[FFmpeg-devel] [PATCH] Fix unaligned access in lcl decoder

Måns Rullgård mans
Wed May 27 23:03:23 CEST 2009


Jeff Downs <heydowns at borg.com> writes:

> On Wed, 27 May 2009, Michael Niedermayer wrote:
>
>> On Wed, May 27, 2009 at 12:22:19PM -0400, Jeff Downs wrote:
>> > Attached fixes by using memcpy instead of integer copies on platforms 
>> > without fast unaligned access.
>> 
>> AV_R/WN32() might be more appropriate than the ifdeffery
>> 
>
> Much cleaner looking, yes. Attached.
>
> Probably a slight speed hit on platforms where AV_R/WN32 are non-optimized 
> though due to all the shifts. I don't have a platform where that is 
> realistically the case to test it on,

The shifts are the last fallback after arch-specific code and compiler
tricks, and then only when fast_unaligned is off.

> but if I force that implementation it is slower than memcpy (2% with
> the test sample on sparc).

I thought it was crashing on Sparc.  Please explain.

-- 
M?ns Rullg?rd
mans at mansr.com



More information about the ffmpeg-devel mailing list