[Halld-offline] 100x Slower MC Generation With version_4.0.0 Software

[jzarling at jlab.org]

Ok, thanks! Unfortunate that it sounds kind of tricky at present... I have
no immediate rush (yet) so perhaps I can wait this out if a more elegant
solution is somewhere on the horizon.


Jon

> Hi Jon,
>
>   The way hdgeant gets its geometry requires the XML be converted to
> fortran and compiled
> into a binary. The mechanism in hdgeant does this, but predates geometry
> info in CCDB.
> To use an alternate geometry with hdgeant, you need to pass the
> -xml=main_HDDS.xml
> command line argument where the main_HDDS.xml refers to an actual file
> accessible on the
> computer. i.e. you will need to have checked out HDDS and point to that.
> Additionally, you
> will need to have built HDDS so the hdds-meant executable is available.
>
>   To make it work with a geometry from CCDB you will need to extract the
> geometry files
> to a directory and point to the main_HDDS.xml there. There may be a script
> or program to
> do this, but I’d have to dig around to find it. Perhaps someone else
> knows?
>
>
> Regards,
> -David
>
> -------------------------------------------------------------
> David Lawrence Ph.D.
> Staff Scientist, Thomas Jefferson National Accelerator Facility
> Newport News, VA
> davidl at jlab.org<mailto:davidl at jlab.org>
> (757) 269-5567 W
> (757) 746-6697 C
>
>
> On Jan 28, 2019, at 12:55 PM, Jonathan Zarling
> <jzarling at iu.edu<mailto:jzarling at iu.edu>> wrote:
>
> Hi all,
>
> I tried changing my JANA_GEOMETRY_URL to ccdb:///GEOMETRY/main_HDDS.xml to
> use the ccdb as my geometry source (and using the mysql ccdb file
> mysql://ccdb_user@hallddb.jlab.org/ccdb<http://ccdb_user@hallddb.jlab.org/ccdb>).
> Still appears quite slow with the same "Warning in savehits: unsaved hit
> information found from sensitive volume XXXX" messages as before. Run
> 30300.
>
> Is there some other way I should be configuring things, or does the ccdb
> by default include the DIRC for runs even in this range at present?
>
>
> Jon
>
>
>
> On Mon, Jan 28, 2019 at 12:09 PM Thomas Britton
> <tbritton at jlab.org<mailto:tbritton at jlab.org>> wrote:
> A little late for the party. This effect appears to be reproduced by
> MCwrapper bot. Submitted projects with the 4.0.0 tag are taking
> significantly longer to finish.
>
> Thomas Britton
>
> On Jan 28, 2019, at 11:21 AM, Richard Jones
> <richard.t.jones at uconn.edu<mailto:richard.t.jones at uconn.edu>> wrote:
>
> Hello Dave,
>
> It is not clear what additional development would be needed, provided that
> it still works as originally designed. Shouldn't it just work, once the
> run number is set to the appropriate run period? How does one turn on this
> behavior in hdgeant?
>
> -Richard J.
>
> On Mon, Jan 28, 2019 at 10:28 AM David Lawrence
> <davidl at jlab.org<mailto:davidl at jlab.org>> wrote:
> Hi Guys,
>
>  Sorry I’m late on this. I somehow missed this conversation earlier.
>
> The feature Richard references is one I put in quite some time ago to
> dynamically compile and link
> an hdds geometry so that hdgeant could effectively use an alternative
> geometry to what was originally
> compiled into the executable. I assume the mechanism still works, but have
> not exercised it in a long time.
> In order to use it as Jon suggested (with a flag to turn off DIRC and run
> number specifying specific
> version from CCDB) will take a little effort, but is not impossible. It
> will be a bit of a Rube-Goldberg though.
> If you can get away with just having the HDDS version of interest checked
> out and available so the user
> can point hdgeant to it, then no additional development would be needed.
>
> The software group can decide if having the automated feature in hdgeant3
> is worth it and I’ll help
> implement it as needed.
>
>
> Regards,
> -David
>
> -------------------------------------------------------------
> David Lawrence Ph.D.
> Staff Scientist, Thomas Jefferson National Accelerator Facility
> Newport News, VA
> davidl at jlab.org<mailto:davidl at jlab.org>
> (757) 269-5567 W
> (757) 746-6697 C
>
>
> On Jan 25, 2019, at 5:38 PM, Richard Jones
> <richard.t.jones at uconn.edu<mailto:richard.t.jones at uconn.edu>> wrote:
>
> Justin,
>
> Yes, that was the decision at that time, but then as I looked into doing
> this, I discovered this code in the hdgeant codebase that David had put in
> to create the capability to dynamically compile and load the hdds library
> from sources in ccdb at runtime. At that point, I stopped doing any
> further work on this until hearing more about that project from David.
>
> -Richard Jones
>
> On Fri, Jan 25, 2019 at 5:22 PM Justin Stevens
> <jrsteven at jlab.org<mailto:jrsteven at jlab.org>> wrote:
> I thought the discussion at the January 8th software meeting was that we
> should make the DIRC inactive in hdgeant3, so it wouldn’t cause these
> kinds of issues
> https://halldweb.jlab.org/wiki/index.php/GlueX_Software_Meeting,_January_8,_2019#Merging_the_DIRC_branches<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fhalldweb.jlab.org%2Fwiki%2Findex.php%2FGlueX_Software_Meeting%2C_January_8%2C_2019%23Merging_the_DIRC_branches&data=02%7C01%7C%7Caab7955521c147b6e6fd08d6854e785a%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636842969405700696&sdata=5YJhSIJPH7m1wuLgbpURB%2FJg7B2KaqNm%2B2pguWJjVp8%3D&reserved=0>.
>
> The suggestion was to use a run-dependent CCDB constant
> https://halldweb.jlab.org/cgi-bin/ccdb/versions?table=/DIRC/install_status<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fhalldweb.jlab.org%2Fcgi-bin%2Fccdb%2Fversions%3Ftable%3D%2FDIRC%2Finstall_status&data=02%7C01%7C%7Caab7955521c147b6e6fd08d6854e785a%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636842969405700696&sdata=RhaCo4RFMZt11rHOWbcRDyk%2BubICbXspZANkb6iSJ2k%3D&reserved=0>
> to determine at run time in hdgeant3 if the DIRC material should be
> included or not.  For install_status = 0 the DIRC should be excluded
> entirely, but for install_status = 1 the DIRC could be included as dead
> material (to see the effects on other subsystems) but should be inactive
> to not trace all the cherenkov photons (which is where the major
> computation slowdown comes from), since we will never simulate the DIRC
> with hdgeant3.
>
> Of course, if it’s possible to get the geometry from CCDB dynamically in
> hdgeant3 (like we can for hdgeant4) this would be ideal.
>
> -Justin
>
> On Jan 25, 2019, at 4:26 PM, Richard Jones
> <richard.t.jones at uconn.edu<mailto:richard.t.jones at uconn.edu>> wrote:
>
> Hello Jon,
>
> There is nothing at the moment. This is a good question to the offline
> software group. One idea might be for the master branch of hdds to contain
> the latest image of the detector, and then have various other named
> branches for earlier run periods, like spring_2017, fall_2018 and the
> like. How does that sound to you? Of course, David's solution where he
> builds the hdgeant geometry dynamically at runtime from sources in ccdb
> would be the most flexible, and would make these special branches
> unnecessary.
>
> -Richard
>
> On Fri, Jan 25, 2019 at 4:20 PM Jonathan Zarling
> <jzarling at iu.edu<mailto:jzarling at iu.edu>> wrote:
> Ah ok, got it. So the hdds geometry includes the dirc regardless of run
> number? Is there any documentation you can point me to for the various use
> cases at present for hdgeant/hdgeant4 with this latest software?
>
> Jon
>
> On Fri, Jan 25, 2019 at 4:08 PM Richard Jones
> <rjones30 at gmail.com<mailto:rjones30 at gmail.com>> wrote:
> Hello Jon,
>
> Can you switch to using ccdb as your source for the geometry? The hdds
> text-file geometry now has the dirc in it, and I don't think that is what
> you want. For now and all times future, you want to use the run-dependent
> geometry from the ccdb, rather than the latest image from hdds, unless you
> have created your own private hdds branch for some study you are working
> on with a custom geometry.
>
> To use hdgeant (not hdgeant4) with ccdb geometry takes a bit of work at
> present, but at one point Dave Lawrence was working on a system for use
> with hdgeant that would dynamically compile and link the geometry package
> from the sources in ccdb at runtime, without having to extract the xml
> files by hand for the desired run number and build a custom hdgeant for a
> given run. I have cc'ed Dave on this email, as he might be able to update
> us on the status of that work. Dave, did that effort ever materialize?
>
> -Richard
>
> On Fri, Jan 25, 2019 at 3:53 PM Jonathan Zarling
> <jzarling at iu.edu<mailto:jzarling at iu.edu>> wrote:
> Hi Richard et al,
>
> Looks like I am using the path $GLUEX_TOP/hdds/hdds-4.0.0/main_HDDS.xml
> for $JANA_GEOMETRY_URL. I'm using run number 30300 as a test case.
>
>
> Jon
>
> On Thu, Jan 24, 2019 at 5:20 PM Richard Jones
> <rjones30 at gmail.com<mailto:rjones30 at gmail.com>> wrote:
> Hello Jon,
>
> What is your value of JANA_GEOMETRY_URL? What is your simulation run
> number?
>
> -Richard Jones
>
> On Thu, Jan 24, 2019 at 4:48 PM Jonathan Zarling
> <jzarling at iu.edu<mailto:jzarling at iu.edu>> wrote:
> Hey gang,
>
> My MC jobs that I run locally at IU have gone from 5 minutes to run to 10
> hours when using software version "version_4.0.0". I'm running geant3. Has
> anyone else tried working with this build yet?
>
> Possibly also related: during the geant step I get many warning messages
> of the form: "Warning in savehits: unsaved hit information found from
> sensitive volume XXXX, hit discarded". These blow up my logfiles to
> ~hundreds of megabytes.
>
> I tried running with version_3.8 and version_3.9 too. These ran fast and
> didn't produce the same warning messages.
>
> I experience this both with a photon gun and an omega->3pi simulation.
>
> Cheers,
> Jon
>
> --
> You received this message because you are subscribed to the Google Groups
> "GlueX Software Help" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to
> gluex-software+unsubscribe at googlegroups.com<mailto:gluex-software+unsubscribe at googlegroups.com>.
> To post to this group, send email to
> gluex-software at googlegroups.com<mailto:gluex-software at googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/gluex-software/CACPp7wHuYrG6i6FakKik%2BT69Q9LxHv-aSJTvQbH_wHiwtBHcpw%40mail.gmail.com<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgroups.google.com%2Fd%2Fmsgid%2Fgluex-software%2FCACPp7wHuYrG6i6FakKik%252BT69Q9LxHv-aSJTvQbH_wHiwtBHcpw%2540mail.gmail.com%3Futm_medium%3Demail%26utm_source%3Dfooter&data=02%7C01%7C%7Caab7955521c147b6e6fd08d6854e785a%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636842969405700696&sdata=wukmiqG4W5EJTNdEOJh1zbrMr%2B%2BOHsgRzh%2B6WVB7wAM%3D&reserved=0>.
> For more options, visit
> https://groups.google.com/d/optout<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgroups.google.com%2Fd%2Foptout&data=02%7C01%7C%7Caab7955521c147b6e6fd08d6854e785a%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636842969405700696&sdata=QQCDlv%2BuN58sx%2BYkwbomj3g4BB%2BdOQeiwuF5U4smwr0%3D&reserved=0>.
>
> --
> You received this message because you are subscribed to the Google Groups
> "GlueX Software Help" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to
> gluex-software+unsubscribe at googlegroups.com<mailto:gluex-software+unsubscribe at googlegroups.com>.
> To post to this group, send email to
> gluex-software at googlegroups.com<mailto:gluex-software at googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/gluex-software/CABfxa3RO%3D49egcVp0hsL%2BO9rZQY-9mJw2R1BYOp5KeQoD0%2BkPw%40mail.gmail.com<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgroups.google.com%2Fd%2Fmsgid%2Fgluex-software%2FCABfxa3RO%253D49egcVp0hsL%252BO9rZQY-9mJw2R1BYOp5KeQoD0%252BkPw%2540mail.gmail.com%3Futm_medium%3Demail%26utm_source%3Dfooter&data=02%7C01%7C%7Caab7955521c147b6e6fd08d6854e785a%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636842969405700696&sdata=KYsPpFcc9yQEgPE3hjR6xaUDgdv%2BL8d5MEuKV5DGBVQ%3D&reserved=0>.
> For more options, visit
> https://groups.google.com/d/optout<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgroups.google.com%2Fd%2Foptout&data=02%7C01%7C%7Caab7955521c147b6e6fd08d6854e785a%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636842969405700696&sdata=QQCDlv%2BuN58sx%2BYkwbomj3g4BB%2BdOQeiwuF5U4smwr0%3D&reserved=0>.
> _______________________________________________
> Halld-offline mailing list
> Halld-offline at jlab.org<mailto:Halld-offline at jlab.org>
> https://mailman.jlab.org/mailman/listinfo/halld-offline<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmailman.jlab.org%2Fmailman%2Flistinfo%2Fhalld-offline&data=02%7C01%7C%7Caab7955521c147b6e6fd08d6854e785a%7Cb4d7ee1f4fb34f0690372b5b522042ab%7C1%7C0%7C636842969405856949&sdata=2DZIaWQoE9jafiCgluu9wapSdC8aKb731Sxk5vaucAE%3D&reserved=0>
>
> _______________________________________________
> Halld-offline mailing list
> Halld-offline at jlab.org<mailto:Halld-offline at jlab.org>
> https://mailman.jlab.org/mailman/listinfo/halld-offline
>
> _______________________________________________
> Halld-offline mailing list
> Halld-offline at jlab.org<mailto:Halld-offline at jlab.org>
> https://mailman.jlab.org/mailman/listinfo/halld-offline
>
> _______________________________________________
> Halld-offline mailing list
> Halld-offline at jlab.org
> https://mailman.jlab.org/mailman/listinfo/halld-offline