Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/andreasholt/cusmc
A CUDA-accelerated Statistical Model Checker for Stochastic Timed Automata
https://github.com/andreasholt/cusmc
cuda smc
Last synced: about 1 month ago
JSON representation
A CUDA-accelerated Statistical Model Checker for Stochastic Timed Automata
- Host: GitHub
- URL: https://github.com/andreasholt/cusmc
- Owner: AndreasHolt
- Created: 2024-09-16T06:21:48.000Z (5 months ago)
- Default Branch: main
- Last Pushed: 2024-12-17T14:49:59.000Z (about 2 months ago)
- Last Synced: 2024-12-17T15:42:52.830Z (about 2 months ago)
- Topics: cuda, smc
- Language: C++
- Homepage:
- Size: 1.51 MB
- Stars: 2
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
A CUDA-accelerated Statistical Model Checker for Stochastic Timed Automata
## Overview
cuSMC implements GPU-accelerated statistical model checking for stochastic timed automata. The tool parallelizes simulations through:
- Mapping model components to CUDA threads within blocks representing simulation runs
- Each thread is responsible for its component's entire behavior including:
- Evaluating guards on outgoing transitions
- Selecting and taking enabled transitions
- Sampling delays
- Updating component state and synchronizing with other components
- Using shared memory for efficient component state management, sampling and synchronization between threads
- Maintaining an optimized model structure in global memory with:
- Array-of-structs design for node, edge, guard, invariant and update data with cache-friendly access patterns
- Contiguous memory layout for components at each network level
- Read-only data caching through restrict and const qualifiers
- Use of __ldg instruction for improved global memory loads of expression data
- Supporting broadcast synchronization through shared memory
- Thread-level sampling and component race resolution through:
- Per-thread uniform distribution sampling for bounded delays
- Per-thread exponential distribution sampling for unbounded delays
- O(log n) parallel reduction to determine race winners
## Acknowledgements
cuSMC utilises [SMACC's](https://github.com/Baksling/P7-SMAcc) UPPAAL XML parser. We also utilise their structure for expressions and their helper functions to evaluate non-PN and PN expressions.
## Requirements
- C++17 compiler
- CUDA Toolkit (≥11.0)
- CMake (≥3.18)
## Building
```bash
mkdir build
cd build
cmake ..
make
```
## Running
To get help with the arguments run:
```bash
./cuSMC -h
```
This could look like:
```bash
./cuSMC -m "../xml_files/agentBaseCovid_100_1.0.xml" -q 1000 100 -1 1 1 1 " "
```