Remove ENABLE_OPENMP in favor of QMC_OMP

CMake configuration files
Source code pre-processor

CMake/ClangCompilers.cmake

@@ -12,7 +12,6 @@ endif()
 
 # Enable OpenMP
 if(QMC_OMP)
-  set(ENABLE_OPENMP 1)
   if(ENABLE_OFFLOAD)
     if (QMC_CUDA2HIP)
       set(OFFLOAD_TARGET_DEFAULT "amdgcn-amd-amdhsa")

CMake/GNUCompilers.cmake

@@ -5,7 +5,6 @@ endif()
 
 # Enable OpenMP
 if(QMC_OMP)
-  set(ENABLE_OPENMP 1)
   set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fopenmp")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fopenmp")
   if(ENABLE_OFFLOAD)

CMake/IBMCompilers.cmake

@@ -23,7 +23,6 @@ if(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64le")
   set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "-g -O3")
 
   if(QMC_OMP)
-    set(ENABLE_OPENMP 1)
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -qsmp=omp")
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -qsmp=omp")
     if(ENABLE_OFFLOAD)

CMake/IntelCompilers.cmake

@@ -17,7 +17,6 @@ endif()
 
 # Enable OpenMP
 if(QMC_OMP)
-  set(ENABLE_OPENMP 1)
   if(CMAKE_CXX_COMPILER_ID MATCHES "IntelLLVM" OR INTEL_ONEAPI_COMPILER_FOUND)
     if(ENABLE_OFFLOAD)
       set(OFFLOAD_TARGET

CMake/NVHPCCompilers.cmake

@@ -3,7 +3,6 @@
 # Specifying 'allcores' will run on all cores if OMP_NUM_THREADS is not set (which seems
 #  to be the default for other OpenMP implementations)
 if(QMC_OMP)
-  set(ENABLE_OPENMP 1)
   set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mp=allcores")
   if(ENABLE_OFFLOAD)
     message(WARNING "QMCPACK OpenMP offload is not ready for NVIDIA HPC compiler.")

CMakeLists.txt

@@ -227,7 +227,6 @@ if(CMAKE_TOOLCHAIN_FILE)
   message(STATUS "Using ${CMAKE_TOOLCHAIN_FILE} toolchain ")
 else(CMAKE_TOOLCHAIN_FILE)
   message(STATUS "Trying to figure out compiler options ....")
-  set(ENABLE_OPENMP 0)
 
   #------------------------
   # On Cray's machine

doxygen/config.h

@@ -1,21 +1,21 @@
 /* src/ohmms-config.h.in.  Generated from configure.in by autoheader.  */
 // -*- c++  -*-
 //
-//Ohmms Configuration Header. Automatically Generated
+// Ohmms Configuration Header. Automatically Generated
 //
-//See the LICENSE file in the top-level directory for copyright notices
+// See the LICENSE file in the top-level directory for copyright notices
 //
 #ifndef QMCPLUSPLUS_CONFIGURATION_H
 #define QMCPLUSPLUS_CONFIGURATION_H
 
 /* define the major version */
-#define QMCPLUSPLUS_VERSION_MAJOR  0
+#define QMCPLUSPLUS_VERSION_MAJOR 0
 
 /* define the minor version */
-#define QMCPLUSPLUS_VERSION_MINOR  6
+#define QMCPLUSPLUS_VERSION_MINOR 6
 
 /* define the patch version */
-#define QMCPLUSPLUS_VERSION_PATCH  1
+#define QMCPLUSPLUS_VERSION_PATCH 1
 
 /* define the release version */
 /* #undef QMCPLUSPLUS_RELEASE */
@@ -24,16 +24,17 @@
 /* #undef QMCPLUSPLUS_LINEARSCALE */
 
 /* define the subversion branch */
-#define QMCPLUSPLUS_BRANCH  5487
+#define QMCPLUSPLUS_BRANCH 5487
 
 /* define the subversion last changed date */
-#define QMCPLUSPLUS_LAST_CHANGED_DATE  "2012-04-22 10:24:14 -0400 (Sun, 22 Apr 2012)"
+#define QMCPLUSPLUS_LAST_CHANGED_DATE                                          \
+  "2012-04-22 10:24:14 -0400 (Sun, 22 Apr 2012)"
 
 /* define PRINT_DEBUG */
 /* #undef PRINT_DEBUG */
 
 /* Enable OpenMP parallelization. */
-#define ENABLE_OPENMP 1
+#define QMC_OMP 1
 
 /* Define to 1 if you have the `hdf5' library (-lhdf5). */
 #define HAVE_LIBHDF5 1
@@ -165,4 +166,3 @@
 /* #undef DEBUG_PSIBUFFER_ON */
 
 #endif // QMCPLUSPLUS_CONFIGURATION_H
-

src/Concurrency/Info.hpp

@@ -17,7 +17,10 @@
  *  @brief Abstraction of information on executor environments
  */
 
+#ifdef QMC_OMP
 #include <omp.h>
+#endif
+
 #ifdef QMC_EXP_THREADING
 #include <thread>
 #endif
@@ -42,7 +45,11 @@ unsigned int maxCapacity();
 template<>
 inline unsigned int maxCapacity<Executor::OPENMP>()
 {
+#ifdef QMC_OMP
   return omp_get_max_threads();
+#else
+  return 1;
+#endif
 }
 
 template<Executor TT = Executor::OPENMP>
@@ -51,10 +58,13 @@ unsigned int getWorkerId();
 template<>
 inline unsigned int getWorkerId<Executor::OPENMP>()
 {
+#ifdef QMC_OMP
   return omp_get_thread_num();
+#else
+  return 0;
+#endif
 }
 
-
 #ifdef QMC_EXP_THREADING
 template<>
 inline unsigned int maxCapacity<Executor::STD_THREADS>()

src/Concurrency/ParallelExecutorOPENMP.hpp

@@ -19,10 +19,10 @@
 
 #include <stdexcept>
 #include <string>
-#include <omp.h>
 
 #include "Concurrency/ParallelExecutor.hpp"
 #include "Platforms/Host/OutputManager.h"
+#include "Message/OpenMP.h"
 
 namespace qmcplusplus
 {

src/Concurrency/UtilityFunctions.hpp

@@ -17,14 +17,13 @@
  *  @brief utility functions for executors
  */
 
-#include <omp.h>
 #include "Concurrency/Info.hpp"
+#include "Message/OpenMP.h"
 
 namespace qmcplusplus
 {
 namespace Concurrency
 {
-
 template<Executor TT = Executor::OPENMP>
 class OverrideMaxCapacity;
 
@@ -33,17 +32,15 @@ class OverrideMaxCapacity<Executor::OPENMP>
 {
 private:
   int original_max_threads_;
+
 public:
   OverrideMaxCapacity(int max_threads)
   {
     original_max_threads_ = omp_get_max_threads();
     omp_set_num_threads(max_threads);
   }
-  
-  ~OverrideMaxCapacity()
-  {
-    omp_set_num_threads(original_max_threads_);
-  }
+
+  ~OverrideMaxCapacity() { omp_set_num_threads(original_max_threads_); }
 };
 
 template<Executor TT>
@@ -52,7 +49,7 @@ class OverrideMaxCapacity
   OverrideMaxCapacity(int max_threads) {}
 };
 
-}
-}
+} // namespace Concurrency
+} // namespace qmcplusplus
 
 #endif

src/Concurrency/tests/test_ParallelExecutorOPENMP.cpp

@@ -57,8 +57,10 @@ TEST_CASE("ParallelExecutor<OPENMP> nested case", "[concurrency]")
     ParallelExecutor<Executor::OPENMP> test_block2;
     test_block2(num_threads, TestTaskOMP, std::ref(my_count));
   };
+#ifdef QMC_OMP
   REQUIRE_THROWS_WITH(test_block(num_threads, nested_tasks, std::ref(count)),
                       Catch::Contains("ParallelExecutor should not be used for nested openmp threading"));
+#endif
 }
 
 } // namespace qmcplusplus

src/Message/OpenMP.h

@@ -18,15 +18,15 @@
 #include "config.h"
 #endif
 
-#if defined(ENABLE_OPENMP)
+#if defined(QMC_OMP)
 #include <omp.h>
 #else
+#include "formic/utils/openmp.h"
+
 typedef int omp_int_t;
-inline omp_int_t omp_get_thread_num() { return 0; }
-inline omp_int_t omp_get_max_threads() { return 1; }
-inline omp_int_t omp_get_num_threads() { return 1; }
 inline omp_int_t omp_get_level() { return 0; }
 inline omp_int_t omp_get_ancestor_thread_num(int level) { return 0; }
+inline void omp_set_num_threads(int num_threads) {}
 #endif
 
 /// get the number of threads at the next parallel level

src/QMCApp/qmcapp.cpp

@@ -273,7 +273,7 @@ void output_hardware_info(Communicate* comm, Libxml2Document& doc, xmlNodePtr ro
   doc.addChild(hardware, "mpi", using_mpi);
 
   bool using_openmp = false;
-#ifdef ENABLE_OPENMP
+#ifdef QMC_OMP
   using_openmp = true;
   doc.addChild(hardware, "openmp_threads", omp_get_max_threads());
 #endif

src/QMCDrivers/WFOpt/QMCFixedSampleLinearOptimize.cpp

@@ -319,7 +319,7 @@ bool QMCFixedSampleLinearOptimize::run()
       app_log() << "  Using XS:" << XS << " " << failedTries << " " << stability << std::endl;
       eigenvalue_timer_.start();
       getLowestEigenvector(Right, currentParameterDirections);
-      Lambda   = getNonLinearRescale(currentParameterDirections, S);
+      Lambda = getNonLinearRescale(currentParameterDirections, S);
       eigenvalue_timer_.stop();
       //       biggest gradient in the parameter direction vector
       RealType bigVec(0);
@@ -515,10 +515,10 @@ bool QMCFixedSampleLinearOptimize::processOptXML(xmlNodePtr opt_xml,
 {
   m_param.put(opt_xml);
   targetExcitedStr = lowerCase(targetExcitedStr);
-  targetExcited = (targetExcitedStr == "yes");
+  targetExcited    = (targetExcitedStr == "yes");
 
   block_lmStr = lowerCase(block_lmStr);
-  block_lm = (block_lmStr == "yes");
+  block_lm    = (block_lmStr == "yes");
 
   auto iter = OptimizerNames.find(MinMethod);
   if (iter == OptimizerNames.end())
@@ -545,7 +545,7 @@ bool QMCFixedSampleLinearOptimize::processOptXML(xmlNodePtr opt_xml,
       current_optimizer_type_ != OptimizerType::DESCENT)
     myComm->barrier_and_abort("targetExcited = \"yes\" requires that MinMethod = \"adaptive or descent");
 
-#ifdef ENABLE_OPENMP
+#ifdef QMC_OMP
   if (current_optimizer_type_ == OptimizerType::ADAPTIVE && (omp_get_max_threads() > 1))
   {
     // throw std::runtime_error("OpenMP threading not enabled with AdaptiveThreeShift optimizer. Use MPI for parallelism instead, and set OMP_NUM_THREADS to 1.");

src/QMCDrivers/WFOpt/QMCFixedSampleLinearOptimizeBatched.cpp

@@ -662,10 +662,10 @@ bool QMCFixedSampleLinearOptimizeBatched::processOptXML(xmlNodePtr opt_xml,
 {
   m_param.put(opt_xml);
   targetExcitedStr = lowerCase(targetExcitedStr);
-  targetExcited = (targetExcitedStr == "yes");
+  targetExcited    = (targetExcitedStr == "yes");
 
   block_lmStr = lowerCase(block_lmStr);
-  block_lm = (block_lmStr == "yes");
+  block_lm    = (block_lmStr == "yes");
 
   auto iter = OptimizerNames.find(MinMethod);
   if (iter == OptimizerNames.end())
@@ -680,7 +680,7 @@ bool QMCFixedSampleLinearOptimizeBatched::processOptXML(xmlNodePtr opt_xml,
   if (targetExcited && current_optimizer_type_ != OptimizerType::ADAPTIVE)
     APP_ABORT("targetExcited = \"yes\" requires that MinMethod = \"adaptive");
 
-#ifdef ENABLE_OPENMP
+#ifdef QMC_OMP
   if (current_optimizer_type_ == OptimizerType::ADAPTIVE && (omp_get_max_threads() > 1))
   {
     // throw std::runtime_error("OpenMP threading not enabled with AdaptiveThreeShift optimizer. Use MPI for parallelism instead, and set OMP_NUM_THREADS to 1.");

src/QMCDrivers/WFOpt/QMCLinearOptimize.cpp

@@ -18,10 +18,10 @@
 #include "Particle/HDFWalkerIO.h"
 #include "OhmmsData/AttributeSet.h"
 #include "Message/CommOperators.h"
-//#if defined(ENABLE_OPENMP)
+
 #include "QMCDrivers/VMC/VMC.h"
 #include "QMCDrivers/WFOpt/QMCCostFunction.h"
-//#endif
+
 //#include "QMCDrivers/VMC/VMCSingle.h"
 //#include "QMCDrivers/QMCCostFunctionSingle.h"
 #include "QMCHamiltonians/HamiltonianPool.h"

src/QMCDrivers/tests/test_DMCBatched.cpp

@@ -45,6 +45,7 @@ private:
 /** Since we check the DMC only feature of reserve walkers perhaps this should be
  *  a DMC integration test.
  */
+#ifdef QMC_OMP
 TEST_CASE("DMCDriver+QMCDriverNew integration", "[drivers]")
 {
   using namespace testing;
@@ -95,6 +96,6 @@ TEST_CASE("DMCDriver+QMCDriverNew integration", "[drivers]")
   CHECK(reserved_walkers == 10);
   // What else should we expect after process
 }
-
+#endif
 
 } // namespace qmcplusplus

src/QMCDrivers/tests/test_QMCDriverNew.cpp

@@ -73,9 +73,11 @@ TEST_CASE("QMCDriverNew tiny case", "[drivers]")
   // What else should we expect after process
 }
 
+#ifdef QMC_OMP
 TEST_CASE("QMCDriverNew more crowds than threads", "[drivers]")
 {
   using namespace testing;
+
   Concurrency::OverrideMaxCapacity<> override(8);
   Communicate* comm;
   comm = OHMMS::Controller;
@@ -160,5 +162,6 @@ TEST_CASE("QMCDriverNew walker counts", "[drivers]")
 
   qmc_batched.testAdjustGlobalWalkerCount();
 }
+#endif
 
 } // namespace qmcplusplus

src/QMCTools/CMakeLists.txt

@@ -31,9 +31,11 @@ target_link_libraries(qmc-extract-eshdf-kvectors PUBLIC qmcio)
 
 add_executable(qmc-get-supercell getSupercell.cpp)
 
-add_executable(qmc-check-affinity check-affinity.cpp)
-if(HAVE_MPI)
-  target_link_libraries(qmc-check-affinity MPI::MPI_CXX)
+if(QMC_OMP)
+  add_executable(qmc-check-affinity check-affinity.cpp)
+  if(HAVE_MPI)
+    target_link_libraries(qmc-check-affinity MPI::MPI_CXX)
+  endif()
 endif()
 
 add_executable(convertpw4qmc convertpw4qmc.cpp XmlRep.cpp WriteEshdf.cpp)
@@ -51,9 +53,15 @@ if(USE_OBJECT_TARGET)
   target_link_libraries(qmcfinitesize qmcparticle qmcutil)
 endif()
 
-install(TARGETS convert4qmc qmc-extract-eshdf-kvectors qmc-get-supercell qmc-check-affinity convertpw4qmc qmcfinitesize
+install(TARGETS convert4qmc qmc-extract-eshdf-kvectors qmc-get-supercell convertpw4qmc qmcfinitesize
         RUNTIME DESTINATION bin)
 
+if(QMC_OMP)
+  install(TARGETS qmc-check-affinity
+          RUNTIME DESTINATION bin)
+endif()
+
+
 add_subdirectory(ppconvert)
 
 if(BUILD_UNIT_TESTS)

src/QMCTools/tests/CMakeLists.txt

@@ -31,26 +31,28 @@ add_unit_test(${UTEST_NAME} 1 1 $<TARGET_FILE:${UTEST_EXE}>)
 set_tests_properties(${UTEST_NAME} PROPERTIES WORKING_DIRECTORY ${UTEST_DIR})
 
 # Minimal test for qmc-check-affinity. Check for "OpenMP" in output, no errorcodes.
-if(HAVE_MPI)
-  add_test(NAME check-affinity_runs COMMAND ${MPIEXEC_EXECUTABLE} ${MPIEXEC_NUMPROC_FLAG} 1 ${MPIEXEC_PREFLAGS}
-                                            $<TARGET_FILE:qmc-check-affinity>)
-else()
-  add_test(NAME check-affinity_runs COMMAND $<TARGET_FILE:qmc-check-affinity>)
+if(QMC_OMP)
+  if(HAVE_MPI)
+    add_test(NAME check-affinity_runs COMMAND ${MPIEXEC_EXECUTABLE} ${MPIEXEC_NUMPROC_FLAG} 1 ${MPIEXEC_PREFLAGS}
+                                              $<TARGET_FILE:qmc-check-affinity>)
+  else()
+    add_test(NAME check-affinity_runs COMMAND $<TARGET_FILE:qmc-check-affinity>)
+  endif()
+  set_tests_properties(
+    check-affinity_runs
+    PROPERTIES TIMEOUT
+               120
+               LABELS
+               "deterministic"
+               PASS_REGULAR_EXPRESSION
+               "OpenMP"
+               PROCESSORS
+               1
+               ENVIRONMENT
+               OMP_NUM_THREADS=3
+               PROCESSOR_AFFINITY
+               TRUE)
 endif()
-set_tests_properties(
-  check-affinity_runs
-  PROPERTIES TIMEOUT
-             120
-             LABELS
-             "deterministic"
-             PASS_REGULAR_EXPRESSION
-             "OpenMP"
-             PROCESSORS
-             1
-             ENVIRONMENT
-             OMP_NUM_THREADS=3
-             PROCESSOR_AFFINITY
-             TRUE)
 
 # Basic test for qmc-get-supercell. Cubic primitive vectors, request what should optimally be a 2x2x2 tiling
 if(HAVE_MPI)

src/Utilities/Clock.h

@@ -46,7 +46,7 @@ class CPUClock
 public:
   double operator()()
   {
-#if defined(ENABLE_OPENMP)
+#if defined(QMC_OMP)
     return omp_get_wtime();
 #else
     struct timeval tv;

src/config.h.cmake.in

@@ -33,7 +33,7 @@
 #cmakedefine PRINT_DEBUG @PRINT_DEBUG@
 
 /* Enable OpenMP parallelization. */
-#cmakedefine ENABLE_OPENMP @ENABLE_OPENMP@
+#cmakedefine QMC_OMP @QMC_OMP@
 
 /* Enable OpenMP offload. */
 #cmakedefine ENABLE_OFFLOAD @ENABLE_OFFLOAD@

src/formic/utils/openmp.h

@@ -9,10 +9,6 @@
 //
 // File created by: Jeongnim Kim, jeongnim.kim@gmail.com, University of Illinois at Urbana-Champaign
 //////////////////////////////////////////////////////////////////////////////////////
-    
-    
-
-
 
 
 #ifndef FORMIC_OHMMS_OPENMP_H
@@ -22,21 +18,12 @@
 #include "config.h"
 #endif
 
-#if defined(ENABLE_OPENMP)
+#if defined(QMC_OMP)
 #include <omp.h>
 #else
 typedef int omp_int_t;
-inline omp_int_t omp_get_thread_num()
-{
-  return 0;
-}
-inline omp_int_t omp_get_max_threads()
-{
-  return 1;
-}
-inline omp_int_t omp_get_num_threads()
-{
-  return 1;
-}
+inline omp_int_t omp_get_thread_num() { return 0; }
+inline omp_int_t omp_get_max_threads() { return 1; }
+inline omp_int_t omp_get_num_threads() { return 1; }
 #endif
 #endif // FORMIC_OHMMS_OPENMP_H