Seaborg Fortran compiler upgraded

From: Michael Stewart (pmstewart_at_lbl.gov)
Date: 11/15/2002

• Next message: Francesca Verdier: "use of Seaborg nodes for Supercomputing this week"
• Previous message: Francesca Verdier: "average review score in NIM"

On the afternoon of 11/14/02 the Fortran compiler on seaborg was 
upgraded to version 8.1.0.02 and several additional fixes were added. 
NERSC advises all seaborg users to recompile and relink all of their 
Fortran programs to take advantage of these fixes.

We found at least one instance where the previous compiler would give 
incorrect answers when -qintsize=8 was specified.  This seems to have 
been restricted to powers of complex numbers.

Here is an example showing the problem:

       program badpwr
! A simple test to show a bug in IBM power calculations
       implicit none
       complex z
       z=(1.,1.)
       print *,"complex z=",z," z**2=",z**2," should be (0,2)"
       end

With the previous compiler, z**2 was calculated incorrectly when the 
-qintsize=8 options was added to the compile line.  (If this option was 
not present the results were correct).

% xlf -o badpwr badpwr.f -qintsize=8
** badpwr   === End of Compilation 1 ===
1501-510  Compilation successful for file badpwr.f.
% ./badpwr
  complex z= (1.000000000,1.000000000)  z**2= 
(1.000000000,0.0000000000E+00)  should be (0,2)

The present compiler gives the correct result:

%  xlf -o badpwr badpwr.f -qintsize=8
** badpwr   === End of Compilation 1 ===
1501-510  Compilation successful for file badpwr.f.
seaborg:/usr/common/homes/m/mstewart/public/pmrs/40548% ./badpwr
  complex z= (1.000000000,1.000000000)  z**2= 
(0.0000000000E+00,2.000000000)  should be (0,2)


Please contact the NERSC consultants at consult@nersc.gov if you have 
any questions.

• Next message: Francesca Verdier: "use of Seaborg nodes for Supercomputing this week"
• Previous message: Francesca Verdier: "average review score in NIM"

This archive was generated by hypermail 2.1.6 : 08/21/2008 PDT