Date: Sat, 30 Apr 2005 15:43:21 +0100 (BST) From: Robert Watson <rwatson@FreeBSD.org> To: current@FreeBSD.org Subject: Kernel memory allocator change (cvs commit: src/sys/vm uma_core.c uma_int.h (fwd) Message-ID: <20050430154103.L31768@fledge.watson.org>
next in thread | raw e-mail | index | archive | help
I've finally committed the UMA critical section per-cpu cache
synchronization change to the CVS HEAD. Two notes:
- This is intended to make things slightly faster. If you notice a
consistent slow-down, ideally characterizable with a highly
reproduceable benchmark, I'd like to hear about it. "Yes, it did get
faster" is also good, as is "it didn't get slower".
- We've done quite a bit of testing with these patches in place. However,
that doesn't preclude a stability proble. For those running regular
stability testing on HEAD, running with this change would be a good
thing.
Thanks,
Robert N M Watson
---------- Forwarded message ----------
Date: Fri, 29 Apr 2005 18:56:36 +0000 (UTC)
From: Robert Watson <rwatson@FreeBSD.org>
To: src-committers@FreeBSD.org, cvs-src@FreeBSD.org, cvs-all@FreeBSD.org
Subject: cvs commit: src/sys/vm uma_core.c uma_int.h
rwatson 2005-04-29 18:56:36 UTC
FreeBSD src repository
Modified files:
sys/vm uma_core.c uma_int.h
Log:
Modify UMA to use critical sections to protect per-CPU caches, rather than
mutexes, which offers lower overhead on both UP and SMP. When allocating
from or freeing to the per-cpu cache, without INVARIANTS enabled, we now
no longer perform any mutex operations, which offers a 1%-3% performance
improvement in a variety of micro-benchmarks. We rely on critical
sections to prevent (a) preemption resulting in reentrant access to UMA on
a single CPU, and (b) migration of the thread during access. In the event
we need to go back to the zone for a new bucket, we release the critical
section to acquire the global zone mutex, and must re-acquire the critical
section and re-evaluate which cache we are accessing in case migration has
occured, or circumstances have changed in the current cache.
Per-CPU cache statistics are now gathered lock-free by the sysctl, which
can result in small races in statistics reporting for caches.
Reviewed by: bmilekic, jeff (somewhat)
Tested by: rwatson, kris, gnn, scottl, mike at sentex dot net, others
Revision Changes Path
1.119 +120 -103 src/sys/vm/uma_core.c
1.30 +0 -10 src/sys/vm/uma_int.h
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050430154103.L31768>