From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-return-200728-listarch-gcc=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 110396 invoked by alias); 25 Nov 2019 16:53:37 -0000
Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc/>
List-Post: <mailto:gcc@gcc.gnu.org>
List-Help: <http://gcc.gnu.org/ml/>
Sender: gcc-owner@gcc.gnu.org
Received: (qmail 110386 invoked by uid 89); 25 Nov 2019 16:53:37 -0000
Authentication-Results: sourceware.org; auth=none
X-Spam-SWARE-Status: No, score=-7.9 required=5.0 tests=AWL,BAYES_00,FREEMAIL_FROM,RCVD_IN_DNSWL_LOW,SPF_HELO_PASS,SPF_PASS autolearn=ham version=3.3.1 spammy=sincos, __builtin_cexpi, H*MI:@protonmail.com, protonmailcom
X-HELO: mail-40135.protonmail.ch
Received: from mail-40135.protonmail.ch (HELO mail-40135.protonmail.ch) (185.70.40.135) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Mon, 25 Nov 2019 16:53:35 +0000
Date: Mon, 25 Nov 2019 16:53:00 -0000
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=protonmail.com;	s=default; t=1574700812;	bh=LtuGL8G8gURCKk6Ctv4V0YcoI+MrpZkvR98V3+RuziY=;	h=Date:To:From:Cc:Reply-To:Subject:In-Reply-To:References:	 Feedback-ID:From;	b=CTtS5kKW55TEX4nANY2i0txLo5afBpyWW4ir3pq+mW/fHxseILuogb+ripE/hn8tz	 tNoxv4te9tzkY1O5yjHQWjFF4YiGowCS5nSztxxhxNI0rILOqS81rFLLQ6VD/JrNJL	 JsGcvI7MLe9P4ipy1T+ckpmJn8BDbZK9Aa+/uuvw=
To: Richard Biener <richard.guenther@gmail.com>
From: GT <tnggil@protonmail.com>
Cc: "gcc@gcc.gnu.org" <gcc@gcc.gnu.org>, Szabolcs Nagy <Szabolcs.Nagy@arm.com>, nd <nd@arm.com>, Bill Schmidt <wschmidt@linux.ibm.com>
Reply-To: GT <tnggil@protonmail.com>
Subject: Re: PPC64 libmvec implementation of sincos
Message-ID: <H7bxRYXZvjXK7ldtctmFF0Yn_APn9JL2aROfOmb6AcDY7FH-hpD6sOakmUfQL0tWzJyUJYlQSZye9d8EVhhibTexqPkDugYdWuvimHAnEbg=@protonmail.com>
In-Reply-To: <3F536966-92DE-49A3-A432-5646981EE26F@gmail.com>
References: <GcaLIHZaxSm2PVhunNuxEwEJB-BC3PeXIBlmYb6eMw54vZQ_KF6uBjlTgWmWtPNIMMeXu7UyYg797u1vV1_GS2Qv3ThxrUeORQD0wcddwwo=@protonmail.com> <0728c06a-d506-e8e8-c4a5-9187a7047f5f@arm.com> <3F536966-92DE-49A3-A432-5646981EE26F@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
X-SW-Source: 2019-11/txt/msg00206.txt.bz2

> >
> > i wonder if gcc can auto-vectorize scalar sincos
> > calls, the vectorizer seems to want the calls to
> > have no side-effect, but attribute pure or const
> > is not appropriate for sincos (which has no return
> > value but takes writable pointer args)
>
> We have __builtin_cexpi for that but not sure if any of the mechanisms ca=
n provide a mapping to a vectorized variant.
>

1. Using flags -fopt-info-all and -fopt-info-internals, the failure to vect=
orize sincos
is reported as "unsupported data-type: complex double". The default GCC beh=
avior is to
replace sincos calls with calls to __builtin_cexpi.

2. Using flags -fno-builtin-sincos and -fno-builtin-cexpi, the failure to v=
ectorize
sincos is different. In this case, the failure to vectorize is due to "numb=
er of iterations
could not be computed". No calls to __builtin_cexpi; sincos calls retained.

Questions:
1. Should we aim to provide a vectorized version of __builtin_cexpi? If so,=
 it would have
to be a PPC64-only vector __builtin-cexpi, right?

2. Or should we require that vectorized sincos be available only when -fno-=
builtin-sincos flag
is used in compilation?

I don't think we need to fix both types of vectorization failures in order =
to obtain sincos
vectorization.

Thanks.
Bert.