v0.9.23

KernelAbstractions v0.9.23

Diff since v0.9.22

Merged pull requests:

• Add GPU reverse mode to EnzymeExt (#454) (@wsmoses)
• Fixing transpose function, a missing "end" and some argument orders. (#487) (@Sixzero)
• add 1.11 to CI (#491) (@vchuravy)
• Improve CPU launch heuristic (#500) (@vchuravy)
• use AirspeedVelocity for benchmark CI (#502) (@vchuravy)
• Bump UnsafeAtomicsLLVM to 0.2 (#503) (@pxl-th)

Closed issues:

• [Enzyme] NTuple index not working on CUDA backend (#494)
• External functions in GPU ODE example (#496)
• Default CPU workgroupsize can be inadequat for higher-dimensionsal kernels (#499)

v0.9.22

KernelAbstractions v0.9.22

Diff since v0.9.21

Merged pull requests:

• Implement KA.functional() (#490) (@vchuravy)

Closed issues:

• Printing results changes results in CPU kernel (#485)
• KA.functional(::Backend) to mimick CUDA.functional (#489)

v0.9.21

KernelAbstractions v0.9.21

Diff since v0.9.20

Merged pull requests:

• Enzyme: simplify via mixedduplicated (#483) (@wsmoses)
• Bump version in Project.toml (#484) (@wsmoses)

Closed issues:

• Small discrepancy (#482)

v0.9.20

KernelAbstractions v0.9.20

Diff since v0.9.19

Merged pull requests:

• Avoid deadlock in EnzymeExt (#478) (@vchuravy)

v0.9.19

KernelAbstractions v0.9.19

Diff since v0.9.18

Merged pull requests:

• Add fallback for EnzymeCore.compiler_job_from_backend (#469) (@vchuravy)
• Bump KA test compat for Enzyme (#473) (@wsmoses)
• Putting output before input in function arguments in some examples (#475) (@evelyne-ringoot)
• Correctly adapt to ABI changes to EnzymeCore (#476) (@wsmoses)

v0.9.18

KernelAbstractions v0.9.18

Diff since v0.9.17

Merged pull requests:

• Bump EnzymeCore (#467) (@wsmoses)

Closed issues:

• MAX support? (#460)

v0.9.17

KernelAbstractions v0.9.17

Diff since v0.9.16

Merged pull requests:

• Bump actions/cache from 3 to 4 (#455) (@dependabot[bot])
• Bump codecov/codecov-action from 3 to 4 (#456) (@dependabot[bot])
• Revert "Bump codecov/codecov-action from 3 to 4" (#457) (@vchuravy)
• Revert "Allow empty return statements" (#463) (@vchuravy)

Closed issues:

• CPU version fails with a kernel that contains a return statement (#459)
• @index(Global, NTuple) Giving incorrect behavior with CPU() backend (#461)

v0.9.16

KernelAbstractions v0.9.16

Diff since v0.9.15

Merged pull requests:

• Allow empty return statements (#446) (@eschnett)
• Add allocate to docs (#450) (@Vaibhavdixit02)
• Replace ROCArrays with AMDGPU in docs (#451) (@pxl-th)
• Publish documentation preview (#452) (@pxl-th)

Closed issues:

• Allow return or return nothing in kernels (#443)
• Replace ROCArrays by AMDGPU in documentation? (#447)

v0.9.15

KernelAbstractions v0.9.15

Diff since v0.9.14

Merged pull requests:

• Test AMDGPU on 1.10 and limit 1.9 to gfx1030 (#441) (@vchuravy)
• CompatHelper: bump compat for Adapt to 4, (keep existing compat) (#445) (@github-actions[bot])

Closed issues:

• Atomix example fails with complex array element type (#444)

v0.9.14

KernelAbstractions v0.9.14

Diff since v0.9.13

Merged pull requests:

• [EnzymeExt] handle active vars to kernels (#439) (@wsmoses)
• Add simple example with Atomix.jl (#440) (@roflmaostc)

Closed issues:

• CUDADevice not defined (#296)