[PROF-10124] Lower value used to clamp very high allocation profiling weights #3793
Conversation
This file contains 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
… weights **What does this PR do?** This PR lowers the profiling internal constant `MAX_ALLOC_WEIGHT` from 65535 to 10000. This value is used when clamping very high allocation profiling weights. Aka: Each allocation sample taken by the profiler has a weight assigned to it; this weight is clamped (e.g. limited to) this maximum value. The remaining weight is assigned to a separate "skipped samples" placeholder (see #3792). **Motivation:** Very large weights on samples can produce biased results; by lowering the maximum we reduce the maximum bias that can ever be introduced. **Additional Notes:** I've gathered data from a number of apps when choosing this value, see <https://docs.google.com/document/d/1lWLB714wlLBBq6T4xZyAc4a5wtWhSmr4-hgiPKeErlA/edit> for details (Datadog-only link, sorry!). **How to test the change?** The tests for #3792 also cover this change (although the changes are otherwise independent).
BenchmarksBenchmark execution time: 2024-07-18 10:52:23 Comparing candidate commit 4cd8e7b in PR branch Found 0 performance improvements and 0 performance regressions! Performance is the same for 10 metrics, 2 unstable metrics. |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #3793 +/- ##
=======================================
Coverage 97.91% 97.91%
=======================================
Files 1246 1246
Lines 75037 75037
Branches 3629 3629
=======================================
+ Hits 73473 73474 +1
+ Misses 1564 1563 -1 ☔ View full report in Codecov by Sentry. |
anmarchenko
approved these changes
Jul 24, 2024
Merged
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.
What does this PR do?
This PR lowers the profiling internal constant
MAX_ALLOC_WEIGHTfrom 65535 to 10000.This value is used when clamping very high allocation profiling weights.
Aka: Each allocation sample taken by the profiler has a weight assigned to it; this weight is clamped (e.g. limited to) this maximum value. The remaining weight is assigned to a separate "skipped samples" placeholder (see #3792).
Motivation:
Very large weights on samples can produce biased results; by lowering the maximum we reduce the maximum bias that can ever be introduced.
Additional Notes:
I've gathered data from a number of apps when choosing this value, see https://docs.google.com/document/d/1lWLB714wlLBBq6T4xZyAc4a5wtWhSmr4-hgiPKeErlA/edit for details (Datadog-only link, sorry!).
How to test the change?
The tests for #3792 also cover this change (although the changes are otherwise independent).