From: | "Devulapalli, Raghuveer" <raghuveer(dot)devulapalli(at)intel(dot)com> |
---|---|
To: | John Naylor <johncnaylorls(at)gmail(dot)com> |
Cc: | "pgsql-hackers(at)lists(dot)postgresql(dot)org" <pgsql-hackers(at)lists(dot)postgresql(dot)org>, "Shankaran, Akash" <akash(dot)shankaran(at)intel(dot)com> |
Subject: | RE: Improve CRC32C performance on SSE4.2 |
Date: | 2025-02-12 21:02:27 |
Message-ID: | PH8PR11MB82860A6C0E5C4B384B52E55AFBFC2@PH8PR11MB8286.namprd11.prod.outlook.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Hi,
> 2. Unfortunately, there is another wrinkle that I failed to consider: If you search
> the web for "VirtualBox pclmulqdq" you can see a few reports from not very long
> ago that some hypervisors don't enable the CPUID for pclmul. I don't know how
> big a problem that is in practice today, but it seems we should actually have
> separate checks, with fallback. Sorry I didn't think of this earlier.
If someone using a VM that doesn't support a 15 yr old feature, then I would argue performance is not
the top priority for them. But that’s up to you. I will work on adding it unless you change your mind.
Also, do we really need to have both USE_SSE42_CRC32C and USE_SSE42_CRC32C_WITH_RUNTIME_CHECK
features support? The former macro is used to enable running the SSE42 version without a runtime check
when someone builds with -msse4.2. The code looks fine now, but the runtime dispatch rules get complicated
as we add the PCLMUL and AVX512 dispatch in the future. IMO, this additional complexity is not worth it.
The cpuid runtime dispatch runs just once when postgres server is first setup and would hardly affect performance.
Let me know what you think.
Raghuveer
From | Date | Subject | |
---|---|---|---|
Next Message | Nathan Bossart | 2025-02-12 21:04:20 | Re: describe special values in GUC descriptions more consistently |
Previous Message | Robert Haas | 2025-02-12 20:57:09 | Re: BitmapHeapScan streaming read user and prelim refactoring |