We might benefit from have GraphBLAS op equivalents of the following NumPy functions:
np.ones
np.eye
np.arange
np.full
Motivation: GraphWave uses np.eye and np.arange. I mentioned np.ones since np.eye(3) == np.diag(np.ones(3)). The others just seemed useful. We'd easily be able to get inspiration for others from the NumPy API.
Paul and I chatted about having these all lower down to a generic sparse matrix creation operation which contains a block that receives coordinates and yields values.
We might benefit from have GraphBLAS op equivalents of the following NumPy functions:
Motivation: GraphWave uses
np.eye
andnp.arange
. I mentionednp.ones
sincenp.eye(3) == np.diag(np.ones(3))
. The others just seemed useful. We'd easily be able to get inspiration for others from the NumPy API.This relates to https://github.com/metagraph-dev/mlir-graphblas/discussions/261.