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
Profile-Guided Optimization (PGO) is already enabled for several packages in the MinGW repo: Clang, GCC. Since PGO improves performance, I suggest enabling PGO for the following packages as well:
Probably there are other packages in the MinGW repo that can benefit from PGO. You can find the existing benchmarks for other projects here. Maybe we even want to integrate PGO into the packages like Ruby, Perl, and other similar packages.
The text was updated successfully, but these errors were encountered:
Biswa96
added
the
enhancement
Package requires more dependencies, build options, different packaging style etc.
label
Dec 5, 2023
For linkers and compilers this makes a lot of sense, but for more general things like Ruby or Python which allow both execution or interactive REPL, the performance might be negatively impacted in use-cases other than what is simulated for PGO.
The alternative is running a lot of different uses for profiling which will take way more time to run and lower the overall performance gain.
PGO is, unfortunately, not a magical performance enhancement. It just provides benefits for the code paths used a lot with instrumentation and penalizes the paths instrumentation didn't touch.
I'd say look at packages case by case to determine if it's worth the tradeoff there
Hi!
Profile-Guided Optimization (PGO) is already enabled for several packages in the MinGW repo: Clang, GCC. Since PGO improves performance, I suggest enabling PGO for the following packages as well:
Probably there are other packages in the MinGW repo that can benefit from PGO. You can find the existing benchmarks for other projects here. Maybe we even want to integrate PGO into the packages like Ruby, Perl, and other similar packages.
The text was updated successfully, but these errors were encountered: