PennyLaneAI / PennyLaneAI/catalyst

Improvements to the build system

Open
#1,600 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

CI/Build housecleaning
Dominant language
Python
Stars
234
Forks
84
Avg merge
2d 15h
Merged PRs (30d)
66

Description

After #1588 and #1589, there are a few issues still present in the build systems. This issue lists them here to not forget:

  • [mlir] test ASAN config on different systems
  • [mlir] fix existing ASAN issues
  • [mlir] clean up build output of non-core build targets (llmv, mlir-hlo, enzyme, plugin)
  • [mlir] do not build and link plugin targets every time, which currently adds significant overhead to the make catalyst command
  • [frontend] resolve any warnings in the Python test suite (fix or silence)
  • [frontend] resolve any XPass tests
  • [frontend] ensure C++ builds from the frontend use the correct compiler setup
    • For instance, the lapack kernel builds triggered from setup.py will pick an arbitrary compiler. On the AWS machine with gcc-9 available (on top of other compilers), it will pick that one. This leads to a failure to build the python package due to the "-std=c++20" flags which aren't supported by the picked compiler.
  • [frontend] do not use python3.10 by default in the frontend lit config
  • enable ASAN CI builds on PRs
  • enable both Clang and GCC CI builds on PRs
  • remove OpenMP dependency (I think it might unused now)

Lower down the priority list:

  • speed up handful of unusually slow pytests (likely the mcm one-shot tests)?
  • unify C++ code coverage tools (gcov on linux vs llvm-cov on mac)
  • unify sourcing of gtest (both llvm-style unit tests and stim depend on it, the latter currently raises a warning about missing gtest)

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by separating the checklist into focused tasks around the make catalyst command, setup.py, the frontend lit config, Python tests, and CI workflows. Check the relevant build and test commands first; this issue is done only when the selected build, compiler, warning, sanitizer, coverage, and CI problems have each been resolved and verified.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
build-system, ci-cd, testing-qa
Issue type
Refactor
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.