alpers-git / CS426-Parallel-Computing-PR4

Simple power iteration method implemented for sparse matrices using CUDA 10.0

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

CS426-Parallel-Computing-PR4

Simple power iteration method implemented for sparse matrices using CUDA 10.0

To Compile

cd into directory of cuda file

nvcc powerIteration.cu

To Run

a.exe num_of_threads num_of_iterations print_param test/test.txt

or (for windows)

.\a.exe num_of_threads num_of_iterations print_param test/test.txt

or (for linux)

./a.out num_of_threads num_of_iterations print_param test/test.txt

parameters:
num_of_threads: Number of cuda threads to be executed in a block
num_of_iterations: Number of iterations that is to be executed for power iterations ie. Number of matrix-vector multiplications.
print_param: 1: prints matrix in the sparse matrix form, initial vector(all 1's) and resulting vector after all the iteratons.
2:prints just the resulting vector after all the iterations.
test/test.txt: directory path of test files in the format given in the project4.pdf(test folder contains tests however you can use your own matrices from matrixmarket etc.)

About

Simple power iteration method implemented for sparse matrices using CUDA 10.0


Languages

Language:Cuda 100.0%