Serial code is too slow? Parallelise!

Split computation over multiple processor cores — various architectures:

Shared Memory

• Shared memory — e.g. multicore processors

Distributed Memory

• Many independent memories/processors connected via a network
• Also increases the total memory available
• We can solve larger problems

GPGPUs

• traditional processor + GPUs (emerging technology)
• GPU = graphics processing unit (typically 100s of cores)