Recent technological trends, such as the multi-core processor, have dramatically changed the design tradeoffs in scalable high performance computation. This course will examine the latest approaches to meet address technological trends, examining both algorithms and implementation techniques. Topics will include
CSE 262 meets the requirement for the CSE Department's MS Concentration area in Parallel and Scientific Computing.
Students will present papers in class on various topics relevant to the course and will also turn in paper summaries for some of the assigned readings. There will also be a course project.
In order to allow time to complete the assigned readings and to participate in discussing the readings in class, the expected programming workload in CSE 262 is lower than that of CSE 260. Possible projects may range for algorithm design with a modest amount of implementation effort to more implementation-oriented efforts involving optimization or implementation strategy.
Resources will be available on a variety of computing systems including the newly released Intel Phi (Knights Corners), GPUs, GPU clusters, NUMA, SSD based platforms and large scale systems with 10,000 to 100,000 cores.
Prior experience with some form of parallel computing is required, and can be met by taking CSE 260, PHY 244, or an equivalent course, or having relevant prior (work) experience. If you are comfortable programming with one of CUDA, threads, or MPI, you have the required background.
Students outside CSE are welcome. Undergraduates are also welcome, but may not be concurrently enrolled in CSE 160, which will also be offered in Winter.
If you aren't sure about your technical background, contact me and we can discuss it.
Maintained by [Mon Jan 7 18:00:38 PST 2013]