Hi,
On 2022-11-05 12:05:43 +0300, Alexander Korotkov wrote:
> On Fri, Nov 4, 2022 at 10:07 PM Andres Freund <andres@anarazel.de> wrote:
> > The use of cmpxchg vs lock inc/lock add/xadd is one of the major reasons why
> > lwlocks are slower than a spinlock (but obviously are better under contention
> > nonetheless).
> >
> >
> > I have a benchmark program that starts a thread for each physical core and
> > just increments a counter on an atomic value.
>
> Thank you for this insight! I didn't know xadd is much cheaper than
> cmpxchg unless there are retries.
The magnitude of the effect is somewhat surprising, I agree. Some difference
makes sense to me, but...
> I also wonder how cmpxchg becomes faster with higher concurrency.
If you're referring to the leading 32/64 that's not concurrency, that's
32/64bit values. Sorry for not being clearer on that.
Greetings,
Andres Freund