CUDA Programming Summer 2013

1. SSH to one of the machines from ug51.eecg.toronto.edu to ug75.eecg.toronto.edu range.

2. Define the necessary environment variables: “source /cad1/CUDA/cuda.csh”

Add the above source command to your .cshrc file so it automatically takes effect at every login. The CUDA compilers and runtime need these variables defined to work properly.

After this stage, you should see CUDA_HOME variable already defined when you run “setenv” command.

3. Install the SDK: “sh /cad1/CUDA/InstallSDK.sh”. Accept the default settings. The script executes and finally gives “* Installation Complete” message. You should see “NVIDIA_GPU_Computing_SDK” directory created in your home directory. It takes around 250 MB of your disk space.

There should be several subdirectories under “NVIDIA_GPU_Computing_SDK”. The CUDA examples are under “NVIDIA_GPU_Computing_SDK/C”.

4. Let’s first build a number of shared libraries.

Change directory by executing “cd NVIDIA_GPU_Computing_SDK/shared”

Compile: “make”

This creates the “release” version of the libraries. Some macros default to nothing in this version. You can also build a version for debugging: “make dbg=1”. If you want to use the emulator you can build the emulated version: “make emu=1” or “make dbg=1 emu=1”. The same options apply to all makefiles found under the CUDA SDK.

Better compile the debug version as well: “make dbg=1”

5. Change directory to “NVIDIA_GPU_Computing_SDK/C/common”.

Compile the libraries: “make” . Ignore the warnings.

Compile the debug version as well: “make dbg=1”. Ignore the warnings.

Read what the library provides: “less cutil_readme.txt”.

6. Now let’s compile some of the examples which can be found under the “NVIDIA_GPU_Computing_SDK/C/src” directory each on its own subdirectory.

Change directory to the bandwidth test example: “cd ../src/bandwidthTest”.

Now you are under NVIDIA_GPU_Computing_SDK/C/src/bandwidthTest”.

Compile the example: “make”

The executable is installed in NVIDIA_GPU_Computing_SDK/C/bin/linux/release

Compile the debug version as well: “make dbg=1”

Run the release version. You should see something like this:

[bandwidthTest] starting… bandwidthTest Starting…

Running on…

Device 0: GeForce GTX 480 Quick Mode

Host to Device Bandwidth, 1 Device(s), Paged memory

 Transfer Size (Bytes)        Bandwidth(MB/s)
 33554432                     2228.2

Device to Host Bandwidth, 1 Device(s), Paged memory

 Transfer Size (Bytes)        Bandwidth(MB/s)
 33554432                     1564.1

Device to Device Bandwidth, 1 Device(s)

 Transfer Size (Bytes)        Bandwidth(MB/s)
 33554432                     119485.0

[bandwidthTest] test results… PASSED

Press ENTER to exit…

---

7. For creating your own new project, follow these steps:

There is a “template” project that you can copy and modify:

(a) Copy the template project: cd ~/NVIDIA_GPU_Computing_SDK/C/src/template mkdir ../myproject cp -r * ../myproject

(b) Edit the filenames of the project to suit your needs mv template.cu myproject.cu mv template_kernel.cu myproject_kernel.cu mv template_gold.cpp myproject_gold.cpp

*** Alternatively, copy the files from the deviceQuery project. This example just uses a .cu file in which you can write C code as well.

© Edit the Makefile and source files. Just search and replace all occurrences of “template” with “myproject”. You’ll need to change the Makefile and file “myproject.cu” that includes the test kernel #include <template_kernel.cu>.

(d) Build the project make

You can build a debug version with “make dbg=1”, an emulation version with “make emu=1”, and a debug emulation with “make dbg=1 emu=1”. Similarly, you can build versions without debugging support using just “make”.

(e) Run the program ../../bin/linux32/release/myproject

(It should print “Test PASSED”)

(f) Now modify the code to perform the computation you require.