Running multiple rpict commands for different scenes

What exactly is your question? Of course you can start multiple rtrace or whatever processes, they will not care about each other. If you run more processes then CPUs, they will slow down, same is true (even worse) if you exceed available memory. If you have lots of scenes, you can write a script to automate the handling. You may also like batch processing queues if you have a lot of processes to distribute. Still, it is impossible to answer your question if it is that general.

Cheers, Lars.

···

--
Dipl.-Ing. Architect Lars O. Grobe

On May 18, 2011, at 11:11, "Sotiris Papantoniou" <[email protected]> wrote:

Hello list again,

I have a new issue concerning multiple simulations. I would run the same command (rpict, rtrace, rad) many times but for different scenes.oct. Do you know if this is possible?

Thank you in advance

Sotiris

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

Hello Lars and Christopher,

My idea is to run many simulations, for now up to 1200, in order to create data for a fuzzy system. I have many simulations because I have a glass part of a building which has 4 sets of blinds which are autonomously controlled. If I am thinking correctly when the angle of each blind changes we have a new scene, which is separately simulated. For each scene I am running a new rpict command to see the result. I found out that in the University we have a grid-computer which is working on Linux. From the grid-computer web-site I read the following:

Translators

* C (gcc 4.1.2 <http://gcc.gnu.org/onlinedocs/gcc-3.4.6/gcc/> gcc44 4.4.0)
* Fortran77 <http://www.ibiblio.org/pub/languages/fortran/> & 98 (gfortran 4.1.2 gfortan44 4.4.0)
* C++ <http://gcc.gnu.org/onlinedocs/gcc-3.4.6/gcc/> (g++ 4.1.2 g++44 4.4.0)
* Java <http://java.sun.com/> (jdk 1.6.0 gcj 4.1.2)
* Python <http://www.python.org/> 2.4.3

Data bases

* MySQL 5.0.77 <http://dev.mysql.com/doc/>

Computational Systems

* Matlab 7.9.0 <http://www.mathworks.com/products/matlab/>
* Octave 3.0.5 <http://www.gnu.org/software/octave/about.html>
* R-2.12.2 <http://www.r-project.org/>

* Management systems and grid operations
* TORQUE 2.3.6 <http://www.clusterresources.com/products/torque-resource-manager.php>
* openMPI 1.4.1 <http://www.open-mpi.org/>
* maui 3.2 <http://www.clusterresources.com/products/maui/>

This is the software of the grid. So I was thinking, to install Radiance on the grid and run the commands from it. So my main question is if I can use radiance on a grid-computer, and If I use the advantage of the grid by running in parallel the same command.

For now, since I was in a hurry to get some results, I am creating the combinations in Matlab and I am running rpict from DOS. I know it is old but I for now I have in the University only an Win OS computer.

I hope I managed to give you some details. If not please tell me so, in order to reply again.

Thank you in advance for your time

Sotiris

···

From: Lars O. Grobe [mailto:[email protected]]
Sent: Wednesday, May 18, 2011 12:14 PM
To: Radiance general discussion
Subject: Re: [Radiance-general] Running multiple rpict commands for different scenes

What exactly is your question? Of course you can start multiple rtrace or whatever processes, they will not care about each other. If you run more processes then CPUs, they will slow down, same is true (even worse) if you exceed available memory. If you have lots of scenes, you can write a script to automate the handling. You may also like batch processing queues if you have a lot of processes to distribute. Still, it is impossible to answer your question if it is that general.

Cheers, Lars.

--

Dipl.-Ing. Architect Lars O. Grobe

On May 18, 2011, at 11:11, "Sotiris Papantoniou" <[email protected]> wrote:

Hello list again,

I have a new issue concerning multiple simulations. I would run the same command (rpict, rtrace, rad) many times but for different scenes.oct. Do you know if this is possible?

Thank you in advance

Sotiris

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

  _____

No virus found in this message.
Checked by AVG - www.avg.com
Version: 10.0.1375 / Virus Database: 1509/3644 - Release Date: 05/17/11

Hi!

My idea is to run many simulations, for now up to 1200, in order to
create data for a fuzzy system. I have many simulations because I have
a glass part of a building which has 4 sets of blinds which are
autonomously controlled. If I am thinking correctly when the angle of
each blind changes we have a new scene, which is separately simulated.
For each scene I am running a new rpict command to see the result. I
found out that in the University we have a grid-computer which is
working on Linux. From the grid-computer web-site I read the following:

Two options:

A) Create your scene octrees, name them according to your parameters. Create a rif-file with the settings you use. In this rif-file, you use dummies names for the critical parameters that should not be shared (e.g. all file names including the ambient file). Write a script that loops through all the possible names (and thus setups), replaces the dummy by the current scene name, runs rad on the rif file. sed is a very useful command for doing so.

B) Create one rif file, and replace the varying parts of your scene (can be material definitions, geometry, ...) by shell commands (preceded by !) that produce the output. In the simplest way, this command can just return the value of an environment variable that is altered by your script. E.g. in your script, you set the variable myVariableMat="void plastic thisTimeItsWhite 0 0 5 1 1 1 0 0", and in your material definition file, you reference it by adding a line !echo $myVariableMat.

Hope this helps.

Cheers, Lars.

Hi Sotiris,

There are several things that you are going to have to figure out in order to take advantage of the compute resources available at your university. Hopefully there is some administrative support that can provide some basic advice. But I think that you will need to figure out the following, and this is likely not going to be an fast learning curve unless you are accustomed to working with this type of system and the scripting that will likely be involved. It will take some trial and error, but once you get it figured out it will probably be well worth the effort to have such a resource available.

   1. You will need to figure out how you need to install your
      application (Radiance) to make it accessible to the cluster.
   2. You will need to figure out how to script radiance to to make the
      scene changes that are needed. This might involve some selection
      of: make, csh, perl, python or whatever other scripting language
      you enjoy. I think that it may also be possible to control through
      Matlab/Octave (see
      http://web.mac.com/geotrupes/iWeb/Main%20site/RadBlog/E549E7F4-6DA2-4D78-8F91-74A4691ED86A.html
      at Giulio Antonutto's site). If the only variation in your scene
      is going to be the orientation of the individual blind slats, then
      there are a number of ways to script this and one option might be
      to consider looking at the radiance program ranimate with the
      ANIMATE variable set to a script to rotates the slat geometry by a
      given increment or parameter, with this you could also change
      other scene conditions such as the position of the sun/sky
      condition. But this could also be done with the other scripting
      methods I indicate.
   3. Lastly you will need to figure out how to submit your radiance
      rendering job to the grid. Based on the information that you
      provide it is most likely that this would be done via "qsub" which
      is a job submital routine for Torque, which looks to be the
      queuing system for the cluster (Maui takes care of how the
      resources are actually allocated). Unfortunately this (job
      submittal with qsub) also requires some scripting, however there
      should be "howto" resources available from the system
      administrators of the cluster. Such a script will have parameters
      indicating what resources (nodes, cpus, cores and time are needed
      for the job) as well as what the job comprises. In this case you
      would probably want to script something that actually generates
      1200 separate "jobs" that can then be submitted individually to
      the cluster as resources become available. If you "owned" the
      cluster you would be able to do things more simply using ranimate
      to distribute the job over selected cluster nodes.

Regards,

-Jack de Valpine

···

--
# Jack de Valpine
# president
#
# visarc incorporated
# http://www.visarc.com
#
# channeling technology for superior design and construction

On 5/18/2011 6:29 AM, Sotiris Papantoniou wrote:

Hello Lars and Christopher,

My idea is to run many simulations, for now up to 1200, in order to create data for a fuzzy system. I have many simulations because I have a glass part of a building which has 4 sets of blinds which are autonomously controlled. If I am thinking correctly when the angle of each blind changes we have a new scene, which is separately simulated. For each scene I am running a new rpict command to see the result. I found out that in the University we have a grid-computer which is working on Linux. From the grid-computer web-site I read the following:

*Translators***

    * C (gcc 4.1.2 gcc44 4.4.0)
      <http://gcc.gnu.org/onlinedocs/gcc-3.4.6/gcc/>
    * Fortran77 & 98 (gfortran 4.1.2 gfortan44 4.4.0)
      <http://www.ibiblio.org/pub/languages/fortran/>
    * C++ (g++ 4.1.2 g++44 4.4.0)
      <http://gcc.gnu.org/onlinedocs/gcc-3.4.6/gcc/>
    * Java (jdk 1.6.0 gcj 4.1.2) <http://java.sun.com/>
    * Python 2.4.3 <http://www.python.org/>

*Data bases*

    * MySQL 5.0.77 <http://dev.mysql.com/doc/>

*Computational****Systems***

    * Matlab 7.9.0 <http://www.mathworks.com/products/matlab/>
    * Octave 3.0.5 <http://www.gnu.org/software/octave/about.html>
    * R-2.12.2 <http://www.r-project.org/>

    * *Management systems****and****grid****operations***
    * TORQUE 2.3.6
      <http://www.clusterresources.com/products/torque-resource-manager.php>
    * openMPI 1.4.1 <http://www.open-mpi.org/>
    * maui 3.2 <http://www.clusterresources.com/products/maui/>

This is the software of the grid. So I was thinking, to install Radiance on the grid and run the commands from it. So my main question is if I can use radiance on a grid-computer, and If I use the advantage of the grid by running in parallel the same command.

For now, since I was in a hurry to get some results, I am creating the combinations in Matlab and I am running rpict from DOS. I know it is old but I for now I have in the University only an Win OS computer.

I hope I managed to give you some details. If not please tell me so, in order to reply again.

Thank you in advance for your time

Sotiris

*From:*Lars O. Grobe [mailto:[email protected]]
*Sent:* Wednesday, May 18, 2011 12:14 PM
*To:* Radiance general discussion
*Subject:* Re: [Radiance-general] Running multiple rpict commands for different scenes

What exactly is your question? Of course you can start multiple rtrace or whatever processes, they will not care about each other. If you run more processes then CPUs, they will slow down, same is true (even worse) if you exceed available memory. If you have lots of scenes, you can write a script to automate the handling. You may also like batch processing queues if you have a lot of processes to distribute. Still, it is impossible to answer your question if it is that general.

Cheers, Lars.

--

Dipl.-Ing. Architect Lars O. Grobe

On May 18, 2011, at 11:11, "Sotiris Papantoniou" > <[email protected] <mailto:[email protected]>> wrote:

    Hello list again,

    I have a new issue concerning multiple simulations. I would run
    the same command (rpict, rtrace, rad) many times but for different
    scenes.oct. Do you know if this is possible?

    Thank you in advance

    Sotiris

    _______________________________________________
    Radiance-general mailing list
    [email protected]
    <mailto:[email protected]>
    http://www.radiance-online.org/mailman/listinfo/radiance-general

------------------------------------------------------------------------

No virus found in this message.
Checked by AVG - www.avg.com <http://www.avg.com>
Version: 10.0.1375 / Virus Database: 1509/3644 - Release Date: 05/17/11

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general