[SR-4814] Smoke benchmark optimization #47391
Labels
bug
A deviation from expected or documented behavior. Also: expected but undesirable behavior.
performance
Comments
|
@swift-ci create |
|
Do I understand this issue correctly as use the |
|
@palimondo Actually this looks like (basically) a dup of SR-4669 to me. I didn't realize someone had already proposed that. Do you see anything substantial in this bug that isn't in that one? If not, let's close it. As for a swift-ci contact, I think @shahmishal is the right person to work with. |
This issue was closed.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Additional Detail from JIRA
md5: 6c1911952bff35f46499c394765f59ed
duplicates:
Issue Description:
For smoke benchmarks (and maybe regular benchmarks too) we should follow a procedure where we quickly run very few iterations and discard the tests that show no change, then run the rest of the benchmarks for more iterations, and repeat the process as long as it is profitable to do so. This would help eliminate what we currently see in smoke benchmarks: differences are often reported that can easily be debunked by another pass at the tests.
/cc @palimondo
The text was updated successfully, but these errors were encountered: