Holidays and GPU

Next week I am visiting London.

In the meanwhile enjoy a code preview of A GPU Exercise part 1 and part 2. To get, compile, and test the code, clone my funproject repository on github.

git clone git://

To compile the code is required the NVIDIA CUDA Toolkit. Then run:

cd funproject
cmake .

If CUDA Toolkit was found by CMake, then the lev_distance target should be built into the experiments directory. To quickly test the performances run:

cd experiments
./lev_distance thread.h

The following result is achieved on my late 2010 MBA size is 8716 bytes
thread.h size is 17818 bytes
elapsed time: 9.240 (s)
elapsed time: 0.872 (s)
Results matches!
Distance: 14332

Please note that this is an educational implementation, and several improvements can be done to exploit all the features of the underlying architecture…

But before talking about this… holidays! :)

– 25/09/2012