stephenbrooks.orgForumMuon1GeneralHigh-res scan project
Username: Password:
Page 1 2 3 4 5 6
Search site:
Subscribe to thread via RSS
Hydrox
2010-10-03 01:39:21
700-749 done and send by mail
Stephen Brooks
2010-10-03 12:46:48
A quick analysis I put up in the directory as a spreadsheet shows that if you fit a quadratic through the 300-550 range, it suggests an optimum of 443 not 490 in this parameter, so if there were a way to get through the noise it could add about 0.025% to the score.
K`Tetch
2010-10-03 18:53:14
Just sent mine back via email
Stephen Brooks
2010-10-03 21:13:41
Time for a graph on this page, including K`tetch 600-649 and Hydrox 700-749. (URL is the same)



The first recheck of 500-549 from [DPC]white_panther has given identical results except for 3 occasions where #runs=5 in one case but 1 in the other.
Stephen Brooks
2010-10-03 22:32:08
Added RGtx 0-99
Stephen Brooks
2010-10-04 18:19:59
Here's a second scan to try. This time, you'll need to download the new Linac900Ext10d2_axial3 lattice, where I've used a simple trick to multiply up the number of original particles in the simulation by three.  So the simulations will probably take longer.  I'm interested in seeing if this will reduce the noise (it won't be by large factor: 1/sqrt(3) at best).
Stephen Brooks
2010-10-04 18:21:08
Quick instructions: put Linac900Ext10d2_axial3.txt in your lattices directory; pick a queueNNN-NNN.txt file and say you've claimed it on this thread; rename your queueNNN to queuecli.txt in the Muon1 directory and then run muon1 -c -q; send me back the results.txt file when it finishes.
Stephen Brooks
2010-10-04 18:26:50
I'm claiming queue270-299 for my workstation.
tomaz
2010-10-04 18:56:28
I took 300-319.
waffleironhead
2010-10-04 19:40:07
Is it strange that I never got credit for doing my chunk yet?
I uploaded shortly after I sent the file in.
[Edited by waffleironhead at 2010-10-04 19:40:40]
Stephen Brooks
2010-10-04 19:53:31
I got an e-mail from you that said "Attached are the results for 900-924" and they are in the graph above.

You mean in your Muon1 Mpts score?  I could have a look tomorrow, I guess, to see if they've made it in.
RGtx
2010-10-04 20:08:02
0-29 Taken.
Hydrox
2010-10-04 21:57:30
950-999 taken
[TN]steinrar
2010-10-04 22:08:55
Just to confirm 500-549 has been rechecked and sent by email to you Stephen
[DPC]white_panther
2010-10-04 23:16:30
0-29 RGtx
30-59
60-89
30-119
120-149
150-179
180-209
210-239
240-269
270-299 Stephen Brooks
300-319 tomaz
320-339
340-359
360-379
380-399
400-419
420-439
440-459
460-479
480-499
500-549
550-599
600-649
650-699
700-749
750-799
800-849
850-899
900-949
950-999 Hydrox

when i'm home ill take also a file
[DPC]white_panther
2010-10-04 23:58:00
30-59 taken

0-29 RGtx
30-59 [dpc]white_panther
60-89
30-119
120-149
150-179
180-209
210-239
240-269
270-299 Stephen Brooks
300-319 tomaz
320-339
340-359
360-379
380-399
400-419
420-439
440-459
460-479
480-499
500-549
550-599
600-649
650-699
700-749
750-799
800-849
850-899
900-949
950-999 Hydrox
waffleironhead
2010-10-05 01:29:32
yeah, stephen I saw my results in your graph, I just havnt seen any credit in the MPTS stats.  It was easy to see as I havnt been active for quite some time.  In 7874 hours to be exact. 
Hydrox
2010-10-05 02:01:33
as the first run took 7573.1 Mpts I would appreciate it, if a result good result could be posted on completion, to minimized reruns.
Boots
2010-10-05 07:21:08
360-379 taken

0-29 RGtx
30-59 [dpc]white_panther
60-89
30-119
120-149
150-179
180-209
210-239
240-269
270-299 Stephen Brooks
300-319 tomaz
320-339
340-359
360-379 Boots
380-399
400-419
420-439
440-459
460-479
480-499
500-549
550-599
600-649
650-699
700-749
750-799
800-849
850-899
900-949
950-999 Hydrox

[Edited by Boots at 2010-10-05 07:26:22]
tomaz
2010-10-05 10:45:39
320-339 taken
Cameron
2010-10-05 11:23:31
Taking Queue420-439.

After the Hi-Res Run Is this Axial3 lattice going to replace Linac900Ext10d2 or is it just for the Hi-Res Only.
[Edited by Cameron at 2010-10-05 11:31:46]
Stephen Brooks
2010-10-05 12:47:30
This lattice is just for the high-res scan.  Here's a 2.14% result for your results.dat if you want it.

sb9r=349;sb9l=647;sb9f=932;sb8r=426;sb8l=149;sb8f=390;sb7r=797;sb7l=295;sb7f=945;sb6r=797;sb6l=295;sb6f=945;sb5r=724;sb5l=457;sb5f=618;sb4r=974;sb4l=049;sb4f=670;sb3r=000;sb3l=999;sb3f=187;sb2r=000;sb2l=999;sb2f=990;sb1r=000;sb1l=631;sb1f=103;sb10r=096;sb10l=647;sb10f=932;s2r=234;s2l=992;s2f=000;s1l=999;s1f=996;rf9v=235;rf9p=000;rf8v=715;rf8p=118;rf7v=571;rf7p=688;rf6v=043;rf6p=583;rf5v=604;rf5p=940;rf4v=991;rf4p=849;rf3v=918;rf3p=702;rf2v=537;rf2p=000;rf1v=992;rf1p=809;rf10v=926;rf10p=855;ls9l=992;ls9f=000;ls8l=431;ls8f=000;ls7l=629;ls7f=995;ls6l=875;ls6f=036;ls5l=275;ls5f=997;ls4l=761;ls4f=999;ls3l=991;ls3f=984;ls2l=898;ls2f=997;ls1l=992;ls1f=000;ls10l=999;ls10f=000;linaccells=997;ld9=509;ld8=231;ld7=751;ld6=999;ld5=048;ld4=483;ld3=552;ld2=480;ld10=000;ld1=000;decaycells=018;db9l=000;db8l=255;db7l=817;db6l=734;db5l=178;db4l=403;db3l=400;db2l=137;db1l=446;db10l=097;d4l=679;d5l=679;d3l=958;d2l=169;d1l=449;tantalumrodz=222;tantalumrodr=000;#runs=1;
2.148691 (7219.2 Mpts) [v4.44e] <Linac900Ext10d2_axial3> {2C1800C5A912C714C9A68C80}
Stephen Brooks
2010-10-05 14:50:06
waffleironhead - Keep your eyes on the stats.  I've just found some of our scan results had been put to one side in the stats checker because they looked "wrong" to it (probably too many Mpts for a low scoring result).  I've relaxed those constraints now and resubmitted a bunch of files.
tomaz
2010-10-05 17:18:16
If there is no need or interest in #runs=5 and if we were clever than we should first compute 490 or something near, get highest result and use it in all results.dat.
Stephen Brooks
2010-10-05 17:46:59
Yes, why don't you take 480-499 next time?
waffleironhead
2010-10-06 02:17:11
Thanks Stephen, The stats have showed up.
Cameron
2010-10-06 05:28:57
should autocli.sav be read in at some point if it exists when your operating muon1 -c -q.

Because I don't think it is.  muon1 seems to ignore the autocli.sav file entirely although when it reads in the queuecli.txt file it is aware it currently needs to recheck the first result.

I noticed this happening with the previous Hi-Res Scan but as there was an existing best result most (all) of the results only ran the once and at approximately 2500 Mpts it didn't take to long to check one result.  Now with 7500Mpts for a single run waiting for the end of the run takes a very long time.

My first result that initiated a recheck (still to recheck and average out) is 2.4% and I've no idea if it will trend up or down over the sample of 20 (I think up but *shrugs*)
[Edited by Cameron at 2010-10-06 05:32:43]
tomaz
2010-10-06 06:05:22
Cameron, trend in your set is probably up (check graph above).  I am computing 490 right now.  I will post result in about 5 hours.  If you use it there will be no runs=5 except maybe very few in 450-550 range.

I could't wach rechecking of 2.2% for 20 hours...
tomaz
2010-10-06 10:18:58
Here is result for 490:
sb9r=349;sb9l=647;sb9f=932;sb8r=426;sb8l=149;sb8f=390;sb7r=797;sb7l=295;sb7f=945;sb6r=797;sb6l=295;sb6f=945;sb5r=724;sb5l=457;sb5f=618;sb4r=974;sb4l=049;sb4f=670;sb3r=000;sb3l=999;sb3f=187;sb2r=000;sb2l=999;sb2f=990;sb1r=000;sb1l=631;sb1f=103;sb10r=096;sb10l=647;sb10f=932;s2r=234;s2l=992;s2f=000;s1l=999;s1f=996;rf9v=235;rf9p=000;rf8v=715;rf8p=118;rf7v=571;rf7p=688;rf6v=043;rf6p=583;rf5v=604;rf5p=940;rf4v=991;rf4p=849;rf3v=918;rf3p=702;rf2v=537;rf2p=000;rf1v=992;rf1p=809;rf10v=926;rf10p=855;ls9l=992;ls9f=000;ls8l=431;ls8f=000;ls7l=629;ls7f=995;ls6l=875;ls6f=036;ls5l=490;ls5f=997;ls4l=761;ls4f=999;ls3l=991;ls3f=984;ls2l=898;ls2f=997;ls1l=992;ls1f=000;ls10l=999;ls10f=000;linaccells=997;ld9=509;ld8=231;ld7=751;ld6=999;ld5=048;ld4=483;ld3=552;ld2=480;ld10=000;ld1=000;decaycells=018;db9l=000;db8l=255;db7l=817;db6l=734;db5l=178;db4l=403;db3l=400;db2l=137;db1l=446;db10l=097;d4l=679;d5l=679;d3l=958;d2l=169;d1l=449;tantalumrodz=222;tantalumrodr=000;#runs=5;
2.489781 (38278.4 Mpts) [v4.44d] <Linac900Ext10d2_axial3> {B26F6D6E110EF2639F253DED}

It is not above 2.5 but it will do the job.
Since I allready got my hands on coworker's i7 for that job I'll do whole 480-499 (it's only 1h per run

0-29 RGtx
30-59 [dpc]white_panther
60-89
30-119
120-149
150-179
180-209
210-239
240-269
270-299 Stephen Brooks
300-319 tomaz
320-339 tomaz
340-359
360-379 Boots
380-399
400-419
420-439 Cameron
440-459
460-479
480-499 tomaz
500-549
550-599
600-649
650-699
700-749
750-799
800-849
850-899
900-949
950-999 Hydrox
[Edited by tomaz at 2010-10-06 10:20:32]
K`Tetch
2010-10-07 02:08:49
I'll take 60-89
[Edited by K`Tetch at 2010-10-07 02:09:17]
tomaz
2010-10-07 14:11:23
460-479 and 500-549 taken
K`Tetch
2010-10-07 17:41:48
bit of a 'yikes' - these are about 6500mpts a run.  I've been running one for 14 hours and it's halfway through the 4th of 5 runs, on a Q6600
Stephen Brooks
2010-10-07 18:10:57
K'tetch - yes, these should be about 3x as long as "normal" results.  I just received 480-499 from TOMAZ.
[DPC]white_panther
2010-10-07 18:49:18
what kind of memorey are you using, mine Q6600 take 1 hour per run
[DPC]white_panther
2010-10-07 20:07:32
sorry i mend 2 hour per run

would be nice a Q6600 preforming like an i7
[Edited by [DPC]white_panther at 2010-10-07 20:08:33]
Cameron
2010-10-08 03:26:32
So Stephen is it alright to use Tomaz 490 result in order to cut-down possible re-checks? 

My ceiling is currently 2.428% so the 490 result maybe not that helpful in cutting down rechecks (but I need a larger population in the results.txt)

I've just started muon1 [4.44d] -c -q for todays session.  [date 08-October-2010]
It's checking the fifth run in the original queue420-439

I'll preserve the autocli.sav file after todays session and see where the simulation starts tomorrow.  If it begins at 0.00 (I'll make sure the save file in the middle of a 7500 Mpt Run) I can send it to you if needed.  Due to it being ignored (and then overwritten by the first autosave of the new session.
K`Tetch
2010-10-08 03:54:16
It's DDR2-6400
(system specs here - http://ktetch.blogspot.com/2010/09/updated-browser-benchmarks.html#recap)

I'm a computer kleptomaniac though, I keep browsers open, and open and open.  My server box has some tabs that ahve been open since september... 2008

The annoying thing is, despite autocli.sav being created, it doesn't seem to use it to resume.
tomaz
2010-10-08 05:59:51
2.489781 is the highest result in 480-499 which also makes it highest result in 0-999. It will void rechecks for sure. 

It seems that results with more particles in this run are considerably lower than in run with "normal" number of particles.  Beside that, zigzaging(noise) of results remains as before.
[TN]opyrt
2010-10-08 10:40:19
Just to make sure there are no misunderstandings regarding the savefiles for Muon.
autocli.sav: Savefile for the cli-version of Muon (Muon -c).  It is used in a single run after each checkpoint (when it says "Auto-saving" ).
queuecli.txt: Savefile for the cli-version of Muon.  It is used when a result requires rechecks until "runs" are equal to the "Rechecks for best-so-far results" setting in config.txt.

So let's say you're running a test that is 1500 Mpts long and has a high score so it goes into rechecking.
1: Up until the first check is done, it will write to autocli.sav each time it is autosaving.
2: When the first check is complete, it will write the result to queuecli.txt, adding "#runs=1" to the end of the result.  It will delete autocli.sav.

If you restart Muon here, it is supposed to start at 0.0, but on the second run.  autocli.sav is non-existing, but queuecli.txt is present in the muon folder.

3: The second test will start writing autocli.sav on every autosave.

If you stop Muon here (after an autosave), it will leave an autocli.sav and a queuecli.txt in the muon folder.  If you then restart, it should pick up in the middle of the second test.  The only way you can tell it's the second test is that you have a queuecli.txt with a results that has "runs=1" in its string.
[Edited by [TN]opyrt at 2010-10-08 10:40:55]
Cameron
2010-10-08 11:59:46
Opyrt what you are saying about muon interacting between the autosave file and the queue file is correct for the usual muon simulation.

Stephen has however asked for some volunteers to assist him as he performs a High Resolution Scan where he is changing the value of a single parameter across all it's possible values.

In this instance Stephen has created a number of queuefiles that have all the parameters for a single simulation run preconfigured and has asked us to tell muon to read from the queuefile these parameters rather than let muon randomly generate or try to optomise the design itself by changing many parameters which is muons usual behaviour.  We tell muon this by including the -q switch.  and Each sequence is concluded with TEST in the queuefile

The lattice file for this SECOND High Resolution Scan is NOT Linac900Ext10d2 which mean there is no existing History in our local Results.Dat file for muon to try and measure against for is this the best result so Far.

What is now happening in this second run is that if a result needs to be rechecked it has the number of runs appened.  runs=3 and if I stop now that it has completed 3 runs of a minimum 5 [in the First High Res Scan and for normal muon simulations I have 7 Rechecks] before it has created an autosave file and I later start Muon up again it will read from the Queue file it is 3 simulations through a 5 simulation run recheck.  OK Fine so far.

The Problem I and K'Tech are experiencing regarding Muons Usual interaction with the queue and autosave files is that with an average single run being on the plus side of 7500Mpts which is taking my machine 90 minutes to do and maybe 2 hours or more for K'Tech is if the simulation is somewhere in the middle of a run and creating it's restore points.

Although the restore point may be at time 1398 and 5303.5Mpts but if Muon is then restarted it reads from the queuefile either that it has completed n of 5 runs to perform a recheck or it reads in the next list of parameters to perform a simulation test on (which isn't removed from the queuefile until it is written to the results.txt file).  And it begins the simulation at time 0.0 .

So muon is ignoring the existence of the autosave file (if it exists) and then overwriting it with the new autosave which is most likely earlier in simulation time to the autosave it automatically ignored and overwrote.

This is not quite normal behaviour (at least as I understand) but we are following the instructions that Stephen has given us and so we are waiting for his guidance.
Cameron
2010-10-08 12:12:06
Oh yeah (sorry if you consider this double posting) but Stephen I'm Ending my Muon Session for the Day Now.

Simulation Time 286.61 1059.4Mpts

Results completed 5
Ceiling so far 2.455642%

autocli.sav now copied outside of muon folder (size 10.4 M
[DPC]white_panther
2010-10-08 13:40:09
0-29 RGtx
30-59 [dpc]white_panther
60-89 K`Tetch
90-119 [dpc]white_panther
120-149
150-179
180-209
210-239
240-269
270-299 Stephen Brooks
300-319 tomaz
320-339 tomaz
340-359
360-379 Boots
380-399
400-419
420-439 Cameron
440-459
460-479 tomaz
480-499 tomaz
500-549 tomaz
550-599
600-649
650-699
700-749
750-799
800-849
850-899
900-949
950-999 Hydrox

i'll take 90-119

@K`Tetch i'm using DDR2 1066 MHz ( PC2-8500 )
Stephen Brooks
2010-10-08 15:35:31
Done 270-299, taking 900-949.

Cameron said:
--[So Stephen is it alright to use Tomaz 490 result in order to cut-down possible re-checks?]--

Yes.

I think when the -q switch is on the client might ignore .sav files, simply because it's not that easy to tell if a given .sav file is from a result from the queue or whether it's some other partly-finished simulation.
Maniacken [US-Distributed]
2010-10-08 23:53:53
Maniacken 000-999. yes the whole range.  Think it would be good to have the range done on a single computer.  Also a second data point.

Using the 490 result to cut down on rechecks
Cameron
2010-10-09 03:28:22
Right I'll add the 490 result to the dat file.  I can cope better with with 90min run per result rather than than 7-8 hour run for a rechecked result that might be every second simulation.
Stephen Brooks
2010-10-11 16:37:17
Hmm, tried plotting the bits of the second scan that I've got back, but so far it's not a particularly informative picture.

Anyone know what happened to eveiga and KWSN grizzly who took the 100-299 range of the original scan?
RGtx
2010-10-11 19:02:55
0-29 Returned
120-149 Taken.
Stephen Brooks
2010-10-11 19:26:51
Errr... scratching my head a bit as to why there are still 30788 particles left over at 2000.00ns in one of my simulations.  Can someone manage a graphics screenshot near the end with energy colouring on?  (It may be a lot of particles that are too low-energy to matter).  You'll need to use queuegfx.txt and muon1 -q and then if you want autogfx.sav to work, drop the -q switch so it loads from the save file.
RGtx
2010-10-11 20:03:10
OK, I'll do a screenshot, probably around 22:15 tonight.
Boots
2010-10-12 09:38:45
Sent 360 379
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io RSS feed

Site has had 25166972 accesses.