[Interest] Thread-To-CPU-Core distribution
thiago.macieira at intel.com
Thu Dec 29 12:40:10 CET 2022
On Thursday, 29 December 2022 07:59:28 -03 coroberti wrote:
> Additionally, to decrease the jitter and get more attention to your threads,
> you may try running as a root and switch to SCHED_FIFO priority and assign
> high priority there.
And, if you set the affinity to a single CPU, malloc() will conclude you re in a
single-core system (because that's what sched_setaffinity does) and therefore
not use arena-per-thread allocation strategies. Your mallocs() will take locks
in the same two global arenas.
On Linux with glibc.
Thiago Macieira - thiago.macieira (AT) intel.com
Cloud Software Architect - Intel DCAI Cloud Engineering
More information about the Interest