Changeset 9265

Timestamp:

Jul 14, 2014, 4:48:21 PM (11 years ago)

Author:

steve

Message:

Moving checkpoint.py to shallow_water

Location:

trunk/anuga_core

Files:

• source/anuga/config.py (modified) (1 diff)
• source/anuga/shallow_water/checkpoint.py (moved) (moved from trunk/anuga_core/source/anuga_parallel/checkpoint.py)
• user_manual/source/anuga_user_manual.tex (modified) (11 diffs)
• user_manual/source/manual_example.tex (modified) (1 diff)
• validation_tests/case_studies/towradgi/Compare_results_with_fieldObs.py (modified) (1 diff)

trunk/anuga_core/source/anuga/config.py

@@ -34 +34 @@
 # Major revision number for use with create_distribution
 # and update_anuga_user_guide
-major_revision = '1.3.0-beta'
+major_revision = '1.3.1'
 
 ################################################################################

trunk/anuga_core/user_manual/source/anuga_user_manual.tex

@@ -29 +29 @@
 \usepackage{graphicx}
 \usepackage{hyperref}
-\usepackage[english]{babel}
+\usepackage[australian]{babel}
 \usepackage{datetime}
 \usepackage[hang,small,bf]{caption}
@@ -246 +246 @@
   \item All spatial coordinates are assumed to be UTM (meters). As such,
   \anuga is unsuitable for modelling flows in areas larger than one UTM zone
-  (6 degrees wide).
+  (6 degrees wide), though we have run over 2 zones by projecting onto one zone and 
+  living with the distortion.
   \item Fluid is assumed to be inviscid -- i.e.\ no kinematic viscosity included.
   \item The finite volume is a very robust and flexible numerical technique,
@@ -275 +276 @@
 
 What follows is a discussion of the structure and operation of a
-script called \file{runup.py}.
+script called \file{runup.py} (which is available in the \file{demos} directory 
+of \file{anuga_core}.
 
 This example carries out the solution of the shallow-water wave
@@ -331 +333 @@
 
 \label{ref:runup_py_code}
-\verbatiminput{demos/runup.py}
+\verbatiminput{../../demos/runup.py}
+
 
 \subsection{Establishing the Domain}\index{domain, establishing}
 
-The first task is to set up the triangular mesh to be used for the
+The very first thing to do is import the various modules, of which the 
+\anuga{} module is the most important. 
+%
+\begin{verbatim}
+import anuga
+\end{verbatim}
+%
+Then we need to set up the triangular mesh to be used for the
 scenario. This is carried out through the statement:
 
@@ -341 +351 @@
 domain = anuga.rectangular_cross_domain(10, 5, len1=10.0, len2=5.0) 
 \end{verbatim}
-
+%
 The above assignment sets up a $10 \times
 5$ rectangular mesh, triangulated in a regular way with boundary tags \code{'left'}, \code{'right'},
@@ -350 +360 @@
 domain = anuga.Domain(points, vertices, boundary)
 \end{verbatim}
-%
 where
 \begin{itemize}
@@ -909 +918 @@
 Here is the code for \file{runcairns.py}:
 
-\verbatiminput{demos/cairns/runcairns.py}
+\verbatiminput{../../demos/cairns/runcairns.py}
 
 In discussing the details of this example, we follow the outline
@@ -970 +979 @@
 \file{project.py}:
 
-\verbatiminput{demos/cairns/project.py}
+\verbatiminput{../../demos/cairns/project.py}
 
 Figure \ref{fig:cairns3d} illustrates the landscape of the region
@@ -1326 +1335 @@
 \end{figure}
 
+
+\section{A Parallel Simulation}
+
+The previous examples were run just using one processor. \anuga also has the option of running in parallel using \mpi.
+
+Such jobs are run using the command
+
+\begin{verbatim}
+mpirun -np n python runParallelCairns.py
+\end{verbatim}
+where \code{n} is the total number of processors being used for the parallel run. 
+
+Essentially we can expect speedups comparable to the number of cores available. This is measured via scalablity. We can expect scalability of 70\% when using up to the number of processors so that the local partitioned domains contain around 2000 triangles. 
+
+\subsection{The Code}
+
+Here is the code for \file{runParallelCairns.py}:
+
+\verbatiminput{../../demos/cairns/runParallelCairns.py}
+
+\subsection{Structure of the Code}
+
+The code is very similar to the sequential code. The same procedures are used to setup the domain, setup the inital conditions, boundary conditions and evolve. 
+
+
+We first import a few procedures  need forthe parallel code.
+\begin{verbatim}
+from anuga import distribute, myid, numprocs, finalize, barrier
+\end{verbatim}
+
+\code{myid} returns the id of the current processor running the code. 
+
+\code{numprocs} returns the total number of processors involved in this parallel jobs (the \code{n} in the original \code{mpirun} command. 
+
+\code{finalize} is called at the end of a script to close down the parallel job.
+
+\code{distribute} is used to partition and setup the parallel domains.
+
+\code{barrier} is a command for processors to wait as all the other processor to catchup to this point.
+
+
+
+
+The creation of the \code{domain} is only done on processor 0. Hence we have the structure:
+
+\begin{verbatim}
+#-------------------------------------
+# Do the domain creation on processor 0
+#-------------------------------------
+if myid == 0: 
+        ....
+        domain = ...
+        
+else:
+        domain = None
+\end{verbatim}
+
+We only need to create the original domain on one processor, otherwise we will have multiple copies of the full domain (which will easily eat up our memory). 
+
+Once we have our code{domain} setup we partition it and send the partitions to each of the other processors, via the command
+
+\begin{verbatim}
+#-------------------------------------
+# Now produce paralle domain
+#-------------------------------------
+domain = distribute(domain)
+\end{verbatim}
+
+This takes the \code{domain} on processor 0 and distributes that domain to each of the processors, (it overwrites the full domain on processor 0).  From this point of the code, there is a different domain on each processor, with each domain comunicating with the other domains to ensure required transfer of inforation to allow flow over the combined domains. 
+
+It is important to apply the boundary conditions after the \code{distribute}
+
+
+
+As the partitoined domain evolve, they will store their data to individual sww files, named as \code{domain_name_Pn_m.sww}, where \code{n} is the total number of processors being used and \code{m} is the specific processor id. 
+
+We have a procedure to merge these individual sww files via the command 
+
+\begin{verbatim}
+domain.sww_merge()
+\end{verbatim}
+
+And we close down the parallel job by issuing the command 
+
+\begin{verbatim}
+finalize()
+\end{verbatim}
+
+
+
+\section{Validation Tests}
+
+We have an extensive suite of validation tests, ranging from simple analytical tests to large scale case studies. These can be found in the \file{validation_tests} directory in the \file{anuga_core} directory. 
+
+All these tests can be run in parallel or in sequential mode, and they also provide the mechanism to produce a report detailing the results of the test. (pdfLatex needs to be available to produce the reports). 
+
+
+
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
@@ -1396 +1503 @@
 use in Python code. For example, suppose we wish to specify that the
 function \function{create\_mesh\_from\_regions} is in a module called
-\module{mesh\_interface} in a subfolder of \module{inundation} called
+\module{mesh\_interface} in a subfolder of \module{anuga} called
 \code{pmesh}. In Linux or Unix syntax, the pathname of the file
-containing the function, relative to \file{inundation}, would be:
+containing the function, relative to \file{anuga}, would be:
 
 \begin{verbatim}
@@ -4217 +4324 @@
 %\begin{center}
 %\pgfplotstableset{% could be used in preamble
-%empty cells with={--}, % replace empty cells with ’--’
+%empty cells with={--}, % replace empty cells with ￜ--ￜ
 %every head row/.style={%
 %   before row={%

trunk/anuga_core/user_manual/source/manual_example.tex

@@ -7 +7 @@
 
 \documentclass{manual}
+
+
+\usepackage{graphicx}
+\usepackage{hyperref}
+%\usepackage[english]{babel}
+\usepackage{datetime}
+\usepackage[hang,small,bf]{caption}
+\usepackage{amsbsy,enumerate}
+
+\usepackage{amsmath, amssymb, amsthm}
 
 \title{Big Python Manual}

trunk/anuga_core/validation_tests/case_studies/towradgi/Compare_results_with_fieldObs.py

@@ -64 +64 @@
                       CellSize=CellSize,EPSG_CODE=32756,output_dir=tif_outdir)
     print 'Made tifs'
-    # Plot depth raster with discrepency between model and data
-    depthFile=tif_outdir+'/Towradgi_historic_flood_depth_max_max.tif'
-    myDepth=scipy.misc.imread(depthFile)
-    X=scipy.arange(p.xllcorner, p.xllcorner+myDepth.shape[1]*CellSize, CellSize)
-    Y=scipy.arange(p.yllcorner, p.yllcorner+myDepth.shape[0]*CellSize, CellSize)
-    X,Y=scipy.meshgrid(X,Y)
-    pyplot.clf()
-    pyplot.figure(figsize=(12,6))
-    pyplot.plot([X.min(),X.max()],[Y.min(),Y.max()],' ')
-    pyplot.imshow(scipy.flipud(myDepth),extent=[X.min(),X.max(),Y.min(),Y.max()],origin='lower',cmap=pyplot.get_cmap('Greys'))
-    pyplot.gca().set_aspect('equal')
-    pyplot.colorbar(orientation='horizontal').set_label('Peak Depth in model (m)')
-    er1=floodLevels[:,3]-modelled_level
-    pyplot.scatter(floodLevels[:,0], floodLevels[:,1], c=er1,s=20,cmap=pyplot.get_cmap('spectral'))
-    pyplot.colorbar().set_label(label='Field observation - Modelled Peak Stage (m)')
-    pyplot.xlim([p.x.min()+p.xllcorner,p.x.max()+p.xllcorner])
-    pyplot.ylim([p.y.min()+p.yllcorner,p.y.max()+p.yllcorner])
-    pyplot.savefig('Spatial_Depth_and_Error.png')
 except:
     print 'Cannot make GIS plot -- perhaps GDAL etc are not installed?'
+
+    
+# Plot depth raster with discrepency between model and data
+depthFile=tif_outdir+'/Towradgi_historic_flood_depth_max.tif'
+myDepth=scipy.misc.imread(depthFile)
+X=scipy.arange(p.xllcorner, p.xllcorner+myDepth.shape[1]*CellSize, CellSize)
+Y=scipy.arange(p.yllcorner, p.yllcorner+myDepth.shape[0]*CellSize, CellSize)
+X,Y=scipy.meshgrid(X,Y)
+pyplot.clf()
+pyplot.figure(figsize=(12,6))
+pyplot.plot([X.min(),X.max()],[Y.min(),Y.max()],' ')
+pyplot.imshow(scipy.flipud(myDepth),extent=[X.min(),X.max(),Y.min(),Y.max()],origin='lower',cmap=pyplot.get_cmap('Greys'))
+pyplot.gca().set_aspect('equal')
+pyplot.colorbar(orientation='horizontal').set_label('Peak Depth in model (m)')
+er1=floodLevels[:,3]-modelled_level
+pyplot.scatter(floodLevels[:,0], floodLevels[:,1], c=er1,s=20,cmap=pyplot.get_cmap('spectral'))
+pyplot.colorbar().set_label(label='Field observation - Modelled Peak Stage (m)')
+pyplot.xlim([p.x.min()+p.xllcorner,p.x.max()+p.xllcorner])
+pyplot.ylim([p.y.min()+p.yllcorner,p.y.max()+p.yllcorner])
+pyplot.savefig('Spatial_Depth_and_Error.png')
+