nZetta Toolkit

---

The nZetta toolkit is the new standard in performance; the easiest and most productive toolkit of derivative pricing components available for the financial industry.  The nZetta toolkit enables quants to rapidly solve pricing problems by providing them with components that fully utilize the latest hardware and software advances to provide blistering computing performance. Whether providing a 15x speed-up on x86 processors, 100x speed-up on GPUs, or over 1000x speed-up using algorithmic differentiation for risk, the nZetta  toolkit allows quant to focus on modeling and products rather than performance. With quick and easy integration into existing libraries, the nZetta  toolkit will rapidly turbocharge your products.

Jointly developed by Zettamatics, experts in financial modeling, and NAG, the world’s leading provider of advanced mathematical analytics. Combining Zettamatics’ team’s decades of front-office banking experience with NAG’s cutting-edge analytics, the nZetta toolkit provides the fastest state-of-the-art components for derivative pricing with a pedigree you can trust.

The components will significantly enhance existing libraries' performance and prepare them for future directions in software and hardware innovations such as Adjoint Algorithmic Differentiation (AAD) and GPUs. You can easily achieve performance gains of orders of magnitude.

The nZetta toolkit is a C++17 collection of components designed to give the best performance possible from your hardware. Software switchable between x86 single-threaded, x86 thread-distributed, or GPU abstracts the hardware so you can concentrate on the finance modelling; write code once and run on x86, thread-distributed, or GPU unchanged. All components have been written using the latest algorithms and approaches to give the fastest possible performance for the finance domain. 

And better still, all components are AAD-ready.

The nZetta toolkit may be integrated as a drop-in functionality via a C-style API for rapid integration or using the nZetta Tensor framework to enable even faster performance.

The toolkit is not a pricing library; it does not provide conventions, curves, or models. 

It does provide all the performance critical maths components necessary for quant pricing 10x, 100x, or even 1000x faster than standard approaches including:

• Monte Carlo: Generators, correlation & repair, Brownian-bridge, antithetic, tranching, skip-ahead, regression, reduction, fast processes
• 1, 2-D PDE: Douglas ADI, Craig-Sneyd, modified Craig-Sneyd and Hunsdorfer-Verwer methods, Rannacher stepping, non-uniform ordinate construction, lattice expansion and contraction, forward Kolmogorov support.
• Basic finance components: interpolators, processes, vol models
• Tensor support: Full tensor mathematics, shaping, regressions, reductions, windowed reductions.
• In addition, the nZetta library provides extensive tensor capabilities to fully vectorize your pricing library. 
• Basic finance components: interpolators, processes, vol models
• Tensor support: Full tensor mathematics, shaping, regressions, reductions, windowed reductions.

In addition, the nZetta library provides extensive tensor capabilities to fully vectorize your pricing library. 

Stochastic local volatility models are the backbone of equity and FX pricing. See how nZetta 's drop-in PDE solver gave a 13x speed-up with minimal code change.

The standard method for multi-ccy baskets requires repeated repairs to correlation matrices. This can make pricing such options very slow. nZetta achieves a massive speed-up on both x86 and GPU.