Triggering huge-page collapse from user space

Posted Mar 15, 2022 5:49 UTC (Tue) by Nikratio (subscriber, #71966)
Parent article: Triggering huge-page collapse from user space

"Processes do not always need large, virtually contiguous memory ranges," - that should be "physically", not "virtually" I think?

Triggering huge-page collapse from user space

Posted Mar 15, 2022 7:01 UTC (Tue) by edeloget (subscriber, #88392) [Link] (4 responses)

Both are true. Some programs are perfectly happy with a lot of small, non-contiguous virtual memory ranges because they never do large allocations.

Triggering huge-page collapse from user space

Posted Mar 16, 2022 0:44 UTC (Wed) by mtaht (subscriber, #11087) [Link] (3 responses)

aggregate, rather than "collapse" might be a better word.

Triggering huge-page collapse from user space

Posted Mar 18, 2022 1:17 UTC (Fri) by ncm (guest, #165) [Link]

The most apt English word is "consolidate".

(That did not stop business reporters from preferring "conglomerate", some decades back.)

Triggering huge-page collapse from user space

Posted Mar 18, 2022 13:34 UTC (Fri) by pitb0ss (subscriber, #137324) [Link] (1 responses)

Perhaps coalesce is another good word

Triggering huge-page collapse from user space

Posted Mar 18, 2022 15:08 UTC (Fri) by zokeefe (subscriber, #140292) [Link]

I thought about "coalesce", but "collapse" is just what the operation is ubiquitously referred to in the kernel and existing apis (e.g. /sys/kernel/mm/transparent_hugepage/khugepaged/pages_collapsed) and stats (e.g. /proc/vmstat:thp_collapse_alloc[_failed]).