Messages in this thread | | | Date | Wed, 17 Dec 2014 10:52:02 -0800 | From | Dave Hansen <> | Subject | Re: post-3.18 performance regression in TLB flushing code |
| |
On 12/17/2014 08:53 AM, Will Deacon wrote: > On Wed, Dec 17, 2014 at 04:28:23PM +0000, Linus Torvalds wrote: >> On Wed, Dec 17, 2014 at 2:08 AM, Will Deacon <will.deacon@arm.com> wrote: >> So why not just this trivial patch, to make the logic be the same it >> used to be (just using "end > 0" instead of the old "need_flush")? > > Looks fine to me... Dave?
First of all, this is quite observable when testing single-threaded on a desktop. This is a mildly crusty Sandybridge CPU from 2011. I made 3 runs with a single thread: ./brk1_processes -s 30 -t 1
fb7332a9fed : 4323385 fb7332a9fed^: 4503736 fb7332a9fed+Linus's fix: 4516761
These things are also a little bit noisy, so we're well within the margin of error with Linus's fix.
This also holds up on the large system.
| |