From: Richard Biener <rguenther@suse.de>
To: Robin Dapp <rdapp.gcc@gmail.com>
Cc: Tamar Christina <Tamar.Christina@arm.com>,
gcc-patches <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH] ifcvt/vect: Emit COND_ADD for conditional scalar reduction.
Date: Fri, 6 Oct 2023 13:43:27 +0000 (UTC) [thread overview]
Message-ID: <nycvar.YFH.7.77.849.2310061333310.5561@jbgna.fhfr.qr> (raw)
In-Reply-To: <f28f3486-a025-59fc-8305-791786b8b6e3@gmail.com>
On Fri, 6 Oct 2023, Robin Dapp wrote:
> > We might need a similar assert
> >
> > gcc_assert (HONOR_SIGNED_ZEROS (vectype_out)
> > && !HONOR_SIGN_DEPENDENT_ROUNDING (vectype_out));?
>
> erm, obviously not that exact assert but more something like
>
> if (HONOR_SIGNED_ZEROS && !HONOR_SIGN_DEPENDENT_ROUNDING...)
> {
> if (dump)
> ...
> return false;
> }
>
> or so.
Yeah, of course the whole point of a fold-left reduction is to
_not_ give up without -ffast-math which is why I added the above.
I obviously didn't fully verify what happens for an original
MINUS_EXPR. I think it's required to give up for -frounding-math,
but I think I might have put the code to do that in a generic
enough place.
For x86 you need --param vect-partial-vector-usage=2 and an
AVX512 enabled arch like -march=skylake-avx512 or -march=znver4.
I think tranforming - x to + (-x) works for signed zeros.
So if you think you got everything correct the patch is OK as-is,
I just wasn't sure - maybe the neutral_element change deserves
a comment as to how MINUS_EXPR is handled.
Richard.
next prev parent reply other threads:[~2023-10-06 13:43 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-09-20 13:51 Robin Dapp
2023-09-27 0:44 ` Tamar Christina
2023-10-04 7:54 ` Robin Dapp
2023-10-04 13:15 ` Robin Dapp
2023-10-04 15:12 ` Tamar Christina
2023-10-05 8:54 ` Robin Dapp
2023-10-05 9:02 ` Robin Dapp
2023-10-05 14:05 ` Robin Dapp
2023-10-05 14:15 ` Tamar Christina
2023-10-06 9:10 ` Richard Biener
2023-10-06 12:28 ` Robin Dapp
2023-10-06 12:30 ` Robin Dapp
2023-10-06 13:43 ` Richard Biener [this message]
2023-10-06 20:54 ` Robin Dapp
2023-10-09 8:25 ` Richard Biener
2023-10-09 12:54 ` Robin Dapp
2023-10-09 13:05 ` Richard Biener
2023-10-09 5:50 ` Richard Sandiford
2023-10-09 12:02 ` Robin Dapp
2023-10-09 14:57 ` Richard Sandiford
2023-10-11 19:15 ` Robin Dapp
2023-10-12 10:47 ` Richard Sandiford
2023-10-12 11:11 ` Richard Biener
2023-10-19 20:07 ` Robin Dapp
2023-10-23 10:53 ` Richard Biener
2023-10-24 11:11 ` Richard Sandiford
2023-10-24 19:56 ` Robin Dapp
2023-10-31 21:04 ` Richard Sandiford
2023-10-31 21:19 ` Robin Dapp
2023-11-02 7:48 ` Richard Biener
2023-09-27 11:42 ` Richard Biener
2023-11-02 23:26 ` Andrew Pinski
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=nycvar.YFH.7.77.849.2310061333310.5561@jbgna.fhfr.qr \
--to=rguenther@suse.de \
--cc=Tamar.Christina@arm.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=rdapp.gcc@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).