The following pages describe a series of test on numerical computations I have encountered in astronomy, especially those that have involved GPGPU computing.
In most cases the tests are based on the use of OpenCL libraries, which enable parallel computations on both CPUs and GPUs (and other accelerators), have a wide device support, and also have open source implementations. OpenCL programs are natively written in C or C++. However, it can be used (more) conveniently from other languages like Python (pyOpenCL ) or Julia (OpenCL.jl ).