Re: optimize atomic exchanges

Hi,

On 2023-11-30 21:18:15 -0600, Nathan Bossart wrote:
> On Wed, Nov 29, 2023 at 03:29:05PM -0600, Nathan Bossart wrote:
> > I haven't done any sort of performance testing on this yet. Some
> > preliminary web searches suggest that there is unlikely to be much
> > difference between cmpxchg and xchg, but presumably there's some difference
> > between xchg and doing cmpxchg in a while loop (as is done in
> > atomics/generic.h today). I'll report back once I've had a chance to do
> > some testing...
>
> Some rudimentary tests show a >40% speedup with this patch on x86_64.

On bigger machines, with contention, the wins are likely much higher. I see
two orders of magnitude higher throughput in a test program that I had around,
on a two socket cascade lake machine. Of course it's also much less
powerfull...

Greetings,

Andres Freund