compiler: Tweak par-tile unrolling and revamp MPI topology #6881

Job	Run time
pytest-ubuntu-py311-gcc11-noomp-base	50m 55s
pytest-ubuntu-py311-gcc11-noomp-adjoint	37m 40s
pytest-ubuntu-py312-gcc12-omp-base	1h 6m 23s
pytest-ubuntu-py312-gcc12-omp-adjoint	51m 9s
pytest-ubuntu-py39-gcc7-omp-base	50m 45s
pytest-ubuntu-py39-gcc7-omp-adjoint	37m 32s
pytest-ubuntu-py310-gcc10-noomp-base	50m 14s
pytest-ubuntu-py310-gcc10-noomp-adjoint	37m 41s
pytest-ubuntu-py312-gcc13-omp-base	1h 7m 48s
pytest-ubuntu-py312-gcc13-omp-adjoint	47m 36s
pytest-ubuntu-py39-gcc9-omp-base	50m 32s
pytest-ubuntu-py39-gcc9-omp-adjoint	37m 40s
pytest-osx-py312-clang-omp-base	45m 41s
pytest-docker-py39-gcc-omp-base	44m 30s
pytest-docker-py39-gcc-omp-adjoint	33m 13s
pytest-docker-py39-icx-omp-base	36m 51s
pytest-docker-py39-icx-omp-adjoint	33m 27s
	12h 59m 37s

Provide feedback