Evaluate using Profile-Guided Optimization (PGO) and Post-Link Optimization (PLO) #990
zamazan4ik
started this conversation in
Ideas
Replies: 2 comments 2 replies
-
Hello, I tested PGO with BOLT like 6 months ago, we got ~15% performance improvement. We have this issue: I have branch somewhere with kind-of basic flow (build with pgo, run performance tests, build again, build with bolt, etc.) For now, we simply don't have enough time to pursuit this :) I know this is really nice low hanging fruit and we will probably do it in future. |
Beta Was this translation helpful? Give feedback.
1 reply
-
Also, thank you! I will ping you if I'll have more questions - did you already join our discord? |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Hi!
Recently I checked many optimizations like PGO and PLO (mostly with LLVM BOLT) improvements on multiple projects including many databases - the results are available here. According to the tests, these optimizations can help to achieve better performance in many cases. I think trying to optimize Iggy's performance with PGO and PLO can be an interesting idea.
I can suggest the following action points:
iggy-server
according to their workloads.Here are some examples of how PGO optimization is integrated into other projects:
Unfortunately, for now, building the project with LTO and PGO enabled at the same time fails due to a bug in the compiler: rust-lang/rust#115344 (comment) . I suggest waiting for the fix in the upstream since disabling LTO is a worse option (IMHO).
I would be happy to answer your questions about all the optimizations above.
Beta Was this translation helpful? Give feedback.
All reactions