[Bug fortran/35285] New: Failures in the NIST test suite FM827 with -m64

[Bug fortran/35285] New: Failures in the NIST test suite FM827 with -m64

[dominiq at lps dot ens dot fr]

I noticed today (gcc version 4.4.0 20080221) two failures in FM827 when running
the NIST test suite with -m64. I have reduced the code to:

      PROGRAM FM827

        DOUBLE PRECISION AVD, BVD, CVD, DVD, DVCORR               
      IVPASS = 0
      IVFAIL = 0
      IVDELE = 0
      IVINSP = 0
      IVTOTN = 0
      ICZERO = 0
      I01 = 05
      I02 = 06
      NUVI = I02
CT007*  TEST 7                                           SQRT(X)**2 = X 
           IVTNUM = 7                                                   
        AVD = DSQRT(9.125D0) ** 2 - 9.125D0                             
           IF (AVD + 0.5000000000D-09) 20070, 10070, 40070              
40070      IF (AVD - 0.5000000000D-09) 10070, 10070, 20070              
10070      IVPASS = IVPASS + 1                                          
           WRITE (NUVI, 80002) IVTNUM                                   
           GO TO 0071                                                   
20070      IVFAIL = IVFAIL + 1                                          
           DVCORR = 0.0D+00                                             
           WRITE (NUVI, 80031) IVTNUM, AVD, DVCORR                      
 0071      CONTINUE                                                     
CT008*  TEST 8                                LN(X) = LN(10) * LOG10(X) 
           IVTNUM = 8                                                   
        BVD = 62.5D0 / 1000.0D0                                         
        AVD = DLOG10(BVD) * DLOG(10.0D0) - DLOG(0.0625D0)               
           IF (AVD + 0.5000000000D-09) 20080, 10080, 40080              
40080      IF (AVD - 0.5000000000D-09) 10080, 10080, 20080              
10080      IVPASS = IVPASS + 1                                          
           WRITE (NUVI, 80002) IVTNUM                                   
           GO TO 0081                                                   
20080      IVFAIL = IVFAIL + 1                                          
           DVCORR = 0.0D+00                                             
           WRITE (NUVI, 80031) IVTNUM, AVD, DVCORR                      
 0081      CONTINUE
CT009*  TEST 9                                    COSH**2 - SINH**2 = 1 
           IVTNUM = 9                                                   
        BVD = 0.125D0                                                   
        CVD = DSINH(2.125D0)                                            
        DVD = DCOSH(2.0D0 + BVD)                                        
        AVD = DVD  ** 2 - CVD ** 2 - DCOSH(0.0D0)                       
           IF (AVD + 0.5000000000D-09) 20090, 10090, 40090              
40090      IF (AVD - 0.5000000000D-09) 10090, 10090, 20090              
10090      IVPASS = IVPASS + 1                                          
           WRITE (NUVI, 80002) IVTNUM                                   
           GO TO 0091                                                   
20090      IVFAIL = IVFAIL + 1                                          
           DVCORR = 0.0D+00                                             
           WRITE (NUVI, 80031) IVTNUM, AVD, DVCORR                      
 0091      CONTINUE                                                     
CT010*  TEST 10                             TANH(X) = 1 - 2/(EXP(2X)+1) 
           IVTNUM = 10                                                  
        BVD = 5.0D0
        CVD = 2.0D0
        DVD = DLOG10(BVD * CVD) - DSQRT(4.0D0) /                        
     1  (DEXP(2.0D0 * (BVD - CVD)) + DCOS(0.0D0))                       
        AVD = DVD - DTANH(3.0D0)                                        
           IF (AVD + 0.5000000000D-09) 20100, 10100, 40100              
40100      IF (AVD - 0.5000000000D-09) 10100, 10100, 20100              
10100      IVPASS = IVPASS + 1                                          
           WRITE (NUVI, 80002) IVTNUM                                   
           GO TO 0101                                                   
20100      IVFAIL = IVFAIL + 1                                          
           DVCORR = 0.0D+00                                             
           WRITE (NUVI, 80031) IVTNUM, AVD, DVCORR                      
 0101      CONTINUE
80002 FORMAT (" ",2X,I3,4X," PASS  ",32X,A31)                           
80031 FORMAT (" ",2X,I3,4X," FAIL  ",/," ",16X,"COMPUTED= " ,           
     1D17.10,/," ",16X,"CORRECT=  " ,D17.10)                            
      STOP
      END

for which the executable gives:

     7     PASS  
     8     FAIL  
                 COMPUTED=  0.2772588722D+01
                 CORRECT=   0.0000000000D+00
     9     PASS  
    10     FAIL  
                 COMPUTED= -0.1000000000D+01
                 CORRECT=   0.0000000000D+00

I have tried to reduce the code further, but I cannot keep at the same time the
logic of the code and the failures. I don't get any failure on PPC. When the
tests fail, DLOG10(BVD)=0.0D0.

I am pretty sure to have run the NIST test suite with -m64 without error when
Jerry DeLisle proposed his first patch for PR35036 three weeks ago, now it
fails even with 4.2.3 and all the old gfortran I have tried. So I don't
understand what's happening. I don't know if this related, but in the last
weeks I have updated mpfr to 2.3.1 (from 2.3.0) and Darwin to 9.2 (from 9.1).


-- 
           Summary: Failures in the NIST test suite FM827 with -m64
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dominiq at lps dot ens dot fr
 GCC build triplet: i686-apple-darwin9
  GCC host triplet: i686-apple-darwin9
GCC target triplet: i686-apple-darwin9


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[jvdelisle at gcc dot gnu dot org]

------- Comment #1 from jvdelisle at gcc dot gnu dot org  2008-02-22 15:45 -------
Works for me.

Can you try this test case:

      program fm827

      double precision avd, bvd, cvd, dvd, dvcorr               
ct010*  test 10                             tanh(x) = 1 - 2/(exp(2x)+1) 
      bvd = 5.0d0
      cvd = 2.0d0
      dvd = dlog10(bvd * cvd) - dsqrt(4.0d0) /
     1 (dexp(2.0d0 * (bvd - cvd)) + dcos(0.0d0))
      avd = dvd - dtanh(3.0d0) 
      print *, "avd=",avd," should be: ", 0.0_8
      if (abs(avd) > 0.5000000000d-09) then
        print *, "fails"
      else
        print *, "passes"
      end if
      end


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[dominiq at lps dot ens dot fr]

------- Comment #2 from dominiq at lps dot ens dot fr  2008-02-22 16:01 -------
Subject: Re:  Failures in the NIST test suite FM827 with
 -m64

The test in comment #1 pass. I have forgotten to say that the arithmetic IF is
needed for the failure and that removing any one of the four left test cases
make the test to pass.

What would be helpful that people having a mac intel with 10.5.2 confirm
the failure and people with earlier OSX check if the test fails or not.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[dominiq at lps dot ens dot fr]

------- Comment #3 from dominiq at lps dot ens dot fr  2008-02-22 16:17 -------
Sorry, I have been too fast. The test in comment #1 fails with -m64.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[dominiq at lps dot ens dot fr]

------- Comment #4 from dominiq at lps dot ens dot fr  2008-02-22 16:21 -------
Forgot to give the result:

 avd=  -1.00000000000000       should be:    0.00000000000000     
 fails


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[jvdelisle at gcc dot gnu dot org]

------- Comment #5 from jvdelisle at gcc dot gnu dot org  2008-03-03 04:59 -------
Was it decided this is a platform specific library issue, not gfortran?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[fxcoudert at gcc dot gnu dot org]

------- Comment #6 from fxcoudert at gcc dot gnu dot org  2008-03-03 23:26 -------
(In reply to comment #5)
> Was it decided this is a platform specific library issue, not gfortran?

I think it's likely, but to be sure we need some more input. Dominique, could
you try to print out the values of the different parts of the calculation of
dvd and avd, to see which function call gives the wrong value?


-- 

fxcoudert at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fxcoudert at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[dominiq at lps dot ens dot fr]

------- Comment #7 from dominiq at lps dot ens dot fr  2008-03-04 21:45 -------
I have done some additional tests: compiled under 10.4.11, passed, then
executed under 10.5.2 failed.
Compiled on 10.5.2 with -S, then did the assembly and link under 10.4.11 (using
gfortran) then the executable passed the test. So this seems to be a bug in (or
exposed by) 10.5.2.

It would be nice to have a C test (if possible using gcc 4.0 from apple).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[fxcoudert at gcc dot gnu dot org]

------- Comment #8 from fxcoudert at gcc dot gnu dot org  2008-03-04 21:58 -------
(In reply to comment #7)
> It would be nice to have a C test (if possible using gcc 4.0 from apple).

It'll be rather easy if you can do what I suggested in comment #6: trim it down
a bit by printing the various quantities involved.


-- 

fxcoudert at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2008-03-04 21:58:33
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[dominiq at lps dot ens dot fr]

------- Comment #9 from dominiq at lps dot ens dot fr  2008-03-04 22:20 -------
(In reply to comment #8)

Sorry, for not replying to your comments. I have been out of town for four days
(remote place without connction to the net) and I am trying to clean up the
backlog.

To answer the question, when the test fail dlog10 returns 0, but to get that
you need some intermediate steps. The problem has the usual signature of
something reading/writing where it should not.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[fxcoudert at gcc dot gnu dot org]

------- Comment #10 from fxcoudert at gcc dot gnu dot org  2008-03-05 14:39 -------
(In reply to comment #9)
> To answer the question, when the test fail dlog10 returns 0, but to get that
> you need some intermediate steps.

Can you show us your program that shows dlog10 returning 0?

> The problem has the usual signature of
> something reading/writing where it should not.

Seems implausible, there's no array. I'm more tempted to suspect a library
numerical issue or an alignment issue.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[dominiq at lps dot ens dot fr]

------- Comment #11 from dominiq at lps dot ens dot fr  2008-03-05 15:29 -------
(In reply to comment #10)

[ibook-dhum] f90/bug% cat nist_827_red_2.f
      program fm827

      double precision avd, bvd, cvd, dvd, dvcorr               
      cvd = 10.0d0
      dvd = 6.0d0
      dvd = dlog10(cvd) - 2.0d0 /
     1 (dexp(dvd) + 1.0d0)
      avd = dvd - dtanh(3.0d0) 
      if (abs(avd) > 0.5000000000d-09) then
        print *, "fails"
        print *, "avd=",avd," should be: ", 0.0_8
        print *, dlog10(cvd), dexp(dvd)
      else
        print *, "passes"
      end if
      end
[ibook-dhum] f90/bug% gfc -m64 nist_827_red_2.f
[ibook-dhum] f90/bug% a.out
 fails
 avd= -1.00000000000000000       should be:    0.0000000000000000     
   0.0000000000000000       0.99506696128579508     

The first real of the last line is dlog10(cvd) with cvd = 10.0d0, so it should
be 1.0 and not 0.0.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[dominiq at lps dot ens dot fr]

------- Comment #12 from dominiq at lps dot ens dot fr  2008-03-05 20:45 -------
I have reported the bug to Apple, it is radar 5782719 with the following test:

[ibook-dhum] f90/bug% cat nist_827_c.c
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <float.h>
int main() {
    double avd, bvd, cvd, dvd, dvcorr;
    cvd = 10.0;
    dvd = 6.0;
    dvd = log10(cvd) - 2.0/(exp(dvd)+1.0);
    avd = dvd - tanh(3.0);
    if (abs(avd) > 0.5e-9)
       {
         printf("fails\n");
         printf("avd=%g, should be 0.0\n", avd);
         printf("%g, %g\n", log10(cvd), exp(dvd));
       }
    else printf("pass\n");
}
[ibook-dhum] f90/bug% gcc -m64 nist_827_c.c
[ibook-dhum] f90/bug% a.out
fails
avd=-1, should be 0.0
0, 0.995067
[ibook-dhum] f90/bug% gcc -v
Using built-in specs.
Target: i686-apple-darwin9
Configured with: /var/tmp/gcc/gcc-5465~16/src/configure --disable-checking
-enable-werror --prefix=/usr --mandir=/share/man
--enable-languages=c,objc,c++,obj-c++
--program-transform-name=/^[cg][^.-]*$/s/$/-4.0/
--with-gxx-include-dir=/include/c++/4.0.0 --with-slibdir=/usr/lib
--build=i686-apple-darwin9 --with-arch=apple --with-tune=generic
--host=i686-apple-darwin9 --target=i686-apple-darwin9
Thread model: posix
gcc version 4.0.1 (Apple Inc. build 5465)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285

[Bug fortran/35285] Failures in the NIST test suite FM827 with -m64

[fxcoudert at gcc dot gnu dot org]

------- Comment #13 from fxcoudert at gcc dot gnu dot org  2008-03-05 21:27 -------
(In reply to comment #12)
> I have reported the bug to Apple, it is radar 5782719

Closing.


-- 

fxcoudert at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35285