General new:
- Helga: docs require updates, example will be shown next time
- Markus: was working on unittests for LLRF
- Other unittests still missing
- Kostis: np.deepcopy() fails in the bunch generation in python3.6 but not in python3.5 -> dreadlock cannot be copied
- Change the code?
- Do to miniconda vs anaconda?
- Kostis: check packaging with Jelena
- Installation guide: just write requirements
Markus: Correctly simulating with the SPS impedance model
- Use fast sine or interp_kick() with the impedance model?
- Summary of objects inside the impedance model, minimum BW required
- Run times for single bunches...
- Alex: extending Juan's test for induced voltage convergence? He had a test for a single resonator
- Comparison of interpolated/continuous kick: squared difference of time coordinates
- Run time is 30-50 % larger for continuous kick
Kostis: Distributed BLonD
- Master-worker model
- Initialisations on Master process, create workers
- Main loop iterations are done with all workers; master broadcasts
- Only exchange bunch profile information, but coordinates are distributed over different workers
- Minimise exchange of information
- Workers don't have to communicate with each other but just with the Master when the histogram is reduced
- First tests with EX_01_Acceleration.py
- CERN cluster for distributed applications -> BE dept.
- First results OMP vs MPI on 1 or 2 nodes
- Next steps: minimise the amount of communication; workflow is fixed -> predefine sequence
Simon: Frequency offset
- For PSB and LEIR injection with a frequency offset w.r.t. design
There are minutes attached to this event.
Show them.