Skip to content

v5.0.5 (May 17, 2021)

Choose a tag to compare

@DrTimothyAldenDavis DrTimothyAldenDavis released this 18 May 13:03
· 3445 commits to stable since this release
bc64ba2

Version 5.0.5, May 17, 2021

* (26) performance bug fix:  reduce-to-vector where A is hypersparse CSR
    with a transposed descriptor (or CSC with no transpose), and some cases
    for GrB_mxm/mxv/vxm when computing C=A*B with A hypersparse CSC and B
    bitmap/full (or A bitmap/full and B hypersparse CSR), the wrong
    internal method was being selected via the auto-selection strategy,
    resulting in a significant slowdown in some cases.