[Diff since v0.9.4](https://github.com/JuliaGPU/KernelAbstractions.jl/compare/v0.9.4...v0.9.5) **Closed issues:** - Defining timing infrastructure that works with events. (#15) - Kernels fail on CPU when waiting on kernels that allocate shared memory (#55) - Use macros in nested functions (#377) - `CPU(static=true)` option (#387) - Need for @inline when using GPU backend (#392) - KA seems to be broken for CUDA (#400) **Merged pull requests:** - Add kernel cpu=false and context accessor (#389) (@vchuravy) - Add reverse CI for oneAPI and AMDGPU (#391) (@vchuravy) - Update readme (#393) (@vchuravy) - Improve clarity of numa_aware example (#397) (@carstenbauer) - Docs: numa aware saxpy example (#398) (@carstenbauer) - Add implementation notes to host functionality (#401) (@vchuravy)