Tags

## KernelAbstractions v0.3.1

[Diff since v0.3.0](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.3.0...v0.3.1)


**Closed issues:**
- Matrix examples fails on Julia 1.4 (#43)
- AssertionError from ndrange=() (#107)
- Kernel compilation for OffsetArrays fails with `KernelError: recursion is currently not supported ` (#110)
- Revise failure with KernelAbstractions (#111)

**Merged pull requests:**
- CompatHelper: add new compat entry for "LLVM" at version "1.5" (#105) (@github-actions[bot])
- add erf + erfc functions (#115) (@simonbyrne)
- Shared memory transpose (#116) (@vchuravy)
- add docs for localmem, private and uniform macros (#118) (@simonbyrne)
- Typo (#120) (@PallHaraldsson)

## KernelAbstractions v0.3.0

[Diff since v0.2.6](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.2.6...v0.3.0)



**Merged pull requests:**
- Update KernelAbstractions to use CUDA 1.0 (#104) (@jakebolewski)

## KernelAbstractions v0.2.6

[Diff since v0.2.5](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.2.5...v0.2.6)

## KernelAbstractions v0.2.5

[Diff since v0.2.4](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.2.4...v0.2.5)



**Merged pull requests:**
- Add `SpecialFunctions.gamma` to kernel language (#99) (@lcw)
- CompatHelper: add new compat entry for "SpecialFunctions" at version "0.10" (#100) (@github-actions[bot])
- Bump version to 0.2.5 (#101) (@simonbyrne)

## KernelAbstractions v0.2.4

[Diff since v0.2.3](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.2.3...v0.2.4)


**Closed issues:**
- LLVM error: Cannot select: 0xca6fd20: f64 = fpow 0x93a9e90 (#89)
- Defining the same function twice? Maybe a typo? (#93)
- private variable not available after an `@synchronize` (#95)
- `@synchronize` in an if statement (#96)

**Merged pull requests:**
- fix function name check (#52) (@GiggleLiu)
- Remove requires (#92) (@vchuravy)
- remove redefinition of kernel (#94) (@vchuravy)
- Allow typeof on @private memory (#97) (@jkozdon)

## KernelAbstractions v0.2.3

[Diff since v0.2.2](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.2.2...v0.2.3)



**Merged pull requests:**
- allow docstrings on kernels (#87) (@simonbyrne)
- test on 1.4 (#88) (@vchuravy)

## KernelAbstractions v0.2.2

[Diff since v0.2.1](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.2.1...v0.2.2)



**Merged pull requests:**
- Don't busy wait on the CPU (#84) (@vchuravy)
- Allow ndrange to be zero (#86) (@lcw)

## KernelAbstractions v0.2.1

[Diff since v0.2.0](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.2.0...v0.2.1)


**Closed issues:**
- Kernel launch overhead when launching lots of small kernels (#75)

**Merged pull requests:**
- CompatHelper: bump compat for "CUDAnative" to "3.0" (#77) (@github-actions[bot])
- Improve launch performance of kernels some more (#82) (@mwarusz)

## KernelAbstractions v0.2.0

[Diff since v0.1.6](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.1.6...v0.2.0)


**Closed issues:**
- Trailing return yields wrong CPU code. (#50)
- Cannot resolve isbits when using function call to isbits (#79)

**Merged pull requests:**
- Error on return statements inside kernels (#74) (@mwarusz)
- Forbid waiting in CUDA on a CPUEvents (#78) (@lcw)
- Improve launch performance of kernels (#80) (@vchuravy)
- Create MultiEvent from tuple of empty MultiEvents (#81) (@lcw)

## KernelAbstractions v0.1.6

[Diff since v0.1.5](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.1.5...v0.1.6)


**Closed issues:**
- Integrate `async_copy!` into the event system (#40)

**Merged pull requests:**
- Recurse into nested scopes containing synchronize (#70) (@mwarusz)
- Implement better error propagation and semaphore (#72) (@vchuravy)

## KernelAbstractions v0.1.5

[Diff since v0.1.4](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.1.4...v0.1.5)



**Merged pull requests:**
- Run Event(f) on main thread (#71) (@vchuravy)

## KernelAbstractions v0.1.4

[Diff since v0.1.3](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.1.3...v0.1.4)



**Merged pull requests:**
- Add CUDA rewrites for sincos(x) and exp(y) for complex y (#67) (@oschub)
- Add `Event(f, args)` to integrate code using at_async better (#68) (@vchuravy)
- async_copy! fixes (#69) (@lcw)

## KernelAbstractions v0.1.3

[Diff since v0.1.2](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.1.2...v0.1.3)



**Merged pull requests:**
- Allow MultiEvent to be created from an Event (#65) (@lcw)

## KernelAbstractions v0.1.2

[Diff since v0.1.1](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.1.1...v0.1.2)



**Merged pull requests:**
- Unified printing (#61) (@leios)
- add multievents (#62) (@vchuravy)
- don't recuse into functions like Base.sin (#63) (@vchuravy)
- make at_print work outside KA (#64) (@vchuravy)

## KernelAbstractions v0.1.1

[Diff since v0.1.0](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.1.0...v0.1.1)



**Merged pull requests:**
- Fix CUDA waiting on CUDA events (#59) (@lcw)

## KernelAbstractions v0.1.0



**Closed issues:**
- Variable live-time counter intuitive on the CPU (#13)
- Using Val as kernel argument triggers an assertion (#21)
- Performance of naive transpose  (#22)
- Initialization error (#23)
- `unroll` not defined inside a kernel (#24)
- Document that private memory works differently than scratch in GPUifyLoops (#31)
- How best sync with the default stream in the CUDA backed? (#46)

**Merged pull requests:**
- Bring up GPU functionality fully (#1) (@vchuravy)
- Cleanup docs and remove ScalarCPU (#2) (@vchuravy)
- CompatHelper: add new compat entry for "CUDAdrv" at version "5.1" (#4) (@github-actions[bot])
- CompatHelper: add new compat entry for "Requires" at version "1.0" (#5) (@github-actions[bot])
- add stream GC and wait with progress function (#10) (@vchuravy)
- Adding a few more examples (#12) (@leios)
- Fix and test local memory (#14) (@vchuravy)
- implement Const memory for GPU and CPU (#16) (@vchuravy)
- Handle type parameters in kernel functions (#25) (@vchuravy)
- be less judicous with escape (#26) (@vchuravy)
- dont't use nested inits (#27) (@vchuravy)
- Blocked iteration (#28) (@vchuravy)
- cleanup examples (#29) (@vchuravy)
- add group index (#32) (@vchuravy)
- Make kernels dispatchable (#33) (@mwarusz)
- Use macrotools (#34) (@vchuravy)
- add a block syntax for uniform (#35) (@vchuravy)
- Fix private memory on the CPU (#36) (@mwarusz)
- handle at_synchronize in blocks (#37) (@vchuravy)
- add ntuple index type (#38) (@vchuravy)
- fix nested unroll macros (#39) (@vchuravy)
- Allow CPU and CUDA kernels to wait on each other (#41) (@lcw)
- Fix tuple destructuring and bors+travis (#42) (@vchuravy)
- Wait for GPU events using synchronize (#45) (@mwarusz)
- [WIP] Infrastructure to sync CuDefaultStream() (#47) (@vchuravy)
- Allow CPU kernels to depend on default events (#51) (@lcw)
- Implement async_copy! (#53) (@vchuravy)
- CompatHelper: bump compat for "CUDAapi" to "4.0" (#56) (@github-actions[bot])
- only create as many tasks as threads and more inference barriers (#57) (@vchuravy)
- Ensure that constify doesn't cause arguments to be captured (#58) (@vchuravy)