stephenbrooks.orgForumMuon1Generaltime difference
Username: Password:
Search site:
Subscribe to thread via RSS
tron
2003-04-22 09:30:51
i just discovered a timedifference between the muon1.exe and muon1_background.exe.  muon1.exe needs 140 seconds to complete a simulation, muon1_background.exe almost needs 6 minutes.  i cant imagine why...
px3
2003-04-22 09:56:19
maybe different simulation settings ?

PX3
Pascal
2003-04-22 10:08:46
The background client should be the fastest, if you compare the clients with identical simulations. 
If you use the console client have a look at it, that it is the active window (in the foreground). 
The task manager tells you something about CPU usage, and if there are any other applications (clients??), that may take some CPU cycles.

___________________________
[DPC]Stephan202
2003-04-22 10:32:06
Are you saying that, on identical simulations, muon1_background.exe is faster than muon1_cmdline.exe?

---
Dutch Power Cow.
MOOH!
px3
2003-04-22 10:34:34
Sure, muon_background is faster.  no screenoutput, that increases speed Cool

PX3
tron
2003-04-22 11:16:21
but thats the thing...
same simulation, 100% cpu usage, and nevertheless something seems to slow it down...
px3
2003-04-22 11:28:50
Hi Tron,

which simulation type was used ?

PX3
tron
2003-04-23 10:26:22
what do you mean by simulation type?  the decay channel is simulated and you cant really change this in the current version, can you?
MaFi
2003-04-23 11:54:48
hi tron
the duration of a simulation depends on how good the design is you are simulating.  the simulation of a "good" design (with a high muon yield) takes longer because more paritcles have to be simulated, while in a bad design lots of particles are lost -> fewer particles -> less to compute -> simulation finishs faster. 
use the comandline version and take a look in your results.txt and you'll soon see what I'm talking about.
markus
tron
2003-04-23 13:07:40
yeah, i know what you mean, but im not talking about a single simulation but about a whole row of 500. nevertheless, it seems to work fine with the command line version.  thx Wink
[DPC]Stephan202
2003-04-23 22:51:05
Even in a row of 500 simulations differences can be seen, because the client mutates previous (good) results, so the first 500 results are likely to be finished faster than the second 500.

This theory of course doesn't work when you deleted results.dat after the first 500 simulations.  But even then an amount of luck (a very high yield in the beginning of the row - e.g. the 20th simulation) can make a huge difference.

---
Dutch Power Cow.
MOOH!
Ben Whitten
2003-04-24 03:34:33
Both My computer have hit 100 results (taking them forever) and the best they have done is 0.5xx, roughly when will they start to get high yield results in your experiance.

Ben
[DPC]Stephan202
2003-04-24 03:39:59
Yields will raise very slow, unless you are either very lucky, or you use a topXXX file, which means you share your results with others and let your client mutate and combine the best results.  Currently Stephen does not distributed such file with top results.  You'll have to wait for that, or join a DC-team that has set up a collection and distribution system for this.

---
Dutch Power Cow.
MOOH!
tron
2003-04-24 09:12:38
it doesnt make much sense to use a topxxx file anyway.  the goal of the whole thing is to discover various different ways...
[AMD Users]Core
2003-04-24 10:18:18
yeah, and it's cheating too.  if you want good results, get them yourself, dont use somebody else's results as a shortcut
[DPC]Stephan202
2003-04-24 11:03:22
Then why did Stephen release a top250 and top1000 file with the previous version (v4.2x) of the client?

I cannot deny that not using a topXXX file is better variety of the results.  However, to find the optimal simulation a LOT of mutations need to be tested.  When there's no input from outside, small users will never achieve this (only with an enormous amount of luck).  If you think that’s the way to go, then you are saying that everyone with only one computer can better go to distributed.net or something.  Sounds kind of dumb to me, less scientific at its least.

Also, besides the fact that users with a lot of computers have a 'natural' advantage in the stats (more CPU cycles), this effect will grow exponentially when the sharing of results is not allowed.  After all, higher yields are better credited in the stats.

Yes, it kind of pisses me off when you say it's cheating.  Coz it isn't. It's NOT illegal to share results/configurations.  That's written nowhere.  Cheating means bypassing the rules in an illegal way.  That's not what's happening.

From the scientific point of view it might seem less preferable to share results, but it's a necessity to small users.  Remember, if those small users wouldn't use a topXXX file, they would be useless.  Then they might just as well stop participating.  Also, it DOESN'T negatively affect the project if (small) users DO use a topXXX file.

Take a look at it like this: these topXXX files, currently created on a team basis (until Stephen releases such a file), bring competition to the project.  Competition means that people get involved.  Assimilate more computers.  Tell friends about it.  Tell team mates about it.  Now THAT'S good for the project.

---
Dutch Power Cow.
MOOH!
tron
2003-04-24 13:57:49
pfff cheating.  are you refering to the stats?  dont worry- they dont work Wink but really- its not the stats why im contributing to this project...
rsarkezi
2003-04-25 12:40:24
Hmm, maybe the stats should be sorted in alphabetical order so it would be regardless wether you have 1 Billion Mpts or only 1 thousand!

-----------------------------
Be efficient!

Team www.it-academy.cc
www.it-academy.cc
Stephen Brooks
2003-04-25 13:23:04
The various things people do to the results (exchange of the best ones plus even some manual seeding has gone on), don't seem to harm the project too much.  Actually the results-sharing can sometimes allow higher percentages to be reached faster (that's why I release a full-project bestNNN file once it's running smoothly).  But it's also true that the users who do it on their own are important because they might go up a different pathway and discover something entirely new.  As it happens, the Mpts scoring system is fair and really is proportional to the work your CPU has done (regardless of muon % score).  In the last project I think using a best-file did get you a few more points for your power.

Today's weather in %region is Sunny/(null), max.  temperature -99999°C
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io

Site has had 16834411 accesses.