For more information, see the ExofastGpu wiki

• Common algorithms for transit light curve modeling
• High-performance code is one to two orders of magnitude faster on selected nVidia GPUs than a CPU core
• Code also runs on CPUs without a GPU (either single or multi-core via OpenMP or TBB)
• Routines easily accesible from C, C++, CUDA, and IDL

These routines have not been validated (yet). If you find bugs, validate a routine, or would like to contribute, please notify Eric Ford.

It's not clear if/when there will be a publication about ExofastGpu. In the mean time, please cite:

• Mandel, K. & E. Agol (2002) ApJ 580, L171-175. (doi:10.1086/345520)
• Ford, E.B. 2009, New Astronomy, 14, 406-412. (doi:10.1016/j.newast.2008.12.001)
• Dindar, S. et al. (2013), New Astronomy, 23, 6-18. (arXiv:1208.1157)
• Eastman, J. et al. (2013), PASP, 125, 83 (arXiv:1206.5798)

• Mandel & Agol code website
• Exofast website
• Eric Ford's website
• [Thrust website] (http://thrust.github.com/)
• [Comparison of nVidia GPU Specifications] (http://en.wikipedia.org/wiki/Comparison_of_Nvidia_graphics_processing_units#Tesla)