high performance computing on graphics processing units: hgpu.org

In the advent of new large galaxy surveys, which will produce enormous datasets with hundreds of millions of objects, new computational techniques are necessary in order to extract from them any two-point statistic, the computational time of which grows with the square of the number of objects to be correlated. Fortunately technology now provides multiple means to massively parallelize this problem. Here we present a free-source code specifically designed for this kind of calculations. Two implementations are provided: one for execution on shared-memory machines using OpenMP and one that runs on graphical processing units (GPUs) using CUDA. The code is available at http://members.ift.uam-csic.es/dmonge/CUTE.html.