akdimitri/nearest-neighbor-cuda-v.1

Nearest-Neighbor-Cuda-v.1

Nearest Neighbor Grid Implementation in CUDA

Author: Dimitris Antoniadis ~ [email protected]

---

This Project has been implemented in the context of Academic Subject
Parallel and Distributed Systems of Aristotle University of Thessaloniki(AUTH)

---

Project 3

This repository includes a serial CPU implementation and a parallel GPU implementation.

Compilation Makefile The following command creates the parallel program

$ make
or $ make parallel

The following command creates the parallel program using Shared Memory

$ make shared

The following command creates the serial program

$make serial

Cleanup

$ make clean

Execution

$ ./cudaSerial argument1

$ ./cudaParallel argument1

$ ./cudaParallelShared argument1

where d = 2^argument1 and grid size is [d X d X d]

*** Important ***

1)Choose the power of the size of the Array N_power.

2)Open matlabScript.m and change N_power value.(If you wish change the name of the files and the output folder)

3)Run matlabScript

4)Open *.cu file and change the value of N_power and set the desired value of d_power(If you have changed matlabScript's output files, make sure to change them in the folder too. You have to change 4 path strings (2 in main function, 2 in validationFromFile function)

5)Compile the file you wish to run

6)Run the file