Messages in this thread | | | Date | Tue, 14 Dec 2010 10:28:42 -0600 | From | Christoph Lameter <> | Subject | [cpuops cmpxchg V2 0/5] Cmpxchg and xchg operations |
| |
Add cmpxchg and xchg operations to the cpu ops and use them for irq handling and for vm statistics.
Requires the per cpu patches for and_return cpu ops.
V1->V2: - Determine that cmpxchg without LOCK is faster than xchg with implied lock (numbers are in the patch description). - Drop 64 bit cpuops operations on 32 bit x86 (resulted in various nasty things such as 64 bit alignment requirements for 64 bit entities on 32 bit, sizable inline code and very complex macros)
| |