- Table of contents
- 10/08/2013 Notes
Craig D., Zukai, Jan, Matthew, Brian, Alec and Martin (taking notes)
- Everyone gave a nice update linked from the event above.
- Based on my update, it seems that we need to set up our processing as following:
- read in one event
- process one event
- read in the next event
- Perhaps there is a configuration parameter for the read queue to give us this behavior. I will look into this.
Time Lag Issue¶
- Jan gives a verbal update on the time lag issue.
- Jan noticed that we started writing out empty events when a process' memory usage became larger than 15%.
- We then did some time print outs and saw that this corresponds to ~25 s. This means that the BNEVB data buffer is around 25 s deep and once we exceed that in processing, we cannot retrive that data anymore.
- We need a way to skip events when we start falling behind.
- Andrew notes that this should be handled in Ron's memory reader.
Plots for CHEP Talk¶
- Andrew needs plots for his talk in Amsterdam.
- Brian ran over the raw 5 ms data.
- He noticed that things can crash when there is an empty HitList.
- He committed a bit of code to deal with this gracefully.
- Matthew asked if we can resolve the associations issue that cropped up last week due to the TriggerDecision instance label.
- We decided that we will remove the instance label from the TriggerDecisions. I will take care of this.
- Zukai would like to have the CompareDAQHit<TDC> function sort by TDC, then plane and finally cell if the TDC values are the same.
- We decided to do this.