Problem with the precision of scalar values

faab · November 25, 2015, 12:11

Dear everyone,

I am working with a modified version of the ChtMultiRegion Solver, and the DP (double precision) is the one I set for the compiler. I am working on a 64-bit unit. My problem is that when the solver computes a very large number - say -1.45 to the power 21, to take an actual example - and stacks it in the allocated memory by passing this value to an existing variable, it is converted to 0. So of course, when I later use this variable in a division I get a segmentation fault. I do not understand how this process works, and because I use the DP, I do not see how to fix the issue. Could anyone give me a hand with this matter please ?

PS. In case it helps: I just ran a test where I state a=-1.45e+21, and "print" the value which is returned correctly by the solver. But when I do not use the "e" exponent and enter the full value (with 19 zeros) I get 0 in return. So I guess the issue/limitation comes from the number of digits, any ideas ?? Thanks !

faab · November 26, 2015, 11:18

After some time I was able to isolate the problem and it seems that the issue is caused by negative values computed in the power function Foam:

ow(). I computed the following statement separately:
Foam:

ow((-27e1), scalar(1)/scalar(3))
The same statement with +27 instead of -27 returns the correct answer (3). But with the negative value as an argument I get the following error message:
Foam::error:

rintStack(Foam::Ostream&)
#1 Foam::sigFpe::sigHandler(int)
#2
at sigaction.c
#3
at interp.c
#4
#5 __libc_start_main
#6

Am I doing something wrong here ? Thank you for your help !

wouter · November 26, 2015, 11:27

Hello Faab,
here is your answer: http://www.cplusplus.com/reference/cmath/pow/
google is your friend.
hope this helps
wouter

faab · November 26, 2015, 11:39

Hi Wouter,

Thank you for your help. Your link only tells me that what I am doing is wrong, which I knew already. As I said I figured that using a negative base generates a domain error. My goal is to solve this, to find a work around.
Cheers, thanks again for your time.

wouter · November 26, 2015, 17:55

hello Faab,
Maybe the use or binairy instead of ascii in controldict. Or you need to change write precision to 23 or something.

hope this helps
Wouter

faab · November 27, 2015, 04:44

Hi Wouter,

Thank you for the tips. It did not solve the issue but I found out that the base is never supposed to be negative, so I must adapt the code to implement new sets of equations when the temperature gets to a limit where the base gets negative. In the end the problem is still unsolved but I don't need to work on this now, however if anyone has a solution as to how to compute negative bases using the Foam:

ow() function, I am very keen to hear it !
Thanks again.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Negative values in scalar transport	jnally	OpenFOAM Running, Solving & CFD	5	August 23, 2018 15:32
sonicFoam Wedge, Max. Iterations Exceeded, hConstThermo,	v199ah	OpenFOAM Running, Solving & CFD	4	March 15, 2016 09:39
Non-Uniform Scalar Values	Shabirmean	OpenFOAM Running, Solving & CFD	0	August 27, 2014 13:12
[swak4Foam] Problem: Defining a linear scalar BC with groovyBC	holp	OpenFOAM Community Contributions	3	July 19, 2012 20:19
NEAGTIVE SCALAR VALUES	Mario	Siemens	1	November 18, 2002 02:01

November 25, 2015, 12:11	Problem with the precision of scalar values	#1
faab Member fouad abi Join Date: Nov 2015 Location: Geneva, Switzerland Posts: 42 Rep Power: 10	Dear everyone, I am working with a modified version of the ChtMultiRegion Solver, and the DP (double precision) is the one I set for the compiler. I am working on a 64-bit unit. My problem is that when the solver computes a very large number - say -1.45 to the power 21, to take an actual example - and stacks it in the allocated memory by passing this value to an existing variable, it is converted to 0. So of course, when I later use this variable in a division I get a segmentation fault. I do not understand how this process works, and because I use the DP, I do not see how to fix the issue. Could anyone give me a hand with this matter please ? PS. In case it helps: I just ran a test where I state a=-1.45e+21, and "print" the value which is returned correctly by the solver. But when I do not use the "e" exponent and enter the full value (with 19 zeros) I get 0 in return. So I guess the issue/limitation comes from the number of digits, any ideas ?? Thanks !

November 26, 2015, 11:18		#2
faab Member fouad abi Join Date: Nov 2015 Location: Geneva, Switzerland Posts: 42 Rep Power: 10	After some time I was able to isolate the problem and it seems that the issue is caused by negative values computed in the power function Foam:ow(). I computed the following statement separately: Foam:ow((-27e1), scalar(1)/scalar(3)) The same statement with +27 instead of -27 returns the correct answer (3). But with the negative value as an argument I get the following error message: Foam::error:rintStack(Foam::Ostream&) #1 Foam::sigFpe::sigHandler(int) #2 at sigaction.c #3 at interp.c #4 #5 __libc_start_main #6 Am I doing something wrong here ? Thank you for your help !

November 26, 2015, 11:27		#3
wouter Senior Member Wouter van der Meer Join Date: May 2009 Location: Elahuizen, Netherlands Posts: 203 Rep Power: 17	Hello Faab, here is your answer: http://www.cplusplus.com/reference/cmath/pow/ google is your friend. hope this helps wouter

November 26, 2015, 11:39		#4
faab Member fouad abi Join Date: Nov 2015 Location: Geneva, Switzerland Posts: 42 Rep Power: 10	Hi Wouter, Thank you for your help. Your link only tells me that what I am doing is wrong, which I knew already. As I said I figured that using a negative base generates a domain error. My goal is to solve this, to find a work around. Cheers, thanks again for your time. Last edited by faab; November 26, 2015 at 11:44. Reason: -

November 26, 2015, 17:55		#5
wouter Senior Member Wouter van der Meer Join Date: May 2009 Location: Elahuizen, Netherlands Posts: 203 Rep Power: 17	hello Faab, Maybe the use or binairy instead of ascii in controldict. Or you need to change write precision to 23 or something. hope this helps Wouter

November 27, 2015, 04:44		#6
faab Member fouad abi Join Date: Nov 2015 Location: Geneva, Switzerland Posts: 42 Rep Power: 10	Hi Wouter, Thank you for the tips. It did not solve the issue but I found out that the base is never supposed to be negative, so I must adapt the code to implement new sets of equations when the temperature gets to a limit where the base gets negative. In the end the problem is still unsolved but I don't need to work on this now, however if anyone has a solution as to how to compute negative bases using the Foam:ow() function, I am very keen to hear it ! Thanks again.