why "NAN"

ztdep · 11-13-2006, 04:27 AM

Hello:
I encounter some problem when i run my fortran program on the linu WS 4(64bit) .
I compile the program with g95. and my cpu is intel xeon 64bit
real*8 var1=0.2539998000
real*8 var2=0.2539970307
when i compute the var1-var2, it gives "NaN"
So what is the reason of getting a NaN.
Regards

Ohesceayar · 11-13-2006, 01:30 PM

The only possible reason I can think of is that your var1 and var2 variables are not being initialized to what you think they are being initialized to.

Try putting a print statement before the calculation, I bet one or both are already NAN even before the subtraction.

asgeirss · 11-15-2006, 01:19 AM

You are setting both var1 and var2 to 8 bits, but you are trying to put values with too many digits into them. They are NaN "not a number" right after they are allocated.

Try shortening out the number of decimal places, one by one, and eventually it will work.

- Dave

jim mcnamara · 11-15-2006, 09:53 AM

Quote:

Originally Posted by asgeirss

You are setting both var1 and var2 to 8 bits, but you are trying to put values with too many digits into them. They are NaN "not a number" right after they are allocated.

Try shortening out the number of decimal places, one by one, and eventually it will work.

- Dave

REAL*8 equates to a C double which for the OP's environment should have a minimum 15 digits of precision.

Offhand, I don't see the problem. He is on 64 bit box, maybe the compiler is creating 32 bit code and calling the wrong (32 bit)runtime library. Just guessing. I don't know if there is 64 bit support for GNU FORTRAN.

chrism01 · 11-16-2006, 06:19 PM

REAL * 8 => 8 bytes => C style long long

jlliagre · 11-17-2006, 01:10 AM

Quote:

Originally Posted by chrism01

REAL * 8 => 8 bytes => C style long long

Nope, real means floating point variable, not integer. Jim is right.

sundialsvcs · 11-17-2006, 09:32 AM

Correct. REAL is floating-point.

Display both operands before you compute with them. I think that you will find that one or both operands are "NaN."

chrism01 · 11-19-2006, 11:04 PM

oops, sorry, brain fart there. I was thinking of something else.