In this talk we will first start with an explanation on how GPUs can be used as an alternative to traditional CPUs in the field of high-performance computing. Subsequently, we will spend some time on discussing the hardware especially the GPU cards as they are installed on several Ember nodes.
In the main part of the talk we will show a few examples where the basic concepts such as thread, block, and grid will be elucidated as well as the concept of shared memory.
The last part of the presentation covers topics such as compilation, mathematical libraries. Experience with C/C++ is strongly advised.