https://github.com/lyynn777/cuda-bitonic-sort
Simple CUDA project to implement Bitonic Sort and compare it with normal CPU sorting.
https://github.com/lyynn777/cuda-bitonic-sort
bitonic-sort cuda gpu-computing gpu-vs-cpu parallel-computing performance-testing pycuda python
Last synced: 2 months ago
JSON representation
Simple CUDA project to implement Bitonic Sort and compare it with normal CPU sorting.
- Host: GitHub
- URL: https://github.com/lyynn777/cuda-bitonic-sort
- Owner: Lyynn777
- Created: 2025-11-02T09:55:13.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2025-11-02T10:12:47.000Z (2 months ago)
- Last Synced: 2025-11-02T11:41:29.847Z (2 months ago)
- Topics: bitonic-sort, cuda, gpu-computing, gpu-vs-cpu, parallel-computing, performance-testing, pycuda, python
- Language: Jupyter Notebook
- Homepage:
- Size: 179 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# CUDA Bitonic Sort — GPU Sorting in Google Colab
This project implements the **Bitonic Sorting Algorithm on GPU using CUDA (PyCUDA)** and compares performance against CPU sorting.
The goal is to understand **parallel sorting networks** and observe when GPU parallelism becomes beneficial.
---
## **Project Objectives**
* Implement **Bitonic Sort** in CUDA using PyCUDA
* Sort arrays and verify correctness vs. CPU sorting
* Measure execution time for multiple input sizes
* Plot **CPU vs GPU runtime graph**
* Run fully in **Google Colab** (Tesla T4 GPU)
---
## **Why Bitonic Sort?**
Bitonic Sort is chosen because:
* Perfectly fits **parallel execution model**
* Regular & predictable memory access
* No recursion or branching complexity
* Ideal for learning GPU sorting architectures
It's used in academic & research demos for **GPU parallel algorithms**.
---
## **Tech Stack**
| Component | Details |
| ------------- | --------------- |
| Language | Python |
| GPU | NVIDIA Tesla T4 |
| CUDA Library | PyCUDA |
| Visualization | Matplotlib |
| Environment | Google Colab |
---
## **Project Structure**
```
cuda-bitonic-sort/
├── README.md
├── bitonic_sort.ipynb
├── images/
│ └── sort_time.png
│ └── sorting_time_comparison..png
```
---
## **Running the Project in Colab**
### 1️⃣ Load GPU
```
Runtime > Change runtime type > GPU
```
### 2️⃣ Install Dependencies
```bash
!pip install pycuda
```
### 3️⃣ Run Notebook
Open: `bitonic_sort.ipynb`
---
## **Results Overview**
| Array Size | CPU Time | GPU Time | Correct? |
| ---------- | -------- | -------- | -------- |
| 512 | ✅ | ✅ | True |
| 1024 | ✅ | ✅ | True |
| 2048 | ✅ | ✅ | True |
| 4096 | ✅ | ✅ | True |
| 8192 | ✅ | ✅ | True |
| 16384 | ✅ | ✅ | True |
**Observation:**
* GPU is slower for small inputs (kernel overhead)
* GPU becomes beneficial as input size increases
* Demonstrates parallel scalability behavior
---
## **Performance Graph**
---
## **Learning Outcomes**
By completing this project, you achieved:
* Understanding of **GPU threads & blocks**
* Knowledge of **parallel sorting networks**
* Experience with **PyCUDA kernel programming**
* Performance benchmarking & graphing
---
## **References**
* NVIDIA CUDA Programming Guide
* PyCUDA Documentation
* Bitonic Sorting Network Theory
---
## **Conclusion**
This project demonstrates how GPU parallelism behaves for sorting tasks.
It’s a **simple but powerful** introduction to CUDA-based parallel computing.
---
### ⭐ If this helped, star the repo!