replicas

[maureeungaro]

replicas are not working when a volume is placed inside replicas - see attached files.
an improvement of the api would be to add the replica pars in the type description, this will remove the need of an additional volume.

replica.gcard.txt
solid_PVDIS_ec_forwardangle__geometry_Original.txt

[zhaozhiwen]

=== emails related to the two files ================
"component" is a GEMC thing, right?
You don't do physical placement for any "component".
I think you only use it when building Boolean operations.
I am not sure how you use it for any thing else.

"copyof" is a GEMC thing also, right?
It does multiple physical placement of with one logic volume.
But it's not geant4 "replicas" which saves mem

I think my modification works.
Can you show me a working example you are referring to?
For example, can you do the B4 in GEMC?

Zhiwen

On 11/23/2015 4:14 PM, Maurizio Ungaro wrote:

Hi Zhiwen,

Oh I think there’s a misunderstanding on how replicas work in gemc.

1. you defined the volume to be replicated. The material is “component”
2. you define a separate volume that build the replicas. That material is not component. This is a different entry in the perl script.

Let me know if that works,

Mauri

On Nov 23, 2015, at 2:31 PM, Zhiwen Zhao zwzhao@jlab.org wrote:

About bug, refer to the example here
http://geant4.web.cern.ch/geant4/UserDocumentation/Doxygen/examples_doc/html/classB4DetectorConstruction.html
logic volume needs to be built first as any other volume, so we would need type and dimension to do it
then we replica instead of regular placement.
In my code, I used "description" to carry the additional info need for replica and thus leave type and dimension intact for
regular logic volume building.
This find of fix may work for Parameterised or Divisions also because they needs additional info also.
I am not sure if using "description" is good for long term, but I don't want to break backward compatilbity.

On 11/23/2015 1:45 PM, Maurizio Ungaro wrote:

Hi Zhiwen,

I think I understand this bug. This can be checked in detector construction. Let me work on that.

I don’t understand why you put the replica constructor pars in “description” instead of “dimensions”, can you elaborate?

I fixed the bugs you reported on the installation scripts and the scons_bm. Testing now on mac, fedora23 and ifarm65.

Mauri

On Nov 23, 2015, at 9:28 AM, Zhiwen Zhao zwzhao@jlab.org wrote:

all right

I fix the bug to built "replica" in detector.cc of gemc 2.2 and I think the same bug is at 2.1 and 2.3
The bug is due to that the logic volume of the replica was not built before its physical placement
refer to attached files

I still found one problem with after my fix
In gui mode,if I use USE_GUI=1, the mem usage is HUGE for the calorimeter example I have,
but USE_GUI=2 works fine with small mem usage
running is batch mode is fine with small mem usage also.
Do you have any insight about it?

I use "description" item for the job now, not sure if this is good for long term,
unless we add a new item, but I don't want to break backward compatibility.
Also, I don't see Parameterised or Divisions volumes are added at all, maybe we should them all together?

Zhiwen

[zhaozhiwen]

=== another email history =================

I am puzzled by
"G4PVReplica represents nReplicas volumes differing only in their positioning, and completely filling the containing mother volume. Consequently if a G4PVReplica is 'positioned' inside a given mother it MUST be the mother's only daughter volume. Replica's correspond to divisions or slices that completely fill the mother volume and have no offsets."
from https://geant4.web.cern.ch/geant4/UserDocumentation/UsersGuides/ForApplicationDeveloper/html/ch04.html

Zhiwen

On 11/25/2015 3:17 PM, Maurizio Ungaro wrote:

These are replicas, not divisions

The first number is axis (which btw I can see in replicas it could be azimuthal as well. Third number is the distance between volumes and last one is starting offset.

Mauri

On Nov 25, 2015, at 3:10 PM, Zhiwen Zhao zwzhao@jlab.org wrote:

Thanks a lot!
I tried devel on github and it shows 10 boxes on screen
(the sgementation fault seems gone somehow, great!)
I guess you are slicing the mother "root" into 10 boxes.
I don't quite understand $detector{"dimensions"} = "0 10counts 20cm 5cm";
0 means pAxis is some default value, not kXaxis/kYaxis/kZaxis/
10 means 10 slices
20cm means the width is 20cm
5cm means offset by 5*cm
I need a bit more time to understand and think about it.

Zhiwen

On 11/25/2015 2:19 PM, Maurizio Ungaro wrote:

Hi Zhiwen,

Can you try the new package in devel? It should handle replicas properly now.

See attached example.

Mauri

replicaExample.pl.txt
config.dat.txt