Auto scheduling

• DF interpreter on each PE

• when idle ask task(automatic load balancing achieved)

• communication overhead masked by multithreading (excess parallelism)

Previous slide

Next slide

Back to first slide

View graphic version