Old Final Exams

Final Exam - Spring 2010

In this final exam we are going to take another look at optimizations under CUDA.

This final exam is a 'take home' exam. You are expected to work on it completely by yourself. It is due Thursday May 6th 2010 at 11:59pm. By that time you should have set up a web page with your solution, and emailed the location of that web page to Andy. During the scheduled final exam time we will meet in class and you will have 10 minutes to briefly describe your work to the class. We won't have time for a question / answer period; this way we can get through everyone within two hours.

The code we are going to look at is the the Matrix Transposition code from CUDA There is a very nice chapter on optimizing this code included with the CUDA examples and available on the web here:

http://developer.download.nvidia.com/compute/cuda/3_0/sdk/website/CUDA/website/C/src/transposeNew/doc/MatrixTranspose.pdf

Your job is to show how much of an affect those different optimizations have compared to a naive CUDA version of the algorithm with no optimizations on your card, or multiple cards if you chose.

The current code relies on square matrices with each side being a multiple of 32. You should relax this requirement in your code. You should also play with the thread size and thread orientation in a block to see what affect they have on the speed of the solution.

You can either code each of the versions in a single file or in multiple files. Your code should be designed to run from the command line (though launching it from a bat file which contains a command line is fine). Your code timing should include the time to do the transpose, and the timing for the setup and readback. I would highly suggest setting up an automated way to do multiply runs with different parameters so you can gather data more efficiently.


Your grade will be based on the number of optimizations that you evaluate and the quality of the web-based documentation for your testing. When reading through your web site someone new to CUDA should get a better idea where they should focus their energy in optimizing their code.

Be sure to detail which GPU(s) you are using.

Final Exam - Spring 2008

In this final exam we are going to take another look at optimizations under CUDA.

This final exam is a 'take home' exam. You are expected to work on it completely by yourself. It is due Monday December 8th, 2008 at 11:59pm. By that time you should have set up a web page with your solution, and emailed the location of that web page to Andy. During the scheduled final exam time we will meet in class and you will have 10 minutes to briefly describe your work to the class. We won't have time for a question / answer period; this way we can get through everyone within two hours.

The code we are going to look at is the the convolution code from CUDA. There is a very nice paper on optimizing this code included with the CUDA examples and available on the web here:

http://developer.download.nvidia.com/compute/cuda/sdk/website/projects/convolutionSeparable/doc/convolutionSeparable.pdf

The CUDA examples come with 3 convolution examples. The two important ones here are convolutionTexture and convolutionSeparable. convolutionTexture has some optimizations; convolutionSeparable has more optimizations. Your job is to show how much of an affect those different optimizations have compared to a naive CUDA version of the algorithm with no optimizations.

Your code should be designed to run from the command line (though launching it from a bat file which contains a command line is fine.) The code should read in a single image in raw format, apply a filter, and write out the new image in raw format. If it makes things easier you can make use of existing libraries to read in, and write out a standard format (jpeg, tiff, png, ppm, etc.) The time used to compare optimizations should be based on the time taken to do the image conversion, not on the time taken to read in and write out the image. In order to be able to get large enough values for comparison you may need to run the convolution kernels multiple times.

Your grade will be based on the number of optimizations that you evaluate and the quality of the web-based documentation for your testing. You should start with a naive non-separated version, and then apply the changes in the paper (separating the horizontal and vertical work, using shared memory, reducing idle threads, coalescing memory access, unrolling the loops) or start from the optimized version and back-off those optimizations in turn. When reading through your web site someone new to CUDA should get a better idea where they should focus their energy in optimizing their code.

Be sure to detail which graphics card you are using.

I've supplied a test image in multiple resolutions in both tif and raw rgb format in
ftp.evl.uic.edu/pub/INcoming/andy/525/hubble.tar.gz and is about 200 MB

This includes images in the following sizes to allow you to test the performance of the code on a variety of image sizes:
    1024 x 1024
    2048 x 2048
    4096 x 4096
    8192 x 8192

The original image is from http://www.spacetelescope.org/goodies/posters/original/earth.tif


-----

The tutorial is due Monday December 8th at 11:59pm.

At the time of the final we will have two hours, and each person will give a brief 10 minute talk on their work.

Final Exam - Spring 2006


There's a signpost up ahead ...

The final exam is a take home exam where you will write a tutorial. The tutorial should be in the form of a series of web pages similar to the Lighthouse3D pages that we used in class.

The tutorial will take the student though the creation of a series of shaders to build a billboard (i.e. one of those big advertising signs that you see on the side of the highway or a smaller rectangular signboard attached to a post planted in the ground)

-----

the billboard can be a simple rectangular sign or fancier

the billboard can be static or have a moving part, or it could spin. If it animates in any way then that animation must be possible with a real current billboard.

in daylight there is an image / text on the billboard

at night it lights up or is lit up (LEDs, neon, light shining on it, flaming torches, etc)

the content of the billboard is up to you

you can place the billboard in some location if you wish (in a field, on a rooftop, etc)

you can type 'billboard' into the google image search and get a bunch of ideas

-----

You should go through a series of steps in the tutorial to create the final billboard

1 - structure of the billboard with daylight imagery
2 - lighting for nightime
3 - something fancy (motion, animation, more creative lighting)
4 - extras - be creative - show off

The code must be in GLSL and while there will obviously need to be some supporting OpenGL code, the shader code is what is important for your grade. You can use other people's algorithms but all of the code must be yours.

Your web page(s) should include links to working code at the end of each stage, and the important sections of the code should be highlighted and explained in the web page. The idea here is that someone that already has GLSL set up on their computer should be able to go through this tutorial and learn more about what they can do with GLSL.

Implementing the basic stuff (1, 2) gets you a C. Implementing something fancy gets you a B, and doing some extra stuff gets you an A. But remember this is about explaining how you did what you did well, not just doing it well.

The code should be able to run under windows/mac-os x/linux in EVL if I ask for a demo of it.

-----

The tutorial is due at the beginning of our final exam period.

At the time of the final we will have two hours, and each person will give a brief 8 minute talk on their tutorial.


last revision 12/12/11