Psana

  • New analysis package xtcav with code written by Alvaro Sanchez-Gonzalez with some help from Christopher O'Grady.  Some initial documentation can be found Here.
  • PSShmemInput: Reduce queue depth in DgramSourceShmem from 32 to 3.  This is because shmem applications want minimal latency, which requires minimal buffering (at the expense of potentially more dropped events).
  • Add support for Epix100a
    • psana_examples: DumpEpix.cpp - add support for new types Psana::Epix::Config100aV1 and Psana::Epix::ElementV2 (see PSAS-7)
    • ImgAlgos: EpixNDArrProducer - add support for Epix::ElementV2 and Epix::Config100aV1 (see PSAS-7)
    • CalibManager
      • ConfigParametersForApp, ConfigFileGenerator add support for Epix100a (see PSAS-6, PSAS-7)

      • use calibtypes from config pars in stead of evaluation from source (PSAS-6)

      • ConfigParametersForApp - support for Epix100a: add 'geometry', update default list of sources (PSAS-7, PSAS-11)
    • PSCalib
      • SegGeometryCspad2x1V1.py - move common methods to SegGeometry.py, improve self-documentation
      • add SegGeometryEpix100V1(.py, .cpp, .h)
      • in SegGeometryStore(.py, .h) add SegGeometryEpix100V1
      • test/ex_seg_geometry_store.cpp - updated for epix100
      • in CalibParsStore.h, GenericCalibPars.cpp - add support for Epix100a
    • pdscalibdata: add Epix100aBaseV1.h

External Packages

  •  

Bugs fixed

  • pypsalg: Throttle warning messages in find_blobs.py. Fix missing imports and add 2D histogram in Histogram.py.
  • psmon: Fix from Dan that prevents psplot from "falling behind" when ZMQ buffers on the client (display side) fill up.
  • PSXtcInput: Fix a bug in the indexing access to calibcycle times: multiple streams' calibcycles were being included in the list. To avoid duplicates, reduced it to just one stream's calibcycles (since they are all identical).

Known Issues

  • Python applications using h5py with parallel HDF5 would hang when exiting. Apparently, there is a deadlock caused by either mpi4py or h5py implementations. We're looking at this problem.

  • No labels