Interchanging Legate Sparse and SciPy Sparse

While developing applications, it might be beneficial to use SciPy Sparse APIs in the beginning while still prototyping the application and then replace them with Legate Sparse APIs later. This way the user can focus on quickly prototyping the application and then later optimize the application for better performance.

All examples under the examples directory can be run using SciPy Sparse or Legate Sparse using the --package option. For instance, to run the pde.py example using SciPy Sparse, you can run the following command:

legate pde.py --package scipy --throughput --max-iters 25 --warmup-iters 3

Similarly, to run the pde.py example using Legate Sparse, you can run the following command:

legate pde.py --package legate --throughput --max-iters 25 --warmup-iters 3

This will internally replace the import statements from

import scipy.sparse as sparse
import numpy as np

to

import legate_sparse as sparse
import cupynumeric as np

The example scripts under the examples directory all support a special --package command line option to make it simple to switch between Legate Sparse and SciPy Sparse. Additionally, you can change the default for the --package option to enable using the package of your choice by default.