[snip]
Then I tried looking at the code in the MessageTallys that were just posted. I tried changing WorldState>>initialize so that the MinCycleLapse is 40 instead of 20 (and re-initializing), expecting CPU usage to be cut in half if I were lucky. But it dropped down to 1%! Squeak's responsiveness still seemed fine from what I could tell, too, although this is a pretty fast machine. Trying some different numbers revealed that there's a big jump (at least on my PC) from setting MinCycleLapse to 31 versus 33. This must be some sort of magic threshold at which the VM starts churning events, or something, I don't know.
Thanks, Bob Arning sent me a fix for this yesterday. I saw similar results - the CPU usage dropped to around 1%. Bob said he was going to drop it in the update stream.
FWIW - MVC still takes all of the machine.
-david
"Pennell, David" wrote:
Then I tried looking at the code in the MessageTallys that were just posted. I tried changing WorldState>>initialize so that the MinCycleLapse is 40 instead of 20 (and re-initializing), ...
Thanks, Bob Arning sent me a fix for this yesterday. I saw similar results - the CPU usage dropped to around 1%. Bob said he was going to drop it in the update stream.
Cool beans.
FWIW - MVC still takes all of the machine.
As far as I know, MVC in Squeak has always done this, so it's not a new problem. (I remember up until 2.4 or so, Morphic did it too...)
- Doug Way dway@riskmetrics.com
squeak-dev@lists.squeakfoundation.org