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 6A8E13858D37; Mon, 26 Jun 2023 18:48:36 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 6A8E13858D37
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1687805316;
	bh=3QA1GXr1uomMA8OB4RY65BNyctz/5DKHyIzhVkh11XM=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=bRkQQ+JfK7e3O6/gmA7FIbc3RThuhQEjX9XdIXw7/i5LR1dLTR6ZXDgcY1riNfS4/
	 oF31IzIQNCImK64U5dF3DG1XpMjGJx0r4RsMG6Bi2MCwWa8rx+Lgq5SEqnf04BqxEG
	 F4Sr2OWdMrlb40eXiSvBqbXSWkd29k43caXZDc7M=
From: "jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug c/102989] Implement C2x's n2763 (_BitInt)
Date: Mon, 26 Jun 2023 18:48:34 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: c
X-Bugzilla-Version: 12.0
X-Bugzilla-Keywords: 
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: jakub at gcc dot gnu.org
X-Bugzilla-Status: NEW
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-102989-4-HuMFR35REn@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-102989-4@http.gcc.gnu.org/bugzilla/>
References: <bug-102989-4@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=3D102989

--- Comment #70 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
For right shifts, I wonder if we shouldn't emit inline (perhaps with except=
ion
of -Os) something like:
__attribute__((noipa)) void
ashiftrt575 (unsigned long *p, unsigned long *q, int n)
{
  int prec =3D 575;
  int n1 =3D n & 63;
  int n2 =3D n / 64;
  int n3 =3D n1 !=3D 0;
  int n4 =3D (-n1) & 63;
  unsigned long ext;
  int i;
  for (i =3D n2; i < prec / 64 - n3; ++i)
    p[i - n2] =3D (q[i] >> n1) | (q[i + n3] << n4);
  ext =3D ((signed long) (q[prec / 64] << (64 - (prec & 63)))) >> (64 - (pr=
ec &
63));
  if (n1 && i =3D=3D prec / 64 - n3)
    {
      p[i - n2] =3D (q[i] >> n1) | (ext << n4);
      ++i;
    }
  i -=3D n2;
  p[i] =3D ((signed long) ext) >> n1;
  ext =3D ((signed long) ext) >> 63;
  for (++i; i < prec / 64 + 1; ++i)
    p[i] =3D ext;
}

__attribute__((noipa)) void
lshiftrt575 (unsigned long *p, unsigned long *q, int n)
{
  int prec =3D 575;
  int n1 =3D n & 63;
  int n2 =3D n / 64;
  int n3 =3D n1 !=3D 0;
  int n4 =3D (-n1) & 63;
  unsigned long ext;
  int i;
  for (i =3D n2; i < prec / 64 - n3; ++i)
    p[i - n2] =3D (q[i] >> n1) | (q[i + n3] << n4);
  ext =3D q[prec / 64] & ((1UL << (prec % 64)) - 1);
  if (n1 && i =3D=3D prec / 64 - n3)
    {
      p[i - n2] =3D (q[i] >> n1) | (ext << n4);
      ++i;
    }
  i -=3D n2;
  p[i] =3D ext >> n1;
  ext =3D 0;
  for (++i; i < prec / 64 + 1; ++i)
    p[i] =3D 0;
}
(for _BitInt(575) and 64-bit limb little endian).  If the shift count is
constant, it will allow further optimizations,
and if e.g. get_nonzero_bits tells us that n is variable but multiple of li=
mb
precision, we can optimize some more as well.
Looking at what LLVM does, they seem to sign extend in memory to twice as m=
any
bits and then just use an unrolled loop without any conditionals, but that
doesn't look well for memory usage etc.=