When we first implemented matrix_multiply, we put in the restriction that argA, argB, and outputC must all have the same dtype. That works for many semirings, but leaves out things like land_gt, any_overlapi, etc. We should remove this restriction.
A good test for this is making BFS a float64 input, but return int64 parent and levels vectors.
When we first implemented
matrix_multiply
, we put in the restriction that argA, argB, and outputC must all have the same dtype. That works for many semirings, but leaves out things likeland_gt
,any_overlapi
, etc. We should remove this restriction.A good test for this is making BFS a float64 input, but return int64 parent and levels vectors.