Econometrics on Gpus

A graphical processing unit (GPU) is a hardware device normally used to manipulate computer memory for the display of images. GPU computing, also known as general purpose GPU computing, is the practice of using a GPU device for scienti c or general purpose computations that are not necessarily related to the display of images. The ability of a GPU to render rapidly changing complex images depends on the ability of the GPU to perform oating point operations related to rotations, scaling, shading, etc. Such operations exhibit data parallelism where di erent computational units operating simultaneously on independent data. GPUs have become extremely powerful for this sort of work. A number of scienti c applications have a data parallel structure, and realizing this, vendors and other parties have developed programming frameworks to make it relatively easy to conduct general purpose computation using GPUs. In this paper, we show how some problems in econometrics have the data parallel structure that allows for successful use of GPU computing. We nd speedups from 9.5 up to 55.4 times, compared to computations done on a single CPU core. These speedups can be obtained with very little expense, energy consumption, and time dedicated to system maintenance, compared to equivalent performance solutions using CPUs. The paper is accompanied by example code that illustrates how econometric computations can be done using one or a number of GPUs. The main intent of the paper is to contribute to the limited number of examples of GPU computing in economics and econometrics, and to provide example code that can be adapted with relative ease to other problems.