An Experiment

[rsbriggs]

For two days now my production has been nearly flat at 200K per day. Using -g 25 on all the machines, new comp is in and set up, everything seems very stable and fairly constant.

I noticed earlier on that dumping seems to give you more points. Maybe 20% more. So, without changing any hardware or other configurations, I'm going to take the apps off-line and just dump them from time to time over the next two days, and compare the production.

As I said, it appeared that I get nearly 20% more by folding offline - so... The experiment is to see how my production changes over the next two days.

Ummm - Dyyryath - if you happen to see this, can you tell me when todays production gets closed out ? Does that happen at midnight EST ?? I'd really like to start doing this on "tomorrows" production, rather than affect todays.....

I want more production !!!! :whip:

[bwkaz]

Originally posted by rsbriggs
I noticed earlier on that dumping seems to give you more points. "Dump"ing defined how? Could it be related to the memory leak / file handle leak that's being tracked elsewhere?

What I'm thinking is, if you define "dump" as "I shut the client down, run it with -ut, and start it back up again", then yeah, it makes perfect sense that doing that from time to time will let you run faster. After all, just plain shutting the client down and restarting it seems to help some people that are getting hit with the memory leak issue. ;)

And incidentally, dfGUI's "Upload" button does exactly that, too -- it shuts the client down, modifies the batch file to include -ut (and other options at the DF client defaults), runs it, waits for it to exit, then resets the batch file, and starts the client back up again. Don't know if that's what you're doing or not, but if so, there you go.

Sound plausible to anyone else?