[Joachim]

Hi everyone!

I have been having an issue with my fortran code, which really does not make sense to me. If somebody could give me his opinion, I would greatly appreciate!

At one point, my code generates a mistake (a variable = NaN). I used write statements to spot the source of the error.

if I use

print *, Mbar -> answer: 2.802

if I use

print *, Mbar, Cavg -> answer: NaN, 0,354

I did not change anything in the code, but the value of the variable Mbar changed to NaN!!!
Why would such a thing happen??

Thank you very much for your help!

Joachim

[FMDenaro]

Originally Posted by Joachim

Hi everyone!

I have been having an issue with my fortran code, which really does not make sense to me. If somebody could give me his opinion, I would greatly appreciate!

At one point, my code generates a mistake (a variable = NaN). I used write statements to spot the source of the error.

if I use

print *, Mbar -> answer: 2.802

if I use

print *, Mbar, Cavg -> answer: NaN, 0,354

I did not change anything in the code, but the value of the variable Mbar changed to NaN!!!
Why would such a thing happen??

Thank you very much for your help!

Joachim

what option you added in compiling your code? Mbar should be an integer? Often this kind of error is just an indication that your source code has some error like mismatch of type or some error in the bound of array.
Use the debug settings..

[cfdnewbie]

I would agree, Debug flags are your best option. I would assume some memory allocation / out of bounds array things, because when an error seems to "jump" when you add a line, that usually means that a different part of the memory becomes corrupted.

Just out of curiosity: what type of code is it? f77 still around?

[Joachim]

Hey! thanks to both of you.
I am not very familiar with makefiles however. What kind of line should I had to spot the mistake?

So far, that is what I have in my Makefile

-------------------------------------------------------------------------------------
FC = gfortran

# List of executables to be built within the package
PROGRAMS = main_euler_v1

-fcheck=all -fbacktrace -ffpe-trap=invalid,zero,overflow

# "make" builds all
all: $(PROGRAMS)

# Dependency rules
main_euler_v1.o: data_package_v1.o
main_euler_v1: data_package_v1.o
main_euler_v1.o: grid_jacobian_v1.o
main_euler_v1: grid_jacobian_v1.o
main_euler_v1.o: euler_solver_v1.o
main_euler_v1: euler_solver_v1.o
main_euler_v1.o: fluxes_v1.o
main_euler_v1: fluxes_v1.o

grid_jacobian_v1.o: data_package_v1.o
euler_solver_v1.o: data_package_v1.o
fluxes_v1.o: data_package_v1.o

# General rules for building the program
%: %.o
$(FC) $(FCFLAGS) -o $@ $^ $(LDFLAGS)

%.o: %.f90
$(FC) $(FCFLAGS) -c $<

%.o: %.F90
$(FC) $(FCFLAGS) -c $<

# Utility targets
.PHONY: clean veryclean

clean:
rm -f *.o *.mod *.MOD

veryclean: clean
-----------------------------------------------------------------------------------

Thank you once again!

Joachim

ps: I am using f90

[Patrick1]

Originally Posted by Joachim

Hi everyone!

I have been having an issue with my fortran code, which really does not make sense to me. If somebody could give me his opinion, I would greatly appreciate!

At one point, my code generates a mistake (a variable = NaN). I used write statements to spot the source of the error.

if I use

print *, Mbar -> answer: 2.802

if I use

print *, Mbar, Cavg -> answer: NaN, 0,354

I did not change anything in the code, but the value of the variable Mbar changed to NaN!!!
Why would such a thing happen??

Thank you very much for your help!

Joachim

Try using a format specifier:

http://www.fortran.com/F77_std/rjcnf0001-sh-13.html

here is an old piece of code of mine. First I define the format with ID 91:

91 format (3X,I4,4X,3(6x,f6.1),2(6x,F10.6))

Then I write onto the screen (6) using the 91 format:

write (6,91) turbno,x_mat(dum_turbno),y_mat(dum_turbno),height, (U(dum_turbno,j)),(TI(j))

turbno etc are just the names of my variables.

The format specifiers can be found on the web but the basic ones are:

Integers of field width w: Iw
Real numbers of field width w and n decimal places: Fw.n
n blank spaces: nX

So in my format 91 I write 3 blank spaces, and integer of field width 4, 3 times 6 blank spaces then a real number of field width 6 and one decimal place..etc

[FMDenaro]

Originally Posted by Joachim

Hey! thanks to both of you.
I am not very familiar with makefiles however. What kind of line should I had to spot the mistake?

So far, that is what I have in my Makefile

-------------------------------------------------------------------------------------
FC = gfortran

# List of executables to be built within the package
PROGRAMS = main_euler_v1

-fcheck=all -fbacktrace -ffpe-trap=invalid,zero,overflow

# "make" builds all
all: $(PROGRAMS)

# Dependency rules
main_euler_v1.o: data_package_v1.o
main_euler_v1: data_package_v1.o
main_euler_v1.o: grid_jacobian_v1.o
main_euler_v1: grid_jacobian_v1.o
main_euler_v1.o: euler_solver_v1.o
main_euler_v1: euler_solver_v1.o
main_euler_v1.o: fluxes_v1.o
main_euler_v1: fluxes_v1.o

grid_jacobian_v1.o: data_package_v1.o
euler_solver_v1.o: data_package_v1.o
fluxes_v1.o: data_package_v1.o

# General rules for building the program
%: %.o
$(FC) $(FCFLAGS) -o $@ $^ $(LDFLAGS)

%.o: %.f90
$(FC) $(FCFLAGS) -c $<

%.o: %.F90
$(FC) $(FCFLAGS) -c $<

# Utility targets
.PHONY: clean veryclean

clean:
rm -f *.o *.mod *.MOD

veryclean: clean
-----------------------------------------------------------------------------------

Thank you once again!

Joachim

ps: I am using f90

but what about FCFLAGS?

[Joachim]

thanks everyone! Actually, I cheated for that makefile. I just took a template and adapted it for my purpose...

I think I found something interesting, but I still don't fully understand it.

In my code, I use a special type called flow, in which I store most of the variables (--> flow(i,j)%u, etc).

1.At one point in the code, I checked and flow(i,j)%u is properly defined.

2. It is then called by a subroutine to compute the fluxes. The flow pointer is a global variable (use simulation_data at the beginning of the subroutine), but for some reason, if I do print *, flow(i,j)%u, I obtain NaN.

3. Pretty weird, if I add a print statement at the very beginning of the subroutine, then everything works just fine!! (if I don't, the flux calculation fails because u = NaN).

I think there is a problem when communicating the flow pointer, but I really don't understand why!! Would anyone have an idea??
Thanks very much for your help!