Cloned SEACAS for EXODUS library with extra build files for internal package management.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

3730 lines
118 KiB

2 years ago
<!-------- @HEADER
!
! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
! Zoltan Toolkit for Load-balancing, Partitioning, Ordering and Coloring
! Copyright 2012 Sandia Corporation
!
! Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
! the U.S. Government retains certain rights in this software.
!
! Redistribution and use in source and binary forms, with or without
! modification, are permitted provided that the following conditions are
! met:
!
! 1. Redistributions of source code must retain the above copyright
! notice, this list of conditions and the following disclaimer.
!
! 2. Redistributions in binary form must reproduce the above copyright
! notice, this list of conditions and the following disclaimer in the
! documentation and/or other materials provided with the distribution.
!
! 3. Neither the name of the Corporation nor the names of the
! contributors may be used to endorse or promote products derived from
! this software without specific prior written permission.
!
! THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY
! EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
! IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
! PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE
! CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
! EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
! PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
! PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
! LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
! NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
! SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
!
! Questions? Contact Karen Devine kddevin@sandia.gov
! Erik Boman egboman@sandia.gov
!
! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
! @HEADER
------->
<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<meta name="GENERATOR" content="Mozilla/4.7 [en] (X11; U; SunOS 5.6 sun4m) [Netscape]">
<meta name="sandia.approval_type" content="formal">
<meta name="sandia.approved" content="SAND2007-4748W">
<meta name="author" content="Zoltan PI">
<title>Zoltan User's Guide: General Zoltan Query Functions</title>
</head>
<body bgcolor="#FFFFFF">
<div align=right><b><i><a href="ug.html">Zoltan User's Guide</a>&nbsp;
|&nbsp; <a href="ug_query_mig.html">Next</a>&nbsp; |&nbsp; <a href="ug_query.html">Previous</a></i></b></div>
<h2>
<a NAME="Load-Balancing Query Functions"></a>General Zoltan Query Functions</h2>
The following registered functions are used by various Zoltan algorithms
in the Zoltan library. No single algorithm uses all the query functions;
the <a href="ug_alg.html">algorithm descriptions</a> indicate which query
functions are required by individual algorithms. These query functions should
NOT contain interprocessor communication.
<blockquote><a href="#General Functions">Object ID Functions</a>
<blockquote><b><a href="#ZOLTAN_NUM_OBJ_FN">ZOLTAN_NUM_OBJ_FN</a></b>
<br><b><a href="#ZOLTAN_OBJ_LIST_FN">ZOLTAN_OBJ_LIST_FN</a></b>
<br><b><a href="#ZOLTAN_FIRST_OBJ_FN">ZOLTAN_FIRST_OBJ_FN</a></b> (deprecated)
<br><b><a href="#ZOLTAN_NEXT_OBJ_FN">ZOLTAN_NEXT_OBJ_FN</a></b> (deprecated)
<br><b><a href="#ZOLTAN_PART_MULTI_FN">ZOLTAN_PART_MULTI_FN</a></b> or
<b><a href="#ZOLTAN_PART_FN">ZOLTAN_PART_FN</a></b>
</blockquote>
<a href="#Geometry-based Functions">Geometry-Based Functions</a>
<blockquote><b><a href="#ZOLTAN_NUM_GEOM_FN">ZOLTAN_NUM_GEOM_FN</a></b>
<br><b><a href="#ZOLTAN_GEOM_MULTI_FN">ZOLTAN_GEOM_MULTI_FN</a></b> or
<b><a href="#ZOLTAN_GEOM_FN">ZOLTAN_GEOM_FN</a></b>
</blockquote>
<a href="#Graph-based Functions">Graph-Based Functions</a>
<blockquote>
<b><a href="#ZOLTAN_NUM_EDGES_MULTI_FN">ZOLTAN_NUM_EDGES_MULTI_FN</a></b> or
<b><a href="#ZOLTAN_NUM_EDGES_FN">ZOLTAN_NUM_EDGES_FN</a></b>
<br><b><a href="#ZOLTAN_EDGE_LIST_MULTI_FN">ZOLTAN_EDGE_LIST_MULTI_FN</a></b> or
<b><a href="#ZOLTAN_EDGE_LIST_FN">ZOLTAN_EDGE_LIST_FN</a></b>
</blockquote>
<a href="#Hypergraph-based Functions">Hypergraph-Based Functions</a>
<blockquote>
<b><a href="#ZOLTAN_HG_SIZE_CS_FN">ZOLTAN_HG_SIZE_CS_FN</a></b>
<b><br><a href="#ZOLTAN_HG_CS_FN">ZOLTAN_HG_CS_FN</a></b>
<b><br><a href="#ZOLTAN_HG_SIZE_EDGE_WTS_FN">ZOLTAN_HG_SIZE_EDGE_WTS_FN</a></b>
<b><br><a href="#ZOLTAN_HG_EDGE_WTS_FN">ZOLTAN_HG_EDGE_WTS_FN</a></b>
<b><br><a href="#ZOLTAN_NUM_FIXED_OBJ_FN">ZOLTAN_NUM_FIXED_OBJ_FN</a></b>
<b><br><a href="#ZOLTAN_FIXED_OBJ_LIST_FN">ZOLTAN_FIXED_OBJ_LIST_FN</a></b>
</blockquote>
<a href="#Tree-based Functions">Tree-Based Functions</a>
<blockquote><b><a href="#ZOLTAN_NUM_COARSE_OBJ_FN">ZOLTAN_NUM_COARSE_OBJ_FN</a></b>
<br><b><a href="#ZOLTAN_COARSE_OBJ_LIST_FN">ZOLTAN_COARSE_OBJ_LIST_FN</a></b>
<br><b><a href="#ZOLTAN_FIRST_COARSE_OBJ_FN">ZOLTAN_FIRST_COARSE_OBJ_FN</a></b>
<br><b><a href="#ZOLTAN_NEXT_COARSE_OBJ_FN">ZOLTAN_NEXT_COARSE_OBJ_FN</a></b>
<br><b><a href="#ZOLTAN_NUM_CHILD_FN">ZOLTAN_NUM_CHILD_FN</a></b>
<br><b><a href="#ZOLTAN_CHILD_LIST_FN">ZOLTAN_CHILD_LIST_FN</a></b>
<br><b><a href="#ZOLTAN_CHILD_WEIGHT_FN">ZOLTAN_CHILD_WEIGHT_FN</a></b>
</blockquote>
<a href="#Hierarchical Partitioning Functions">Hierarchical Partitioning Functions</a>
<blockquote><b><a href="#ZOLTAN_HIER_NUM_LEVELS_FN">ZOLTAN_HIER_NUM_LEVELS_FN</a></b>
<br><b><a href="#ZOLTAN_HIER_PART_FN">ZOLTAN_HIER_PART_FN</a></b>
<br><b><a href="#ZOLTAN_HIER_METHOD_FN">ZOLTAN_HIER_METHOD_FN</a></b>
</blockquote>
</blockquote>
<hr>
<!------------------------------------------------------------------------->
<h2>
<a NAME="General Functions"></a><hr>Object ID Functions</h2>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_NUM_OBJ_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_NUM_OBJ_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;*<i>ierr</i>);</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Num_Obj</i></b>(<i>data, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Num_Obj&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_NUM_OBJ_FN</b> query function returns the number
of objects that are currently assigned to the processor.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NUM_OBJ_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>The number of objects that are assigned to the processor.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_OBJ_LIST_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void<b> ZOLTAN_OBJ_LIST_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_ids</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_ids</i>,
int&nbsp;<i>wgt_dim</i>,
float&nbsp;*<i>obj_wgts</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Obj_List</i></b>(<i>data, num_gid_entries, num_lid_entries, global_ids, local_ids,
wgt_dim, obj_wgts, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: global_ids&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: local_ids&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: wgt_dim&nbsp;
<br>REAL(Zoltan_FLOAT), INTENT(OUT), DIMENSION(*) :: obj_wgts&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_OBJ_LIST_FN</b> query function fills two (three
if weights are used) arrays with information about the objects currently
assigned to the processor. Both arrays are allocated (and subsequently
freed) by Zoltan; their size is determined by a call to a <b><a href="#ZOLTAN_NUM_OBJ_FN">ZOLTAN_NUM_OBJ_FN</a></b>
query function to get the array size. For many algorithms, either a <b>ZOLTAN_OBJ_LIST_FN</b>
query function or a <b><a href="#ZOLTAN_FIRST_OBJ_FN">ZOLTAN_FIRST_OBJ_FN</a></b>/<b><a href="#ZOLTAN_NEXT_OBJ_FN">ZOLTAN_NEXT_OBJ_FN</a></b>
query-function pair must be registered;
however, both query options need not
be provided. The <b>ZOLTAN_OBJ_LIST_FN</b> is preferred for efficiency.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_OBJ_LIST_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_ids</i></td>
<td>Upon return, an array of unique global IDs for all objects assigned to the processor.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_ids</i></td>
<td>Upon return, an array of local IDs, the meaning of which can be determined by the
application, for all objects assigned to the processor. (Optional.)</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>wgt_dim</i></td>
<td>The number of weights associated with an object (typically 1), or 0
if weights are not requested. This value is set through the parameter
<a href="ug_param.html#OBJ_WEIGHT_DIM">OBJ_WEIGHT_DIM</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>obj_wgts</i></td>
<td>Upon return, an array of object weights. Weights for object <i>i</i> are stored in <i>obj_wgts[(i-1)*wgt_dim:i*wgt_dim-1]</i>.&nbsp;
If <i>wgt_dim</i>=0, the return value of <i>obj_wgts</i> is undefined and
may be NULL.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_FIRST_OBJ_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_FIRST_OBJ_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>first_global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>first_local_id</i>,
int&nbsp;<i>wgt_dim</i>,
float&nbsp;*<i>first_obj_wgt</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_First_Obj</i></b>(<i>data, num_gid_entries, num_lid_entries, first_global_id, first_local_id,
wgt_dim, first_obj_wgt, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_First_Obj&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: first_global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: first_local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: wgt_dim&nbsp;
<br>REAL(Zoltan_FLOAT), INTENT(OUT), DIMENSION(*) :: first_obj_wgt&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">
(Deprecated) A <b>ZOLTAN_FIRST_OBJ_FN</b> query function initializes an
iteration over objects assigned to the processor. It returns the global
and local IDs of the first object on the processor. Subsequent calls to
a <b><a href="#ZOLTAN_NEXT_OBJ_FN">ZOLTAN_NEXT_OBJ_FN</a></b> query function iterate
over and return other objects assigned to the processor.
For many algorithms, either a <b><a href="#ZOLTAN_OBJ_LIST_FN">ZOLTAN_OBJ_LIST_FN</a></b>
query function or a <b>ZOLTAN_FIRST_OBJ_FN</b><i>/</i><b><a href="#ZOLTAN_NEXT_OBJ_FN">ZOLTAN_NEXT_OBJ_FN</a></b>
query-function pair must be registered;
however, both query options need not
be provided.
The <b><a href="#ZOLTAN_OBJ_LIST_FN">ZOLTAN_OBJ_LIST_FN</a></b>) is preferred
for efficiency.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_FIRST_OBJ_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; first_global_id</i></td>
<td>The returned value of the global ID for the first object; the value
is ignored if there are no objects.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; first_local_id</i></td>
<td>The returned value of the local ID for the first object; the value
is ignored if there are no objects.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>wgt_dim</i></td>
<td>The number of weights associated with an object (typically 1), or 0
if weights are not requested. This value is set through the parameter
<a href="ug_param.html#OBJ_WEIGHT_DIM">OBJ_WEIGHT_DIM</a>.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>first_obj_wgt</i></td>
<td>Upon return, the first object's weights; an array of length <i>wgt_dim. </i>Undefined if <i>wgt_dim=0.</i></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; 1</td>
<td>If <i>first_global_id</i> and <i>first_local_id</i> contain valid IDs
of the first object.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; 0</td>
<td>If no objects are available.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_NEXT_OBJ_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_NEXT_OBJ_FN</b>
(void&nbsp;*<i>&nbsp;data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>next_global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>next_local_id</i>,
int&nbsp;<i>wgt_dim,</i>
float&nbsp;*<i>next_obj_wgt</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Next_Obj</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id, next_global_id,
next_local_id, wgt_dim, next_obj_wgt, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Next_Obj&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: next_global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: next_local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: wgt_dim&nbsp;
<br>REAL(Zoltan_FLOAT), INTENT(OUT), DIMENSION(*) :: next_obj_wgt&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">
(Deprecated) A <b>ZOLTAN_NEXT_OBJ_FN</b> query function is an iterator
function which, when given an object assigned to the processor, returns
the next object assigned to the processor. The first object of the iteration
is provided by a <b><a href="#ZOLTAN_FIRST_OBJ_FN">ZOLTAN_FIRST_OBJ_FN</a></b>
query function.
For many algorithms, either a <b><a href="#ZOLTAN_OBJ_LIST_FN">ZOLTAN_OBJ_LIST_FN</a></b>
query function or a <b><a href="#ZOLTAN_FIRST_OBJ_FN">ZOLTAN_FIRST_OBJ_FN</a></b>/<b>ZOLTAN_NEXT_OBJ_FN</b>
query-function pair must be registered;
however, both query options need not
be provided.
The <b><a href="#ZOLTAN_OBJ_LIST_FN">ZOLTAN_OBJ_LIST_FN</a></b>) is preferred
for efficiency.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NEXT_OBJ_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_id</i></td>
<td>The global ID of the previous object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_id</i></td>
<td>The local ID of the previous object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; next_global_id</i></td>
<td>The returned value of the global ID for the next object; the value
is ignored if there are no more objects.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; next_local_id</i></td>
<td>The returned value of the local ID for the next object; the value is
ignored if there are no more objects.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>wgt_dim</i></td>
<td>The number of weights associated with an object (typically 1), or 0
if weights are not requested. This value is set through the parameter
<a href="ug_param.html#OBJ_WEIGHT_DIM">OBJ_WEIGHT_DIM</a>.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>next_obj_wgt</i></td>
<td>Upon return, the next object's weights; an array of length <i>wgt_dim. </i>Undefined if <i>wgt_dim=0.</i></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; 1</td>
<td>If <i>next_global_id</i> and <i>next_local_id</i> contain valid IDs
of the next object.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; 0</td>
<td>If no more objects are available.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_PART_MULTI_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_PART_MULTI_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
int&nbsp;<i>num_obj</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_ids</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_ids</i>,
int&nbsp;*<i>parts</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Part_Multi</i></b>(<i>data, num_gid_entries, num_lid_entries, num_obj, global_ids, local_ids,
ierr</i>)
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries, num_obj
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_ids
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_ids
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: parts
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_PART_MULTI_FN</b> query function returns a
list of
parts to which
given objects are currently assigned.
If a <b>ZOLTAN_PART_MULTI_FN</b>
or <b><a href="#ZOLTAN_PART_FN">ZOLTAN_PART_FN</a></b> is not
registered, Zoltan assumes the part numbers are the processor
number of the owning processor.
Valid part numbers are non-negative
integers.
This function is used when parameter <a href="ug_alg.html#REMAP">REMAP</a>=1
and for certain methods with parameter
<a href="ug_alg.html#LB_APPROACH">LB_APPROACH</a>=<i>Repartition</i>.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_PART_MULTI_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_obj</i></td>
<td>The number of object IDs in arrays <i>global_ids</i> and <i>local_ids</i>.
</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>global_ids</i></td>
<td>The global IDs of the objects for which the part numbers should be
returned.</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>local_ids</i></td>
<td>The local IDs of the objects for which the part numbers should be
returned. (Optional.)</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>parts</i></td>
<td>Upon return, an array of part numbers corresponding to the global
and local IDs.
</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<p>
<hr><hr>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_PART_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_PART_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Part</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id,
ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Part&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_PART_FN</b> query function returns the
part to which
a given object is currently assigned.
If a <b>ZOLTAN_PART_FN</b>
or <b><a href="#ZOLTAN_PART_MULTI_FN">ZOLTAN_PART_MULTI_FN</a></b>
is not
registered, Zoltan assumes the part numbers are the processor
number of the owning processor.
Valid part numbers are non-negative
integers.
This function is used when parameter <a href="ug_alg.html#REMAP">REMAP</a>=1
and for certain methods with parameter
<a href="ug_alg.html#LB_APPROACH">LB_APPROACH</a>=<i>Repartition</i>.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_PART_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>global_id</i></td>
<td>The global ID of the object for which the part number should be
returned.</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>local_id</i></td>
<td>The local ID of the object for which the part number should be
returned.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td VALIGN=TOP><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>The part number for the object identified by <i>global_id</i> and
<i>local_id</i>.</td>
</tr>
</table>
<p>
<hr>
<!------------------------------------------------------------------------->
<h2>
<a NAME="Geometry-based Functions"></a><hr>Geometry-based Functions</h2>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_NUM_GEOM_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_NUM_GEOM_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Num_Geom</i></b>(<i>data, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Num_Geom&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_NUM_GEOM_FN</b> query function returns the number
of values needed to express the geometry of an object. For example, for
a two-dimensional mesh-based application, (<i>x</i>,<i>y</i>) coordinates
are needed to describe an object's geometry; thus the <b>ZOLTAN_NUM_GEOM_FN</b>
query function should return the value of two. For a similar three-dimensional
application, the return value should be three.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NUM_GEOM_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td VALIGN=TOP><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>The number of values needed to express the geometry of an object.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_GEOM_MULTI_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_GEOM_MULTI_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
int&nbsp;<i>num_obj</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_ids</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_ids</i>,
int&nbsp;<i>num_dim</i>,
double&nbsp;*<i>geom_vec</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Geom_Multi</i></b>(<i>data, num_gid_entries,
num_lid_entries, num_obj, global_ids, local_ids, num_dim, geom_vec, ierr</i>)
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_obj, num_dim
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_ids
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_ids
<br>REAL(Zoltan_DOUBLE), INTENT(OUT), DIMENSION(*) :: geom_vec
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_GEOM_MULTI FN</b> query function returns a
vector of
geometry values for a list of given objects. The geometry vector is allocated by
Zoltan to be of size <i>num_obj</i> * <i>num_dim</i>; its format is described
below.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_GEOM_MULTI_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_obj</i></td>
<td>The number of object IDs in arrays <i>global_ids</i> and <i>local_ids</i>.
</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_ids</i></td>
<td>Array of global IDs of objects whose geometry values should be returned.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_ids</i></td>
<td>Array of local IDs of objects whose geometry values should be returned. (Optional.)</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>num_dim</i></td>
<td>Number of coordinate entries per object (typically 1, 2, or 3).
</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>geom_vec</i></td>
<td>Upon return, an array containing geometry values.
For object <i>i</i>
(specified by <i>global_ids[i*num_gid_entries]</i> and
<i>local_ids[i*num_lid_entries]</i>, <i>i=0,1,...,num_obj-1</i>),
coordinate values should be stored
in <i>geom_vec[i*num_dim:(i+1)*num_dim-1]</i>.
</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<br>&nbsp;
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_GEOM_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_GEOM_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
double&nbsp;*<i>geom_vec</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Geom</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id, geom_vec,
ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_id&nbsp;
<br>REAL(Zoltan_DOUBLE), INTENT(OUT), DIMENSION(*) :: geom_vec&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_GEOM_FN</b> query function returns a vector of
geometry values for a given object. The geometry vector is allocated by
Zoltan to be of the size returned by a <b><a href="#ZOLTAN_NUM_GEOM_FN">ZOLTAN_NUM_GEOM_FN</a></b>
query function.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_GEOM_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_id</i></td>
<td>The global ID of the object whose geometry values should be returned.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_id</i></td>
<td>The local ID of the object whose geometry values should be returned.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>geom_vec</i></td>
<td>Upon return, an array containing geometry values.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<br>&nbsp;
<hr>
<!------------------------------------------------------------------------->
<h2>
<a NAME="Graph-based Functions"></a><hr>Graph-based Functions</h2>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_NUM_EDGES_MULTI_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_NUM_EDGES_MULTI_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
int&nbsp;<i>num_obj</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_ids</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_ids</i>,
int&nbsp;*<i>num_edges</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Num_Edges_Multi</i></b>(<i>data, num_gid_entries, num_lid_entries, num_obj, global_ids, local_ids, num_edges,
ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Num_Edges&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries, num_obj
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_ids
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_ids
<br>INTEGER(Zoltan_INT), INTENT(OUT),DIMENSION(*) :: num_edges
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_NUM_EDGES_MULTI_FN</b> query function returns
the number
of edges in the communication graph of the application for each object in a
list of objects. That is, for each object in the <i>global_ids/local_ids</i>
arrays, the number of objects with which the given object must share
information is returned.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NUM_EDGES_MULTI_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_obj</i></td>
<td>The number of object IDs in arrays <i>global_ids</i> and <i>local_ids</i>.
</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>global_ids</i></td>
<td>Array of global IDs of objects whose number of edges should be returned.</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>local_ids</i></td>
<td>Array of local IDs of objects whose number of edges should be returned. (Optional.)</td>
</tr>
<tr>
<td valign=top>&nbsp;&nbsp;&nbsp; <i>num_edges</i></td>
<td>Upon return, an array containing numbers of edges.
For object <i>i</i>
(specified by <i>global_ids[i*num_gid_entries]</i> and
<i>local_ids[i*num_lid_entries]</i>, <i>i=0,1,...,num_obj-1</i>),
the number of edges should be stored
in <i>num_edges[i]</i>.
</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_NUM_EDGES_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_NUM_EDGES_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Num_Edges</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id,
ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Num_Edges&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_NUM_EDGES_FN</b> query function returns the number
of edges for a given object in the communication graph of the application
(i.e., the number of objects with which the given object must share information).
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NUM_EDGES_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>global_id</i></td>
<td>The global ID of the object for which the number of edges should be
returned.</td>
</tr>
<tr>
<td>&nbsp;&nbsp;&nbsp; <i>local_id</i></td>
<td>The local ID of the object for which the number of edges should be
returned.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td VALIGN=TOP><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>The number of edges for the object identified by <i>global_id</i> and
<i>local_id</i>.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_EDGE_LIST_MULTI_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_EDGE_LIST_MULTI_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
int&nbsp;<i>num_obj</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_ids</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_ids</i>,
int&nbsp;*<i>num_edges</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>nbor_global_id</i>,
int&nbsp;*<i>nbor_procs</i>,
int&nbsp;<i>wgt_dim</i>,
float&nbsp;*<i>ewgts</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Edge_List_Multi</i></b>(<i>data, num_gid_entries, num_lid_entries, num_obj, global_ids, local_ids, num_edges,
nbor_global_id, nbor_procs, wgt_dim, ewgts, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries, num_obj
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_ids
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_ids
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: num_edges
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: nbor_global_id
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: nbor_procs
<br>INTEGER(Zoltan_INT), INTENT(IN) :: wgt_dim
<br>REAL(Zoltan_FLOAT), INTENT(OUT), DIMENSION(*) :: ewgts
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_EDGE_LIST_MULTI_FN</b> query function returns lists
of global IDs, processor IDs, and optionally edge weights for objects sharing
edges with objects specified in the <i>global_ids</i> input array; objects share
edges when they must share information
with other objects. The arrays for the returned neighbor lists are
allocated by Zoltan; their size is determined by a
calls to <b><a href="#ZOLTAN_NUM_EDGES_MULTI_FN">ZOLTAN_NUM_EDGES_MULTI_FN</a></b> or <b><a href="#ZOLTAN_NUM_EDGES_FN">ZOLTAN_NUM_EDGES_FN</a></b> query functions.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_EDGE_LIST_MULTI_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_obj</i></td>
<td>The number of object IDs in arrays <i>global_ids</i> and <i>local_ids</i>.
</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_ids</i></td>
<td>Array of global IDs of objects whose edge lists should be returned.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_ids</i></td>
<td>Array of local IDs of objects whose edge lists should be returned. (Optional.)</td>
</tr>
<tr>
<td valign=top>&nbsp;&nbsp;&nbsp; <i>num_edges</i></td>
<td>An array containing numbers of edges for each object in <i>global_ids</i>.
For object <i>i</i>
(specified by <i>global_ids[i*num_gid_entries]</i> and
<i>local_ids[i*num_lid_entries]</i>, <i>i=0,1,...,num_obj-1</i>),
the number of edges is stored
in <i>num_edges[i]</i>.
</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>nbor_global_id</i></td>
<td>Upon return, an array of global IDs of objects sharing edges with the
objects specified in <i>global_ids</i>.
For object <i>i</i>
(specified by <i>global_ids[i*num_gid_entries]</i> and
<i>local_ids[i*num_lid_entries]</i>, <i>i=0,1,...,num_obj-1</i>),
edges are stored in
<i>nbor_global_id[sum*num_gid_entries]</i> to
<i>nbor_global_id[(sum+num_edges[i])*num_gid_entries-1]</i>, where
<i>sum</i> = the sum of <i>num_edges[j]</i> for <i>j=0,1,...,i-1</i>.
</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>nbor_procs</i></td>
<td>Upon return, an array of processor IDs that identifies where the neighboring
objects reside. For neighboring object <i>i</i> (stored in
<i>nbor_global_id[i*num_gid_entries]</i>), the processor owning the neighbor
is stored in <i>nbor_procs[i]</i>.
</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>wgt_dim</i></td>
<td>The number of weights associated with an edge (typically 1), or 0 if
edge weights are not requested. This value is set through the parameter
<a href="ug_param.html#EDGE_WEIGHT_DIM">EDGE_WEIGHT_DIM</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>ewgts</i></td>
<td>Upon return, an array of edge weights, where <i>ewgts[i*wgt_dim:(i+1)*wgt_dim-1]</i>
<br>corresponds to the weights for the <i>i</i>th edge. If <i>wgt_dim=0</i>, the
return value of <i>ewgts</i> is undefined and may be NULL.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<br>&nbsp;
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_EDGE_LIST_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_EDGE_LIST_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>nbor_global_id</i>,
int&nbsp;*<i>nbor_procs</i>,
int&nbsp;<i>wgt_dim</i>,
float&nbsp;*<i>ewgts</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Edge_List</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id,
nbor_global_id, nbor_procs, wgt_dim, ewgts, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: nbor_global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: nbor_procs&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: wgt_dim&nbsp;
<br>REAL(Zoltan_FLOAT), INTENT(OUT), DIMENSION(*) :: ewgts&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_EDGE_LIST_FN</b> query function returns lists
of global IDs, processor IDs, and optionally edge weights for objects sharing
an edge with a given object (i.e., objects that must share information
with the given object). The arrays for the returned neighbor lists are
allocated by Zoltan; their size is determined by a
call to <b><a href="#ZOLTAN_NUM_EDGES_MULTI_FN">ZOLTAN_NUM_EDGES_MULTI_FN</a></b> or
<b><a href="#ZOLTAN_NUM_EDGES_FN">ZOLTAN_NUM_EDGES_FN</a></b> query functions.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_EDGE_LIST_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_id</i></td>
<td>The global ID of the object for which an edge list should be returned.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_id</i></td>
<td>The local ID of the object for which an edge list should be returned.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>nbor_global_id</i></td>
<td>Upon return, an array of global IDs of objects sharing edges with the
given object.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>nbor_procs</i></td>
<td>Upon return, an array of processor IDs that identifies where the neighboring
objects reside.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>wgt_dim</i></td>
<td>The number of weights associated with an edge (typically 1), or 0 if
edge weights are not requested. This value is set through the parameter
<a href="ug_param.html#EDGE_WEIGHT_DIM">EDGE_WEIGHT_DIM</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>ewgts</i></td>
<td>Upon return, an array of edge weights, where <i>ewgts[i*wgt_dim:(i+1)*wgt_dim-1]</i>
<br>corresponds to the weights for the <i>i</i>th edge. If <i>wgt_dim=0</i>, the
return value of <i>ewgts </i>is undefined and may be NULL.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<br>&nbsp;
<hr>
<!------------------------------------------------------------------------->
<h2>
<a NAME="Hypergraph-based Functions"></a><hr>Hypergraph-based Functions</h2>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_HG_SIZE_CS_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_HG_SIZE_CS_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;*<i>num_lists</i>,
int&nbsp;*<i>num_pins</i>,
int&nbsp;*<i>format</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_HG_Size_CS</i></b>(<i>data, num_lists, num_pins,
format, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: num_lists&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: num_pins&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: format&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A hypergraph
can be supplied to the Zoltan library in
one of two compressed storage formats. Although hypergraphs
are often used to represent the structure of sparse matrices,
the Zoltan/PHG terminology is purely in terms of vertices and hyperedges
(not rows or columns). The two compressed formats are analogous to CRS and CCS
for matrices. In compressed hyperedge format
(<b>ZOLTAN_COMPRESSED_EDGE</b>)
a list of global hyperedge IDs is provided. Then a single
list of the hypergraph pins, is provided.
A pin is the connection between a vertex and a hyperedge
(corresponds to a nonzero in a sparse matrix).
Pins do not have separate IDs but are rather
identified by the
global ID of the vertex containing the pin, and implicitly also
by the hyperedge ID. An example is provided below.
<p>
The other format
is compressed vertex (<b>ZOLTAN_COMPRESSED_VERTEX</b>). In this
format a list of vertex global IDs is provided. Then a
list of pins ordered by vertex and then by hyperedge is provided.
The pin ID in this case is the global ID of the hyperedge in which the
pin appears. In both formats, an array must be provided pointing to
the start in the list of pins where each hyperedge or vertex begins.
<p>
The purpose of this query function is to tell Zoltan in which format the
application will supply the hypergraph, how many vertices and hyperedges
there will be, and how many pins. The actual hypergraph is supplied
with a query function of the type
<b><a href="#ZOLTAN_HG_CS_FN">ZOLTAN_HG_CS_FN_TYPE</a></b>.
<br> <br>
This query function is required by all applications using the hypergraph methods
of Zoltan (unless they are using the
<a href="#Graph-based Functions">graph-based</a>
functions with hypergraph code instead).
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_HG_SIZE_CS_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lists</i></td>
<td>Upon return, the number of vertices (if using compressed vertex storage) or
hyperedges (if using compressed hyperedge storage) that will be supplied to Zoltan by the application process.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_pins</i></td>
<td>Upon return, the number of pins (connections between vertices and hyperedges) that will be
supplied to Zoltan by the application process.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>format</i></td>
<td>Upon return, the format in which the application process will provide the
hypergraph to Zoltan. The options are
<b>ZOLTAN_COMPRESSED_EDGE</b> and <b>ZOLTAN_COMPRESSED_VERTEX</b>.
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_HG_CS_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_HG_CS_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_vtx_edge</i>,
int&nbsp;<i>num_pins</i>,
int&nbsp;<i>format</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>vtxedge_GID</i>,
int&nbsp;*<i>vtxedge_ptr</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>pin_GID</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_HG_CS</i></b>(<i>data,
num_gid_entries, num_vtx_edge, num_pins, format,
vtxedge_GID, vtxedge_ptr, pin_GID, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_vtx_edge, num_pins, format
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: vtxedge_GID
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: vtxedge_ptr
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: pin_GID
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_HG_CS_FN</b> returns
a hypergraph in a compressed storage (CS) format. The size and format
of the data to be returned must have been supplied to Zoltan using a
<b><a href="#ZOLTAN_HG_SIZE_CS_FN">ZOLTAN_HG_SIZE_CS_FN_TYPE</a></b>
function.
<br> <br>
When a hypergraph is distributed across multiple processes, Zoltan
expects that all processes share a consistent global numbering
scheme for hyperedges and vertices. Also, no two
processes should return the same pin (matrix non-zero) in this
query function. (Pin ownership is unique.)
<br> <br>
This query function is required by all applications using the hypergraph methods
of Zoltan (unless they are using the
<a href="#Graph-based Functions">graph-based</a>
functions with hypergraph code instead).
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_HG_CS_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_vtx_edge</i></td>
<td>The number of global IDs
that is expected to appear on return in <i>vtxedge_GID</i>. This may
correspond to either vertices or (hyper-)edges. </td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_pins</i></td>
<td>The number of pins
that is expected to appear on return in <i>pin_GID</i>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>format</i></td>
<td>If <i>format</i> is <b>ZOLTAN_COMPRESSED_EDGE</b>, Zoltan expects
that hyperedge global IDs will be returned in <i>vtxedge_GID</i>,
and that vertex global IDs will be returned in
<i>pin_GIDs</i>. If it is <b>ZOLTAN_COMPRESSED_VERTEX</b>, then
vertex global IDs are expected to be returned in <i>vtxedge_GID</i> and
hyperedge global IDs are expected to be returned in <i>pin_GIDs</i>.
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>vtxedge_GID</i></td>
<td>Upon return, a list of <i>num_vtx_edge</i> global IDs.
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>vtxedge_ptr</i></td>
<td>Upon return, this array contains <i>num_vtx_edge</i> integers
such that the number of pins specified for hyperedge <i>j</i>
(if format is <b>ZOLTAN_COMPRESSED_EDGE</b>) or vertex <i>j</i>
(if format is <b>ZOLTAN_COMPRESSED_VERTEX</b>)
is
<i>vtxedge_ptr[j+1]-vtxedge_ptr[j]</i>.
If <i>format</i> is <b>ZOLTAN_COMPRESSED_EDGE</b>,
<i>vtxedge_ptr[j]*num_gid_entries</i> is the index into the
array <i>pin_GID</i> where edge <i>j</i>'s pins
(vertices belonging to edge <i>j</i>)
begin;
if <i>format</i> is <b>ZOLTAN_COMPRESSED_VERTEX</b>,
<i>vtxedge_ptr[j]*num_gid_entries</i> is the index into the
array <i>pin_GID</i> where vertex <i>j</i>'s pins (edges to which
vertex <i>j</i> belongs)
begin.
Array indices begin at zero.
</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>pin_GID</i></td>
<td>Upon return, a list of <i>num_pins</i> global IDs. This is the list
of the pins contained in the hyperedges or vertices
listed in <i>vtxedge_GID</i>.
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<br> <br>
<center>
<table border=5 cellspacing=10 cellpadding=10>
<caption>Example</caption>
<tr>
<td>
<table border=2>
<tr>
<th colspan=1></th><th colspan=5>vertex</th>
</tr>
<tr>
<th>hyperedge</th> <th>10</th> <th>20</th> <th>30</th> <th>40</th> <th>50</th>
</tr>
<tr>
<th>1</th> <td>0</td> <td>0</td> <td>X</td> <td>X</td> <td>0</td>
</tr>
<tr>
<th>2</th> <td>0</td> <td>X</td> <td>X</td> <td>0</td> <td>0</td>
</tr>
<tr>
<th>3</th> <td>X</td> <td>0</td> <td>0</td> <td>0</td> <td>X</td>
</tr>
</table>
</td>
<td>
Compressed hyperedge storage:<br><br>
vtxedge_GID = {1, 2, 3}<br>
vtxedge_ptr = {0, 2, 4}<br>
pin_GID = {30, 40, 20, 30, 10, 50}<br><br>
Compressed vertex storage:<br><br>
vtxedge_GID = {10, 20, 30, 40, 50}<br>
vtxedge_ptr = {0, 1, 2, 4, 5}<br>
pin_GID = {3, 2, 1, 2, 1, 3}
</td>
</tr>
</table>
</center>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_HG_SIZE_EDGE_WTS_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_HG_SIZE_EDGE_WTS_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;*<i>num_edges</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_HG_Size_Edge_Wts</i></b>(<i>data, num_edges, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: num_edges
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_HG_SIZE_EDGE_WTS_FN</b> returns
the number of hyperedges for which a process will supply edge weights.
The number of weights per hyperedge was supplied by the application
with the <a href="ug_param.html#EDGE_WEIGHT_DIM">EDGE_WEIGHT_DIM</a>
parameter. The actual edge weights will be supplied with a
<b><a href="#ZOLTAN_HG_EDGE_WTS_FN">ZOLTAN_HG_EDGE_WTS_FN_TYPE</a></b>
function.
<br> <br>
This query function is not required. If no hyperedge weights are
supplied, Zoltan will assume every hyperedge has weight 1.0.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_HG_SIZE_EDGE_WTS_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_edges</i></td>
<td>Upon return, the number of hyperedges for which edge weights
will be supplied.
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_HG_EDGE_WTS_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_HG_EDGE_WTS_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
int&nbsp;<i>num_edges</i>,
int&nbsp;<i>edge_weight_dim</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>edge_GID</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>edge_LID</i>,
float &nbsp;*<i>edge_weight</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_HG_Edge_Wts</i></b>(<i>data, num_gid_entries, num_lid_entries, num_edges, edge_weight_dim, edge_GID, edge_LID, edge_weight,
ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries, num_edges, edge_weight_dim&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: edge_GID&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: edge_LID&nbsp;
<br>REAL(Zoltan_FLOAT), INTENT(OUT), DIMENSION(*) :: edge_weight&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_HG_EDGE_WTS_FN</b> returns
edges weights for a set of hypergraph edges. The number of weights
supplied for each hyperedge should equal the value of the
<a href="ug_param.html#EDGE_WEIGHT_DIM">EDGE_WEIGHT_DIM</a>
parameter. In the case of a hypergraph which is distributed across
multiple processes, if more than one process supplies edge weights for the
same hyperedge, the different edge weights will be resolved according
to the value of the
<a href="ug_alg_phg.html#PHG_EDGE_WEIGHT_OPERATION">PHG_EDGE_WEIGHT_OPERATION</a>
parameter.
<br><br>
This query function is not required. If no hyperedge weights are
supplied, Zoltan will assume every hyperedge has weight 1.0.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_HG_EDGE_WTS_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_edges</i></td>
<td>The number of hyperedges for which edge weights
should be supplied in the <i>edge_weight</i> array.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>edge_weight_dim</i></td>
<td>The number of weights which should be supplied for each hyperedge. This
is also the value of the
<a href="ug_param.html#EDGE_WEIGHT_DIM">EDGE_WEIGHT_DIM</a> parameter.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>edge_GID</i></td>
<td>Upon return, this array should contain the global IDs of the
<i>num_edges</i> hyperedges for which the application is supplying
edge weights.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>edge_LID</i></td>
<td>Upon return, this array can optionally contain the local IDs of the
<i>num_edges</i> hyperedges for which the application is supplying
edge weights.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>edge_weight</i></td>
<td>Upon return, this array should contain the weights for each
edge listed in the <i>edge_GID</i>. If <i>edge_weight_dim</i> is
greater than one, all weights for one hyperedge are listed before
the weights for the next hyperedge are listed.
</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<br>&nbsp;
<p>
<!------------------------------------------------------------------------->
<hr>
<a NAME="ZOLTAN_NUM_FIXED_OBJ_FN"></a>
<hr>
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_NUM_FIXED_OBJ_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Num_Fixed_Obj</i></b>(<i>data, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Num_Fixed_Obj&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
</td>
</tr>
</table>
<hr WIDTH="100%">
A <b>ZOLTAN_NUM_FIXED_OBJ_FN </b> returns
the number of objects on a given processor fixed to particular parts.
These objects will not be assigned to a part other than the user specified
part by the parallel hypergraph algorithm.
<br><br>
This query function is not required. If it is not defined, all
objects are candidates for migration to new parts.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NUM_FIXED_OBJ_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td VALIGN=TOP><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>The number of objects on this processor that are to be fixed to a specific part.</td>
</tr>
</table>
<br>&nbsp;
<hr>
<!------------------------------------------------------------------------->
<a NAME="ZOLTAN_FIXED_OBJ_LIST_FN"></a>
<hr>
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void<b> ZOLTAN_FIXED_OBJ_LIST_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_fixed_obj</i>,
int&nbsp;<i>num_gid_entries</i>,
<b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>fixed_gids</i>,
int&nbsp;*<i>fixed_parts</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Fixed_Obj_List</i></b>(<i>data, num_fixed_obj, num_gid_entries, fixed_gids, fixed_parts, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_fixed_obj&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: fixed_gids&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: fixed_parts&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_FIXED_OBJ_LIST_FN</b> query function fills two
arrays with information about the objects that should not be moved from
their user assigned parts.
These arrays are allocated (and subsequently
freed) by Zoltan; their size is determined by a call to a <b><a href="#ZOLTAN_NUM_FIXED_OBJ_FN">ZOLTAN_NUM_FIXED_OBJ_FN</a></b>
query function to get the array size.
<p>
A process should only fix the part of objects that it <i>owns</i>, that is,
objects that it had previously supplied in a
<a href=#ZOLTAN_OBJ_LIST_FN>ZOLTAN_OBJ_LIST_FN</a> query function. It is
an error to list the global ID of an object owned by another process.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_FIXED_OBJ_LIST_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_fixed_obj</i></td>
<td>The number of objects you will list in the two output arrays.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; fixed_gids</i></td>
<td>Upon return, an array of unique global IDs for all objects assigned to the processor which are to be fixed to a part.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>fixed_parts</i></td>
<td>Upon return, an array of part numbers, one for each object listed
in the global ID array. These objects will not be migrated from this assigned
part.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<p>
<hr>
<p>
<!------------------------------------------------------------------------->
<h2>
<a NAME="Tree-based Functions"></a><hr>Tree-based Functions</h2>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_NUM_COARSE_OBJ_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_NUM_COARSE_OBJ_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Num_Coarse_Obj</i></b>(<i>data, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Num_Coarse_Obj&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_NUM_COARSE_OBJ_FN</b> query function returns
the number of objects (elements) in the initial coarse grid.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NUM_COARSE_OBJ_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td VALIGN=TOP><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>The number of objects in the coarse grid.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_COARSE_OBJ_LIST_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_COARSE_OBJ_LIST_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_ids</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_ids</i>,
int&nbsp;*<i>assigned</i>,
int&nbsp;*<i>num_vert</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>vertices</i>,
int&nbsp;*<i>in_order</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>in_vertex</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>out_vertex</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Coarse_Obj_List</i></b>(<i>data, num_gid_entries, num_lid_entries, global_ids, local_ids,
assigned, num_vert, vertices, in_order, in_vertex, out_vertex, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: global_ids&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: local_ids&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: assigned, num_vert, vertices,
in_vertex, out_vertex&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: in_order, ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_COARSE_OBJ_LIST_FN</b> query function returns
lists of global IDs, local IDs, vertices, and order information for all
objects (elements) of the initial coarse grid. The vertices are designated
by a global ID such that if two elements share a vertex then the same ID
designates that vertex in both elements and on all processors.
The user may choose to provide the order in which the elements should be
traversed or have Zoltan determine the order. If the user provides the
order, then entry and exit vertices for a path through the elements may
also be provided. The arrays for the returned values are allocated by
Zoltan; their size is determined by a call to a <b><a href="#ZOLTAN_NUM_COARSE_OBJ_FN">ZOLTAN_NUM_COARSE_OBJ_FN</a></b>
query function.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_COARSE_OBJ_LIST_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_ids</i></td>
<td>Upon return, an array of global IDs of all objects in the coarse grid.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_ids</i></td>
<td>Upon return, an array of local IDs of all objects in the coarse grid. (Optional.)</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>assigned</i></td>
<td>Upon return, an array of integers indicating whether or not each object
is currently assigned to this processor. A value of 1 indicates it is assigned
to this processor; a value of 0 indicates it is assigned to some other
processor. For elements that have been refined, it is ignored unless weights
are assigned to interior nodes of the tree.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; num_vert</i></td>
<td>Upon return, an array containing the number of vertices for each object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; vertices</i></td>
<td>Upon return, an array of global IDs of the vertices of each object. If the
number of vertices for objects 0 through <i>i</i>-1 is <i>N</i>, then the
vertices for object <i>i</i> are in <i>vertices</i>[<i>N*num_gid_entries:
(N+num_vert</i>[<i>i</i>]<i>)*num_gid_entries</i>]</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; in_order</i></td>
<td>Upon return, 1 if the user is providing the objects in the order in
which they should be traversed, or 0 if Zoltan should determine the order.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; in_vertex</i></td>
<td>Upon return, an array of global IDs of the vertices through which to enter
each element in the user provided traversal. It is required only if the
user is providing the order for the coarse grid objects (i.e., <i>in_order</i>==1)
and allowing Zoltan to select the order of the children in at least one
invocation of
<b><a href="#ZOLTAN_CHILD_LIST_FN">ZOLTAN_CHILD_LIST_FN</a></b>.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; out_vertex</i></td>
<td>Upon return, an array of global IDs of the vertex through which to exit each
element in the user provided traversal. The same provisions hold as for
<i>in_vertex</i>.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_FIRST_COARSE_OBJ_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_FIRST_COARSE_OBJ_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
int&nbsp;*<i>assigned</i>,
int&nbsp;*<i>num_vert</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>vertices</i>,
int&nbsp;*<i>in_order</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>in_vertex</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>out_vertex</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_First_Coarse_Obj</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id,
assigned, num_vert, vertices, in_order, in_vertex, out_vertex, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_First_Coarse_Obj&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: assigned, num_vert, in_order, ierr&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: vertices, in_vertex, out_vertex&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_FIRST_COARSE_OBJ_FN</b> query function initializes
an iteration over the objects of the initial coarse grid. It returns the
global ID, local ID, vertices, and order information for the first object
(element) of the initial coarse grid. Subsequent calls to a <b><a href="#ZOLTAN_NEXT_COARSE_OBJ_FN">ZOLTAN_NEXT_COARSE_OBJ_FN</a></b>
iterate over and return other objects from the coarse grid. The vertices
are designated by a global ID such that if two elements share a vertex then
the same ID designates that vertex in both elements and on all processors.
The user may choose to provide the order in which the elements
should be traversed, or have Zoltan determine the order. If the user provides
the order, then entry and exit vertices for a path through the elements
may also be provided.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_FIRST_COARSE_OBJ_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_ids</i></td>
<td>Upon return, the global ID of the first object in the coarse grid.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_ids</i></td>
<td>Upon return, the local ID of the first object in the coarse grid. (Optional.)</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>assigned</i></td>
<td>Upon return, an integer indicating whether or not this object is currently
assigned to this processor. A value of 1 indicates it is assigned to this
processor; a value of 0 indicates it is assigned to some other processor.
For elements that have been refined, it is ignored unless weights are assigned
to interior nodes of the tree.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; num_vert</i></td>
<td>Upon return, the number of vertices for this object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; vertices</i></td>
<td>Upon return, an array of global IDs of the vertices of this object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; in_order</i></td>
<td>Upon return, 1 if the user is providing the objects in the order in
which they should be traversed, or 0 if Zoltan should determine the order.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; in_vertex</i></td>
<td>Upon return, the vertex through which to enter this element in the
user provided traversal. It is required only if the user is providing the
order for the coarse grid objects (i.e., <i>in_order</i>==1) and allowing
Zoltan to select the order of the children in at least one invocation of
<b><a href="#ZOLTAN_CHILD_LIST_FN">ZOLTAN_CHILD_LIST_FN</a></b>.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; out_vertex</i></td>
<td>Upon return, the vertex through which to exit this element in the user
provided traversal. The same provisions hold as for
<i>in_vertex</i>.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; 1</td>
<td>If <i>global_id</i> and <i>local_id</i> contain valid IDs of the first
object in the coarse grid.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; 0</td>
<td>If no coarse grid is available.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_NEXT_COARSE_OBJ_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_NEXT_COARSE_OBJ_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>next_global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>next_local_id</i>,
int&nbsp;*<i>assigned</i>,
int&nbsp;*<i>num_vert</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>vertices</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>in_vertex</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>out_vertex</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Next_Coarse_Obj</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id,
next_global_id, next_local_id,
assigned, num_vert, vertices, in_vertex, out_vertex, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Next_Coarse_Obj&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: next_global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: next_local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: assigned, num_vertex, ierr&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: vertices, in_vertex, out_vertex&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_NEXT_COARSE_OBJ_FN</b> query function is an iterator
function that returns the next object in the initial coarse grid.
The first object
of the iteration is provided by a <b><a href="#ZOLTAN_FIRST_COARSE_OBJ_FN">ZOLTAN_FIRST_COARSE_OBJ_FN</a></b>
query function.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NEXT_COARSE_OBJ_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_id</i></td>
<td>The global ID of the previous object in the coarse grid.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_id</i></td>
<td>The local ID of the previous object in the coarse grid.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; next_global_id</i></td>
<td>Upon return, the global ID of the next object in the coarse grid.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; next_local_id</i></td>
<td>Upon return, the local ID of the next object in the coarse grid.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>assigned</i></td>
<td>Upon return, an integer indicating whether or not this object is currently
assigned to this processor. A value of 1 indicates it is assigned to this
processor; a value of 0 indicates it is assigned to some other processor.
For elements that have been refined, it is ignored unless weights are assigned
to interior nodes of the tree.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; num_vert</i></td>
<td>Upon return, the number of vertices for this object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; vertices</i></td>
<td>Upon return, an array of global IDs of the vertices of this object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; in_vertex</i></td>
<td>Upon return, the vertex through which to enter this element in the
user provided traversal. It is required only if the user is providing the
order for the coarse grid objects (i.e., <i>in_order</i>==1) and allowing
Zoltan to select the order of the children in at least one invocation of
<b><a href="#ZOLTAN_CHILD_LIST_FN">ZOLTAN_CHILD_LIST_FN</a></b>.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; out_vertex</i></td>
<td>Upon return, the vertex through which to exit this element in the user
provided traversal. The same provisions hold as for
<i>in_vertex</i>.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; 1</td>
<td>If <i>global_id</i> and <i>local_id</i> contain valid IDs of the next
object in the coarse grid.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; 0</td>
<td>If no more objects are available.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_NUM_CHILD_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_NUM_CHILD_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Num_Child</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id,
ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Num_Child&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_NUM_CHILD_FN</b> query function returns the number
of children of the element with the given global and local IDs. If the
element has not been refined, the number of children is 0.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_NUM_CHILD_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_id</i></td>
<td>The global ID of the object for which the number of children is requested.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_id</i></td>
<td>The local ID of the object for which the number of children is requested.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td VALIGN=TOP><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>The number of children.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_CHILD_LIST_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_CHILD_LIST_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>parent_gid</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>parent_lid</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>child_gids</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>child_lids</i>,
int&nbsp;*<i>assigned</i>,
int&nbsp;*<i>num_vert</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>vertices</i>,
<b>ZOLTAN_REF_TYPE</b>&nbsp;*<i>ref_type</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>in_vertex</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>out_vertex</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Child_List</i></b>(<i>data, num_gid_entries,
num_lid_entries, parent_gid, parent_lid,
child_gids, child_lids, assigned, num_vert, vertices, ref_type, in_vertex,
out_vertex, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: parent_gid&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: parent_lid&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: child_gids&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: child_lids&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT), DIMENSION(*) :: assigned, num_vert, vertices,
in_vertex, out_vertex&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ref_type, ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_CHILD_LIST_FN</b> query function returns lists
of global IDs, local IDs, vertices, and order information for all children
of a refined element. The vertices are designated by a global ID
such that if two elements share a vertex then the same ID designates that
vertex in both elements and on all processors. The user may choose to
provide the order in which the children should be traversed, or have Zoltan
determine the order based on the type of element refinement used to create
the children. If the user provides the order, then entry and exit vertices
for a path through the elements may also be provided. The arrays for the
returned values are allocated by Zoltan; their size
is determined by a call to a <b><a href="#ZOLTAN_NUM_CHILD_FN">ZOLTAN_NUM_CHILD_FN</a></b>
query function.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_CHILD_LIST_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; parent_gid</i></td>
<td>The global ID of the object whose children are requested.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; parent_lid</i></td>
<td>The local ID of the object whose children are requested.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; child_gids</i></td>
<td>Upon return, an array of global IDs of all children of this object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; child_lids</i></td>
<td>Upon return, an array of local IDs of all children of this object. (Optional.)</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>assigned</i></td>
<td>Upon return, an array of integers indicating whether or not each child
is currently assigned to this processor. A value of 1 indicates it is assigned
to this processor; a value of 0 indicates it is assigned to some other
processor. For children that have been further refined, it is ignored unless
weights are assigned to interior nodes of the tree.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; num_vert</i></td>
<td>Upon return, an array containing the number of vertices for each object.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; vertices</i></td>
<td>Upon return, an array of global IDs of the vertices of each object. If the
number of vertices for objects 0 through <i>i</i>-1 is <i>N</i>, then the
vertices for object <i>i</i> are in <i>vertices</i>[<i>N*num_gid_entries:
(N+num_vert</i>[<i>i</i>]<i>)*num_gid_entries</i>]</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ref_type</i></td>
<td>Upon return, a value indicating what type of refinement was used to
create the children. This determines how the children will be ordered.
The values currently supported are:</td>
</tr>
<tr>
<td></td>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>ZOLTAN_TRI_BISECT</i> Bisection of triangles.</td>
</tr>
<tr>
<td></td>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>ZOLTAN_QUAD_QUAD</i> Quadrasection of quadrilaterals.</td>
</tr>
<tr>
<td></td>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>ZOLTAN_HEX3D_OCT</i> Octasection of hexahedra.</td>
</tr>
<tr>
<td></td>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>ZOLTAN_OTHER_REF</i> All other forms of
refinement.</td>
</tr>
<tr>
<td></td>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>ZOLTAN_IN_ORDER</i> Traverse the children
in the order in which they are provided.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; in_vertex</i></td>
<td>Upon return, an array of global IDs of the vertex through which to enter
each element in the user provided traversal. It is required only if the
user is providing the order for the children of this element (i.e.,
<i>ref_type</i>==<i>ZOLTAN_IN_ORDER</i>)
but does not provide the order for the children of at least one of those
children.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; out_vertex</i></td>
<td>Upon return, an array of global IDs of the vertex through which to exit each
element in the user provided traversal. The same provisions hold as for
<i>in_vertex</i>.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_CHILD_WEIGHT_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void<b> ZOLTAN_CHILD_WEIGHT_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>num_gid_entries</i>,
int&nbsp;<i>num_lid_entries</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>global_id</i>,
<a href="ug_usage.html#Data Types for Object IDs"><b>ZOLTAN_ID_PTR</b></a>&nbsp;<i>local_id</i>,
int&nbsp;<i>wgt_dim</i>,
float&nbsp;*<i>obj_wgt</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Child_Weight</i></b>(<i>data, num_gid_entries, num_lid_entries, global_id, local_id,
wgt_dim, obj_wgt, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: num_gid_entries, num_lid_entries&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: global_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN), DIMENSION(*) :: local_id&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: wgt_dim&nbsp;
<br>REAL(Zoltan_FLOAT), INTENT(OUT), DIMENSION(*) :: obj_wgt&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_CHILD_WEIGHT_FN</b> query function returns the
weight of an object. Interior nodes of the refinement tree as well as the
leaves are allowed to have weights.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_CHILD_WEIGHT_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_gid_entries</i></td>
<td>The number of array entries used to describe a single
global ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_GID_ENTRIES">NUM_GID_ENTRIES</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>num_lid_entries</i></td>
<td>The number of array entries used to describe a single
local ID.&nbsp; This value is the maximum value over all processors of
the parameter <a href="ug_param.html#NUM_LID_ENTRIES">NUM_LID_ENTRIES</a>. (It should be zero if local ids are not used.)</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; global_id</i></td>
<td>The global ID of the object whose weight is requested.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; local_id</i></td>
<td>The local ID of the object whose weight is requested.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>wgt_dim</i></td>
<td>The number of weights associated with an object (typically 1), or 0
if weights are not requested. This value is set through the parameter
<a href="ug_param.html#OBJ_WEIGHT_DIM">OBJ_WEIGHT_DIM</a>.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>obj_wgt</i></td>
<td>Upon return, an array containing the object's weights. If <i>wgt_dim</i>=0, the return
value of <i>obj_wgts</i> is undefined and may be NULL.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<!------------------------------------------------------------------------->
<h2>
<hr>
<a NAME="Hierarchical Partitioning Functions"></a>
<hr>
Hierarchical Partitioning Functions (used only by method <B>HIER</B>)</h2>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_HIER_NUM_LEVELS_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_HIER_NUM_LEVELS_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Hier_Num_Levels</i></b>(<i>data, nbor_proc, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Hier_Num_Levels
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_HIER_NUM_LEVELS_FN</b> query function
returns, for the calling processor, the number of levels of hierarchy
for hierarchical load balancing.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_HIER_NUM_LEVELS_FN_TYPE</b></td>
</tr>
<tr>
<td VALIGN=TOP><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td VALIGN=TOP><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>the number of levels of balancing hierarchy for method HIER.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_HIER_PART_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef int <b>ZOLTAN_HIER_PART_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>level</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>FUNCTION <b><i>Get_Hier_Part</i></b>(<i>data, level, ierr</i>)&nbsp;
<br>INTEGER(Zoltan_INT) :: Get_Hier_Part
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) ::: level&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_HIER_PART_FN</b> query function gets the part number to be used for the given level of a hierarchical balancing procedure.
<br>&nbsp;
<table WIDTH="100%" NOSAVE >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_HIER_PART_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>level</i></td>
<td>The level of a hierarchical balancing for which the part ID is requested.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
<tr>
<td VALIGN=TOP><b>Returned Value:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; int</td>
<td>The part number the process is to compute for this level.</td>
</tr>
</table>
<p>
<!------------------------------------------------------------------------->
<hr WIDTH="100%">
<a NAME="ZOLTAN_HIER_METHOD_FN"></a>
<hr WIDTH="100%">
<table WIDTH="100%" NOSAVE >
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>C and C++:</td>
<td WIDTH="85%">
typedef void <b>ZOLTAN_HIER_METHOD_FN</b>
(void&nbsp;*<i>data</i>,
int&nbsp;<i>level</i>,
struct Zoltan_Struct *&nbsp;<i>zz</i>,
int&nbsp;*<i>ierr</i>);&nbsp;</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP WIDTH="15%" NOSAVE>FORTRAN:</td>
<td>SUBROUTINE <b><i>Get_Hier_Method</i></b>(<i>data, level, zz, ierr</i>)&nbsp;
<br>&lt;<i>type-data</i>>, INTENT(IN) :: data&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(IN) :: level&nbsp;
<br>TYPE(Zoltan_Struct), INTENT(IN) :: zz&nbsp;
<br>INTEGER(Zoltan_INT), INTENT(OUT) :: ierr&nbsp;
<p>&lt;<i>type-data</i>> can be any of INTEGER(Zoltan_INT), DIMENSION(*) or
REAL(Zoltan_FLOAT), DIMENSION(*) or REAL(Zoltan_DOUBLE), DIMENSION(*) or TYPE(Zoltan_User_Data_<i>x</i>)
where <i>x</i> is 1, 2, 3 or 4. See the section on
<a href="ug_fortran_api.html#fortran ug api query">Fortran
query functions</a> for an explanation.&nbsp;</td>
</tr>
</table>
<hr WIDTH="100%">A <b>ZOLTAN_HIER_METHOD_FN</b> query function provides
to the calling process the Zoltan_Struct
to be used to guide the partitioning and load balancing at the
given level in the hierarchy. This Zoltan_Struct can be passed to
Zoltan_Set_Param to set load balancing parameters for this level
in the hierarchical balancing.
<br>&nbsp;
<table WIDTH="100%" >
<tr>
<td VALIGN=TOP WIDTH="20%"><b>Function Type:</b></td>
<td WIDTH="80%"><b>ZOLTAN_HIER_METHOD_FN_TYPE</b></td>
</tr>
<tr>
<td><b>Arguments:</b></td>
<td></td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; data</i></td>
<td>Pointer to user-defined data.</td>
</tr>
<tr NOSAVE>
<td VALIGN=TOP NOSAVE>&nbsp;&nbsp;&nbsp; <i>level</i></td>
<td>Level in the hierarchy being considered.</td>
</tr>
<tr>
<td VALIGN=TOP>&nbsp;&nbsp;&nbsp; <i>zz</i></td>
<td>Zoltan_Struct to use to set parameters.</td>
</tr>
<tr>
<td VALIGN=TOP><i>&nbsp;&nbsp;&nbsp; ierr</i></td>
<td>Error code to be set by function.</td>
</tr>
</table>
<hr><hr>
<p>
<hr WIDTH="100%">[<a href="ug.html">Table of Contents</a>&nbsp; |
<a href="ug_query_mig.html">Next:&nbsp;
Migration Query Functions</a>&nbsp; |&nbsp; <a href="ug_query.html">Previous:&nbsp;
Application-Registered Query Functions</a>&nbsp; |&nbsp; <a href="https://www.sandia.gov/general/privacy-security/index.html">Privacy and Security</a>]
</body>
</html>