[FLASH-USERS] Problem with HDF5 output on Ranger

[Sean Lynch]

On 12/07/2012 11:53 AM, Sean Lynch wrote:
> I've been having some problems with a Flash4 simulation creating HDF5
> output files on TACC Ranger.
>
> I wasn't sure what was happening so I tried the default Sedov problem
> and got the same error:
>
>   WARNING: you have called IO_writePlotfile but no plot_vars are defined.
>   Put the vars you want in the plotfile in your flash.par (plot_var_1 =
> "dens")
>   *** Wrote plotfile to sedov_hdf5_plt_cnt_0000 ****
>   Initial plotfile written
>
> I haven't changed the default flash.par so the plot variable are defined
> as:
>
>   plot_var_1 = "dens"
>   plot_var_2 = "pres"
>   plot_var_3 = "temp"
>
> The plot file is written to disk but contains no data for density,
> pressure, or temperature.
>
> I am using the default makefile for ranger.tacc.utexas.edu and have
> loaded the hdf5/1.6.5 module.
>
> Has anyone experienced this behavior?

This seems to be related to the Portland compiler.

With pgi I get the following warning during compilation (I didn't 
noticed it before)

PGF90-W-0164-Overlapping data initializations of haveplotvars (IO_init.F90)

I did some debugging in IO_init and 'haveplotvars' is always .FALSE. 
even right after it is declared as .TRUE.. So this seems to be why I was 
getting no output. It didn't have anything to do with the HDF5 library.

When I switched the compiler to Intel it compiled fine. It also ran 
correctly and produced the correct output for the Sedov problem.

Is this a bug in the pgi compiler or a problem with Flash?