
> 2. Clement's idea AFAIK does not work, since the names must be
> identical to the standard C names for the build to work with/without
> hardcoded tables.
... yes it can. It doesn't work without more work, but it it sure can
work.
Anyway,
On Sat, Nov 28, 2015 at 3:19 PM, Derek Buitenhuis
wrote:
> On 11/28/2015 7:51 PM, Ganesh Ajjanagadde wrote:
>> In principle of course, e.g with more ifdefry, configure, or something
>> of that sort. I do not believe this is what he meant.
>
> I will await his reply to
On 11/28/2015 9:26 PM, Derek Buitenhuis wrote:
> Changing all of them sounds fine to me.
[21:31] <@ubitux> re: i%4 vs i&3; you should ask him to make sure the compiler
is smart enough regarding the signess of i
- Derek
___
ffmpeg-devel mailing list
On Nov 28, 2015, at 5:07 PM, Timothy Gu wrote:
>
> On Sat, Nov 28, 2015 at 02:06:07PM -0500, Rick Kern wrote:
>> vtenc.c was updated to match the coding style of the project.
>>
>> I used several other files to get a feel for the project's coding
>> style. Is there
On Sat, Nov 28, 2015 at 5:26 PM, Paul B Mahol wrote:
> Signed-off-by: Paul B Mahol
> ---
> doc/filters.texi | 57 ++
> libavfilter/Makefile | 1 +
> libavfilter/af_apulsator.c | 270
> +
>
On Sat, Nov 28, 2015 at 6:21 PM, Ganesh Ajjanagadde wrote:
> On Sat, Nov 28, 2015 at 5:26 PM, Paul B Mahol wrote:
>> Signed-off-by: Paul B Mahol
> [...]
>
>> +case SQUARE:
>> +val = (phs < 0.5) ? -1 : +1;
>
> Something I missed:
42 matches
Mail list logo