Dakkak, Abdul; Pearson, Carl; Li, Cheng

WebGPU is an online GPU programming environment used in various online courses offered by the University of Illinois and the IMPACT group. It is a scalable online GPU programming environment accessible though the web. By removing access to fully-featured system environment and developing targeted labs, WebGPU simplifies management without unduly impacting educational capability. The number of GPUs available through WebGPU can be dramatically fewer than the expected number of concurrent users, and can be dynamically scaled as the course participation changes. It has been used to teach introductory parallel programming courses, the introductory GPU programming course offered at UIUC, an advanced GPU programming course offered in cohorts with 5 other universities, a parallel programing summer school at Barcelona Supercomputing Center, three iterations of the Coursera CUDA course, and multiple lab sessions at tech conferences. It is battle-hardened, with the following on the system:

  • 4 million CUDA programs stored in the system
  • 1.75 million CUDA program compilations made
  • 700K CUDA related questions answered
  • 150K CUDA labs grades

Related papers:

"WebGPU: A Scalable Online Development Platform for GPU Programming Courses", Abdul Dakkak, Carl Pearson, Wen-mei Hwu, Proceedings of the 6th NSF/TCPP Workshop on Parallel and Distributed Computing Education (EduPar-16), 2016. [more...]