I fell like there are a couple of conversations going on in this thread, but then again, I'm also easily confused.
Anyway, if the goal is multi-threading, I only had to disembowel a couple of functions to get the desired results. I've attached a patch that shows what I've done. Andrew may want to avert his eyes... it isn't pretty, but it works. Basically you replace GLPK's memory management with regular malloc() and free() calls within glplib07.c and then you make a couple of hacks in glplib10.c. I left all of the original lines as comments to better see the changes. If anyone tries this, please verify that the patch works as advertised as I may have forgotten some other change I made. Oh, my changes were based on 4.28 I think. My application is a column generation method where there are several subproblems running in parallel communicating with a master problem. It works. I'm running it on Linux, so I can't speak to any other platform. As a side note, to hunt down these issues, I used Valgrind (for memory problems) and Helgrind (for threading issues) (http://valgrind.org/). Very nice open source tools. Also, these conversations from a couple years ago are helpful: http://lists.gnu.org/archive/html/help-glpk/2006-12/msg00018.html http://lists.gnu.org/archive/html/help-glpk/2007-08/msg00040.html Joey
multi_thread.patch
Description: multi_thread.patch
_______________________________________________ Help-glpk mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-glpk
