Title : Technical assistance on using Runner in # machines with Christophe Tribes. Also Waïl Tahmaoui and Edoh.

• Solving library path issues used by Runner (static → dynamic paths adequate to linux 7 & 8).
• I managed thanks to Edoh to run tests on multiple machines without taking the burden to separate manually problems selection.
• Changing parameters/enabling a block of the code in Speculative Search (src files) leads to recompiling Nomad. For testing with Runner, it is better to add flag/double params in src file, compile once & for all.
• Previous benchmarks of MADS w/out Speculative search shows poor performance of this option. That is because of NM & Quad searches which are powerful tools and thus hinder the true performance of Speculative search.
• To plot performance profiles, need to enable/complete code in Runner src files.

ToDo:
* Conducting sensitivity analysis of MADS to the Speculative Search parameters (BASEFACTOR, SEARCHMAX).
* Maybe, documenting tests on # machines with mpirun.