High-Performance Computing and Concurrency

It's deja vu all over again. In the old days (35 plus years ago), developers wanting to write half-decent programs had to know their hardware. Those days are back. Clock frequencies have peaked and hardware can no longer be abstracted behind high-level languages. Designed for developers with high performance requirements (games, finance analysis, scientific computation, etc.), this course teaches you what really happens when programs are executed and the subtle details that make a program go slow or fast.

With a focus on concurrency, specifically local concurrency (multi-threading), the course is all about writing efficient programs that make the best use of the computing resources available to you. While the sample code is written in C++, the course is not C++ specific. If you can read C++ code, but don’t use it in your work, you will still learn from this class.

• Learn how programs execute in hardware and the subtle details that affect program speed
• Practice writing efficient programs that get the most out of today’s CPUs, caches, and memory
• Discover how single and multi-core CPUs interact with memory and how to avoid memory slowness
• Explore memory models, concurrent data structures, lock-free concurrency, and lock-based concurrency
• Acquire the tools needed to measure the performance of programs and their components

Fedor G. Pikus is a chief engineering scientist in the Design-to-Silicon division of Mentor Graphics and a former senior software engineer at Google. Fedor builds the design automation tools used by the people who build the chips in your computers, cars, and more. He has over 25 patents, and over 90 papers and conference presentations on physics, EDA, software design, and the C++ language. He holds a Ph.D. in Applied Physics from Peter the Great St. Petersburg Polytechnic University.