Open rachitnigam opened 6 years ago
maybe this doesn't matter. we don't need PLDI-grade performance any more. integrate with pyret and we can use pyret's performance benchmarks as a proxy.
Would that work for codeHS? Isolating major performance regressions at the Stopify level would be better than using Pyret as a proxy.
To be clear, not all PRs need perf testing. We just need to maintain a baseline and test against it on every major PR (eg: stack traces, adaptive deep stacks).
We're making some changes to the instrumentation this summer. We should bring back performance testing for Pull requests to make sure we don't have big regressions.
We used to have a cluster for doing the performance testing. Consider bringing that back up.