In my last blog I outlined primitive data types to enable easy parallel programming using my C++ library. This weekend I began to implement my “lists with a twist”. The implementation has an elegant thought behind it, but there are inevitable messy low-level details. In this blog, I will outline these messy low-level details and in my next blog I will demonstrate the completed interface.
My parallel programming research has culminated with the design of a few elementary data structures, each of which will be implemented optimally for specific parallel hardware. Each data structure interface will operate on the same data, and the parallel hardware to use for execution is selected at runtime. This design provides users with a high level abstraction which is guaranteed to execute efficiently at a low level. The next step is to prove whether or not the result of composing these data structures in arbitrary ways is still efficient. Read the rest of this entry »