All pull requests

Pull requests

• Use GPUArrays 12 for sorting, scans, reductions and findall

  #3356·maleadt opened Oct 7, 2026Member·

  ·

• Fix some bugs and add tests for EVEN MORE device math

  #3355·kshyatt opened Oct 6, 2026Member·

  ·

• Pad allocations and shared memory to whole 32-bit words

  bugfixcuda kernels
  #3353·maleadt opened Oct 6, 2026Member·

  ·
  2

• Use generic LLVM IR for device intrinsics where the back-end produces the same code

  bugfixcuda kernels
  #3351·maleadt opened Oct 6, 2026Member·

  ·
  1

• Implement the atomic functions with UnsafeAtomics

  #3350·maleadt opened Oct 6, 2026Member·

  ·
  1

• Report kernel lookups to the compile hook instead of recompiling

  #3348·maleadt opened Oct 6, 2026Member·

  ·

• Capture the operations of tasks spawned during graph capture

  #3337·maleadt opened Oct 5, 2026Member·4/7·

  ·
  2

• Launch trylock managed

  #3335·sheevy opened Oct 4, 2026·

  ·
  1

• KernelInterface: sub-group shuffles, votes and a constant width

  #3331·vchuravy opened Oct 4, 2026Member·

  ·

• Hand off arrays between streams on the GPU instead of the host

  #3329·maleadt opened Oct 4, 2026Member·6/7·

  ·
  2

• Don't wait for work that an event already ordered

  #3328·maleadt opened Oct 3, 2026Member·7/7·

  ·
  2

• Make CUDA graphs usable

  #3326·maleadt opened Oct 3, 2026Member·3/7·

  ·
  3

  4

• Keep unified memory host-attached on devices without concurrent managed access

  #3325·maleadt opened Oct 3, 2026Member·5/7·

  ·
  1

  2

• Don't wait for the GPU when releasing resources from finalizers

  #3318·maleadt opened Oct 1, 2026Member·1/7·

  ·
  2

• Recycle the streams of finished tasks

  #3317·maleadt opened Oct 1, 2026Member·2/7·

  ·
  7

• Implement KernelInterface, and support KernelAbstractions 0.10

  #3314·maleadt opened Sep 30, 2026Member·1/2·

  ·
  1

• Keep kernel launches type-stable with index-typed device arrays

  #3294·maleadt opened Sep 24, 2026Member·4/4·

  ·
  2

• Add an index type parameter to CuDeviceArray

  #3293·maleadt opened Sep 24, 2026Member·3/4·

  ·
  2

• Add hostcalls for device-to-host services

  #3243·maleadt opened Aug 20, 2026Member·

  ·
  2

• Try enabling fwd broadcating Enzyme test

  #3214·kshyatt opened Jul 24, 2026Member·

  ·
  4

• Implement unsafe_wrap! for in-place re-wrapping of non-owning CuArrays

  #3194·ameligrana opened Jul 12, 2026·

  ·
  1

• Remove some allocations from unsafe_wrap(::Type{CuArray{T,N,M}}, p::P…

  #3193·ameligrana opened Jul 12, 2026·

  ·
  3

• Expose directed rounding for Float64 WMMA tensor cores

  #3143·orkolorko opened May 16, 2026Contributor·4 tasks·

  ·
  2

• Use Republic.jl to forward subpackage APIs

  #3103·AntonOresten opened Apr 15, 2026Contributor·

  ·
  2

• CUPTI Profiler Host, Range Profiler, and PM Sampling APIs

  #3059·gbaraldi opened Mar 23, 2026Member·1 of 2 tasks·

  ·
  3