perf(profiling): separate queue and waking thread #3611
+91
−28
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
On every sample we send data to another thread and attempt to wake that thread. The problem is the syscall here is almost as expensive as collecting the sample if the other thread is asleep, which is often the case. This branch avoids that. Instead it writes to a queue and samples are handled when that background thread wakes up for other reasons. Notably one of those reasons is every 10ms when wall-time is enabled (default) so _probably_ this should work pretty well without filling the queues.
|
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #3611 +/- ##
==========================================
- Coverage 62.20% 62.12% -0.08%
==========================================
Files 141 141
Lines 13387 13387
Branches 1753 1753
==========================================
- Hits 8327 8317 -10
- Misses 4263 4270 +7
- Partials 797 800 +3 see 3 files with indirect coverage changes Continue to review full report in Codecov by Sentry.
🚀 New features to boost your workflow:
|
Under sustained load in ZTS, this could otherwise run forever.
Benchmarks [ profiler ]Benchmark execution time: 2026-01-31 18:02:19 Comparing candidate commit b622e0f in PR branch Found 9 performance improvements and 6 performance regressions! Performance is the same for 14 metrics, 7 unstable metrics. scenario:php-profiler-exceptions-with-profiler
scenario:php-profiler-exceptions-with-profiler-and-timeline
scenario:php-profiler-timeline-memory-with-profiler
scenario:php-profiler-timeline-memory-with-profiler-and-timeline
scenario:walk_stack/1
scenario:walk_stack/50
scenario:walk_stack/99
scenario:walk_stack_instructions/1
scenario:walk_stack_instructions/50
scenario:walk_stack_instructions/99
|
morrisonlevi
force-pushed
the
levi/messages
branch
from
b5a4330 to
07df26e
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
On master, we send every sample to another thread and attempt to wake that thread to process the sample. The problem is the syscall here is almost as expensive as collecting the sample if the other thread is asleep, which is often the case in NTS builds. Here's a profile showing this (but note it's not always this close):
This git branch avoids that. Instead it writes to a queue and samples are handled when that background thread wakes up for other reasons. Notably one of those reasons is every 10ms when wall-time is enabled (default) so probably this should work pretty well without filling the queues. The queues are larger on ZTS builds, though by a fixed multiplier (16).
Reviewer checklist