From: Alexander Bluhm <alexander.bluhm@gmx.net>
Subject: Re: improve spinning in mtx_enter
To: Jeremie Courreges-Anglas <jca@wxcvbn.org>
Cc: Mark Kettenis <mark.kettenis@xs4all.nl>, mpi@openbsd.org, mjguzik@gmail.com, tech@openbsd.org
Date: Mon, 25 Mar 2024 15:22:43 +0100

On Sun, Mar 24, 2024 at 01:23:39PM +0100, Jeremie Courreges-Anglas wrote:
> Le Sun, Mar 24, 2024 at 12:39:28PM +0100, Mark Kettenis a ?crit :
> > > I assume that's an amd64 machine?  Can we test this on some other
> > > architectures too before we commit it?  I don't expect a regression,
> > > but it would be interesting to see what the effect is on an
> > > architectures that has better atomic operations than amd64.
> > 
> > FWIW, I see no measureable difference on a kernel build on my 10-core
> > M2 Pro Mac mini with this diff.  (And that is fine).
> 
> Same on my single-socket 8 cores amd64 builder.

I did run regress with this diff on sparc64.  Apart from the usual
failures on sparc64 I see no issues.

With "time make -j64 bsd" on SPARC Enterprise T5220 I get:

without mutex diff:
    6m08.67s real   197m09.82s user    56m52.57s system

with mutex diff:
    6m07.02s real   203m42.07s user    45m33.08s system

I have no idea, why user increased.  Real time got slightly better,
and less kernel time is an improvement.

bluhm