[Bug target/25305] New: [4.0]: -O2 miscompiled fma3d in SPEC CPU 2K

[Bug target/25305] New: [4.0]: -O2 miscompiled fma3d in SPEC CPU 2K

[hjl at lucon dot org]

Gcc 4.0.3 20051207 miscompiles fma3d in SPEC CPU 2K when -O2 on i686 and
x86-64.
I got
Running Benchmarks
  Running 191.fma3d ref base o2 default
Child returned with invalid return code(0)
*** Miscompare of fma3d.out, see
/export/spec/src/2000/spec/benchspec/CFP2000/191.fma3d/run/00000002/fma3d.out.mis

Gcc 4.1 and 4.2 are OK.


-- 
           Summary: [4.0]: -O2 miscompiled fma3d in SPEC CPU 2K
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: hjl at lucon dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug target/25305] [4.0 regression]: -O2 miscompiled fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #1 from hjl at lucon dot org  2005-12-08 01:56 -------
It looks like a regression. Gcc 4.0.2 20050912 works fine.


-- 

hjl at lucon dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.0]: -O2 miscompiled fma3d|[4.0 regression]: -O2
                   |in SPEC CPU 2K              |miscompiled fma3d in SPEC
                   |                            |CPU 2K


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug target/25305] [4.0 regression]: -O2 miscompiled fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #2 from hjl at lucon dot org  2005-12-08 02:04 -------
Gcc 4.0.3 20051007 is also OK.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug target/25305] [4.0 regression]: -O2 miscompiled fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #3 from hjl at lucon dot org  2005-12-08 02:11 -------
Gcc 4.0.3 20051103 is also OK.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug target/25305] [4.0 regression]: -O2 miscompiled fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #4 from hjl at lucon dot org  2005-12-08 03:16 -------
Gcc 4.0.3 20051108 is OK.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug target/25305] [4.0 regression]: -O2 miscompiled fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #5 from hjl at lucon dot org  2005-12-08 03:28 -------
Gcc 4.0.3 20051111 is OK.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug target/25305] [4.0 regression]: -O2 miscompiled fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #6 from hjl at lucon dot org  2005-12-08 03:30 -------
Gcc 4.0.3 20051113 is bad. So this regression was introduced between 20051111
and 20051113.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #7 from hjl at lucon dot org  2005-12-08 06:55 -------
I have verified that

http://gcc.gnu.org/ml/gcc-patches/2005-11/msg00874.html

is the cause. Since gcc 4.1 and 4.2 are OK, the problem may be in the backport.


-- 

hjl at lucon dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jvdelisle at gcc dot gnu dot
                   |                            |org
          Component|target                      |libfortran
            Summary|[4.0 regression]: -O2       |[4.0 regression]: libfortran
                   |miscompiled fma3d in SPEC   |failed fma3d in SPEC CPU 2K
                   |CPU 2K                      |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

Re: [Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[Jerry DeLisle]

hjl at lucon dot org wrote:
> ------- Comment #7 from hjl at lucon dot org  2005-12-08 06:55 -------
> I have verified that
> 
> http://gcc.gnu.org/ml/gcc-patches/2005-11/msg00874.html
> 
> is the cause. Since gcc 4.1 and 4.2 are OK, the problem may be in the backport.
> 
> 
OK, I see now this is not a problem in 4.1 or 4.2.  Yes, backport problem.

Thanks,

Jerry

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #8 from hjl at lucon dot org  2005-12-08 07:32 -------
Revert

@@ -293,7 +292,7 @@ write_block (int length)
 {
   char *dest;

-  if (!is_internal_unit() && current_unit->bytes_left < length)
+  if (current_unit->bytes_left < length)
     {
       generate_error (ERROR_EOR, NULL);
       return NULL;

which wasn't even mentioned in ChangeLog, fixed this regression.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[jvdelisle at verizon dot net]

------- Comment #9 from jvdelisle at verizon dot net  2005-12-08 09:01 -------
Subject: Re:  [4.0 regression]: libfortran failed fma3d
 in SPEC CPU 2K

hjl at lucon dot org wrote:
> ------- Comment #7 from hjl at lucon dot org  2005-12-08 06:55 -------
> I have verified that
> 
> http://gcc.gnu.org/ml/gcc-patches/2005-11/msg00874.html
> 
> is the cause. Since gcc 4.1 and 4.2 are OK, the problem may be in the backport.
> 
> 
OK, I see now this is not a problem in 4.1 or 4.2.  Yes, backport problem.

Thanks,

Jerry


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #10 from hjl at lucon dot org  2005-12-12 15:04 -------
A patch is posted at

http://gcc.gnu.org/ml/gcc-patches/2005-12/msg00594.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[jvdelisle at gcc dot gnu dot org]

------- Comment #11 from jvdelisle at gcc dot gnu dot org  2005-12-19 07:25 -------
Fixed in 4.0, not a bug in 4.1 and 4.2


-- 

jvdelisle at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #12 from hjl at lucon dot org  2005-12-19 14:40 -------
Can you tell me which check in fixes this bug for 4.0?


-- 

hjl at lucon dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|FIXED                       |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[grigory_zagorodnev at linux dot intel dot com]

------- Comment #13 from grigory_zagorodnev at linux dot intel dot com  2005-12-19 14:41 -------
Here is the small reproducer, extracted from cpu2000/191.fma3d, so you can
update regression test base.
       CHARACTER RD*8
       CHARACTER WR*8
       CHARACTER(1) C1

       RD='N 1'
       READ (RD(1:2),*) C1

       WRITE (WR,*) 'passed'
       WRITE (*,*) WR
       END

Fails with the message:
        At line 8 of file foo.f90
        Fortran runtime error: End of record


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[jvdelisle at gcc dot gnu dot org]

------- Comment #14 from jvdelisle at gcc dot gnu dot org  2005-12-20 00:50 -------
My apologies, I was looking at the wrong place.

Grigory, thanks for test case.

I have regression tested the patch and see no new failures.  There are some
NIST failures, but these are not affected by the patch.  We have already fixed
those failures in 4.1 and 4.2.

I will commit the patch tonight as "obvious"

Again, sorry for the confusion.


-- 

jvdelisle at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |jvdelisle at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-12-20 00:50:48
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[jvdelisle at gcc dot gnu dot org]

------- Comment #15 from jvdelisle at gcc dot gnu dot org  2005-12-20 05:06 -------
Evidently I do not have authorization to commit to the 4.0 branch. (I tried
twice)  Here is a completed and ready testcase.  I strongly suggest people move
on to 4.1 if at all possible.  If someone else wants to commit this, please do.

! { dg-do run }
! PR25305 Check for no end-of-record error on write after read.
! Derived from example given in PR by Grigory Zagorodnev.
       CHARACTER RD*8
       CHARACTER WR*8
       CHARACTER(1) C1
       RD='N 1'
       READ (RD(1:2),*) C1
       IF (C1.NE.'N') CALL ABORT
       WRITE (WR,*) 'passed'
       IF (WR.NE.' passed') CALL ABORT()
       END


-- 

jvdelisle at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|jvdelisle at gcc dot gnu dot|unassigned at gcc dot gnu
                   |org                         |dot org
             Status|ASSIGNED                    |NEW


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[jvdelisle at gcc dot gnu dot org]

------- Comment #16 from jvdelisle at gcc dot gnu dot org  2005-12-20 06:06 -------
Subject: Bug 25305

Author: jvdelisle
Date: Tue Dec 20 06:06:28 2005
New Revision: 108844

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=108844
Log:
2005-12-20  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

        PR fortran/25305
        * io/transfer.c (write_block): Revert previous commit. 

Modified:
    branches/gcc-4_0-branch/libgfortran/ChangeLog
    branches/gcc-4_0-branch/libgfortran/io/transfer.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[jvdelisle at gcc dot gnu dot org]

------- Comment #17 from jvdelisle at gcc dot gnu dot org  2005-12-20 06:12 -------
Subject: Bug 25305

Author: jvdelisle
Date: Tue Dec 20 06:12:53 2005
New Revision: 108845

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=108845
Log:
2005-12-20  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

        PR libgfortran/25305
        * gfortran.dg/read_write.f: New test.

Added:
    branches/gcc-4_0-branch/gcc/testsuite/gfortran.dg/read_write.f
Modified:
    branches/gcc-4_0-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[jvdelisle at gcc dot gnu dot org]

------- Comment #18 from jvdelisle at gcc dot gnu dot org  2005-12-20 06:19 -------
Fixed 4.0 only.  Not a bug in 4.1 or 4.2


-- 

jvdelisle at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #19 from hjl at lucon dot org  2005-12-20 06:22 -------
Shouldn't that case also be added to 4.1 and mainline to prevent this bug
from happening there?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[steven at gcc dot gnu dot org]

------- Comment #20 from steven at gcc dot gnu dot org  2005-12-20 06:48 -------
HJ, see comment #18.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[hjl at lucon dot org]

------- Comment #21 from hjl at lucon dot org  2005-12-20 14:44 -------
Steven, see comment #1. I was talking about the testcase.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[kargl at gcc dot gnu dot org]

------- Comment #22 from kargl at gcc dot gnu dot org  2005-12-20 17:01 -------
The testcase isn't needed and should not be committed.
As explained elsewhere, the problem was caused by merging
one line from a 4.1 patch into 4.0 that should not have
been committed.  Jerry has fixed that problem.  With very
limited gfortran manpower, gfortran 4.0.x will see very few
if any new changes.

In fact, using gfortran 4.0.x to obtain SPEC benchmarks is
actually rather stupid.  Do yourself a favor and upgrade to
gfortran 4.1 (pre-release).  


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305

[Bug libfortran/25305] [4.0 regression]: libfortran failed fma3d in SPEC CPU 2K

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.0.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25305