From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugzilla@gcc.gnu.org>
Received: by sourceware.org (Postfix, from userid 48)
	id 70DA0383A0F1; Fri,  9 Dec 2022 14:51:38 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 70DA0383A0F1
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1670597498;
	bh=LNtxzCZsv+qkfkpIqztNl8bZt9Trbtn+oe7BKEg5JKU=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=JRlEqauXzOorKxVtpIwWWZ4XL2tvoTE3MTtlOME/N6rE6YDcrca77ImvNVLbY7HCR
	 H5JlcgLajOtcNXkm5REmrsCDw4tNSZHemb+WOf9ltyK4T3nV8SuXEVvrpt+9tbBH80
	 YD9M/q2iy6vl0z26I3+Hox4D+ndjgrEkpy+Wsen8=
From: "mkretz at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug libstdc++/108030] `std::experimental::simd` not inlined
Date: Fri, 09 Dec 2022 14:51:37 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: libstdc++
X-Bugzilla-Version: 12.2.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: mkretz at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: mkretz at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 13.0
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-108030-4-rIBfuzcNu0@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-108030-4@http.gcc.gnu.org/bugzilla/>
References: <bug-108030-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
List-Id: <gcc-bugs.sourceware.org>

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D108030

--- Comment #3 from Matthias Kretz (Vir) <mkretz at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #2)
> I bet by adding too many always_inline functions that call normal inlines
> that is what is bound to happen, one runs into inline growth limits.  It =
is
> better to use always_inline on the leaf functions rather than on what cal=
ls
> them.

How is the inline growth limit determined? I mean, in the cases where it re=
ally
hurts, the resulting function compiles down to a single instruction (plus
parameter passing boilerplate). The optimizer cannot know about the number =
of
instructions, so what is the measure it uses?

Especially with the helper functions necessary to work with parameter packs=
 /
index_sequence, it's not enough to use always_inline on the leaf functions.
E.g. any simd binary operator basically should be [[gnu::always_inline,
gnu::flatten]]. However, simd maybe shouldn't use 'flatten' for functions t=
hat
call a user-provided callable (e.g. the simd generator constructor).=