From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugs-return-445189-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 27470 invoked by alias); 1 Mar 2014 22:41:40 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Received: (qmail 27154 invoked by uid 48); 1 Mar 2014 22:41:37 -0000
From: "glisse at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/58897] Improve 128/64 division
Date: Sat, 01 Mar 2014 22:41:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: target
X-Bugzilla-Version: 4.9.0
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: glisse at gcc dot gnu.org
X-Bugzilla-Status: UNCONFIRMED
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-58897-4-wL9SdJhKjq@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58897-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58897-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2014-03/txt/msg00058.txt.bz2

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58897

--- Comment #1 from Marc Glisse <glisse at gcc dot gnu.org> ---
Untested, but this shows what the 128/64 division could look like (with obvious
variants for mod and divmod). I don't really know how to get gcc to generate
anything like that though. I have seen udivmodtidi3 in s390 but that seems to
return (ql,r) packed in an __int128, not what I want. And gcc doesn't show
ul->ux as a zero_extend and forgets extremely fast the REG_EQUAL note on the
call to __udivti3, so I can't rely on combine.

typedef unsigned __int128 ux;
typedef unsigned long ul;
#define udiv_qrnnd(q, r, n1, n0, dx) \
  __asm__ ("divq %4" : "=a" (q), "=d" (r) \
      : "0" ((ul)(n0)), "1" ((ul)(n1)), "rm" ((ul)(dx)))
ux div128by64(ux a, ul b){
  ul ah = a >> 64;
  ul al = a;
  ul qh, ql, r1, r;
  qh=ah/b; r1=ah%b;
  udiv_qrnnd(ql,r,r1,al,b);
  return (ux)qh << 64 | ql;
}

/*
a=ah*2^64+al
ah=qh*b+r1
a=qh*2^64*b+r1*2^64+al
r1*2^64+al<2^64*b
r1*2^64+al=ql*b+r
a=(qh*2^64+ql)*b+r
*/