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 6C28C3858C35; Sun,  3 Dec 2023 13:58:02 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 6C28C3858C35
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1701611882;
	bh=Flj4NdqkUxCE3/uwuHq9PYllskM+vfe0XeqjoZFFERE=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=SRq6VSSEKuuykhew4g7aiFgq9S4wsGUT3AOfLsHGh5obYOBSomUvBgDrQFRE0pK8l
	 aVnyrHnO7xlOEWap3Tr94FxNZnFVXoS7So3PUKzgGCkslEhSPQtwEe717neozrJMgw
	 L9pPY/A3LaHBbbhJuARbu0nBYi3BJ9DbNiMQtlZM=
From: "elrodc at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug middle-end/112824] Stack spills and vector splitting with
 vector builtins
Date: Sun, 03 Dec 2023 13:58:02 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: middle-end
X-Bugzilla-Version: 14.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: elrodc at gmail dot com
X-Bugzilla-Status: UNCONFIRMED
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-112824-4-qVKFOtjwPQ@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-112824-4@http.gcc.gnu.org/bugzilla/>
References: <bug-112824-4@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=3D112824

--- Comment #3 from Chris Elrod <elrodc at gmail dot com> ---
> I thought I hit the important cases, but my non-minimal example still get=
s unnecessary register splits and stack spills, so maybe I missed places, o=
r perhaps there's another issue.

Adding the unroll pragma to the `Vector`'s operator + and *:

template<typename T, ptrdiff_t N>
[[gnu::always_inline]] constexpr auto operator+(Vector<T,N> x, Vector<T,N> =
y)
-> Vector<T,N> {
    Vector<T,N> z;
#pragma GCC unroll 16
    for (ptrdiff_t n =3D 0; n < Vector<T,N>::L; ++n) z.data[n] =3D x.data[n=
] +
y.data[n];
    return z;
}
template<typename T, ptrdiff_t N>
[[gnu::always_inline]] constexpr auto operator*(Vector<T,N> x, Vector<T,N> =
y)
-> Vector<T,N> {
    Vector<T,N> z;
#pragma GCC unroll 16
    for (ptrdiff_t n =3D 0; n < Vector<T,N>::L; ++n) z.data[n] =3D x.data[n=
] *
y.data[n];
    return z;
}
template<typename T, ptrdiff_t N>
[[gnu::always_inline]] constexpr auto operator+(T x, Vector<T,N> y) ->
Vector<T,N> {
    Vector<T,N> z;
#pragma GCC unroll 16
    for (ptrdiff_t n =3D 0; n < Vector<T,N>::L; ++n) z.data[n] =3D x + y.da=
ta[n];
    return z;
}
template<typename T, ptrdiff_t N>
[[gnu::always_inline]] constexpr auto operator*(T x, Vector<T,N> y) ->
Vector<T,N> {
    Vector<T,N> z;
#pragma GCC unroll 16
    for (ptrdiff_t n =3D 0; n < Vector<T,N>::L; ++n) z.data[n] =3D x * y.da=
ta[n];
    return z;
}


does not improve code generation (still get the same problem), so that's a
reproducer for such an issue.=