| From: | Andres Freund <andres(at)anarazel(dot)de> |
|---|---|
| To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
| Cc: | 吴亚飞 <wuyf41619(at)hundsun(dot)com>, "pgsql-hackers(at)lists(dot)postgresql(dot)org" <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: spinlock support on loongarch64 |
| Date: | 2022-11-02 21:04:52 |
| Message-ID: | 20221102210452.ydontvnukjrkewp6@awork3.anarazel.de |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi,
On 2022-11-02 14:55:04 -0400, Tom Lane wrote:
> I wrote:
> > So about like this, then.
>
> After actually testing (by removing the ARM stanza on a macOS machine),
> it seems that placement doesn't work, because of the default definition
> of S_UNLOCK at the bottom of the "#if defined(__GNUC__)" stuff. Putting
> it inside that test works, and seems like it should be fine, since this
> is a GCC-ism.
Looks reasonable. I tested it on x86-64 by disabling that section and it
works.
FWIW, In a heavily spinlock-contending workload it's a tad slower, largely due
to to loosing spin_delay. If I define that it's very close. Not that it
matters hugely, I just thought it'd be good to validate.
I wonder if it's worth keeing the full copy of this in the arm section? We
could just define SPIN_DELAY() for aarch64?
Greetings,
Andres Freund
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Peter Eisentraut | 2022-11-02 21:30:42 | Re: pg_dump: Refactor code that constructs ALTER ... OWNER TO commands |
| Previous Message | Peter Smith | 2022-11-02 21:01:48 | Re: create subscription - improved warning message |