Arguments to be supplied:
	<operation> <type> <cpus> <minSecs> <density> <transposeA> <transposeB> {sizes}
where
	operation = the operation to benchmark; in this case: dgemm
	type = matrix type to be used; e.g. dense, sparse or rowCompressed
	cpus = #cpus available; e.g. 1 or 2 or ...
	minSecs = #seconds each operation shall at least run; e.g. 2.0 is a good number giving realistic timings
	density = the density of the matrices to be benchmarked; e.g. 0.999 is very dense, 0.001 is very sparse
	transposeA = false or true
	transposeB = false or true
	sizes = a list of problem sizes; e.g. 100 200 benchmarks squared 100x100 and 200x200 matrices
