stephenbrooks.orgForumMuon1GeneralLattice download batch file
Username: Password:
Search site:
Subscribe to thread via RSS
[TWH]emveiga
2008-08-09 03:03:21
Would it be possible to create a batch file that goes online and downloads updated lattice files.  The idea is to have something similar to manual_send.bat.

I have some computers that can not be connected to the internet all the time and it would be great to retrieve current lattice files when I do a manual send.
yoyo
2008-08-09 20:37:33
Should be possible to make a Perl script which does it.  Do you also need the sample files for these lattices?
yoyo
[TWH]emveiga
2008-08-10 21:29:22
Perhaps I misunderstand the nature of the lattice files.  For some reason I am under the impression that these change regularly.  The options in the config.txt file suggest that the client should check regularly for updated lattice files and has the option to download sample files as well.

I am just concerned that computers that do not regularly check for new data will return results that are out of date and no longer relevant to the project.

Would the process work better if the script also retrieved sample files?
yoyo
2008-08-12 19:40:06
This should be mainly answered by Stephen.
I think it is ok if your client downloads new lattices later and returns results also for old lattices.
The samplefiles are important, they contain the best simulations of all users.  These are used as starting point for new simulations which leads to better results in a shorter time.  But even without them your client simulates something and uses it's best own results as input for new simulations and gets better and better results over the time.
yoyo
Stephen Brooks
2008-08-13 10:59:19
The lattice files change every few months and are the instructions for the designs the network is currently working on (if you don't have these, you'll end up working on obsolete lattices eventually).  The sample files are good results from elsewhere in the network that may help your program generate better scores.  Some people run without sample files, which can give more "diverse" results, but their optimisation may be slower.

There isn't currently a commandline switch for Muon1 that specifically retrieves the lattice files, though what you could try is running muon1 -init on a machine with network and configured to download these files.  Then it will just do its updating of files as required without any running of simulations.
yoyo
2008-08-14 19:58:41
You can also download the lattices there http://www.stephenbrooks.org/muon1/lattices/
[TWH]emveiga
2008-08-23 03:37:34
Thank you Stephen Brooks.  Running muon1 -init did the trick.  It turns out that the lattices that I had were out of date and I was missing the two newer ones.

Your suggestion is especially helpful in my case because I can ensure that lattices are up to date and return work in one shot.  I can then put away my long network cable before someone trips. 
Zerberus
2008-08-23 23:54:26
Speaking of lattices, is there a place where I can download all the old ones?
Stephen Brooks
2008-08-26 11:17:19
Hmm, funny that -init was originally a debugging switch for my use but has now come in useful to someone else!

I uploaded an archive of my lattices, though there's no "official" way of getting the old ones.  I thought there was no need.
Zerberus
2008-08-27 00:04:19
Thanks!  Yes, sometimes I like to play a little... I set up an extra client, so you won't get rogue results back.

Just have to add the Backend component to the older ones to make them compatible with the recent client, I guess...


Though for some of them I seem to miss something (datafiles?). 
Zerberus
2008-08-27 02:49:21
One example is pi_plus_30GeV.csv.
K`Tetch
2008-09-01 23:58:37
Yoyo said "The samplefiles are important, they contain the best simulations of all users.  These are used as starting point for new simulations which leads to better results in a shorter time.  But even without them your client simulates something and uses it's best own results as input for new simulations and gets better and better results over the time."

Personally, I don't use samplefiles.  While it's good for many people to use them, it's still nice for some people not to use them.  My reasoning is that if everyone uses the samplefiles, we end up concentrating too much on the top 2-3 designs, which may not be the best ones overall.  With me, and others, working independantly, we're looking for other hot spots that might develop into worthwhile designs.  For the same reason, I've urged Stephen many times over the past few years, to delay the introduction of the samplefiles for new lattices, to allow the breadth, or base, of the lattice to spread, and give more variety of slightly developed ones to work from.

In short, everyone using the samplefiles means that if you find a decent area, people are going to focus mainly on it, but it may only be a '5th best area' of the whole lattice design space.  You still need people looking for the 4 (or maybe 20, or maybe none - you don't know) better ones, as well as some concentrating on the known 'hot' areas.  The sad thing is, I'm often the highest person not using the samplefiles, and my systems aren't particularly powerfull (right at this minute, my computers are a 1ghz athlon, and a 2x1.1ghz P3 server)

If people with larger groups of computers would run just one in 20, or even one in 50 of their systems without the samplefiles, to add some low-end spread, and make sure there's no huge gaps, it would probably help identify hotspots all the sooner.
Zerberus
2008-09-02 22:34:53
Running 2 clients on my Dual-Core.  One with samplefiles, one without.
K`Tetch
2008-09-02 22:38:52
good-oh zerberus!
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io RSS feed

Site has had 25161165 accesses.