GPU Computing using Python
The core topics covered are as follows:
- Understanding how OpenCL and CUDA map work to compute devices
- Making efficient use of memory on compute devices
- Using the Reikna library to schedule kernels and bridge the gap between Python Numpy
arrays and compute device memory allocations - Writing compute kernels to process data in Numpy arrays. Training in “survival C” for
kernels is included - Computing Fourier transforms over multi-dimensional arrays
- Optimisation tips for getting the best out of kernels
- Interleaving compute and IO for optimal compute device usage
Prerequisites:
- Attendees require a firm grasp Python
- Attendees require a first-year, tertiary-level familiarity of mathematical concepts surrounding scientific library topics
This course is designed for intermediate and advanced Python developers who need to accelerate parallelisable algorithms.
This workshop is now full booked, register for the waitlist here:
This course is planned for on-site delivery at the ARRC in Kensington, Western Australia. If the on-site situation changes due to COVID restrictions, a virtual conduct will replace the in-person class.