Moderator: Minesh B. Amin, Engineer/Founder, MBA Sciences, Inc.

Exploiting parallelism when starting from scratch is hard enough; doing so by leveraging the code-base of existing commercial serial software running upwards of a million lines of code in an industry (Electronic Design Automation) littered with NP Complete problems, well ... It is as if a doctor were about to perform a precise surgical operation that involved giving a patient both a new heart and a new circulatory system, and also replacing a good portion of the existing nervous system ... all while said patient is running in a marathon.

In this BOF session, we shall review four main challenges when exploiting parallelism, and a battle-tested methodology that is predicated on the following:

(a) any successful, scalable parallel solution must exploit non-determinism at some fundamental level, and

(b) the debugging, validation, verification, testing, and as much of the benchmarking as possible of the aforementioned parallel solution must be done using O(1) -- as opposed to O(N) -- compute resources.

The end result: An ability to leverage (your) existing serial methodologies for validation, verification and testing of serial software as much as possible. For in doing so, we would do a long way toward making the exploitation of parallelism easily accessible. No need for N windows, one for each remote/slave process. No need for movie-screen-size monitors.