Perf Tests: Type 450 times for type metric instead of 20
#47852
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
|
Size Change: 0 B Total Size: 1.33 MB ℹ️ View Unchanged
|
|
Flaky tests detected in 17ea27e. 🔍 Workflow run URL: https://github.com/WordPress/gutenberg/actions/runs/4237827291
|
dmsnell
changed the title
type metric instead of 20type metric instead of 20
dmsnell
force-pushed
the
perf/type-200-times-instead-of-20
branch
2 times, most recently
from
714949b to
eee76d9
Compare
|
@sgomes @youknowriad seems like the new-page-per-test dramatically improved the shape of the distribution on the |
|
What happens to the bias if we disable/comment all the other metrics. I mean a job that only compares and logs the "type" metric. Is this something you tested in the past? The hypothesis is that if a single metric job does better than multiple ones, maybe we could rewrite the workflow to perform the metrics in different jobs. |
I tried this once and it didn't seem to have a notable impact, but that was before the new-page-per-test change. with the reported mean difference here being 0.32 ms, less than 1% difference, it concerns me far less from a data-quality standpoint. I'll update this PR to only track the
yeah this is probably a good idea overall. if I were spending more time on the tests I think I would try and do this more broadly, particularly since the recent introduction of theme perf tests bumped up the runtime again. the upside to splitting is that things will move that much faster in PR flows, but the downside is of course that local runs of the perf tests might get more complicated; instead of running one command we would potentially have to run a suite of commands. |
|
@youknowriad results for only collecting I'll rebase this PR and upload again, but for completeness sake, here is the distribution when only running the
|
By typing 200 times instead of 20 we will examine if this impacts the measured variability and bias in the `type` performance metric.
In this commit we're storing the raw results of all the performance test runs as an artifact in GitHub Actions so that we can perform more extensive analysis outside of workflow runs. For every _round_ of testing we end up with a copy of the results in a JSON file with the branch ref, the test suite name, and the test run index in its name. This makes it possible to analyze each run separately without making it hard to analyze everything together.
dmsnell
force-pushed
the
perf/type-200-times-instead-of-20
branch
from
0a6df2e to
17ea27e
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.
Status
Please ignore for now, this is for testing.
Saw an 8% difference in same-code branches for
type. Completed in 36 min for 1 round.Description
By typing 200 times instead of 20 we will examine if this impacts the measured variability and bias in the
typeperformance metric.Results