[x264-devel] Re: Sparc optimization (8x8 SAD) & test sequences

[Mauricio Briceno]

Tuukka Toivonen wrote:
> On Sat, 16 Jul 2005, Phil Jensen wrote:
> 
> 
>>I've finished a first stab at an 8x8 SAD using VIS for a significant speedup on
>>UltraSPARC hardware (31% overall, ~400% improvement within the SAD).  I was
>>wondering if someone could kindly send me a link to a sample raw file I could
>>use when testing the CLI x264 encoder.  As soon as I can verify all is well,
> 
> 
> What about
>    dd if=/dev/urandom of=random-176x144.yuv bs=38016 count=100
> Encode it with and without VIS and see if the results are identical.
> Actually you need to modify the original x264 as well (just add a comment 
> somewhere) otherwise the bitstream headers differ by one byte.

Hi, what do you mean by that difference in the bitstream headers? Thanks.

> 
> Solaris probably doesn't have /dev/urandom, but I think you got the
> idea: just generate random data and cmp the results.
> 
> Here are some real test videos:
> http://media.xiph.org/video/derf/
> but you need to convert the y4m files to yuv with
> mencoder or something. Or Google for the yuv versions.
> 


-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html