| print, long(1.0e10) [message #16541] |
Tue, 03 August 1999 00:00  |
Gary Fu
Messages: 9
Registered: April 1997
|
Junior Member |
|
|
Hi,
I got different results from the following example on SGI IRIX and PC
Linux:
a = 1.0e10
b = long(a)
print, b ; 2147483647 for IRIX, -2147483648 for Linux
It looks like that IRIX and Linux handle it differently when assigning a
floating value greater (or less) than the boundary of a long integer to
a long variable. Should I always check the range of 'a' before
assigning it to a 'long' variable to make the result consistent on IRIX
and Linux ?
TIA,
Gary
--
*****************************************
* Gary Fu, GSC (301) 286-7107 *
* email : "gfu@seadas.gsfc.nasa.gov" *
* NASA/Goddard Space Flight Center *
*****************************************
|
|
|
|
| Re: print, long(1.0e10) [message #16620 is a reply to message #16541] |
Wed, 04 August 1999 00:00  |
thompson
Messages: 584
Registered: August 1991
|
Senior Member |
|
|
Gary Fu <gfu@seadas.gsfc.nasa.gov> writes:
> Hi,
> I got different results from the following example on SGI IRIX and PC
> Linux:
> a = 1.0e10
> b = long(a)
> print, b ; 2147483647 for IRIX, -2147483648 for Linux
> It looks like that IRIX and Linux handle it differently when assigning a
> floating value greater (or less) than the boundary of a long integer to
> a long variable. Should I always check the range of 'a' before
> assigning it to a 'long' variable to make the result consistent on IRIX
> and Linux ?
I get yet a different result with IDL/v5.2 on Digital Unix.
IDL> a = 1.0e10
IDL> b = long(a)
% Program caused arithmetic error: Floating illegal operand
% Detected at $MAIN$
IDL> print, b
1410065408
This is a completely sensible result, since it's equal to (A mod 2.^32),
i.e. the 32 lowest most bits of A expressed as an integer.
Since different machines seem to act in different ways, it's best to catch the
error, either by testing the value beforehand, or by using something like
CHECK_MATH.
William Thompson
|
|
|
|
| Re: print, long(1.0e10) [message #16626 is a reply to message #16541] |
Wed, 04 August 1999 00:00 |
Liam Gumley
Messages: 473
Registered: November 1994
|
Senior Member |
|
|
Gary Fu wrote:
> I got different results from the following example on SGI IRIX and PC
> Linux:
> a = 1.0e10
> b = long(a)
> print, b ; 2147483647 for IRIX, -2147483648 for Linux
>
> It looks like that IRIX and Linux handle it differently when assigning a
> floating value greater (or less) than the boundary of a long integer to
> a long variable. Should I always check the range of 'a' before
> assigning it to a 'long' variable to make the result consistent on IRIX
> and Linux ?
Gary,
I guess it depends on what you are trying to do. If you want to
interpret the binary representation of 1.0e10 as a 32-bit signed LONG,
then the syntax is
result = long(1.0e10, 0)
However if you want to represent 1.0e10 as an integer value, I think the
only option (other than a range check) is to use IDL 5.2, which allows
64-bit signed and unsigned integers, e.g.
result = long64(1.0e10)
Cheers,
Liam.
--
Liam E. Gumley
Space Science and Engineering Center, UW-Madison
http://cimss.ssec.wisc.edu/~gumley
|
|
|
|
| Re: print, long(1.0e10) [message #16644 is a reply to message #16541] |
Tue, 10 August 1999 00:00 |
R.Bauer
Messages: 1424
Registered: November 1998
|
Senior Member |
|
|
Gary Fu wrote:
> Hi,
>
> I got different results from the following example on SGI IRIX and PC
> Linux:
> a = 1.0e10
> b = long(a)
> print, b ; 2147483647 for IRIX, -2147483648 for Linux
>
Results of WIN NT
IDL> a=1.0e10
IDL> b = long(a)
IDL> print, b
1410065408
Results of AIX
IDL> a=1.0e10
IDL> b=long(a)
% Program caused arithmetic error: Floating illegal operand
% Program caused arithmetic error: Conversion to integer error
IDL> print,b
2147483647
R.Bauer
|
|
|
|
| Re: print, long(1.0e10) [message #16659 is a reply to message #16541] |
Mon, 09 August 1999 00:00 |
luthi
Messages: 20
Registered: March 1999
|
Junior Member |
|
|
Hold on
> I got different results from the following example on SGI IRIX and PC
> Linux:
> a = 1.0e10
> b = long(a)
> print, b ; 2147483647 for IRIX, -2147483648 for Linux
The same commands evaluated with PV-Wave (6.21, Solaris 2.6) gives
% Program caused arithmetic error: Floating divide by 0
% Program caused arithmetic error: Floating underflow
% Program caused arithmetic error: Floating illegal operand
% Detected at $MAIN$ (LONG).
print,b
1343554297
While I don't understand the "divide by 0" error at least there is some error
messge. Is this due to the wonderful world of IEEE numerics?
M. Luethi
--
============================================================
Martin Luethi Tel. +41 1 632 40 92
Glaciology Section Fax. +41 1 632 11 92
VAW ETH Zuerich
CH-8092 Zuerich mail luthi@vaw.baum.ethz.ch
Switzerland
============================================================
|
|
|
|
| Re: print, long(1.0e10) [message #16660 is a reply to message #16541] |
Mon, 09 August 1999 00:00 |
thompson
Messages: 584
Registered: August 1991
|
Senior Member |
|
|
Gary Fu <gfu@seadas.gsfc.nasa.gov> writes:
> I got different results from the following example on SGI IRIX and PC
> Linux:
> a = 1.0e10
> b = long(a)
> print, b ; 2147483647 for IRIX, -2147483648 for Linux
> It looks like that IRIX and Linux handle it differently when assigning a
> floating value greater (or less) than the boundary of a long integer to
> a long variable. Should I always check the range of 'a' before
> assigning it to a 'long' variable to make the result consistent on IRIX
> and Linux ?
It appears that Digital Unix gives yet another answer:
IDL> a = 1.0e10
IDL> b = long(a)
% Program caused arithmetic error: Floating illegal operand
IDL> print, b
1410065408
which makes sense, since that's A modulo 2^32.
William Thompson
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
