Spack installation of ASPECT on cluster

ASPECT
1

Hello,

I’ve recently had ASPECT installed and compiled on a new computer at my university. The developers used spack for the installation.

However, now the program does not output the ASPECT error messages when the model fails but rather I’m given a generic failure message.

I spoke to the research computing team and they are wondering if there is a known way to flag verbosity when doing the installation using spack.

If not I’m sure it can be manually installed but I thought I’d check here first.

Thanks,
Stephanie

2

Stephanie,
when you start ASPECT, it shows an output header that looks like this:

-----------------------------------------------------------------------------
-- This is ASPECT, the Advanced Solver for Problems in Earth's ConvecTion.
--     . version 2.5.0-pre (wb, 54ebb9a5b)
--     . using deal.II 9.5.0-pre (master, f8d87ffe95)
--     .       with 32 bit indices and vectorization level 1 (128 bits)
--     . using Trilinos 13.2.0
--     . using p4est 2.0.2
--     . running in DEBUG mode
--     . running with 1 MPI process
-----------------------------------------------------------------------------

You will only get good error messages if you are in DEBUG mode as shown above. Does your build show that it’s in debug, or in release mode?

Best
W.

3

Hi Wolfgang,

Thanks for your quick response. I am using optimized mode but I still get error outputs in my error file when I introduced an intentional error for example in the parameter file I just ran on the old computer.

I got this error:

Exception ‘dealii::ExcMessage (“Invalid input parameter file.”)’ on rank 0 on processing:

An error occurred in line <487> of file </scratch/root/builds/aspect/15584/src/source/main.cc> in function
void parse_parameters(const string&, dealii::ParameterHandler&)
The violated condition was:
false
Additional information:
Invalid input parameter file.

The research computing team is also installing the debug mode for me, per my request but I was used to getting some error output even with optimized mode so I’m a little confused.

Thanks,
Stephanie

4

Hi Stephanie,
you should expect to get the same quality of error messages from release and debug mode. The general rule is that your model should be able to run in debug mode without errors, and only then should you switch. The reason why release mode is 4x to 10x faster than debug mode is principally because it checks far far fewer potential error conditions.

I don’t know why you do or don’t get some error messages in release mode, but it’s a rabbit hole not worth chasing down.
Best
W.

5

OK, sounds good.

Because I was able to fix errors as they arose using the release mode because they were typically simple and the error messages were descriptive enough, I was doing that. But now that you remind me of the proper practice of using debug mode to - well, debug - I can just do that.

Thanks,
Stephanie