In this talk I will give a short introduction how systems of ordinary differential equations (ODEs) can be solved on modern GPU devices. The talk is based on odeint [1] a C++ library devoted to solve ODEs and which is accepted as a Boost library. I will mainly show how different CUDA and OpenCL libraries can be used in odeint and compare these libraries by usability and performance. In more detail I will consider the frameworks VeXCL[2], ViennaCL[3], Thrust[4] and MTL[5] which are all OSS libraries for GPU technologies.
[1] www.odeint.com
[2] https://github.com/ddemidov/vexcl
[3] viennacl.sourceforge.net
[4] thrust.github.com
[5] www.mtl4.org