From: Jeff Law <law@redhat.com>
To: Richard Guenther <richard.guenther@gmail.com>
Cc: Ian Lance Taylor <iant@google.com>, GCC <gcc@gcc.gnu.org>
Subject: Re: Latent bug in update_equiv_regs?
Date: Thu, 20 Aug 2009 14:27:00 -0000 [thread overview]
Message-ID: <4A8D519D.5020705@redhat.com> (raw)
In-Reply-To: <84fc9c000908200145l53d0e86dt47ddf3ecb8f02ee3@mail.gmail.com>
[-- Attachment #1: Type: text/plain, Size: 930 bytes --]
On 08/20/09 02:45, Richard Guenther wrote:
>
> It looks indeed bogus. Do you have a testcase at hand?
>
Compile the attached testcase with -O3 -mopenmp on i686-pc-linux-gnu.
Find MAIN__.omp_fn.2 in the .expand dump.
Within that function, you're looking for this sequence of insns:
;; D.3137_29 = REALPART_EXPR <.omp_data_i_55(D)->c>;
(insn 242 241 0 j.f90:138 (set (reg:SF 74 [ D.3137 ])
(mem/s:SF (plus:SI (reg/f:SI 247 [ .omp_data_i ])
(const_int 32 [0x20])) [2 .omp_data_i_55(D)->c+0 S4
A64])) -1 (nil))
[ ... ]
(insn 247 246 0 j.f90:138 (set (mem/s:SF (plus:SI (reg/f:SI 247 [
.omp_data_i ])
(const_int 32 [0x20])) [13 S4 A64])
(reg:SF 351)) -1 (nil))
This doesn't cause a libgomp testsuite failure in the mainline because
(reg:SF 74) gets a hard reg and thus reload doesn't utilize the bogus
REG_EQUIV note that eventually gets added to insn 242.
Jeff
[-- Attachment #2: reduction1.f90 --]
[-- Type: text/plain, Size: 4309 bytes --]
! { dg-do run }
!$ use omp_lib
integer :: i, ia (6), n, cnt
real :: r, ra (4)
double precision :: d, da (5)
complex :: c, ca (3)
logical :: v
i = 1
ia = 2
r = 3
ra = 4
d = 5.5
da = 6.5
c = cmplx (7.5, 1.5)
ca = cmplx (8.5, -3.0)
v = .false.
cnt = -1
!$omp parallel num_threads (3) private (n) reduction (.or.:v) &
!$omp & reduction (+:i, ia, r, ra, d, da, c, ca)
!$ if (i .ne. 0 .or. any (ia .ne. 0)) v = .true.
!$ if (r .ne. 0 .or. any (ra .ne. 0)) v = .true.
!$ if (d .ne. 0 .or. any (da .ne. 0)) v = .true.
!$ if (c .ne. cmplx (0) .or. any (ca .ne. cmplx (0))) v = .true.
n = omp_get_thread_num ()
if (n .eq. 0) then
cnt = omp_get_num_threads ()
i = 4
ia(3:5) = -2
r = 5
ra(1:2) = 6.5
d = -2.5
da(2:4) = 8.5
c = cmplx (2.5, -3.5)
ca(1) = cmplx (4.5, 5)
else if (n .eq. 1) then
i = 2
ia(4:6) = 5
r = 1
ra(2:4) = -1.5
d = 8.5
da(1:3) = 2.5
c = cmplx (0.5, -3)
ca(2:3) = cmplx (-1, 6)
else
i = 1
ia = 1
r = -1
ra = -1
d = 1
da = -1
c = 1
ca = cmplx (-1, 0)
end if
!$omp end parallel
if (v) call abort
if (cnt .eq. 3) then
if (i .ne. 8 .or. any (ia .ne. (/3, 3, 1, 6, 6, 8/))) call abort
if (r .ne. 8 .or. any (ra .ne. (/9.5, 8.0, 1.5, 1.5/))) call abort
if (d .ne. 12.5 .or. any (da .ne. (/8.0, 16.5, 16.5, 14.0, 5.5/))) call abort
if (c .ne. cmplx (11.5, -5)) call abort
if (ca(1) .ne. cmplx (12, 2)) call abort
if (ca(2) .ne. cmplx (6.5, 3) .or. ca(2) .ne. ca(3)) call abort
end if
i = 1
ia = 2
r = 3
ra = 4
d = 5.5
da = 6.5
c = cmplx (7.5, 1.5)
ca = cmplx (8.5, -3.0)
v = .false.
cnt = -1
!$omp parallel num_threads (3) private (n) reduction (.or.:v) &
!$omp & reduction (-:i, ia, r, ra, d, da, c, ca)
!$ if (i .ne. 0 .or. any (ia .ne. 0)) v = .true.
!$ if (r .ne. 0 .or. any (ra .ne. 0)) v = .true.
!$ if (d .ne. 0 .or. any (da .ne. 0)) v = .true.
!$ if (c .ne. cmplx (0) .or. any (ca .ne. cmplx (0))) v = .true.
n = omp_get_thread_num ()
if (n .eq. 0) then
cnt = omp_get_num_threads ()
i = 4
ia(3:5) = -2
r = 5
ra(1:2) = 6.5
d = -2.5
da(2:4) = 8.5
c = cmplx (2.5, -3.5)
ca(1) = cmplx (4.5, 5)
else if (n .eq. 1) then
i = 2
ia(4:6) = 5
r = 1
ra(2:4) = -1.5
d = 8.5
da(1:3) = 2.5
c = cmplx (0.5, -3)
ca(2:3) = cmplx (-1, 6)
else
i = 1
ia = 1
r = -1
ra = -1
d = 1
da = -1
c = 1
ca = cmplx (-1, 0)
end if
!$omp end parallel
if (v) call abort
if (cnt .eq. 3) then
if (i .ne. 8 .or. any (ia .ne. (/3, 3, 1, 6, 6, 8/))) call abort
if (r .ne. 8 .or. any (ra .ne. (/9.5, 8.0, 1.5, 1.5/))) call abort
if (d .ne. 12.5 .or. any (da .ne. (/8.0, 16.5, 16.5, 14.0, 5.5/))) call abort
if (c .ne. cmplx (11.5, -5)) call abort
if (ca(1) .ne. cmplx (12, 2)) call abort
if (ca(2) .ne. cmplx (6.5, 3) .or. ca(2) .ne. ca(3)) call abort
end if
i = 1
ia = 2
r = 4
ra = 8
d = 16
da = 32
c = 2
ca = cmplx (0, 2)
v = .false.
cnt = -1
!$omp parallel num_threads (3) private (n) reduction (.or.:v) &
!$omp & reduction (*:i, ia, r, ra, d, da, c, ca)
!$ if (i .ne. 1 .or. any (ia .ne. 1)) v = .true.
!$ if (r .ne. 1 .or. any (ra .ne. 1)) v = .true.
!$ if (d .ne. 1 .or. any (da .ne. 1)) v = .true.
!$ if (c .ne. cmplx (1) .or. any (ca .ne. cmplx (1))) v = .true.
n = omp_get_thread_num ()
if (n .eq. 0) then
cnt = omp_get_num_threads ()
i = 3
ia(3:5) = 2
r = 0.5
ra(1:2) = 2
d = -1
da(2:4) = -2
c = 2.5
ca(1) = cmplx (-5, 0)
else if (n .eq. 1) then
i = 2
ia(4:6) = -2
r = 8
ra(2:4) = -0.5
da(1:3) = -1
c = -3
ca(2:3) = cmplx (0, -1)
else
ia = 2
r = 0.5
ra = 0.25
d = 2.5
da = -1
c = cmplx (0, -1)
ca = cmplx (-1, 0)
end if
!$omp end parallel
if (v) call abort
if (cnt .eq. 3) then
if (i .ne. 6 .or. any (ia .ne. (/4, 4, 8, -16, -16, -8/))) call abort
if (r .ne. 8 .or. any (ra .ne. (/4., -2., -1., -1./))) call abort
if (d .ne. -40 .or. any (da .ne. (/32., -64., -64., 64., -32./))) call abort
if (c .ne. cmplx (0, 15)) call abort
if (ca(1) .ne. cmplx (0, 10)) call abort
if (ca(2) .ne. cmplx (-2, 0) .or. ca(2) .ne. ca(3)) call abort
end if
end
next prev parent reply other threads:[~2009-08-20 13:37 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-08-20 0:01 Jeff Law
2009-08-20 0:24 ` Ian Lance Taylor
2009-08-20 1:20 ` Jeff Law
2009-08-20 4:48 ` Ian Lance Taylor
2009-08-20 14:03 ` Jeff Law
2009-08-20 9:48 ` Richard Guenther
2009-08-20 14:27 ` Jeff Law [this message]
2009-08-20 15:11 ` Richard Guenther
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4A8D519D.5020705@redhat.com \
--to=law@redhat.com \
--cc=gcc@gcc.gnu.org \
--cc=iant@google.com \
--cc=richard.guenther@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).