stephenbrooks.orgForumMuon1GeneralRechecks not done when necessary
Username: Password:
Search site:
Subscribe to thread via RSS
[DPC] Mr. Aldi
2010-11-05 16:59:21
Now I ran Tomaz best result so far for 6tc2 while taking decay part from best 1900 Mpts result.  After 4 checks we see this in queuecli:

d1l=517;d2l=107;d3l=766;d4l=878;d5l=027;d6l=000;db1l=411;db2l=079;db3l=137;db4l=695;db5l=201;db6l=996;decaycells=018;ld1=000;ld2=571;ld3=999;ld4=999;ld5=002;ld6=892;ld7=000;linaccells=999;ls1f=000;ls1l=800;ls2f=996;ls2l=999;ls3f=947;ls3l=996;ls4f=990;ls4l=621;ls5f=980;ls5l=476;ls6f=000;ls6l=841;rf1p=996;rf1v=992;rf2p=004;rf2v=575;rf3p=804;rf3v=877;rf4p=918;rf4v=986;rf5p=661;rf5v=820;rf6p=966;rf6v=670;s1f=999;s1l=901;s2f=000;s2l=999;s2r=278;s3f=709;s3l=219;s3r=004;s4f=996;s4l=000;s4r=351;s5f=992;s5l=002;s5r=999;s6f=999;s6l=988;s6r=000;sb1f=031;sb1l=886;sb1r=000;sb2f=113;sb2l=218;sb2r=992;sb3f=585;sb3l=004;sb3r=999;sb4f=997;sb4l=488;sb4r=855;sb5f=030;sb5l=921;sb5r=000;sb6f=814;sb6l=008;sb6r=252;tantalumrodz=145;tantalumrodr=000;#queued=1;#runs=4;
2.555231,2.614857,2.525143,2.598351 (10505.7 Mpts) [v4.44d] <Linac900Ext6tc2> {21629C09D55753344F5C4C6E}

You see the first result?  2.555!  The final result was 2.585339. Say muon1 finds a new configuration with a yield of 2.57xxx using 'seed 0'. Now the final result after rechecks could be 2.60xxx, but no rechecks will be done as 'the best rechecked score for this linac score is 2.583339'!

Muon1 should keep the scores of the different seeds itself, instead of taking the average of this result and save this to the results.dat.

Only then Muon1 would find 2.57xxx with seed 0 worthy to recheck.

Also think that this can be the other way around.  Sometimes the average after rechecks is always lower, thus giving constant rechecks when muon1 finds a inital high yield with seed 0. There is a topic here somewhere with a complaint about the constant rechecks

I guess it is needed to edit the way results are accepted by the server for this.
Stephen Brooks
2010-11-05 23:24:41
The main reason for rechecks is if there is a computer error and the first result comes up far too high.  Then four more simulations are done and the highest and lowest thrown out.  Increasing the accuracy of high results is only a secondary objective, and as you've noticed, it's a fairly crude way of doing it.
[DPC] Mr. Aldi
2010-11-06 00:45:37
Thanks for the quick answer

The results are quite inaccurate sometimes then.  I now have an example of what I meant in my TS.

d1l=517;d2l=107;d3l=766;d4l=878;d5l=027;d6l=000;db1l=411;db2l=079;db3l=137;db4l=695;db5l=201;db6l=996;decaycells=018;ld1=000;ld2=571;ld3=999;ld4=999;ld5=002;ld6=892;ld7=000;linaccells=999;ls1f=000;ls1l=800;ls2f=996;ls2l=999;ls3f=947;ls3l=996;ls4f=993;ls4l=621;ls5f=980;ls5l=476;ls6f=000;ls6l=841;rf1p=996;rf1v=992;rf2p=004;rf2v=575;rf3p=804;rf3v=877;rf4p=918;rf4v=986;rf5p=661;rf5v=820;rf6p=966;rf6v=670;s1f=999;s1l=901;s2f=000;s2l=999;s2r=278;s3f=709;s3l=219;s3r=004;s4f=996;s4l=000;s4r=351;s5f=992;s5l=002;s5r=999;s6f=999;s6l=988;s6r=000;sb1f=031;sb1l=886;sb1r=000;sb2f=113;sb2l=218;sb2r=992;sb3f=585;sb3l=004;sb3r=999;sb4f=997;sb4l=488;sb4r=855;sb5f=030;sb5l=921;sb5r=000;sb6f=814;sb6l=008;sb6r=252;tantalumrodz=145;tantalumrodr=000;#queued=1;#runs=4;
2.563849,2.620140,2.528810,2.623759 (10517.5 Mpts) [v4.44d] <Linac900Ext6tc2> {D35308F768398939D58220CF}

After the final check I got 2.600148 (13138.2 Mpts) and beating Tomaz

When you look at the result with seed 0 (2.563849) it would not get rechecked as Tomaz result was in my samples-file.  As the seed 0 result was higher than Tomaz' result: I renamed the results.dat and deleted the sample-file.  Now it got rechecked to finally give 2.600148. It now also gets into the sample-file to be used by other clients (it would not have otherwise!!)

So I think it is still necessary to save the individual results from the different seeds so that my best result now automagically would have been rechecked

Btw, I found this candidate with muon1 @ -/10.
Cameron
2010-11-06 04:18:56
Are the highest and lowest still discarded Stephen.  I thought you decided (a long time ago) to go with the [sum of total runs]/runs.
Stephen Brooks
2010-11-07 00:31:48
No, I changed it in the other direction!  And that was in 2005 or something.
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io

Site has had accesses.