SciComp Speeds Derivatives Performance with New Parallel Code Styles and SciXpress Features
Rapid response to new market opportunities is critical for financial institutions in today’s markets. The derivatives market requires fast model development and accelerated execution speeds for derivatives pricing models. SciComp Inc. has enhanced SciFinance, its flagship derivatives pricing software, to help quantitative developers significantly shorten model development time and create models with fast execution speeds. SciFinance 5.0 features automatic parallel Monte Carlo code generation and a more concise specification dialect, SciXpress, for PDE (partial differential equation) models.
"New over-the-counter (OTC) contracts or new market conditions require new pricing models, and our customers must respond by producing models rapidly, but with the attention to detail required for accurate results," said Curt Randall, executive vice president of SciComp. "SciFinance was developed to replace days or weeks of error-prone hand coding and debugging; it automatically generates C or C++ pricing model source code in minutes."
Generation of Parallel Code with SciFinance 5.0
Among major enhancements in version 5.0 of SciFinance is automatic generation of parallel codes for Monte Carlo pricing models that run up to 30 to 200 times faster than serial code (support for PDE codes scheduled for release later this year). This acceleration is achieved by taking advantage of the highly parallel structure of high-end NVIDIA GPUs.
"Our customers can benefit immediately from SciFinance’s new ability to generate CUDA-enabled code," added Randall. "Without having to become experts in parallel coding, they can quickly create code that delivers up to 200X execution speed increases. A single PC equipped with several relatively inexpensive NVIDIA GPU cards can replace many racks of blades with a single box, reducing physical footprint and energy consumption by large factors."
SciFinance 5.0 also includes the ability to automatically create OpenMP parallel code for Monte Carlo pricing models. Nearly all modern desktop computers have multiple CPUs, usually from two to eight, while workstations may have many more. The synthesized parallel code is compliant with the OpenMP standard and with existing Windows and Unix compilers. It executes in the multi-processor environment with nearly linear speed-up, e.g. a factor of 3.9X on a quad-core PC or 22X on a 24 CPU workstation.
Shortened Specifications with SciFinance 5.0
With SciFinance, pricing models are created with concise, keyword-rich model specifications; hundreds of examples are provided as jumping off points. SciFinance version 5.0 now includes a new, concise SciXpress dialect for PDE specifications. This dialect shortens even the most complex pricing problems to a few dozen lines. Many numerical parameters and algorithm choices (e.g. defining difference grids, solvers etc.) can be defaulted to SciFinance’s expert system or, if preferred, remain exposed for complete user control. SciFinance version 5.0 includes the new SciXpress Catalog containing new concise example specifications across all asset classes.