ARCA absolute pointing taskforce
https://cern.zoom.us/j/62833072398?pwd=YTZFSXpmaXluYzNGTHBtY2crdnM0Zz09
JG presentation of the production of the tripod files as input for the fit (see slides).
After a suggestion from Lilian, JG will change the nomenclature of the produced tripod files to be snakemake compatible.
First idea would be something like: tripod_nominalX_nominalY_nominalZ_dX_dY_dZ.txt
Marion's presentation of the fit procedure (see slides).
One open question is to check if we can remove the 3c step. Results already produced can answer that; we just need to check the comparison between 3a and 3c.
Open discussion of the efficiency; it seems like we are limited to ~50% because of the way Jpp works, the gradient estimation being a single-core serial procedure. Out of our scope to develop another version of that; this is an open question for later on.
Lilian opened a discussion to run a profiler on the fit procedure.
Do we need to rerun the acoustic event if we consider perturbation? As 1m ~ 1.5 ms, the answer is mostly no as long as the window used is large enough.
Overall, the results shown in Marion's slides show that we are already at the computing cost we wanted to reach with the tested options; hence the next steps will be:
- Confirm if we can remove the 3c step.
- Confirm that the combination of starting from optimized positions, using 1/2 data, and 3c step only is providing consistent results in a bunch (~10) trial runs and re-confirm the computing time.
If this is confirmed, we have a defined setup for the runs.
.