23288: [CEA 1626] Meshgems v2.3

[plugins/ghs3dprlplugin.git] / doc / salome / gui / GHS3DPRLPLUGIN / input / ghs3dprl_hypo.doc

/*!

\page ghs3dprl_hypo_page MG-Tetra Parallel Parameters hypothesis

\n MG-Tetra Parameters hypothesis works only with <b>MG-Tetra</b>
meshing algorithm which uses <b>MG-Tetra-hpc</b> code (formerly tepal)
which is the parallel implementation of MG-Tetra (formerly TetMesh-GHS3D) algorithm. 
This algorithm is a DISTENE commercial software, its use requires a license.
\n
See http://www.distene.com and http://www.meshgems.com/volume-meshing-meshgems-tetra.html.
\n MG-Tetra-hpc (Tepal V3 in fact) gives the possibility to generate a partitioned
mesh with more than 200 million tetrahedrons on computers using MPI.
The launch of this version is described below.
\n This is a serious alternative to MG-Tetra, which requires a much less common
configuration with 64Go RAM to only try to make a partition of a mesh with
200 million tetrahedrons, no result guaranteed (in 2010).
\n
\note The plug-in doesn't load in the memory the supposedly large resulting meshes. 
The meshes are saved in MED files and can be imported in the user-defined location via menu File-Import-MED Files.
\n Pay attention, that Salome GUI needs 2Go RAM to load a MED
file with 5 million tetrahedrons.

\image html ghs3dprl_parameters_basic.png

<ul>
<li>
<b>Name</b> - allows to define the name of the hypothesis (MG-Tetra Parallel Parameters by default).
</li>
<li>
<b>MED Name</b> - allows to define the path and the prefix of the 
resulting MED files ("DOMAIN" by default).
If the path is not defined, the environment variable $SALOME_TMP_DIR
is used. If $SALOME_TMP_DIR is not defined as well, the environment
variable $TMP is used.
</li>
<li>
<b>Nb Partitions</b> - allows to define the number of generated MED files.
The initial skin (triangles) will be meshed (tetrahedrons) and partitioned 
in Nb_Part by the elementary algorithm implemented in Tepal.<br>
</li>
<li>
<b>Keep Files</b> - if this box is checked, input files of MG-Tetra-hpc
(GHS3DPRL.points and GHS3DPRL.faces) are not deleted after use (if the
background mode was not used).
</li>
<li>
<b>Tetra_hpc in Background</b> - if this box is checked, MG-Tetra-hpc execution
and MED file generation are launched in background mode and the user
can even exit Salome. Pay attention that in this case MG-Tetra-hpc algorithm works
independently of "killSalome.py", and sometimes on another host.
</li>
<li>
<b>Tetra_hpc Multithread</b> - if this box is checked, MG-Tetra-hpc execution
is with mg_tetra_hpc.exe (multithread version), else mg_tetra_hpc_mpi.exe (MPI distributed version).
</li>
<li>
<b>Merge subdomains</b> - if this box is checked, merge the sub-domains 
into one mesh and write the output .mesh(b).
</li>
<li>
<b>Tag subdomains</b> - if this box is checked, use the parallel sub-domain 
index as tag into the merged output mesh or not (used in combination with the 
<b>Merge subdomains</b> option).
</li>
<li>
<b>Output interfaces</b> - if this box is checked, write the parallel
sub-domains interface triangles into the merged output mesh (used in
combination with the <b>Merge subdomains</b> option).
</li>
<li>
<b>Discard subdomains</b> - if this box is checked, discard the parallel sub-domains 
(mesh, global numbering and interfaces).
</li>

<h1>Modifying MG-Tetra-hpc Advanced Parameters</h1><br>
MG-Tetra Parallel plug-in launches a standalone binary
executable <b>tetrahpc2med</b>.<br>
tetrahpc2med launches MG_Tetra-hpc, waits for the end of computation, and
converts the resulting output files into MED files.<br>
Some advanced optional parameters are accessible as arguments.<br>

If <b>Keep Files</b> option is checked, it is possible to re-launch 
\a tetrahpc2med or MG-Tetra-hpc in the Terminal as a command with
custom parameters.<br> 

<li>
<b>Advanced tetrahpc2med Parameters</b> - type <b>tetrahpc2med --help</b> in the Terminal. <p>

\verbatim
myname@myhost > /export/home/myname/salome_7/GHS3DPRLPLUGIN/bin/salome/tetrahpc2med --help
tetrahpc2med V3.0 (MED3+tetra-hpc) Available options:
   --help               : produces this help message
   --casename           : path and name of input tetrahpc2med files which are
                           - output files of GHS3DPRL_Plugin .mesh
                           - output file of GHS3DPRL_Plugin casename_skin.med (optional)
                          with initial skin and its initial groups
   --number             : number of partitions
   --medname            : path and name of output MED files
   --limitswap          : max size of working cpu memory (Mo) (before swapping on .temp files)
   --verbose            : trace of execution (0->6)
   --test               : more tests about joints, before generation of output files
   --menu               : a GUI menu for option number
   --launchtetra        : also launch tetra-hpc on files casename.mesh and option number
   --merge_subdomains   : merge the subdomains into one mesh and write the output .mesh(b) file
   --tag_subdomains     : use the parallel subdomain index as tag into the merged output mesh
                            to identify the parallel subdomains (used in combination with the merge_subdomains option)
   --output_interfaces  : write the parallel subdomains interface triangles into the merged output mesh
                            (used in combination with the merge_subdomains option)
   --discard_subdomains : discard the parallel subdomains informations output (mesh, global numbering and interfaces)
   --background         : force background mode from launch tetra-hpc and generation of final MED files (big meshes)
   --deletegroups       : regular expression (see QRegExp) which matches unwanted groups in final MED files
                            (try --deletegroups="(\bJOINT)"
                            (try --deletegroups="(\bAll_Nodes|\bAll_Faces)"
                            (try --deletegroups="((\bAll_|\bNew_)(N|F|T))"
example:
   tetrahpc2med --casename=/tmp/GHS3DPRL --number=2 --medname=DOMAIN --limitswap=1000 --verbose=0 --test=yes --menu=no --launchtetra=no

\endverbatim
\n
</li>
<li>
<b>Advanced tetra_hpc parameters (2014)</b> <p>

\verbatim

Usage: tetra_hpc.exe [options]

Options: 

     Short option (if it exists)
    /    Long option
   |    /   Description
   |   |   /
   v   v  v

     --help
          print this help

     --in <input mesh file name>
          Sets the input file
          (MANDATORY)

     --out <output mesh file name>
          Sets the output file
          (MANDATORY)

     --merge_subdomains <merge>
          Describes whether to merge the subdomains into one mesh and write the
          output .mesh(b) file or not.
          if <merge> is
             yes : the subdomains will be merged into one mesh and written to
          the output .mesh(b),
             no : the subdomains will not be merged.
          default : no

     --tag_subdomains <tag>
          Describes whether to use the parallel subdomain index as tag into the
          merged output mesh or not (used in combination with the
          merge_subdomains option).
          if <tag> is
             yes : the tags of the tetrahedra in the merged output will
          identify the parallel subdomains,
             no : the tag will keep its standard meaning of volume domain.
          default : no

     --output_interfaces <output_interfaces>
          Describes whether to write the parallel subdomains interface
          triangles into the merged output mesh or not (used in combination
          with the merge_subdomains option).
          if <output_interfaces> is
             yes : the parallel subdomains interface triangles will be written
          into the merged output mesh,
             no : they will not be added to the merged output mesh.
          default : no

     --verbose <verbose>
          Set the verbosity level, increasing from 0 to 10.
           <verbose> values are increasing from 0 to 10 :
             0 : no details
            10 : very detailed
          default : 3

     --discard_subdomains <discard>
          Describes whether to discard the parallel subdomains (mesh, global
          numbering and interfaces) or not.
          if <discard> is
             yes : the subdomain informations (mesh, global numbering and
          interfaces) will be discarded,
             no : they will be written to disk as output.
          default : no

\endverbatim
\n
</li>

<h1>Saving user's preferred MG-Tetra Parallel Advanced Parameters</h1><br>
MG-Tetra Parallel plug-in launches standalone binary executable tetrahpc2med.<br>
You may rename file tetrahpc2med as tetrahpc2med.exe for example, and replace
tetrahpc2med by a shell script at your convenience to overriding parameters.
<br>... or else $PATH modification... .<br>
<li>
<b>Advanced tetrahpc2med Parameters</b> - overriding parameter deletegroups<p>
You may rename tetrahpc2med as tetrahpc2med.exe for example.

\code
#!/bin/bash
#script tetrahpc2med overriding parameter deletegroups
#we have renamed binary executable tetrahpc2med as tetrahpc2med.exe
#echo tetrahpc2med initial parameters are $1 $2 $3 $4 ... or $*
#$0 is ignored

tetrahpc2med.exe $* --deletegroups="(\bAll_Nodes|\bAll_Faces)"

\endcode
\n
</li>
 
<h1>tetra_hpc and MPI use.</h1><br>
This 2014 beta-version needs MPI, (openmpi was used). To use it you have to proceed as below.

<li>
<b>Obsolete example tepal_v2_mpirun.</b><p>

\code

#!/bin/bash
#script tepal overriding launching Tepal_V2.0 with MPI (tepal run 64 bits only).
#we have renamed binary executable tepal as tepal64_v2.exe.
#typical command to launch tepal v1 :
#tepal -f /tmp/myname/GHS3DPRL -n 16 > /tmp/myname/tepal.log
#this file is an example to transform this call for tepal v2.0, 
#   (beta version using .mesh input file)
#you have to adapt for your convenience.

#first problem  is convert v1 input files GHS3DPRL.faces and GHS3DPRL.points 
#               to v2 input file GHS3DPRL.mesh.
#second problem is to launch on heterogeneous system linux cluster of 
#               2 hosts (64 bits) of 8 nodes (by example)
#               with different 2 executables codes linked on 2 different
#               openmpi shared library codes.
#third problem  is convert tepal v2 output files GHS3DPRL*.mesh
#               to v1 input files GHS3DPRL*.faces an GHS3DPRL*.points.

#you have to work on the same physical disk and same path input and output files : $SAME_DIR
#you have to work on different physical disk but same path and name for executable files 
#    (and shared libraries) : $DIFF_DIR

echo "parameter 0="$0
echo "parameter 1="$1
echo "parameter 2="$2
echo "parameter 3="$3
echo "parameter 4="$4

export SAME_DIR=/same_physical_disk_and_same path/tmp
export DIFF_DIR=/different_physical_disk_but_same path/myname

#copy input local files from local current directory (something like /tmp/myname)
#in this case we need /tmp/myname and $SAME_DIR different
cd $SAME_DIR
rm *
cp $2* .

export IN_FILES=`basename $2`
export IN_DIR=`dirname $2`
#created .mesh from .faces et .points
/through_salome_path/facespoints2mesh.py $IN_FILES

#there are 2 executable openmpi and library through 2 physical DIFF_DIR
export PATH=$DIFF_DIR/openmpi-1.3.1_install/bin:${PATH}
export LD_LIBRARY_PATH=$DIFF_DIR/openmpi-1.3.1_install/lib:${LD_LIBRARY_PATH}

#there are 2 executables tepal_v2 through 2 physical DIFF_DIR
export LD_LIBRARY_PATH=$DIFF_DIR/tepal-2.0.0/bin/Linux_64:${LD_LIBRARY_PATH}
export PATH=$DIFF_DIR/tepal-2.0.0/bin/Linux_64:$PATH

#small test between friends
#rm hostnames.log
#mpirun -n $4 hostname >> hostnames.log

#there necessary set env licence file for tepal v2
export DISTENE_LICENSE_FILE="Use global envvar: DLIM8VAR"
export DLIM8VAR="dlim8 1:1:29030@is142356/0016175ef08c::a1ba...9e19"
export SIMULOGD_LICENSE_FILE=29029@is142356 
export LICENSE_FILE=/product/distene/dlim8.var.sh

#mpirun with necessary set environment
export TMP_ENV="-x PATH -x LD_LIBRARY_PATH -x DISTENE_LICENSE_FILE -x DLIM8VAR \
                -x SIMULOGD_LICENSE_FILE -x LICENSE_FILE"
#mpirun $TMPENV -n $4 which tepal64_v2.exe >> hostnames.log

#real mpirun uncomment after verify small test
mpirun $TMPENV -n $4 tepal64_v2.exe --in $IN_FILES.mesh --out $IN_FILES.mesh --verbose 100

#convert output files tepalv1 format
/through_salome_path/mesh2facespoints.py $IN_FILES

#copy output files from $SAME_DIR to local current directory (something like /tmp/myname)
cp -f hostnames.log $IN_DIR
cp -f $IN_FILES* $IN_DIR

#ls -al $SAME_DIR
#cat $SAME_DIR/hostnames.log
#cat /tmp/myname/tepal.log

\endcode
\n
</li>

<h1>TUI use.</h1><br>

<li>
<b>example ex30_tepal.py.</b><p>

\code

#!/bin/python
import os

import GEOM
from salome.geom import geomBuilder
geompy = geomBuilder.New(salome.myStudy)

import SMESH
from salome.smesh import smeshBuilder
smesh = smeshBuilder.New(salome.myStudy)

# Parameters
# ----------

results = "/tmp/ZZ"

radius =  50
height = 200

# Build a cylinder
# ----------------

base = geompy.MakeVertex(0, 0, 0)
direction = geompy.MakeVectorDXDYDZ(0, 0, 1)

cylinder = geompy.MakeCylinder(base, direction, radius, height)

geompy.addToStudy(cylinder, "Cylinder")

# Define a mesh on a geometry
# ---------------------------

m = smesh.Mesh(cylinder)

# 2D mesh with BLSURF
# -------------------

algo2d = m.Triangle(smeshBuilder.BLSURF)

algo2d.SetPhysicalMesh(1)
algo2d.SetPhySize(5)

algo2d.SetGeometricMesh(0)

# 3D mesh with tetra-hpc (formerly tepal v3 (2014))
# ----------------------------------------------------

algo3d = m.Tetrahedron(smeshBuilder.MG_Tetra_Parallel)

algo3d.SetMEDName(results)
algo3d.SetNbPart(4)
algo3d.SetBackground(False)
algo3d.SetMultithread(False)
algo3d.SetKeepFiles(False)
algo3d.SetGradation(1.05)
algo3d.SetMinSize(0)
algo3d.SetMaxSize(0)


# Launch meshers
# --------------

status = m.Compute()

# Test if ok
# ----------

if os.access(results+".xml", os.F_OK):
    print "Ok: tetra_hpc"
else:
    print "KO: tetra_hpc"
\endcode
\n
</li>
</ul>


*/