[xine-devel] Re: memcpy.c speedup patch

[Nick K. <nic...@ma...>]

Hello, Jonathan!

On Thu, 09 Jan 2003 07:34:39 +0000 you wrote:

> I found the same mistake in both sse_memcpy and mmx2_memcpy. They both 
> presume that prefetchnta prefetches 64 bytes. In actual fact, the p3 
> prefetches 32 bytes and the p4 prefetches 128 bytes. The patch optimizes 
> it correctly for the p3. If you want to optimize for the p4 you should 
> really use movdqa/movdqu.
> 
> Please apply to the tree.
> 
> CC any replies to me as I am not on the list.
> 
Thanks for the patch!
I've improved your patch (see mplayerxp's CVSlog) to perform run-time
cache-line size detection. What about movdqa/movdqu - these insns perform
serialized memory storing (means - they can't perform unordeder memory writing
like movntps, movntpd, mopnti insns so they are not the best solution for
fastest memory writing.

WBR! Nick