Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
-
None
Description
Without the spark reblock, all blocksizes in singlenode remain 0, which requires robustness fixes all over the place. Instead we should properly set the blocksize, and use a dedicated in-memory reblock operation that similar to the existing special case for small data, just reads the matrix in.