New Atlas-based variant using BlockStructured FunctionSpace #372

Job	Run time
Test on gnu/9.4.0	54s
Test on gnu/9.4.0 --with-gpu --with-loki --with-atlas	14m 30s
Test on gnu/9.4.0 --with-gpu --with-loki --with-atlas --with-mpi	14m 38s
Test on gnu/9.4.0 --single-precision	38s
Test on gnu/9.4.0 --single-precision --with-gpu --with-loki --with-atlas	16m 44s
Test on gnu/9.4.0 --single-precision --with-gpu --with-loki --with-atlas --with-mpi	14m 6s
Test on gnu/9.4.0 --with-serialbox	1m 39s
Test on gnu/9.4.0 --with-serialbox --with-gpu --with-loki --with-atlas	3m 55s
Test on gnu/9.4.0 --with-serialbox --with-gpu --with-loki --with-atlas --with-mpi	4m 11s
Test on gnu/9.4.0 --with-serialbox --single-precision	1m 56s
Test on gnu/9.4.0 --with-serialbox --single-precision --with-gpu --with-loki --with-atlas	3m 24s
Test on gnu/9.4.0 --with-serialbox --single-precision --with-gpu --with-loki --with-atlas --with-mpi	3m 30s
Test on gnu/9.4.0 --cloudsc-fortran-pyiface=ON --cloudsc-python-f2py=ON	2m 48s
Test on nvhpc/21.9 --with-gpu --with-loki --with-cuda --with-atlas	18m 11s
Test on nvhpc/21.9 --with-serialbox --with-gpu --with-loki --with-cuda --with-atlas	46m 30s
Test on nvhpc/23.5 --with-gpu --with-loki --with-cuda --with-atlas	15m 21s
Test on nvhpc/23.5 --with-serialbox --with-gpu --with-loki --with-cuda --with-atlas	45m 50s
	3h 28m 45s

Provide feedback