Fix #2319: MueLu research build

cmake/std/atdm/shiller/environment.sh

@@ -15,7 +15,7 @@ echo "Using hansen/shiller compiler stack $ATDM_CONFIG_COMPILER to build $ATDM_C
 
 module load ninja/1.7.2
 
-export ATDM_CONFIG_KOKKOS_ARCH=BDW
+export ATDM_CONFIG_KOKKOS_ARCH=HSW
 if [ "$ATDM_CONFIG_COMPILER" == "GNU" ]; then
     module load devpack/openmpi/2.1.1/gcc/4.9.3/cuda/8.0.61
     export OMPI_CXX=`which g++`

cmake/std/atdm/shiller/tweaks/CUDA-DEBUG-CUDA.cmake

@@ -0,0 +1 @@
+INCLUDE("${CMAKE_CURRENT_LIST_DIR}/CUDA_COMMON_TWEAKS.cmake")

cmake/std/atdm/shiller/tweaks/CUDA-RELEASE-CUDA.cmake

@@ -0,0 +1 @@
+INCLUDE("${CMAKE_CURRENT_LIST_DIR}/CUDA_COMMON_TWEAKS.cmake")

cmake/std/atdm/shiller/tweaks/CUDA_COMMON_TWEAKS.cmake

@@ -0,0 +1,9 @@
+ATDM_SET_ENABLE(Panzer_ENABLE_EXAMPLES OFF)
+ATDM_SET_ENABLE(PanzerCore_ENABLE_EXAMPLES OFF)
+ATDM_SET_ENABLE(PanzerDofMgr_ENABLE_EXAMPLES OFF)
+ATDM_SET_ENABLE(PanzerDiscFE_ENABLE_EXAMPLES OFF)
+ATDM_SET_ENABLE(PanzerAdaptersSTK_ENABLE_EXAMPLES OFF)
+ATDM_SET_ENABLE(PanzerAdaptersIOSS_ENABLE_EXAMPLES OFF)
+ATDM_SET_ENABLE(PanzerMiniEM_ENABLE_EXAMPLES OFF)
+# ToDo: Change TriBITS so that, by default, Panzer_ENABLE_EXAMPLES=OFF will
+# disable examples in the Panzer subpackages!q

packages/muelu/research/luc/region_algorithms/CMakeLists.txt

@@ -13,12 +13,6 @@ INCLUDE_DIRECTORIES(${CMAKE_CURRENT_SOURCE_DIR}/../../../test/unit_tests)
       COMM serial mpi
       )
 
-    TRIBITS_ADD_EXECUTABLE(
-      ImportTest
-      SOURCES Import.cpp
-      COMM serial mpi
-      )
-
     # TRIBITS_COPY_FILES_TO_BINARY_DIR(tawiesn_cp
     #   SOURCE_FILES driver.xml
     #   )

packages/teuchos/core/src/Teuchos_BigUInt.hpp

@@ -176,7 +176,7 @@ BigUInt<n>& BigUInt<n>::operator>>=(std::uint32_t b) {
   auto ndigits = b / 32;
   auto nbits = b - (ndigits * 32);
   for (int i = 0; i < n; ++i) {
-    std::uint32_t xi;
+    std::uint32_t xi = 0;
     if (i + ndigits < n) xi = x[i + ndigits] >> nbits;
     if (nbits && i + ndigits + 1 < n) xi |= x[i + ndigits + 1] << (32 - nbits);
     x[i] = xi;

packages/zoltan2/example/block/block.cpp

@@ -51,8 +51,6 @@
 #include <Zoltan2_PartitioningProblem.hpp>
 #include <Zoltan2_PartitioningSolution.hpp>
 
-using namespace std;
-
 /*! \example block.cpp
     An example of the use of the Block algorithm to partition data.
     \todo error handling
@@ -80,18 +78,20 @@ int main(int argc, char *argv[])
   ///////////////////////////////////////////////////////////////////////
   // Generate some input data.
 
-  size_t localCount = 40*(rank+1);
-  globalId_t *globalIds = new globalId_t [localCount];
+  int localCount = 40*(rank+1);
+  int totalCount = 20*nprocs*(nprocs+1);
+  int targetCount = totalCount / nprocs;
+  globalId_t *globalIds = new globalId_t[localCount];
 
   if (rank==0)
-    for (int i=0, num=40; i <= nprocs ; i++, num+=40)
-      cout << "Rank " << i << " has " << num << " ids." << endl;
+    for (int i=0, num=40; i < nprocs ; i++, num+=40)
+      std::cout << "Rank " << i << " generates " << num << " ids." << std::endl;
 
   globalId_t offset = 0;
   for (int i=1; i <= rank; i++)
     offset += 40*i;
 
-  for (size_t i=0; i < localCount; i++)
+  for (int i=0; i < localCount; i++)
     globalIds[i] = offset++;
 
   ///////////////////////////////////////////////////////////////////////
@@ -132,10 +132,35 @@ int main(int argc, char *argv[])
   problem->solve();
 
   ///////////////////////////////////////////////////////////////////////
-  // Check the solution.
+  // Check and print the solution.
+  // Count number of IDs assigned to each part; compare to targetCount
+
+  const globalId_t *ids = NULL;
+  ia.getIDsView(ids);
+  std::vector<int> partCounts(nprocs, 0), globalPartCounts(nprocs, 0);
+
+  for (size_t i = 0; i < ia.getLocalNumIDs(); i++) {
+    int pp = problem->getSolution().getPartListView()[i];
+    std::cout << rank << " LID " << i << " GID " << ids[i]
+                << " PART " << pp << std::endl;
+    partCounts[pp]++;
+  }
+
+#ifdef HAVE_ZOLTAN2_MPI
+  MPI_Allreduce(&(partCounts[0]), &(globalPartCounts[0]), nprocs,
+                MPI_INT, MPI_SUM, MPI_COMM_WORLD);
+#endif
 
-  if (rank == 0)
-    cout << "PASS" << endl;
+  if (rank == 0) {
+    int ierr = 0;
+    for (int i = 0; i < nprocs; i++)
+      if (globalPartCounts[i] != targetCount) 
+        std::cout << "FAIL: part " << i << " has " << partCounts[i]
+                  << " != " << targetCount << "; " << ++ierr << " errors"
+                  << std::endl;
+    if (ierr == 0)
+      std::cout << "PASS" << std::endl;
+  }
 
   delete [] globalIds;
   delete problem;