Giters

postgresml / postgresml

The GPU-powered AI application database. Get your app to market faster using the simplicity of SQL and the latest NLP, ML + LLM models.

Home Page:https://postgresml.org

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Evaluate using Profile-Guided Optimization (PGO) and LLVM BOLT

zamazan4ik opened this issue · comments

commented

Hi! A few days ago I read the article about PostgresML migration from Python to Rust. I was especially interested in your performance results since I have an additional idea of how it's possible to potentially improve the PostgresML performance.

Recently I checked Profile-Guided Optimization (PGO) improvements on multiple projects - the results are available here. According to the tests, PGO helps with achieving better performance in many software applications like databases, compilers, network applications, etc. I think trying to optimize PostgresML with PGO can be a good idea.

I can suggest the following action points:

  • Perform PGO benchmarks on PostgresML. If it shows improvements - add a note to the documentation about possible improvements in PostgresML's performance with PGO. PGO documentation examples in different projects you can find here.
  • Providing an easier way (e.g. via a build option) to build PostgresML with PGO can be helpful for the end-users and maintainers since they will be able to optimize PostgresML according to their own workloads.
  • Optimize pre-built PostgresML binaries with PGO.

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.

For the Rust projects, I recommend starting with cargo-pgo. More details about PGO support in Rust can be found in the official docs.

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