Scribd
Upload
13 views

Image Blurring Report

Uploaded by

omarobeidd03
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
13 views

Image Blurring Report

Uploaded by

omarobeidd03
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Image Blurring with CUDA - Tiling vs.

Non-Tiling

Objective

The purpose of this lab is to implement and optimize an image blur filter using CUDA. The

optimization is achieved through tiling to reduce global memory accesses. The tiled and non-tiled

versions are compared in terms of performance, memory bandwidth utilization, and global memory

access patterns.

Implementation

The non-tiled version accesses global memory directly for each pixel in the BLUR_SIZE x

BLUR_SIZE region. The tiled version uses shared memory to load a tile of the image, reducing

redundant global memory accesses and improving performance.

Experimentation

Experiments were conducted using a 1024x1024 grayscale image. Execution times were measured

for different tile sizes and blur sizes. Results showed significant performance improvements in the

tiled version.

Results

Execution Times (in milliseconds):

| Tile Size | Blur Size | Non-Tiled Version | Tiled Version |

|-----------|-----------|-------------------|---------------|

| N/A |1 | 23.5 | 15.2 |

| N/A |2 | 45.1 | 28.6 |

| N/A |3 | 71.3 | 45.9 |

| 16x16 |1 | N/A | 12.8 |

| 16x16 |2 | N/A | 24.5 |

| 16x16 |3 | N/A | 39.7 |


Conclusion

The tiled implementation significantly reduced execution times compared to the non-tiled version.

Shared memory usage reduced global memory accesses, leading to better performance. The

experiments demonstrated that tiling is a powerful optimization technique for image processing tasks

on the GPU.

You might also like