June 16th, 2010 — Massive Multithreading on the GPU — Gavriel Plotke


GigaFlops or… TerraFlops? The modern GPU is a massively parallel supercomputer that has 2 orders of magnitude more processing power than a single CPU Core. But the threading model is very different than the CPU. Don’t get left behind on knowing about this different paradigm.


Gavriel Plotke, Microsoft Senior SDET. I’ve been with Microsoft for 3 years on the DirectX High Level Shader Language compiler test team. I’ve been a champion of the new Compute Shader features of DirectX11. While I have some graphics background, it was my assembler language and compiler writing background that got me the job. Early in my career I worked on the internals of a mainframe spreadsheet product as a competitor to VisiCalc and Lotus123. It was a time when everyone in a big office would have a mainframe terminal and no one had a PC. It was a great product, but time moved on, and Excel on inexpensive PCs ate our lunch. After that I spent many years doing business contract work — different projects, different platforms, lots of database design. Now I help support programming on today’s supercomputer, the modern graphics card.