Fiendishly clever computer trading of financial instruments using hideously complex algorithms is all the rage on Wall Street and increasingly in Tokyo, and now researchers at the London Parallel Applications Centre, based at London’s Queen Mary & Westfield College, are developing special library tools for financial modelling with the support of Digital Equipment Co Ltd and Oxford mathematical library supplier Numerical Algorithms Group Ltd, or NAG. Their two-year project initially involves the implementation of software on a massively parallel Single Instruction, Multiple Data system, with a view to converting it for Multiple Instruction, Multiple Data systems at a later date. The finished result will be a Fortran-90 mathematical library tool that NAG will likely market and that it is hoped will encourage the use of parallel machines in non-linear programming such as scheduling, planning and optimisation. A version of the software will also be developed in parallel C++. After consulting with financial experts in the City of London, the researchers, who are from University College London, Queen Mary & Westfield College, Imperial College and City University, have decided to take options pricing as a starting point for their project. This can then be extended to more complicated options such as Asian options and conditions, baskets of options and eventually portfolio evaluation and hedging. The researchers will first develop demonstration programs to show which routines are required and how they will interact. It is expected that the first demonstration will compute the exercise boundary of US options with continuously paid dividends and will be tested using a simulated data feed. The project will use the Centre’s recently delivered DECmpp 12000Sx massively parallel computer – from Maspar Computer Corp marketed by DEC – which has 8,192 CPUs. And it is hoped work will continue after the initial two years, perhaps with support from the business community nearby.
