README (1) - CUDA lab: In this lab, you will use the CUDA...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
CUDA lab: In this lab, you will use the CUDA environment to accelerate the registration of two images. The code attached to this assignment was provided by one of the PhD students in the CS department. He had translated this code from Matlab, and then went on to obtain a 120X speedup over this simple code using a CUDA-enhanced GPU. The code you will be using is not his final version, but a somewhat naive intermediate version that has some inefficiencies beyond simple parallelism. Your job is to accelerate this code, but still achieve the same output for the data set we provide. The image registration problem tries to align two images (a reference and a target) to optimize certain metrics about the two images - in this case, the metric will be similarity. This is an important problem in medical imaging - consider the case where an older MRI of a patient is compared against a new MRI of the same patient. Organs are squishy, moving within the body and patient orientation may not be identical from one MRI to another. For both of these reasons, in order to accurately compare two MRI images, it is first necessary to align these images. If the organ in question is the brain for example, you would like to arrange the images so the two brains (one from each image) are in the same orientation in space. It is not necessary to understand everything that is going on in this code - nor is it necessary to even understand what registration is. You can feel free to treat the code as a black box and simply try and ensure that whatever changes you make do not impact the outcome of the code. Steps: 1. Unpack the tar file downloaded from Courseweb: tar xzf cuda.tar.gz This should create a new directory called cuda - enter the directory cd cuda 2. Compile the program - here you have multiple options: Option a. Default options to compile and run on SEAS machines: To compile in our class environment, you should add the following to your .cshrc file (assuming you are using csh or tcsh) in your home directory: setenv PATH /usr/local/cuda/bin:$PATH
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 06/13/2011 for the course EE 102 taught by Professor Levan during the Spring '08 term at UCLA.

Page1 / 3

README (1) - CUDA lab: In this lab, you will use the CUDA...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online