The 2.6 and 2.7 releases of AMDGPU.jl broaden the package's linear-algebra coverage — GPU Cholesky, LU and now a singular value decomposition via rocSOLVER, mixed-precision matmul, and a sparse-array interface — add WMMA matrix-core support…
cuTile.jl v0.3 integrates with CUDA.jl, making it even easier to write and run CUDA Tile kernels in Julia. Performance has also been greatly improved, closing the gap with cuTile Python on every benchmark we ship. Added features include a r…
Metal.jl adds initial support for MPSGraph, with the matrix multiplication functions wrapped, resolving some matrix multiplication issues in the previous method. PR #526 enabled the automatic generation of wrappers for all enums, structs, a…
CUDA.jl v5.8 brings several enhancements, most notably the introduction of broadcasting support for CuSparseVector. The release also includes support for CUDA 12.9, and updates to key CUDA libraries like cuTENSOR, cuQuantum, and cuDNN. A si…
CUDA.jl v5.6 adds support for the new GPUArrays.jl caching allocator interface, which should improve performance of repetitive, memory-heavy applications. CUDA.jl v5.7 brings a greatly improved CuRef type, which enables fully asynchronous C…
The latest version of GPUArrays.jl involved a port of all vendor-neutral kernels to KernelAbstractions.jl. This should make it easier to add new functionality and improve the performance of existing kernels. Back in the day, we created GPUA…
CUDA.jl 5.5 is a minor release that comes with a couple of small improvements and new features. The only important change is that the minimal required Julia version has been bumped to 1.10, in anticipation of it becoming the next LTS releas…
CUDA.jl 5.4 comes with many memory-management related changes that should improve performance of memory-heavy applications, and make it easier to work with heterogeneous set-ups involving multiple GPUs or using both the CPU and GPU. Before
oneAPI.jl v1.5 is a significant release that brings many new features, from extended hardware support to greatly improved wrappers of the oneMLK math library. In oneAPI.jl v1.5 we introduce support for the Intel Ponte Vecchio (PVC) architec…
CUDA.jl 5.1 greatly improves the support of two important parts of the CUDA toolkit: unified memory, for accessing GPU memory on the CPU and vice-versa, and cooperative groups which offer a more modular approach to kernel programming. Unifi…
CUDA.jl 5.0 is an major release that adds an integrated profiler to CUDA.jl, and reworks how tasks are synchronized. The release is slightly breaking, as it changes how local toolkits are handled and raises the minimum Julia and CUDA versio…
Profiling GPU applications is hard, so this post shows how to use Intel's VTune Profiler to profile GPU applications written in Julia with oneAPI.jl. Because of the asynchronous nature of GPU execution, profiling GPU applications with Julia…