| From: | Andres Freund <andres(at)anarazel(dot)de> |
|---|---|
| To: | Nathan Bossart <nathandbossart(at)gmail(dot)com> |
| Cc: | John Naylor <john(dot)naylor(at)enterprisedb(dot)com>, John Naylor <jcnaylor(at)gmail(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: optimize lookups in snapshot [sub]xip arrays |
| Date: | 2022-08-05 22:04:34 |
| Message-ID: | 20220805220434.6h2aa7eww5rxird6@awork3.anarazel.de |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi,
On 2022-08-05 13:25:10 -0700, Nathan Bossart wrote:
> I went ahead and renamed it to pg_lfind32() and switched it back to
> returning the pointer. That felt the cleanest from the naming perspective,
> but as Andres noted, it might not be as fast as just looking for the
> presence of the element. I modified my small testing program to perform
> many searches on small arrays, and I wasn't able to identify any impact, so
> perhaps thіs is good enough.
Why on small arrays? I'd expect a difference mainly if it there's at least a
few iterations.
But mainly I'd expect to find a difference if the SIMD code were optimized a
further on the basis of not needing to return the offset. E.g. by
replacing _mm_packs_epi32 with _mm_or_si128, that's cheaper.
- Andres
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Andres Freund | 2022-08-06 00:03:11 | Re: Cleaning up historical portability baggage |
| Previous Message | Justin Pryzby | 2022-08-05 21:51:39 | Re: Proposal to provide the facility to set binary format output for specific OID's per session |