Details
-
Task
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
-
None
Description
Operations over large matrices with dimensions > max integer, are always forced to spark which implies missing nnz information due to lazy evaluation in spark. This task aims to proactively recompute the nnz for such large matrices (which are often ultra-sparse) to improve size propagation during recompilation and repeated access.