Re: [PATCH] selftests/seccomp: Pin benchmark to single CPU

[Mark Brown]

On Tue, Feb 06, 2024 at 03:04:32AM -0800, Kees Cook wrote:
> On Tue, Feb 06, 2024 at 10:16:19AM +0000, Mark Brown wrote:
> > On Tue, Feb 06, 2024 at 01:56:47AM -0800, Kees Cook wrote:

> > > +	/* Set from highest CPU down. */
> > > +	for (cpu = ncores - 1; cpu >= 0; cpu--) {
> > > +		CPU_ZERO_S(setsz, setp);
> > > +		CPU_SET_S(cpu, setsz, setp);

> > Is there some particular reason to go from the highest CPU number down?
> > Not that it super matters but the default would be to iterate from 0 and
> > there's a comment but it just says the what not the why.

> I was arbitrarily picking a direction and all the examples I could find
> started at 0, so this would be more (?) out of the way. :P

> Without a cpu cgroup, I can't _exclude_ the pinned CPU from other
> processes, so I was pretending the last CPU will be less likely to be
> used.

That feels like it should go in a comment so it's a bit less mysterious.

Attachment: signature.asc
Description: PGP signature