Register
Abstract: R is no
t a fast language. Poorly written R is really slow! Faced with s
low code\, people tend to think “parallel” or “GPU” (which is an a
djacent topic since GPUs allow to run many simple calculations in parallel
). Parallel programming can indeed greatly help speed up some types of cod
e. A lot of hardware however is not the answer to poorly written code. Bef
ore considering parallelization\, you should think of ways to optimize you
r code sequentially because not all programs can be parallelized\, paralle
l programming has costs and overheads\, and an optimized serial code will
also benefit your parallel code. In many cases\, writing better code will
save you more computing time than parallelization. This course will cover
just that: how to benchmark code\, how to identify bottlenecks\, and how t
o greatly improve code performance without using additional hardware.