ocl: revised device-split, additional tuning param, and other improvements #1386

Job	Run time
build-on-opencl (OPENMP=ON, SMM=libxsmm)	11m 35s
pre-commit	1m 26s
build-on-cuda (OPENMP=ON)	11m 29s
build-and-test (MPI=ON, OPENMP=ON, SMM=blas, openmpi)	7m 28s
build-and-test (MPI=ON, OPENMP=ON, SMM=blas, mpich)	6m 26s
build-on-rocm (OPENMP=ON)	17m 0s
build-and-test (MPI=ON, OPENMP=ON, SMM=libxsmm, openmpi)	7m 54s
build-and-test (MPI=ON, OPENMP=ON, SMM=libxsmm, mpich)	6m 24s
build-and-test (MPI=ON, OPENMP=OFF, SMM=blas, openmpi)	4m 28s
build-and-test (MPI=ON, OPENMP=OFF, SMM=blas, mpich)	3m 58s
build-and-test (MPI=ON, OPENMP=OFF, SMM=libxsmm, openmpi)	4m 39s
build-and-test (MPI=ON, OPENMP=OFF, SMM=libxsmm, mpich)	4m 0s
build-and-test (MPI=OFF, OPENMP=ON, SMM=blas, openmpi)	7m 12s
build-and-test (MPI=OFF, OPENMP=ON, SMM=libxsmm, openmpi)	7m 4s
build-and-test (MPI=OFF, OPENMP=OFF, SMM=blas, openmpi)	4m 22s
build-and-test (MPI=OFF, OPENMP=OFF, SMM=libxsmm, openmpi)	3m 48s
Combine & check coverage.	52s
	1h 50m 5s

Provide feedback