Messages in this thread | | | Subject | Re: [PATCH v5]: Improve performance of LZO hibernation | From | Bojan Smojver <> | Date | Wed, 28 Sep 2011 17:57:29 +1000 |
| |
On Wed, 2011-09-28 at 10:48 +0300, Pekka Enberg wrote: > > We want to keep at least one CPU free for that I/O and for pulling > the > > other threads into sync when they are done (that is if we have more > than > > one), right? > > Well, dunno if it matters much. Did you see performance improvement > with that?
Haven't tried, to be honest. Just thought it may make sense.
> Is the CPU binding really needed?
Don't really know, but I would think it would help with compression/decompression code. We don't want these threads bouncing between CPUs unless they have to. I would guess the caches would work better that way and all that.
Again, just guessing.
> Anyway, if you want to keep the existing behavior, maybe something > like > > nr_other_cpus = min(1, num_online_cpus()-1); > > nr_threads = min(nr_other_cpus, LZO_THREADS); > > would do the trick?
Yeah, makes sense. The first one should be max() though.
-- Bojan
| |