|
HPS-MC
|
|
|
HPS-MC
|
|
This example uses pede_job.py to optimize the alignment parameters. The inputs are recursively searched for *.bin files that we assume were prepared by hps-java during the tracking step. These data files are then used to run the minimization scheme within pede. Finally, the results form pede are applied to the detector description file (compact.xml), generating a new iteration of the detector.
There are several special parameters that are separated depending on if they are required and how often they change between jobs.
These parameters are required but do not change often so they are put into a config file (e.g. .hpsmc in this directory) rather than the job JSON.
| param | description |
|---|---|
| param_map | map of SVT sensors listing names, ID numbers, and other location details |
| pede_minimization | pede minimization settings to append to its steering file |
| java_dir | root directory of hps-java for applying pede results |
These parameters are required and change often depending on the iteration and what alignment you are studying, so they are in the job JSON.
| param | description |
|---|---|
| input_files | text file listng bin files or the list of bin files themselves |
| output_files | mapping the pede output files to their destination names |
| to_float | list of pede parameters to allow to float during minimization (see below) |
| detector | name of detector that tracking was run on |
These parameters are optional and only some of them are tested. They are provided in the job JSON as well.
| param | description |
|---|---|
| subito | if true, use the -s pede flag which is a rougher minimization focusing on speed |
| constraint_file | path to external constraint file pede should load before running minimizer |
| previous_fit | path to previous results (res) file to start fit from |
| beamspot_constraints | if true, apply the beamspot constraints |
| survey_constraints | if true, apply the survey constraints |
| res_file | name of millepede.res file (if not "millepede.res", only helpful if separate) |
| bump | if false, do not update the iteration number of detector when applying results |
| force | if true, ignore any existing detector when creating a new directory |
| next_detector | if set, use this name for the new iteration of the detector (with parameters applied) |
| no_copy | don't use the destination file as inputs to pede, use the source |
The no\_copy parameter is helpful when you have a local set of bin files that you want to run over. Often times, the pede program tries to run over the input files before they are done being loaded into the kernel cache and so you see a "Open error" (example below). This parameter can avoid this error by just having pede look at the original source file rather than the copy in the scratch directory.
The to_float job parameter is more complicated than the others, so it deserves its own section. It is a list of different so-called Pattern each of which can be an ID number of a composition of several different boolean operations in order to construct a group of paramters to float. If a single alignment parameter "matches" any of the Patterns in the list, then it will be floated.
More detail about the Pattern syntax is given in its documentation: hpsmc.alignment._pattern.Pattern
Within each pattern, different operations are separated by & to reflect that all of them must be true for a Parameter to match that Pattern.
"direction=u & operation=t" would match Parameters that translate along the local u axis"tu""top & front & tu" would match Parameters that are individual sensors in front (first 4 layers), top half of the detector and translate along the local u axisYou can also select parameters using logical or by separating Patterns into different elements of the list.
["top & front & tu", "top & front & rw"] would match Parameters that are individual sensors in the front, top half of the detector and either translate along u or rotate around w."rw" is an alias for "direction=w & operation=r" 
1.9.8