Default optimization level of intel compiler

We appreciate your help for providing a very fast and accurate interatomic potential MLIP.

Default optimization level of intel compiler is "-O3", and a HPC webpage recommends "-xHost -O2" as it gives a good performance (Ref.: https://wiki.bwhpc.de/e/Development/Intel_Compiler). For the server that I use, "-xHost -O2" gave faster test result than "-O3". Will it also apply to other servers generally?

With "-O3", test time from "time make test" follows:

real    1m26.665s
user    3m38.355s   <----
sys     0m1.968s

With "-xHost -O2" the test time follows:

real    1m13.136s
user    3m4.188s    <----
sys     0m1.703s

Note that the configure command was done with the following command ./configure --compiler=intel --blas=mkl with the intel compiler 19.0, intel mpi 2019.5, and numlib/mkl 2019, and then -O3 option was manually changed in make/config.mk file.