![]() | |||||||||||
Messages in this thread |
> This patch adds blacklist of buggy chip, and if chip is not buggy, > this uses fast normal version instead of slow workaround version. I can confirm that this patch solves the problem I originally complained about. Timings of 10,000,000 gettimeofday calls on my pentium 4 3GHz (which I won at OLS, ha!): tsc --- real 0m2.504s user 0m0.700s sys 0m1.804s 2.6.16-mainline --------------- real 0m36.973s user 0m1.440s sys 0m34.130s 2.6.16-ogawa ------------ real 0m13.697s user 0m1.712s sys 0m11.117s For reference, baseline 2.6.16 on my Athlon64 3200+ --------------------------------------------------- real 0m1.994s user 0m0.990s sys 0m0.990s I'm still going to recommend all 'power' pdns_recursor users on single CPU and not using frequency scaling to boot with clock=tsc - saves 2.2usec/packet. Or run on an opteron of course. Thanks for the work everybody! -- http://www.PowerDNS.com Open source, database driven DNS Software http://netherlabs.nl Open and Closed source services - 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: 2006-03-25 13:02 [from the cache] ©2003-2008 | |||||||||||