This article describes how research into processor design boosts speed over today's superscalar offerings. The research is trying to work out whether it is possible to create a processor that looks to the programmer as though it is just one blisteringly fast machine but which is made up of lots of little sub-processors. The work of distributing the work of parallelising, and therefore speeding up code execution, is divided between the compiler and the machine, but the compiler gets to do more of the work.