On 1/4/24 09:38, Jonathan Perkin wrote:
Yeah, like current load. We always had issues with things like this in the HPC environment, where programs would use all available cores on a compute node, when only some of them had been allocated by the scheduler. Unfortunately, some use cases rely on auto-detection, so we need to make it work as well as possible.* On 2024年01月04日 at 15:26 GMT, Jason Bacon wrote:In general these "use $number_of_cores" things are terrible anyway as they completely ignore many other very important factors. I wouldn't spend too much effort investing in them.If not, using hw.ncpu is not a catastrophe. I just means NetBSD cores will be somewhat oversubscribed by default.