Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-08-21 Thread Michal Hocko
On Thu 20-08-15 14:42:40, Mel Gorman wrote: > On Thu, Aug 20, 2015 at 03:18:43PM +0200, Michal Hocko wrote: > > On Wed 12-08-15 11:45:26, Mel Gorman wrote: > > [...] > > > 4-node machine stutter > > > 4-node machine stutter > > > 4.2.0-rc1 4.2.0-rc1 > > >

Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-08-20 Thread Vlastimil Babka
On 08/20/2015 04:17 PM, Mel Gorman wrote: On Thu, Aug 20, 2015 at 03:30:54PM +0200, Vlastimil Babka wrote: These hunks appear unrelated to zonelist cache? Also they move the evaluation of cpuset_current_mems_allowed Ah forgot to delete the "Also" part. I wanted to write that it moves the eval

Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-08-20 Thread Mel Gorman
On Thu, Aug 20, 2015 at 03:30:54PM +0200, Vlastimil Babka wrote: > >Note the maximum stall latency which was 6 seconds and becomes 67ms with > >this patch applied. However, also note that it is not guaranteed this > >benchmark always hits pathelogical cases and the milage varies. There is > >a seco

Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-08-20 Thread Mel Gorman
On Thu, Aug 20, 2015 at 03:18:43PM +0200, Michal Hocko wrote: > On Wed 12-08-15 11:45:26, Mel Gorman wrote: > [...] > > 4-node machine stutter > > 4-node machine stutter > > 4.2.0-rc1 4.2.0-rc1 > >vanilla nozlc-v1r20

Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-08-20 Thread Vlastimil Babka
On 08/12/2015 12:45 PM, Mel Gorman wrote: The zonelist cache (zlc) was introduced to skip over zones that were recently known to be full. This avoided expensive operations such as the cpuset checks, watermark calculations and zone_reclaim. The situation today is different and the complexity of zl

Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-08-20 Thread Michal Hocko
On Wed 12-08-15 11:45:26, Mel Gorman wrote: [...] > 4-node machine stutter > 4-node machine stutter > 4.2.0-rc1 4.2.0-rc1 >vanilla nozlc-v1r20 > Min mmap 53.9902 ( 0.00%) 49.3629 ( 8.57%) > 1st-qrtl

[PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-08-12 Thread Mel Gorman
The zonelist cache (zlc) was introduced to skip over zones that were recently known to be full. This avoided expensive operations such as the cpuset checks, watermark calculations and zone_reclaim. The situation today is different and the complexity of zlc is harder to justify. 1) The cpuset check

Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-07-23 Thread Mel Gorman
On Tue, Jul 21, 2015 at 04:47:35PM -0700, David Rientjes wrote: > On Mon, 20 Jul 2015, Mel Gorman wrote: > > > From: Mel Gorman > > > > The zonelist cache (zlc) was introduced to skip over zones that were > > recently known to be full. At the time the paths it bypassed were the > > cpuset checks

Re: [PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-07-21 Thread David Rientjes
On Mon, 20 Jul 2015, Mel Gorman wrote: > From: Mel Gorman > > The zonelist cache (zlc) was introduced to skip over zones that were > recently known to be full. At the time the paths it bypassed were the > cpuset checks, the watermark calculations and zone_reclaim. The situation > today is differ

[PATCH 01/10] mm, page_alloc: Delete the zonelist_cache

2015-07-20 Thread Mel Gorman
From: Mel Gorman The zonelist cache (zlc) was introduced to skip over zones that were recently known to be full. At the time the paths it bypassed were the cpuset checks, the watermark calculations and zone_reclaim. The situation today is different and the complexity of zlc is harder to justify.