Evaluate using Profile-Guided Optimization (PGO) and Post Link Optimization (PLO)

[zamazan4ik]

Hi!

Recently I checked Profile-Guided Optimization (PGO) improvements on multiple projects. The results are here. According to the multiple tests, PGO can help with improving performance in many cases. There is at least one performance report where Naga performed faster than glslang, and for Naga PGO effects can be checked here. Since this, I think PGO integration into the glslang would be a good thing to have.

I can suggest the following action points:

• Perform PGO benchmarks on glslang. And if it shows improvements - add a note to the documentation about possible improvements in glslang performance with PGO.
• Providing an easier way (e.g. a build option) to build scripts with PGO can be helpful for the end-users and maintainers since they will be able to optimize glslang according to their workloads.
• Optimize pre-built binaries (if it's possible to do)

Maybe testing Post-Link Optimization techniques (like LLVM BOLT) would be interesting too (Clang and Rustc already use BOLT as an addition to PGO) but I recommend starting from the usual PGO.

Here are some examples of how PGO optimization is integrated in 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

[arcady-lunarg]

Better performance would definitely be nice to have, but I don't think PGO is going to get us a 30x speedup, for that we would probably need to do some actual profiling and look at algorithmic improvements.

[zamazan4ik]

Better performance would definitely be nice to have, but I don't think PGO is going to get us a 30x speedup, for that we would probably need to do some actual profiling and look at algorithmic improvements.

Algorithmic improvements are always nice, of course. However, enabling PGO helps even for well-optimized algorithms. Like better hot/cold splits, better inlining, etc. - all these optimizations are helpful for algorithmically-optimized programs too.

How many improvements PGO can bring to glslang - it should be tested. As an estimation, right now I have only Naga benchmarks - more tests should be performed on glslang with PGO.