When
you enable a trap without providing a trap handler (a mechanism
for handling it), the trap causes a SIGFPE
signal. The signal, in turn, causes your program to abort with an
error message that is more or less informative, depending on the
method you use to enable the trap. If you want your program to abort
when it encounters a trap, then enabling the trap may be all you
want to do. However, you may want to handle the trap gracefully;
see “Handling Traps” for
information on trap handling.
HP 9000 systems provide several methods of enabling traps:
The +FP
compiler option (all compilers)
The fesettrapenable
function
The +fp_exception
(Fortran 90) and +T
(HP FORTRAN/9000) compile-line options
We discuss these briefly in the following subsections.
Using the +FP Compiler Option |
 |
The +FP
option, described in “Command-Line Mode Control: The
+FP Compiler Option”, allows you to enable traps from
the compiler command line. No change in your program is required.
The disadvantage of using this option is that you cannot know
exactly where in your program the exception occurred. Moreover,
unless you enable a trap for only one exception, you cannot know
the type of exception that occurred. You merely get a core dump.
The following Fortran program, for example, generates an overflow.
Example 6-1 Sample Program: overflow.f
PROGRAM OVERFLOW
DOUBLE PRECISION X, Y, Z
X = 1.79D308
Y = 2.2D-308
Z = X / Y
PRINT 30, X, Y, Z
30 FORMAT (1PE11.4, ' divided by', 1PE11.4, ' = ', 1PE11.4)
END |
If you compile this program with traps disabled (the default),
it produces the following output:
1.7900+308 divided by 2.2000-308 = +INF |
If you compile it with the +FP
option, however, you get a stack trace and core dump. (The O
flag of the +FP
option enables traps for overflow exceptions.)
$ f90 +FPO overflow.f
overflow.f
program OVERFLOW
11 Lines Compiled
$ ./a.out
( 0) 0xc1602c5c traceback + 0x14 [/usr/lib/pa1.1/libcl.1]
( 1) 0xc0126500 _sigreturn [/usr/lib/libc.1]
( 2) 0x000024b4 _start + 0x6c [./a.out]
Floating exception (core dumped) |
Using the fesettrapenable Function |
|
The fesettrapenable
function is part of the fenv(5) suite and
is described in detail in “Exception Bits”. It is provided in the C math library.
The fesettrapenable
routine can enable one or more traps in any combination.
Using fesettrapenable
to enable traps has the same disadvantages as the +FP
option. You get a core dump, and you cannot determine the exception
type. Moreover, the use of these routines lacks the simplicity of
the +FP option,
because you must modify your code in order to use them.
Here is the program from “Using the +FP Compiler Option”, modified to enable a trap for the
overflow exception using fesettrapenable.
It does this by using an !$HP$ ALIAS
or $ALIAS directive
to tell the Fortran compiler that the function's arguments
are passed by value.
Example 6-2 Sample Program: overflow_trap.f
PROGRAM OVERFLOW_TRAP
C F90:
!$HP$ ALIAS FESETTRAPENABLE = 'fesettrapenable' (%val)
C F77:
C $ALIAS FESETTRAPENABLE = 'fesettrapenable' (%val)
PARAMETER (FE_OVERFLOW = Z'20000000')
EXTERNAL FESETTRAPENABLE
DOUBLE PRECISION X, Y, Z
CALL FESETTRAPENABLE(FE_OVERFLOW)
X = 1.79D308
Y = 2.2D-308
Z = X / Y
WRITE(*,*) X, ' divided by', Y, ' = ', Z
END |
You do not need to specify any options when you compile this
program, but you need to link in the C math library. The result
is even less useful than in the preceding section:
$ f90 overflow_trap.f -lm
overflow_trap.f
program OVERFLOW_TRAP
13 Lines Compiled
$ ./a.out
Floating exception (core dumped) |
Using the +fp_exception or +T Compiler Option (Fortran
only) |
|
The +fp_exception
option, available with the HP Fortran 90 compiler, enables traps
for four IEEE exceptions: invalid operation, division by zero, overflow,
and underflow. (Few programs need to trap for the inexact exception,
which occurs often and conveys little information.) The HP FORTRAN/9000
equivalent is +T.
For Fortran applications, these
options have the simplicity of +FP.
They also provides more information than either +FP
or fesettrapenable.
They tell what kind of exception occurred and the virtual address
of the statement that triggered the exception. Moreover, they do
not cause a core dump.
For example, if you use +fp_exception
or +T to compile
the program in “Using the +FP Compiler Option”,
it produces a result like the following:
$ f77 +T overflow.f
overflow.f:
MAIN overflow:
$ ./a.out
PROGRAM ABORTED : IEEE overflow
PROCEDURE TRACEBACK:
( 0) 0x000024e8 _start + 0x88 [./a.out] |
The effect is similar with +fp_exception:
$ f90 +fp_exception overflow.f
overflow.f
program OVERFLOW
11 Lines Compiled
$ ./a.out
PROGRAM ABORTED : IEEE overflow
PROCEDURE TRACEBACK:
( 0) 0x000024ec _start + 0x74 [././a.out] |
See “Command-Line Mode Control: The
+FP Compiler Option” for information about how +fp_exception
and +T interact
with +FP if you
use them together. See the f90(1) man page
and the HP Fortran 90 Programmer's Reference
for details about using +fp_exception.
See the f77(1) man page and the HP
FORTRAN/9000 Programmer's Guide for details about using
+T.
Printable version
