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 5A6AA3858D20; Tue,  3 Jan 2023 21:02:10 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 5A6AA3858D20
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1672779730;
	bh=6B94UaeR+rhkJJGvWNSGp6HCZmNuxMjf3vjiLU8Ql+A=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=xj9nSRAo7b81ZOEft4WgAHuNyyGp8B8GFDSc/LSzIDVhwexXqCUMoYWUh7U6Uv1S5
	 PEXNI8JOtHPUIfLj+9Mp2hI78QW5DARjZABT8ysOzhfx9AeGGsetYclTN0wv8SDYH/
	 SxNb3qarIyaJpVcaYkfGt6J0ruPXqYoHQ56kYug0=
From: "tkoenig at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug libgcc/108279] Improved speed for float128 routines
Date: Tue, 03 Jan 2023 21:02:09 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: libgcc
X-Bugzilla-Version: unknown
X-Bugzilla-Keywords: 
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: tkoenig at gcc dot gnu.org
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: attachments.created
Message-ID: <bug-108279-4-oisHkanNUQ@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-108279-4@http.gcc.gnu.org/bugzilla/>
References: <bug-108279-4@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=3D108279

--- Comment #1 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
Created attachment 54183
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=3D54183&action=3Dedit
Example patch with Michael S's code just pasted over the libgcc implementat=
ion,
for a test

A benchmarks: Just pasting over the code from the github
repo yields an improvement of gfortran's matmul by almost a factor of two,
so significant speedups are possible:

module tick
interface
function rdtsc() bind(C,name=3D"rdtsc")
use iso_c_binding
integer(kind=3Dc_long) :: rdtsc
end function rdtsc
end interface
end module tick

program main
use tick
use iso_c_binding
implicit none
integer, parameter :: wp =3D selected_real_kind(30)
! integer, parameter :: n=3D5000, p=3D4000, m=3D3666
integer, parameter :: n =3D 1000, p =3D 1000, m =3D 1000
real (kind=3Dwp) :: c(n,p), a(n,m), b(m, p)
character(len=3D80) :: line
integer(c_long) :: t1, t2, t3
real (kind=3Dwp) :: fl =3D 2.d0*n*m*p
integer :: i,j

print *,wp

line =3D '10 10'
call random_number(a)
call random_number(b)
t1 =3D rdtsc()
t2 =3D rdtsc()
t3 =3D t2-t1
print *,t3
t1 =3D rdtsc()
c =3D matmul(a,b)
t2 =3D rdtsc()
print *,1/(fl/(t2-t1-t3)),"Cycles per operation"
read (unit=3Dline,fmt=3D*) i,j
write (unit=3Dline,fmt=3D*) c(i,j)
end program main

showed

tkoenig@gcc188:~> ./original
16
32
^C
tkoenig@gcc188:~> time ./original
16
32
90.5696151959999999999999999999999997 Cycles per operation

real 1m2,148s
user 1m2,123s
sys 0m0,008s
tkoenig@gcc188:~> time ./modified
16
32
52.8148391719999999999999999999999957 Cycles per operation

real 0m36,296s
user 0m36,278s
sys 0m0,008s=20

where "original" is the current libgcc soft-float implementation, and
"modified" is with the code from the repro.

It does not handle exceptions, so this causes a few regressions, but certai=
nly
shows the potential=