Strange consistent 2nd diagonal behaviour for UCP Timings
Now that the UCP timings work. I did some additional tests on JUSUF. There I got weird timing results for a while, see attached files. Normally timing matrices are transpose symmetric up to errors. In these results, however, this was not the case. The upper right second diagonal showed consistent strange timings, for a variety of messages size and number of nodes. This should be further investigated to make sure this is not an artifact of the code. linktest_jusuf_ucp_32nx1c_1kiB.pdf
linktest_jusuf_ucp_64nx1c_v1.pdf