Opticks Testing, Geocache Creation, Python setup

Testing Installation with opticks-t

The opticks-t function runs ctests for all the opticks projects:

simon:opticks blyth$ opticks-
simon:opticks blyth$ opticks-t
Test project /usr/local/opticks/build
      Start  1: SysRapTest.SEnvTest
 1/65 Test  #1: SysRapTest.SEnvTest ........................   Passed    0.00 sec
      Start  2: SysRapTest.SSysTest
 2/65 Test  #2: SysRapTest.SSysTest ........................   Passed    0.00 sec
      Start  3: SysRapTest.SDigestTest
 3/65 Test  #3: SysRapTest.SDigestTest .....................   Passed    0.00 sec
      Start 59: cfg4Test.CPropLibTest
59/65 Test #59: cfg4Test.CPropLibTest ......................   Passed    0.05 sec
      Start 60: cfg4Test.CTestDetectorTest
60/65 Test #60: cfg4Test.CTestDetectorTest .................   Passed    0.04 sec
      Start 61: cfg4Test.CGDMLDetectorTest
61/65 Test #61: cfg4Test.CGDMLDetectorTest .................   Passed    0.45 sec
      Start 62: cfg4Test.CG4Test
62/65 Test #62: cfg4Test.CG4Test ...........................   Passed    5.06 sec
      Start 63: cfg4Test.G4MaterialTest
63/65 Test #63: cfg4Test.G4MaterialTest ....................   Passed    0.02 sec
      Start 64: cfg4Test.G4StringTest
64/65 Test #64: cfg4Test.G4StringTest ......................   Passed    0.02 sec
      Start 65: cfg4Test.G4BoxTest
65/65 Test #65: cfg4Test.G4BoxTest .........................   Passed    0.02 sec

100% tests passed, 0 tests failed out of 65

Total Test time (real) =  59.89 sec
opticks-ctest : use -V to show output

How to proceed when tests fail

See the instructions for running tests individually and running them in the gdb debugger at debugging_failing_tests.

Creating a geocache geometry directory with geocache-create

Many of the tests will fail in the absence of a geocache. Geometries are created using the geocache- bash functions. To create a geometry use:

geocache-              # run precursor function which defines the others
type geocache-create   # take a look at geocache-create
geocache-create        # create geocache from GDML geometry file, default geometry is Dayabay near site detector

The bash functions essentially boil down to the below commandline with additional command line options for metadata recording:

OKX4Test --deletegeocache --gdmlpath /path/to/your/detector.gdml

The OKX4Test executable loads a GDML file and translates it into an Opticks GGeo geometry instance and persists that into a geocache.

Geocache creation is time and memory consuming, taking about 1 minute for the JUNO geometry on a workstation with lots of memory. Fortunately this only needs to be done once per geometry, and as the geocache is composed of binary .npy files they are fast to load and upload to the GPU.

Near to the end of the logging from geocache creation you should find output similar to the below which reports the OPTICKS_KEY value of the geometry:

2020-10-28 09:24:26.523 INFO  [6138491] [OpticksProfile::dump@482]  npy 69,4 /tmp/blyth/opticks/source/evt/g4live/torch/OpticksProfile.npy
2020-10-28 09:24:26.523 INFO  [6138491] [OpticksProfile::accumulateDump@339] Opticks::postgeocache nacc 0
2020-10-28 09:24:26.523 INFO  [6138491] [Opticks::reportKey@992] OKX4Test

KEYDIR  /usr/local/opticks/geocache/OKX4Test_World0xc15cfc00x40f7000_PV_g4live/g4ok_gltf/50a18baaf29b18fae8c1642927003ee3/1
KEY     OKX4Test.X4PhysicalVolume.World0xc15cfc00x40f7000_PV.50a18baaf29b18fae8c1642927003ee3

To reuse this geometry include below export in ~/.opticks_config::

    export OPTICKS_KEY=OKX4Test.X4PhysicalVolume.World0xc15cfc00x40f7000_PV.50a18baaf29b18fae8c1642927003ee3

2020-10-28 09:24:26.523 INFO  [6138491] [Opticks::reportKey@1011] This key matches that of the current envvar
2020-10-28 09:24:26.523 INFO  [6138491] [Opticks::dumpRC@240]  rc 0 rcmsg : -

Opticks executables and scripts read the OPTICKS_KEY envvar to determine the geometry to load. Add some lines to ~/.opticks_config exporting the OPTICKS_KEY:

export OPTICKS_KEY=OKX4Test.X4PhysicalVolume.lWorld0x4bc2710_PV.f6cc352e44243f8fa536ab483ad390ce

The corresponding geocache directory is derived from this OPTICKS_KEY and can be found with the geocache-keydir bash function (python and C++ equivalents are opticks.ana.key/Key and optickscore/OpticksResource.cc):

epsilon:ana blyth$ geocache-;geocache-keydir

Common issue : GEOCACHE_CODE_VERSION MISMATCH causes assert at geocache loading

A code version integer Opticks::GEOCACHE_CODE_VERSION is persisted with the geocache metadata during geocache-create. On loading the geocache this persisted int is compared with the current static int from optickscode/Opticks.hh. This is done to track code changes that would invalidate previously written geocache. When there is a mismatch the geocache loading asserts with error:

2020-10-28 21:15:59.717 INFO  [28974] [Opticks::loadOriginCacheMeta@1897] ExtractCacheMetaGDMLPath /home/blyth/local/opticks/opticksaux/export/DayaBay_VGDX_20140414-1300/g4_00_CGeometry_export_v0.gdml
2020-10-28 21:15:59.717 FATAL [28974] [Opticks::loadOriginCacheMeta@1916]
 (current) Opticks::GEOCACHE_CODE_VERSION 7
 (loaded)  m_origin_geocache_code_version 6
 -> RECREATE THE CACHE EG WITH geocache-create

The solution as suggested is to recreate the geocache and ensure that the OPTICKS_KEY exported into the environment is updated if necessary. When running opticks-t against an invalid or non-existing geocache a total of ~86 out of 440 tests are expected to fail. The solution is to recreate the geocache with the updated code with:


Another issue : geocache-create fails to find GDML input file

If geocache-create gives an error like the below:

2020-10-28 21:32:46.910 INFO  [54688] [main@72]  digestextra2 NONE
2020-10-28 21:32:46.910 INFO  [54688] [CGDML::Parse@39] path /home/blyth/local/opticks/opticksaux/export/DayaBay_VGDX_20140414-1300/g4_00_CGeometry_export_v1.gdml
G4GDML: Reading '/home/blyth/local/opticks/opticksaux/export/DayaBay_VGDX_20140414-1300/g4_00_CGeometry_export_v1.gdml'...

-------- EEEE ------- G4Exception-START -------- EEEE -------

*** ExceptionHandler is not defined ***
*** G4Exception : InvalidRead
      issued by : G4GDMLRead::Read()
Unable to open document: /home/blyth/local/opticks/opticksaux/export/DayaBay_VGDX_20140414-1300/g4_00_CGeometry_export_v1.gdml
*** Fatal Exception ***

You probably need to update your clone of the opticksaux repository which is used to share test GDML files. Do this with:


Inspect the geocache with opticks-kcd

The bash function opticks-kcd changes directory to the geocache directory which is derived from the OPTICKS_KEY envvar. Searching for NumPy arrays in the geocache reveals a large number of them.

epsilon:1 blyth$ find . -name '*.npy' | wc -l

epsilon:1 blyth$ find . -name volume_transforms.npy

IPython session loading the placement transform of a particular volume in the geometry:

epsilon:1 blyth$ ipython

In [1]: import numpy as np
In [2]: vt = np.load("GNodeLib/all_volume_transforms.npy")
In [3]: vt.shape
Out[3]: (12230, 4, 4)
In [3]: vt[3154]
array([[      0.54317,      -0.83962,       0.     ,       0.     ],
       [      0.83962,       0.54317,       0.     ,       0.     ],
       [      0.     ,       0.     ,       1.     ,       0.     ],
       [ -18079.453  , -799699.44   ,   -7100.     ,       1.     ]],

Opticks Analysis and Debugging using Python, IPython, NumPy and Matplotlib : manage with miniconda

Opticks uses the NumPy (NPY) buffer serialization format for geometry and event data, thus analysis and debugging requires python and ipython with numpy and matplotib extensions. Optionally pyvista/VTK is useful for fast 3D plotting of large datasets. For management of these and other python packages it is convenient to use miniconda.

Opticks is in the process of migrating scripts from python2 to python3, currently using Python 3.7.8. Please report problems from unmigrated scripts.

After installing miniconda the additional packages can be installed with commands such as:

conda install ipython
conda install numpy
conda install sympy
conda install matplotlib

In addition to your PATH you can also control which python Opticks uses with the optional OPTICKS_PYTHON envvar, for example:

export OPTICKS_PYTHON=python3