General:
- Updated BLonD website http://blond.web.cern.ch/
Kostis update
- PSB tests on approximate computing (double precision) and single precision w/o approximate computing
- Precision good
- Speed-up less good than in other cases, but up to factor 2
- Pull request: FFTW should be built for single and double precision
BLonD common: RF objects
- Pseudocode of usage for RFSystem and RFStation -- how should these be split?
- "combine" system (voltage and phase) -- rename to "collect"?
- for same harmonics: proper vector sum cannot be done at this stage, only interpolation in time
- for different harmonics: returns various systems
- RFStation could have vectorSum=True/False option
- for vector sum, need cycle_time from Ring
- for tracking, should we build in some intelligence? e.g. sum voltages if same harmonic and no cavity controller. NO: user controlled better
- RFSystem: should be input, not allowing several arrays of voltages etc.
- can use direct_input for data_types handling to input several harmonics
- For tracking, need some default, automatic behaviour e.g. to sum different voltages if no cavity controller present (Assembler?)
- Next steps: continue developing these objects with the above logic, work out the pseudocode
- data_types should be checked
- make input handling like in Ring
- Next meeting in 3 weeks
There are minutes attached to this event.
Show them.