v11.5.8

GPUArrays v11.5.8

Diff since v11.5.7

Merged pull requests:

• [1.14] normalize is now gpu-safe (#735) (@christiangnrd)

v11.5.7

GPUArrays v11.5.7

Diff since v11.5.6

• Support findmin(f, itr) and findmax(f, itr) on complex-valued arrays if f's codomain is entirely real
• Reduce the time taken by the testsuite by merging similar tests related to mapreducedim

Merged pull requests:

• Reduce testsuite runtime by merging similar reduction testsets (#728) (@christiangnrd)
• Run Metal tests with --all (#736) (@christiangnrd)
• Update Buildkite pipeline for the new JuliaGPU cluster (#737) (@maleadt)
• Let findmax/min work with Complex and real-valued f (#738) (@kshyatt)
• Update Project.toml (#740) (@kshyatt)

v11.5.6

GPUArrays v11.5.6

Diff since v11.5.5

Merged pull requests:

• Keep Symmetric/Hermitian matmul with BLAS eltypes off BLAS.symm!/hemm!. (#734) (@maleadt)

v11.5.5

GPUArrays v11.5.5

Diff since v11.5.4

Merged pull requests:

• Support and test mul(C, A, B, alpha, beta) for triangular mats (#727) (@kshyatt)
• Update Project.toml (#729) (@kshyatt)

Closed issues:

• mul! on AbstractTriangulars fails if beta is not 0 (#726)

v11.5.4

GPUArrays v11.5.4

Diff since v11.5.3

Merged pull requests:

• JLArrays: Disallow conversions to CPU pointers. (#549) (@maleadt)
• Specialize ReshapedArray to resolve setindex! ambiguities (#680) (@shreyas-omkar)
• Use grid-stride loop for fill! (#691) (@christiangnrd)
• Test (and maybe fix) Strided matmatmul (#692) (@kshyatt)
• Remove inconsistent Base.copy! method (#696) (@nhz2)
• Bump codecov/codecov-action from 5 to 6 (#701) (@dependabot[bot])
• Bump julia-actions/cache from 2 to 3 (#702) (@dependabot[bot])
• Update PhiloxRNG.jl to v1.1.1 and fix counter overflow with UInt64 (#717) (@nhz2)
• Bump julia-actions/setup-julia from 2 to 3 (#718) (@dependabot[bot])
• broadcast: split ComposedFunction on AbstractGPUArrayStyle (#719) (@ChrisRackauckas)
• Generalize norm and normalize to AnyGPUArray. (#720) (@maleadt)

Closed issues:

• GPUArrays and Base have different resizing behaviors for copy! (#695)
• Wrap around in rand_batched_kernel! can lead to repeating patterns (#713)

v11.5.3

GPUArrays v11.5.3

Diff since v11.5.2

Merged pull requests:

• Reduce code of testsuite (#715) (@maleadt)

v11.5.2

GPUArrays v11.5.2

Diff since v11.5.1

Merged pull requests:

• Fixup/simplify coo_type (#709) (@kshyatt)
• Shrink high-RSS tests in testsuite. (#711) (@maleadt)
• Update Project.toml (#714) (@kshyatt)

v11.5.1

GPUArrays v11.5.1

Diff since v11.5.0

Merged pull requests:

• Add workspaces for 1.12. (#708) (@maleadt)
• Restore CPU-RNG to GPU-array fallback for rand!/randn!. (#710) (@maleadt)

v11.5.0

GPUArrays v11.5.0

Diff since v11.4.1

Merged pull requests:

• Fix typo in runtests.jl (#688) (@christiangnrd)
• Move bcast check earlier (#689) (@kshyatt)
• feat: add GPU-native kron support for Diagonal matrices (#690) (@shreyas-omkar)
• Simplify some type mappings for sparse arrays (#700) (@kshyatt)
• Store JLArray offset in bytes instead of elements (#703) (@maleadt)
• Detect contiguous views with mixed unit-range indices (#704) (@maleadt)
• Improve CI (#705) (@maleadt)
• Avoid scalar indexing in LinearAlgebra.normalize (#706) (@maleadt)
• Replace Tausworthe+LCG RNG with stateless Philox4x32-10 (#707) (@maleadt)

v11.4.1

GPUArrays v11.4.1

Diff since v11.4.0

Merged pull requests:

• [JLArrays] Update Julia version requirement to 1.10 (#682) (@christiangnrd)
• [CI] Use init_worker_code to reduce testset initialization overhead (#685) (@christiangnrd)
• Fix for sparse-to-sparse matrix broadcast (#686) (@kshyatt)
• Update Project.toml (#687) (@kshyatt)