SymphonyX 2004-06-11 21:42:01 | Nice result Pollock!! |
Pollock[Romulus2] 2004-06-11 23:18:17 | Thanks Actually the 1.47 result was sent by mistake. I had a perfectly matched pair at 1.44 that were sent in under my name and my team member Herb's. The one sent under Herb[Romulus2] never appeared. It may have been lost with the rest of his results which have recently been lost. They were intended to seed the sample file with two different breeds. Unfortunately I don't think the sample file could handle them. They are both expanded and hopefully more accurate versions of the recent top breed. The current and even more refined versions of them are currently running: 1.589764 (2385.3 Mpts) 1.582106 (491.0 Mpts) I would send them, but it may be better to wait until the sample file can handle them. |
Herb[Romulus2] 2004-06-11 23:38:08 | making again little surprises Joe, aren't ya? Nice tweaking, congrats You know I'm doing it from the statisical side and it's much different in behaviour. The solenoids seem to be maxed out now, the Phaserot has a lot more room to perform, but the Chicane stuff I can't figure out, my math doesn't work on that parameters. All my best values I had from the standard clients gen selection. My own seeds always stop below 0.2%, duh. |
Pollock[Romulus2] 2004-06-12 00:31:34 | quote: I have'nt been working on the chicane lately. Just a few trials to see what would work. Max at this point is 1.119010, but it will not grow as fast as phaserot. The solenoid channel will improve, but in a different direction. The accelerator is a different challenge. There seems to be two different parts. The ls**l=000;ls**f=000; and ld=000; parameters are a standard solenoid channel. The rf**p=000;rf**v=000; parameters are the electrical field for the accelerator. Those two work together. If you make a big adjustment in one of them, the other must be adjusted to balance the electrical field. The linaccells=000 is the Parameter which controls the overall length of the accelerator channel. The key that I have found to tweaking them is simple. If one part (or one group) of parameters is changed, then usually another parameter (or group) needs to be changed with it. I have got a good handle on the tantalumrodr=000; and phaserotcells=000; in phaserot. The one that I need to understand is tantalumrodz=000; at this point. I think 'z' is related to s1l=000; but not sure how. The phaserot run at 1.58 is nearly finished with the current tweaks, just need to get it more accurate. That is being done now. I still have a few more unusual ideas that I want to try, before I forget them. It still needs a lot of minor adjustment, though. Both breeds in phaserot will climb higher. |
Pollock[Romulus2] 2004-06-13 13:30:33 | Everything reasonable that I try seems to work in phaserot. It's getting scary. Current recheck: 1.683614,1.670058 The 'z' 'r' and s1l experiment is working better than I expected, it could be interesting. |
kitsura 2004-06-13 19:58:03 | ChicaneLinac...i can't figure it out too I have about 10 clients running it 24/7 and all are stuck at -1.7x |
Maniacken [US-Distributed] 2004-06-14 10:44:18 | nice job pollock you almost got the baseline beat. |
Pollock[Romulus2] 2004-06-14 17:43:09 | Phaserot will beat the baseline. It has some growth left in it. I am off to play with the chicane, so most likely it will be somebody else who finds the leader. I do still have a few ideas, though. |
SymphonyX 2004-06-15 13:03:00 | "The computed design is 0.0483% better than the baseline one" |
[TA]z 2004-06-15 14:16:08 | nice, congrats |
Nexus[Free-DC.org] 2004-06-15 16:49:42 | Good work work on PhaseotB, guys. I wouldn't get too dpressed about SolenoidsTo15cm just yet. Remember when it got stuck around 9.9% for ages, and then suddenly shot up again? Who knows, maybe it'll happen again. |
AySz88 2004-06-15 16:54:32 | I think SolenoidsTo15cm is going to be killed off soon - Stephen said something about ending it if it leveled off for several weeks. |
Herb[Romulus2] 2004-06-16 01:09:37 | I wouldn't do so in the moment. We have currently 3 out of the top 5 values representing relatively wide variations. Not exaxtly own breeds, but close. That needs to be settled a bit more. |
Stephen Brooks 2004-06-16 03:55:10 | I reckon I might leave SolenoidsTo15cm around for a while - the behaviour of it "sticking" and then moving off means that it's worth keeping it since I'll be improving the optimiser in August and the new methods might squeeze more out of it again. |