12 ms·
from what I can tell he was looking hard for problems. He also noticed that if you ran 9 "yes" commands on a 8 core machine, the CPU wasn't allocated in 8/9ths
by dekhn 5y ago
from what I can tell he was looking hard for problems.
He also noticed that if you ran 9 "yes" commands on a 8 core machine, the CPU wasn't allocated in 8/9ths units (one process always got a lot less CPU). he was used to much more predictible CPU scheduler on multiprocessor systems (spoilt by TruCluster). Linux later adopted a better scheduler (fair share) so imho he was right on that one, even if "yes" is a terribly performance metric.
- fragmede 5y agoIt seems the lesson here is that changing things out from under someone isn't a good look. What would have been the cost in not changing users' shells for them?
- tedunangst 5y agoEverybody else has a worse shell.