TSMP runs with ParFlow on GPU

[chartick]

When running any testcase with ParFlow on GPU, ParFlow does not properly terminate. The model creates the last output, prints Node 0: Problem solved and the other models terminate but not ParFlow. Instead, there is a segmentation fault Caught signal 11 (Segmentation fault: address not mapped to object at address (nil))

Production and benchmark workflows suffer from this because they usually stop after an error.

[chartick]

It seems that the problem only occurs with the Intel compiler. Unfortunately, our GNU environment files are not up-to-date. I will assemble a new one.