Surf the Exponentials

Wired Magazine UK Edition, Jun-12 Issue. In the article  How to Spot the Future it is recommended to apply 7 rules. One is Surf the Exponentials. Moore's Law (chips will exponentially become faster and cheaper) has been the driver of the information age.

Nothing for a mathematician to roll the eyes? To catch the wave of faster-and-cheaper, Mathematicians let their solvers dance with symbols and clever numerical schemes - symbolically enhanced numeric computing as Wolfram calls the related differential advantage of Mathematica.

But catching the wave of exponentials of the new computing muscles - heterogeneous CPU/GPU architectures - might give us even a better way?

We are Reinventing UnRisk. I have announced this also here.
We made our present pricing and calibration engines blazingly fast by applying clever numerical schemes to the financial PDEs from the simpler to the most complex ones. This includes adaptive integration, finite elements with streamline diffusion, FourierCosine methods and more. All methods not so common in quant finance circles. This and combining grid computing with massive parallelism in the CUDA framework we have reduced difficult practical portfolio-across-scenario analysis cases from 8h to 8sec. This was the start: Taming the Machine Infernal and our engines became steadily faster from release to release.

The All-New UnRisk engines will be re-implemented in OpenCL and now having a major portion of the work done we recognize that simpler massive parallel algorithms fit even better. With an amazing code reduction and extra flexibility. The code is platform-agnostic.

We have caught the wave of faster-and-cheaper again and surf the exponentials with unbelievable little one-for-all code. This will enable quant developers and us to hone the experiences of programming the high level financial objects and events in a high-level domain-specific language, but rely on bank-proof models and solvers that utilize future computing muscles. Mathematica is supporting OpenCL.