https://github.com/soumik12345/several-days-of-cuda

https://github.com/soumik12345/several-days-of-cuda

cuda gpgpu

Last synced: 3 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/soumik12345/several-days-of-cuda
• Owner: soumik12345
• License: mit
• Created: 2021-01-16T09:02:06.000Z (about 4 years ago)
• Default Branch: master
• Last Pushed: 2021-07-27T05:34:19.000Z (over 3 years ago)
• Last Synced: 2025-01-07T01:41:23.811Z (about 1 month ago)
• Topics: cuda, gpgpu
• Language: Jupyter Notebook
• Homepage:
• Size: 1.48 MB
• Stars: 2
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Several Days of Cuda

## Instructions

### Instructions for building locally

1. Download and install cuda toolkit for your platform.

2. Check the GPU you're using by `nvidia-smi -L`.

3. For building on Linux, use `mkdir build && cd build && cmake ../ && make`.

### Instructions for building and running on Google Colab

- In order to build on Google Colab, run the `colab_execution.ipynb` Notebook [![](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/soumik12345/Several-Days-of-Cuda/blob/refactor/notebooks/colab_execution.ipynb)

- In order to develop on Google Colab:

   

   - Run the `Cuda_Workspace.ipynb` Notebook on Google Colab [![](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/soumik12345/Several-Days-of-Cuda/blob/master/notebooks/Cuda_Workspace.ipynb)

   

   - Check the GPU you're using by `nvidia-smi -L`.

   

   - If you wish to ssh into the colab instance from your local VSCode, follow the instructions for `VSCode Remote SSH`. For setting up ssh connection between local VSCode to Google Colab, please follow the instructions this article: [Connect Local VSCode to Google Colab’s GPU Runtime](https://medium.com/swlh/connecting-local-vscode-to-google-colabs-gpu-runtime-bceda3d6cf64)

![](./assets/sample_execution_example.gif)

### Instructions for Development

In order to add a new program to the codebase, you can use a simple Python CLI. Simply install the required 

dependencies for the CLI using `python3 -m pip install -r requirements.txt`. Now you can use the CLI to add a new 

example to the codebase. Executing the `create_kernel.py` using the necessary parameters would add a cuda header file 

with some simple starter code. The usage of the CLI is described below:

```

Usage: create_example.py [OPTIONS]

Options:

  -k, --example_name TEXT  Example Name in Camel Case

  --help                   Show this message and exit.

```

## Examples

   

      Example

      Run Instructions (inside build directory)

   

   

      1

      Hello World

      ./src/hello-world/hello_world

   

   

      2

      Thread-ID Demo

      ./src/thread-id-demo/thread_id_demo

   

   

      3

      Block-Dim Demo

      ./src/block-dim-demo/block_dim_demo

   

   

      4

      Unique Index Calculation

      ./src/unique-index/unique_index

   

   

      5

      Unique Index Calculation in 2D

      ./src/unique-index-2d/unique_index_2d

   

   

      6

      Parallelized Loops

      ./src/parallel-loop/parallel_loop

   

   

      7

      Parallelized Loops in Multiple Blocks

      ./src/parallel-loop-multi-blocks/parallel_loop_multi_blocks

   

   

      8

      Array Manipulation

      ./src/array-manipulation/array_manipulation

   

   

      9

      Array Manipulation with Manual Memory Allocation

      ./src/array-manipulation-manual-memory/array_manipulation_manual_memory

   

   

      10

      Vector Addition

      ./src/vector-addition/vector_addition

   

   

      11

      Vector Addition with Unified Memory

      ./src/vector-addition-unified-memory/vector_addition_unified_memory