On Thu, Oct 30, 2008 at 06:01:26AM +0000, Ganesh Sittampalam wrote: > On Wed, 29 Oct 2008, Jason Dagit wrote: > >Have you retimed things with the full set of patches you submitted? > >Do you know what the overall improvement would be? > > Nope - without some really good "fire and forget" infrastructure and a > dedicated machine that can be guaranteed quiescent, benchmarking is quite > fiddly and time-consuming, so I've only been doing it for things where it > seemed particularly warranted.
I've just reviewed this one, and it looks correct, but I couldn't predict whether its performance behavior. So I'd rather not apply it, unless either you can explain it to me in such a way that I can understand the improvement is, or you have benchmarks demonstrating the improvement. I can see that you replace (+>+) with (:>:) using some clever tricks (which is definitely always a bonus), but that only affects the scaling when many, many changes are made to a single file, in which case this is almost certainly not a bottleneck (since we're diffing said file, which is a slow operation). The other change (and I think these two changes are separable?) is a switch from foldl' to foldr, and I must admit that these fold functions almost always confuse the heck out of me... -- David Roundy http://www.darcs.net _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
