An execution model for adaptive load-balancing on multicore and multi-GPU systems