My Community

One thing I noticed in your makefile.include is that you use -mkl=cluster. I did not encounter this setting yet; how did you come up with it as it is also not present in the template makefile.include?
I am not sure if you are aware of the link line advisor for Intel. I tried to put in the settings based on what I read about your setup and got:
Link line: -L${MKLROOT}/lib/intel64 -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64 -lpthread -lm -ldl
Compiler options: -mkl=sequential
which seems to match what the default makefile.include sets, but notice the mkl=sequential compare to your setup.

I cannot tell you how I got to this, someone must have recommended it or I must have stumbled upon it in some forum ... Maybe I thought this was related to the sequential intel make file.
I will redo this with the default file, but again, it contains CUDA compiler instructions, shouldn't these be removed if CUDA is no longer supported by VASP?

I still get the same tests as failed when using the default makefile.include. Can I just remove the CUDA section in the makefile.include to skip GPU support?

If you use make std, make gam, or make ncl the CUDA code will not be compiled. Only make gpu and make gpu_ncl use that part.

Are you sure the machine has sufficient memory to run the tests?

If you are sure can you compile the code adding the following to the end of your makefile.include
FREE += -traceback -debug -g -check assume -check bounds -check format -check pointers -check output_conversion -check stack -check udio_iostat -check uninit
and see if that spits out an error when you run any of these tests.