WARN: open /dev/cpu_dma_latency: Permission deniedįATAL: invalid cpu passed to numa_node_of_cpu(8) cyclictest -t20 (where 20 > number of cpus on your machine) Version-Release number of selected component (if applicable):Įverytime when requesting more threads than cpus
It wasn't possible to simply move the numa code after this though since the numa node obtained was applied to other following calls, so we need to apply the affinity code earlier, adjust the requested cpu, and then go through the numa settings. We were invoking this code after applying numa settings, resulting in some inconsistencies. However, there was code that adjusted the requested cpu in a round robin fashion after taking affinity into consideration.
I noticed that when requesting more threads than cpus we would get an error
0 kommentar(er)
