Grid
Grid copied to clipboard
Add hint to shm error message
Hi,
I've been recently running Benchmark_ITT
and ran into
ShmBufferMalloc exceeded shared heap size -- try increasing with --shm <MB> flag
I was very pleased to immediately get a suggestion for a solution but then wasted hours trying out different values for shm
. In the end, it turned out that I had a typo in --mpi=1.1.2.2
instead of --mpi=2.2.1.1
. I presume that this led to a very weird (likely strided) memory layout that tried to allocate a huge amount of shared memory? Given the already helpful suggestion in the error message, I'd suggest to supplement this by a hint to also check the --mpi
layout.
Best,
Julian