Quote* your post is about having a flexible engine, which gets configured (runtime/compiletime), and then need to numbercrunch with high efficiency...It's about the archictecture and design drawings and interfaces that wil be mapped to C++, C# , Matlab etc. This process is well-known and documented. A well-designed system leads to an efficient (and maintainable!) one.Quote* a nice set of information containers and manipulation bits that setup the engine.Maybe, this look like detailed design, which is the easy part. I want to avoid premature optimisation. QuoteCan we work on a case? Eg the price of digital barrier option? With implicit, explicit Euler and with MC?That would allow use to define elements, and put things at the correct place? This is possible (and has been done, see bespoke chapter 0 MC!) but I don't need code examples in order to discover requirements as I have a number of frameworks at design level which I want to 'stress-test' by inspecting the UML class and component diagrams. Simultaneously, I have a number of Proof-Of-Concept (POC) programs to test the design.QuoteEg, a digital barrier contract has:* barrier* expiration date...we can use doubles for both for now.It has a payoff of $1 at expiration if max(St) 0<t<=T is at least B We could decide to sample the max(St) distribution, or we can make various path samples with discrete time steps, and approximate the distribution.There is also a process, eg gbm or Heston, that's another element... That's no problem and low-risk; more pressing is finding the components and their interfaces. See again Chap 0 POC.Question:Do quants 'see' a system as multiple lines of code or have they some _back_of_envelope_ design model??
Last edited by Cuchulainn
on October 22nd, 2011, 10:00 pm, edited 1 time in total.