 |
| tomaz 2008-02-29 10:51:53 | Like back in Phaseroteb6, I started new computation with few highest Mpts results as a seeds. It seems good so far but computations are rather long. Is anybody willing to put some of thise high Mpts results in his own result.dat file or maybe make new file only from thise? It should overtake mainstream design because mainstream is close to local maximum I think...(look at eb6 graph what happened) |
| [TA]JonB 2008-03-01 19:06:55 | I'd be willing on one of my home PCs (dual core Opteron 165). How do you propose to transfer the results? |
| tomaz 2008-03-03 08:15:58 | Nice. My results.dat file can be downloaded from here: www.dhd.si/muon/results.dat Stephen's Mpts/muon% graphs are also clickable and a single results can be obtained from them, too. Good luck!  |
| tomaz 2008-03-10 08:40:51 | Hey, why nobody trust me ?  There were only two downloads of my results.dat. Anyway in a week I got to 4.849 % at 1250 Mpts/run. It is realy simple logic-high Mpts designs have higher potencial for optimisation and yield higher % at last. So results with high Mpts should be treated as perspective ones and put in samplefiles... My new results.dat can be downloaded from above address. |
| [TA]JonB 2008-03-10 18:45:15 | Core two of my Opteron is working on just your results.dat file. Also limited it to that specific optimisation (phaserotlinac1). I haven't checked it for output yet. It has been several days already and I'm not sure that it has dumped yet. Time will tell! |
| tomaz 2008-03-18 11:52:35 | It gave quite a kick to opimization, did't it ? Stephen, which part of design is responsible for such good results ? (we are on 6% muon now) |
| tomaz 2008-03-21 08:27:09 | I got a 6.122 % result. It has 241 linaccells, 1446 Mpts/run. I noticed before that linaccells have great influence on Mpts (higherl linaccells higher Mpts) and also on results (lower linaccells tends to give higher %), so I tested same design with linaccels=191 and 000. I got 6.133% (1423 Mpts/run) in first case and 6.174% (1340 Mpts/run). All three were queued (5 runs). But on the other hand I think things are not so simple-I think designs with linaccells are better and that higher results with low or zero linaccells are just local extrems. Few week ago I started computation from only few desigs I picked from graph. They all have high Mpts and linaccels. Mainstream design stucked at 4.6% (800 Mpts) and has linaccells=000. I think design of other parts of accelerator adopted to higher linaccells and that's why there have been such a good development in a short time. |
| Stephen Brooks 2008-03-24 12:52:45 | What about phaserotcells? I started this optimisation off expecting phaserotcells to have a nonzero value but last time I looked it was still on 000! |
| tomaz 2008-03-24 19:32:21 | It must be some time ago Stephen. Lately it is around 999 (high Mpts branch). |
| Stephen Brooks 2008-03-25 10:46:44 | The current top result for Linac88MHz900MeV1:...the cell-length parameters being: |
