Published May 21, 2026 | Version v2.4.0
Software Open

JuliaGPU/AMDGPU.jl: v2.4.0

  • 1. University of Lausanne
  • 2. Advanced Micro Devices
  • 3. KIT
  • 4. @JuliaComputing
  • 5. @AMD
  • 6. Curtin Institute of Data Science
  • 7. @mit
  • 8. @atdepth
  • 9. Argonne National Laboratory
  • 10. @UCL-ARC
  • 11. @dwavesystems
  • 12. RUB University
  • 13. UIUC/Google Deepmind
  • 14. DLR
  • 15. University of Birmingham

Description

AMDGPU v2.4.0

Diff since v2.3.0

Merged pull requests:

  • Replace ExceptionInfo with lightweight packed UInt64 atomic (#894) (@gbaraldi)
  • Add graph capture (#895) (@pxl-th)
  • Update compat bounds for 1.13 (#896) (@vchuravy)
  • Simplify memcpy handling (#897) (@pxl-th)
  • Store ROCArray offset in bytes instead of elements (#898) (@maleadt)
  • Test 1.13 and nightly (#899) (@kshyatt)
  • Fix stream printing (#900) (@luraess)
  • Bump julia-actions/setup-julia from 2 to 3 in the all-github-actions group (#902) (@dependabot[bot])
  • Fix sparse matrix adaption (#905) (@Abdelrahman912)
  • Adapt to GPUCompiler changes (#907) (@luraess)
  • Address RNG deprecation warning (#910) (@luraess)
  • Skip test for missing HIP libs (#911) (@luraess)
  • Rename master to main (#912) (@luraess)
  • Fix ROCSparseMatrix constructor (#913) (@luraess)
  • Drain sticky HIP error (#914) (@luraess)
  • Rework stream finalizer (#915) (@luraess)
  • Update wrappers to ROCm 7.13 (#916) (@luraess)

Closed issues:

  • GPUArrays test failure with ROCm 7 (#869)
  • Mi300x + room 7.1 fails roc solver tests (#879)
  • hipMallocFromPoolAsync broken on ROCm 7.2 (#901)

Files

JuliaGPU/AMDGPU.jl-v2.4.0.zip

Files (809.7 kB)

Name Size Download all
md5:e85d9ce795e5433325482cb443c527e1
809.7 kB Preview Download

Additional details

Related works