Optimise decommitting memory from the spare chunk
Categories
(Core :: Memory Allocator, enhancement)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox145 | --- | fixed |
People
(Reporter: pbone, Assigned: pbone)
References
(Blocks 1 open bug)
Details
(Whiteboard: [sp3])
Attachments
(4 files)
In mozjemalloc when we decommit memory it may be better to decommit from the spare chunk first - since that's typically the last used memory. Also we can possibly commit multiple non-contiguous runs of pages with a single system call.
| Assignee | ||
Comment 1•1 year ago
|
||
| Assignee | ||
Comment 2•1 year ago
|
||
Decommit non-contigious runs of dirty pages when other free pages separate
them. For example if there are pages:
__DDDFFDFDD__
D - Dirty
F - Free
_ - other/used.
Decommit all the dirty and free pages above with a single system call.
| Assignee | ||
Comment 3•1 year ago
|
||
Updated•1 year ago
|
Updated•1 year ago
|
| Assignee | ||
Comment 4•1 year ago
|
||
This improves "pages returned per system call" from 11.39 to 15.78, the median doesn't move as much so we can say that it's catching more outliers, but when it does it is helpful. The median time for a purge improves but I can't tell if it's significant.
| Assignee | ||
Comment 5•1 year ago
|
||
Purge stats now show the number of purged and dirty pages separately.
Updated•1 year ago
|
Updated•1 year ago
|
Updated•1 year ago
|
Updated•1 year ago
|
| Assignee | ||
Comment 6•11 months ago
|
||
https://hg.mozilla.org/mozilla-central/rev/d98190178515
https://hg.mozilla.org/mozilla-central/rev/0936d18965f4
https://hg.mozilla.org/mozilla-central/rev/6691b36a60f7
https://hg.mozilla.org/mozilla-central/rev/a3c6826a7f8c
Updated•10 months ago
|
Description
•