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
In WGUNDERWOOD/tex-fmt#22 (comment) we found an issue that people are running their PGO training workloads according to the recommendations from the cargo-pgo's README file "for at least a minute". I completely understand your intention why did you write such a recommendation in the README file? However, as we see, this recommendation can be misleading for people - such a recommendation can be at least useless in some cases.
Maybe we can think about changing the wording a bit to something else. Honestly, I don't know what to write instead since choosing the proper PGO training workload is more like an art than a simple step-by-step algorithm. Or we can just remove it at all from the README file :)
The text was updated successfully, but these errors were encountered:
It's true that the minimum duration recommendation is more applicable when sampling is used, however cargo-pgo uses instrumentation, so that's a bit different indeed. I guess that the main point is to exercise all the important parts of the codebase (in the coverage sense).
Feel free to send a PR that rewords this.
zamazan4ik
added a commit
to zamazan4ik/cargo-pgo
that referenced
this issue
Oct 1, 2024
Hi!
In WGUNDERWOOD/tex-fmt#22 (comment) we found an issue that people are running their PGO training workloads according to the recommendations from the
cargo-pgo
's README file "for at least a minute". I completely understand your intention why did you write such a recommendation in the README file? However, as we see, this recommendation can be misleading for people - such a recommendation can be at least useless in some cases.Maybe we can think about changing the wording a bit to something else. Honestly, I don't know what to write instead since choosing the proper PGO training workload is more like an art than a simple step-by-step algorithm. Or we can just remove it at all from the README file :)
The text was updated successfully, but these errors were encountered: