stephenbrooks.orgForumMuon1GeneralWhat does DPAD do on encountering h/w related errors?
Username: Password:
Search site:
Subscribe to thread via RSS
[TA]Assimilator1
2009-02-06 20:04:54
I'm referring to either failing hardware or an unstable overclock.

I know F@H ends WUs early on finding errors, IIRC SETI reports 'compute error' after the WU is validated but with DPAD I have no idea.

What does it do?
Stephen Brooks
2009-02-24 09:28:56
If it's a rare single bit error, it will only affect one particle in the beam and produce a very small yield error.  The most dangerous error condition would be one where it produced a false high yield, but these are rerun with the quarantine mechanism and the highest and lowest of 5 runs discarded.
RGtx
2009-02-25 15:47:10
Rerunning the current highest scoring genome of Linac900Removable12, I cannot reproduce its reputed 3.669259% yield

d1l=000;d2l=000;d3l=998;d4l=998;d5l=801;db10l=013;db11l=000;db12l=000;db1l=022;db2l=022;db3l=000;db4l=000;db5l=089;db6l=089;db7l=000;db8l=000;db9l=013;decaycells=018;ld1=000;ld10=000;ld11=102;ld12=102;ld2=000;ld3=000;ld4=000;ld5=000;ld6=000;ld7=002;ld8=002;ld9=000;linaccells=999;ls10f=990;ls10l=985;ls10r=000;ls11f=000;ls11l=999;ls11r=000;ls12f=000;ls12l=999;ls12r=000;ls1f=000;ls1l=999;ls1r=000;ls2f=000;ls2l=999;ls2r=000;ls3f=000;ls3l=999;ls3r=000;ls4f=000;ls4l=999;ls4r=000;ls5f=006;ls5l=964;ls5r=000;ls6f=006;ls6l=964;ls6r=000;ls7f=000;ls7l=991;ls7r=000;ls8f=000;ls8l=991;ls8r=000;ls9f=990;ls9l=985;ls9r=000;pd1=999;pd10=999;pd11=999;pd12=999;pd2=999;pd3=999;pd4=999;pd5=999;pd6=999;pd7=999;pd8=999;pd9=999;phaserotcells=000;prf10p=446;prf10v=218;prf11p=870;prf11v=215;prf12p=870;prf12v=215;prf1p=433;prf1v=000;prf2p=433;prf2v=000;prf3p=439;prf3v=226;prf4p=439;prf4v=226;prf5p=440;prf5v=227;prf6p=440;prf6v=227;prf7p=000;prf7v=000;prf8p=000;prf8v=000;prf9p=446;prf9v=218;ps10f=276;ps10l=056;ps11f=286;ps11l=000;ps12f=286;ps12l=000;ps1f=000;ps1l=021;ps2f=000;ps2l=021;ps3f=000;ps3l=021;ps4f=000;ps4l=021;ps5f=015;ps5l=000;ps6f=015;ps6l=000;ps7f=278;ps7l=000;ps8f=278;ps8l=000;ps9f=276;ps9l=056;rf10p=413;rf10v=989;rf11p=000;rf11v=998;rf12p=000;rf12v=998;rf1p=384;rf1v=994;rf2p=384;rf2v=994;rf3p=406;rf3v=999;rf4p=406;rf4v=999;rf5p=341;rf5v=968;rf6p=341;rf6v=968;rf7p=490;rf7v=987;rf8p=490;rf8v=987;rf9p=413;rf9v=989;s1f=999;s1l=999;s2f=999;s2l=977;s2r=573;s3f=985;s3l=999;s3r=406;s4f=999;s4l=290;s4r=999;sb10f=000;sb10l=999;sb10r=999;sb11f=000;sb11l=999;sb11r=819;sb12f=000;sb12l=999;sb12r=819;sb1f=002;sb1l=999;sb1r=999;sb2f=002;sb2l=999;sb2r=999;sb3f=007;sb3l=999;sb3r=806;sb4f=007;sb4l=999;sb4r=806;sb5f=000;sb5l=959;sb5r=780;sb6f=000;sb6l=959;sb6r=780;sb7f=000;sb7l=969;sb7r=999;sb8f=000;sb8l=969;sb8r=999;sb9f=000;sb9l=999;sb9r=999;tantalumrodr=000;tantalumrodz=448;#runs=5;
3.667213 (8393.4 Mpts) [v4.44d] <Linac900Removable12> {EE11CACBFAE5EE35E1E2B332}


[TA]Assimilator1
2009-02-27 00:03:35
Thanks stephen, what would happen if a machine had failing RAM with errors all over the place?, would the '5' runs average out the huge variations or would something else happen?

I can see that the '5 run' check would cut out the odd error as produced by a slightly unstable overclock but I don't see how it would help where a string of errors occureed.....

RGtx
How is that related to my question?
Stephen Brooks
2009-02-27 11:34:01
RGTx, what "3.669" genome are you talking about in the ...12 optimisation?  The highest on the score board right now is amd.borg's 3.667558.

Assimilator1, yes, if the computer is really messed up it will produce wrong results.  However, in practice I've never seen any result submitted that is wrong by more than a part in 1000. Note that there is no exactly "right" answer, if people voluntarily increase their recheck number, statistics will produce similar small changes in the yield.  However, with the same number of rechecks and a fixed random seed (as in recent lattices), the calculation should be identical.
RGtx
2009-02-27 17:52:47
What "3.669" genome am I talking about?  Why this one (from the 100-result sample file):

#gen=11;#runs=5;d1l=000;d2l=000;d3l=998;d4l=998;d5l=801;db10l=013;db11l=000;db12l=000;db1l=022;db2l=022;db3l=000;db4l=000;db5l=089;db6l=089;db7l=000;db8l=000;db9l=013;decaycells=018;ld1=000;ld10=000;ld11=102;ld12=102;ld2=000;ld3=000;ld4=000;ld5=000;ld6=000;ld7=002;ld8=002;ld9=000;linaccells=999;ls10f=990;ls10l=985;ls10r=000;ls11f=000;ls11l=999;ls11r=000;ls12f=000;ls12l=999;ls12r=000;ls1f=000;ls1l=999;ls1r=000;ls2f=000;ls2l=999;ls2r=000;ls3f=000;ls3l=999;ls3r=000;ls4f=000;ls4l=999;ls4r=000;ls5f=006;ls5l=964;ls5r=000;ls6f=006;ls6l=964;ls6r=000;ls7f=000;ls7l=991;ls7r=000;ls8f=000;ls8l=991;ls8r=000;ls9f=990;ls9l=985;ls9r=000;pd1=999;pd10=999;pd11=999;pd12=999;pd2=999;pd3=999;pd4=999;pd5=999;pd6=999;pd7=999;pd8=999;pd9=999;phaserotcells=000;prf10p=446;prf10v=218;prf11p=870;prf11v=215;prf12p=870;prf12v=215;prf1p=433;prf1v=000;prf2p=433;prf2v=000;prf3p=439;prf3v=226;prf4p=439;prf4v=226;prf5p=440;prf5v=227;prf6p=440;prf6v=227;prf7p=000;prf7v=000;prf8p=000;prf8v=000;prf9p=446;prf9v=218;ps10f=276;ps10l=056;ps11f=286;ps11l=000;ps12f=286;ps12l=000;ps1f=000;ps1l=021;ps2f=000;ps2l=021;ps3f=000;ps3l=021;ps4f=000;ps4l=021;ps5f=015;ps5l=000;ps6f=015;ps6l=000;ps7f=278;ps7l=000;ps8f=278;ps8l=000;ps9f=276;ps9l=056;rf10p=413;rf10v=989;rf11p=000;rf11v=998;rf12p=000;rf12v=998;rf1p=384;rf1v=994;rf2p=384;rf2v=994;rf3p=406;rf3v=999;rf4p=406;rf4v=999;rf5p=341;rf5v=968;rf6p=341;rf6v=968;rf7p=490;rf7v=987;rf8p=490;rf8v=987;rf9p=413;rf9v=989;s1f=999;s1l=999;s2f=999;s2l=977;s2r=573;s3f=985;s3l=999;s3r=406;s4f=999;s4l=290;s4r=999;sb10f=000;sb10l=999;sb10r=999;sb11f=000;sb11l=999;sb11r=819;sb12f=000;sb12l=999;sb12r=819;sb1f=002;sb1l=999;sb1r=999;sb2f=002;sb2l=999;sb2r=999;sb3f=007;sb3l=999;sb3r=806;sb4f=007;sb4l=999;sb4r=806;sb5f=000;sb5l=959;sb5r=780;sb6f=000;sb6l=959;sb6r=780;sb7f=000;sb7l=969;sb7r=999;sb8f=000;sb8l=969;sb8r=999;sb9f=000;sb9l=999;sb9r=999;tantalumrodr=000;tantalumrodz=448;
3.669259 (8393.4 Mpts) [v4.44d] <Linac900Removable12> #time=91394; by [TN]steinrar


Perhaps you ought to look on this page, http://stephenbrooks.org/muon1/
[TA]Assimilator1
2009-02-28 10:08:19
What do you mean by 'fixed random seed'?
RGtx
2009-02-28 13:19:06
RGtx
How is that related to my question?

It was related to Stephen's answer, an example of a false high yield.


Stephen Brooks
2009-03-03 14:39:04
My browser was returning me a cached version of the page from two weeks ago - now I see the result!  I also see a lot of results close to it in yield, for instance 3.669225 - some of those should be correct.  If the highest one is causing problems I can remove it, though I think in fact these optimisations have stopped for their own reasons.
[TA]Assimilator1
2009-04-10 10:07:01
RGtx
Ah ok, fair enough.

Stephen
>>>However, in practice I've never seen any result submitted that is wrong by more than a part in 1000<<<

How do you know that?  do you manually check the highest results or do you have some sort of automatic error checking for high results?
[TA]Assimilator1
2010-01-16 18:32:13
Err, Stephen? 
Stephen Brooks
2010-01-17 00:04:09
I manually check the highest results when I need them for a paper.
[TA]Assimilator1
2010-01-17 13:13:59
Thanks for the replie , that sounds like a time consuming job, unless you just check the top few.
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io

Site has had 17129161 accesses.