Precision problems in Fortran in KIVA4

zdz0309 · December 4, 2017, 10:24

I use CHEMKIN DVODE to integrate the chemistry in KIVA, where the species concentrations are used as the inputs into DVODE. These species concentrations (Vay) are calculated by
Vay(: ) = spd(i4c, : )*rmw(: )
And this way to integrate the chemistry works fine. When I manipulate this to (this manipulation is due to an algorithm)
Vay(: ) = spd(i4c, : )*rmw(: )*vol(i4c)
Vay(: ) = Vay(: )/vol(i4c)

And the in-cylinder pressures and temperatures show huge error compared with the results before the manipulation.
I doubt that the small values of vol(i4c) is the source of the error.

I used gfortran compiler. Can anyone tell me how to avoid this kind of error in Fortran?

Ravindra Shende · December 5, 2017, 07:39

You are right. small values of 'vol' could be the source of the errors.
Using the governing equations in non-dimensionalised form could help. Or you can use quad precision (available with intel fortran compiler, not sure about gfortran).

December 4, 2017, 10:24	Precision problems in Fortran in KIVA4	#1
zdz0309 New Member zdz Join Date: Dec 2017 Posts: 2 Rep Power: 0	I use CHEMKIN DVODE to integrate the chemistry in KIVA, where the species concentrations are used as the inputs into DVODE. These species concentrations (Vay) are calculated by Vay(: ) = spd(i4c, : )rmw(: ) And this way to integrate the chemistry works fine. When I manipulate this to (this manipulation is due to an algorithm) Vay(: ) = spd(i4c, : )rmw(: )*vol(i4c) Vay(: ) = Vay(: )/vol(i4c) And the in-cylinder pressures and temperatures show huge error compared with the results before the manipulation. I doubt that the small values of vol(i4c) is the source of the error. I used gfortran compiler. Can anyone tell me how to avoid this kind of error in Fortran?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Single precision better than double precision?	140raiders	CFX	1	July 30, 2015 03:32
User Fortran Problems	Silmaril	CFX	6	June 26, 2012 10:35
CFX11 + Fortran compiler ?	Mohan	CFX	20	March 30, 2011 19:56
Fortran 90 faster than C/C++	B. R. Guirguis	Main CFD Forum	48	March 6, 2006 14:49
'C' or FORTRAN or 'C++'	Yogesh Talekar	Main CFD Forum	20	October 21, 1999 05:00

December 5, 2017, 07:39		#2
Ravindra Shende Member Ravindra Shende Join Date: Feb 2011 Location: Pune, India Posts: 45 Rep Power: 15	You are right. small values of 'vol' could be the source of the errors. Using the governing equations in non-dimensionalised form could help. Or you can use quad precision (available with intel fortran compiler, not sure about gfortran).