Skip to content

Various XHA improvements#18

Draft
edwintorok wants to merge 2 commits intoxenserver:masterfrom
edwintorok:private/edvint/lockinglatency
Draft

Various XHA improvements#18
edwintorok wants to merge 2 commits intoxenserver:masterfrom
edwintorok:private/edvint/lockinglatency

Conversation

@edwintorok
Copy link
Contributor

Might need splitting into separate PRs (there are about 8 independent fixes here). For now keep them together until I test them.

Copy link

@andyhhp andyhhp left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think you do want to split the mechanical fixes from the logical fixes (esp the lock contention one).

@edwintorok edwintorok force-pushed the private/edvint/lockinglatency branch 5 times, most recently from b96eb51 to c2098f5 Compare February 19, 2025 10:24
@edwintorok edwintorok force-pushed the private/edvint/lockinglatency branch from c2098f5 to 446f61e Compare February 21, 2025 19:01
@edwintorok
Copy link
Contributor Author

edwintorok commented Feb 21, 2025

I think you do want to split the mechanical fixes from the logical fixes (esp the lock contention one).

Split into:

@edwintorok edwintorok force-pushed the private/edvint/lockinglatency branch from 446f61e to 447a64b Compare March 10, 2026 18:28
Reduce the scope of the global com mutex, otherwise every attempt to
acquire a reader lock results in contention on this global mutex.

Use a per-object-table lock instead.

Signed-off-by: Edwin Török <edwin.torok@cloud.com>
@edwintorok edwintorok force-pushed the private/edvint/lockinglatency branch from 447a64b to efa7b6f Compare March 10, 2026 18:34
TODO: use pthread_condattr_setclock and switch to a monotonic clock
instead

Signed-off-by: Edwin Török <edwin.torok@cloud.com>
@edwintorok edwintorok force-pushed the private/edvint/lockinglatency branch from efa7b6f to c5777e7 Compare March 10, 2026 18:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants