[PATCH] i386: pte xchg optimization
authorZachary Amsden <zach@vmware.com>
Wed, 2 May 2007 17:27:19 +0000 (19:27 +0200)
committerAndi Kleen <andi@basil.nowhere.org>
Wed, 2 May 2007 17:27:19 +0000 (19:27 +0200)
commit142dd975911fdd82b1b6f6617cd20ac90a8ccf00
tree1e72e96bce64d9301d3305b4dd0b30675c8676b0
parentc2c1accd4b2f9c82fb89d40611c7f581948db255
[PATCH] i386: pte xchg optimization

In situations where page table updates need only be made locally, and there is
no cross-processor A/D bit races involved, we need not use the heavyweight
xchg instruction to atomically fetch and clear page table entries.  Instead,
we can just read and clear them directly.

This introduces a neat optimization for non-SMP kernels; drop the atomic xchg
operations from page table updates.

Thanks to Michel Lespinasse for noting this potential optimization.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andi Kleen <ak@suse.de>
include/asm-i386/pgtable-2level.h
include/asm-i386/pgtable-3level.h