Computes column sums of the matrix stored in a
resident_handle, using a GPU-resident reduction when the
backend supports it. Falls back to base::colSums on the
materialized matrix when no resident reduction is available.
Computes column sums of the matrix stored in a
resident_handle, using a GPU-resident reduction when the
backend supports it. Falls back to base::colSums on the
materialized matrix when no resident reduction is available.