ChangeLog

********************************************************************************
                            ChangeLog File
********************************************************************************

Please add at the top of this file, a description of the latests changes
you have made to the source code.  Please follow the following format:

YYYY/MM/DD Author name <email@domain.foo>

	* filename_1: 
	
	  General description of changes done in filename_1. Reaching column 80
	  go to the next line please. The star separation lines have exactly
	  80 stars to help you with this.

	* filename_2 (function_A):

	  General description of changes done in filename_2, assocaited with 
	  function_A.

	* filename_2 (function_B):

	- Itemized description of changes in function_B of filename_1.

	- Continued itemized description of changes in function_B of filename_1.

	* filename_3:

	- Itemized description of changes done in filename_3.

	- Continued itemized description of changes done in filename_3.

	(function_C):
	
	- Itemized description of changes to function_C in filename_3.

	(function_D):
	
	- Itemized description of changes to multiple functions in filename_3. 
	
	
********************************************************************************
	
2012-07-14 Yigit Isbiliroglu <yisbilir@andrew.cmu.edu>

	A major bug which crashes program(building module) with the increaased 
	number of processors is resolved with this version. Other small fixes are
	applied to buildings module.
		
	* quake/forward/buildings.c:
		
	(get_airprops):
	
	- x,y and z coordinates of the air elements are adjusted wrt shifted origin
	(if any).
	
	* quake/forward/drm.c:
	
	(searchBucketList):
	
	- return is included to avoid warning from the compiler.
	
	* quake/forward/meshformatlab.c:
	
	(saveMeshCoordinatesForMatlab):
	
	- changes to  print the damping ratios .(if wanted)
	
	* quake/forward/psolve.c:
		
	(setrec):
	
	- new parameters are fed into bldgs_setrec function.
	
	(main):
	
	- new parameters are fed into saveMeshCoordinatesForMatlab function.
	
	(mesh_generate):
	
	- second pass to octor_carvebuildings is eliminated. Since timer still
	exists it is not deleted completely.
	
	* octor.c:
	
	(oct_getnextleaf):
	
	- second if statement is modified in order not to go to octants that I dont
	possess.
	
	(oct_installleaf):
	
	- second if statement is modified and another line is added in order not to
	get a 'false next leaf' in  oct_getnextleaf.
	
	(node_setproperty):
	
	- a bug is fixed for touches == 2 case(2).
	
	(octor_carvebuildings):
	
	- built-in function oct_shrink is now used to eliminate the air elements.
	- an unneccesary call to tree_setcom is deleted.
	
	(octor_partitiontree):
	
	- another variable is added (irecv).
	- PIs with 0 elements(due to carving) are not included in the MPI 
	communications. For this, another for loop is included(line 4682). Also
	if statement(line 4723) is modified.
	
	- NOTE: all changes related to fix are tagged by the name 'yigit'.
	
	
********************************************************************************

2012-06-19 Dorian Restrepo <drestrep@andrew.cmu.edu>

	All changes in this commit are related to new alternative for rate-independent
	plasticity in the nonlinear module.

	* quake/forward/nonlinear.h:

	- Added option for rate-independent plasticity.
	- Added new structs vector_t and qpvectors_t
	- Added new variables to the structs nlsolver_t, nlconstants_t and nlstation_t
	- Added new method compute_dLambdaII

	* quake/forward/nonlinear.c:
	
	(isThisElementNonLinear):
	
	- New lower limit for nonlinear parameters.
	
	(nonlinear_init):
	
	- added broadcasting of the new variables: theNonLinVsMin, thePlasticityModel, theHardeningModulus
	
	(nonlinear_initparameters):
	
	- Added parsing lines, sanity checks and modification of previous arrays sizes for 
	  storing the three new variables mentioned in (nonlinear_init)
	
	(nonlinear_solver_init):
	
	- Memory allocation of new structs myNonlinSolver->ep1 and myNonlinSolver->ep2
	- added interpolation line for the hardening modulus 
	
	(compute_dLambda):
	
	- Added module for the calculation of the plastic multiplier according with rate-independant theory.

	(compute_dLambdaII):
	
	- This is an updated version of (compute_dlambda) and the one that currently is been used for the nonlinear module. Its -
	  expression for the material update is much more simple, easy and accurate than the one used in (compute_dlambda). In the-
      future the previous function (compute_dLambda) should be deprecated.
      
    (tensor_t compute_pstrain2):
    
    - added lines for the updated plastic strains corresponding to rate-independant theory
    
    (compute_nonlinear_state):
    
    - added module for the material update algorithm corresponding to rate-independant theory.
    - line: if ( ( theMaterialModel == LINEAR ) || ( J2 == 0 ) ); was replaced by: if ( ( theMaterialModel == LINEAR ) )
    
    (nonlinear_stations_init):
    
    - added pointer for the equivalent plastic strain.
    
    (print_nonlinear_stations):
    
    - added variables and module for the material update algorithm corresponding to rate-independant theory.
    - last column printed was changed from "k" to "k + hrd * (*ept)"
      
      
	  
	
********************************************************************************

2011-10-19 Yigit Isbiliroglu <yisbilir@andrew.cmu.edu>

	Some bugs are fixed in psolve.c. And timer statistics are added for drm.
	
	* quake/forward/psolve.c:
		
	(toexpand):
	
	- Part1 of drm also goes for drm_toexpand rule.
	
	(mesh_correct_properties):
	
	- theXForMeshOrigin, theYForMeshOrigin and theZForMeshOrigin are added to
	query points to correct the mesh properties (to north_m ,east_m and depth_m
	respectively. This fixes the bug.
	
	(solver_run_collect_timers):
	
	- Timer Functions are added here for drm.
	
	(print_timing_stat):
	
	- Similarly code is included to print the drm statistics.
	
********************************************************************************
2011-10-16 Yigit Isbiliroglu <yisbilir@andrew.cmu.edu>

	Some changes have been done to drm module.Now one can define drm_print_rate
	which makes i/o of drm coordinates faster. Also a bug in psolve.c is fixed.

	* parameters.in

	- Added lines defining the parameters for drm.

	part1_delta_t  = 0.001
	drm_print_rate = 10
	( These are also added to below for convenience.)
	
	part1_delta_t : is the deltaT used in part1. One can choose a lower deltaT 
	for part2.
		
	* quake/forward/psolve.c:
	
	- return value of open_cvmdb is changed to static void from static int.
	
	(read_parameters):
	
	- Param.theRegionLat, Param.theRegionLong and Param.theRegionDepth  are
	 broadcasted to all processors, which now  prevents the program to crash.
		
	(parse_parameters):
	
	- Definitions of unused variables localFourDOutFile,
	localcvmdb_input_file, localmesh_etree_output_file are deleted.
	
	(open_cvmdb):
	
	- Exit codes in case of an error have been fixed.
	

********************************************************************************
2011-10-07 Yigit Isbiliroglu <yisbilir@andrew.cmu.edu>

	A new feature is added to program. Now, Domain Reduction Method (DRM) is
	fully implemented in Hercules.For this, drm.c and drm.h files are created.
	It's worth it to mention that one needs to run 3 simulations related to DRM
	in order to take advantage of this method, namely Part0, Part1 and Part2.
	This method is useful if one needs to conduct a more detailed (repeated) 
	study in a small sub-region within the original big domain.In Part0,drm-box 
	boundary is defined and coordinates of the drm-nodes are stored(you only 
	need the small subregion and only let the meshing process to happen). In 
	Part1, displacements for the drm-nodes (read from the Part0) are stored at 
	each time step (this is the regular Hercules run except that you store 
	displacements at some points to be used in Part2). And finally in Part2, 
	by using the displacements stored in Part1 and converting them to effective 
	forces, one can do repeated simulations in the small subregion. So Part0
	and Part1 runs are done ONLY once, then you have everything necessary to do
	multiple Part2 runs. It is impossible to explain all the details here,but 
	i tried to explain steps within the code. 
	
	* parameters.in

	- Added lines defining the parameters for drm.
	  One of the following two options should be added to parameters.in
	
	# DRM
	implement_drm  = no
	
	OR
	
	# DRM
	implement_drm  = yes
	drm_directory  = outputfiles/DRM

	which_drm_part = part2

	#                   part0  part1 or part2
	#............................................................................

	drm_edgesize = 50.0
	
	drm_offset_x =  10000.0
	drm_offset_y =  20000.0

	part1_delta_t  = 0.001 **
	drm_print_rate = 10  **

	drm_boundary =

     	200      200     600       600      525
	#   min_x    min_y    max_x    max_y    depth
	# ............................................................................
	
	** Explanations for these two are above.
	
	If the implement_drm is no, the program will work as usual.
	
	If you want to use DRM, you need to make sure that you created part0,part1
	and part2 directories under outputfiles/DRM.
	
	drm_offset_x and y: distance between origin of the original domain and
	origin of your sub-domain(in m).

	drm_edgesize: for some reasons (to avoid dangling nodes) we force the
	drm boundary to consist of same sized elements.This is the desired element
	size for the drm elements(in m). 

	drm_boundary:limits of your drm boundary(in m). This is according to origin
	of your small subregion. This is NOT used in PART1.
		
	* quake/forward/psolve.c:
	
	- drm.h is added in header.
	
	- noyesflag_t  drmImplement is defined in Params.
	
	- drm_part_t   theDrmPart is defined in Params.
	
	(read_parameters):
	
	- Param.drmImplement variable is broadcasted to all processors.
	
	(parse_parameters):
	
	- Added char implement_drm[64].
		
	- Added noyesflag_t implementdrm.
	
	- implement_drm is parsed by PE0 from parameters.in and sanity
	  check is done to make sure it is either yes or no. Then, it is assigned
	  to implementdrm. Also printed on the screen.
	
	(mesh_generate):

	- A call to drm_fix_coordinates is added .
		
	(toexpand):
		
	- An if statement was added to check whether the drm is on or off(except for
	  Part1).If it is on and the element is in the drm boundary, it forces to 
	  have same sized elements in the drm boundary,and skips the regular Vs 
	  rule.
	
	(source_init):
		
	- An if statement was added at the very beginning of the function.It skips
	source init if this is a DRM Part0 or Part2 run.
	
	(solver_run):
	
	- Calls to solver_output_drm_nodes, solver_read_drm_displacements,
	 solver_compute_effective_drm_force and solver_drm_close are added.I think
	 names are self explanatory.
		
	(main):
	
	- An if statement is added to initialize parameters for drm if drm_implement
	is yes.
	
	- And other drm related calls are added after the meshing process.

********************************************************************************
2011-07-31 John Urbanic <urbanic@.psc.edu>

	Move planes array data out of parameter input file and into own
	file.  This is first of three such changes and will be soon
	followed by similar fixes for stations and buildings.  At that time,
	I will document the new/final input file format.

	* quake/forward/psolve.c:

	- Add new planes array filename to Params.  Passed this to
	Planes_setup() routine.

	* quake/forward/io_planes.c/h:

	(new/oldplanes_setup):

	- Added in new input parameter to pass planes filename.

	- Now open and parse this new input file instead of reading from
	general input file data.
	

********************************************************************************
2011-07-20 John Urbanic <urbanic@.psc.edu>

	Implementing new input file format and cleaning up psolve.c

	* parameters.in (parameter input file - not a real source file):

	Three previous command line arguments have been moved into the input
	file.  They are cvmdb_input_file, mesh_etree_output_file, and
	4D_output_file.  They are currently required.

	* quake/forward/psolve.c:

	- Removed all references, save one, to argv[].  Replaced with proper
	parameters.
	
	- Added last few loose parameters to Param, and removed remote
	initialization of Global variables Global.myOctree, Global.myMesh and
	Global.mySolver.
	
	(main)
	
	- Sole input argument is now the data input file.  This includes, or
	points to, all other parameters.

	(load_parameters)

	- includes old local_init, but moves out unrelated pieces.  Mostly the
	cvmdb open and create routines.  Also consolidates all parameter
	broadcasts.

	(open_cvbdb)

	- Moved all cvmdb initialization code into this.  This replaces code in
	radom other places, primarily local_init, replicateDB and OpenDB.

	(parse_parameters)

	- just a slightly modified version of old read_params to include new
	input arguments.  Slight re-arrangement to accomodate heavy
	modifications to immediately follow.

********************************************************************************	
2011-07-05 John Urbanic <urbanic@.psc.edu>

	Major changes throughout psolve.c, psolve.h, io_planes.c and ioplanes.h

	* quake/forward/psolve.c:

	All new parameters should be added to Param, and should be passed as
	individual structures. No new globals should be added to Global, or
	anywhere.  Those here should eventually be eliminated as globals.

	- All global variables, bar MPI_Comm, have been gathered into Params and
	Global structures.  They have retained their original names there.
	Initializers are set as previously.

	- External functions that required passing any of these variables have
	been changed to pass them as parameters.

	- All VIS defined real-time visualization structures were removed.

	- Some unused (completely unreferenced) variables were removed.  Some
	variables that look unused were left as they look potentially useful.

	* quake/forward/psolve.h:

	- changes to maintain consistancy with the above.

	* quake/forward/io_planes.c:

	- Eliminated all extern global variables.  All data now passed from
	outside as parameters.  Only local static private data.

	* quake/forward/io_planes.h:

	- changes to maintain consistancy with the above.


********************************************************************************
2011-05-08 Haydar Karaoglu <hkaraogl@andrew.cmu.edu>

	A new attenuation model (BKT) is implemented into Hercules.

	* quake/forward/stiffness.c
	
	* quake/forward/damping.c
	
	- New methods to incorporate BKT model.
	
	* quake/forward/psolve.c
	
	- theQTABLE is added as hard-coded. 
	- New method to read theQTABLE.
	- Changes have been made to setrec method to read and store the coefficients related to introduced model for each element.
	(Velocity profile in the model is being corrected according to the attenuation model within setrec)
	
	*quake/forward/quake_util.c
	
	- New method to search theQTABLE to get the corresponding coefficients for each element in the mesh.
	
	*parameters.in
	
	- "bkt" is a new option for the type of damping.
	- "simulation_velocity_profile_freq_hz" (The frequency for which the given velocity profile holds) should be given if bkt model is to be used, otherwise it should be given as "zero".

********************************************************************************
2011-05-07 John Urbanic <urbanic@.psc.edu>

	*/quake/forward/psolve.c
	
        First definition of Param and Global structures to eliminate all wild
	global variables.  The process of herding all variables into these
	has begun.

	
********************************************************************************
2010-12-11 John Urbanic <urbanic@.psc.edu>
	
	* quake/forward/quakesource.c:

	(compute_myForces_srfh)

	- Changed so that non-participating nodes do not write source force files

********************************************************************************
2010-11-03 John Urbanic <urbanic@.psc.edu>
	
	* quake/forward/psolve.c:

	- re-enabled calls to damping routines

	* quake/forward/io_planes.c:

	(IO_PE_main)

	- accomodated odd combo of number of planes and IO_Pool count

	* quake/forward/checkpointing.c/h:

	- added stdlib.h and copyright headers

********************************************************************************
2010-10-29 John Urbanic <urbanic@.psc.edu>

	Changes to implement single or double precision selectable compilation
	
	* quake/forward/psolve.h:

	- Added SINGLE_PRECISION_SOLVER ifdef to switch between types

	- Changed structures f_vector and n_t to use this type, solver_float

	* quake/forward/psolve.c:

	(compute_adjust, schedule_sendata):

	- Changed doublesperentry to itemsperentry to be size agnostic

	- Changed all relevent doubles to new type, solver_float
	
********************************************************************************	
2010-07-20 John Urbanic <urbanic@.psc.edu>

	* quake/forward/psolve.c:

	- Moved checkpoint timers to io_checkpoint

	* quake/forward/io_checkpoint.c:

	(checkpoint_read, checkpoint_write):

	- Added striping code to scale on Lustre
	
********************************************************************************

2010-07-17 John Urbanic <urbanic@.psc.edu>

	* quake/forward/psolve.c:

	- Added header io_checkpoint.h

	- Changed checkpoint_read() and checkpoint_write() parameters

	- Moved all checkpoint related code to io_checkpoint.  Only
	  one call to checkpoint read and write left in psolve.

	* quake/forward/io_checkpoint.h:

	- Created for io_checkpoint.c.  Only included in psolve.c

	* quake/forward/io_checkpoint.c:

	- Move all checkpoint code here.  Refactored.  Will be
	  changed a more shortly to accomodate new large scaling
	  algorithm.

	* quake/forward/Makefile:

	- Modified to accomodate io_checkpoint.c/h

********************************************************************************

2010-07-13 Ricardo Taborda <rtaborda@.cmu.edu>

	* octor/octor.c:

	(node_setproperty):

	- Fixed bug for 6 touches case for a building resting on the surface
	  with a node in the interior of the bottom face.  This case was
	  missing.

	* quake/forward/buildings.c:

	- Added header file commutil.h.

	- Added a bunch of global variables to handle the fixed base options.
	  Will need to re-arrange them in a different way later.

	- Added local prototypes for fixed base methods. (some still missing
	  to add).

	(basenode_t):
	
	- New struct to handle nodes at the base of the buildings.

	(get_fixedbase_flag):

	- New public method to pass the fixed-base flag controller to psolve.

	(basenode_search):

	- New method to determine if a node is on the contact/cut of a building
	  with the surface. It returns 0 if the node is not in the building
	  and N > 0, N being the number of the building the node relates to.

	(bldgs_init):

	- Added broadcasting of the base fixed input parameters

	(buildings_initparameters):

	- Added instructions to read the flag to control fixed base options.

	- Added missing instructions to close parametersin file.

	(fixedbase_read, count_base_nodes, map_base_nodes, read_base_input, ...)
	(bldgs_fixedbase_init, interpolatedisp, bldgs_get_base_disp, ...)
	(bldgs_load_fixedbase_disps):

	- All new methods to force input at the base of buildings to reproduce
	  fixed base conditions.

	* quake/forward/buildings.h:

	- Added header file quake_util.h

	- Added public prototypes for get_fixedbase_flag, bldgs_fixedbase_init,
	  and bldgs_load_fixedbase_disps.

	* quake/forward/psolve.c:

	(solver_load_fixedbase_displacements):

	- Auxiliar method to invoque and load the displacements at the base
	  of the building nodes.

	(solver_run):

	- Added call to solver_load_fixedbase_displacements.

	(main):

	- Added call to bldgs_fixedbase_init.

********************************************************************************

2010-06-30 Ricardo Taborda <rtaborda@.cmu.edu>

	* octor/octor.c:

	(tree_t):

	- Added new element to know the tick address of the surface. Also
	  done in .h file for octree_t.

	(tree_setcom):

	- Added bldgs_nodesearch plugin to the parameters and pass it along
	  to com_allocpstl.

	(com_allocpctl):

	- Added bldgs_nodeseearch plugin and surface tick point and ticksize
	  to the method's parameters.

	- Style editing on dicard out of bounds if statements.

	- Added new provision to discard 'air' points.

	(node_setproperty):

	- Fixed bug for 6 touches and improved 5 touches case.

	(octor_newtree):

	- Added assignation of tree->surfacep.

	- Changed parameters being sent to com_allocpctl so that in this
	  first occasion it does not consider the case of buildings yet.

	(octor_carvebuildings):

	- Added new parameters: (1) a plugin from the buildings application;
	  and (2) a flag to handle the two different calls to this method
	  differently.

	- Added new sanity checks and temporarily commented out abort
	  exits.

	(octor_trimfirstleaf):

	- Eliminated! obsolete!

	(octor_partitiontree):

	- Added buildings plugin to the parameters and passes it to tree_setcom.

	(octor_extractmesh):

	- Fractioned the order of ecvaluation to discard 'air' points.

	- Improved error messages.
	
	* octor/octor.h:

	- Modified prototypes in accordance with changes in octor.c, i.e.
	  added parameters to balancetree, carvebuildings, and partitiontree.

	- Eliminated the prototype of the obsolete method octor_trimfirstleaf.
	
	(octree_t):

	- Added new item to this structure to have information about the tick
	  corresponding to the surface for the case of buildings. This is in
	  agreement with changes in octor.c related to tree_t.
	
	* quake/forward/qmesh:

	(main):
	
	  Added NULL as the second parameter in octor_partitiontree for 
	  compliance with recent changes in octor related to buildings meshing.
	  This is only to get avoid compilation errors. It will have to be 
	  reviwed in the future for running qmesh.

	* quake/forward/psolve.c:

	- Changed timers related to carving buildings.

	(mesh_generate):

	- Added new parameters requested by octor_* methods.

	- Eliminated call to onsolete method octor_trimfirstleaf and replaced
	  it with the variation of octor_carvebuildings with flag = 0.

	* quake/forward/buildings.c

	(get_airprops)

	- Added immediate return in case of a failed cvm_query so that queries
	  out of the domain do not spoil the meshing process for non-cubic
	  simulation domains.

	(bldgs_nodesearch):

	- Expanded description to be in compliance with onboundarysearch
	  method.

********************************************************************************

2010-06-17 Ricardo Taborda <rtaborda@.cmu.edu>

	* ..............................checkpoint..............................

	* parameters.in:

	  New and modified lines apply to whether include buildings/foundations
	  into the model. The whole module is now controlled by the following
	  line

	  include_buildings = <yes> OR <no>

	  If the above parameter is <yes>, the following input lines apply:

	  number_of_buildings = <INTEGER>
	  min_octant_size_m   = <DOUBLE>
	  surface_shift_m     = <DOUBLE>
	  buildings_n_factor  = <INTEGER>
	  
	  If the number of buildings above is > 0 then the following table
	  applies:

	  building_properties =

	  min_x max_x min_y max_y depth height Vp Vs Rho Vp Vs Rho

	  All values are real.  Coordinates are in local XYZ coordinates
	  where X=North, Y=East, and Z=Depth.  Coordinates do not have to comply
	  with etree power of 2 grid. They are adjusted according to the closest
	  grid point using the minimum octant size in physical coordinates (or
	  meters).

	  The minimum octant size DO HAVE to be such that:

	  min_octant_size = L / ( 2 ^ N )
	 
	  where

	  L is the maximum (dominant) domain dimension, and
	  N is a positive integer.

	* matlab-utils/examples

	  Simplified file name: Replaced parameters_for_matlab.in for 
	  plotmeshinput.in

	* matlab-utils/scripts

	  Simplified file name: Replaced plot3d_Hercules_v2.m for plotmesh.m

	* octor/octor.h

	(bldgs_nodesearch_t)

	- New application provided plugin. Helps finding if a vertex does or
	  does not exist in the simulation domain.

	(octor_newtree)

	- Added input parameter surface_shift

	(octor_carvebuildings, octor_trimfirstleaf)

	- New public functions

	(octor_extractmesh)

	- Added input parameter (function plugin) bldgs_nodesearch.

	* octor/octor.c

	- Added two global variables: theSurfaceShift and theTotalDepth.
	  The former indicates for how much is the surfaces push down to 
	  have space for buildings. The latter is the dimension of the
	  model in depth or Z-direction.

	- Added a typedef struct that is a copy of the edata_t from psolve.h.
	  We will need to think about these copies in a future re-ingineering
	  of the whole program.

	(node_setproperty)

	- Completely refactored.  It now allows many more cases of 'touches'
	  to satisfy with the different options posed by pushing the surface
	  down and including buildings that protrude from the surface after
	  they are carved off from the mesh including air elements.

	  A vertex can now be touched by 8, 6, 5, 4, 3, 2 and 1 elements.
	  The case for 7 touches is there but throws an immediate error.

	  Some of these cases will help in the future to implement topography.

	  Each case is well documented with comments at the beginning and
	  throughout their sub-cases.

	(octor_carvebuildings)

	- This new method identifies which elements or leaves have a payload 
	  with a negative value of Vp and eliminates those leaves from the tree.
	  When a leaf is eliminated, other tree structures and counters are
	  updated.

	- If a processor is to be left without any element, the method prevents
	  itself of allowing this by leaving at least one (first/leftmost) leaf
	  in the local tree, even if this leaf has a negative Vp.

	(octor_trimfirstleaf)

	- This new method eliminates the first leaf of the local tree if
	  this leaf has a negative value of Vp.  It corrects the special case
	  describe above.  It occurs after partitioning, thus it assumes that
	  all processors have now more than 1 leaf/element.

	(backfire_parent)

	- Is a new recursive method invoked by octor_carvebuildings and
	  octor_trimfirstleaf to make sure that a parent for whom all children
	  leafs have been eliminated because of them having a negative Vp
	  is also erased from the tree. Once this happens it calls itself
	  recursively to make sure that grandpas, great-grandpas and so one
	  do have childrens with data.

	(octor_partitiontree)

	- Added a temporary sanity check to make sure that after partition,
	  the first leaf is not pointing to NULL.  May be eliminated later on
	  when we get sure the buildings carving is working properly.

	- Changed the tree_showstat to 'details' instead of 'brief'.

	(octor_extractmesh)

	- Added new parameter. A plugin to be sent to the node_setproperty
	  method.

	- Added two sanity checks before traversing the elements to create 
	  the vertices.  May be eliminated later on when we proof the 
	  buildings package.

	- Improved error messaging.

	- Additional sanity check before doing the direct sharing. For later
	  elimination.

	- Also in direct sharing loop.  Corrected style of discards for
	  potential neighboring vertices.

	- Added a provision to discard vertices that are in the 'air'.

	- Added a sanity check to make sure that the returned procid from
	  match_zsearch exists.

	- Improved error message for when node_setproperty returns a negative
	  value.

	* quake/cvm

	  cvmpayload_t is now a typedef from the go.

	* quake/forward/io_planes.h

	  Added copyright header.

	(planes_setup)

	  Added parameter surfaceShift

	* quake/forward/io_planes.c

	- Added copyright header.

	(planes_setup)

	  Added parameter surfaceShift

	(Old_planes_setup, New_planes_setup)

	- Now take the surface shift and adjust the planes origin accrdingly.

	* quake/forward/qmesh.c

	  ***WARNING***  
	  Changed octor_ calls to comply with the new buildings package.
	  Not tested.  Only changed to ensure compilation.

	* quake/forward/quake_util.c

	  Cosmetic changes.

	* quake/forward/quakesource.h

	(compute_print_source)

	  Added new parameter: surfaceShift.

	* quake/forward/quakesource.c:

	  ***WARNING*** Buildings are only working properly with SRF source.

	(read_srf_source, source_init_parameters, compute_print_source)

	  Now accepting surfaceShift to adjust source location in the case
	  of buildings.

	* quake/forward/psolve.c

	- New controller for buildings is a No/Yes flag.

	- Renamed controller for nonlinear analysis.

	(setrec)

	- Added provisions to capture data in octants that belong to a
	  building.

	- Added an if statement to check if the query point needs to be
	  pushed down by the surface-shift.

	(mesh_generate)

	- Added parameter in octor_newtree

	- Adedd calls to octor_carvebuildings and octor_trimfirstleaf.

	- Added plugin-passing parameter to octor_extractmesh.

	(toexpand, mesh_correct_properties)

	- Refactored for the case of buildings.

	(print_timing_stat)

	- Added new octor_ buildings-methods timers.

	(source_init)

	- Now passing the surfaceShift.

	(read_stations_info)

	- Now adjusting depths for the case of buildings.

	(main)

	- Modified to make all calls to comply with buildings new provisions
	  and passing parameters.

	* quake/forward/buildings.h
	* quake/forward/buildings.c

	  Almost completely refactored.  Took all the work Yigit had done
	  for the special case of foundations and converted it in what we have
	  envisioned for the case of full rectangular-prismatic buildings.

	  It would be pointless to describe the many changes done. It is well
	  commented and organized.  Refer to the header of each method for
	  explanations.
	

********************************************************************************

2010-06-02 Yigit Isbiliroglu <yisbilir@.andrew.cmu.edu>

	* meshformatlab.c
	
	(saveMeshCoordinatesForMatlab)
	
	- This function is modified such that a processor will create output binary 
	  files if that processor has elements within the specified boundary. In
	  other words, no empty binary files will be created. For this int counter
	  is added.

	* matlab-utils/
	
	- plot3d_Hercules.m is also modified to be compatible with the changes in
	  saveMeshCoordinatesForMatlab. Its name is changed to plot3d_Hercules_v2.m
		
********************************************************************************

2010-05-17 Ricardo Taborda <rtaborda@.cmu.edu>

	* parameters.in:

	- New lines applicable only if nonlinear flag is active. The parameters
	  input file should have these two lines:

	  geostatic_loading_time_sec   = 0
	  geostatic_cushion_time_sec   = 0

	  Values have to be 0 or greater than 0, and their sum cannot be
	  greater that the total simulation time.  Their sum will effectively
	  delay the starting of the source (srfh only) and this time should be
	  considered for defining the simulation time.

	* quake/forward/psolve.c:

	(local_init):

	- Added definition LOCAL_INIT_DOUBLE_MESSAGE_LENGTH to control the
	  number of doubles to be sent and recieved.

	- Eliminated the USECVMDB specific extension to double_message and
	  created a double_message_extra instead, that is only created if
	  the compilation flag is defined.  Changed the casting and collection
	  of data accordingly.

	(solver_nonlinear_state):

	- Modified input parameters to reflect changes related to the new
	  geostatic loading functions described below.

	- Added call to compute_bottom_reactions conditioned to the
	  geostatic loading time.

	(solver_compute_force_gravity):

	- New method that controls the call to compute_addforce_gravity
	  described below.

	(solver_compute_displacement):

	- Eliminated some commented lines that no longer apply.

	(solver_geostatic_fix):

	- New method that controls the call to geostatic_displacements_fix
	  in the nonlinear module described below.

	(solver_run_collect_timers):

	- Added new timer for addforces gravity.

	(solver_run):

	- Added/modified calls to solver_nonlinear_state, solver_compute_gravity
	  and solver_geostatic_fix.

	(printing_timing_stat):

	- Added new timer for gravity forces calculation.

	(source_init):

	- Added globalDelayT (global delay in time for the source in seconds).
	  For now, this is ONLY VALID for the srfh source type.  It delays the
	  starting time of the source to accomodate time for the geostatic
	  loading and cushion. Its value is passed to compute_print_source.

	(main):

	- Added parameters to call of nonlinear_init().

	- Moved nonlinear_solver_init and nonlinear_stations_init to before
	  calling source_init.

	- Added call to check_balance().  This function checks if the weight
	  of the model matches that of the reactions for after geostatic loading
	  is applied.  Will need to be relocated and added provisions for
	  handleing errors.

	* quake/forward/quakesource.h:

	(compute_print_source):
	
	- Added parameter globalDelayT

	* quake/forward/quakesource.c:

	(read_srfh_source):

	- Added parameter globalDelayT.

	- This parameter is added to all theSourceTinitArray entries.

	(source_init_parameters, compute_print_source):

	- Added parameter globalDelayT

	* quake/forward/nonlinear.h:

	- Added typedef struct bottom_elements_f. It has an id that maps to the
	  table of elements and a vector with the nodal forces to be used as
	  reactions at the bottom.

	(get_geostatic_total_time, compute_addforce_gravity,...)
	(compute_bottom_reactions, geostatic_displacement_fix, check_balance):

	- New method prototypes.

	* quake/forward/nonlinear.c:

	- New global variables: theGeostaticLoadingT, theGeostaticCushionT,
	  theGeostaticFinalStep, myBottomElementsCount, myBottomElements,
	  totalWeight.

	- Some cosmetic changes.

	(get_geostatic_total_time):

	- Returns the sum of the two variables defining the total
	  geostatic loading time.

	(nonlinear_init):

	- Added new parameters to apss on to nonlinear_initparameters.

	- Increased the double and int messages arrays.

	- Added boradcasting of new global variables mentioned above.

	(nonlinear_initparameters):

	- Added new arguments and local variables.

	- Now reading geostatic loading parameters.

	(isThisElementAtTheBottom):

	- New method that returns YES if an element is at the bottom, NO
	  otherwise.

	(nonlinear_print_stats, nonlinear_stats):

	- Now prints/collect infor for how many elements there are at the 
	  bottom of the model.

	(nonlinear_solver_init):

	- Now calls for counting and mapping the bottom elements for geostatic
	  loading.

	(check_yield_limit):

	- Set the limit back to 1.5---need more research before settling for
	  a good value.

	(smooth_rise_factor):

	- New method. Returns a factor between 0 and 1 depending on the current 
	  step in the pre-simulation stage for geostatic loading. This factor
	  is used to calculate the progressive addition og gravitational
	  forces.

	(add_force_reactions):

	- New method. It adds the reactions at the bottom to compensate the 
	  weight of the model once it is released for the simulation.

	(check_balance):

	- New method. Does not actually checks (it should in the future) but
	  for now it simply prints to stdout the result and difference of
	  the total weight of the model and the sum of the reactions at the
	  bottom.  They should match or be sufficiently close to avoid
	  numerical errors.

	(compute_addforce_gravity):

	- New method. It computes the contribution of each elememt's weight
	  for geostatic loading and once total pre-loading is reached it also
	  calls for adding the reactions at the bottom.

	(compute_bottom_reactions):

	- New method. Computes the reactions at the bottom elements due to
	  the deformation these elements suffered because of geostatic
	  loading. It uses the conventional stiffness matrix - displacements
	  product.

	(geostatic_displacements_fix):

	- New method. Fixes the vertical displacement of the bottom nodes
	  for applying geostatic gravitational loading.

********************************************************************************

2010-05-20 Yigit Isbiliroglu <yisbilir@.andrew.cmu.edu>

	* matlab-utils/

	A new feature is added to program. A matlab function 'plot3d_Hercules.m'
	that plots 3d meshing of any specified region within domain is added 
	into newly created file 'matlab-utils'. Its usage is decribed in the 
	script itself. A sample 'parameters.in' file is also added in the 
	subfolder 'examples' under matlab-utils.
	
********************************************************************************

2010-05-19 Yigit Isbiliroglu <yisbilir@.andrew.cmu.edu>

	A new feature is added to program. Now, any specific portion of domain 
	can be plotted in 3D via MATLAB. For this, meshformatlab.c and 
	meshformatlab.h are created. Element coordinates and data are stored 
	in binary format. 4th dimension in plot can be Vp, Vs or rho depending 
	on user preferences. It also can be id of processor which possesses 
	the data of corresponding element.
	
	* parameters.in

	- Added lines defining the parameters for mesh properties of matlab.
	  One of the following two options should be added to parameters.in
	
	# Mesh properties for matlab
	
	mesh_coordinates_for_matlab            = no
	
	OR
	
	# Mesh properties for matlab

	mesh_coordinates_for_matlab            = yes
	mesh_coordinates_directory_for_matlab  = outputfiles/For_Matlab

	#         min_x       min_y       max_x     max_y     min_z      max_z
	# .....................................................................

	mesh_corners_matlab =

	0 0 800 800  0  800
	
	mesh_corners_matlab defines the portion of the domain of which element
	properties will be written in a binary file.
			
	* quake/forward/psolve.c:
	
	- meshformatlab.h is added in header.
	
	- static  noyesflag_t  storeMeshCoordinatesForMatlab is defined.
	
	(local_init):
	
	- storeMeshCoordinatesForMatlab  is broadcasted to all processors.
	
	(initparameters):
	
	- Added char mesh_coordinates_for_matlab[64].
	
	- Added  noyesflag_t  meshCoordinatesForMatlab and initialized as -1.
	
	- mesh_coordinates_for_matlab is parsed by PE0 from parameters.in and 
	  sanity check is done to make sure it is either 'yes' or 'no'. Then, 
	  assigned to meshCoordinatesForMatlab which is assigned to 
	  storeMeshCoordinatesForMatlab afterwards. Also printed on the screen.
	
	(toexpand):
	
	- int returnval is removed.
	
	- Regular Vs rule is moved into a new function called vsrule which is 
	  located in quake_util.c.
	
	- A new function 'buildingmagic' is called to check whether an element 
	  is in	foundation and need to be refined. 
	
	(main):
	
	- An if statement is added to call saveMeshCoordinatesForMatlab if 
	  storeMeshCoordinatesForMatlab is 'yes'.
	  
	* quake/forward/buildings.c:
	
	- A new function 'buildingmagic' is added. It is not doing anything new,
	  it just calls either foundationSearch or vsrule.
	
	(isThisElementInFoundation):
	
	- Code repetition is overcomed.For this, incoming parameters of 
	  isThisElementInFoundation is modified. Again, no change in 
	  functionality. Now, this function is called by both 
	  foundationSearch and foundationCorrectProperties.
	
	(foundationCorrectProperties):
	
	- This function is modified such that now it calls 
	  isThisElementInFoundation in order to avoid code repetition.
	
	* quake/forward/quake_util.c:
	
	- A new function 'vsrule' is created and regular Vs rule is moved into 
	  that function. Now, functions toexpand and buildingmagic calls 
	  'vsrule' to decide on expansion of octants.
		  
********************************************************************************

2010-05-18 Ricardo Taborda <rtaborda@.cmu.edu>

	* quake/forward/psolve.c:

	(main):

	- Added parameter in nonlinear_yield_stats call.

	* quake/forward/nonlinear.h:

	- Added parameter in nonlinear_yield_stats call.

	* quake/forward/nonlinear.c:

	(nonlinear_yield_stats):

	- Entirely new.  It now does not take statistics for all values
	  of Vs for each processor but does it globally for all processors
	  and using the ranges defined by the nonlinear table of properties.
	
********************************************************************************

2010-05-17 Ricardo Taborda <rtaborda@.cmu.edu>

	* quake/forward/psolve.c:

	(interpolate_station_displacements):

	- Added an if statement to force files to be flushed every 10 times
	  the stations printing rate. Will need to be improved so that
	  that factor of 10 is a user defined parameter with certain
	  default value.

	(main):

	- Added an if statement after solver run that goes to the nonlinear
	  module for printing yield statistics. Might need to be improved to
	  capture more statistics.

	* quake/forward/nonlinear.h:

	- Commented out for later elimination the swrtJ2 element in the
	  nonlinear element constants structure.

	- Added the maxFs and avgFs elements to the constants structure.

	- Added nonlinear_yield_stats prototype.

	* quake/forward/nonlinear.c:

	(compute_nonlinear_state):

	- Moved some variable declarations to the beginning.

	- Fs, J2, and the deviatoric and elastic strain tensor are now 
	  computed and used in a per quadrature-point form.

	- Fs is now computed also for the linear case.

	- Added commands for storing yield statistics.

	(nonlinear_yield_stats):

	- New method that collects and prints the statistics (max and avg) of
	  the Fs values. Designed to run in linear case and see the range
	  of Fs and then in the nonlinear case to check how much above the
	  limit it goes.

********************************************************************************

2010-05-05 Ricardo Taborda <rtaborda@.cmu.edu>

	- Activated stations print rate. Should have done a long time ago.

	* quake/forward/psolve.c:

	(solver_output_stations):

	- Moved timers to within the conditional.

	(solver_nonlinear_state):

	- Moved timers to within the conditional and separate them into the two 
	  different timers each call should belong to.  Maybe later, nonlinear
	  stations printing should have its own timer.

	- Added step and rate parameters.

	(solver_run):

	- Added step parameter in call to solver_nonlinear_state.

	(broadcast_stations_info):

	- Added line to broadcast theStationsPrintRate.

	* quake/forward/nonlinear.h,c:

	(print_nonlinear_stations):
	
	- Added step and rate parameters

	* quake/forward/nonlinear.c:

	(print_nonlinear_stations):

	- Boundle the printing in a conditional that checks if the residual
	  of step/rate is zero.

********************************************************************************

2010-05-05 Ricardo Taborda <rtaborda@.cmu.edu>

	* quake/forward/nonlinear.h:

	(init_tensorptr, get_displacements):
	
	- New method prototypes.

	* quake/forward/nonlinear.c:

	(init_tensorptr):
	
	- New method. Initializes a tensor pointer to have all tensor 
	  components equal zero.

	(compute_yield_surface_state):

	- Commented out all options.  Now returns the complete Fs value
	  regardless of the material model. Required for printing linear
	  state.

	(get_displacements):
	
	- New method. Returns the displacements on the nodes of a given element.
	  Created to avoid code duplication. Authomatically checks if displace-
	  ments are all zero. Could be of use in other parts of the code.

	(compute_nonlinear_state)

	- Now capturing displacements using new methdo described above and
	  skipping for the next element if displacements are zero.

	(nonlinear_stations_init):

	- Now initializes all tensors in a station to zero.

	(print_nonlinear_stations):

	- Moved mapping index declaration to out of the loop. Moved declaration
	  of lx, ly, lz, and bulk tensors to the beginning. Same for the mapping
	  index assignation.

	- Now capturing displacements with new method described above and 
	  jumping to printing if all disps are zero.

	- Moved Fs calculation to out of the conditional to have results for
	  the linear case.

********************************************************************************

2010-05-04 Ricardo Taborda <rtaborda@.cmu.edu>

	* quake/forward/nonlinear.c:

	- Eliminated commented structs that were moved to psolve.h a while ago.

	(solver_nonlinear_entities, solver_nonlinear_state):

	- Changed name of method, before ..._entities, now ..._state.
	
	- Included additional parameter in print_nonlinear_entities, now
	  called print_nonlinear_stations.

	- Changed to nonlinear module changes for former 
	  compute_nonlinear_entities, now compute_nonlinear_state.

	(solver_run):

	- Reflects change above.

	* quake/forward/nonlinear.c:

	- Added new global variable to store the nonlinear state of stations

	(qptensors_I1, qptensors_octahedrals, qptensors_deviators,...)
	(qptensors_J2, compute_qp_strains, copy_qp_tensors)

	- Eliminated. These were obsolete now that I have merged everything to
	  per-point methods.

	(compute_nonlinear_entities, compute_nonlinear_state):

	- Former compute_nonlinear_entities is now compute_nonlinear_state.

	- Eliminated TODO tag that was completed.

	(nonlinear_stations_init):

	- Added allocation of memory for nonlinear state of stations.

	(interextrapolate):

	- Eliminated. Was not accurate and became obsolete now that stations
	  are done correctly for each coordinate.

	(print_nonlinear_stations):

	- Completely refactored to do nonlinear state calculations at the
	  exact coordinates of the station.
	
	* quake/forward/nonlinear.h:

	- Refactored to reflect changes in nonlinear.c

	- Added new typedef struct nlstation_t.

********************************************************************************

2010-05-04 Ricardo Taborda <rtaborda@.cmu.edu>

	* quake/forward/nonlinear.c:

	(pqtensors_I1, qptensors_octahedrals, qptensors_deviators,...)
	(qptensors_J2, compute_cp_strains, copy_qp_tensors):

	- Became obsolete with the move to do computations per (quadrature)
	  point for later correct implementation of stations.  For now only
	  commented out. For later removal.

	(compute_nonlinear_entities):

	- Copy of pstrains 2 to 1 moved to correct place.

	- All calculations modified to be per quadrature point and not all 
	  at once per element.

	- A more natural 'continue' statement that accounts at once for
	  linearity.

********************************************************************************

2010-05-04 Ricardo Taborda <rtaborda@.cmu.edu>

	* quake/forward/psolve.c:

	(setup_stations_data):

	- Changed header for nonlinear stations.

	* quake/forward/nonlinear.h:

	(nlconstants_t):

	- Changed fsk to fs.

	(stability_methods):

	- Changed name of still_stable to check_yield_limit, and added
	  check_strain_stability.

	* quake/forward/nonlinear.c:

	- Changes reflected from .h file

	(compute_yield_surface, compute_dLambda, compute_dfds,...)
	(compute_pstrain2, check_strain_stability):

	- New methods for more modularity and cleaner code for manipulating
	  nonlinear points and tensors, and check stability.

	(check_yield_limit):

	- Old still_stable improved.

	(compute_nonlinear_entities):

	- Compressed the last section in calls to methods introduced above.
	  
	(print_nonlinear_entities):

	- Reduced the number of parameters to print by one, eliminatig 
	  unnecessary redundancy between Fsk and Fs.

********************************************************************************

2010-04-29 Ricardo Taborda <rtaborda@.cmu.edu>

	* quake/forward/psolve.c:

	- Identation and cosmetic changes.

	(solver_compute_force_damping):

	- Re-activated damping.  It had temporarily been shut down for the 
	  SC10 paper.

	* quake/forward/nonlinear.h:

	(typedef_structs):

	- New approach to tensors with explicit components and a higher level
	  structure for the quadrature points.

	- Cosmetic changes in the constants struct and the nonlinear solver 
	  struct.

	(methods):

	- Changed numerical.in to parameters.in.

	- Rearrange them in the correct order (still need some editing)

	- Added a whole new set of methods to deal with tensors and modified
	  old ones. See larger explanation below.

	* quake/forward/nonlinear.c:

	- Added definitions for quadrature point coordinate factor QC and
	  for shape functions table XI.

	(nonlinear_init):

	- Changed to parametersin

	(nonlinear_solver_init):

	- Adapted allocations to the changes in typedefs and structs in .h.

	(tensor_I1, qptensors_I1, tensor_octahedral, qptensors_octahedrals,...)
	(tensor_deviator, qptensors_deviators, tensor_J2, qptensors_J2,...)
	(init_tensor, point_strain, compute_qip_strains, point_stress,...)
	(compute_qp_stresses, subtract_tensors, subtract_qptensors,...)
	(copy_tensor, copy_qp_tensors):

	- New and changed methods for operations on tensors.

	(point_dxi, compute_qp_dxi):

	- New methods that replace compute_xi_derivatives.

	(compute_addforce_nl, compute_nonlinear_entities,...)
	(print_nonlinear_entities):

	- Addapted procedures for new treatment of tensors.
	

********************************************************************************

2010-04-21 Ricardo Taborda <rtaborda@.cmu.edu>

	* parameters.in

	- New options for printing the stiffness matrices to standard output and
	  to inlcude the velocities and accelerations in the station files. The
	  new input lines should go like this...

	  print_matrix_k              = <yes> OR <no>
	  print_station_velocities    = <yes> OR <no>
	  print_station_accelerations = <yes> OR <no>

	* quake/forward/buildings.h:
	* quake/forward/buildings.c:

	- Mainly cosmetic changes.

	- Renamed some methods to start with small letters.

	- Also added two TODO messages for future reference.

	* quake/forward/nonlinear.h:

	- Removed the 'dononlinear' type. It is being replaced by a lower level
	  type called the 'noyesflag' in quake_util.h

	- Added the strainrate and sensitivity variables to the nonlinear
	  elements constants.

	- theGroupSize is now being passed in nonlinear_solver_init.

	* quake/forward/nonlinear.c:

	- Added global variables: theStrainRates and theSensitivities

	(isThisElementNonLinear):

	- Now passes a YES, NO flag.

	- The Vs cut is now inclusive. An element is declared nonlinear if
	  its value of Vs <= theNonLinVsCut.

	(interpolate_property_value):

	- Fixed bug for the above ceiling case

	(nonlinear_local_init):

	- Added memory allocation and broadcasting for the two new global 
	  arrays for the strain rates and sensitivity factors.

	(nonlinear_initparameters):

	- Changed input file name to parametersin.

	- Added memory allocation for the two new global arrays.

	- Change the dimension diven to parsearray from 3 to 5 and added the
	  lines to collect the array columns and sort them properly.

	(nonlinear_elements_count):
	(nonlinear_elements_mapping):

	- Changed 0,1 flags to NO,YES flags.

	(nonlinear_solver_init):

	- Now receives theGroupSize as a parameter for doing statistics.

	- Added calls to interpolate the values of strainrate and sensitivity
	  to be assigned to each element.

	(compute_nonlinear_entities):

	- Now, strainRate and sensitivity values are now those assigned to each
	  element. Before, they were hard coded.

	- eliminated the use of lambdaZero, now strainRate.

	* quake/forward/psolve.h:

	(mysolver_t):
	
	- Added TODO message.

	- Added the 3rd vector of displacements. Necessary for accelerations
	  to be recorded in stations.

	* quake/forward/psolve.c:

	- Commented out unused global variables. To be deleted next time aorund.

	- Created a 'Some Additional Options' section of global variables
	  for controling printing K, velocities, and accelerations.

	- Changed dononlinear_t to noyes_t type in the declaration of the
	  doNonlinear variable.

	(local_init):

	- Total of integer messages is now 16.

	- Now also broadcasting printK, printStationVelocities, and 
	  printStationAccelerations.

	(initparameters):

	- Changed some flags to the noyesfalg_t.

	- Now reading and storing flags for printing the stiffness matrices, and
	  printing stations velocities and accelerations.

	(setrec):

	- Deleted unused vartiables and methods that had been commented out.

	(toexpand):

	- Style and cosmetic changes.

	(print_K_stdoutput):

	- Now printing all three K1, K2, and K3 matrices in clean format to 
	  standard output.

	(solver_init):

	- Printing K matrices is now a user controled option.

	- Added control of memory allocation for tm3 for accelerations depending
	  on the user selection for printing accelerations to stations.

	(solver_update_status):

	- Moved the star '*' to before the new line control for uniformity
	  on the standard output file.

	(solver_compute_displacement)

	- Added control for updating tm3 if the print accelerations option is
	  ON.

	(read_stations_info):

	- Fixed bug in memory allocation check for theStationZ (theStationY
	  was twice)

	(setup_stations_data):

	- Refactor formating of stations header.

	- Made velocities and accelerations header optional as controlled
	  by the user.

	(interpolate_station_displacements):

	- Refactor variables of displacements for shorter name and
	  added new variables for velocities and accelerations.

	- Activated option of printing velocities and accelerations.

	NOTE: If accelerations are printed, so are velocities by default.

	(output_stations_init):

	- Deleted unused variable init_time.

	(main):

	- Added call to stiffness_init for mapping elastic element indices.
	  This was necessary for compatibility with the nonlinear module.

	* quake/forward/quake_util.h

	- Added the noyesflag_t typedef enumarator for controling no, yes 
	  options trasparently.

	- Other cosmetic changes.

	* quake/forward/quake_util.c:

	- Expanded comment on how to use the UNDERFLOW_CAP definition.

	- Change the UNDERFLOW_CAP back to 1e-20

	- Re-activated the vector_is_zero and vector_is_all_zero methods.
	  Left the immediate returns commented and added a comment on how to
	  use them for scalability tests.

	* quake/forward/quakesource.c:

	(print_myForces):
	(search_nodes_octant):

	- Were unused. Commented them out.

	* quake/forward/stiffness.h:

	- Added three new methods stiffness_init, linear_elelemts_count, and
	  linear_elements_mapping, described in stiffness.c changes.
	
	* quake/forward/stiffness.c:

	- Included nonlinear.h for calls to isThisElementNonLinear.

	- Included util.h

	- Created Global variables myLinearElementsCount and 
	  myLinearElementsMapper

	(linear_elements_count):
	
	- New method. It counts how many elements in the processor are to be 
	  treated as linear elastic

	(linear_elements_mapper):

	- New method. It creates a mapper between the table of lements and the 
	  indices of those elements that will be considered as linera elastic.

	(stiffness_init):

	- New method. It calls the counter and mapper for initialization of 
	  global variables.
	
	(compute_addforces_conventional):
	(compute_addforces_effective):

	- Changed indices in the main for loops to be only for the amount of
	  linear elastic elements.

	- Added line that assigns the right element index depending on the 
	  mapper.

********************************************************************************

2010-04-16 Yigit Isbiliroglu <yisbilir@andrew.cmu.edu>

	A new feature is added to program. Now, foundation of buildings can be
	refined with desirable resolution for any given Vs value.For this, 
	buildings.c and buildings.h files are created. 
	
	* parameters.in

	- Added lines defining the parameters for building's foundations.
	  One of the following two options should be added to parameters.in
	
	  # Buildings
	
	  number_of_buildings           = 0
	
	  OR
	
	  # Buildings

	  number_of_buildings           = 2
	  buildings_n_factor            = 4
	  buildings_Vp_factor           = 2
	  buildings_Rho_factor          = 3

	  # min_x  min_y  max_x  max_y  depth    Vs
	  # .......................................

	  building_properties =
	    
	    350.0  350.0  450.0  450.0  100.0  1000
	    100.1  100.1  130.0  130.0  100.0  1000
	
	
	  If the number_of_buildings is 0, the program will work as usual.
	
	  building_properties: foundation bounds location and Vs to be assigned
	  buildings_n_factor:  minimum number of subdivisions per foundation
	                       edge. Currently effective for all 3 dirs.
		
	* quake/forward/psolve.c:
	
	- buildings.h is added in header.
	
	- static int theNumberOfBuildings is defined.
	
	(local_init):
	
	- theNumberOfBuildings variable is broadcasted to all processors.
	
	(initparameters):
	
	- Added int number_of_buildings.
	
	- number_of_buildings is parsed by PE0 from parameters.in and sanity
	  check is done to make sure its value is greater than 0. Then, assigned
	  to theNumberOfBuildings. Also printed on the screen.
	
	(toexpand):
	
	- int returnval is defined.
	
	- An if statement was added to check whether the PE has buildings or 
	  not. If it does, it proceeds according to the foundation rules of
	  expansion and skips the regular Vs rule.
	
	(mesh_correct_properties):
	
	- An if statement was added at the very beginning of the function. 
	  It checks whether the element is in a foundation. If yes, the main 
	  for loop continues with the next element after the properties of 
	  foundation element are corrected in an independent method.
	
	- At the very end of the function, allocated memories for 
	  foundation_coordinates are freed.
	
	(main):
	
	- An if statement is added to initialize parameters for buildings and
	  foundations if theNumberOfBuildings is greater than 0.

********************************************************************************

2010-03-31 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c:

	(print_timing_stat):
	
	- Mesh stats now total correctly

********************************************************************************
	
2010-03-26 Julio Lopez <jclopez@ece.cmu.edu>

	* quake/forward/psolve.c

	(mesh_generate):

	Relocated etree_correct_properties in order to reuse the already
	open CVM database

	(etree_correct_properties):

	Changed parameter to take a pointer to a etree file instead of
	a file name.

	(mesh_correct_properties):

	Renamed etree_correct_properties to mesh_correct_properties
	
********************************************************************************
	
2010-03-21 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c:

	(print_timing_stat)
	(solver_update_status)

	-Fixed display error for simulation rate maximum time
	
********************************************************************************

2010-03-20 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c:

	Added globals fastestTimeSteps, slowestTimeSteps and MONITOR_STATS_RATE
	to track simulation print rate and make solver_update_status() more
	integrated.

	(solver_update_status):

	- Added in tracking for run rate and estimated time to finish.

	(print_timing_stat):
	
	- Added in lots of new stats and averages

********************************************************************************

2010-03-19 Julio Lopez <jclopez@ece.cmu.edu>

	* quake/forward/psolve.c (solver_run):

	Restructured and moved pieces to separate functions in preparation
	for new communication functionality.

	
2010-03-15 Julio Lopez <jclopez at ece dot cmu dot edu>

	* quake/forward/psolve.c (local_init):

	BUG fix: theTimingBarriersFlag parameter was not being propagated
	to the compute nodes.  It is now added to the 'int_message'
	broadcast
	
********************************************************************************

2010-03-16 John Urbanic <urbanic@psc.edu>

        * quake/forward/psolve.c

	(solver_run):
	
	- removed temporaty barriers around compute addforce routines.

	(print_timing_stat):

	- Added in prints for new timers and stats.

	(main):

	- Added in timers for Mesh and Schedule stats printing.

********************************************************************************

2010-03-15 Julio Lopez <jclopez at ece dot cmu dot edu>

	* quake/forward/psolve.c 

	(solver_run):

	- Fixed timer placement for 'compute_addforces_e'
	- Added timer for 'damping_addforces'

********************************************************************************
	
2010-03-15 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/psolve.c:

	(local_init):
	
	- Corrected a bug with broadcasting the stiffness method.

	(solver_run):

	- Commented the call of addforce_damping.  We are forcing NO DAMPING
	  in order to be able to run performace/scalability problems toward
	  SC'10 paper.

********************************************************************************

2010-03-10 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c

	(solver_run):
	
	- added in temporaty barriers around compute addforce routines
	  for debugging

	(print_timing_stat):
	
	- added in Timer_Printall to dump raw timer data

	* quake/forward/stiffness.c
	
	(compute_addforce_effective):

	- added in temporary timers for debugging
	
	
********************************************************************************
2010-03-08 John Urbanic <urbanic@psc.edu>

	* quake/forward/quakesource.c

	(source_compute_print_stat)

	- Won't print out data for PEs with no forces

********************************************************************************
	
2010-03-05 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/stiffness.c:
	* quake/forward/psolve.c:
	
	- I commented out some of the changes made in the previous commit 
	  because they could affect performance and linear runs.  

********************************************************************************

2010-03-03 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/nonlinear.h:
	* quake/forward/nonlinear.c:

	- Changed invalid formulation using rate-independent plasticity
	  for the so-far-correct implementation of rate-dependent plasticity.
	  Changes are mainly reflected on the compute_addforce_nl and
	  compute_nonlinear_entities methods.

	- New methods were also added to make the above more transparent and
	  easy to follow.

	* quake/forward/stiffness.c:

	- Added include nonlinear.h for call to the isThisElementNonlinear 
	  method.

	(compute_addforce_conventional)
	(compute_addforce_effective)

	- Added a check to see if an element is nonlinear.  If it is, then 
	  it goes to the next element.  Now, elements tagged as nonlinear
	  do not require to go through these two methods.

	- Clean remanents of old way of computing stiffness contributions
	  and timers that were left commented out.

	* quake/forward/psolve.c:

	(solver_run):

	- Added aditional parameter in call to compute_nonlinear_entities.

	- Deleted old comments that no longer apply and were related to
	  an obsolete way of solving the next time step displacements.

	(setup_stations_data):

	- Added header for temporarily permanent printing of velocity.
	  TODO: Need to come back to this later to make it optional.

	- Also corrected the heading of the nonlinear option.

	(interpolate_station_displacements):

	- Activated printing of velocities.
	
********************************************************************************

2010-02-02 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/psolve.c: 

	(main):

	- Added call to the nonlinear stats after the nonlinear stations init.

	* quake/forward/nonlinear.c:

	(nonlinear_solver_init):

	- Eliminated call to nonlinear stats---moved to psolve to have it
	  after the nonlinear stations are done.

	(compute_nonlinear_entities):

	- Added the 'zero' as the check for the change in the flow rule (fsk).

	(nonlinear_stations_init):

	- Commented out unused variables.

********************************************************************************

2010-02-01 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/psolve.c: 

	(interpolate_station_displacements):

	- Changed the format for printing the time in the station files to a
	  larger number of significant figures to avoid cut numbers when a very
	  small delta t is required.

	- Added a 'cor-now-commented' section to print velocities as well.

	(main):

	- Added extra argument in the call of nonlinear_solver_init.

	- Moved the call to nonlinear_stations_init to after the nonlinear
	  solver has been done.

	* quake/forward/nonlinear.h:

	- Added four new method declarations related to initialization of non-
	  linear variables and structs and printing stats. More details below.

	* quake/forward/nonlinear.c:

	- Added golval variables for counting nonlinear elements per PE and
	  for mapping nonlinear elements data to the regular list in myMesh.

	(nonlinear_elements_count):

	- New method. Counts the number of elements that have a Vs less than
	  the nonlinear Vs cut especified in the parameters file.

	(nonlinear_elelemts_mapping):

	- New method. Allocates memory for the mapping vector in accordance with
	  previous method and then loops over all elements sotirng the local
	  element index to the complete list in myMesh.

	(nonlinear_print_stats):

	- New method. Prints stats about the number of nonlinear elements per
	  PE and the number of stations in nonlinear elements.

	(nonlinear_stats):

	- New method. Gathers the number of nonlinear elelemts and stations in
	  each PE and pass them to PE0 for printing.

	(nonlinear_solver_init):

	- Added theGroupSize as paremeter.

	- Added the call to the nonlinear elelemts count and mapping methods 
	  and stats method.

	- Change the allocation of memory for nonlinear solver to be dependant
	  on the local number of nonlinear elements and not on the total local
	  nomber of elements.

	- Change loop over all elements to be over nonlinear elements only and
	  subsequent indexing accordingly.

	(compute_addforce_nl, compute_nonlinear_entities)
	(nonlinear_stations_init, print_nonlinear_entities):

	- Change loop over all elements to be over nonlinear elements only and
	  subsequent indexing accordingly.  Also commented out check to the 
	  nonlinear vs cut because the above makes it unnecessary.

********************************************************************************

2010-01-29 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c

	(main)
	
	- Eliminated all traces of argv[3], the old physics.in file and shifted
	  all arguments accordingly in all subroutines called here.  Also fixed
	  comments.

	(local_init)

	- Combined 2 and 3 args as per above.

	(initparamters)

	- Collapsed both arguments into one as per above.  Just aliased 
	  physics.in to numerical.in for simplicity.
	
********************************************************************************

2010-01-28 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/nonlinear.c:

	- Added global variables for mapping the stations that fall in nonlinear
	  elements with the complete list of stations. The variables are the
	  integer myNumberOfNonlinStations and the vector 
	  myNonlinStationsMapping.

	(compute_addforce_nl):
	
	- Moved the changed of the sign in the forces to the final passing to
	  the nodalForce vector and keep original commented out until this is
	  well understood.
	
	(compute_nonlinear_entities):
	
	- Deleted the old code for printing that was commented out. The new
	  methods for printing are well tested now.

	(nonlinear_stations_init):

	- Added a first pass loop over the stations to determine how many fall
	  in nonlinear elements.

	- Added allocation for the mapping vector and changed the allocations
	  to be the number of nonlinear stations and not the number of stations.

	- Modified loops and assignation indices accordingly, and added sanity
	  checks for the number of stations. 

	(print_nonlinear_entities):
	
	- Changed loop limit to the number of nonlinear stations.

	- Erased old average methods that were commented out.

	- Changed stations indices where necessary to map to the correct 
	  stations.

********************************************************************************

2010-01-26 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/nonlinear.c:

	(compute_addforce_nl):
	
	- Changed the sign in the forces. This apparently was the error we have
	  been carrying for too long now.
	
	(compute_nonlinear_entities):
	
	- Improved the section that checks if an element is failing to meet
	  nonlinear rules.

	(print_nonlinear_entities):
	
	- Added printing invariant for bulk deformations and stresses.

********************************************************************************

2010-01-20 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/nonlinear.c:

	(compute_addforce_nl):
	
	- The vs cut is now correctly checked with the input parameter
	
	(compute_nonlinear_entities):
	
	- Added a provision to detect if the nonlinear run needs a smaller delta t.
	  It stops the simulation and prints out the location of the critical
	  element.
	  
	- Added a provision for the impossed LINEAR case and corrected the hard-
	  coded check on the vs cut to be the one provided in the input parameters.
	  
	(print_nonlinear_entities):
	
	- Override the localcoords of the station to print the average/middle point
	  of the element. Inter- extra-polation outside the quadrature points is
	  causing erroneous data. We need a better method here.

********************************************************************************

2010-01-20 John Urbanic <urbanic@psc.edu>

	* quake/forward/timers.c

	(Timers_PrintAll)

	- Reformatted output to make grep friendly

	* quake/forward/psolve.c

	(main)
	
	- Added in timer for source initialization

	(print_timing_stats)

	- Reformatted output for above timer

	(source_init)

	- Removed obsolete timer
		
********************************************************************************

2010-01-19 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/psolve.c:

	(solver_run):

	- Added call to print_nonlinear_entities.

	(setup_stations_data):

	- Wrapped the printing of nodes in a DEBUG ifdef block. Printing this
	  is irrelevant unless one is doing some debugging.

	- Changed the headings in the stations file to be aligned with the
	  data and added a header for when nonlinear flag is ON.

	(interpolate_station_displacements)

	- Changed back the format for printing displacements to how it was
	  originally done by Leo.  Julio had changed this a while ago. It turns
	  out that the original way is advantageous for how the nonlinear
	  module adds data to be printed out. Added a very visible warning
	  to avoid someone changes it again in the future.

	(main):

	- Added call to nonlinear_stations_init.
	
	* quake/forward/nonlinear.c: 

	- Added global variable myStationsElementIndices to handle the element 
	  indices for printing nonlinear data to stations.

	(nonlinear_solver_init):

	- Commented out global variables and code for getting global maxima
	  parameters. This will be refactored later in a more clever way.

	- Eliminated some old and commented out code.

	(compute_addforce_nl):

	- Added a provision to return without changes to the force vector for
	  the LINEAR case.

	- Cleaned up and did cosmetic changes.

	(compute_nonlinear_entities):

	- Added loop to store the sqrtJ2 values in the nonlinear element's 
	  data structure (enlcons).

	- Commented out all code related to printing and maxima and averages
	  calculation.  Will be later removed.

	(nonlinear_stations_init):

	- New method for setting up the element indices of stations.

	(interextrapolate):

	- New method to inter- and extra-polate nonlinear related quantities 
	  from the quadrature points to any other point in the element.

	(print_nonlinear_entities):

	- New method to handle printing nonlinear data to stations. There is
	  pending work to identify if the station has indeed an element with
	  nonlinear data or not.

	* quake/forward/nonlinear.h:

	- Added sqrtJ2 to the enlconstants_t struct. Needed it for printing.

	- Added declaration of methods for printing and interpolation.
	
********************************************************************************

2010-01-11 John Urbanic <urbanic@psc.edu>

	* quake/forward/timers.c

	- Added function Timers_Exists() to query if timer is instantiated

	* quake/forward/timers.h

	- changed to reflect above

	* quake/forward/psolve.c

	- Changed all old timer types to use timers.c lib

	- Eliminated all old timer variables and various timer routines and headers

	(print_timing_stat)

	- Reformatted for updated timer info

	(solver_run)

	-simplified monitor/progress output

********************************************************************************

2010-01-06 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/io_planes.c: 
	
	- Added inlcude quake_util.h to conform with parsedarray
	  having been moved there. This eliminates a warning.
	  
	* quake/forward/quakesource.c:
	
	- Added inlcude quake_util.h to conform with parsedarray
	  having been moved there. Also rearrange inlcudes in alphabetical order.
	  
	- Eliminated two unnecessary definitions (LINESIZE and ERROR)

	(parsedarray):
		
	- Erased the parsedarray method and moved it to quake_util.c.

	* quake/forward/quakesource.h:
	
	- Eliminated unnecesary definition of LINESIZE
	
	- Eliminated the declaration of method parsedarray.
	
	* quake/forward/quake_util.c:
	
	- Added and rearrange includes.
	
	(parsedarray):
	
	- Added parsedarray method. Was originally in quakesource. It is now also
	  used by nonlinear.c.  Having it at a higher level sourcecode makes it
	  vissible to necessary parties.
	
	* quake/forward/quake_util.h:
	
	- Added definition of LINESIZE.  This is defined at several places in the
	  code. We should try to unify this and others (e.g., PI).
	
	- Added declaration of method parsedarray.
	
	* quake/forward/nonlinear.c:
	
	- Added include quake_util.h
	
	- Change the global matrix theMaterialProperties to three vectors that are
	  easier to broadcast and their names are more meaningful.
	  
	(interpolate_property_value):
	(get_alpha):
	(get_kay):
	
	- Added these three new methods and place them in a utilities block.  They
	  help to read information from numerical.in that was initially hard-coded.
	  
	(nonlinear_local_init):
	
	- Change and corrected the memory allocation of the new vectors and their
	  broadcasrting to all processors after being read by PE0.
	  
	(nonlinear_initparameters):
	
	- Corrected typo in string comparison for DruckerPrager material model.
	
	- Added returns to the sanity checks so they do stop the program.
	
	- Change and corrected the memory allocation of the new vectors for PE0.
	  Before I was using a matrix.
	  
	(nonlinear_solver_init):
	
	- Changed from hard-coded assignment of nonlinear material properties to
	  the new user-friendly methods to read this information from numerical.in.
	
	* quake/forward/nonlinear.h:
	
	- Eliminated unnecessary definitions that had been commented out.
	
	- Added new declarations for methods: interpolate_property_value, get_alpha,
	  and get_kay. 
	
********************************************************************************

2010-01-05 Ricardo Taborda <rtaborda@cmu.edu>

	  Added new feature to the program. Now we can soften the material
	  properties according to the mesh final structure. Elements with 
	  stiff soil or rock-like properties in small elements (due to the
	  2-to-1 rule) are softened. By lowering Vs and Vp values down to 
	  a certain factor times the 'ideal' values, the cirtical delta t
	  may double, bringing simulation time down.

	* numerical.in:
	
	  A new line needed.
	  
	  softening_factor = <double>
	  
	  <double> must be 0 if you wan't no changes, or >= 1 for mesh
	  softening to take place. Read more below.

	* quake/forward/psolve.c: 

	- Added global variable theSofteningFactor. This variable can take
	  0 for no softening, and values greater or equal to 1. A value of
	  1 results in maximum softening. We need more testing on the 
	  consecuences of this factor, for now farctors > 4 or 5
	  are suggested.

	(local_init):

	- increased double_message size to 17

	- added theSofteningFactor to the passing message array for the
	  way forward and back and changed the corresponding number of
	  doubles passed in the MPI_Bcast.

	(initparameters):

	- added local variable for reading the softening_factor and a
	  line in the onvocation of the parsing.

	- added sanity check for the softening factor and instruction to
	  feed the global variable of PE0.

	- Did other cosmetic changes.

	(setrec):

	- correct an error in the definiton of points array.

	(etree_correct_properties):

	- Added local variable RhoVpRatio for correction of material
	  properties. I have pending discussion with Jacobo about it.

	- Re-arrange the adjustment of properties and added the new 
	  section that actually does the softening of the mesh 
	  transitions.

********************************************************************************

2010-01-05 John Urbanic <urbanic@psc.edu>

	* quake/forward/timers.c

	- changed declarations to elimate bogus gcc error

	- added in stdlib.h to surpress error

	- reformatted for 4 space indents

	* quake/forward/io_planes.c

	- eliminated unused variables

	- reformatted for 4 space indents

********************************************************************************

2010-01-04 John Urbanic <urbanic@psc.edu>

	* quake/forward/timers.c

	- Made timer default to cumulative timers.

	* quake/forward/timers.h

	- changed to reflect above

	* quake/forward/psolve.c

	- added in "timers.h"

	* quake/forward/Makefile

	- added timers.c/h to project
	
********************************************************************************

2009-12-30 John Urbanic <urbanic@psc.edu>

	* quake/forward/timers.c

	- Added in new timer library.  Initial routines are Timer_Start(),
	  Timer_Stop(), Timer_Reset(), Timer_Config(), Timer_Value(),
	  Timer_PrintAll(), and Timer_Reduce().

	* quake/forward/timers.h

	- added to support timers.c

********************************************************************************

2009-12-17 John Urbanic <urbanic@psc.edu>

	* quake/forward/io_planes.c

	- Finished all "New" routines.  IO server mode is now implemented.
	  This checkin involved New_Planes_setup() and planes_IO_PES_main().

	- Eliminated new_print_plane_displacements(). Rolled into _main()
	
	- Fixed unstable structure bcasts is Old_planes_setup()

	* quake/forward/psolve.c

	- Eliminated one more global from Globals section.

********************************************************************************

2009-12-16 John Urbanic <urbanic@psc.edu>

	* quake/forward/io_planes.c

	- Finished New_planes_print(), New_Planes_setup(), 
	  New_planes_close() and New_output_planes_construct_strips()

********************************************************************************

2009-12-15 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c

	- updated call to io_planes_close() to eliminate need for global myID


	* quake/forward/io_planes.h

	- updated call to io_planes_close()


	* quake/forward/io_planes.c

	- updated planes_IO_PES_main(), New_planes_close(), New_planes_print()

********************************************************************************

2009-12-14 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c

	- eliminated a few unused globals

	- added in a new communicator, comm_io, so that we do not need to
	  use comm_world for anything

	* quake/forward/io_planes.h

	- fixed def for planes_IO_PES_main

	* quake/forward/io_planes.c

	- updated planes_t struct for additional IO server variables

	- added in new server versions of New_planes_setup(),
	  planes_IO_PES_main() and New_output_planes_construct_strips()
	
********************************************************************************
	
2009-12-11 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/nonlinear.h
	
	- Commented out some definitions soon to disapear (SENSITIVITY and 
	  LAMBDAZERO)
	
	- Change dononlinear_type_t for dononlinear_t.  This has implications in 
	  psolve.c and at other nonlinear places.
	  
	- Added new structures: materialmodel_t and material_properties_t
	
	- Refactor some internal names and did other cosmetic changes.
	
	* quake/forward/nonlinear.c
	
	- Included quakesource.h.  This will later be repared by moving some
	  methods from there to quake_util
	  
	- Refactored the way everything is about reading data from input files.
	  The nonlinear_init is now done in two methods nonlinear_local_init,
	  nonlinear_initparameters and nonlinear_solver_init, emulating the way
	  we do it for regular linear solver. 
	
	- (nonlinear_solver_init): Old hard coded decisions about material model
	  are now in user control through parameters given in .in file.
	  
	- (nonlinear_solver_init): nonlinear Vs cut is now user controlled. 
	
	* quake/forward/psolve.c
	
	- Change the donolinear_t global variable
	
	- Erased commented lines that were no longer necessary for the nonlinear 
	  module.
	
	- (initiparameters) Change the dononlinear_t and made cosmetic changes 
	  for the damping and nonlinear sanity checks.
	  
	- (main) Added call to nonlinear_local_init and change the call to
	  nonlinear_init for the new nonlinear_solver_init

********************************************************************************

2009-12-10 Haydar Karaoglu <hkaraogl@andrew.cmu.edu>

	* quake/forward/stiffness.c
	
	- Input arguments to compute_addforce_effective are changed (Some are 
	  not needed anymore).
	
	* quake/forward/psolve.c
	
	- Changes in compute_addforce_effective is repeated.
	
********************************************************************************

2009-12-10 Haydar Karaoglu <hkaraogl@andrew.cmu.edu>

	* quake/forward/damping.c
	
	- This is a new file including the method used for the calculation of 
	  damping contribution to the force.
	
	* quake/forward/damping.h
	
	- This is a new file including the declarations and definitions used in 
	  damping.c
	
	* quake/forward/quake_util.c
	
	- This is a new file including the methods common to damping.c and 
	  stiffness.c.
	
	* quake/forward/quake_util.h
	
	- This is a new file including the declarations and definitions used in 
	  quake_util.c.
	
	* quake/forward/psolve.c
	
	- compute_addforce_e method which is related to the calculation of stiffness 
	  matrix contibution is taken into stiffness.c and renamed as 
	  compute_addforce_conventional.
	
	- vector_is_all_zero method is taken into quake_util.c.
	
	* quake/forward/psolve.h
	
	- damping_type_t definition is taken into damping.h 

********************************************************************************

2009-12-09 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c

	- Added IO_pool_count as global to determine size of IO server pool

	- In main() added in code to read IO_PES environment variable.  Based 
	  upon this the flow through main will activate the IO server or not.  
	  This will allow us to	maintain backwards compatibility with non-
	  server 1 core IO code.

	* quake/forward/io_planes.c 

	- Duplicated all routines with Old_/New_ versions and added a wrapper 
	  to redirect calls based upon server activation.  Fixed Old_ calls for 
	  internal consistancy (Old_ only call Old_).  Need to do that with 
	  New_ and put code into server main loop.

********************************************************************************

2009-12-09 Haydar Karaoglu <hkaraogl@andrew.cmu.edu>

	* numerical.in
	
	- Added a new paramter to define the method that will be used for the 
	  calculation of stiffness matrix contribution. The corresponding line in
	  numerical.in file should be one of the two following options: 

	      stiffness_calculation_method = effective
	      
	  or
	      
	      stiffness_calculation_method = conventional

	* quake/forward/Makefile:
     
        - Added stiffness.o in the PSOLVE_OBJECTS and added stiffness.h for 
	  psolve.o target.

	* quake/forward/stiffness.c:

	- This is a new file. It has the methods used for the effective 
	  calculation of the stiffness matrix contribution. 
	  compute_addforce_effective(..) method is in this file.
	
	* quake/forward/stiffness.h:

	- This is a new file. It contains the definitions and declarations 
	  used in stiffness.c.
	
	* quake/forward/psolve.c:
	
	- Added include stiffness.h
	
	- Added "stiffness_type_t stiffness" global variable to keep the method 
	  used for the calculation of stiffness matrix contribution. Struct is 
	  defined in stiffness.h. 
	
	(init_parameters)
	
	- Added char stiffness_calculation_method to be read from numerical.in 
	  and stiffness_method flag.
	
	- Added the line (parsetext call) that reads the stiffness flag from 
	  numerical.in.

********************************************************************************

2009-12-09 John Urbanic <urbanic@psc.edu>

	* quake/forward/psolve.c:

	- Removed two unused globals, added IO_POOL_PE_COUNT and
	  IO_POOL_PES[] as statics to track IO PEs.

	- Added final main loop logic for IO PEs.  They call planes_IO_
	  PES_main() and then return via an evil goto to the end of main().
	  This whole routine should be a case statement to avoid the goto.

	* quake/forward/io_planes.c:

	- Added planes_IO_PES_main() for IO Pool PEs.  Deleted two
	  unused globals.

	* quake/forward/io_planes.h:

	- Added planes_IO_PES_main()

********************************************************************************

2009-12-08 John Urbanic <urbanic@psc.edu>

	* created quake/forward/io_planes.c:

	- Moved all planes related functions from psolve.c to here: 

	- print_planeinfo(),print_plane_displacements(),
	  output_planes_construct_strips() are now static/private.

	- planes_interpolate_communicate_and_print(), output_planes_setup();
	  extern void output_planes_finalize() and now renamed to 
	  planes_print(),planes_setup(),planes_close() and added some
	  new arguments to pass previous globals in psolve.

	* created quake/forward/io_planes.h:

	- planes_print(),planes_setup(),planes_close() defined.

	* quake/forward/psolve.c:

	- added io_planes.h and deleted code that was moved to io_planes.c

	- updated planes_ calls to conform to new names and args

	- changed compute_csi_eta_dzeta() to extern so that it could be
	called from io_planes.c as well as here.  This should be moved to
	a utility file.

	- Made these variables global as they are used by io_planes and
	  stations code at least.  They should be collected into into two
	  structures (DomainDescription and InputParams for ex.) along with
	  related globals and passed as structures.  I will do this when I
	  redo the input parsing:

	  theRegionLong, theRegionLat, theRegionDepth, theNumberOfPlanes,
	  thePlaneDirOut[256], theSurfaceCornersLong[4], theSurfaceCornersLat[4],
	  thePlanePrintRate, theNumberOfPlanes, thePrintPlaneTime = 0,
	  theCollectPlaneTime = 0, theDomainX, theDomainY, theDomainZ
	
	* quake/forward/Makefile:

	- Added io_planes.c and io_planes.h
	
********************************************************************************

2009-11-25 Ricardo Taborda <rtaborda@cmu.edu>

	* quake/forward/Makefile:

	- Included nonlinear.o in the PSOLVE_OBJECTS and added nonlinear.h
	  for the psolve.o target.

	* quake/forward/quakesource.c:

	- Commented out the definition of ERROR. This definition is now at
	  psolve.h and instead of -100, its value is set to HERC_ERROR.

	- Typedef struct edata_t was erased. This structure is now at
	  psolve.h and is visible from there to the nonlinear module.

	- Commented out the search_point method because this code already
	  existed in psolve.c.

	* quake/forward/geometrics.h:

	- Added include "psolve.h" to allow visibility to some common
	  definitions.

	- Commented out typedef struct vector3D_t that is now in psolve.h

	* quake/forward/geometrics.c:

	- Moved the inlcude "geometrics.h" statement to after the system
	  libraries.

	* quake/forward/nonlinear.h:

	- This file is all new. It contains the definitions and
	  declarations used in nonlinear.c.

	* quake/forward/nonlinear.c:

	- All new. This file has the methods relative to the incorporation
	  of nonlinear soil conditions.

	* quake/forward/psolve.h:

	- Added include "octor.h" and the definition of ERROR as
	  HERC_ERROR.

	- Added the typedef struct edata_t, messenger_t, schedule_t,
	  my_solver_t, and station_t.  All these structs were in psolve.c
	  but needed to be visible to the nonlinear module---thus the
	  decision to move them here.

	- Brought the struct vector3D_t from geometrics.h to here to make
	  it available to the nonlinear module. In the future it would be
	  good if we can unify this structure with fvector_t. They are
	  almost identical and used in very similar conditions all over the
	  code.

	- Added declaration of search_point method (no longer a static
	  method).

	- Added declaration of method mu_and_lambda.

	* quake/forward/psolve.c:

	- Added include "nonlinear.h" statement.

	- Eliminated the definion of ERROR, now at psolve.h.

	- Added timers variables for the nonlinear module (NL,NL2).

	- Eliminated edata_t struct, now at psolve.h.

	- Commented out (for later elimination): messenger_t, schedule_t,
	  and mysolver_t. These typedefs are now at psolve.h---so they may
	  be totally eliminated from psolve.c later on once we agree
	  psolve.h is the right place for them to be.

	- Included some declarations for the nonlinear module. Most remain
	  commented out and will be erased later.

	- Commented out three global variables related to planes writint
	  that were not being used and they were triggering unnecessary
	  compilation warnings.

	- Erased typedef struct station_t, now at psolve.h. This structure
	  is also being used by the nonlinear module---thus the need for
	  moving it.

	(local_init):
	
	- Made int_message larger for including the nonlinear flag.

	- Added the doNonlinear message passing ints.

	(initparameters):

	- Added char inlcude_nonlinear_analysis for be read from
	  numerical.in and the inlcudeNonlinear flag.

	- Reformat the parsetext calls for easier reading. Added the line
	  that reads the nonlinear flag from numerical.in

	- Added a sanity check for the nonlinear flag read from
	  numerical.in.

	- Corrected an undetectable bug with missing check in the
	  parsetext call for reading the region_azimuth.

	- Added a sequence of monitor_print calls for printing very basic
	  data about the run at the top of the monitor file.

	(replicateDB): 

	  moved the declaration of char curdir[256] into the
	  compilation-controlled section CVM_DESTDIR that does use it. This
	  eliminates a compilation warning.

	(solver_init): 

	  eliminated the section that was calculating the
	  Lame constants mu and lambda, now done in an independent method.

	(planes_interpolate_communicate_and_print): 

	  Commented out several variables that were not being used and were 
	  causing unnecessary warnings during compilation.

	(solver_run):

	- Changed the progress monitor_print to include the NL and NL2
	  timers and to avoid the single star printed for the startingstep.

	- Added statements to go into compute_nonlinear_entities and
	  compute_addforce_nl.

	- Reformat indentation.

	(print_timing_stat):

	- Added timers NL and NL2 to the final report.

	- Changed search_point to be a non-static method so that it can be
	  shared with the nonlinear module.

	(output_planes_construct_strips): 

	  Commented unused variables.

	(main): 

	  Added the statement to call nonlinear_init.

********************************************************************************

2009-11-24 Ricardo Taborda <rtaborda@cmu.edu>

	* ChangeLog: 

	  Changed the format of this file so that we keep
	  adding the changes at the top and not at the bottom.  This makes
	  it easier to read/write the latest changes.

********************************************************************************

2009-11-19 Ricardo Taborda <rtaborda@cmu.edu>

	* systemdef.mk: 

	  Added a new system configuration for running
	  locally at hooke.cml.cs.cmu.edu.  It is basically a copy of the
	  linux-mpich configuration but it is being used regularly enough to
	  make it to the repository.

********************************************************************************

2009-10-28 John Urbanic <urbanic@psc.edu>

	- All changes this checkin were to enable creation of non-solver
	  pools of PEs.  The overall theme was to eliminate references to
	  MPI_COMM_WORLD.  The numerous exceptions are for Abort calls.

	* octor/octor.h: 

	  Include <mpi.h>.  New function declaration for
	  octor_newtree() to pass in communicator.  New member of octree
	  type to hold it.

	* octor/octor.c: 

	  New communicator member added in tree type to
	  compensate for overlay on octree type. Added MPI_Comm argument to
	  com_OrchestrateExchange() to pass in communicator. Replaced all
	  MPI_COMM_WORLDs with tree comm member except for Aborts.

	* quake/forward/psolve.c: 

	  Changed Octor_newtree() call as per above. Added in comm_solver as 
	  global to replace MPI_COMM_WORLD, and did so everywhere except Abort 
	  calls and communicator creation.  Added communicator creation in 
	  main().  This uses IO_POOL_SIZE now, but will be superceded by a 
	  user input parameter.

	* quake/forward/psolve.h: 

	  Added in comm_solver global to replace MPI_COMM_WORLD. 
	  Included <mpi.h>.

	* quake/forward/output.c: 

	  Replaced all MPI_WORLD_COMMs except for Aborts.

	* quake/forward/util.h: 

	  Replaced MPI_WORLD_COMMs in HU_COND_EXEC macro.

	* quake/forward/quakesource.c: 

	  Replaced all MPI_WORLD_COMMs except for Aborts.

	* quake/forward/qmesh.c: 

	  Updated call to octor_newtree() with additional parameter as per 
	  above.

********************************************************************************

2009-09-07 John Urbanic <urbanic@psc.edu>

	General changes:

	- Added copyright header to all .c and .h files
	- Added XT5 system type to systemdef.mk
	- Added in sample user.mk for reference

********************************************************************************

2009-09-03 Julio Lopez <jclopez@ece.cmu.edu>

	* examples/simple:

	  Added new simple test scenario for aiding regression testing and 
	  debugging.

	* quake/forward/psolve.c:

	(setup_stations_data):

	  Change in the format for the stations output.

	(interpolate_station_displacements): 

	  Change in the format for the stations output.

	* version:

	  Increased version number.

	* octor/octor.c:

	- In tree_getmaxleaflevel, tree_getminleaflevel, octor_getmaxleaflevel,
	  and octor_getminleaflevel: Made parameter 'const'.

	- In octor_getmaxleaflevel and octor_getminleaflevel: Made parameter 
	  'const'.

	* quake/forward/util.h:

	- In DEF_XFREE_TYPE: Added new macro to define functions following the 
	  pattern 'xfree_type( type** v )'.  Used the macro to create xfree 
	  functios for int32_t* and char*.

	(hu_fclosep): 

	  New function declaration, see description below in util.c.

	* quake/forward/util.c (hu_fclosep): 

	  New function with parameter checks to close a file and then make 
	  the FILE pointer NULL to prevent future operations on the stream 
	  descriptor.

	* quake/forward/psolve.h:

	  Added new data types for global and local node ids.  However, these 
	  types are not widely used throughout the code yet.

	* quake/forward/psolve.c (mesh_print_stat): 

	  New implementation to print the mesh statistics that completely 
	  replaces the old implementation. The new features include:

	- Saving the stats to a file with a user chooseable file name.

	- Use of simpler MPI collective functions instead of the idiotic 
	  complex implementation using single message point-to-point 
	  primitives.

	  The purpose of saving the mesh stats to a separate file is two-fold:

	  1. It'll make it easier to go through the solver output.  The mesh 
	     stats become large when running on thousands of processors.
	  2. Having this stats in a separate file should make it easier to write 
	     scripts to process or plot the mesh data if needed.

********************************************************************************

2009-07-30 Julio Lopez <jclopez@ece.cmu.edu>

	* quake/forward/psolve.c (initparameters): 

	  Bug fix.  Corrected the parameter name and default value for the file 
	  name for mesh statistics.

********************************************************************************

2009-07-29 Julio Lopez <jclopez@ece.cmu.edu>

	* version:

	* newfile:

	  New file to keep track of a global version number ala svn.

	* tag-version.sh:

	  Helper script to manage the version file and create the appropriate 
	  tags in the repository.

	* quake/forward/quakesorce.c (filter_myForce): 

	  Bug fix.  An incorrectly place curly brace was causing the signal to 
	  be filtered multiple times, thus attenuating the signal.  After 
	  testing, the solver produces the same output as the version from
	  January 2009.

********************************************************************************

2009-05-13  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/cvm/cvm.h:

    * In cvm_query: Changed signature, removed "field" parameter and added type 
      for the payload parameter (was void*).

  quake/cvm/cvm.c:

    * In cvm_query: Changed function to match signature in the cvm.h 
      declaration.

  quake/forward/psolve.c:

    * In setrec: Modified call to cvm_query to match new signature.

    * In etree_correct_properties: Modified call to cvm_query to match new 
      signature.

    * In setrec: Added check for the return value of the cvm_query call.
      Additional test to handle corner case where all the point queries fail 
      (no point is found).

  quake/forward/qmesh.c:

    * In setrec: Modified call to cvm_query to match new signature.

  quake/forward/quakesource.c:

    * In parsedarray: Fixed bug with an incorrectly placed curly brace. Argh!

********************************************************************************

2009-05-10  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/psolve.c:

    * In solver_printstat: Changed function signature to take as parameter the
      solver structure. Changed implementation to print the scheduler stats to
      a file instead of stdout. Added configuration parameters in the config 
      file to change the default name for the file to where the schedule is
      saved.

    * In solver_printstat_to_stream: New function to support the changes 
      mentioned in solver_printstat().

    * In setup_stations_data: Changed informational message print statement 
      (printf) so only PEs that have stations print the message. Also, the 
      format of the message was changed slightly, for instance to have the new 
      line at the end of the message, not at the beginning.

  quake/forward/quakesource.c

    * In source_broadcast_parameters: Bug fix. The value of 
      theThresholdFrequency was not being set after the broadcast. Instead 
      theSourceFunctionType value was incorrectly overwritten.

********************************************************************************

2009-04-20  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/quakesource.c:
  
    * In broadcast_info: Refactored as follows:
      - renamed as broadcast_source_parameters.
      - split into separate functions according to function type:
        broadcast_point_source_parameters, broadcast_plane_source_parameters 
        and broadcast_srfh_parameters.
      In the process, cleaned up the way some parameters were passed and group 
      them in as few arrays as possible based on the data type (double, int,
      etc).

    * In isforceinprocessor: Removed global variable and replaced with local 
      variables respectively.

    * In load_myForces_with_point_source: Changed signature. The function now 
      takes as parameters the bitmap indicating whether or not a fault force 
      falls in this processor. The name of the bitmap parameter is: 
      is_force_in_processor. The function does not reference the 
      isforceinprocessor global variable.

    * In open_filescomputemyForces: Removed function.

    * In source_get_local_forces_filename: New function, see documentation in 
      source file.

    * In compute_myForces_planes: Removed reference to isforceinprocessor global
      variable.  Replaced with local var.

    * In compute_myForces_srfh: Removed reference to isforceinprocessor global 
      variable.  Replaced with local var.

    * In compute_myForces_point: Removed reference to isforceinprocessor global 
      variable.  Replaced with local var. Modified to fix bug. The PEs that did 
      not contain the source were closing the file with the source data without 
      writing a header to it. Then, psolve.c:source_init was trying to read the 
      header from the file, but the file did not contain any data, so it was 
      aborting.  The change is to make PE that contains the source write the 
      file as before.  The ones that don't contain the source do nothing, i.e., 
      they do not create a file, which in turn reduces the stress on the file 
      system (FS).  Then, in psolve.c: source_init, the PEs that don't have 
      nodes loaded by the source do not even try to open the file, this further 
      avoids innecesary stress to the FS.  This oughta be the approach for the
      other source types, but not now, will implement at a later time.

  octor/octor.c:

    * In struct hdr_t: This is a stuct used to represent a chunk in the memory 
      pool manager. Added a pointer to the next memory chunk, and moved a 
      pointer to the next available object from the tlr_t struct to the hdr_t 
      struct. Modified the appropriate functions for the memory pool to match 
      the change in the structs.

    * In com_OrchestrateExchange: Removed reference to an object pointer in the 
      tlr_t struct, since (a) the field was not used and (b) the field was moved
      to the hdr_t struct. Note: Perhaps it should be initialized to NULL in 
      hdr_t as well.

  quake/forward/commutil.*: 
  
    * New files with utility helper functions for communication.

  quake/forward/Makefile:

    * Modified to include new 'commutil' module as appropriate.

  quake/forward/util.c:

    * In hu_fopen: Changed the message in the call to solver abort to include 
    the name of the file that the caller requested to open. Before it was 
    printing "fopen(...)".

  quake/forward/util.h:

    * XMALLOC_STRING: New macro to allocate character strings.

  quake/forward/psolve.c:

    * broadcast_string: Moved to commutil.c

    * broadcast_char_array: Moved to commutil.c

    * setrec: Modified to return early when a cvm query returns a secondary 
      wave velocity (Vs) under the minimum velocity considered for the 
      simulation (theVsCut).

    * source_init: Changed behavior. The file with the sources is now ONLY 
      opened if the number of loaded sources for the local PE != 0.  This is to 
      prevent thousands of processes storming the file system with requests to 
      open a file, just to perform useless work, i.e., read a header to obtain 
      information the PE already had (the number of loaded nodes).  This is also 
      part of a bug fix for the forces generation for sources of type 'point'.

    * theSchedulePrintToFile: Changed default value to 0.

    * Other cosmetic-only miscellaneous changes, i.e., fixing line breaks and 
      indentation, etc.

  examples/test1:

    * Updated configuration files to include currently required parameters.
      Also, this example is for the LA basin.

  examples/terashake:

    * Created sample configuration for the terashake scenario.  Not tested yet.

********************************************************************************

2009-01-17  Julio Lopez  <jclopez at ece dot cmu dot edu>

  octor/octor.c:
  
    * Replaced the name of a structure and corresponding typedef to avoid name 
      clash conflict on Darwin / Mac OS X.

  systemdef.mk:

    * Minor cleanup.  Removed unneeded statements, etc.

  common.mk:
  
    * Cleanup.  Moved declaration of flags that are only used in the solver to 
      quake/forward/Makefile.

  Makefile:
  
    * Added support for including / excluding the compilation of the 
      visualization module via the ENABLE_VIZ flag, which can be overriden 
      in user.mk.

  quake/forward/funcname.h:
  
    * Added definition for _GNUC_PREREQ macro for cases where the compilation 
      is not done with gcc nor against glibc.

  quake/forward/util.h:
  
    * Added macro for fdatasync() for system that do not have this function.

  quake/forward/util.c:

    * Added getline replacement implementation for systems that do not have it.

  visualize/Makefile:
  
    * Minor cleanup.

  quake/forward/Makefile:
  
    * Cleanup and added support for including / excluding the visualization 
      module at compilation time. Moved variable declaration from common.mk to 
      here, for variables that are only related to the solver, e.g., 
      SOLVE_CPPFLAGS, etc.

********************************************************************************

2007-11-05  Julio Lopez  <jclopez at ece dot cmu dot edu>

  visualize/a3dmatrix.h: 
  
    * Changes in compiler version (to gcc 4.1.2) prompted a change.  Added a 
      declaration for ZeroMatrix().

  visualize/Makefile: 
  
    * modified the setting for the CXX variable in the LINUX-MPICH system 
      type -- C++ compiler specification.  It is now set to 'mpicxx' instead 
      of 'g++' directly.

  systemdef.mk: 
  
    * Added variables for the MPI_INCLUDE variable

********************************************************************************

2007-07-10  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/psolve.c:

    * Changes in various functions to coalesce freads and fwrites, eliminating 
      some of the inner loops and replacing the appropriate count parameter in 
      the call to fread/fwrite.

    * Changes in the checkpointing routines to achieve the following:

      - BUG FIX: compute the appropriate offset for files larger than 2GB.
      - The two displacement vectors for a PE are now contiguous in the 
        checkpointing file.  Before, they were spaced far apart.
      - replaced fopen/fclose/fread/fwrite with corresponding hu_fopen,...  
        utility routines from util.c.  These functions perform error checking 
        and abort on error by calling hu_solver_abort with a hopefully 
        meaningful message.

  quake/forward/util.c:

    * New file with utility routines. Corresponding utility functions for 
      fopen/fclose/fread/fwrite named hu_fopen, hu_fclose, etc. These functions 
      perform error checking and abort on error by calling hu_solver_abort with 
      a hopefully meaningful message.  Other miscellanous functions have been 
      moved to util.c In addition, there are a new set of functions for reading 
      variables from the configuration files.  The new functions are type safe, 
      more flexible and perform appropriate error checking.

********************************************************************************

2007-01-26  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/psolve.c:

    * In load_output_parameters: Added new flags "output_debug" and 
      "output_debug_file" to the configuration file in order to enable debug 
      statements at runtime.  Notice that in order to use these statements, the 
      DEBUG compilation flag must be defined.

    * In output_init_parameters: Fixed the following bug, theNTotal value is 
      not known in processors other than PE 0.  This affected the computation 
      of params->output_size.  Also, as a side-effect added code to update 
      theNTotal.

  quake/forward/output.c:

    * generate_local_string: New function used in the generation of per-PE 
      filenames.

    * po_init_debug: New function, it initializes per-PE debugging output files.

    * po_fini_debug: New function, it finalizes per-PE debuggin output, i.e., 
      it closes debuggin output files.

    * po_debug_msg: New support function to conditionally print out debug 
      messages.  Debug messages are enabled through a config file option 
      "output_debug".

    * po_print: New funtion to print the contents of a par_output_t structure.  
      Quite useful for debugging.

    * po_print_debug: Convenience wrapper function to print the contents of a 
      par_output_t structure to the debbuging output file when debugging is 
      enabled.

    * po_check_file_size: New function, it checks whether the size of a file 
      matches an expected size.

    * po_check_file_sizes: New function, it checks whether the displacement 
      and velocity output files have the expected file size.

    * compute_current_offset: New function, consolidated the offset computation 
      for the output in a single function.  This function is called from 
      write_displacement(), write_velocity() and po_do_output().

    * In po_gather_base_offset, po_init, write_displacement, and po_do_output: 
      Added DEBUG_MSG statements.

    * In po_create_file: Fixed bug in metadata header creation.  An incorrect 
      pointer was being passed to fwrite.

  quake/forward/q4showmeta.c:

    * q4_print_output_header: Cosmetic fix, output format aligment.

********************************************************************************

2006-08-14  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/psolve.c

    * In initparameters: New flag "output_mesh" added to the configuration file 
      to select whether or not the mesh should be stored to disk.

********************************************************************************

2006-08-11  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/psolve.c

    * compute_addforce_s_old: Removed this function, it was inside a #if 0 
      preprocessor macro.  The routine did not compile anymore.

    * In main: Moved the output stats collection further down towards the end 
      of the execution to obtain a better estimate of the end-to-end time.

  quake/forward/output.c

    * In po_collect_stats and po_collect_detailed_stats: Added IOPS stats.

********************************************************************************

2006-08-09  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/output.c:

    * Created code to collect aggregate I/O statisctics when performing 
      parallel I/O.

  quake/forward/psolve.c:

    * Changes to call the new routine that collects the parallel I/O stats (in 
      output.c).  Other minor changes, like deletion of old (commented out) 
      code.  Moved Ricardo's comments to the ChangeLog file.

********************************************************************************

2006-07-31  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/quakesource.c
  
    * In broadcast_info: There was an assumption about the layout of static 
      data that would not hold for other platform / compiler combinations. The 
      problem and fix is similar to the one applied in revision 1.3 of this 
      file, but for different static variables. Now data is copied to a 
      temporary (contiguous) buffer in the stack, and the MPI calls use that 
      buffer instead.  The specific part for this fix is in 
      broadcast_plane_info.

********************************************************************************

2006-07-29  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/psolve.h:

    * Added #ifdef __cplusplus to allow inclusion of this header file from C++ 
      code.  Also added #include <stdio.h>.

********************************************************************************

2006-07-24  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/psolve.c:

    * In parsetext: Declared parsetext() as extern (instead of static) in both 
      in psolve.c and psolve.h.

  quake/forward/quakesource.c

    * In parsetext: deleted the duplicated parsetext() code from quakesource.c.

********************************************************************************

2006-07-23  Julio Lopez  <jclopez at ece dot cmu dot edu>

  quake/forward/*:
  
    * All the files previous to this change were tagged with the following 
      label in the CVS repository: BEFORE_PARALLEL_OUTPUT_MOD.  All the files 
      after this change were tagged with the label: NEW_PARALLEL_IO.

  quake/forward/output.c:
  
    * This is the major change in this revision.  This file contains a new 
      version of the parallel output code.  The sequential code is still in 
      psolve.c Note: No new warning messages were introduced by this change 
      (on Linux with MPICH2).  Various warning messages related to the 
      previous output code were removed.

  quake/forward/output.h:

    * Created this file with relevant definitions for the parallel output.

  quake/forward/psolve.h:

    * Modified out_hdr_t structure to contain other relevant (and not relevant) 
    metadata for the 4D output files.

  quake/forward/psolve.c:

    * Added new configuration parameters for numerical.in.  These parameters 
      are read from the config file:

      - output_displacement
      - output_velocity
      - output_parallel
      - output_displacement_file = "filename"
      - output_velocity_file = "velocity"

      These parameters are backwards compatible in the following sense:

      1. If any of: simulation_displacement_out, simulation_velocity_out, 
         output_parallel, are not found in the configuration file they are 
         assumed to be zero (false / disabled) and it is not an error.  This 
         might change in the future.

      2. The respective file name parameters (output_displacement_file,
         output_velocity_file) are read only if output_parallel is 1 and 
         simulation_displacement_out and/or simulation_velocity_out are 
         respectively set.

      3. These variables are only read if output is not disabled at compile 
         time i.e., -DNO_OUTPUT.  Similarly, the output function is not called 
         if NO_OUTPUT is defined at compile time.

    * Modified parsetext(), so the TAB character is considered as a token 
      separator.

    * BTW, while I was at it, I fixed the following bug: A static vector 
      (theK3) was being freed in compute_K(), removed call to free().

********************************************************************************

2006-06-20  Ricardo Taborda  <rtaborda at cmu dot edu>

  quake/forward/psolve.c:

    * In the calculation of the stiffness matrix, submatrices K1 and K3 have 
      been merged into only K1.  This changes has reduced solving time in about 
      30%.  Old code is still around but commented out.

    * The offdiagonal stiffness matrices are no longer included in the solution 
      of the next step displacement after the changes in the algorithm to 
      backwards differences.  Julio removed comments and commented out variable 
      definition on 2006-08-09.

    * Two new global variables have been introduced.  These thresehold values 
      are read from the numerical file and define the top values the target 
      damping and the Vp over Vs ratio can have.  The former affects only the 
      damping terms but the later affects in general the stiffnesses matrices 
      and its influence in the damping terms.  This as shown improvements in 
      the delta_t of up to double to what was needed before without 
      compromising the solution.

    * In solver_init: Old code for damping on calculating lambda: 
      lambda = edata->rho * edata->Vp * edata->Vp - 2 * mu.

    * Changes in the n_t structure due to changes on the solving algorithm 
      because of the damping problems.

********************************************************************************

2006-06-01  Ricardo Taborda  <rtaborda at cmu dot edu>
	* quake/forward/psolve.c:

	- A new algorithm for the solution of the next time step
	displacement has been implemented in order to have control over
	the damping influence in the solution.  Old computations are still
	in the code but commented out.  We will keep old code segments
	until we're done with the Terashake runs.

	- After reviewing different things regarding the damping and
	delta_t problems, two new input parameter have been included in
	the numericals file theThresholdDamping and theThresholdVpVs.

	- In local_init: double_message has been changed to 16 to include
	the two new threshold values.

********************************************************************************

2006-05-31  Ricardo Taborda  <rtaborda at cmu dot edu>

	* quake/forward/psolve.c (solver_init):

	- Old code for damping since the np structure has changed:
	np->mass2x += M * 2;

	- Old code for damping K_diag is no longer needed. The C matrix is
	no longer constructed as one by itself and the np structure has
	changed as well as the way its components are calculated:

	double K_diag;

	K_diag = (edata->edgesize / 9) * (mu * theK1[j][j].f[axis][axis]
	+ lambda * theK2[j][j].f[axis][axis]
	+ mu * theK3[j][j].f[axis][axis]);

	C = a * M + b * K_diag;

	np->mplus[axis] += (theDeltaT / 2 * C);
	np->mminus[axis] -= (theDeltaT / 2 * C);

**************************************************************************

	S T O P ! ! !

	Please add the entries describing your changes at the TOP of this
	file after the instructions and before the last entry.

**************************************************************************