Abstract: We propose a hybrid parallelization scheme for matrix inversion on multicore processors that combines a look-ahead technique to extract task-parallelism, at a high level, with loop-level ...