Safety measures around AppleAccelerate #676
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
```julia Enzyme Derivative Rules: Error During Test at /home/runner/.julia/packages/SafeTestsets/raUNr/src/SafeTestsets.jl:30 Got exception outside of a @test LoadError: could not load library "/System/Library/Frameworks/Accelerate.framework/Accelerate" /System/Library/Frameworks/Accelerate.framework/Accelerate.so: cannot open shared object file: No such file or directory Stacktrace: [1] fix_ptr_lookup(name::String) @ Enzyme.Compiler.JIT ~/.julia/packages/Enzyme/LMVya/src/compiler/orcv2.jl:63 [2] add!(mod::LLVM.Module) @ Enzyme.Compiler.JIT ~/.julia/packages/Enzyme/LMVya/src/compiler/orcv2.jl:251 [3] _link(job::GPUCompiler.CompilerJob{<:Enzyme.Compiler.EnzymeTarget}, mod::LLVM.Module, edges::Vector{Any}, adjoint_name::String, primal_name::Union{Nothing, String}, TapeType::Any, prepost::String) @ Enzyme.Compiler ~/.julia/packages/Enzyme/LMVya/src/compiler.jl:5661 [4] cached_compilation @ ~/.julia/packages/Enzyme/LMVya/src/compiler.jl:5750 ``` Basically, you can't do option dependencies, but BinaryBuilder just simply sends no binary if the platform isn't supported. So LinearSolve.jl always has a dependency on AppleAccelerate, and if the binary exists it defaults to it (since it's pretty much always the fastest on M-series mac) and otherwise it disables it. Enzyme still seems to want to try to call it, so this should fix that.
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
| @@ -34,6 +34,8 @@ function aa_getrf!(A::AbstractMatrix{<:ComplexF64}; | |||
| ipiv = similar(A, Cint, min(size(A, 1), size(A, 2))), | |||
| info = Ref{Cint}(), | |||
| check = false) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
| @@ -54,6 +56,8 @@ | |||
| ipiv = similar(A, Cint, min(size(A, 1), size(A, 2))), | |||
| info = Ref{Cint}(), | |||
| check = false) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
| @@ -74,6 +78,8 @@ | |||
| ipiv = similar(A, Cint, min(size(A, 1), size(A, 2))), | |||
| info = Ref{Cint}(), | |||
| check = false) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
| @@ -94,6 +100,8 @@ | |||
| ipiv = similar(A, Cint, min(size(A, 1), size(A, 2))), | |||
| info = Ref{Cint}(), | |||
| check = false) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
| @@ -116,6 +124,8 @@ | |||
| ipiv::AbstractVector{Cint}, | |||
| B::AbstractVecOrMat{<:ComplexF64}; | |||
| info = Ref{Cint}()) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
| @@ -140,6 +150,8 @@ | |||
| ipiv::AbstractVector{Cint}, | |||
| B::AbstractVecOrMat{<:ComplexF32}; | |||
| info = Ref{Cint}()) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
| @@ -165,6 +177,8 @@ | |||
| ipiv::AbstractVector{Cint}, | |||
| B::AbstractVecOrMat{<:Float64}; | |||
| info = Ref{Cint}()) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
| @@ -190,6 +204,8 @@ | |||
| ipiv::AbstractVector{Cint}, | |||
| B::AbstractVecOrMat{<:Float32}; | |||
| info = Ref{Cint}()) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
| @@ -236,6 +252,8 @@ | |||
|
|
|||
| function SciMLBase.solve!(cache::LinearCache, alg::AppleAccelerateLUFactorization; | |||
| kwargs...) | |||
| __appleaccelerate_isavailable() || | |||
There was a problem hiding this comment.
[JuliaFormatter] reported by reviewdog 🐶
Suggested change
| __appleaccelerate_isavailable() || | |
| __appleaccelerate_isavailable() || |
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.
Basically, you can't do option dependencies, but BinaryBuilder just simply sends no binary if the platform isn't supported. So LinearSolve.jl always has a dependency on AppleAccelerate, and if the binary exists it defaults to it (since it's pretty much always the fastest on M-series mac) and otherwise it disables it. Enzyme still seems to want to try to call it, so this should fix that.
Checklist
contributor guidelines, in particular the SciML Style Guide and
COLPRAC.
Additional context
Add any other context about the problem here.