LAURENU2
02-27-2009, 10:20 PM
I'm reducing the workunit queue to 6 per core. Increasing it to 20 seemed to make the work availability problem worse. I'm thinking that with a smaller number it will take more work requests to clear out the scheduler's work queue. This is just a temporary fix so don't get too worked up about it. Once we get the scheduler's queue increased we'll up the WU queue again.
More... (http://milkyway.cs.rpi.edu/milkyway/all_news.php#160)
More... (http://milkyway.cs.rpi.edu/milkyway/all_news.php#160)