Timothy Williams comments

Results 4 comments of


                                            Timothy Williams

Memory growth with GPU-aware MPICH on Intel PVC GPUs

@abrooks98 I don't have a small reproducer yet. Where we observed it is pretty deep down in XGC functionality, and involves a number of template instances as well as Kokkos...

mpl/gpu/ze: add ipc cache limit enforcement and replacement policy

@abrooks98 As discussed with you and Renzo Bustamante, there's no small reproducer for this, and it's far from trivial to construct one because it's an alltoallv() pattern and only shows...

mpl/gpu/ze: add ipc cache limit enforcement and replacement policy

> Hi @zippylab, thanks for clarifying. Unfortunately the test suite does not currently support alltoallv (or any vector-based collectives) for GPU buffers; It only supports the non-vector collectives. Though, I...

mpl/gpu/ze: add ipc cache limit enforcement and replacement policy

@abrooks98 I reran with `export MPIR_CVAR_CH4_IPC_GPU_MAX_ENTRIES=4`, and the behavior was the same as without setting this environment variable: the FDs increased monotically (and the GPU memory consumed grew rapidly).