Thread partitioning and value prediction for exploiting speculative thread-level parallelism | Publicación