High performance simulations of turbulent clouds on a desktop PC: exploiting the GPU

Processor clock speeds have increased exponentially over the last several decades. This has gone a long way toward supplying the necessary computational power for running these numerical simulations. Computational clusters and supercomputing facilities now have computing nodes with traditional processors, specialized processors, or both. The number of processing cores in computing nodes is also increasing. The graphics processing unit (GPU) is becoming a mature platform for running numerical simulations. It was designed to perform the intensive matrix projection calculations associated with gaming graphics. In order to efficiently and quickly perform such calculations, modern GPUs are designed as massively parallel calculating devices. In order to effectively utilize the full power of today' supercomputer through large-scale parallelization, relatively large problem sizes are required.

@Article { SGPJ12,
  author       = "Schalkwijk, Jerome and Griffith, Eric and Post, Frits H. and Jonker, H.J.J.",
  title        = "High performance simulations of turbulent clouds on a desktop PC: exploiting the GPU",
  journal      = "Bulletin of the American Meteorological Society (BAMS)",
  number       = "3",
  volume       = "93",
  pages        = "307-314",
  month        = "March ",
  year         = "2012",
  url          = "http://graphics.tudelft.nl/Publications-new/2012/SGPJ12"
}