• Transfer the code to CUDA

  • Issues:
    • Code does not work for large 'R' (Max test-value of R = 700)
    • Error % lies between 10% to 25%, which is too high, as compared to serial processing.
    • Wait for fixes. These issues are linked to the implementation of the curand() functions, but mainly "warp divergence".

• Obtain a larger dataset.

• Then, tweak the theoretical model, if necessary.

P.S. This code is in no way meant to be used in any kind of serious/production work, at least, not in its current state.