CHAPTER 1Effective C++

It is often said that quantitative analysts and developers should focus on algorithms and produce a readable, modular code and leave optimization to the compiler. It is a fact that substantial progress was made recently in the domain of compiler optimization, as demonstrated by the massive difference in speed for code compiled in release mode with optimizations turned on, compared to debug mode without the optimizations. It is also obviously true that within a constantly changing financial and regulatory environment, quantitative libraries must be written with clear, generic, loosely coupled, reusable code that is easy to read, debug, extend, and maintain. Finally, better code may produce a linear performance improvement while better algorithms increase speed by orders of magnitude. It is a classic result that 1D finite differences converge in images and images while Monte Carlo simulations converge in images; hence FDM is preferable whenever possible. We will also demonstrate in Part III that AAD can produce thousands of derivative sensitivities for a given computation in constant time. No amount of code magic will ever match such performance. Even in Python, which is quite literally ...

Get Modern Computational Finance now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.