| Arithmetic error from NR_MACHAR() [message #12177] |
Mon, 29 June 1998 00:00  |
Heiko H�nnefeld
Messages: 7
Registered: June 1998
|
Junior Member |
|
|
Hi all,
I tried to find out the machine precision on my LINUX PC with IDL 5.0.
The result is the following:
IDL> help, /struc, NR_MACHAR(/DOUBLE)
** Structure DMACHAR, 13 tags, length=68:
IBETA LONG 2
IT LONG 64
IRND LONG 2
NGRD LONG 0
MACHEP LONG -63
NEGEP LONG -64
IEXP LONG 2
MINEXP LONG -1075
MAXEXP LONG -1073
EPS DOUBLE 1.0842022e-19
EPSNEG DOUBLE 5.4210109e-20
XMIN DOUBLE 0.0000000
XMAX DOUBLE Inf
% Program caused arithmetic error: Floating divide by 0
It seems, that MAXEXP has got the wrong value.
Also XMAX, which is according to the manual calculated as
(1-EPSNEG)*IBETA^MAXEXP, is Infinity, what may be the origin of the
error message.
But then I tried this:
IDL> prec = NR_MACHAR(/DOUBLE)
% Program caused arithmetic error: Floating divide by 0
IDL> print, (1-prec.EPSNEG)*prec.IBETA^prec.MAXEXP
0.0000000
There I dont get 'Infinity' as result, but '0.0'.
If, however, I try the same on a UNIX-system, first everything seems ok.
IDL> help, /struc, NR_MACHAR(/DOUBLE)
** Structure DMACHAR, 13 tags, length=72:
IBETA LONG 2
IT LONG 53
IRND LONG 2
NGRD LONG 0
MACHEP LONG -52
NEGEP LONG -53
IEXP LONG 11
MINEXP LONG -1022
MAXEXP LONG 1024
EPS DOUBLE 2.2204460e-16
EPSNEG DOUBLE 1.1102230e-16
XMIN DOUBLE 2.2250739e-308
XMAX DOUBLE 1.7976931e+308
But - if I calculate XMAX myself again - the result is 0.0:
IDL> prec = NR_MACHAR(/DOUBLE)
IDL> print, (1-prec.EPSNEG)*prec.IBETA^prec.MAXEXP
0.0000000
If I use MACHAR() instead of NR_MACHAR(), exactly the same happens.
The keyword DOUBLE also is not the origin of the problem.
Maybe anybody else understands, what happens?
Thanks,
Heiko
--
Heiko H�nnefeld
HASYLAB / DESY
Notkestr. 85
22603 Hamburg
Tel.: 040 / 8998-2698
Fax.: 040 / 8998-2787
e-mail: hhuenne@desy.de
|
|
|
|
| Re: Arithmetic error [message #29073 is a reply to message #12177] |
Fri, 01 February 2002 02:12  |
Nigel Wade
Messages: 286
Registered: March 1998
|
Senior Member |
|
|
Ed Wright wrote:
> in article a3b58p$kb2i$1@rook.le.ac.uk, Nigel Wade at nmw@ion.le.ac.uk
> wrote on 1/31/02 2:11 AM:
>
>> Ed Wright wrote:
>>
>>>
>>> To: IDL
>>> From: Ed Wright, JPL
>>>
>>> I require assistance with an odd problem. My current projects involves
>>> creating a dlm module of some 200+ routines from a library of 1500+
>>> routines. Several of the IDL callable routines replicate intrinsic IDL
>>> calls. While testing one such routine that calculates determinate of a
>>> 3X3 I encountered a problem. The numerical result between my determinant
>>> function and determn matches to 10^(-16).
>>>
>>> Now the odd thing. Under certain circumstances the use of my determinant
>>> function as an argument in another routine call causes an error:
>>>
>>> % Program caused arithmetic error: Floating illegal operand.
>>>
>>
>> That error means somewhere in your code you have generated a NaN (Not a
>> Number). The operations which generate NaNs include 0/0, Inf*0, sqrt(<0)
>> Inf-Inf,Inf/Inf (there may be others...).
>
> By code, should I take it to mean the external C code?
>
It could be either. Both IDL and C would generate the same floating point
exception. I can't offhand remember whether an exception in your C code
would generate this exception in IDL. But it might be that the C code
generates a Nan/Inf and then IDL generates the exception when it tries to
use it.
>>
>> Since you don't have a preceding 'Floating divide by zero', it's a fair
>> bet your error doesn't involve Inf. So, all you have to do is look
>> through your code to find where you might have a 0/0 or sqrt(<0). ;-)
>
> Oh, is that all. No problem. 300k lines - let me at them.... Mr.
> Codebuster, that's me.
Debugging external code is always fun.
I forsee extensive use of printf...
--
-----------------------------------------------------------
Nigel Wade, System Administrator, Space Plasma Physics Group,
University of Leicester, Leicester, LE1 7RH, UK
E-mail : nmw@ion.le.ac.uk
Phone : +44 (0)116 2523568, Fax : +44 (0)116 2523555
|
|
|
|
| Re: Arithmetic error [message #29097 is a reply to message #12177] |
Thu, 31 January 2002 10:41 |
Paul van Delst
Messages: 364
Registered: March 1997
|
Senior Member |
|
|
Ed Wright wrote:
>
> in article a3b58p$kb2i$1@rook.le.ac.uk, Nigel Wade at nmw@ion.le.ac.uk wrote
> on 1/31/02 2:11 AM:
>
>> Ed Wright wrote:
>>
>>>
>>> To: IDL
>>> From: Ed Wright, JPL
>>>
>>> I require assistance with an odd problem. My current projects involves
>>> creating a dlm module of some 200+ routines from a library of 1500+
>>> routines. Several of the IDL callable routines replicate intrinsic IDL
>>> calls. While testing one such routine that calculates determinate of a 3X3
>>> I encountered a problem. The numerical result between my determinant
>>> function and determn matches to 10^(-16).
>>>
>>> Now the odd thing. Under certain circumstances the use of my determinant
>>> function as an argument in another routine call causes an error:
>>>
>>> % Program caused arithmetic error: Floating illegal operand.
>>>
>>
>> That error means somewhere in your code you have generated a NaN (Not a
>> Number). The operations which generate NaNs include 0/0, Inf*0, sqrt(<0)
>> Inf-Inf,Inf/Inf (there may be others...).
>
> By code, should I take it to mean the external C code?
>
>>
>> Since you don't have a preceding 'Floating divide by zero', it's a fair bet
>> your error doesn't involve Inf. So, all you have to do is look through your
>> code to find where you might have a 0/0 or sqrt(<0). ;-)
>
> Oh, is that all. No problem. 300k lines - let me at them.... Mr. Codebuster,
> that's me.
Try !EXCEPT=2 and let IDL do the searching for you. Be prepared for a lot of output.....
paulv
--
Paul van Delst Religious and cultural
CIMSS @ NOAA/NCEP purity is a fundamentalist
Ph: (301)763-8000 x7274 fantasy
Fax:(301)763-8545 V.S.Naipaul
|
|
|
|
| Re: Arithmetic error [message #29104 is a reply to message #12177] |
Thu, 31 January 2002 07:53 |
Ed Wright
Messages: 39
Registered: February 1999
|
Member |
|
|
in article a3b58p$kb2i$1@rook.le.ac.uk, Nigel Wade at nmw@ion.le.ac.uk wrote
on 1/31/02 2:11 AM:
> Ed Wright wrote:
>
>>
>> To: IDL
>> From: Ed Wright, JPL
>>
>> I require assistance with an odd problem. My current projects involves
>> creating a dlm module of some 200+ routines from a library of 1500+
>> routines. Several of the IDL callable routines replicate intrinsic IDL
>> calls. While testing one such routine that calculates determinate of a 3X3
>> I encountered a problem. The numerical result between my determinant
>> function and determn matches to 10^(-16).
>>
>> Now the odd thing. Under certain circumstances the use of my determinant
>> function as an argument in another routine call causes an error:
>>
>> % Program caused arithmetic error: Floating illegal operand.
>>
>
> That error means somewhere in your code you have generated a NaN (Not a
> Number). The operations which generate NaNs include 0/0, Inf*0, sqrt(<0)
> Inf-Inf,Inf/Inf (there may be others...).
By code, should I take it to mean the external C code?
>
> Since you don't have a preceding 'Floating divide by zero', it's a fair bet
> your error doesn't involve Inf. So, all you have to do is look through your
> code to find where you might have a 0/0 or sqrt(<0). ;-)
Oh, is that all. No problem. 300k lines - let me at them.... Mr. Codebuster,
that's me.
As always,
Ed Wright
|
|
|
|
| Re: Arithmetic error [message #29107 is a reply to message #12177] |
Thu, 31 January 2002 02:11 |
Nigel Wade
Messages: 286
Registered: March 1998
|
Senior Member |
|
|
Ed Wright wrote:
>
> To: IDL
> From: Ed Wright, JPL
>
> I require assistance with an odd problem. My current projects involves
> creating a dlm module of some 200+ routines from a library of 1500+
> routines. Several of the IDL callable routines replicate intrinsic IDL
> calls. While testing one such routine that calculates determinate of a 3X3
> I encountered a problem. The numerical result between my determinant
> function and determn matches to 10^(-16).
>
> Now the odd thing. Under certain circumstances the use of my determinant
> function as an argument in another routine call causes an error:
>
> % Program caused arithmetic error: Floating illegal operand.
>
That error means somewhere in your code you have generated a NaN (Not a
Number). The operations which generate NaNs include 0/0, Inf*0, sqrt(<0)
Inf-Inf,Inf/Inf (there may be others...).
Since you don't have a preceding 'Floating divide by zero', it's a fair bet
your error doesn't involve Inf. So, all you have to do is look through your
code to find where you might have a 0/0 or sqrt(<0). ;-)
--
-----------------------------------------------------------
Nigel Wade, System Administrator, Space Plasma Physics Group,
University of Leicester, Leicester, LE1 7RH, UK
E-mail : nmw@ion.le.ac.uk
Phone : +44 (0)116 2523568, Fax : +44 (0)116 2523555
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
