README.md

CUDA_Box-filter

This repository has been created for college assignments in the Advanced Computer Architecture subject during the summer semester of 2022/2023.

Exercises

• Exercise 1 - CPU threads
• Exercise 2 - Thread synchronization and mutex

Project

Project is demonstrating implementation of the box filter for

• CPU
  • Single thread
• GPU
  • Without shared memory usage
  • With shared memory usage

Project is written in the form of Jupyter Notebook and developed with Google Colab (GPU).

👉 Project Notebook

Key features

• filter results generated by different algorithms are asserted using checksum inside Google Colab Notebook, to skip slow file download from Google Colab
• CPU/GPU speed comparison analytics are build automatically with python script

Results

Following 4 images demonstrate filter usage (original images were resized).

Original	Filter radius 4 px
Filter radius 7 px	Filter radius 14 px