Add AMDGPUOffloadFactorization algorithm support #708
Merged
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
This commit adds support for AMD GPU-accelerated linear solving through the new AMDGPUOffloadFactorization algorithm: - Added AMDGPUOffloadFactorization struct in src/extension_algs.jl with proper error handling when AMDGPU.jl is not loaded - Created LinearSolveAMDGPUExt extension in ext/LinearSolveAMDGPUExt.jl implementing GPU-offloaded LU factorization using AMDGPU.rocSOLVER - Added AMDGPU as weak dependency and extension configuration in Project.toml - Exported AMDGPUOffloadFactorization in src/LinearSolve.jl The implementation follows the same pattern as CudaOffloadFactorization, using rocSOLVER for LU factorization and solve operations on AMD GPUs. 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <[email protected]>
…zation - Renamed AMDGPUOffloadFactorization to AMDGPUOffloadLUFactorization for clarity - Added AMDGPUOffloadQRFactorization for QR-based solving - Updated extension to support both LU and QR factorizations - LU uses rocSOLVER.getrf\! and getrs\! - QR uses rocSOLVER.geqrf\!, ormqr\!, and rocBLAS.trsv\! 🤖 Generated with Claude Code Co-Authored-By: Claude <[email protected]>
|
Updated the PR with the following changes:
The implementation now provides two factorization options for AMD GPU offloading, allowing users to choose based on their numerical stability and performance requirements. |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
| using LinearSolve.LinearAlgebra, LinearSolve.SciMLBase | ||
|
|
||
| # LU Factorization | ||
| function SciMLBase.solve!(cache::LinearSolve.LinearCache, alg::AMDGPUOffloadLUFactorization; |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| function SciMLBase.solve!(cache::LinearSolve.LinearCache, alg::AMDGPUOffloadLUFactorization; | |
| function SciMLBase.solve!( | |
| cache::LinearSolve.LinearCache, alg::AMDGPUOffloadLUFactorization; |
| cache.cacheval = fact | ||
| cache.isfresh = false | ||
| end | ||
|
|
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
|
|
||
| A_gpu, ipiv = cache.cacheval | ||
| b_gpu = AMDGPU.ROCArray(cache.b) | ||
|
|
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| b_gpu = AMDGPU.ROCArray(cache.b) | ||
|
|
||
| AMDGPU.rocSOLVER.getrs!('N', A_gpu, ipiv, b_gpu) | ||
|
|
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| end | ||
|
|
||
| # QR Factorization | ||
| function SciMLBase.solve!(cache::LinearSolve.LinearCache, alg::AMDGPUOffloadQRFactorization; |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| function SciMLBase.solve!(cache::LinearSolve.LinearCache, alg::AMDGPUOffloadQRFactorization; | |
| function SciMLBase.solve!( | |
| cache::LinearSolve.LinearCache, alg::AMDGPUOffloadQRFactorization; |
| cache.cacheval = (A_gpu, tau) | ||
| cache.isfresh = false | ||
| end | ||
|
|
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
|
|
||
| A_gpu, tau = cache.cacheval | ||
| b_gpu = AMDGPU.ROCArray(cache.b) | ||
|
|
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
|
|
||
| # Apply Q^T to b | ||
| AMDGPU.rocSOLVER.ormqr!('L', 'T', A_gpu, tau, b_gpu) | ||
|
|
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| # Solve the upper triangular system | ||
| m, n = size(A_gpu) | ||
| AMDGPU.rocBLAS.trsv!('U', 'N', 'N', n, A_gpu, b_gpu) | ||
|
|
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| (A_gpu, tau) | ||
| end | ||
|
|
||
| end |
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| end | |
| end |
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.
Summary
This PR adds support for AMD GPU-accelerated linear solving through the new
AMDGPUOffloadFactorizationalgorithm:• Added
AMDGPUOffloadFactorizationstruct insrc/extension_algs.jlwith proper error handling when AMDGPU.jl is not loaded• Created
LinearSolveAMDGPUExtextension inext/LinearSolveAMDGPUExt.jlimplementing GPU-offloaded LU factorization using AMDGPU.rocSOLVER• Added AMDGPU as weak dependency and extension configuration in
Project.toml• Exported
AMDGPUOffloadFactorizationinsrc/LinearSolve.jlImplementation Details
The implementation follows the same pattern as
CudaOffloadFactorization, usingrocSOLVER.getrf\!for LU factorization androcSOLVER.getrs\!for solve operations on AMD GPUs via ROCArrays. The algorithm provides GPU acceleration for sufficiently large matrices where the computation benefits outweigh the data transfer costs.Test plan
🤖 Generated with Claude Code