Support numba 0.61 #37
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
- Add patches for 15.0.7, 16.0.6
- Avoid ffi dlopen namespace issues with initialization
- Recent versions of LLVM use opaque pointers. The numba frontend must communicate the pointee types to the LLVM plugin pass for lowering
- Avoids using other installation directories of LLVM in the path
- Ensure CUDA context is initalized once
- Fix single tests to nest in parallel (as it must) - Skip target data tests that are problematic if no target region executes - Clean up
There was a problem hiding this comment.
Pull request overview
This pull request adds support for Numba 0.61 by updating PyOMP to be compatible with LLVM versions 15, 16, and 17 (beyond the previously required LLVM 14). The changes involve substantial updates to both Python and C++ codebases to handle API changes and the migration to opaque pointers in LLVM.
Changes:
- Updated build system to support dynamic LLVM version selection via environment variable
- Migrated from deprecated Numba APIs (compile_isolated → njit, cffi → ctypes)
- Added opaque pointer support throughout C++ codebase with LLVM version conditionals
- Refactored CUDA codegen into singleton pattern
- Updated tests and documentation to reflect new device numbering (0=CUDA, 1=host)
Reviewed changes
Copilot reviewed 22 out of 23 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| setup.py | Refactored to dynamically download OpenMP sources based on LLVM_VERSION env var; added patch support for multiple LLVM versions |
| src/numba/openmp/init.py | Removed hard-coded LLVM 14 version check; moved decorators to separate module |
| src/numba/openmp/decorators.py | New module for jit/njit decorators extracted from init.py |
| src/numba/openmp/tags.py | Migrated from compile_isolated to njit API; added poison value operands for opaque pointer support |
| src/numba/openmp/omp_runtime.py | Replaced cffi with ctypes+llvmlite.binding for runtime function calls |
| src/numba/openmp/omp_ir.py | Refactored OpenMPCUDACodegen into singleton; removed nogil flag; updated to use new pass manager API |
| src/numba/openmp/libs/pass/.cpp/.h | Added LLVM version conditionals for opaque pointer support; updated OMPIRBuilder API calls for LLVM 16+ |
| src/numba/openmp/libs/pass/CMakeLists.txt | Made LLVM_DIR required parameter; removed LLVM version check |
| src/numba/openmp/tests/test_openmp.py | Updated device IDs in comments; added test for firstprivate arrays; uncommented test code; added skip decorators |
| buildscripts/cibuildwheel/setup-miniconda3.sh | Parameterized LLVM version via environment variable |
| README.md | Added compatibility matrix showing PyOMP 0.3.x supports Numba 0.57.x - 0.61.x |
| examples/hello-target.py | Changed device ID from 1 to 0 (CUDA backend) |
| src/numba/openmp/libs/libomp/patches/* | Added versioned patch directories for LLVM 14.0.6, 15.0.7, and 16.0.6 |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Todd A. Anderson <drtodd13@comcast.net>
- Append when x is a str - Cleanup dead code in arg_to_str
- Use LLVM 15 for building to be compatible with numba 0.61 - Pin all actions to the commit ref - Add least permissions at the workflow level
- Use command to check if nvidia-smi exists
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.