GPU exercises
This lab assignment is designed to give you experience leveraging OpenCL to convert a standard C
program for parallel execution on multiple CPU and GPU cores. You will start by profiling a model PDE
solver. Using those performance results you will move portions of the computation over to an OpenCL
GPU device to accelerate the computation. At each step you will analyze the resulting performance and
answer a series of questions to provoke you to think about what is going on.
Instructions
Follow these instructions to complete the computer exercise.Code
Download and unpack this file which contain the code for this exercise.
Work sheet
Links
- Lecture notes: GPU Architectures for Non-Graphics People, Introduction to OpenCL
- The OpenCL Specification v. 1.0
- OpenCL Quick Reference Card
- Nvidia CUDA C Programming Guide v. 3.2
Preparing and submitting a job.
Don't forget to replace "username" and directory names according to your account and own directory names!
| 1. Login to Zorn |
ssh username@zorn.pdc.kth.se |
| 2. Change to your working directory |
cd /cfs/zorn/nobackup/u/username |
| 3a. Copy your source code, |
cp –r ~/yourcode ./ cd ./yourcode/ |
| 3b. or use the sample solution | cd solution_code/ |
| 4. Compile the code |
make clean; make all |
| 5a Open a job script in an editor |
nano job.sh |
| 5b Write into this file the test: |
#!/bin/bash cd /cfs/zorn/nobackup/u/username/yourcode make run |
| 6. Make it executable and submit the job |
chmod +x job.sh qsub –l walltime=10:00 ./job.sh |
| 7. Watch occasionally the job status (every few minutes is enough - it does not change so faster) and wait for the output in your working directory. It will be a file with the name "job.e<job-number>". You saw the job number during the submission. |
qstat |
