Architectures and Compilation Techniques for Fine and Medium Grain Parallelism