Profile-Guided Optimization (PGO) results

[zamazan4ik]

Writing this for the history. Maybe these results will be interesting to someone who trying to achieve better performance with quick-xml.

I test Profile-Guided Optimization (PGO) on different kinds of software - the current results are here(with a lot of other PGO-related information). That's why I tried to optimize quick-xml with PGO too.

Test setup

My test setup is:

• Macbook M1 Pro
• macOS Ventura 13.4
• Rustc version: rustc 1.71.1 (eb26296b5 2023-08-03)
• quick-xml version: https://github.com/tafia/quick-xml/commit/1be35e1645dad7d807347b490b4806339984a38f commit from the master branch

Benchmarks

As a benchmark, I used built-in into the quick-xml crate benchmarks. For PGO optimization I use cargo-pgo.

Results

• Instrumented run compared to Release: https://pastebin.com/CAYJUrQw
• PGO-optimized compared to Release: https://pastebin.com/hkVziJcZ

So as you see, for some workloads the performance has improved, for others - regressed. So you can estimate PGO benefits for your scenarios.

[Mingun]

@zamazan4ik , feel free to add a note to documentation that users that needs incredible performance in their applications could consider enabling PGO with links to your investigations