From: Tom Lane <tgl@sss.pgh.pa.us>
To: Nathan Bossart <nathandbossart@gmail.com>
cc: Jeff Davis <pgsql@j-davis.com>, pgsql-hackers@postgresql.org,
        Xiang.Gao@arm.com
Subject: Re: always use runtime checks for CRC-32C instructions
In-reply-to: <20231031195331.GA77436@nathanxps13>
References: <20231030161706.GA3011@nathanxps13> <2427946.1698683963@sss.pgh.pa.us> <c1e9436426cb0368c6664130c3e89b8923bc449d.camel@j-davis.com> <20231031033601.GA68409@nathanxps13> <20231031155518.GA70172@nathanxps13> <2613682.1698779776@sss.pgh.pa.us> <20231031195331.GA77436@nathanxps13>
Comments: In-reply-to Nathan Bossart <nathandbossart@gmail.com>
	message dated "Tue, 31 Oct 2023 14:53:31 -0500"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2620793.1698783160.1@sss.pgh.pa.us>
Date: Tue, 31 Oct 2023 16:12:40 -0400
Message-ID: <2620794.1698783160@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/2620794.1698783160%40sss.pgh.pa.us>
Precedence: bulk

Nathan Bossart <nathandbossart@gmail.com> writes:
> Okay.  With that in mind, I think the path forward for new instructions is
> as follows:

> * If the special CRC instructions can be used with the default compiler
>   flags, we can only use newer instructions if they can also be used with
>   the default compiler flags.  (My M2 machine appears to add +crypto by
>   default, so I bet your buildfarm animals would fall into this bucket.)
> * Otherwise, if the CRC instructions can be used with added flags (i.e.,
>   the runtime check path), we can do a runtime check for the new
>   instructions as well.  (Most other buildfarm animals would fall into this
>   bucket.)

This seems like a reasonable proposal.

> Any platform that can use the CRC instructions with default compiler flags
> but not the new instructions wouldn't be able to take advantage of the
> proposed optimization, but it also wouldn't be subject to the small
> performance regression.

Check.  For now I think that's fine.  If we get to a place where this
policy is really leaving a lot of performance on the table, we can
revisit it ... but that situation is hypothetical and may remain so.

(It's worth noting also that a package builder can move the goalposts
at will, since our idea of "default flags" is really whatever the user
says to use.)

			regards, tom lane