Newsletter ~ Spring 2009

Here is the latest news from SciComp on SciFinance®, the complete solution that automates coding and delivers source code (C/C++) for derivatives pricing and risk models. For more information on SciFinance, contact SciComp sales.

Table of Contents

New Version of SciFinance - 5.0

SciFinance 5.0 speeds derivatives performance with new CUDA and OpenMP parallel code styles and SciXpress features

New Movie

Check out our new movie, detailing our CUDA parallel computing support

Upcoming Events

SciComp to exhibit at ICBI Global Derivatives 2009 in Rome

New SciFinance Examples

Expanded SciXpress and SciMC Examples Catalogs

New SciFinance ASPEN Language/ Numerical Method Features

New specification statements Project and CompuTime

New Version of SciFinance - 5.0

CUDA-Enabled Code

New enhancements in version 5.0 of SciFinance include automatic parallel code generation for Monte Carlo pricing models that run 30 to 200 times faster than serial code (support for PDE codes scheduled for release later this year). The speedup is achieved by accessing the graphics processing unit (GPU) on high-end NVIDIA graphics cards. SciFinance customers can benefit immediately from SciFinanceÕs new ability to generate CUDA-enabled code, without having to become experts in parallel coding, simply by adding one keyword, "CUDA" to a SciFinance specifcation. One PC equipped with several inexpensive NVIDIA GPU cards can replace many racks of blades reducing the physical footprint and energy consumption.

OpenMP code

SciFinance 5.0 also includes the ability to automatically output OpenMP parallel code style for Monte Carlo pricing models. Nearly all modern desktop computers have multiple CPUs (typically two to eight processors). The SciFinance generated OpenMP compliant code can be compiled on standard Windows and Unix compilers. Performance improvement is near 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 SciXpress

SciFinance verison 5.0 now includes a new, concise SciXpress dialect for PDE specifications, and a catalog of examples. This dialect shortens even the most complex pricing problems to a few dozen lines, and requires less mathematical expertise.

A general working knowledge of the partial differential equations of finance is assumed. A rudimentary knowledge of finite difference techniques is helpful but not required. Many numerical parameters (e.g. defining regions, coordinate systems, grid parameters, etc.) can be defaulted to SciFinance's expert system or, if preferred, remain exposed for complete user control.

New Movie

Check out our new movie, detailing our new CUDA parallel computing support

Featuring Curt Randall of SciComp, Inc. and John Milner of NVIDIA, this new movie starts with an overview of SciFinance, quickly moves on to describing the new CUDA functionality, the benefits provided from CUDA-enabled codes and the ease with which SciFinance end users can generate this new code style. Have a colleague that would be interested in automated coding for derivatives? Send them this link, it's a quick and entertaining way to get a complete picture of SciFinance.

Watch it here

Upcoming Events

SciComp to Exhibit at ICBI Global Derivatives 2009 -- Rome

SciComp will demonstrate its products and have a speaking slot at the event. Visit us at our booth, we would like to discuss with you how we can help to take your derivatives pricing to the next level. We look forward to seeing you there.

• ICBI Global Derivatives 2009
• Rome April 28-30, 2009
• Sheraton Roma Hotel & Conference Center.
  

As a special discount to readers of our newsletters, quote VIP code: KR2205SCIEM and receive a 25% discount. Download form here.

New SciFinance Examples

The SciFinance Examples Catalogs provide hundreds of pricing and risk model specifications for all asset classes. The latest release includes a new SciXpress Catalog with approximately 100 of these concise specifications for PDEs. In addition, the SciMC catalog has undergone a major upgrade with the addition of many new examples and revision of existing ones.

The SciXpress Catalog includes these PDE examples:

• Basics
• Common Features
• Special Algorithms
• Digitals
• Asians and Lookbacks
• Barriers
• Short Rate Models
• CrossCurrency
• Stochastic Volatility
• Multi-Assets
• Convertible Bonds
• Fokker-Planck
• Vector PDEs

The SciMC Catalog includes these Monte Carlo examples:

• Basics
• Common Features
• Special Features
• Asians and Lookbacks
• Barriers
• Stochastic Volatility and Jumps
• Multivariate Options
• Mountain Range Options
  
• Other Path Dependent Options
• American Options
• Fixed Income (Short Rate Models)
• Fixed Income (LIBOR Market Models)
• Cliquet and Forward Options
• Volatility and Variance Swaps/Swaptions
• Structured Notes
  

New SciFinance ASPEN Language/ Numerical Method Features

Project

The "Project;" specification statement is intended to generate code files that interface smoothly to standard C projects. In a specification named "spec", a Project specification causes the main function to be placed in a separate specMain.c(pp) file, causes all other headers to be included in a file spec.h, and causes all other code to be included in spec.c(pp). The header files do extern "C" when necessary. Instead of SciStandard and SciInternal files, the auxilliary files are organized into SciStdIncludes.h, SciErrorCode.h, and when needed, SciErrorMessage.h.

The generated .c* files #include SciStdIncludes.h, which does #includes of the C or C++ library files and of SciErrorCode.h. Unlike SeparateModules, Project has no specShared file because it omits array validations in the Main routine. The Project specification also causes a <spec>.vcproj file to be generated.

CompuTime

The CompuTime specification statement allows you to measure the average run time of the pricing function. The value is output in the file CompuTime.out. The optional arguments Nskip and Niter allow you to specify the number of runs to skip before taking the measurement and the number of runs over which to average.

CompuTime[<Nskip>, <Niter>];

If you leave out the arguments, Nskip and Niter default to 0 and 1 respectively. You can set higher values for the arguments to get more stable output (e.g. Nskip=1 and Niter=10).

SciComp Inc., 5806 Mesa Drive
Suite 250, Austin, TX 78731
512-451-1050
Get more information on SciComp products