State of the IT
To speed up the models produced by its SciFinance automated coding application for rapidly developing Monte Carlo or partial differential equation-based derivatives pricing and risk models, the company introduced the automatic generation of OpenMP or Nvidia CUDA-enabled parallel code. CUDA-based code runs on Nvidia's graphics processors 30-100 times faster than serial code on a standard PC, the company claims. It is also applicable to a range of Monte Carlo pricing models, including those with complex path dependency and Bermudan exercise features. SciCalibrator translates pricing model calibration specifications into C/C++ source code for use with a given model or as a stand-alone routine.
This article appeared in Risk Magazine