Why does SIS2 report a negative zero maximum CFL? #230
Description
For some cases, SIS2 is reporting a maximum CFL of -0.000 at the first timestep (when all velocities are zero) for many of our test cases with the gnu and Nvidia compilers. This is easy enough to trap and fix in SIS_sum_output.F90 (with a line like if (max_CFL == 0.0) max_CFL = abs(maxCFL), but I am very curious to know why these negative zeros are coming out of the FMS mpp_max() routine (which we wrap with max_across_PEs() in MOM6/config_src/infra/FMS[12]/MOM_coms_infra.F90. Curiously, this does not appear to impact MOM6.