v2.3.0

AMDGPU v2.3.0

Diff since v2.2.1

Merged pull requests:

• Add KA.functional (#884) (@Abdelrahman912)
• Improving memory requirements on AMDGPU (#885) (@neoblizz)
• Properly parse the arch string (#886) (@neoblizz)
• Prefer aligned_sizeof (#887) (@vchuravy)
• Support Symbols on the GPU (#888) (@vchuravy)
• Update memory-related docs. (#891) (@neoblizz)
• Rework WMMA (#892) (@pxl-th)

Closed issues:

• WMMA fails tests on mi300x + rocm7.1 (#880)
• AMDGPU does not load anymore without error when installation not functional (#882)
• Segfaults on Strix Halo (Framework Desktop) (#889)

v2.2.1

AMDGPU v2.2.1

Diff since v2.2.0

Merged pull requests:

• update rocRAND wrappers (#867) (@simeonschaub)
• Move tests to ParallelTestRunner.jl (#876) (@gbaraldi)
• Improve testing on ParallelTestRunner (#878) (@luraess)

Closed issues:

• Miscompilation with nested loops and control flow (#672)

v2.2.0

AMDGPU v2.2.0

Diff since v2.1.4

Merged pull requests:

• Bump actions/checkout from 5 to 6 (#855) (@dependabot[bot])
• Group GitHub Actions in a single PR (#856) (@luraess)
• Integrate GPUArrays sparse interfaces (#859) (@pxl-th)
• Define GPUArrays tests dynamically (#865) (@pxl-th)
• Implement WMMA (#871) (@pxl-th)
• Add WMMA docs (#873) (@pxl-th)
• Fix matmatmul wrapper ambiguity (#874) (@luraess)

Closed issues:

• [Question] AI accelerators (#483)
• rocWMMA support (#560)
• GPUArrays testing (#858)

v2.1.4

AMDGPU v2.1.4

Diff since v2.1.3

Merged pull requests:

• Support mul!(Diagonal, A, B) (#853) (@kshyatt)

v2.1.3

AMDGPU v2.1.3

Diff since v2.1.2

Merged pull requests:

• Coverage for Strided Arrays for syrk, trsm, trmm and potrf (#826) (@evelyne-ringoot)
• Tweak CI timeout (#847) (@luraess)
• Update GPUArrays (#850) (@luraess)
• Update GPUArrays version (#851) (@luraess)

Closed issues:

• using AMDGPU fails after updating to ROCm 7.0.2 (#841)
• call to unknown function jl_f_issubtype but CUDA works (#846)
• Here is my errors in test(AMDGPU) with 780M. (#849)

v2.1.2

AMDGPU v2.1.2

Diff since v2.1.1

Merged pull requests:

• Even more BLAS tests and fixes (#832) (@kshyatt)
• Update buildkite for 1.12 (#833) (@kshyatt)
• Support for norm of Diagonal (#834) (@kshyatt)
• Support GPUToolbox 1.0 (#835) (@amontoison)
• Update discovery.jl with minimal approach to fix Windows discovery process (#839) (@jbiffl)

Closed issues:

• Memory leak on AMD MI250X GPU when tranferring data (#836)
• AMDGPU cannot find HIP/rocm install on Windows (#838)

v2.1.1

AMDGPU v2.1.1

Diff since v2.1.0

Merged pull requests:

• Add Julia 1.11 Enzyme job to pipeline (#828) (@wsmoses)
• Updates for PrettyTables 3 (#829) (@kshyatt)
• fix ambiguity error in Julia 1.12 (#830) (@simeonschaub)
• A few more blas tests and fix (#831) (@kshyatt)

Closed issues:

• matrix matrix multiplication gives error (#827)

v2.1.0

AMDGPU v2.1.0

Diff since v2.0.1

Merged pull requests:

• [rocSOLVER] Interface geblttrf_npvt and geblttrs_npvt (#746) (@amontoison)
• update hip wrappers (#813) (@simeonschaub)
• [rocSPARSE] Support more conversions (#816) (@amontoison)
• support cooperative groups (#817) (@simeonschaub)
• Support GPUToolbox 0.3 (#818) (@kshyatt)
• Bump actions/checkout from 4 to 5 (#820) (@dependabot[bot])
• Support lmul/rmul for ROCMatrix and Diagonal (#821) (@kshyatt)
• use correct wavefrontsize for device-side rng (#822) (@simeonschaub)
• fix CI: add JLD2 dependency (#823) (@simeonschaub)
• enable cooperative group tests only on 1.12 (#825) (@simeonschaub)

Closed issues:

• AMDGPU.jl v2 forces downgrade of MPI.jl (#812)
• Support GPUToolbox 0.3.0 (#815)
• device-side rng assumes warp size of 32 (#819)
• Cooperative group triggers out of bound memory error in CI (#824)

v2.0.1

AMDGPU v2.0.1

Diff since v2.0.0

Merged pull requests:

• Revert InexactError exception (#809) (@pxl-th)
• Update docs & fix own kwarg handling for unsafe_wrap (#811) (@pxl-th)

Closed issues:

• reinterpret now leads to invalid IR (#808)
• Update doc to reflect changes in memory (#810)

v2.0.0

AMDGPU v2.0.0

Diff since v1.3.6

Breaking changes:

Update unsafe_wrap to match Base definition: remove lock kwargs and add own kwarg.

Changes:

• Better kernel exception reporting mechanism:

julia> using AMDGPU

julia> function ker!(x)
           x[0] = 1
           return
       end
ker! (generic function with 1 method)

julia> x = ROCArray(zeros(Int, 1));

julia> @roc ker!(x);

julia> AMDGPU.synchronize()
ERROR: GPU Kernel Exception:
BoundsError: Out-of-bounds array access
workitemIdx: (x = 0x00000001, y = 0x00000001, z = 0x00000001)
workgroupIdx: (x = 0x00000001, y = 0x00000001, z = 0x00000001)
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:35
 [2] throw_if_exception(dev::HIPDevice)
   @ AMDGPU ~/.julia/dev/AMDGPU/src/exception_handler.jl:149
 [3] synchronize(stm::HIPStream; blocking::Bool, stop_hostcalls::Bool)
   @ AMDGPU ~/.julia/dev/AMDGPU/src/highlevel.jl:40
 [4] synchronize
   @ ~/.julia/dev/AMDGPU/src/highlevel.jl:36 [inlined]
 [5] synchronize()
   @ AMDGPU ~/.julia/dev/AMDGPU/src/highlevel.jl:36
 [6] top-level scope
   @ REPL[5]:1

• Disable eager GC by default (instead users should use GPUArrays.@cached).
• Implement KA.pagelock!.
• Remove old nonblocking sync (pre ROCm 6.0).
• Docs: Add FAQ entry for AMDGPU on Archlinux.
• Docs: Add SIMD docs & re-include installation tips.

Merged pull requests:

• Re-work exception reporting (take 2) (#800) (@pxl-th)
• Add FAQ entry for AMDGPU on Archlinux (#802) (@vchuravy)
• Update unsafe_wrap, implement KA.pagelock! & remove old stuff (#803) (@pxl-th)
• Fix gpu-less CI (#805) (@pxl-th)
• Disable eager GC by default (#806) (@pxl-th)
• Add SIMD docs & re-include installation tips (#807) (@pxl-th)

Closed issues:

• Improve exception reporting (#270)
• Support for atomic max on Float (#339)
• Implement revised KA functions copyto! and pagelock! (#724)
• 2D cumsum throwing GPU Kernel Exception (#742)
• CI Enzyme Julia 1.10 test failure (#765)
• CI failure on Julia 1.12 with GPUArrays linalg (#766)
• Noisy failure if ROCM not installed (#794)
• [Feature Request] Support KernelAbstractions pagelock! (#799)