Trilinos: Zoltan2: reversal of Cuthill McKee ordering is incorrect

Created on 4 Feb 2020  路  2Comments  路  Source: trilinos/Trilinos

The reversal of the Cuthill McKee ordering in Zoltan2_AlgRCM.hpp is incorrect. This can result in orderings that make the matrix bandwidth worse than the original bandwidth.

Zoltan2 bug
Source
picture jhux2
👀1

Most helpful comment

This fixed all of our regression tests where reordering was the difference between the solves converging or not with Aztec. Thanks!

picture vbrunini on 4 Feb 2020
🎉1 👍1

All 2 comments

@trilinos/zoltan2

jhux2 picture jhux2 on 4 Feb 2020

This fixed all of our regression tests where reordering was the difference between the solves converging or not with Aztec. Thanks!

vbrunini picture vbrunini on 4 Feb 2020
🎉1 👍1
Was this page helpful?
0 / 5 - 0 ratings

Related issues

bartlettroscoe picture bartlettroscoe  路  3Comments
jjellio picture jjellio  路  3Comments
davydden picture davydden  路  4Comments
kddevin picture kddevin  路  4Comments
klevzoff picture klevzoff  路  3Comments