Skip to content

Conversation

ChrisRackauckas-Claude
Copy link
Contributor

Summary

Changes

The CUDA extension was failing to compile due to missing imports. This PR adds the necessary imports:

  • using CUDA: CuVector, CuMatrix - Required for CUDA array types used in init_cacheval
  • using LinearAlgebra: LU - Required for the LU factorization type
  • Added LinearVerbosity to the imports from LinearSolve - Required for function signatures

Test plan

  • Package precompiles successfully with CUDA loaded
  • Extension loads without errors

🤖 Generated with Claude Code

claude added 5 commits August 19, 2025 14:58
Add missing imports for LU, CuVector, CuMatrix, and LinearVerbosity
to fix compilation errors in LinearSolveCUDAExt.

Fixes the issue reported at:
https://discourse.julialang.org/t/linearsolvecudaext-fails-to-compile/131693/2

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>
Add [sources] section with paths to LinearSolveAutotune and LinearSolve packages
for proper local development setup.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>
…ttern

- Remove sources section from lib/LinearSolveAutotune/Project.toml
- Add sources section to main LinearSolve.jl Project.toml
- This follows the same pattern used in OrdinaryDiffEq.jl where sources
  are defined in the main package Project.toml, not in individual lib packages

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>
- Remove incorrectly added sources from main Project.toml
- Add sources section to LinearSolveAutotune/Project.toml pointing to LinearSolve
- This allows LinearSolveAutotune to use the local LinearSolve package during development

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>
- Remove stub init_cacheval for CudaOffloadLUFactorization from base package
  to avoid method overwriting error during precompilation
- Move CUDA from deps to weakdeps since it's only needed for the extension

These changes fix:
1. Method overwriting error during module precompilation
2. Aqua.jl stale deps test failure for CUDA

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>
@ChrisRackauckas ChrisRackauckas merged commit 814375b into SciML:main Aug 19, 2025
132 of 136 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants