![]() | ||||||||||
Messages in this thread |
On December 31, 2001 11:47, Dave Jones wrote: > Completly puzzled right now. Moving the prefetching to copy_to_user > (and doing the tlb preload & prefetching the whole chunk to be copied > (or cachesize if smaller)) results in a performance drop instead of a win. > > My initial guess is that some of the callers of copy_to_user are > doing something that is harmed the prefetching. > (Maybe they are doing additional prefetch() calls) Maybe syscalls that only have to move a very small chunk of data (gettimeofday(2), for instance), are hurt because of the wasted bytes they are prefetching after the intended data? Also, for sizes greater than 512, copy_to_user will call mmx_copy_user, which might call mmx_memcpy, which does prefetching already on x86 CPUs that support it. -Ryan - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ | |||||||||
| Last update: 2005-03-22 12:18 [from the cache] ©2003-2008 | ||||||||||