Re: Fix performance of generic atomics

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> On Wed, Sep 6, 2017 at 9:42 AM, Sokolov Yura
> <funny(dot)falcon(at)postgrespro(dot)ru> wrote:
>> But I think, generic version still should be "fixed".
>> If generic version is not reached on any platform, then why it is kept?
>> If it is reached somewhere, then it should be improved.

> This seems like a pretty sound argument to me. I think Tom's probably
> right that the changes in generic-gcc.h are the important ones, but
> I'm not sure that's an argument against patching generics.h. Given
> that pg_atomic_compare_exchange_u32_impl is defined to update *old
> there seems to be no reason to call pg_atomic_read_u32_impl every time
> through the loop.

Probably not. I'm not quite 100% convinced of that, because of my
observation that gcc is capable of generating different and better
code for some of these primitives if it can prove that the return
value is not needed. It's not clear that that could apply in any
of these uses of pg_atomic_compare_exchange_u32_impl, though.
In any case, by my own argument, it shouldn't matter, because if
any of these are really performance-critical then we should be
looking for better ways.

regards, tom lane