I am trying to wrap the calls for Apple’s Accelerate Sparse Solver and it is my first try at wrapping apple routines. I am following some of the code structure in Accelerate.jl and LinearSolvers.jl but I keep failing at wrapping the SparseMatrixStructure object from Accelerate in Julia. I am follo…

This requires a little knowledge of C to understand. There is no function called SparseMatrixStructure. For Julia, you need to redeclare the struct. Clang.jl can automate the process of recreating the struct for you.

Thanks for your answer. What I don’t get is why the clang call will produce this method for example function SparseMultiply(A, X, Y) ccall((:SparseMultiply, libSparse), Cvoid, (SparseMatrix_Double, DenseMatrix_Double, DenseMatrix_Double), A, X, Y) end and I look for the symbol in the library i…

This was the code that I ran with clang options = load_options(joinpath(@__DIR__, "wrap.toml")) args = get_default_args() includedir = " /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libSparse.dylib" push!(args, "-I$includedir") …

This was an interesting rabbit hole to go down. It seems that all the dylibs are now aggregated into some kind of cache as of Big Sur: New in macOS Big Sur 11.0.1, the system ships with a built-in dynamic linker cache of all system-provided libraries. As part of this change, copies of dynamic li…

Here’s a quick and dirty demo. First, let’s start with the setup. struct DenseMatrix_Double rowCount::Cint columnCount::Cint columnStride::Cint attributes::UInt16 data::Ptr{Cdouble} end dense_data = zeros(Float64, 3,3) dense = DenseMatrix_Double(3, 3, 3, 0, pointer(dense_data)…

Thanks a lot for the prototype. I wonder how much work would it be to expose the linear solver. Any suggestions on how to handle the mangling? Or should I just try to work it out manually?

For mangling or demangling there are existing tools such as the following c++filt (GNU Binary Utilities) c++filt (GNU Binary Utilities) https://llvm.org/docs/CommandGuide/llvm-cxxfilt.html The LLVM version seems advisable given that we are dealing with an …

It seems to be packaged as part of LLVM_Full_jll under tools/llvm-cxxfilt

Progress report in case someone comes across this in the future: I successfully added Julia bindings for some of libSparse’s functionality ( GitHub link ). My approach requires manually working out the name mangling and individually writing the Julia wrappers. There’s bash script for demangling here …

We now have official support! The Julia bindings I made have been integrated with AppleAccelerate.jl, file here . Curiously, still no response on the Apple developer forum. I also tried reaching out through side channels, via a friend-of-a-friend who works at Apple, but all they could say was “yeah …

Apple Accelerate Sparse Solvers

General Usage

gdalle May 27, 2026, 8:18am 14

This looks awesome, thank you so much! Do you have any idea whether something similar is possible for Metal.jl?

Topic		Replies	Views
Wraping a MKL handle in julia General Usage ccall , mkl	16	1740	October 18, 2020
Cannot solve A\b = x with sparse matrix A - Julia V 0.6.2 Specific Domains	11	3175	May 20, 2018
Making efficient some small subroutines Performance	22	1254	May 8, 2020
Sparse matrix-vector product: much more slow than Matlab Performance matlab , optimization	24	4781	December 20, 2017
Apple M1 GPU from Julia? GPU question	20	6084	March 31, 2023

Apple Accelerate Sparse Solvers

Related topics