Just want to clarify - the ISPC implementation shown in the previous slide isn't necessarily using this message passing model under the hood, and is solving the problem using an SPMD programming model. In this lecture we are talking about a different way to parallelize the same grid solver problem using inter-process communication.

I just wanted to meta-clarify @ajayram's clarification comment. I don't think the previous slide was actually using an SPMD programming model. SPMD programs are the ones that use calls to getThreadID (or something similar) to determine what work each processing element needs to do. The program on the previous slide I believe falls more generally under "data-parallel thinking" from the previous lecture.