<div dir="ltr"><div dir="ltr"><div dir="ltr"><div>Dear SCIP team,</div><div><span style="white-space:pre-wrap">we</span><span style="white-space:pre-wrap"> have been </span><span style="white-space:pre-wrap">successfully</span><span style="white-space:pre-wrap"> </span><span style="white-space:pre-wrap">using</span><span style="white-space:pre-wrap"> </span><span style="white-space:pre-wrap">SCIP & FiberSCIP & ParaSCIP</span><span style="white-space:pre-wrap"> for a </span><span style="white-space:pre-wrap">long</span><span style="white-space:pre-wrap"> </span><span style="white-space:pre-wrap">time</span><span style="white-space:pre-wrap">.</span></div><div><span style="white-space:pre-wrap">But sometimes some problems arise...<br></span></div><div>Recently we found that FiberSCIP works very very slowly in deterministic mode.</div><div><span style="white-space:pre-wrap">Moreover, it looks like <b><i>Deterministic=TRUE</i></b> (in ug.set file) switches off multithreading and regardless of <b><i>-sth</i></b> setting total CPU load becomes the same as for the case <b><i>-sth=1</i></b>.</span></div><div>But performance becomes even worse than for one threaded SCIP... See example below.</div><div><br></div><div>The question is: maybe there are some other ways to get a deterministic behaviour of FiberSCIP keeping performance? The question arose when we tried to compare FiberSCIP performance and results on different computing environments...</div><div><br></div><div>Below is an example of results for some NLP global optimization problem without discrete variables but with SOS2 constraints:</div><div><span style="font-family:arial,sans-serif">___________________________________________________________</span></div><div><span style="font-family:arial,sans-serif">Presolved Problem :<br> Variables : 145 (0 binary, 0 integer, 0 implicit integer, 145 continuous)<br> Constraints : 61<br>Constraints : Number<br> SOS2 : 15 <br> linear : 30 <br> nonlinear : 16 <br></span></div><div>--------------------------------------------------------------------------------------------------</div><div>It was solved with total 10% gap, <b><i>-sth=16</i></b> and <b><i>Deterministic=FALSE</i></b> in 87 sec:</div><div><div><span style="font-family:arial,sans-serif">_____________________________________________________</span></div>SCIP Status : solving was interrupted [given gap reached]<br>Total Time : 86.80<br></div><div>B&B Tree : nodes (total) : 202112<br>Solution :<br> Primal Bound : +3.10185241045302e-03<br> Dual Bound : +2.98514588645749e-03<br>Gap : 3.90958 %</div><div>----------------------------------------------------------------------------------------</div><div><br></div><div>But solving time with the same settings and <i><b>Deterministic=TRUE</b></i> became almost 50 times bigger:</div><div><span style="font-family:arial,sans-serif">_____________________________________________________</span></div><div><div>SCIP Status : solving was interrupted [given gap reached]<br>Total Time : 4216.07<br>B&B Tree : nodes (total) : 141474<br>Solution :<br> Primal Bound : +3.09812857597093e-03<br> Dual Bound : +2.98180690970866e-03<br>Gap : 3.90105 %<br>----------------------------------------------------------------------------------------</div><div><br></div>Sincerely yours,</div><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div>Vladimir V. Voloshinov,<br>web: <a href="https://scholar.google.ru/citations?hl=en&user=-m4QhNEAAAAJ&view_op=list_works&sortby=pubdate" target="_blank">GoogleScholar profile</a></div></div></div></div></div></div></div></div></div></div>
</div>
</div>