[Bug fortran/18714] New: Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[Bug fortran/18714] New: Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[paul dot richard dot thomas at cea dot fr]

The full source is available by anonymous ftp from: 
ftp://ftp.cs.unm.edu/pub/smith-quetzal/codes.tar.Z
Directory lu (or I can provide directory lu, with gfc makefile)

The programme hangs in this routine, unless the print * , "" indicated below is 
included.

      SUBROUTINE SLAMC1( BETA, T, RND, IEEE1 )
*
*  -- LAPACK auxiliary routine (version 1.0) --
*     Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
*     Courant Institute, Argonne National Lab, and Rice University
*     February 29, 1992

      etc., etc. until

*
*+       WHILE( C.EQ.ONE )LOOP
   10    CONTINUE
         IF( C.EQ.ONE ) THEN
            A = 2*A
            C = SLAMC3( A, ONE )
            C = SLAMC3( C, -A )
            GO TO 10
         END IF
*+       END WHILE
*
*        Now compute  b = 2.0**m  with the smallest positive integer m
*        such that
*
*           fl( a + b ) .gt. a.
*
         B = 1
         C = SLAMC3( A, B )
         print *,""               !!!!!The code hangs without this line
                                  !!!!!Putting C = C does not cure the problem 
*
*+       WHILE( C.EQ.A )LOOP
   20    CONTINUE
         IF( C.EQ.A ) THEN
            B = 2*B
            C = SLAMC3( A, B )
            print *, C , C.EQ.A
            GO TO 20
         END IF
*+       END WHILE
*

and so on......

-- 
           Summary: Runtime hang in LAPACK routine SLAMC1 - in Quetzal
                    benchmark suite
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: fortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: paul dot richard dot thomas at cea dot fr
                CC: gcc-bugs at gcc dot gnu dot org,paulthomas2 at wanadoo
                    dot fr


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[paul dot richard dot thomas at cea dot fr]

------- Additional Comments From paul dot richard dot thomas at cea dot fr  2004-11-30 12:46 -------
(In reply to comment #0)

I should have said somewhere that I have been quite unable to isolate this 
problem with a short example.  Retaining the structure of the subprogramme 
results in code that compiles and runs properly.  (BTW g95 does the same...)

Paul Thomas



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[tobi at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|                            |19292
              nThis|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[giovannibajo at libero dot it]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[coudert at clipper dot ens dot fr]

------- Additional Comments From coudert at clipper dot ens dot fr  2005-01-28 12:19 -------
Did you try to compile this file without optimization (as is indicated in:
http://www.netlib.org/lapack/faq.html#1.26)? The code in the ?lam*.f files is
used to determine the machine precision (which could be done easily in Fortran >
77), and any level of optimization other than -O0 usually causes infinite loops.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |coudert at clipper dot ens
                   |                            |dot fr


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[prthomas at drfccad dot cea dot fr]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1682 bytes --]


------- Additional Comments From prthomas at drfccad dot cea dot fr  2005-01-28 17:02 -------
Subject: RE:  Runtime hang in LAPACK routine SLAMC1 - i
	n Quetzal benchmark suite

Bon soir François-Xavier,

Tu as raison!  Même -O1 fait coincer le programme.

Je suis étonné quand-même que le "fix" est d'injecter d'un PRINT *,
"n'importe quoi" et, aussi, qu'il est légèrement impossible isoler le bogue.
Tu as des explications là-dessous?

Je vais afficher une réponse sur Bugzilla, avant de sortir.

Paul  

-----Message d'origine-----
De : coudert at clipper dot ens dot fr [mailto:gcc-bugzilla@gcc.gnu.org]
Envoyé : vendredi 28 janvier 2005 13:20
À : paul.richard.thomas@cea.fr
Objet : [Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in
Quetzal benchmark suite



------- Additional Comments From coudert at clipper dot ens dot fr
2005-01-28 12:19 -------
Did you try to compile this file without optimization (as is indicated in:
http://www.netlib.org/lapack/faq.html#1.26)? The code in the ?lam*.f files
is
used to determine the machine precision (which could be done easily in
Fortran >
77), and any level of optimization other than -O0 usually causes infinite
loops.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |coudert at clipper dot ens
                   |                            |dot fr


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

------- You are receiving this mail because: -------
You reported the bug, or are watching the reporter.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[paulthomas2 at wanadoo dot fr]

------- Additional Comments From paulthomas2 at wanadoo dot fr  2005-01-28 17:08 -------
Francois-Xavier Coudert's comment is spot on.  Reducing the optimisation to -O0 
removes the need for this peculiar fix.

We could do with a repository for funnies that are not quite bugs.  I SAY 
REMOVE THIS ONE.

I am still left bemused, as I accidently posted in french, by the fact that a 
print statement releases the programme for any optimisation level and that when 
the section of code is isolated it runs perfectly.  Does anybody have an 
explanation?  

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[coudert at clipper dot ens dot fr]

------- Additional Comments From coudert at clipper dot ens dot fr  2005-01-28 17:20 -------
(reply to comments #3 and #4)

The print statement in the code induces I/O, which in some sense "disrupt the
flow of code" and has the effect of making unaware of possible optimizations. I
can't offer a deeper explanation than this statement, only dictated by
experience and not actual knowledge of the in-depths of compilers.

Removed the "wrong-code" keyword and the blocking of PR18714, since this is
actually not a bug.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|19292                       |
              nThis|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[coudert at clipper dot ens dot fr]

------- Additional Comments From coudert at clipper dot ens dot fr  2005-02-10 22:28 -------
Could this one be closed?

slmac1.f from the LAPACK library is designed to be compiled without any
optimization (it's stated in the FAQ and everywhere). Infinite loop is what is
expected to happen when optimization is turned on.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[paulthomas2 at wanadoo dot fr]

------- Additional Comments From paulthomas2 at wanadoo dot fr  2005-02-10 23:53 -------
Subject: Re:  Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

Yep, I am convinced.  I'm afraid that I missed the "everywhere"...

----- Original Message ----- 
From: "coudert at clipper dot ens dot fr" <gcc-bugzilla@gcc.gnu.org>
To: <paulthomas2@wanadoo.fr>
Sent: Thursday, February 10, 2005 11:28 PM
Subject: [Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in 
Quetzal benchmark suite


>
> ------- Additional Comments From coudert at clipper dot ens dot fr 
> 2005-02-10 22:28 -------
> Could this one be closed?
>
> slmac1.f from the LAPACK library is designed to be compiled without any
> optimization (it's stated in the FAQ and everywhere). Infinite loop is 
> what is
> expected to happen when optimization is turned on.
>
> -- 
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714
>
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.
> 




-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[paulthomas2 at wanadoo dot fr]

------- Additional Comments From paulthomas2 at wanadoo dot fr  2005-02-10 23:56 -------
Subject: Re:  Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

PS It seems to me that it is odd that any level of optimisation sends the 
programme into an infinite loop.  Both ifc and DF6 run find with the maximum 
level.  I have to agree, though, with the empirical result that this is so 
for gfortran! 




-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714

[Bug fortran/18714] Runtime hang in LAPACK routine SLAMC1 - in Quetzal benchmark suite

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2005-02-11 11:47 -------
Not a GCC bug.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18714