qt
Joined: 23 Aug 2005 Posts: 46 Location: Berlin, Germany
|
Posted: Mon Aug 29, 2005 4:52 am Post subject: invalid floaiting point operation |
|
|
Dear Paul,
A customer reported a problem with FTN95
v4.7. He found this problem when comparing results which
were calculated by different compilers (Lahey LF90, LF95,
FTN77) against FTN95 whose result differed significantly
from the others.
Please check this test program:
PROGRAM TEST
IMPLICIT REAL(A - Z)
INTEGER WSP, Z2
COMPLEX H0, H1
Z2 = 1
WSP = 8
!in SDBG: invalid floaiting point operation follows
H0 = (1.0,0.0) * (Z2-1) * WSP
H1 = (1.0,0.0) * REAL((Z2-1) * WSP)
PRINT*, '(1.0,0.0) * (Z2-1) * WSP = ', H0
PRINT*, '(1.0,0.0) * REAL((Z2-1) * WSP) = ', H1
END
I have run this through v4.8 and can confirm that the problem is also there with
this release.
When debugging you see the following at the line "invalid
floaiting point operation"
when stepping to the line
H0 = (1.0,0.0) * (Z2-1) * WSP
Changing this to
H0 = (1.0,0.0) * REAL((Z2-1) * WSP)
removes the problem, but I think FTN95 should handle the
first statement too. Please check. Thank you.
Kind regards
Joerg Kuthe
------------------------------------------------------------
--
QT software GmbH Gaeblerstrasse 2 D-13086
Berlin
Tel. 030/9290087-0 Fax. -2 eMail:
admin@qtsoftware.de
WWW http://www.qtsoftware.de
------------------------------------------------------------
--
|
|