Support for double-precision (64bit) operations and two simultaneous single-precision (32bit) operations. It is also possible to generate only a subset of operations if e.g. divisions are not needed.