Date: Wed, 3 Jan 2024 09:29:54 -0600
From: Nathan Bossart <nathandbossart@gmail.com>
To: John Naylor <johncnaylorls@gmail.com>
Cc: Ants Aasma <ants@cybertec.at>, pgsql-hackers@postgresql.org
Subject: Re: add AVX2 support to simd.h
Message-ID: <20240103152954.GA1300001@nathanxps13>
References: <20231129171526.GA857928@nathanxps13>
 <CANWCAZZHEHufYD7nPNX6sPZ1mHo+30qTiV5UPXL_evh-N=1n5g@mail.gmail.com>
 <20240102161123.GA955987@nathanxps13>
 <CANWCAZavit+92_2E4K04RD7ErzY2T1Cc=3tysLYLk17t_GnM-g@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <CANWCAZavit+92_2E4K04RD7ErzY2T1Cc=3tysLYLk17t_GnM-g@mail.gmail.com>
Archived-At: <https://www.postgresql.org/message-id/20240103152954.GA1300001%40nathanxps13>
Precedence: bulk

On Wed, Jan 03, 2024 at 09:13:52PM +0700, John Naylor wrote:
> On Tue, Jan 2, 2024 at 11:11 PM Nathan Bossart <nathandbossart@gmail.com> wrote:
>> I'm tempted to propose that we move forward with this patch as-is after
>> adding a buildfarm machine that compiles with -mavx2 or -march=x86-64-v3.
> 
> That means that we would be on the hook to fix it if it breaks, even
> though nothing uses it yet in a normal build. I have pending patches
> that will break, or get broken by, this, so minus-many from me until
> there is an availability story.

How will this break your patches?  Is it just a matter of adding more AVX2
support, or something else?

If the requirement is that normal builds use AVX2, then I fear we will be
waiting a long time.  IIUC the current proposals (building multiple
binaries or adding a configuration option that maps to compiler flags)
would still be opt-in, and I'm not sure we can mandate AVX2 support for all
x86_64 builds anytime soon.

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com