On Thu, 2012-02-23 at 17:51 +0400, Konstantin Khlebnikov wrote:v3 changes:
* inactive-ratio reworked again, now it always calculated from from scratch
* hierarchical pte reference bits filter in memory-cgroup reclaimer
* fixed two bugs in locking, found by Hugh Dickins
* locking functions slightly simplified
* new patch for isolated pages accounting
* new patch with lru interleaving
This patchset is based on next-20120210
git: https://github.com/koct9i/linux/commits/lruvec-v3
---
I am seeing an improvement of about 7% in throughput in a workload where
I am doing parallel reading of files that are mmaped. The contention on
lru_lock used to be 13% in the cpu profile on the __pagevec_lru_add code
path. Now lock contention on this path drops to about 0.6%. I have 40
hyper-threaded enabled cpu cores running 80 mmaped file reading
processes.
So initial testing of this patch set looks encouraging.
Tim