To do setup, submit job:
sbatch setup.sh
To run benchmark with synthetically generated data (fp32)
cd synthetic
sbatch job.sh
To use real data interpolation_products_reduced_correct_pulse.npz (fp64):
cd real_data
sbatch job.sh
The npz file is assumed to be in the real_data directory.
This will generate a dump of the interpolation results called cupy_dump.txt showing interpolation results, including absolute and relative errors compared to the expected results."