[FFmpeg-devel] [RFC/PATCH] bitpacked_dec: Optimization for bitpacked_dec decoder performance

[Lance Wang]

On Sat, May 6, 2023 at 8:41 PM Devin Heitmueller <
devin.heitmueller at ltnglobal.com> wrote:

> On Sat, May 6, 2023 at 8:16 AM James Almer <jamrial at gmail.com> wrote:
> > Can you bench with the START_TIMER and STOP_TIMER macros in timer.h?
> > Also, define CACHED_BITSTREAM_READER in bitpacked_dec.c before including
> > git_bits.h and test the actual implementation again, to see if it makes
> > any difference.
>
> Original code:
> 671661910 decicycles in bitpacked_dec,       1 runs,      0 skips
> 669736380 decicycles in bitpacked_dec,       1 runs,      0 skips
> 669370700 decicycles in bitpacked_dec,       1 runs,      0 skips
>
> Original code with CACHED_BITSTREAM_READER defined
> 352599030 decicycles in bitpacked_dec,       1 runs,      0 skips
> 336163810 decicycles in bitpacked_dec,       1 runs,      0 skips
> 344628350 decicycles in bitpacked_dec,       1 runs,      0 skips
>
> My proposed versioned:
> 257353330 decicycles in bitpacked_dec,       1 runs,      0 skips
> 271527000 decicycles in bitpacked_dec,       1 runs,      0 skips
> 252701500 decicycles in bitpacked_dec,       1 runs,      0 skips
>
>
Yes, it's show better performance, so LGTM if nobody have plan to optimize
the bitstream
function.

Devin
>
> --
> Devin Heitmueller, Senior Software Engineer
> LTN Global Communications
> o: +1 (301) 363-1001
> w: https://ltnglobal.com  e: devin.heitmueller at ltnglobal.com
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>