2016-Aug-21 Comp. Phys. Biweekly Report
Revision as of 14:10, 9 February 2017 by Anichols (Created page with "He Zhang <ul> <li>LEM Upgrade: successfully implemented four different optimizer (nsga_II, spea2, sms_meoa, and nspso, provided by PyGMO) on the trip rate and heat load optimi...")
- LEM Upgrade: successfully implemented four different optimizer (nsga_II, spea2, sms_meoa, and nspso, provided by PyGMO) on the trip rate and heat load optimization problem. Performed efficiency studies between the various optimizers. nsga_II seems to be the most optimal one.
- IMP data analysis: A python script file that automatically separate the data bunch by bunch has been finished. The program also calculates the rms bunch size. Comparing the rms bunch size of the cooled ion bunch with that of the uncooled ion bunch, we can clearly show the cooling effect and get an idea of the cooling efficiency. We compared our result with the result from Zhao, the graduate student at IMP. An error in Zhao’s formula is found. After fixing the error, the results agree with each other. I’m working on the bi-Gaussian fitting for the data, preliminary results agree with Zhao’s result.
- Cooling simulation code: River has put a great effort trying to insert our FMM kernel code to Rio’s parallel FMM frame. But he could not work it out. We asked Jie Chen to help him. After reading through Rio’s code in a few days, Jie concluded that it is difficult to integrate our kernels into the framework. The major obstacle is the data structure in the framework is far different from the data structure and the lack of documentation about the framework code. Eventually we can figure it out, but it is difficult to finished in a short time with limited manpower. In such a case, I decided to concentrate on tracking a few ions inside an electron sea. Friction force can be calculated according to the momentum change of the ions, which can be used to benchmark the theoretical formulas. Once it is done, we will go back to the parallel FMM algorithm. By now, a parallel RK4 integrator has been finished. We will also parallelize the symplectic Hermite integrator that River developed before. I want to compare the two integrators. Because we are not tracking a repeated system, I wonder whether a non-symplectic RK4 integrator is good enough for this single pass tracking. The non-symplectic integrator is much easier to implemented and more efficient.
He Zhang & River Huang
- Advanced Accelerator Concept Workshop. River and I attended the AAC2016 workshop. I gave a talk on the FMM using traceless totally symmetric Cartesian tensor in group B on computation
- Worked with summer student Gabriel Nowak on the one way speed of light experiment by analysis the long term stability in the ARC1 arc.
- Set up Balsa's students on the GPU cluster. They are installing GHOST for preliminary testing and debugging.
- Spend time consulting with G. Wei for setting him up on the qcd supercomputer for studies of ground motion using parallel elegant.
- Attended a meeting with the computer center people regarding the needs of the accelerator division for computational resources.
- started optimizing the 5 up/down setup for the cebaf er in collaboration with pierre korysko (bnl)
- reworked the 5 pass vertical extraction for the 3 halls at 5 pass config (needed next spring)
next two weeks:
- redo horizontal 5 pass extraction with new setup
- learn optics for the lerf dark light and take shifts.