[FFmpeg-devel] [PATCH 2/3] lavu/internal: add avpriv_exp10

[Ganesh Ajjanagadde]

On Fri, Dec 25, 2015 at 9:47 AM, Hendrik Leppkes <h.leppkes at gmail.com> wrote:
> On Fri, Dec 25, 2015 at 6:27 PM, Ganesh Ajjanagadde <gajjanag at mit.edu> wrote:
>> On Fri, Dec 25, 2015 at 9:26 AM, James Almer <jamrial at gmail.com> wrote:
>>> On 12/25/2015 2:11 PM, Ganesh Ajjanagadde wrote:
>>>> Fast, reasonably accurate 10^x. Alternative of detection of libm exp10 at configure
>>>> time is not worth the trouble, since it is anyway not POSIX or ISO C,
>>>> and currently only the GNU libm has it. Furthermore, GNU libm's variant
>>>> is ~ 2x slower, and is ironically not correctly rounded (2 ulp off) to justify all
>>>> that slowdown.
>>>>
>>>> Reviewed-by: Michael Niedermayer <michael at niedermayer.cc>
>>>> Signed-off-by: Ganesh Ajjanagadde <gajjanagadde at gmail.com>
>>>> ---
>>>>  libavutil/internal.h | 19 +++++++++++++++++++
>>>>  1 file changed, 19 insertions(+)
>>>>
>>>> diff --git a/libavutil/internal.h b/libavutil/internal.h
>>>> index f86b7fb..ae11601 100644
>>>> --- a/libavutil/internal.h
>>>> +++ b/libavutil/internal.h
>>>> @@ -292,6 +292,25 @@ static av_always_inline av_const int64_t ff_rint64_clip(double a, int64_t amin,
>>>>      return res;
>>>>  }
>>>>
>>>> +/**
>>>> + * Compute 10^x for floating point values. Note: this function is by no means
>>>> + * "correctly rounded", and is meant as a fast, reasonably accurate approximation.
>>>> + * For instance, maximum relative error for the double precision variant is
>>>> + * ~ 1e-13 for very small and very large values.
>>>> + * This is ~2x faster than GNU libm's approach, which is still off by 2ulp on
>>>> + * some inputs.
>>>> + * @param x exponent
>>>> + * @return 10^x
>>>> + */
>>>> +static av_always_inline double avpriv_exp10(double x)
>>>
>>> It's an inline function in a header, and internal at that. Just call it ff_exp10.
>>
>> It is used in avcodec and avfilter. I thought this meant that it
>> should be avpriv?
>> Personally, I like ff_exp10 as it is shorter.
>>
>
> avpriv is a hack for private functions that need to be exported in
> shared libraries, since we have a rule to only export av* functions.

Makes sense now.

> inline functions in headers don't need to be exported in a library, so
> no need for avpriv. ff* is fine there, and used for other similar
> situations.

Guess the patch is ok (once this is changed), and likewise the
configure one. Both of these going in soon. Thanks.

>
> - Hendrik
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel