Skip to content

Latest commit

 

History

History
52 lines (32 loc) · 1.8 KB

README.md

File metadata and controls

52 lines (32 loc) · 1.8 KB

Final Project

Youtube Video Link: https://www.youtube.com/watch?v=Wg8UzDGd_WI&feature=youtu.be

Proposal

  • Teammate 1: Aashish Thyagarajan
  • Teammate 2: Harman Preet Singh
  • Teammate 3: Vipul Sharma

In one sentence, describe your project.

Generating higher order prime numbers utilizing GPU cluster(s).

What is the minimum deliverable for this project?

  • A working kernel implementation to calculate prime numbers over multiple GPUs with focus on accuracy and correctness (we would need the access to the Discovery Lab for the GPU cluster).

  • We will be using the Sieve of Eratosthenes for the prime number calculation.

What is the maximum deliverable for this project?

  • Our focus will be on performance optimization of generating prime numbers.

  • We will try to push the upper limit on calculating higher order prime numbers.

What tools or environment will you be using?

We will need to use:

  • C
  • CUDA API
  • Discovery Lab cluster (multiple GPUs)

What are your first step(s)?

  • Go through the cited publication below to get a better understanding of the project.
  • Learn about parallel programming with GPUs and CUDA.
  • Try and generate small prime numbers in single GPU implementations on our laptops/desktops.

Find at least one related publication or tool and reference it here.

A New High Performance GPU-based Approach to Prime Numbers Generation

What is the biggest problem you foresee or question you need to answer to get started?

  • How to implement parallel programming using CUDA.
  • Ensuring system stability in a multiple GPU environment.
  • Effectively avoiding hardware/kernel implementation bottlenecks.
  • Getting Discovery Lab access and utilizing it effectively for our project.