sync issues on AMDGPU backend

omlins / ParallelStencil.jl

Package for writing high-level code for parallel high-performance stencil computations that can be deployed on both GPUs and CPUs

BSD 3-Clause "New" or "Revised" License

301 stars 31 forks source link

Closed luraess closed 10 months ago

luraess commented 11 months ago

Running large kernels on AMDGPU backend seem to experience issues with syncing. The behaviour is similar to https://github.com/omlins/ParallelStencil.jl/pull/107#discussion_r1271285037.

The issue is that AMDGPU.HIPStream constructor is used instead of getting the default stream for the current task AMDGPU.stream().