[FlopCpp] FLOP C++ Generation Time Issue

[Tim Hultberg]

Hi Rohith,
   Normally the generation time should be proportional to the number of non-zeros – that also seems to apply quite well when going from your case-1 to case-3. Would it be possible to try cases much bigger than case-5? One reason for a sudden increase in generation time could be related to memory consumption.

There were recently some mails on this list related to generation time – they might be relevant for you as well. It is possible that you could speed up things by removing the ‘such_that’ and use sparse sets (I.e. MP_subset )instead, but it is hard to tell without looking at the complete model.

Regards, Tim

From: flopcpp-bounces at list.coin-or.org [mailto:flopcpp-bounces at list.coin-or.org] On Behalf Of Rohith Kumar.H.C
Sent: Monday, October 04, 2010 10:37 AM
To: flopcpp at list.coin-or.org
Cc: rohitkumar.hc at in.abb.com
Subject: [FlopCpp] FLOP C++ Generation Time Issue

Hello,
We are incrementally building a complex model using FLOP C++.
It has been noticed that, as new constraints and additional terms for objective function are added the Flop C++ Generation time increases.
The point to worry is that the increase is quiet sharp with change in problem size.

The analysis data has been presented below:

                                      Case-1    Case-2    Case-3    Case-4    Case-5
Rows                              6409       6409       12793     13177    13897
Columns                         6433       7657        14089    14233    14953
Non-Zeroes                     14729     18401       42081    43881    45465
Generation Time (secs)    0.145      0.150       0.420     0.562     1.049

We would like to know if this generation time can be reduced?
If yes, then what measure need to be taken to bring this down?

The constraints modeled use such_that to restrict the range of constraint. Does this effect the overall performance?

Also what could be the other possible reasons for increased generation times.

Regards
Rohith
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/flopcpp/attachments/20101006/c350e38e/attachment.html