Ludovic Räss

Indeed - thanks! So this should land in ROCm 6.1.1 right

It may also depend if you need stream/queue or device sync. Currently, it implements stream/queue sync for GPU backends.

> I'm not sure about AMDGPU.jl The closest to `CUDA.device!` you can get in AMDGPU is `AMDGPU.device_id!` (see https://github.com/JuliaGPU/AMDGPU.jl/commit/b80dc31ddb3097030aed253341a8ae4e4a943c36). Note that device indexing is 1-based in AMDGPU vs 0-based in...

@vchuravy following this as ability to handle ranges passed to kernels is also a feature that we would necessitate (FD MPI code) to allow for communication computation overlap (in a...

Yeah - having something more automatised could be a nice thing. @utkinis may have a small MWE on what we did recently which would be handy to have as well...

Re-running the Enzyme (v0.11.15) test suite now (Julia 1.10, Apple M2, macOS 14.2.1) I am getting ```julia-repl Test Summary: | Pass Total Time Attributor issues | 2 2 0.2s Test...

Can we re-open the issue?

Here are the outputs from the `device_code` for dynamic (dyn) and static (stat) expressions. [out_dyn.zip](https://github.com/JuliaGPU/KernelAbstractions.jl/files/14862435/out_dyn.zip) [out_stat.zip](https://github.com/JuliaGPU/KernelAbstractions.jl/files/14862437/out_stat.zip)

Should one do more globally what was done for Metal in there?

Maybe one could add a sentence similar to CUDA (https://github.com/JuliaParallel/MPI.jl/blob/master/docs/src/usage.md) ``` If using OpenMPI, the status of CUDA support can be checked via the [MPI.has_cuda()](https://github.com/JuliaParallel/MPI.jl/blob/master/docs/src/@ref) function. ``` in the ROCm-aware...