From Functional to Parallel: Stochastic Modeling in C++
summary
Takes a legacy MFC single threaded financial app and refactors huge global row arrays into a three tier class hierarchy (CAccount, loans/investments, fixed/variable) while trimming reporting code by thousands of lines and cutting globals. Implements stochastic rate paths but runs them as a side module and uses OpenMP and thread pools to parallelize accounts, not yet inner-account cells. Targets 86 rate environments, needs double precision for correctness, experiments with C++ AMP show hardware and precision limits, GitHub demo forthcoming.
AI-written summary · about
chapters (11)
- 0:00Introduction and speaker background
- 4:13Application overview and legacy codebase
- 12:23Functional coding in legacy C++
- 14:08Current issues and global variables
- 27:03Challenges in legacy code structure
- 33:02Reporting improvements and refactoring
- 37:13Static analysis module and parallel testing
- 41:28Implementing stochastic modeling
- 48:17Hardware limitations and future plans
- 50:00Lessons learned over six years
- 55:40Open source plans and Q&A
related talks
50mLeveraging Modern C++ in Quantitative Finance
01h 02mFunctional C++
58mC++ Standard Parallelism
50mC++ Parallel Programming Models
01h 21mImproving Performance and Maintainability in Modern C++
01h 08mBetter Code: Concurrency
53mMulti-Paradigm Programming and Beyond
01h 02mInside Visual C++' Parallel Algorithms
