On Fri, Nov 1, 2024 at 7:16 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I wrote:
> > Hmm. So it seems like we could do this:
> > * Linux, FreeBSD, OpenBSD: do run-time test as recommended
> > * Windows, macOS: assume that supported ARM hardware can do this
> > * anything else: assume no hardware CRC
>
> Actually, after chewing on that second point awhile longer,
> how about this modest proposal:
>
> * Drop all code for run-time determination of ARM CRC support.
> Assume it's there unless user builds with a -march option that
> says it definitely isn't.
For CRC32, that would work (and already does via configure tests) for
macOS (I'd forgotten about that in my earlier message, Macs never even
compile the CRC32 "choose" code because the system compiler targets M1
by default), recent RHEL (see below), and presumably others, but it
would penalise Debian, FreeBSD (using the standard binary package
repo) and others if we didn't have the runtime check due to their
conservative choice of baseline arch (as Bastien just said about
Debian in a crossed email). I've been wondering about what to do
about all this for a while... There are also other features we aren't
using yet, but should, or single instructions that we are calling
through function pointers, preventing automatic vectorisation etc.
> Realistically, exactly who is going to be running Postgres 18+
> on ARM hardware that lacks CRC support? I can think of lots of
> projects that are more worthy of our time than this.
>
> (Perhaps it's time to apply the same mindset to x86 CRC support
> too?)
I think it's quite interesting that RHEL has moved its baseline to
x86_86-v2, having recently worked with Intel and AMD to standardise
new levels "-vN" that group and "linearise" the feature sets, and
others were already talking about standarding on -v3 last time I
looked, while Debian still targets x86_86 which means the instruction
set of the first AMD K8 chip from 2003. Debian has a nice wiki page
(see thread ↓) explaining various workaround strategies, ranging from
runtime tests like ours, per-sub-architecture library selection,
through to (last resort) depending on pseudo-packages like
"x86-64-v2-support" so that your package becomes uninstallable on
ancient hardware. I'm not involved in packaging but I was wondering
out loud whether the PGDG repos might potentially consider making a
different choice than the official Debian repos, always or as an
option, or ... various other ideas proposed by others. Perhaps we
could continue the topic on that thread:
https://www.postgresql.org/message-id/flat/CA%2BhUKGKS64zJezV9y9mPcB-J0i%2BfLGiv3FAdwSH_3SCaVdrjyQ%40mail.gmail.com
In the meantime I'll see about the Linux/*BSD patch for this thing,
more tomorrow hopefully. And I might see if I can find a NetBSD
person to ask...
