Messages in this thread | | | Date | Mon, 6 Mar 2006 11:51:35 -0800 | From | Mark Fasheh <> | Subject | Re: [Ocfs2-devel] Ocfs2 performance bugs of doom |
| |
On Sun, Mar 05, 2006 at 08:59:50PM -0800, Daniel Phillips wrote: > So that hack apparently improves the bucket distribution quite a bit, but > look, the bad old linear systime creep is still very obvious. For that > there is no substitute for lots of buckets. Yes, I think the way to go right now is to allocate an array of pages and index into that. We can make the array size a mount option so that the default can be something reasonable ;)
> Speaking of that, the qstr is so far away from the list.next pointer that > you're most probably getting an extra cache line load there too. Just move > the qstr up next to the hash list. Ahh, good catch.
> Notice, those wildly gyrating real times are still manifesting. Seen > them over there? Yes - times from a test run are below.
Real User Sys 14.35 3.82 11.99 16.75 3.62 13.90 32.12 4.06 15.81 18.94 3.99 16.69 35.18 3.96 15.80 18.66 3.89 16.44 52.09 4.24 16.66 21.25 3.85 17.09 19.73 4.00 17.20 35.40 3.88 16.60
So this definitely reproduces elsewhere. --Mark
-- Mark Fasheh Senior Software Developer, Oracle mark.fasheh@oracle.com - 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/
| |