Evaluate Profile-Guided Optimization (PGO) usage for better performance

[zamazan4ik]

Hi!

Recently I checked optimizations like Profile-Guided Optimization (PGO) and Post-Link Optimizations (PLO) improvements on multiple projects. The results are available here. According to the tests, all these optimizations can achieve better performance in many cases for many applications. Trying to enable them for the project could be a good idea since PGO shows great results on compilers.

I can suggest the following action points:

• Perform PGO benchmarks on the project. If it shows performance improvements - add a note to the documentation (README file?) about possible improvements in the library performance with PGO.
• Optimize prebuilt binaries with PGO (if any).
• Test more advanced approaches like Post-Link Optimization (PLO) with tools like LLVM BOLT.

Here are some examples of how PGO optimization is integrated into other projects:

• Rustc: a CI script for the multi-stage build
• GCC:
  • Official docs, section "Building with profile feedback" (even AutoFDO build is supported)
  • A part in a "wonderful" configure script
• Clang: Docs
• Python:
  • CPython: README
  • Pyston: README
• Go: Bash script
• V8: Bazel flag
• ChakraCore: Scripts
• Chromium: Script
• Firefox: Docs
  • Thunderbird has PGO support too
• PHP - Makefile command and old Centminmod scripts
• MySQL: CMake script
• YugabyteDB: GitHub commit
• FoundationDB: Script
• Zstd: Makefile
• Foot: Scripts
• Windows Terminal: GitHub PR
• Pydantic-core: GitHub PR
• file.d: GitHub PR
• OceanBase: CMake flag

I have some examples of how PGO information looks in the documentation:

• ClickHouse: https://clickhouse.com/docs/en/operations/optimizing-performance/profile-guided-optimization
• Databend: https://databend.rs/doc/contributing/pgo
• Vector: https://vector.dev/docs/administration/tuning/pgo/
• Nebula: https://docs.nebula-graph.io/3.5.0/8.service-tuning/enable_autofdo_for_nebulagraph/
• GCC: Official docs, section "Building with profile feedback" (even AutoFDO build is supported)
• Clang:
  • https://llvm.org/docs/HowToBuildWithPGO.html
  • https://llvm.org/docs/AdvancedBuilds.html
• tsv-utils: https://github.com/eBay/tsv-utils/blob/master/docs/BuildingWithLTO.md

Please, do not treat the discussion like a bug or smth like that. It's just an idea about usage compiler tweaks for the project.

[photonique]

Feels spammy

[zamazan4ik]

However, according to my findings (and not only mine), PGO improves compilers' performance in practice very well - you can find the results for Clang, GCC, Rustc, and other compilers in the awesome-pgo repo (or just check directly the Clang docs). From this perspective, I think raising the attention in the upstream about this optimization is a good thing (IMHO) since it can help with delivering better compiler performance to end-users.

I don't think that proposing an "Idea" for the project improvement is a bad thing, tbh.

[j2kun]

I think we definitely want to do this in HEIR at some point, especially once we have more custom hardware accelerator integrations, but I have no experience in PGO, and this project is rather young, so I'm of the mind that this should wait until we have some more established lowering paths to hardware.

[j2kun]

I think there's also some possibility of doing the Google-style thing where we just run a massive search over possible compiler configuration options and platforms to see what options work best.