| From: | Andres Freund <andres(at)anarazel(dot)de> |
|---|---|
| To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
| Cc: | Mark Dilger <mark(dot)dilger(at)enterprisedb(dot)com>, Arjan van de Ven <arjan(at)linux(dot)intel(dot)com>, pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: src/port/snprintf.c: Optimize the common base=10 case in fmtint |
| Date: | 2021-10-26 18:58:31 |
| Message-ID: | 20211026185831.2r4bdyoojzpgu5gg@alap3.anarazel.de |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi,
On 2021-10-26 14:33:08 -0400, Tom Lane wrote:
> Andres Freund <andres(at)anarazel(dot)de> writes:
> > On 2021-10-26 13:51:55 -0400, Tom Lane wrote:
> >> I'd be inclined to just hard-wire the three allowed cases, and not have
> >> an arbitrary-divisor code path at all.
>
> > Yea, I came to the same conclusion. But I'd implement it by moving the
> > division into a separate inline function called from the switch. I tested that
> > locally and it works, but I got sidetracked by [1].
>
> Uh, why not just a "switch (base)" around three copies of the loop?
> Don't overthink this.
Well, putting the loop into its own function isn't really much more
complicated than duplicating the body. And there's also a few more
"unnecessarily run-time" branches that we could get rid of that way.
But I'm also ok with duplicating, at least for now.
Greetings,
Andres Freund
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Jonathan S. Katz | 2021-10-26 18:59:19 | allowing "map" for password auth methods with clientcert=verify-full |
| Previous Message | Andrew Dunstan | 2021-10-26 18:50:49 | Re: CREATEROLE and role ownership hierarchies |