Create mkmf.template.nvhpc

[ann-norcio]

Description:

Created mkmf.template.nvhpc in build_templates to run quickbuild.sh with nvhpc compiler for all models on Derecho. Modified compiler flag FFLAGS = -O -Mbackslash $(INCS) to include -Mbackslash, which addresses issue #555

Fixes issue

Fixes: #554 #555

Types of changes

• [X] Bug fix (non-breaking change which fixes an issue)
• [X] New feature (non-breaking change which adds functionality)
• [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
• [ ] Documentation update

Documentation changes needed?

• [ ] My change requires a change to the documentation.
  • [ ] I have updated the documentation accordingly.

Tests

Copy mkmf.template.nvhpc and set as your mkmf.template. Set compiler to nvhpc module load nvhpc. Then, run quickbuild.sh on DART/models/lorenz_63/work.

Checklist for merging

• [ ] Updated changelog entry
• [ ] Documentation updated
• [ ] Update conf.py

Checklist for release

• [ ] Merge into main
• [ ] Create release from the main branch with appropriate tag
• [ ] Delete feature-branch

Testing Datasets

• [ ] Dataset needed for testing available upon request
• [ ] Dataset download instructions included
• [X] No dataset needed

[hkershaw-brown]

Nice work testing out all the Derecho compilers and finding this, Ann.

There's code in dart that relies on IEEE standard that for a NaN, a /= a is .true. Background reading if you're interested (#176, #341)

Can you trying building and running the following code on Derecho with nvhpc:

code test_nan.f90

program test_nan

implicit none
real :: a, x

x = -1
a = sqrt(x)

print*, a

print*, a /= a
print*, a == a

if (a /= a) then
  print*, "you have a NaN"
endif

end program test_nan

You can compile and run it like this:

ftn -Kieee test_nan.f90 -o test_nan
./test_nan

vs

ftn test_nan.f90 -o test_nan
./test_nan

We want to make sure a /= a gives .true. for a NaN (we may need to add -Kieee to the compile flags in your mkmf.template.)

And bonus points, can you add some suggested flags if you want to turn on debugging for nvhpc?

[hkershaw-brown]

@hkershaw-brown the nan test should go in a developer test, since we are relying on the IEEE behaviour.