v11.3.1

GPUArrays v11.3.1

Diff since v11.3.0

Merged pull requests:

• Sparse array README fixes (#649) (@christiangnrd)
• Return early (don't kernel launch) for copytrito and zero-sized arrays (#660) (@kshyatt)
• Test all buildkite runs on LTS and latest (#661) (@kshyatt)

v11.3.0

GPUArrays v11.3.0

Diff since v11.2.6

• Sparse GPU array and broadcasting support (#628)
• Add support for diagm (#657)

Merged pull requests:

• Shortcut findfirst for empty arrays. (#610) (@tam724)
• Graceful exit for transpose/adjoint on size 0 arrays (#613) (@kshyatt)
• Another mul specialization for diagonals (#627) (@kshyatt)
• Sparse GPU array and broadcasting support (#628) (@kshyatt)
• Updates to GH workflow for 1.12 release (#629) (@kshyatt)
• Try to unbreak linalg test on nightly (#630) (@kshyatt)
• More diag mul methods (#631) (@kshyatt)
• Fix tr in Julia 1.12 (#633) (@mtfishman)
• Use ParallelTestRunner.jl (#634) (@vchuravy)
• use more precise isdefined check in linalg (#636) (@simeonschaub)
• Minor test improvements (#638) (@christiangnrd)
• Add generic_mul! (#639) (@michel2323)
• Fix deepcopy invalidation (#640) (@christiangnrd)
• Fix indexing for gpu rand (#641) (@christiangnrd)
• edge case: tril!, triu! and I constructors with empty matrices (#642) (@tam724)
• Bump peter-evans/find-comment from 3 to 4 (#644) (@dependabot[bot])
• Bump peter-evans/create-or-update-comment from 4 to 5 (#645) (@dependabot[bot])
• Fix and test opnorm for sparse (#650) (@kshyatt)
• Implement concurrency for Test.yml workflow (#651) (@giordano)
• Update to ParallelTestRunner v2 (#652) (@christiangnrd)
• More timeouts for test (#654) (@kshyatt)
• Add support for diagm (#657) (@kshyatt)
• Add AMDGPU tests via Buildkite (#658) (@luraess)
• Fix construction of empty identity matrix (#659) (@christiangnrd)

Closed issues:

• Indexing with arrays (#90)
• Linear solvers (#91)
• Broadcast of function with errors results in CPU execution (#146)
• Vector transpose (#238)
• sum of zero-array fails (#362)
• Introduce AbstractGPUSparseArray (#431)
• Out-of-bounds access in randn! (#614)
• New scalar indexing error for LinearAlgebra.tr in Julia 1.12 (#632)
• generic_mul! support for Julia 1.12 mul! dispatch (#637)

v11.2.6

GPUArrays v11.2.6

Diff since v11.2.5

Merged pull requests:

• CompatHelper: bump compat for JLD2 in [weakdeps] to 0.6, (keep existing compat) (#612) (@github-actions[bot])
• KernelAbstractions 0.10 support (#623) (@christiangnrd)
• Nightly fixes (#624) (@christiangnrd)
• Specialized mul for 3 diagonal arguments (#625) (@kshyatt)
• Simplify test runner. (#626) (@maleadt)

v11.2.5

GPUArrays v11.2.5

Diff since v11.2.4

Merged pull requests:

• Fixes for Julia 1.12 (#619) (@kshyatt)
• Update the README build status badges (#620) (@kshyatt)
• Try using the new with-testset for nightly (#621) (@kshyatt)
• Bump patch version (#622) (@kshyatt)

Closed issues:

• JLD dependency of GPUArrays breaks AMDGPU tests (#618)

v11.2.4

GPUArrays v11.2.4

Diff since v11.2.3

Merged pull requests:

• JLD2 extension take 2 (#597) (@christiangnrd)
• Defining neutral_element for xor (⊻) and fixing it for and (&). (#602) (@Hamiltonian-Action)
• reflect change in rotate! from LinearAlgebra.jl (#603) (@tam724)
• Test a few more corner cases in reductions (#604) (@christiangnrd)
• Test OpenCL with julia 1.11 (#605) (@christiangnrd)
• Bump actions/checkout from 4 to 5 (#616) (@dependabot[bot])
• Fix findminmax in the presence of NaN. (#617) (@maleadt)

Closed issues:

• kron!(::Diagonal, ::Diagonal, ::Diagonal) falls back to generic LinearAlgebra implementation (#601)

v11.2.3

GPUArrays v11.2.3

Diff since v11.2.2

Merged pull requests:

• Speed up fill for high dimensional arrays (#591) (@GiggleLiu)
• Fix version bounds for broken test (#593) (@christiangnrd)
• Fix CI warnings (#594) (@christiangnrd)
• Preserve 0-dim GPU arrays in map (#599) (@mtfishman)
• improve kron implementation (#600) (@simeonschaub)

Closed issues:

• New scalar indexing issue in JLArrays v0.2 (#571)
• map doesn't preserve 0-dim GPU arrays (#598)

v11.2.2

GPUArrays v11.2.2

Diff since v11.2.1

Merged pull requests:

• Restore map! behaviour (#584) (@vchuravy)

v11.2.1

GPUArrays v11.2.1

Diff since v11.2.0

Merged pull requests:

• Alloc cache: Add type assertion. (#581) (@maleadt)
• Alloc cache: preserve the cache across the cached macro. (#582) (@maleadt)
• More alloc cache improvements (#583) (@maleadt)

v11.2.0

GPUArrays v11.2.0

Diff since v11.1.0

GPUArrays.jl 11.2 adds two new features:

• a generic version of kron: #575
• a memory alocator cache: #576

Merged pull requests:

• Accomodate for rectangular matrices in copytrito! (#538) (@amontoison)
• Bump JLArrays to 0.2 (#569) (@pxl-th)
• [JLArrays] Add KA compat (#570) (@pxl-th)
• Bump codecov/codecov-action from 4 to 5 (#573) (@dependabot[bot])
• Add/unify broadcast in mul! with Diagonal (#574) (@tam724)
• Add vector-vector and matrix-matrix Kronecker product (#575) (@albertomercurio)
• Add caching allocator interface (#576) (@pxl-th)

Closed issues:

• Support LinearAlgebra.kron for all GPU backends (#558)
• Broadcast in overload of mul! for Diagonal (#572)

v11.1.0

GPUArrays v11.1.0

Diff since v11.0.0

Merged pull requests:

• Test-suite improvements. (#566) (@maleadt)
• Add storage query to get an array's DataRef. (#567) (@maleadt)