You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm experiencing significant slowdown (~10x at least) when collecting observability data vs. not. At first I thought the data collection itself was slow, but even a no-op TESTCASE_CALLBACK causes the performance degradation. I was able to narrow the issue down to tracing; if this if-statement is disabled, performance returns to normal.
I think this kind of performance hit is expected for coverage measurement, and that's fine in a lot of cases, but Tyche needs to operate in a pretty tight feedback loop. Can we add a flag that makes coverage measurement optional?
The text was updated successfully, but these errors were encountered:
#3776 means collecting coverage info should be pretty fast on CPython 3.12+, but I'd be happy to add another environment variable and make the _trace_obs calculation a bit more complicated. Want to open a PR?
I'm experiencing significant slowdown (~10x at least) when collecting observability data vs. not. At first I thought the data collection itself was slow, but even a no-op
TESTCASE_CALLBACK
causes the performance degradation. I was able to narrow the issue down to tracing; if this if-statement is disabled, performance returns to normal.I think this kind of performance hit is expected for coverage measurement, and that's fine in a lot of cases, but Tyche needs to operate in a pretty tight feedback loop. Can we add a flag that makes coverage measurement optional?
The text was updated successfully, but these errors were encountered: