22 Apr
2012
22 Apr
'12
2:58 p.m.
Thank you all for thoughts and suggestions. I've been tracking them as they appeared, but being busy with university assignments, couldn't try them out yet. In the meantime, however, Karol Samborski investigated the issue further and found the cause of poor performance -- somehow using HashTable.newHint doesn't play well with HashTable.update. Simply changing newHint to hint gives me execution time of about 12s (on my rather slowish AMD Fusion 1.6GHz laptop) with a dataset of ~1.35mln words. It's evidently a bug, as newHint would hardly be expected to slow things down. I'm yet to discover how to properly report it. Thanks again to all who responded and especially to Karol for solving the problem. Best regards, Radek Szymczyszyn