Harnessing massive parallelism in the era of parallelism for the masses