All of lore.kernel.org
 help / color / mirror / Atom feed
* Memory Compaction Question
@ 2021-10-12 18:35 Daniel Hoffman
  2021-10-13  7:45 ` Vlastimil Babka
  0 siblings, 1 reply; 2+ messages in thread
From: Daniel Hoffman @ 2021-10-12 18:35 UTC (permalink / raw)
  To: linux-mm

[-- Attachment #1: Type: text/plain, Size: 437 bytes --]

In the Linux Kconfig documentation for memory compaction, you mention
that almost nobody should turn the option off, and if so that you'd
like to hear about why at this email address.

Has anybody email'd you to explain why they'd need it turned off? I'm
doing R&D in this area and understanding these edge cases (if they
exist) would be great.

Thanks!

-- 
Daniel Hoffman
University of Illinois Urbana-Champaign
Mobile: (618)-980-5804

[-- Attachment #2: Screenshot from 2021-10-12 13-35-03.png --]
[-- Type: image/png, Size: 48069 bytes --]

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Memory Compaction Question
  2021-10-12 18:35 Memory Compaction Question Daniel Hoffman
@ 2021-10-13  7:45 ` Vlastimil Babka
  0 siblings, 0 replies; 2+ messages in thread
From: Vlastimil Babka @ 2021-10-13  7:45 UTC (permalink / raw)
  To: Daniel Hoffman, linux-mm

On 10/12/21 20:35, Daniel Hoffman wrote:
> In the Linux Kconfig documentation for memory compaction, you mention
> that almost nobody should turn the option off, and if so that you'd
> like to hear about why at this email address.
> 
> Has anybody email'd you to explain why they'd need it turned off? I'm

I don't personally recall about a mail triggered specifically by this
kconfig hint. There have been cases where people complained about high
overhead (e.g. CPU usage) which was sometimes due to bugs. A common non-bug
reason for high usage has been transparent hugepages (THP). The default
settings for THPs today should mean that compaction will not be tried as
hard in that context, but many people also learned from various old blogs
that they should be disabling THP altogether and keep doing it.

> doing R&D in this area and understanding these edge cases (if they
> exist) would be great.
> 
> Thanks!
> 
> 
> Screenshot from 2021-10-12 13-35-03.png
> 



^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-10-13  7:45 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-10-12 18:35 Memory Compaction Question Daniel Hoffman
2021-10-13  7:45 ` Vlastimil Babka

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.