Advancing ray-traced ultra-high vacuum simulations : enhanced algorithms and data structures in molflow / Pascal Bähr. Wuppertal, 06. Dezember 2023
Inhalt
- Kurzfassung
- Contents
- Nomenclature
- Introduction
- Background
- Vacuum
- Monte Carlo simulations with Molflow
- Motivation
- The Test Particle Monte Carlo method
- The Monte Carlo model
- Time-dependent simulation
- Pseudo Random Number Generator
- State of the art
- Case studies
- Synrad
- Ultra high vacuum and Synchrotron radiation problems
- Ray tracing
- Basics
- Coordinate system
- Ray tracing algorithm
- Line-plane intersection
- Point in polygon
- Line-box intersection
- Acceleration Techniques
- Acceleration Data Structures
- State of the art - Ray tracing
- Contributions
- Software Development Process
- Code Refactoring
- Molflow CLI
- Automated testing and GitLab CI/CD
- Parallel and distributed computing
- OS x Compiler
- Profiling
- Test cases
- Design and Development
- Iterative Simulations
- Advanced Ray Tracing techniques
- Construction
- Splitting heuristics
- General Cost Function for ADS Splitting Heuristics
- Surface Area Heuristic (SAH)
- Ray Distribution Heuristic (RDH)
- Hit Rate Heuristic
- Hybrid heuristics
- Binning
- Discussion
- Evaluation and Benchmarks
- Outlook
- GPGPU Kernel
- GPU Basics
- Threads
- Blocks
- Warps
- Streaming Multiprocessors
- Memory Space
- Cooperation and Synchronisation
- Best Practices in CUDA Programming
- Ray tracing on GPUs
- Hardware accelerated RT
- GPU ray tracing with OptiX 7
- OptiX Basics
- Integration Process
- Challenges in OptiX integration for Molflow
- GPU Kernel development
- Neighbour Aware Offset (NAO)
- Conclusion and Future Work
- Bibliography
- CERN hardware