diff --git a/CMakeLists.txt b/CMakeLists.txt
index 8edfcc9d0c..2a73790b5f 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -111,7 +111,7 @@ include ( SuiteSparsePolicy )
 
 if ( SUITESPARSE_USE_SYSTEM_GRAPHBLAS )
     list ( REMOVE_ITEM SUITESPARSE_ENABLE_PROJECTS "graphblas" )
-    find_package ( GraphBLAS 9.3.0 REQUIRED )
+    find_package ( GraphBLAS 9.3.1 REQUIRED )
 else ( )
     if ( "lagraph" IN_LIST SUITESPARSE_ENABLE_PROJECTS )
         # LAGraph requires GraphBLAS.
@@ -137,7 +137,7 @@ endif ( )
 
 if ( SUITESPARSE_USE_SYSTEM_UMFPACK )
     list ( REMOVE_ITEM SUITESPARSE_ENABLE_PROJECTS "umfpack" )
-    find_package ( UMFPACK 6.3.3 REQUIRED )
+    find_package ( UMFPACK 6.3.4 REQUIRED )
 else ( )
     if ( "paru" IN_LIST SUITESPARSE_ENABLE_PROJECTS )
         # ParU requires UMFPACK.
@@ -168,7 +168,7 @@ endif ( )
 
 if ( SUITESPARSE_USE_SYSTEM_AMD )
     list ( REMOVE_ITEM SUITESPARSE_ENABLE_PROJECTS "amd" )
-    find_package ( AMD 3.3.2 REQUIRED )
+    find_package ( AMD 3.3.3 REQUIRED )
 else ( )
     if ( "cholmod" IN_LIST SUITESPARSE_ENABLE_PROJECTS
             OR "ldl" IN_LIST SUITESPARSE_ENABLE_PROJECTS
@@ -184,7 +184,7 @@ endif ( )
 
 if ( SUITESPARSE_USE_SYSTEM_COLAMD )
     list ( REMOVE_ITEM SUITESPARSE_ENABLE_PROJECTS "colamd" )
-    find_package ( COLAMD 3.3.3 REQUIRED )
+    find_package ( COLAMD 3.3.4 REQUIRED )
 else ( )
     if ( "cholmod" IN_LIST SUITESPARSE_ENABLE_PROJECTS
             OR "spex" IN_LIST SUITESPARSE_ENABLE_PROJECTS )
@@ -198,7 +198,7 @@ endif ( )
 
 if ( SUITESPARSE_USE_SYSTEM_CAMD )
     list ( REMOVE_ITEM SUITESPARSE_ENABLE_PROJECTS "camd" )
-    find_package ( CAMD 3.3.2 REQUIRED )
+    find_package ( CAMD 3.3.3 REQUIRED )
 else ( )
     if ( CHOLMOD_CAMD AND "cholmod" IN_LIST SUITESPARSE_ENABLE_PROJECTS )
         # CHOLMOD can optionally use CAMD.
@@ -211,7 +211,7 @@ endif ( )
 
 if ( SUITESPARSE_USE_SYSTEM_CCOLAMD )
     list ( REMOVE_ITEM SUITESPARSE_ENABLE_PROJECTS "ccolamd" )
-    find_package ( CCOLAMD 3.3.3 REQUIRED )
+    find_package ( CCOLAMD 3.3.4 REQUIRED )
 else ( )
     if ( CHOLMOD_CAMD AND "cholmod" IN_LIST SUITESPARSE_ENABLE_PROJECTS )
         # CHOLMOD can optionally use CCOLAMD.
@@ -224,7 +224,7 @@ endif ( )
 
 if ( SUITESPARSE_USE_SYSTEM_SUITESPARSE_CONFIG )
     list ( REMOVE_ITEM SUITESPARSE_ENABLE_PROJECTS "suitesparse_config" )
-    find_package ( SuiteSparse_config 7.8.0 REQUIRED )
+    find_package ( SuiteSparse_config 7.8.2 REQUIRED )
 else ( )
     if ( "mongoose" IN_LIST SUITESPARSE_ENABLE_PROJECTS
             OR "amd" IN_LIST SUITESPARSE_ENABLE_PROJECTS
diff --git a/ChangeLog b/ChangeLog
index 8ac73d6169..d8a7ed8b1f 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,32 @@
+Aug 20, 2024: version 7.8.2
+
+    * LAGraph 1.1.4: bug fix for LAGraph_MMWrite when matrix is dense
+    * SPEX 3.2.1: release date revised, sync with primary SPEX repo
+    * SuiteSparse_config, Example: modified to reflect the release of
+        LAGraph 1.1.4 and SPEX 3.2.1
+    * ParU 0.3.0: added parameter to ParU_Get
+    * Package versions in this release: (* denotes a new version)
+        SuiteSparse_config 7.8.2    *
+        AMD 3.3.3
+        BTF 2.3.2
+        CAMD 3.3.3
+        CCOLAMD 3.3.4
+        CHOLMOD 5.3.0
+        COLAMD 3.3.4
+        CSparse 4.3.2
+        CXSparse 4.4.1
+        Example 1.8.2               *
+        GraphBLAS 9.3.1
+        KLU 2.3.4
+        LDL 3.3.2
+        LAGraph 1.1.4               *
+        SuiteSparse_Mongoose 3.3.4
+        ParU 0.3.0                  *
+        RBio 4.3.3
+        SPEX 3.2.1                  *
+        SPQR 4.3.4
+        UMFPACK 6.3.4
+
 Aug 12, 2024: version 7.8.1
 
     * GraphBLAS 9.3.1: bug fix in creation of JIT package
diff --git a/Example/CMakeLists.txt b/Example/CMakeLists.txt
index 5114449843..e72965d349 100644
--- a/Example/CMakeLists.txt
+++ b/Example/CMakeLists.txt
@@ -53,10 +53,10 @@ message ( STATUS "MY prefix path:    ${CMAKE_PREFIX_PATH}" )
 #-------------------------------------------------------------------------------
 
 # cmake inserts the date and version number into Include/my.h:
-set ( MY_DATE "Aug 12, 2024" )
+set ( MY_DATE "Aug 20, 2024" )
 set ( MY_VERSION_MAJOR 1 )
 set ( MY_VERSION_MINOR 8 )
-set ( MY_VERSION_PATCH 1 )
+set ( MY_VERSION_PATCH 2 )
 
 message ( STATUS "Building MY library version: v"
     ${MY_VERSION_MAJOR}.
@@ -87,7 +87,7 @@ project ( my
 #-------------------------------------------------------------------------------
 
 # look for all SuiteSparse packages:
-find_package ( SuiteSparse_config 7.8.1 REQUIRED )
+find_package ( SuiteSparse_config 7.8.2 REQUIRED )
 find_package ( AMD 3.3.3 REQUIRED )
 find_package ( BTF 2.3.2 REQUIRED )
 find_package ( CAMD 3.3.3 REQUIRED )
@@ -99,11 +99,11 @@ find_package ( GraphBLAS 9.3.1 )
 find_package ( KLU 2.3.4 REQUIRED )
 find_package ( KLU_CHOLMOD 2.3.4 REQUIRED )
 find_package ( LDL 3.3.2 REQUIRED )
-find_package ( LAGraph 1.1.3 )
+find_package ( LAGraph 1.1.4 )
 find_package ( SuiteSparse_Mongoose 3.3.4 REQUIRED )
-find_package ( ParU 0.2.0 REQUIRED )
+find_package ( ParU 0.3.0 REQUIRED )
 find_package ( RBio 4.3.3 REQUIRED )
-find_package ( SPEX 3.2.0 REQUIRED )    # requires GMP and MPFR
+find_package ( SPEX 3.2.1 REQUIRED )    # requires GMP and MPFR
 find_package ( SPQR 4.3.4 REQUIRED )
 find_package ( UMFPACK 6.3.4 REQUIRED )
 
diff --git a/Example/Include/my.h b/Example/Include/my.h
index 50def62c35..8cd74c74e5 100644
--- a/Example/Include/my.h
+++ b/Example/Include/my.h
@@ -11,10 +11,10 @@
 // file, since it is constructed from Config/my.h.in by cmake.
 
 // version and date for example user library
-#define MY_DATE "Aug 12, 2024"
+#define MY_DATE "Aug 20, 2024"
 #define MY_MAJOR_VERSION 1
 #define MY_MINOR_VERSION 8
-#define MY_PATCH_VERSION 1
+#define MY_PATCH_VERSION 2
 
 #ifdef __cplusplus
 extern "C" {
diff --git a/Example/Include/my_internal.h b/Example/Include/my_internal.h
index 6a9f1fee4f..7a10b994dc 100644
--- a/Example/Include/my_internal.h
+++ b/Example/Include/my_internal.h
@@ -14,8 +14,8 @@
 
 // SuiteSparse include files for C/C++:
 #include "SuiteSparse_config.h"
-#if !defined (SUITESPARSE__VERSION) || SUITESPARSE__VERSION < SUITESPARSE__VERCODE(7,8,1)
-#error "This library requires SuiteSparse_config 7.8.1 or later"
+#if !defined (SUITESPARSE__VERSION) || SUITESPARSE__VERSION < SUITESPARSE__VERCODE(7,8,2)
+#error "This library requires SuiteSparse_config 7.8.2 or later"
 #endif
 
 #include "amd.h"
@@ -63,8 +63,8 @@
 
 #if ! defined (NO_LAGRAPH)
     #include "LAGraph.h"
-    #if SUITESPARSE__VERCODE(LAGRAPH_VERSION_MAJOR,LAGRAPH_VERSION_MINOR,LAGRAPH_VERSION_UPDATE) < SUITESPARSE__VERCODE(1,1,3)
-    #error "This library requires LAGraph 1.1.3 or later"
+    #if SUITESPARSE__VERCODE(LAGRAPH_VERSION_MAJOR,LAGRAPH_VERSION_MINOR,LAGRAPH_VERSION_UPDATE) < SUITESPARSE__VERCODE(1,1,4)
+    #error "This library requires LAGraph 1.1.4 or later"
     #endif
 #endif
 
@@ -84,8 +84,8 @@
 #endif
 
 #include "SPEX.h"
-#if !defined (SPEX__VERSION) || SPEX__VERSION < SUITESPARSE__VERCODE(3,2,0)
-#error "This library requires SPEX 3.2.0 or later"
+#if !defined (SPEX__VERSION) || SPEX__VERSION < SUITESPARSE__VERCODE(3,2,1)
+#error "This library requires SPEX 3.2.1 or later"
 #endif
 
 #include "SuiteSparseQR_C.h"
diff --git a/LAGraph/CMakeLists.txt b/LAGraph/CMakeLists.txt
index b8685d669b..90a9350697 100644
--- a/LAGraph/CMakeLists.txt
+++ b/LAGraph/CMakeLists.txt
@@ -39,10 +39,10 @@
 cmake_minimum_required ( VERSION 3.20 ) # LAGraph can be built stand-alone
 
 # version of LAGraph
-set ( LAGraph_DATE "Mar 22, 2024" )
+set ( LAGraph_DATE "Aug 20, 2024" )
 set ( LAGraph_VERSION_MAJOR 1 CACHE STRING "" FORCE )
 set ( LAGraph_VERSION_MINOR 1 CACHE STRING "" FORCE )
-set ( LAGraph_VERSION_SUB   3 CACHE STRING "" FORCE )
+set ( LAGraph_VERSION_SUB   4 CACHE STRING "" FORCE )
 
 message ( STATUS "Building LAGraph version: v"
     ${LAGraph_VERSION_MAJOR}.
diff --git a/LAGraph/ChangeLog b/LAGraph/ChangeLog
index 995f8d43d6..758ed824e3 100644
--- a/LAGraph/ChangeLog
+++ b/LAGraph/ChangeLog
@@ -1,3 +1,7 @@
+Aug 20, 2024: version 1.1.4
+
+    * fix MMWrite when matrix is dense
+
 Mar 22, 2024: version 1.1.3
 
     * minor updates to build system
diff --git a/LAGraph/data/comments_full.mtx b/LAGraph/data/comments_full.mtx
index 430b7756bb..097bb7a77a 100644
--- a/LAGraph/data/comments_full.mtx
+++ b/LAGraph/data/comments_full.mtx
@@ -2,7 +2,7 @@
 %%GraphBLAS type double
 % comments for full.mtx
 % this file was created by test_MMRead.c
-3 3 9
+3 3
 .646
 .709
 .754
diff --git a/LAGraph/experimental/test/test_BF.c b/LAGraph/experimental/test/test_BF.c
index ad48ca5d68..e5e88ff4e6 100644
--- a/LAGraph/experimental/test/test_BF.c
+++ b/LAGraph/experimental/test/test_BF.c
@@ -396,8 +396,6 @@ void test_BF (void)
                     // since d5 is a dense vector filled with infinity, we have
                     // to compare it against d seperaterly
                     OK (GrB_Vector_extractElement (&di, d5, i)) ;
-                    printf ("di %g d[i] %g difference %g\n",
-                        di, d [i], di - d [i]) ;
                     TEST_CHECK (di == d[i]) ;
 
                     // since d5a is a dense vector filled with infinity, we
diff --git a/LAGraph/github_workflows/workflows/build.yml b/LAGraph/github_workflows/workflows/build.yml
index 26fa99157e..f1562e1b0f 100644
--- a/LAGraph/github_workflows/workflows/build.yml
+++ b/LAGraph/github_workflows/workflows/build.yml
@@ -1,4 +1,4 @@
-name: LAGraph CI
+name: LAGraph CI with Builtin GraphBLAS
 
 on:
   workflow_dispatch:
@@ -13,38 +13,34 @@ jobs:
     strategy:
       matrix:
         config:
-         - {grb_version: 7.1.0, conda_grb_package_hash: h27087fc, conda_extension: tar.bz2}
-         - {grb_version: 7.3.0, conda_grb_package_hash: h27087fc, conda_extension: tar.bz2}
-         - {grb_version: 7.4.1, conda_grb_package_hash: hcb278e6, conda_extension: conda}
+        # if there are multiple items in this list, only use should
+        # deployit=true for just one of them.
+         - {grb_version: 9.3.1, deployit: true}
     steps:
     - name: Checkout
       uses: actions/checkout@v2.0.0
     - name: Install tools for build
       run: |
         sudo apt install -y lcov
-    - name: Get GraphBLAS binaries
+    - name: Build GraphBLAS
       run: |
-        mkdir graphblas-binaries
-        cd graphblas-binaries
-        wget --quiet https://anaconda.org/conda-forge/graphblas/${{ matrix.config.grb_version }}/download/linux-64/graphblas-${{ matrix.config.grb_version }}-${{ matrix.config.conda_grb_package_hash }}_0.${{ matrix.config.conda_extension }}
-        if [ ${{ matrix.config.conda_extension }} == "tar.bz2" ]; then
-          tar xf graphblas-${{ matrix.config.grb_version }}-${{ matrix.config.conda_grb_package_hash }}_0.${{ matrix.config.conda_extension }}
-        else
-          unzip graphblas-${{ matrix.config.grb_version }}-${{ matrix.config.conda_grb_package_hash }}_0.${{ matrix.config.conda_extension }}
-          tar xf pkg-graphblas-${{ matrix.config.grb_version }}-${{ matrix.config.conda_grb_package_hash }}_0.tar.zst
-        fi
+        git clone https://github.com/DrTimothyAldenDavis/GraphBLAS.git
+        cd GraphBLAS
+        git checkout tags/v${{ matrix.config.grb_version }}
+        make compact
+        sudo make install
         cd ..
     - name: Build project
       run: |
-        export GRAPHBLAS_INCLUDE_DIR=`pwd`/graphblas-binaries/include
-        export GRAPHBLAS_LIBRARY=`pwd`/graphblas-binaries/lib/libgraphblas.so
+        export GRAPHBLAS_INCLUDE_DIR=`pwd`/GraphBLAS/include/suitesparse
+        export GRAPHBLAS_LIBRARY=`pwd`/GraphBLAS/lib/libgraphblas.so
         cd build
         cmake .. -DCOVERAGE=1 -DGRAPHBLAS_INCLUDE_DIR=${GRAPHBLAS_INCLUDE_DIR} -DGRAPHBLAS_LIBRARY=${GRAPHBLAS_LIBRARY}
         JOBS=2 make
         make test_coverage
     - name: Deploy
       uses: JamesIves/github-pages-deploy-action@4.1.1
-      if: matrix.config.grb_version == '7.4.1' && github.event_name == 'push' && github.ref == 'refs/heads/stable'
+      if: matrix.config.deployit && github.event_name == 'push' && github.ref == 'refs/heads/stable'
       with:
         branch: gh-pages
         folder: build/test_coverage/
@@ -59,9 +55,7 @@ jobs:
     strategy:
       matrix:
         config:
-         - {grb_version: 7.1.0, conda_grb_package_hash: h7881ed4, conda_extension: tar.bz2}
-         - {grb_version: 7.3.0, conda_grb_package_hash: ha894c9a, conda_extension: tar.bz2}
-         - {grb_version: 7.4.1, conda_grb_package_hash: ha894c9a, conda_extension: conda}
+         - {grb_version: 9.3.1}
     steps:
     - name: Checkout
       uses: actions/checkout@v2.0.0
@@ -70,22 +64,18 @@ jobs:
         brew tap-new libomp/cask
         brew extract --version=14.0.6 libomp libomp/cask
         brew install libomp@14.0.6
-    - name: Get GraphBLAS binaries
+    - name: Build GraphBLAS
       run: |
-        mkdir graphblas-binaries
-        cd graphblas-binaries
-        wget --quiet https://anaconda.org/conda-forge/graphblas/${{ matrix.config.grb_version }}/download/osx-64/graphblas-${{ matrix.config.grb_version }}-${{ matrix.config.conda_grb_package_hash }}_0.${{ matrix.config.conda_extension }}
-        if [ ${{ matrix.config.conda_extension }} == "tar.bz2" ]; then
-          tar xf graphblas-${{ matrix.config.grb_version }}-${{ matrix.config.conda_grb_package_hash }}_0.${{ matrix.config.conda_extension }}
-        else
-          unzip graphblas-${{ matrix.config.grb_version }}-${{ matrix.config.conda_grb_package_hash }}_0.${{ matrix.config.conda_extension }}
-          tar xf pkg-graphblas-${{ matrix.config.grb_version }}-${{ matrix.config.conda_grb_package_hash }}_0.tar.zst
-        fi
+        git clone https://github.com/DrTimothyAldenDavis/GraphBLAS.git
+        cd GraphBLAS
+        git checkout tags/v${{ matrix.config.grb_version }}
+        make compact
+        sudo make install
         cd ..
     - name: Build project
       run: |
-        export GRAPHBLAS_INCLUDE_DIR=`pwd`/graphblas-binaries/include
-        export GRAPHBLAS_LIBRARY=`pwd`/graphblas-binaries/lib/libgraphblas.dylib
+        export GRAPHBLAS_INCLUDE_DIR=`pwd`/GraphBLAS/include/suitesparse
+        export GRAPHBLAS_LIBRARY=`pwd`/GraphBLAS/lib/libgraphblas.dylib
         # adding an extra line to the CMakeLists.txt file to locate the libomp instance installed by brew
         echo 'include_directories("/usr/local/opt/libomp/include")' | cat - CMakeLists.txt
         cd build
diff --git a/LAGraph/include/LAGraph.h b/LAGraph/include/LAGraph.h
index baa9bbaf62..8357c8bdb5 100644
--- a/LAGraph/include/LAGraph.h
+++ b/LAGraph/include/LAGraph.h
@@ -37,10 +37,10 @@
 // See also the LAGraph_Version utility method, which returns these values.
 // These definitions are derived from LAGraph/CMakeLists.txt.
 
-#define LAGRAPH_DATE "Mar 22, 2024"
+#define LAGRAPH_DATE "Aug 20, 2024"
 #define LAGRAPH_VERSION_MAJOR  1
 #define LAGRAPH_VERSION_MINOR  1
-#define LAGRAPH_VERSION_UPDATE 3
+#define LAGRAPH_VERSION_UPDATE 4
 
 //==============================================================================
 // include files and helper macros
diff --git a/LAGraph/src/utility/LAGraph_MMWrite.c b/LAGraph/src/utility/LAGraph_MMWrite.c
index 51bd5841c2..4c4c8828f9 100644
--- a/LAGraph/src/utility/LAGraph_MMWrite.c
+++ b/LAGraph/src/utility/LAGraph_MMWrite.c
@@ -459,8 +459,18 @@ int LAGraph_MMWrite
         nvals_to_print = nself_edges + (nvals - nself_edges) / 2 ;
     }
 
-    FPRINTF (f, "%" PRIu64 " %" PRIu64 " %" PRIu64 "\n",
-        nrows, ncols, nvals_to_print) ;
+    if (MM_fmt == MM_array)
+    {
+        // write `nrows ncols` if the array format is used
+        FPRINTF (f, "%" PRIu64 " %" PRIu64 "\n",
+            nrows, ncols) ;
+    }
+    else
+    {
+        // otherwise write `nrows ncols nvals` for the coordinate format
+        FPRINTF (f, "%" PRIu64 " %" PRIu64 " %" PRIu64 "\n",
+            nrows, ncols, nvals_to_print) ;
+    }
 
     if (nvals_to_print == 0)
     {
diff --git a/ParU/CMakeLists.txt b/ParU/CMakeLists.txt
index c273196d3b..683d63f8c7 100644
--- a/ParU/CMakeLists.txt
+++ b/ParU/CMakeLists.txt
@@ -13,9 +13,9 @@
 # cmake 3.22 is required to find the BLAS in SuiteSparse_config
 cmake_minimum_required ( VERSION 3.22 )
 
-set ( PARU_DATE "Aug 2, 2024" )
+set ( PARU_DATE "Aug 20, 2024" )
 set ( PARU_VERSION_MAJOR  0 CACHE STRING "" FORCE )
-set ( PARU_VERSION_MINOR  2 CACHE STRING "" FORCE )
+set ( PARU_VERSION_MINOR  3 CACHE STRING "" FORCE )
 set ( PARU_VERSION_UPDATE 0 CACHE STRING "" FORCE )
 
 message ( STATUS "Building PARU version: v"
@@ -423,14 +423,20 @@ if ( SUITESPARSE_DEMOS )
 
     if ( PARU_HAS_OPENMP )
         add_executable ( paru_demo "Demo/paru_demo.cpp" )
+        add_executable ( paru_benchmark "Demo/paru_benchmark.cpp" )
         if ( BUILD_SHARED_LIBS )
             target_link_libraries ( paru_demo PUBLIC ParU )
+            target_link_libraries ( paru_benchmark PUBLIC ParU )
         else ( )
             target_link_libraries ( paru_demo PUBLIC ParU_static )
+            target_link_libraries ( paru_benchmark PUBLIC ParU_static )
         endif ( )
         target_link_libraries ( paru_demo
             PUBLIC SuiteSparse::CHOLMOD SuiteSparse::UMFPACK
             SuiteSparse::SuiteSparseConfig OpenMP::OpenMP_CXX )
+        target_link_libraries ( paru_benchmark
+            PUBLIC SuiteSparse::CHOLMOD SuiteSparse::UMFPACK
+            SuiteSparse::SuiteSparseConfig OpenMP::OpenMP_CXX )
 
         add_executable ( paru_democ "Demo/paru_democ.c" )
         if ( BUILD_SHARED_LIBS )
diff --git a/ParU/Config/ParU.h.in b/ParU/Config/ParU.h.in
index d9ef6506a0..2fc6d1eca0 100644
--- a/ParU/Config/ParU.h.in
+++ b/ParU/Config/ParU.h.in
@@ -139,6 +139,7 @@ typedef enum
     // int64_t parameter, for ParU_Get only:
     PARU_CONTROL_OPENMP = 1013,               // if ParU compiled with OpenMP;
                                               // (for ParU_Get only, not set)
+    PARU_CONTROL_NUM_THREADS = 1014,          // actual number of threads used
 
     // double parameters for ParU_Set and ParU_Get:
     PARU_CONTROL_PIVOT_TOLERANCE = 2001,      // pivot tolerance
diff --git a/ParU/Demo/.gitignore b/ParU/Demo/.gitignore
new file mode 100644
index 0000000000..4f30d42c90
--- /dev/null
+++ b/ParU/Demo/.gitignore
@@ -0,0 +1,3 @@
+# Ignore the following:
+Res
+!.gitignore
diff --git a/ParU/Demo/paru_benchmark.cpp b/ParU/Demo/paru_benchmark.cpp
new file mode 100644
index 0000000000..6f64fd488e
--- /dev/null
+++ b/ParU/Demo/paru_benchmark.cpp
@@ -0,0 +1,444 @@
+// ==========================================================================  /
+// =======================  paru_benchmark.cpp ==============================  /
+// ==========================================================================  /
+
+// ParU, Copyright (c) 2022-2024, Mohsen Aznaveh and Timothy A. Davis,
+// All Rights Reserved.
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+/*
+ * @brief    benchmark ParU and UMFPACK
+ *
+ * @author Aznaveh
+ * */
+
+#include <iostream>
+#include <iomanip>
+#include <ios>
+#include <cmath>
+
+#include "ParU.h"
+#include <stdlib.h>
+#ifdef _OPENMP
+#include <omp.h>
+#else
+#error "OpenMP not available!"
+#endif
+
+#define FREE_ALL_AND_RETURN(info)               \
+{                                               \
+    if (b != NULL) free(b);                     \
+    if (xx != NULL) free(xx);                   \
+    if (x != NULL) free(x);                     \
+    if (B != NULL) free(B);                     \
+    if (X != NULL) free(X);                     \
+    umfpack_dl_free_symbolic(&Symbolic);        \
+    umfpack_dl_free_numeric(&Numeric);          \
+    ParU_FreeNumeric(&Num, Control);            \
+    ParU_FreeSymbolic(&Sym, Control);           \
+    ParU_FreeControl(&Control);                 \
+    cholmod_l_free_sparse(&A, cc);              \
+    cholmod_l_finish(cc);                       \
+    if (argc > 1 && fp != NULL) fclose (fp) ;   \
+    fp = NULL ;                                 \
+    return (info) ;                             \
+}
+
+static int compar (const void *p1, const void *p2)
+{
+    double x1 = *((double *) p1) ;
+    double x2 = *((double *) p2) ;
+    return (x1 < x2 ? -1 : ((x1 > x2) ? 1 : 0)) ;
+}
+
+int main(int argc, char **argv)
+{
+    cholmod_common Common, *cc;
+    cholmod_sparse *A;
+    ParU_Symbolic Sym = NULL;
+    ParU_Numeric Num = NULL ;
+    ParU_Control Control = NULL ;
+    ParU_Info info;
+    double *b = NULL, *xx = NULL, *B = NULL, *X = NULL, *x = NULL ;
+    void *Symbolic = NULL, *Numeric = NULL ;
+    FILE *fp = stdin ;
+
+    //~~~~~~~~~Reading the input matrix and test if the format is OK~~~~~~~~~~~~
+    // start CHOLMOD
+    cc = &Common;
+    int mtype;
+    cholmod_l_start(cc);
+
+    if (argc > 1)
+    {
+        std::cout << "Matrix: " << argv [1] << std::endl ;
+        fp = fopen (argv [1], "r") ;
+    }
+    else
+    {
+        std::cout << "Matrix: stdin" << std::endl ;
+    }
+
+    // A = mread (fp) ; read in the sparse matrix A
+    A = (cholmod_sparse *)cholmod_l_read_matrix(fp, 1, &mtype, cc);
+    if (A == NULL)
+    {
+        std::cout << "ParU: input matrix is invalid" << std::endl;
+        FREE_ALL_AND_RETURN (PARU_INVALID) ;
+    }
+
+    if (mtype != CHOLMOD_SPARSE)
+    {
+        std::cout << "ParU: input matrix must be sparse" << std::endl;
+        FREE_ALL_AND_RETURN (PARU_INVALID) ;
+    }
+
+    if (A->xtype != CHOLMOD_REAL)
+    {
+        std::cout << "ParU: input matrix must be real" << std::endl;
+        FREE_ALL_AND_RETURN (PARU_INVALID) ;
+    }
+
+    //~~~~~~~~~~~~~~~~~~~Starting computation~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+    int ver[3];
+    char date[128];
+    ParU_Version(ver, date);
+    std::cout << "================= ParU "
+        << ver[0] << "." << ver[1] << "."  << ver[2]
+        << " ================================== " << date << std::endl;
+
+    info = ParU_InitControl (&Control) ;
+    if (info != PARU_SUCCESS)
+    {
+        FREE_ALL_AND_RETURN (info) ;
+    }
+
+    const char *blas_name ;
+    info = ParU_Get (PARU_CONTROL_BLAS_LIBRARY_NAME, &blas_name, Control) ;
+    if (info != PARU_SUCCESS)
+    {
+        FREE_ALL_AND_RETURN (info) ;
+    }
+    std::cout << "BLAS: " << blas_name << std::endl ;
+
+    const char *tasking ;
+    info = ParU_Get (PARU_CONTROL_FRONT_TREE_TASKING, &tasking, Control) ;
+    if (info != PARU_SUCCESS)
+    {
+        FREE_ALL_AND_RETURN (info) ;
+    }
+    std::cout << "frontal tree tasking: " << tasking << std::endl ;
+
+    int64_t using_openmp ;
+    info = ParU_Get (PARU_CONTROL_OPENMP, &using_openmp, Control) ;
+    if (info != PARU_SUCCESS)
+    {
+        FREE_ALL_AND_RETURN (info) ;
+    }
+    std::cout << "OpenMP in ParU: " << (using_openmp ? "yes" : "no" )
+        << std::endl ;
+
+    int64_t max_nthreads ;
+    #ifdef _OPENMP
+    max_nthreads = omp_get_max_threads ( ) ;
+    #else
+    max_ntreads = 1 ;
+    #endif
+    std::cout << "max # threads: " << max_nthreads << std::endl ;
+
+    // allocate workspace
+    int64_t n, anz ;
+    n = A->nrow ;
+    const int64_t nrhs = 16;  // number of right handsides
+    b = (double *)malloc(n * sizeof(double));
+    xx = (double *)malloc(n * sizeof(double));
+    B = (double *)malloc(n * nrhs * sizeof(double));
+    X = (double *)malloc(n * nrhs * sizeof(double));
+    x = (double *)malloc(n * sizeof(double));
+    double rcond ;
+
+    #define NTRIALS 5
+    int middle = NTRIALS / 2 ;
+
+    for (int ord = 0 ; ord <= 1 ; ord++)
+    {
+        int ordering = (ord == 0) ? PARU_ORDERING_AMD :
+            PARU_ORDERING_METIS_GUARD ;
+        printf ("===== ParU ordering: %d\n", ordering) ;
+        ParU_Set (PARU_CONTROL_ORDERING, ordering, Control) ;
+        for (int nthreads = max_nthreads ; nthreads > 0 ; nthreads = nthreads/2)
+        {
+            printf ("# threads: %d\n", nthreads) ;
+            ParU_Set (PARU_CONTROL_MAX_THREADS, (int64_t) nthreads, Control) ;
+            int64_t nthreads2 = 0 ;
+            ParU_Get (PARU_CONTROL_NUM_THREADS, &nthreads2, Control) ;
+            if (nthreads != (int32_t) nthreads2)
+            {
+                std::cout << "ParU: invalid # of threads!" << std::endl;
+                FREE_ALL_AND_RETURN (PARU_INVALID) ;
+            }
+
+            double ParU_sym_times [NTRIALS] ;
+            double ParU_num_times [NTRIALS] ;
+            double ParU_sol_times [NTRIALS] ;
+            for (int trial = 0 ; trial < NTRIALS  ; trial++)
+            {
+                printf ("Trial: %d\n", trial) ;
+
+                double my_start_time = SUITESPARSE_TIME ;
+                info = ParU_Analyze(A, &Sym, Control);
+                double my_time_analyze = SUITESPARSE_TIME - my_start_time;
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << "ParU: analyze failed" << std::endl;
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+
+                info = ParU_Get (Sym, Num, PARU_GET_N, &n, Control) ;
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << "ParU: stats failed" << std::endl;
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+
+                info = ParU_Get (Sym, Num, PARU_GET_ANZ, &anz, Control) ;
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << "ParU: stats failed" << std::endl;
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+
+                if (trial == 0)
+                {
+                    std::cout << "n: " << n << " anz: " << anz << std::endl ;
+                }
+
+                double my_start_time_fac = SUITESPARSE_TIME;
+                info = ParU_Factorize(A, Sym, &Num, Control);
+                double my_time_fac = SUITESPARSE_TIME - my_start_time_fac;
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << std::scientific << std::setprecision(6)
+                        << "ParU: factorization was NOT successful: "
+                        << my_time_fac << " seconds\n";
+                    if (info == PARU_OUT_OF_MEMORY)
+                        std::cout << "Out of memory\n";
+                    if (info == PARU_INVALID)
+                        std::cout << "Invalid!\n";
+                    if (info == PARU_SINGULAR)
+                        std::cout << "Singular!\n";
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+
+                info = ParU_Get (Sym, Num, PARU_GET_RCOND_ESTIMATE, &rcond,
+                    Control) ;
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << "ParU: stats failed" << std::endl;
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+
+                //~~~~~~~~~~Test the results ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+                double my_time, my_solve_time;
+
+                for (int64_t i = 0; i < n; ++i) b[i] = i + 1;
+                double my_solve_time_start = SUITESPARSE_TIME;
+                info = ParU_Solve(Sym, Num, b, xx, Control);
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << "ParU: solve failed" << std::endl;
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+                my_solve_time = SUITESPARSE_TIME - my_solve_time_start;
+                my_time = SUITESPARSE_TIME - my_start_time;
+
+                double resid, anorm, xnorm;
+                info = ParU_Residual(A, xx, b, resid, anorm, xnorm, Control);
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << "ParU: resid failed" << std::endl;
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+                double rresid = (anorm == 0 || xnorm == 0 ) ? 0 :
+                    (resid/(anorm*xnorm));
+
+                for (int64_t i = 0; i < n; ++i)
+                {
+                    for (int64_t j = 0; j < nrhs; ++j)
+                    {
+                        B[j * n + i] = (double)(i + j + 1);
+                    }
+                }
+
+                my_solve_time_start = SUITESPARSE_TIME;
+                info = ParU_Solve(Sym, Num, nrhs, B, X, Control);
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << "ParU: solve failed" << std::endl;
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+                double my_solve_time2 = SUITESPARSE_TIME - my_solve_time_start;
+
+                info = ParU_Residual(A, X, B, nrhs, resid, anorm, xnorm,
+                    Control);
+                if (info != PARU_SUCCESS)
+                {
+                    std::cout << "ParU: solve failed" << std::endl;
+                    FREE_ALL_AND_RETURN (info) ;
+                }
+                double rresid2 = (anorm == 0 || xnorm == 0 ) ? 0 :
+                    (resid/(anorm*xnorm));
+
+                if (trial == 0)
+                {
+                    std::cout << std::scientific << std::setprecision(6)
+                        << "Relative residual: " << rresid << " rcond: "
+                        << rcond << std::endl;
+                    std::cout << std::scientific << std::setprecision(6)
+                        << "Multiple right hand side: relative residual is |"
+                        << rresid2 << "|." << std::endl;
+                }
+
+                std::cout << std::scientific << std::setprecision(6)
+                    << "ParU: time: sym: " << my_time_analyze
+                    << " num: " << my_time_fac
+                    << " solve (1 rhs): " << my_solve_time
+                    << " solve (16 rhs): " << my_solve_time2 << std::endl ;
+
+                ParU_sym_times [trial] = my_time_analyze ;
+                ParU_num_times [trial] = my_time_fac ;
+                ParU_sol_times [trial] = my_solve_time ;
+
+                ParU_FreeNumeric(&Num, Control);
+                ParU_FreeSymbolic(&Sym, Control);
+            }
+
+            qsort (ParU_sym_times, NTRIALS, sizeof (double), compar) ;
+            qsort (ParU_num_times, NTRIALS, sizeof (double), compar) ;
+            qsort (ParU_sol_times, NTRIALS, sizeof (double), compar) ;
+
+            std::cout << std::scientific << std::setprecision(6)
+                << "\nParU ordering " << ordering
+                << " threads " << nthreads
+                << " median sym: " << ParU_sym_times [middle]
+                << " num: " << ParU_num_times [middle]
+                << " sol: " << ParU_sol_times [middle]
+                << " total: " << ParU_sym_times [middle] +
+                ParU_num_times [middle] + ParU_sol_times [middle]
+                << std::endl << std::endl ;
+        }
+    }
+
+    //~~~~~~~~~~~~~~~~~~~End computation~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+    //~~~~~~~~~~~~~~~~~~~Calling umfpack~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+    double umf_time = 0;
+    double status,           // Info [UMFPACK_STATUS]
+        Info[UMFPACK_INFO],  // Contains statistics about the symbolic analysis
+        umf_Control[UMFPACK_CONTROL];  // it is set in umfpack_dl_defaults and
+    // is used in umfpack_dl_symbolic; if
+    // passed NULL it will use the defaults
+    umfpack_dl_defaults(umf_Control);
+
+    int64_t *Ap = (int64_t *)A->p;
+    int64_t *Ai = (int64_t *)A->i;
+    double *Ax = (double *)A->x;
+
+    for (int ord = 0 ; ord <= 1 ; ord++)
+    {
+        int ordering = (ord == 0) ?  UMFPACK_ORDERING_AMD :
+            UMFPACK_ORDERING_METIS_GUARD ;
+        printf ("\n===== UMFPACK ordering: %d\n", ordering) ;
+        for (int nthreads = max_nthreads ; nthreads > 0 ; nthreads = nthreads/2)
+        {
+            printf ("# threads: %d\n", nthreads) ;
+            #ifdef _OPENMP
+            omp_set_num_threads (nthreads) ;
+            #endif
+
+            double UMF_sym_times [NTRIALS] ;
+            double UMF_num_times [NTRIALS] ;
+            double UMF_sol_times [NTRIALS] ;
+            umf_Control[UMFPACK_ORDERING] = ordering ;
+            for (int trial = 0 ; trial < NTRIALS ; trial++)
+            {
+                printf ("Trial: %d\n", trial) ;
+
+                double umf_start_time = SUITESPARSE_TIME;
+                status = umfpack_dl_symbolic(n, n, Ap, Ai, Ax, &Symbolic,
+                    umf_Control, Info);
+                if (status < 0)
+                {
+                    umfpack_dl_report_info(umf_Control, Info);
+                    umfpack_dl_report_status(umf_Control, status);
+                    std::cout << "umfpack_dl_symbolic failed\n";
+                    FREE_ALL_AND_RETURN (PARU_INVALID) ;
+                }
+                double umf_symbolic = SUITESPARSE_TIME - umf_start_time;
+                double umf_fac_start = SUITESPARSE_TIME;
+                status = umfpack_dl_numeric(Ap, Ai, Ax, Symbolic, &Numeric,
+                    umf_Control, Info);
+                if (status < 0)
+                {
+                    umfpack_dl_report_info(umf_Control, Info);
+                    umfpack_dl_report_status(umf_Control, status);
+                    std::cout << "umfpack_dl_numeric failed\n";
+                    FREE_ALL_AND_RETURN (PARU_INVALID) ;
+                }
+
+                double umf_time_fac = SUITESPARSE_TIME - umf_fac_start;
+
+                for (int64_t i = 0; i < n; ++i) b[i] = i + 1;
+
+                double solve_start = SUITESPARSE_TIME;
+                status = umfpack_dl_solve(UMFPACK_A, Ap, Ai, Ax, x, b,
+                    Numeric, umf_Control, Info);
+                double umf_solve_time = SUITESPARSE_TIME - solve_start;
+                umf_time = SUITESPARSE_TIME - umf_start_time;
+
+                double umf_resid, umf_anorm, umf_xnorm;
+                info = ParU_Residual(A, x, b, umf_resid, umf_anorm, umf_xnorm,
+                    Control);
+                double umf_rresid = (umf_anorm == 0 || umf_xnorm == 0 )
+                    ? 0 : (umf_resid/(umf_anorm*umf_xnorm));
+                if (trial == 0)
+                {
+                    std::cout << std::scientific << std::setprecision(6)
+                        << "UMFPACK relative residual: " << umf_rresid
+                        << std::endl;
+                }
+
+                std::cout << std::scientific << std::setprecision(6)
+                    << "UMFPACK time: sym " << umf_symbolic
+                    << " num: " << umf_time_fac
+                    << " solve: " << umf_solve_time << std::endl ;
+
+                UMF_sym_times [trial] = umf_symbolic ;
+                UMF_num_times [trial] = umf_time_fac ;
+                UMF_sol_times [trial] = umf_solve_time ;
+
+                umfpack_dl_free_symbolic(&Symbolic);
+                umfpack_dl_free_numeric(&Numeric);
+            }
+
+            qsort (UMF_sym_times, NTRIALS, sizeof (double), compar) ;
+            qsort (UMF_num_times, NTRIALS, sizeof (double), compar) ;
+            qsort (UMF_sol_times, NTRIALS, sizeof (double), compar) ;
+
+            std::cout << std::scientific << std::setprecision(6)
+                << "\nUMF  ordering " << ordering
+                << " threads " << nthreads
+                << " median sym: " << UMF_sym_times [middle]
+                << " num: " << UMF_num_times [middle]
+                << " sol: " << UMF_sol_times [middle]
+                << " total: " << UMF_sym_times [middle] +
+                UMF_num_times [middle] + UMF_sol_times [middle]
+                << std::endl << std::endl ;
+        }
+    }
+
+    //~~~~~~~~~~~~~~~~~~~Free Everything~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+    FREE_ALL_AND_RETURN (PARU_SUCCESS) ;
+}
+
diff --git a/ParU/Doc/paru_user_guide.tex b/ParU/Doc/paru_user_guide.tex
index 53968be19b..a55b9b61e1 100644
--- a/ParU/Doc/paru_user_guide.tex
+++ b/ParU/Doc/paru_user_guide.tex
@@ -337,7 +337,8 @@ \subsection{{\sf ParU\_Set}: set parameters in the Control object}
     ) ; \end{verbatim}}
 
 The \verb'ParU_Control_enum parameter' defines which parameter to set,
-described below.
+described below.  These are also used for \verb'ParU_Get', to read back these
+parameters from the \verb'Control' object.
 
     {\footnotesize
     \begin{verbatim}
@@ -362,6 +363,7 @@ \subsection{{\sf ParU\_Set}: set parameters in the Control object}
         // int64_t parameter, for ParU_Get only:
         PARU_CONTROL_OPENMP = 1013,               // if ParU compiled with OpenMP;
                                                   // (for ParU_Get only, not set)
+        PARU_CONTROL_NUM_THREADS = 1014,          // actual number of threads used
 
         // double parameters for ParU_Set and ParU_Get:
         PARU_CONTROL_PIVOT_TOLERANCE = 2001,      // pivot tolerance
@@ -763,7 +765,7 @@ \subsection{{\sf ParU\_Get}: get information from a ParU opaque object}
     ) ; \end{verbatim}}
 
 The parameters that can be returned are the same as those described in
-Section~\ref{set}, with three additional parameters that can be queried by
+Section~\ref{set}, with four additional parameters that can be queried by
 \verb'ParU_Get' but not set with \verb'ParU_Set'.  Two cases return string that
 is owned by the library and must not be modified:
 
@@ -773,12 +775,22 @@ \subsection{{\sf ParU\_Get}: get information from a ParU opaque object}
     ParU_Get (PARU_CONTROL_BLAS_LIBRARY_NAME, &blas_name, Control)) ;
     ParU_Get (PARU_CONTROL_FRONT_TREE_TASKING, &front_tasking, Control)) ; \end{verbatim}}
 
-A final case returns true if ParU was compiled with OpenMP, or false otherwise:
+This case returns true if ParU was compiled with OpenMP, or false otherwise:
 
     {\footnotesize
     \begin{verbatim}
     int64_t openmp_used ;
-    ParU_Get (PARU_CONTROL_OPENMP , &openmp_used, Control)) ; \end{verbatim}}
+    ParU_Get (PARU_CONTROL_OPENMP, &openmp_used, Control)) ; \end{verbatim}}
+
+Finally, the number of threads actually to be used by ParU can be queried as
+follows.  If the \verb'Control' is set to its default, this will be the value
+from \verb'omp_get_max_threads'.  Otherwise, the value is smaller of
+\verb'omp_get_max_threads' and \verb'PARU_CONTROL_MAX_THREADS'.
+
+    {\footnotesize
+    \begin{verbatim}
+    int64_t nthreads_used ;
+    ParU_Get (PARU_CONTROL_NUM_THREADS, &nthreads_used, Control)) ; \end{verbatim}}
 
 %-------------------------------------------------------------------------------
 \subsection{{\sf ParU\_Analyze}: symbolic analysis}
diff --git a/ParU/Doc/paru_version.tex b/ParU/Doc/paru_version.tex
index bf4ab9fcab..801498c306 100644
--- a/ParU/Doc/paru_version.tex
+++ b/ParU/Doc/paru_version.tex
@@ -1,2 +1,2 @@
 % version of SuiteSparse/ParU
-\date{VERSION 0.2.0, Aug 2, 2024}
+\date{VERSION 0.3.0, Aug 20, 2024}
diff --git a/ParU/Include/ParU.h b/ParU/Include/ParU.h
index 52eda95f82..2348782d2c 100644
--- a/ParU/Include/ParU.h
+++ b/ParU/Include/ParU.h
@@ -53,25 +53,25 @@ typedef enum ParU_Info
     PARU_TOO_LARGE = -4         // problem too large for the BLAS
 } ParU_Info ;
 
-#define PARU_DATE "Aug 2, 2024"
+#define PARU_DATE "Aug 20, 2024"
 #define PARU_VERSION_MAJOR  0
-#define PARU_VERSION_MINOR  2
+#define PARU_VERSION_MINOR  3
 #define PARU_VERSION_UPDATE 0
 
-#define PARU__VERSION SUITESPARSE__VERCODE(0,2,0)
+#define PARU__VERSION SUITESPARSE__VERCODE(0,3,0)
 #if !defined (SUITESPARSE__VERSION) || \
     (SUITESPARSE__VERSION < SUITESPARSE__VERCODE(7,8,0))
-#error "ParU 0.2.0 requires SuiteSparse_config 7.8.0 or later"
+#error "ParU 0.3.0 requires SuiteSparse_config 7.8.0 or later"
 #endif
 
 #if !defined (UMFPACK__VERSION) || \
     (UMFPACK__VERSION < SUITESPARSE__VERCODE(6,3,4))
-#error "ParU 0.2.0 requires UMFPACK 6.3.4 or later"
+#error "ParU 0.3.0 requires UMFPACK 6.3.4 or later"
 #endif
 
 #if !defined (CHOLMOD__VERSION) || \
     (CHOLMOD__VERSION < SUITESPARSE__VERCODE(5,3,0))
-#error "ParU 0.2.0 requires CHOLMOD 5.3.0 or later"
+#error "ParU 0.3.0 requires CHOLMOD 5.3.0 or later"
 #endif
 
 //  the same values as UMFPACK_STRATEGY defined in UMFPACK/Include/umfpack.h
@@ -139,6 +139,7 @@ typedef enum
     // int64_t parameter, for ParU_Get only:
     PARU_CONTROL_OPENMP = 1013,               // if ParU compiled with OpenMP;
                                               // (for ParU_Get only, not set)
+    PARU_CONTROL_NUM_THREADS = 1014,          // actual number of threads used
 
     // double parameters for ParU_Set and ParU_Get:
     PARU_CONTROL_PIVOT_TOLERANCE = 2001,      // pivot tolerance
diff --git a/ParU/Makefile b/ParU/Makefile
index 447902343d..3864fe02a6 100644
--- a/ParU/Makefile
+++ b/ParU/Makefile
@@ -57,6 +57,9 @@ demos: library
 	( cd build && [ -f paru_democ$(EXEEXT) ] && ./paru_democ$(EXEEXT) < ../Matrix/west0067.mtx || true )
 	( cd build && [ -f paru_demo$(EXEEXT) ] && ./paru_demo$(EXEEXT) < ../Matrix/xenon1.mtx || true )
 
+bench: library
+	( cd build && [ -f paru_benchmark$(EXEEXT) ] && ./paru_benchmark$(EXEEXT) < ../Matrix/xenon1.mtx || true )
+
 V = valgrind --leak-check=full --show-leak-kinds=all
 
 vdemos: library
diff --git a/ParU/Source/ParU_Get.cpp b/ParU/Source/ParU_Get.cpp
index 359f30bfb0..78400190e0 100644
--- a/ParU/Source/ParU_Get.cpp
+++ b/ParU/Source/ParU_Get.cpp
@@ -249,10 +249,17 @@ ParU_Info ParU_Get          // get an int64_t parameter from the Control object
     {
 
         case PARU_CONTROL_MAX_THREADS:            // max number of threads
+            // 0 is the default, which lets ParU use the # of threads
+            // determined by omp_get_max_threads.
             (*c) = (Control == NULL) ? PARU_DEFAULT_MAX_THREADS :
                 Control->paru_max_threads ;
             break ;
 
+        case PARU_CONTROL_NUM_THREADS:            // actual # of threads
+            // This is the actual # of threads ParU will use.
+            (*c) = (int64_t) paru_nthreads (Control) ;
+            break ;
+
         case PARU_CONTROL_OPENMP:                 // 1 if OpenMP, 0 if not
             #if defined ( _OPENMP )
             (*c) = 1 ;
diff --git a/ParU/Tcov/paru_quick_test.cpp b/ParU/Tcov/paru_quick_test.cpp
index be20e5eaf6..90d17afd1f 100644
--- a/ParU/Tcov/paru_quick_test.cpp
+++ b/ParU/Tcov/paru_quick_test.cpp
@@ -78,6 +78,10 @@ int main(int argc, char **argv)
     ParU_Set (PARU_CONTROL_MAX_THREADS, 4, Control) ;
     ParU_Set (PARU_CONTROL_STRATEGY, PARU_STRATEGY_SYMMETRIC, Control) ;
 
+    int64_t nthreads2 = 0 ;
+    ParU_Get (PARU_CONTROL_NUM_THREADS, &nthreads2, Control) ;
+    printf ("nthreads in use: %d\n", (int32_t) nthreads2) ;
+
     // read in the sparse matrix A
     A = (cholmod_sparse *)cholmod_l_read_matrix(stdin, 1, &mtype, cc);
     int64_t n = A->nrow ;
diff --git a/README.md b/README.md
index 4401c34ca4..c8239dbf3c 100644
--- a/README.md
+++ b/README.md
@@ -2,7 +2,7 @@
 SuiteSparse:  A Suite of Sparse matrix packages at http://suitesparse.com
 -----------------------------------------------------------------------------
 
-Aug 12, 2024, SuiteSparse VERSION 7.8.1
+Aug 20, 2024, SuiteSparse VERSION 7.8.2
 
 SuiteSparse is a set of sparse-matrix-related packages written or co-authored
 by Tim Davis, available at https://github.com/DrTimothyAldenDavis/SuiteSparse .
diff --git a/SPEX/CMakeLists.txt b/SPEX/CMakeLists.txt
index 9762da88d8..ef8e7570d1 100644
--- a/SPEX/CMakeLists.txt
+++ b/SPEX/CMakeLists.txt
@@ -13,10 +13,10 @@
 
 cmake_minimum_required ( VERSION 3.22 )
 
-set ( SPEX_DATE "July 2, 2024" )
+set ( SPEX_DATE "Aug 20, 2024" )
 set ( SPEX_VERSION_MAJOR 3 CACHE STRING "" FORCE )
 set ( SPEX_VERSION_MINOR 2 CACHE STRING "" FORCE )
-set ( SPEX_VERSION_SUB   0 CACHE STRING "" FORCE )
+set ( SPEX_VERSION_SUB   1 CACHE STRING "" FORCE )
 
 message ( STATUS "Building SPEX version: v"
     ${SPEX_VERSION_MAJOR}.
@@ -74,22 +74,22 @@ endif ( )
 #-------------------------------------------------------------------------------
 
 if ( NOT SUITESPARSE_ROOT_CMAKELISTS )
-    find_package ( SuiteSparse_config 7.8.0
+    find_package ( SuiteSparse_config 7.8.2
         PATHS ${CMAKE_SOURCE_DIR}/../SuiteSparse_config/build NO_DEFAULT_PATH )
     if ( NOT TARGET SuiteSparse::SuiteSparseConfig )
-        find_package ( SuiteSparse_config 7.8.0 REQUIRED )
+        find_package ( SuiteSparse_config 7.8.2 REQUIRED )
     endif ( )
 
-    find_package ( AMD 3.3.2
+    find_package ( AMD 3.3.3
         PATHS ${CMAKE_SOURCE_DIR}/../AMD/build NO_DEFAULT_PATH )
     if ( NOT TARGET SuiteSparse::AMD )
-        find_package ( AMD 3.3.2 REQUIRED )
+        find_package ( AMD 3.3.3 REQUIRED )
     endif ( )
 
-    find_package ( COLAMD 3.3.3
+    find_package ( COLAMD 3.3.4
         PATHS ${CMAKE_SOURCE_DIR}/../COLAMD/build NO_DEFAULT_PATH )
     if ( NOT TARGET SuiteSparse::COLAMD )
-        find_package ( COLAMD 3.3.3 REQUIRED )
+        find_package ( COLAMD 3.3.4 REQUIRED )
     endif ( )
 endif ( )
 
diff --git a/SPEX/Config/SPEX.h.in b/SPEX/Config/SPEX.h.in
index 622494228d..f90ec2ab15 100644
--- a/SPEX/Config/SPEX.h.in
+++ b/SPEX/Config/SPEX.h.in
@@ -121,8 +121,8 @@
 
 #define SPEX__VERSION SUITESPARSE__VERCODE(@SPEX_VERSION_MAJOR@,@SPEX_VERSION_MINOR@,@SPEX_VERSION_SUB@)
 #if !defined (SUITESPARSE__VERSION) || \
-    (SUITESPARSE__VERSION < SUITESPARSE__VERCODE(7,8,0))
-#error "SPEX @SPEX_VERSION_MAJOR@.@SPEX_VERSION_MINOR@.@SPEX_VERSION_SUB@ requires SuiteSparse_config 7.8.0 or later"
+    (SUITESPARSE__VERSION < SUITESPARSE__VERCODE(7,8,2))
+#error "SPEX @SPEX_VERSION_MAJOR@.@SPEX_VERSION_MINOR@.@SPEX_VERSION_SUB@ requires SuiteSparse_config 7.8.2 or later"
 #endif
 
 #if defined ( __cplusplus )
diff --git a/SPEX/Doc/ChangeLog b/SPEX/Doc/ChangeLog
index 9ec8f9a80a..973189218a 100644
--- a/SPEX/Doc/ChangeLog
+++ b/SPEX/Doc/ChangeLog
@@ -1,4 +1,8 @@
-June 20, 2024: version 3.2.0
+Aug 20, 2024: version 3.2.1
+
+    * Python interface: bug fix for string length, and Python Inf
+
+July 2, 2024: version 3.2.0
 
     * added SPEX_ldl_* methods: allowing D to have negative diagonal entries;
         SPEX_cholesky_* methods revised to require D to have positive entries.
diff --git a/SPEX/Doc/SPEX_UserGuide.pdf b/SPEX/Doc/SPEX_UserGuide.pdf
index 981621b659..d8bdf99d07 100644
Binary files a/SPEX/Doc/SPEX_UserGuide.pdf and b/SPEX/Doc/SPEX_UserGuide.pdf differ
diff --git a/SPEX/Doc/SPEX_version.tex b/SPEX/Doc/SPEX_version.tex
index d5dba69a6d..a07795e883 100644
--- a/SPEX/Doc/SPEX_version.tex
+++ b/SPEX/Doc/SPEX_version.tex
@@ -1,2 +1,2 @@
 % version of SuiteSparse/SPEX
-VERSION 3.2.0, July 2, 2024
+VERSION 3.2.1, Aug 20, 2024
diff --git a/SPEX/Include/SPEX.h b/SPEX/Include/SPEX.h
index 8a9b693b24..c08bc5c651 100644
--- a/SPEX/Include/SPEX.h
+++ b/SPEX/Include/SPEX.h
@@ -106,11 +106,11 @@
 //------------------------------------------------------------------------------
 
 // Current version of the code
-#define SPEX_DATE "July 2, 2024"
-#define SPEX_VERSION_STRING "3.2.0"
+#define SPEX_DATE "Aug 20, 2024"
+#define SPEX_VERSION_STRING "3.2.1"
 #define SPEX_VERSION_MAJOR 3
 #define SPEX_VERSION_MINOR 2
-#define SPEX_VERSION_SUB   0
+#define SPEX_VERSION_SUB   1
 
 #define SPEX_VERSION_NUMBER(major,minor,sub) \
     (((major)*1000ULL + (minor))*1000ULL + (sub))
@@ -119,10 +119,10 @@
                          SPEX_VERSION_MINOR, \
                          SPEX_VERSION_SUB)
 
-#define SPEX__VERSION SUITESPARSE__VERCODE(3,2,0)
+#define SPEX__VERSION SUITESPARSE__VERCODE(3,2,1)
 #if !defined (SUITESPARSE__VERSION) || \
-    (SUITESPARSE__VERSION < SUITESPARSE__VERCODE(7,8,0))
-#error "SPEX 3.2.0 requires SuiteSparse_config 7.8.0 or later"
+    (SUITESPARSE__VERSION < SUITESPARSE__VERCODE(7,8,2))
+#error "SPEX 3.2.1 requires SuiteSparse_config 7.8.2 or later"
 #endif
 
 #if defined ( __cplusplus )
diff --git a/SuiteSparse_config/CMakeLists.txt b/SuiteSparse_config/CMakeLists.txt
index 55a627fee2..f76fd0e012 100644
--- a/SuiteSparse_config/CMakeLists.txt
+++ b/SuiteSparse_config/CMakeLists.txt
@@ -15,10 +15,10 @@
 cmake_minimum_required ( VERSION 3.22 )
 
 # version of both SuiteSparse and SuiteSparse_config
-set ( SUITESPARSE_DATE "Aug 12, 2024" )
+set ( SUITESPARSE_DATE "Aug 20, 2024" )
 set ( SUITESPARSE_VERSION_MAJOR 7 )
 set ( SUITESPARSE_VERSION_MINOR 8 )
-set ( SUITESPARSE_VERSION_SUB   1 )
+set ( SUITESPARSE_VERSION_SUB   2 )
 set ( SUITESPARSE_CONFIG_VERSION_MAJOR ${SUITESPARSE_VERSION_MAJOR} CACHE STRING "" FORCE )
 set ( SUITESPARSE_CONFIG_VERSION_MINOR ${SUITESPARSE_VERSION_MINOR} CACHE STRING "" FORCE )
 set ( SUITESPARSE_CONFIG_VERSION_PATCH ${SUITESPARSE_VERSION_SUB} CACHE STRING "" FORCE )
diff --git a/SuiteSparse_config/SuiteSparse_config.h b/SuiteSparse_config/SuiteSparse_config.h
index dec7019cc8..b1d9b56e84 100644
--- a/SuiteSparse_config/SuiteSparse_config.h
+++ b/SuiteSparse_config/SuiteSparse_config.h
@@ -434,10 +434,10 @@ int SuiteSparse_version     // returns SUITESPARSE_VERSION
 
 #define SUITESPARSE_HAS_VERSION_FUNCTION
 
-#define SUITESPARSE_DATE "Aug 12, 2024"
+#define SUITESPARSE_DATE "Aug 20, 2024"
 #define SUITESPARSE_MAIN_VERSION    7
 #define SUITESPARSE_SUB_VERSION     8
-#define SUITESPARSE_SUBSUB_VERSION  1
+#define SUITESPARSE_SUBSUB_VERSION  2
 
 // version format x.y
 #define SUITESPARSE_VER_CODE(main,sub) ((main) * 1000 + (sub))
@@ -446,7 +446,7 @@ int SuiteSparse_version     // returns SUITESPARSE_VERSION
 // version format x.y.z
 #define SUITESPARSE__VERCODE(main,sub,patch) \
     (((main)*1000ULL + (sub))*1000ULL + (patch))
-#define SUITESPARSE__VERSION SUITESPARSE__VERCODE(7,8,1)
+#define SUITESPARSE__VERSION SUITESPARSE__VERCODE(7,8,2)
 
 //==============================================================================
 // SuiteSparse interface to the BLAS and LAPACK libraries