I am working on relatively large dense linear programming problems using IMSL (the Intel bundle version, I think it is 7.0.1) LP routines.
I'm getting good results with DENSE_LP but it is a bit slow, as it isn't parallelized. I'd like to try LPRS. But it fails with a "maximum number of iteration reached" error. Is there any way to modify the maximum number of iterations? I can't find anything in the docs provided by Roguewave and it seems like a silly thing to have a hard-coded max number of iterations. Is there a hidden parameter?
If anyone has any info on that, I'd be grateful.
There is a chapter named "Reference Material" in the IMSL reference manual. In that chapter you will find descriptions of service routines that can be called to alter the handling of warnings, errors and to control output of messages.
Thanks for the comments.
mecej4, the problem is not so much in the handling of the error as in preventing he error from happening by increasing the max number of iteration. But yes, the routines you describe are very useful to prevent full blown crash of the program when the IMSL routines fail in critical mode...
Steve, thanks for the suggestion. I started with Intel as the Readme of the library mentions that support is to be provided by Intel (which I find a bit strange, you have no control over the content of IMSL!). I'll try to contact Roguewave.
Yes, we provide support in that we are your first contact in case of a problem with IMSL that you bought from us. But when you start asking "why" questions and are looking for details on the algorithms, it seems best to have you talk to Rogue Wave directly, otherwise we'd just be the "middlemen" in an exchange with RW.