Evaluate using Profile-Guided Optimization (PGO) and Post-Link Optimization (PLO) for optimizing Dyon interpreter performance

[zamazan4ik]

Hi!

I evaluate Profile-Guided Optimization (PGO) effects on software performance in different software domains - current results can be found here. Since PGO helps in many workloads including interpreters (like V8, CPython - PGO is already integrated into them), I think trying to optimize Dyon interpreter can improve its performance.

I can suggest trying to optimize Dyon with PGO. Since Dyon is written in Rust, you can use cargo-pgo tool. If will be positive results on performance - docs about PGO with Dyon can be created, and, I suppose, prebuilt binaries can be preoptimized with PGO.

After PGO, the Post-Link Optimization can be tested via tools like LLVM BOLT (also can be done with cargo-pgo).

I would be happy to answer all your questions about PGO and PLO.

Treat the issue just like an improvement idea, not an actual bug or something like that.