PGI User Forum

[davidadams]

Hi group,
this is probably a straight-forward problem, but I am at a loss.
My relatively simple model code works fine on my AMD Athlon XP 3000 linux box (Red Hat 9.0) using pgf90 5.0.
I have moved the code to my laptop Intel Pentium M Processor 760 (Fedora Core 4, pgf90 6.2-4 32-bit) and the model blows up after a few time steps.

The code is indentical for both machines and it compiles in exactly the same manner on my laptop. Am I missing something in my pgf90 6.2-4 compiling line? Any suggestions where to look?

pgf90 -O2 -o main_program main_program.o utilities.o spec_core_driver.o constants.o transforms.o spherical_fourier.o spherical.o windows.o gauss_and_legendre.o grid_fourier.o fft.o fft99.o vert_coordinate.o update_atmosphere.o atmos_tendencies.o simple_physics.o spectral_damping.o implicit.o grid_dynamics.o matrix_invert.o vertical_advection.o spec_core_nml.o spectral_data_io.o


thanks,
Dave

[mkcolg]

Hi Dave,

How does your model "blow up"? Does it seg fault, get a stack overflow, or get wrong answers? Does the error occur at "-O0 -g"?

Try running again, this time setting your stack size to 'unlimited'. If this doesn't try compiling with "-g" and running your code in the debugger (pgdbg).

- Mat