Case Study: Financial Analysis
Financial Analysis on multi-core processors
Download the full document as a PDF
The dramatic increase in algorithmic trading has created increased demand for high performance computing solutions within financial organizations. To remain competitive, portfolio and risk managers must look to innovations in computational finance, especially in modeling and optimization of financial management models and pricing financial instruments. The benefits of algorithmic trading are highly dependent on deploying the smartest, fastest algorithms that take full advantage of the latest computer hardware.
Multi-core processors and accelerators present significant potential for performance gains in these software applications. Since few applications today take full advantage of this new technology, there is an opportunity for financial organizations to gain a competitive advantage. Applications that are not multi-core enabled will not scale as processor vendors continue to add additional cores; in fact, those applications may get slower as the per core performance decreases.
The RapidMind Multi-Core Development Platform allows software organizations to quickly build multi-core capable applications using existing practices, tools, and compilers. Whereas other techniques, such as tools for multi-threading, are difficult, time-consuming, and error prone, use of the RapidMind platform enables a safe, reliable, and maintainable application. Furthermore, RapidMind-enabled applications will automatically scale to additional cores and can easily migrate to future multi-core processors or accelerators (such as GPUs or the Cell). RapidMind lets financial organizations focus on their internal algorithmic expertise yet quickly deploy applications on the best possible hardware available at any time. A number of benchmarks (detailed later in this paper) specific to financial analysis are available that demonstrates the tremendous advantage provided by the RapidMind Platform. These include the following pricing algorithms:
- Quasi Monte Carlo: 32x speed-up on 8-cores
- Binomial Tree: 7.3x speed-up on 8-cores
- Closed Form over portfolio: 10x speed-up on 8-cores