RE: [PATCH 1/1] mm/swap.c: flush lru_add pvecs on compound page arrival

[Odzioba, Lukasz]

On 09-06-16 17:42:00, Dave Hansen wrote:
> Does your workload put large pages in and out of those pvecs, though?
> If your system doesn't have any activity, then all we've shown is that
> they're not a problem when not in use.  But what about when we use them?

It doesn't. To use them extensively I guess we would have to
craft a separate program for each one, which is not trivial.

> Have you, for instance, tried this on a system with memory pressure?

Not then, but here are exemplary snapshots with system using swap to handle 
allocation requests with patch applied: (notation: pages = sum in bytes):
LRU_add              336 =     1344kB
LRU_rotate           158 =      632kB
LRU_deactivate         0 =        0kB
LRU_deact_file         0 =        0kB
LRU_activate           1 =        4kB
---
LRU_add             3262 =    13048kB
LRU_rotate           142 =      568kB
LRU_deactivate         0 =        0kB
LRU_deact_file         0 =        0kB
LRU_activate           6 =       24kB
---
LRU_add             3689 =    14756kB
LRU_rotate            81 =      324kB
LRU_deactivate         0 =        0kB
LRU_deact_file         0 =        0kB
LRU_activate          19 =       76kB

While running idle os we have:
LRU_add             1038 =     4152kB
LRU_rotate             0 =        0kB
LRU_deactivate         0 =        0kB
LRU_deact_file         0 =        0kB
LRU_activate           0 =        0kB

I know those are not representative in overall.

Thanks,
Lukas