search

AIDASoft / DD4hep

Detector Description Toolkit for High Energy Physics
http://dd4hep.cern.ch
GNU Lesser General Public License v3.0
50 stars 100 forks source link

Abnormal material budget of DDCAD-imported files #1134

Closed armin-ilg closed 1 year ago

armin-ilg commented 1 year ago

I also tried to check the material budget with ddsim:

cd FCCee/compact/IDEA/IDEA_o1_v01/
ddsim --compactFile FCCee_IDEA_o1_v01.xml  --runType shell

which however results in a seg fault (ddsim_material_budget.log).

I also tried to use the example STL file from examples/DDCAD/models/STL/IR_assy_all_colour.stl instead of my shape, and also the material budget check in ddsim doesn't run and the material depth is absurdely large (1800 cm, see depth_IR.pdf).

Am I doing something wrong? I tried putting the DDCAD import inside of a box using DD4hep_SingleShape and tried using CAD_Assembly, CAD_MultiVolume and CAD_Shape, but to no avail.

Thank you for your help!

andresailer commented 1 year ago

To make sure the baseline works as expected, what do you get for the beampipe of https://github.com/key4hep/k4geo/tree/master/FCCee/compact/FCCee_o2_v02 ?

armin-ilg commented 1 year ago

Hi André

For my vertex detector and beampipe it works fine, I just also checked for the CLD version you linked: beampipe_depth.pdf (don't know what "beam" is, but the rest looks okay) beampipe_x0.pdf (looks meaningful) beampipe+vertex_x0.pdf (looks meaningful)

Cheers, Armin

MarkusFrankATcernch commented 1 year ago

Concerning the material budget with ddsim: You have a problem there which is noit dd4hep related:

PersistencyIO    INFO  +++ Set Streamer to dd4hep::OpaqueDataBlock
Info in <TGeoManager::TGeoManager>: Geometry default, Detector Geometry created
Info in <TGeoNavigator::BuildCache>: --- Maximum geometry depth set to 100
XMLLoader        INFO  +++ Processing XML file: file:FCCee_IDEA_o1_v01.xml
XercesC          FATAL +++ FATAL Error at file "", Line 0 Column: 0 Message:unable to open primary document entity '/afs/cern.ch/user/a/afehr/lcgeo/FCCee/compact/IDEA/IDEA_o1_v01/file:FCCee_IDEA_o1_v01.xml'
Traceback (most recent call last):
  File "/cvmfs/sw.hsf.org/spackages7/dd4hep/1.23/x86_64-centos7-gcc11.2.0-opt/q4qtg/bin/ddsim", line 25, in <module>
    RUNNER.run()
  File "/cvmfs/sw.hsf.org/spackages7/dd4hep/1.23/x86_64-centos7-gcc11.2.0-opt/q4qtg/lib/python3.10/site-packages/DDSim/DD4hepSimulation.py", line 311, in run
    kernel.loadGeometry(str("file:" + compactFile))
cppyy.gbl.std.runtime_error: void dd4hep::sim::Geant4Kernel::loadGeometry(const string& compact_file) =>
    runtime_error: dd4hep: Failed to parse the XML file file:FCCee_IDEA_o1_v01.xml [Invalid XML ROOT handle]
dd4hep: while parsing file:FCCee_IDEA_o1_v01.xml
dd4hep: with plugin:DD4hep_XMLLoader
 *** Break *** segmentation violation
MarkusFrankATcernch commented 1 year ago

We cannot really support problems embedded inside lcgeo, Gaudi etc. It would take way off more time to accommodate to all frameworks than to actually solve the underlying problems.

Is there no way to strip down the problem to only use dd4hep, ROOT and Geant4? This would greatly help to debug this problem.

armin-ilg commented 1 year ago

Concerning the material budget with ddsim: You have a problem there which is noit dd4hep related:

PersistencyIO    INFO  +++ Set Streamer to dd4hep::OpaqueDataBlock
Info in <TGeoManager::TGeoManager>: Geometry default, Detector Geometry created
Info in <TGeoNavigator::BuildCache>: --- Maximum geometry depth set to 100
XMLLoader        INFO  +++ Processing XML file: file:FCCee_IDEA_o1_v01.xml
XercesC          FATAL +++ FATAL Error at file "", Line 0 Column: 0 Message:unable to open primary document entity '/afs/cern.ch/user/a/afehr/lcgeo/FCCee/compact/IDEA/IDEA_o1_v01/file:FCCee_IDEA_o1_v01.xml'
Traceback (most recent call last):
  File "/cvmfs/sw.hsf.org/spackages7/dd4hep/1.23/x86_64-centos7-gcc11.2.0-opt/q4qtg/bin/ddsim", line 25, in <module>
    RUNNER.run()
  File "/cvmfs/sw.hsf.org/spackages7/dd4hep/1.23/x86_64-centos7-gcc11.2.0-opt/q4qtg/lib/python3.10/site-packages/DDSim/DD4hepSimulation.py", line 311, in run
    kernel.loadGeometry(str("file:" + compactFile))
cppyy.gbl.std.runtime_error: void dd4hep::sim::Geant4Kernel::loadGeometry(const string& compact_file) =>
    runtime_error: dd4hep: Failed to parse the XML file file:FCCee_IDEA_o1_v01.xml [Invalid XML ROOT handle]
dd4hep: while parsing file:FCCee_IDEA_o1_v01.xml
dd4hep: with plugin:DD4hep_XMLLoader
 *** Break *** segmentation violation

Ah sorry, yep, I forgot to add the ./ in front of the xml file name, but still it doesn't work for me (seg fault), here's the updated log from ddsim --compactFile ./FCCee_IDEA_o1_v01.xml --runType shell ddsim_material_budget_updated.log

armin-ilg commented 1 year ago

We cannot really support problems embedded inside lcgeo, Gaudi etc. It would take way off more time to accommodate to all frameworks than to actually solve the underlying problems.

Is there no way to strip down the problem to only use dd4hep, ROOT and Geant4? This would greatly help to debug this problem.

For the ddsim material check I understand, but for the other way of estimating the material budget (cd FCCee/compact/IDEA/IDEA_o1_v01/, sh scripts/check_material_budget.sh), I'm using material_scan.py which is running the DD4hep geometry service under the hood and not much more, how should I further simplify that?

Thank you, Armin

MarkusFrankATcernch commented 1 year ago

-- Which is the xml file you use to feed the material scans? -- I simply have a problem to reproduce all the problems for debugging. If there are external (unresolved) dependencies then debugging is virtually impossible and any answer can only be speculation. Thank you for understanding.

armin-ilg commented 1 year ago

-- Which is the xml file you use to feed the material scans? -- I simply have a problem to reproduce all the problems for debugging. If there are external (unresolved) dependencies then debugging is virtually impossible and any answer can only be speculation. Thank you for understanding.

The xml file I use is FCCee_IDEA_o1_v01.xml, which then includes Vertex_IDEA_o1_v01.xml.

-- I simply have a problem to reproduce all the problems for debugging. If there are external (unresolved) dependencies then debugging is virtually impossible and any answer can only be speculation.

I think for reproducing the error, you could put the STL shape into any working file of a detector (e.g FCCee_o2_v02.xml, removing the other detector parts) with

<detectors>
    <detector id="0" name="Shape_STL" type="DD4hep_SingleShape" vis="CarbonFiberVis">
            <material name="CarbonFiber"/>
            <box x="10*cm" y="10*cm" z="22*cm" vis="VXDVis"/>
            <shape type="CAD_Shape" ref="Assieme_supporti_carbonio.stl" unit="cm">
                <position x="104.3837*cm"  y="-1.7417*cm"        z="-33.7602*cm"/>
            </shape>
            <position x="0*cm" y="0*cm" z="0*cm"/>
            <rotation phi="0*rad" theta="-pi/2*rad"   psi="0*rad"/>                    
   </detector>
</detectors>

using the Assieme_supporti_carbonio.stl STL file from here. This should work in any DD4hep installation, without further dependencies if I understand correctly.

Thank you for your time, Armin

andresailer commented 1 year ago

Hi @armin-ilg

Can you just set /tracking/verbose 1 and then shoot a geantino at different angles (in the geant4 shell via ddsim for example), and see what that reports? That is kind of what (some of) the material scan features do as well, maybe that makes something obvious.

armin-ilg commented 1 year ago

/tracking/verbose 1

Hi @andresailer

Ok, I did the following:

ddsim --compactFile ./FCCee_IDEA_o1_v01.xml --enableG4Gun --runType shell
Idle> /gun/particle geantino
Idle> /tracking/verbose 1
Idle> /gun/number 10        
Idle> /run/beamOn

The output again is a *** Break *** segmentation violation. You can find the complete log here: ddsim_geantino_DDCAD.log

When commenting out the DDCAD part, it runs without issue. I also again tried using the STL example file (IR_assy_all_colour.stl) instead of my STL file, with the same segmentation violation error coming out.

Thank you for your time, Armin

andresailer commented 1 year ago

Thanks for trying this out!

Does the simulation always fail with the CAD geometry or does it only fail for geantinos or for /tracking/verbose > 0?

MarkusFrankATcernch commented 1 year ago

If I summarize: The CAD simulation fails here:


#9  0x00007f8f39aabb29 in vecgeom::cxx::HybridManager2::GetABBoxes_v (numberOfNodes=<optimized out>, size=<optimized out>, structure=..., this=<optimized out>) at /cvmfs/sw.hsf.org/spackages7/vecgeom/1.2.1/x86_64-centos7-gcc11.2.0-opt/4fr2x/include/VecGeom/management/HybridManager2.h:102
#10 vecgeom::cxx::HybridNavigator<false>::GetHitCandidates_v (this=this
entry=0x7f8f39c071f0 <vecgeom::cxx::HybridNavigator<false>::Instance()::instance>, accstructure=..., point=..., dir=..., maxstep=4.73747262e+30, hitlist=hitlist
entry=0x7ffdcdd126e0) at /cvmfs/sw.hsf.org/spackages7/vecgeom/1.2.1/x86_64-centos7-gcc11.2.0-opt/4fr2x/include/VecGeom/navigation/HybridNavigator2.h:122
#11 0x00007f8f39aabf5c in vecgeom::cxx::HybridNavigator<false>::BVHSortedIntersectionsLooper<vecgeom::cxx::HybridManager2::HybridBoxAccelerationStructure, vecgeom::cxx::TessellatedImplementation::DistanceToSolid<double, false>(vecgeom::cxx::TessellatedStruct<3ul, double> const&, vecgeom::cxx::Vector3D<double> const&, vecgeom::cxx::Vector3D<double> const&, double const&, double&, int&, double&, int&)::{lambda(std::pair<int, double>)#1}&>(vecgeom::cxx::HybridManager2::HybridBoxAccelerationStructure const&, vecgeom::cxx::Vector3D<double> const&, vecgeom::cxx::Vector3D<double> const&, double, vecgeom::cxx::TessellatedImplementation::DistanceToSolid<double, false>(vecgeom::cxx::TessellatedStruct<3ul, double> const&, vecgeom::cxx::Vector3D<double> const&, vecgeom::cxx::Vector3D<double> const&, double const&, double&, int&, double&, int&)::{lambda(std::pair<int, double>)#1}&) const (userhook=<synthetic pointer>..., maxstep=<optimized out>, localdir=..., localpoint=..., accstructure=..., this=0x7f8f39c071f0 <vecgeom::cxx::HybridNavigator<false>::Instance()::instance>) at /cvmfs/sw.hsf.org/spackages7/vecgeom/1.2.1/x86_64-centos7-gcc11.2.0-opt/4fr2x/include/VecGeom/navigation/HybridNavigator2.h:179
#12 vecgeom::cxx::TessellatedImplementation::DistanceToSolid<double, false> (tessellated=..., point=..., direction=..., stepMax=
0x7ffdcdd60948: 1.7976931348623157e+308, distance=
0x7ffdcdd60938: 1.7976931348623157e+308, isurf=
0x7ffdcdd60930: -1, distother=
0x7ffdcdd60940: 1.7976931348623157e+308, isurfother=
0x7ffdcdd60934: -1) at /cvmfs/sw.hsf.org/spackages7/vecgeom/1.2.1/x86_64-centos7-gcc11.2.0-opt/4fr2x/include/VecGeom/volumes/kernel/TessellatedImplementation.h:255
#13 0x00007f8f39b4e904 in vecgeom::cxx::TessellatedImplementation::Inside<double, int> (inside=<synthetic pointer>: 3, point=..., tessellated=...) at /cvmfs/sw.hsf.org/spackages7/vecgeom/1.2.1/x86_64-centos7-gcc11.2.0-opt/4fr2x/include/VecGeom/volumes/kernel/TessellatedImplementation.h:80
#14 vecgeom::cxx::CommonUnplacedVolumeImplHelper<vecgeom::cxx::TessellatedImplementation, vecgeom::cxx::VUnplacedVolume>::Inside (p=..., this=<optimized out>) at /cvmfs/sw.hsf.org/spackages7/vecgeom/1.2.1/x86_64-centos7-gcc11.2.0-opt/4fr2x/include/VecGeom/volumes/UnplacedVolumeImplHelper.h:145
#15 G4UAdapter<vecgeom::cxx::UnplacedTessellated>::Inside (this=0x103a3660, p=...) at /tmp/gitlab-runner/spack-stage/spack-stage-geant4-11.1.1-jlctn2ni5o7xwzvcby5ckdrhox2s4pol/spack-src/source/geometry/management/include/G4UAdapter.hh:314
#16 0x00007f8f3997dd73 in G4AuxiliaryNavServices::CheckPointOnSurface (locatedOnEdge=false, sampleTransform=..., globalDirection=0x177ec3d0, localPoint=..., sampleSolid=0x103a3660) at /tmp/gitlab-runner/spack-stage/spack-stage-geant4-11.1.1-jlctn2ni5o7xwzvcby5ckdrhox2s4pol/spack-src/source/geometry/navigation/include/G4AuxiliaryNavServices.icc:41

But it is not understood why.... This will be a very difficult one to sort out. I will have to consult Andrei how to get at this in detail.

armin-ilg commented 1 year ago

Thanks for trying this out!

Does the simulation always fail with the CAD geometry or does it only fail for geantinos or for /tracking/verbose > 0?

It fails in the same way with /tracking/verbose 0, and also when I shoot mu- instead of geantino.

MarkusFrankATcernch commented 1 year ago

I tried to reproduce the results of Armin with his file: Assieme supporti carbonio_1.stl.zip Please see the results here: https://github.com/AIDASoft/DD4hep/pull/1139

MarkusFrankATcernch commented 1 year ago

@armin-ilg Armin, I do not exactly know what you are doing wrong, but something is very fishy in your setup. Please study merge request 1139

I did all stuff like material scan, overlap checks etc. using your STL file without problems. Though I have to say the coordinate system of your STL is quite weird and I first had to move the entire setup somewhere close to the origin.

I do not see any problems.

armin-ilg commented 1 year ago

@armin-ilg Armin, I do not exactly know what you are doing wrong, but something is very fishy in your setup. Please study merge request 1139

I did all stuff like material scan, overlap checks etc. using your STL file without problems. Though I have to say the coordinate system of your STL is quite weird and I first had to move the entire setup somewhere close to the origin.

I do not see any problems.

Thank you very much for your help @MarkusFrankATcernch and @andresailer. It is indeed an issue of my local setup. I must have done something wrong in the installation of my local k4geo repo. It works for me as well now by installing k4geo with

source /cvmfs/sw-nightlies.hsf.org/key4hep/releases/2023-06-29/x86_64-centos7-gcc12.2.0-opt/key4hep-stack/2023-06-29-c6wpxe/setup.sh

# Central dd4hep
source $DD4hep_DIR/bin/thisdd4hep.sh

cmake -DBoost_NO_BOOST_CMAKE=ON ..
make -j4 install
source ../bin/thislcgeo.sh

export PYTHONPATH=${LCIO}/src/python:${ROOTSYS}/lib:$PYTHONPATH

Overlap checks work now. Material budget scans using g4MaterialScan are meaningful as well, though using the material budget estimation from FCCSW is still producing wrong results, I'll open an issue there.

Thank you very much for your support, Armin