[cig-commits] [commit] devel, master: updated README files for examples. also added go_decomposer_pbs.bash to remind user not to run huge partition jobs on the head node, but rather on a (high-memory) compute node (7fbec60)

cig_noreply at geodynamics.org cig_noreply at geodynamics.org
Wed Jun 18 16:48:14 PDT 2014


Repository : https://github.com/geodynamics/specfem3d

On branches: devel,master
Link       : https://github.com/geodynamics/specfem3d/compare/6026e367984905ab133865f62fa6293b343759b9...47f703851338234f96397e7da9fbff63d8178b8a

>---------------------------------------------------------------

commit 7fbec602590ccacd6a5ca4adc82eb6ce08de6886
Author: Carl Tape <carltape at gi.alaska.edu>
Date:   Tue Feb 1 07:00:23 2011 +0000

    updated README files for examples. also added go_decomposer_pbs.bash to remind user not to run huge partition jobs on the head node, but rather on a (high-memory) compute node


>---------------------------------------------------------------

7fbec602590ccacd6a5ca4adc82eb6ce08de6886
 homogeneous_halfspace/README | 18 ++++++++------
 layered_halfspace/README     | 56 ++++++++++++++++++++++++++++++++------------
 tomographic_model/README     |  3 +++
 3 files changed, 55 insertions(+), 22 deletions(-)

diff --git a/homogeneous_halfspace/README b/homogeneous_halfspace/README
index 7da7664..329452a 100644
--- a/homogeneous_halfspace/README
+++ b/homogeneous_halfspace/README
@@ -22,7 +22,7 @@ step-by-step tutorial:
      If successful, this will generate the files in src/shared/:
      Makefile, constants.h, and precision.h, among others
 
-     NOTE: the same compiler used for SCOTCH must be used for SPECFEM3D (ifort or gfortran)
+     note: the same compiler used for SCOTCH must be used for SPECFEM3D (ifort or gfortran)
 
    - copy two run scripts from utils/Cluster/ into SPECFEM3D/, e.g.,
      pbs/go_generate_databases_pbs.bash
@@ -40,7 +40,7 @@ step-by-step tutorial:
      then run meshing script:
      claro -> Menu "Tools" -> "Play Journal File" ... and select file: "block_mesh.py"
 
-     if everything goes fine, this creates all the mesh files in a subdirectory MESH/:
+     if everything goes fine, this creates the ten mesh files in subdirectory MESH/:
         MESH/absorbing_surface_file_bottom
         MESH/absorbing_surface_file_xmax
         MESH/absorbing_surface_file_xmin
@@ -71,18 +71,22 @@ step-by-step tutorial:
      (for your convenience, these resulting database files are also provided in the example's
       DATABASES-default/ directory for comparison.)
 
+     note 1: the program xdecompose_mesh_SCOTCH is a serial program (runs on 1 core)
+     note 2: alternatively, run decomposer on a compute node using
+             utils/Cluster/pbs/go_partition_pbs.bash
+
 
 4. generate databases:
 
-   - compile generate_databases from SPECFEM3D/ :
-     > cd SPECFEM3D
+   - compile generate_databases in directory SPECFEM3D/ :
      > make xgenerate_databases
 
    - submit job script
      > qsub go_generate_databases_pbs.bash
 
-     NOTE: this script will need to be tailored to your cluster, e.g.,
+     note 1: this script will need to be tailored to your cluster, e.g.,
      > bsub < go_generate_databases_lsf.bash
+     note 2: you may want to explicity specify the desired queue
 
      this will create binary mesh files, e.g. "proc000***_external_mesh.bin"
      in directory in_out_files/DATABASES_MPI/.
@@ -96,9 +100,9 @@ step-by-step tutorial:
    - submit job script:
      > qsub go_solver_pbs.bash
 
-     NOTE 1: this script will need to be tailored to your cluster, e.g.,
+     note 1: this script will need to be tailored to your cluster, e.g.,
              > bsub < go_solver_lsf.bash
-     NOTE 2: the simulation runs on 4 cores and should take about 5 minutes,
+     note 2: the simulation runs on 4 cores and should take about 5 minutes,
              and you can track the progress with the timestamp files
              generated in in_out_files/OUTPUT_FILES/
 
diff --git a/layered_halfspace/README b/layered_halfspace/README
index ba8817f..6a08efb 100644
--- a/layered_halfspace/README
+++ b/layered_halfspace/README
@@ -20,32 +20,54 @@ step-by-step tutorial:
      claro -> Menu "Tools" -> "Play Journal File" ... and select file: "2lay_mesh_boundary_fig8.py"
 
      (or select: "2lay_mesh_boundary_fig8-nodoubling.py" for regular mesh without doubling layer)
-
+open read_cubit_
      this creates all the mesh files in subdirectory MESH/
 
 
-  - optional:
-        if you want to check mesh quality (element skewdness, stability criteria)
-        using check_mesh_quality_CUBIT_Abaqus/check_mesh_quality_CUBIT_Abaqus.f90,
-        then Export to Abacus (*.inp) file:
-        claro -> Menu "File" -> "Export..." and select Abaqus (*.inp)
+  - optional: check mesh quality using check_mesh_quality_CUBIT_Abaqus.f90
+      
+      using the CUBIT GUI: 
+        claro -> Menu "File" -> "Export...", select Abaqus (*.inp)
 
         e.g. export as "layered_halfspace_mesh.inp"
 
-        note: only output blocks for volumes, i.e. block ids 1,2,3 for this example here,
-              not for faces
+        note 1: only output blocks for volumes, i.e. block ids 1,2,3 for this example here
+        note 2: do NOT select "Export Using Cubit IDs"
+
+      alternatively, using the CUBIT command line:
+        > export Abaqus "layered_halfspace_mesh.inp" Block 1,2,3 overwrite cubitids 
+
+      change directory
+        > cd ../..//src/check_mesh_quality_CUBIT_Abaqus/
 
-        save number of nodes and number of elements and edit in header of
-        check_mesh_quality_CUBIT_Abaqus.f90:
+      save number of nodes and number of elements and edit in header of
+      check_mesh_quality_CUBIT_Abaqus.f90:
 
             character(len=100), parameter :: cubit_mesh_file = 'layered_halfspace_mesh.inp'
-            integer, parameter :: NPOIN = 44436                   ! number of nodes
-            integer, parameter :: NSPEC = 40500                   ! number of elements (only volumes, i.e. block ids 1,2,3 )
-            integer, parameter :: NGNOD = 8                        ! hexahedral elements
+            integer, parameter :: NPOIN = 76819    ! number of nodes
+            integer, parameter :: NSPEC = 70200    ! number of elements (only volumes, i.e. block ids 1,2,3 )
+            integer, parameter :: NGNOD = 8        ! hexahedral elements
             logical, parameter :: IGNORE_OTHER_HEADERS = .false.
             double precision, parameter :: delta_t = 0.005         ! arbitrary, initial guess
             double precision, parameter :: VP_MAX = 7500.d0        ! maximum vp in volume block id 3
 
+      link Abaqus file
+        > ln -s ../../examples/layered_halfspace/layered_halfspace_mesh.inp .
+
+      compile programs
+        > make
+
+      run program
+        > ../../bin/xcheck_mesh_quality_CUBIT_Abaqus
+      at prompts, enter, for example, 1, then 0.6 to list elements with high skewness
+
+      plot skewness histogram as ps file
+        > gnuplot
+        > load "plot_mesh_quality_histogram.gnu"
+        > set term postscript color solid
+        > set output 'hist_skewness.ps'
+        > replot
+        > quit
 
 2. decompose mesh files:
 
@@ -61,7 +83,7 @@ step-by-step tutorial:
 
 3. generate databases:
 
-    - copy Par_file, CMTSOLUTION & STATIONS files provided in this examples directory to SPECFEM3D/in_data_files/
+    - copy example files Par_file, CMTSOLUTION & STATIONS to SPECFEM3D/in_data_files/
       > cd examples/layered_halfspace/in_data_files
       > cp CMTSOLUTION Par_file STATIONS ../../../in_data_files/
       > cd SPECFEM3D
@@ -72,9 +94,13 @@ step-by-step tutorial:
     - submit job script:
       > qsub go_generate_databases_pbs.bash
 
-    (note: if execution fails due to memory shortage - most likely fails when calling routine to
+      this will create binary mesh files, e.g. "proc000***_external_mesh.bin"
+      in directory in_out_files/DATABASES_MPI/.
+      (note: if execution fails due to memory shortage - most likely fails when calling routine to
            create regional mesh - then try to increase process memory stack size: ulimit -s 2000000  (2GB) )
 
+      optional: load a set of .vtk files into paraview to view model, mesh, and partitioning
+      (note: must have Par_file option SAVE_MESH_FILES = .true.)
 
 4. run simulation:
 
diff --git a/tomographic_model/README b/tomographic_model/README
index 1bf492b..0ab7dd7 100644
--- a/tomographic_model/README
+++ b/tomographic_model/README
@@ -34,6 +34,9 @@ step-by-step tutorial:
 
    - copy this file into directory SPECFEM3D/in_data_files/
      > cp tomography_model.xyz ../../in_data_files/
+     note: this file is read in by subroutine read_model_tomography in
+        SPECFEM3D/src/generate_databases/model_tomography.f90,
+        which is called by get_model.f90
 
    - copy Par_file, CMTSOLUTION & STATIONS files provided in this examples directory to SPECFEM3D/in_data_files/
      > cd in_data_files/



More information about the CIG-COMMITS mailing list