[x264-devel] Re: [PATCH] Altivec optimizations for quant4x4, quant4x4dc, quant8x8, sub8x8_dct8, sub16x16_dct8, pixel_sa8d_8x8, pixel_sa8d_16x16

[Guillaume POIRIER]

Hi,

Guillaume POIRIER a écrit :

[...]
> Next step port FFmpeg's START/STOP_TIMER macros to be able to use them 
> with PMCs. As noted in the doc 
> ffmpeg_powerpc_performance_evaluation_howto.txt, Time Base Registers 
> (TBL), which is somewhat the equivalent of x86's rdtsc register, has 
> the disadvantge that it can't provide an accurate measurement: the 
> registers increment by one every four *bus* cycles. That's too rough 
> for precise measurement.

I attached 2 files today.
First and foremost, is the 12th revision of my patch. It features PMC 
support and an example of how to use them to benchmark in pixel.c (which 
should be stripped off if committed).
Nothing new except that.


The 2nd patch is just the header file that adds the START/STOP_TIMER 
macros for PPC (it's also present in the whole patchset).
It's meant to be put in common/ppc.
I'm putting it aside to ease reviewing and because it's really a 
self-contained "patch".

The question that probably needs to be asked now is: do you guys care to 
have
the START/STOP_TIMER also available on other arches, or just for PPC is 
fine?
Also note that I licensed this file under LGPL as it's essentially 
several pieces of code taken from FFmpeg and melted..

Guilaume
-------------- next part --------------
A non-text attachment was scrubbed...
Name: performance_mesure.h
Type: text/x-chdr
Size: 4306 bytes
Desc: not available
Url : http://mailman.videolan.org/pipermail/x264-devel/attachments/20060910/5b76f413/attachment.h 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Altivec_quant+dct+pixel_routines+PMC_12.diff
Type: text/x-patch
Size: 47841 bytes
Desc: not available
Url : http://mailman.videolan.org/pipermail/x264-devel/attachments/20060910/5b76f413/attachment.bin