Ph.D. School in Scientific GPU Computing
As part of the Ph.D. schools ITMAN and DCAMM, DTU Informatics hosts this summer school about utilizing massively parallel processors (GPUs) for general purpose desktop scientific computing.
Due to thermal restrictions further performance gains of microprocessors do no longer mainly depend on clock frequency increases but parallelization of the processor into multiple cores. Soon there will be tens of cores in each CPU with hundreds to follow. Graphics Processing Units (GPUs) already contain hundreds of scalar processing cores and thus enable us to explore this realm of massively parallel computing today.
The high number of parallel cores poses a great challenge for software design that must expose massive parallelism to benefit from the new hardware. The main purpose of this course is to teach practical algorithm design for such parallel hardware.
Focus will be on both CUDA and OpenCL programming in C, GPU architecture, parallelization of linear algebra algorithms, and how this can be leverage into advanced applications in scientific computing.
The following speaker have been invited to give the course:
- Assoc. Prof. Tim Warburton, Department of Computational and Applied Math, Rice University.
Learning Objectives
A student who has met the objectives of the course will be able to:
- Write CUDA and/or OpenCL programs for GPUs.
- Use CUDA numerics libraries (CUBLAS and CUFFT).
- Parallelize dense and sparse linear algebra computations.
- Solve scientific problems using the GPU.
- Estimate accuracy vs. speedup of numeric algorithms running on GPUs.
- Identify parallelism in a scientific computing problems.
- Arrange threads for parallel execution.
- Reduce global memory traffic in device code.
- Evaluate applicability of massively parallel processing to a specific scientific problem.
News update
Preliminary course program is now available (see Program tab).
The DCAMM flyer for this event is now online.
Organization
This event is organized by
Registration and practical matters
