You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This seminar is a direct result of my dissertation work titled ``Load Balancing on the GPU,'' a GPU-based load-balancing programming model for fine-grained computations. This talk is intended as a tutorial, in which I will present lessons learned on how \emph{not} to program a GPU. I will focus on three key aspects: performance, program abstractions and software engineering. Together we will build the fastest General Matrix-Multiplication (GEMM) kernel, ever.
% Memory is everything, but load-balancing is better.