PDA

View Full Version : dfGUI v1.4 for Windows available (With Source)



Digital Parasite
07-03-2002, 07:41 PM
A new version of dfGUI has been released (v1.4) just in time for the 4th of July holidays for all you Americans out there. It fixes a couple of small bugs and hopefully solves the strange benchmarking numbers that a few of you have seen.

You can get it in the regular location:
http://gilchrist.ca/jeff/dfGUI/

v1.4 (July 3, 2002):
- Fixed small bug in refresh rate timing on startup.

- Modified benchmark code to reset and start over if unexpected data was received from DF client. The benchmark data will now show after it has received three progress updates from the DF client.

MD5 Sums:
03c7ac7647c60cbd88bb5cc51092b1f3 dfGUIv14.zip
78e64246ec76b37cee42ee4bb0bdf9d4 dfGUIv14src.zip

If you have any problems please let me know.

Jeff.

stappel
07-06-2002, 06:25 PM
Sorry bad news. after two days of running the 1.4 version I'm again on 404.000 structures a day. And as far as I know there was no upload problem.

Digital Parasite
07-07-2002, 08:07 AM
Well I don't understand. I haven't seen this happen to me so I can't see the circumstances of what may have gone wrong.

What I do for the benchmarking is to take the "Building structure #" value when you first start dfGUI (if the DF client is running) or if dfGUI is already running, when you start the dfClient. I also store the date/time at that point. Now every time dfGUI "refreshes" if it sees the # structures of complete has changed it will look at the current time and calculate the # of structs you have generated since it recorded the first value. It can then determine how long the benchmark code has been running, how many structures you have generated in that time and determine your structs per sec/min/hour/day.

The only what that I can see the benchmark data getting screwed up is if your date/time change or the "Building structure #" value changes to something unexpected. There is now code in v1.4 that should reset the benchmark if your "Building structure #" becomes a value lower than the intial one it recorded (like if you stopped the DF client and re-started it again).

So stappel, just to confirm you left dfGUI running for a couple of days and the DF client itself is set up to automatically upload results so you haven't actually touched it at all (ie: didn't start/stop the client) or click any other buttons on dfGUI itself?

Jeff.

stappel
07-08-2002, 01:59 PM
Originally posted by Digital Parasite
So stappel, just to confirm you left dfGUI running for a couple of days and the DF client itself is set up to automatically upload results so you haven't actually touched it at all (ie: didn't start/stop the client) or click any other buttons on dfGUI itself?

Jeff.

correct. The client is running 4 days and 4 hours now. The gui is running 3 days and 20 hours. no touching on buttons and no time changes. (atleast non that I know of, but I do run ntp). no starting, stopping or any thing else.

Angus
07-12-2002, 01:39 PM
Here's another sample of the 'runaway' structures/time calc:

Digital Parasite
07-12-2002, 01:53 PM
So it seems you are off by about a factor of 10. If your client was running for 2 days 18 minutes and you started dfGUI right away then after doing the number of structures you did it should work out to about 0.57 structs/second or 34 structs/minute.

Hmmmm, I wonder if everyone else's benchmark data is also off by almost exactly a factor of 10. I will look through the code again to see if I can see anywhere it might be affected like using an int when I should be using a long or something.

I'm just not sure why I have never seen it. I don't have any 2GHz machines so maybe my machines aren't fast enough and it doesn't calculate as many numbers as yours in a couple of days so I need see the bug.

I will keep investigating... :crazy:

Thanks for the pic.
Jeff.

Angus
07-12-2002, 05:21 PM
It started out OK, and was still good yesterday evening, showing about 49K per day.

7/11/02 4:29 PM benchmark file....
------------------------------------------------------------
Distributed Folding Windows dfGUI v1.4 Benchmark

Sample Size: 99255 structures over 174065 seconds.

Structures Per Second: 0.57
Structures Per Minute: 34.2
Structures Per Hour : 2053
Structures Per Day : 49268

OS : Microsoft Windows 2000 MHz: 1993
CPU: Intel(R) Pentium(R) 4 CPU 2.00GHz
------------------------------------------------------------

The DF screen was minimized to the toolbar last night.

It's currently showing about 333K per day...

Digital Parasite
07-12-2002, 06:25 PM
Angus, thanks for your benchmark.txt file, if you are still running dfGUI and the benchmark data is still messed up, could you write another benchmark and post that as well?

Thanks,
Jeff.

Angus
07-12-2002, 07:21 PM
Sorry, Jeff....

I just stopped and restarted everything before the weekend.

Bruce