[Unicon-group] Unicon set size problem

[Richard H. McCullough]

I have been experiencing a severe slow-down problem as MKE processes large knowledge files.  Most of my tests have been on WinXP, and I had convinced myself (yesterday) that the problem was the WinXP task manager. MKE seemed to be stuck in a repeating cycle of 3 minutes of doing nothing, followed by 30 seconds of creating results.   Today, I ran some Linux tests which suggest that the problem is probably that the sets I am creating are "too big".  If my test (see below) had completed, I would have had a set of 72,000 existents, and a set of 3,300 classes, and a set of 110 properties.  All of the other sets, linking concepts on adjacent levels of the ECP hierarchy, would be much smaller.   I am going to change MKE so that it does not (explicitly) create the two largest sets.  These sets can be counted and created on demand by walking the ECP hierarchy.   However, it seems worthwhile to ask whether the Unicon implementation could be changed to solve this problem.     Here are my Linux results.   input file: tapClass.mcf             2 MB             81,000 lines             72,000 existents              3,300 classes   total time    total lines    lines/15 min. ==========    ===========    =============   0 min.           0              -  15           30,000         30,000  30           40,000         10,000  45           45,600          5,600  60           47,400          1,800  75           49,800          2,400  90           50,900          1,100 105           51,500            600 120           52,200            700 135            150           53,300            550 165           53,900            600 180           54,400            500 195           54,800            400 210           55,200            400   Dick McCullough knowledge := man do identify od existent done; knowledge haspart proposition list;