Hi list, I seek an answer to the following question:
Is there an easy way to limit the memory allocations of a process bound to the processors of a specific NUMA node to only that specific NUMA node? Instead of allocating memory from other NUMA nodes, I would prefer for the process to stop running. In my experiments, I launch HPX processes per NUMA node (8 per cluster node). This is supposed to be an easy way to keep the process data near the process, in the same NUMA node. It works fine, but currently nothing prevents a process to allocate additional memory outside the NUMA node it is running on. This happens when the process needs more memory than the NUMA node provides, I think. I use SLURM to schedule the processes. In my code I do not use special memory allocators or anything like that. I just distribute processes over NUMA nodes, which is convenient for me at this moment. Thanks for any hints! Kor _______________________________________________ hpx-users mailing list hpx-users@stellar-group.org https://mail.cct.lsu.edu/mailman/listinfo/hpx-users
