[FFmpeg-devel] MPEG-2 Acceleration Refactor

[Greg Hulands]

Hi,

On 17/06/2007, at 4:13 AM, Michael Niedermayer wrote:
[...]
>
>>
>> As of why the new vs. old is slower, I think that assembler output
>> would give us more clues.
>
> yes
>
> and as long as svn + patch is slower this patch is rejected
> its also still a total mess whitespace wise

I have just finished analysing the assembler output. Apart from the  
offset values changing slightly and label names changing as well,  
there are only a few things that are different. I don't know whether  
they support me in speeding up seen when adding the timers or make it  
slower - you'll have to determine that. I have put the two assembly  
file up on the net so you can download and compare yourself (http:// 
idisk.mac.com/ghulands/Public/ffmpeg.tgz). This is what I found:

L1350: One less register load

LBB1493: Looks to have merged 2 labelled sections together

Line 20926 to 21570: .stabs and .stabn differences, but from what I  
can gather this is just for symbolic debuggers

Line 22909 to 23068: more .stabs and .stabn differences

Line 27777 to 28247: more .stabs and .stabn differences

If you need me to anything else, just let me know.

Cheers,
Greg.