A CUDA implementation of L0-sampler.
To test performance, edit stream_generator.py as appropriate, then generate stream data file with python3 stream_generator.py. By default, it generates 1 million updates, with indices in the range (0, 100,000).
Then run make l0_sampler, which will compile and run the l0 sampler to process the generated data file.