Issues

• CPU backend: a 4-dimensional ndrange makes kernels up to 100× slower

  JuliaGPU/KernelAbstractions.jl#913 · Filter by author CarloLucibello opened on Oct 21, 2026

• CPU backend: invalid SPIR-V ("Block … appears in the binary before its dominator") for nested loops around @synchronize()

  kernelspoclupstream
  Bug

  JuliaGPU/KernelAbstractions.jl#894 · Filter by author CarloLucibello opened on Oct 13, 2026

• CPU backend: locals that live across a @synchronize() in a loop make a kernel 14–78× slower

  kernelsperformancepoclupstream
  Bug

  JuliaGPU/KernelAbstractions.jl#893 · Filter by author CarloLucibello opened on Oct 13, 2026

• CPU backend: kernels are vectorized with 256-bit vectors on AVX-512 CPUs where Julia uses 512-bit ones

  kernelsperformancepoclregressionupstream
  Bug

  JuliaGPU/KernelAbstractions.jl#891 · Filter by author CarloLucibello opened on Oct 13, 2026 · 0.10.1
  1

• CPU backend: a loop in a kernel with unsafe_indices=true is run with the work-items innermost, up to 100x slower

  kernelsperformancepoclregression
  Bug

  JuliaGPU/KernelAbstractions.jl#890 · Filter by author CarloLucibello opened on Oct 13, 2026 · 0.10.1
  3

• KernelAbstractions 0.10 doubles load time

  installationlatencypoclregression
  Bug

  JuliaGPU/KernelAbstractions.jl#889 · Filter by author maleadt opened on Oct 13, 2026
  1

• TagBot: Manual intervention needed for releases

  citagbot-manual
  Task

  JuliaGPU/KernelAbstractions.jl#887 · Filter by author github-actions opened on Oct 13, 2026

• Launch ranges past the back end's grid limits as grid-stride loops

  runtime
  Enhancement

  JuliaGPU/KernelAbstractions.jl#883 · Filter by author maleadt opened on Oct 12, 2026

• KernelInterface: describe grid limits that bound work-items, not work-groups

  runtime
  Enhancement

  JuliaGPU/KernelAbstractions.jl#882 · Filter by author maleadt opened on Oct 12, 2026

• POCL: contract fast-math flags lost through SPIR-V, no FMAs emitted

  kernelsperformancepoclupstream
  Enhancement

  JuliaGPU/KernelAbstractions.jl#865 · Filter by author vchuravy opened on Oct 10, 2026 · 0.10.1
  2

• CPU backend: math functions are scalar libm calls per work-item; PoCL never uses its vector math library

  kernelsperformancepocl
  Enhancement

  JuliaGPU/KernelAbstractions.jl#860 · Filter by author vchuravy opened on Oct 10, 2026 · 0.10.1
  1

• Proposal: portable compiler options for kernels (fastmath, always_inline, maxthreads, …)

  kernelsspeculative
  Enhancement

  JuliaGPU/KernelAbstractions.jl#839 · Filter by author vchuravy opened on Oct 4, 2026

• KernelInterface 1.0: a better API for writing kernels

  kernels
  Task

  JuliaGPU/KernelAbstractions.jl#810 · Filter by author maleadt opened on Sep 30, 2026
  3

• [CI] BusyBox timeout causing failures with Windows CI

  ci
  Task

  JuliaGPU/KernelAbstractions.jl#723 · Filter by author christiangnrd opened on Jul 2, 2026
  2

• Implement KernelAbstractions.get_backend(::Memory{Float64})

  good first issueruntime
  Enhancement

  JuliaGPU/KernelAbstractions.jl#691 · Filter by author FerreolS opened on Mar 25, 2026
  2

• Usage of ::Int

  kernels
  Enhancement

  JuliaGPU/KernelAbstractions.jl#685 · Filter by author AntonReinhard opened on Mar 16, 2026
  2

• Implement KernelAbstractions.get_backend(::PyArray)

  extensions
  Enhancement

  JuliaGPU/KernelAbstractions.jl#677 · Filter by author mattsignorelli opened on Jan 30, 2026

• add Documentation link in About section

  documentation
  Task

  JuliaGPU/KernelAbstractions.jl#662 · Filter by author raman-maker opened on Nov 22, 2025

• Interesting OpenCL extensions

  kernels
  Task

  JuliaGPU/KernelAbstractions.jl#658 · Filter by author vchuravy opened on Nov 19, 2025
  1

• Support dynamic shared memory in kernels

  kernels
  Enhancement

  JuliaGPU/KernelAbstractions.jl#655 · Filter by author christiangnrd opened on Nov 7, 2025

• Improve error on Enzyme inability to differentiate

  extensionskernels
  Bug

  JuliaGPU/KernelAbstractions.jl#629 · Filter by author KookiesNKareem opened on Jul 17, 2025
  5

• Support for allocate(backend, desired_axes) in addition to allocate(backend, desired_size)

  runtime
  Enhancement

  JuliaGPU/KernelAbstractions.jl#619 · Filter by author dzhang314 opened on Jun 26, 2025
  3

• @Const macro raises an error if using import KernelAbstractions

  kernels
  Bug

  JuliaGPU/KernelAbstractions.jl#616 · Filter by author Hamiltonian-Action opened on Jun 18, 2025

• [fr] Make histogram example more realistic

  documentation
  Enhancement

  JuliaGPU/KernelAbstractions.jl#605 · Filter by author Moelf opened on Jun 7, 2025

• Support OrdinalRanges in ndrange argument

  kernels
  Enhancement

  JuliaGPU/KernelAbstractions.jl#603 · Filter by author MasonProtter opened on Jun 2, 2025