WSL2: Significant slowdown with Kokkos CUDA build
On WSL2 (Windows 11), the ProxyApp built with Kokkos 4.6.1 + CUDA (SM_89) shows abnormally long runtimes. The build is Release (-O3 -DNDEBUG), the binary and I/O are on ext4 (not /mnt/c), and CUDA_LAUNCH_BLOCKING is unset. Looking for guidance on WSL2-specific pitfalls or configuration tweaks for Kokkos/CUDA that could explain this slowdown.
Edited by Alexandre ROGET