Changed USE_NVTX to USE_NVTX_API and removed add_definition

CMakeLists.txt

@@ -678,15 +678,6 @@ ELSE(QMC_CUDA)
   MESSAGE(STATUS "Disabling CUDA")
 ENDIF(QMC_CUDA)
 
-SET(USE_NVTX 0 CACHE BOOL "Enable/disable NVTX regions in CUDA code.")
-IF(USE_NVTX)
-  IF(HAVE_CUDA)
-    ADD_DEFINITIONS(-DUSE_NVTX)
-    LINK_DIRECTORIES("${CUDA_TOOLKIT_ROOT_DIR}/lib64")
-    LINK_LIBRARIES("nvToolsExt")
-  ENDIF(HAVE_CUDA)
-ENDIF(USE_NVTX)
-
 
 #INCLUDE(${PROJECT_CMAKE}/FindPkgConfig.cmake)
 ##################################################################
@@ -790,6 +781,14 @@ IF (USE_VTUNE_API)
   LINK_LIBRARIES("${VTUNE_ITTNOTIFY_LIBRARY}")
 ENDIF()
 
+SET(USE_NVTX 0 CACHE BOOL "Enable/disable NVTX regions in CUDA code.")
+IF(USE_NVTX_API)
+  IF(HAVE_CUDA)
+    LINK_DIRECTORIES("${CUDA_TOOLKIT_ROOT_DIR}/lib64")
+    LINK_LIBRARIES("nvToolsExt")
+  ENDIF(HAVE_CUDA)
+ENDIF(USE_NVTX_API)
+
 
 #include(ExternalProject)
 #  set(einspline_PREFIX "${CMAKE_CURRENT_BINARY_DIR}/einspline")

src/QMCDrivers/DMC/DMC_CUDA.cpp

@@ -26,7 +26,7 @@
 #include "Utilities/RunTimeManager.h"
 #include "Message/CommOperators.h"
 #include "type_traits/scalar_traits.h"
-#ifdef USE_NVTX
+#ifdef USE_NVTX_API
 #include <nvToolsExt.h>
 #endif
 
@@ -78,7 +78,7 @@ void DMCcuda::checkBounds (std::vector<PosType> &newpos,
 
 bool DMCcuda::run()
 {
-#ifdef USE_NVTX
+#ifdef USE_NVTX_API
   nvtxRangePushA("DMC:run");
 #endif
   bool scaleweight = ScaleWeight == "yes";
@@ -330,7 +330,7 @@ bool DMCcuda::run()
     }
   }
   while(block<nBlocks && enough_time_for_next_iteration);
-#ifdef USE_NVTX
+#ifdef USE_NVTX_API
   nvtxRangePop();
 #endif
   //finalize a qmc section

src/QMCDrivers/VMC/VMC_CUDA.cpp

@@ -25,7 +25,7 @@
 #include "type_traits/scalar_traits.h"
 #include "Utilities/RunTimeManager.h"
 #include "qmc_common.h"
-#ifdef USE_NVTX
+#ifdef USE_NVTX_API
 #include <nvToolsExt.h>
 #endif
 
@@ -129,7 +129,7 @@ bool VMCcuda::run()
 {
   if (UseDrift == "yes")
     return runWithDrift();
-#ifdef USE_NVTX
+#ifdef USE_NVTX_API
   nvtxRangePushA("VMC:run");
 #endif
   resetRun();
@@ -248,7 +248,7 @@ bool VMCcuda::run()
     std::cerr << "At the end of VMC" << std::endl;
     gpu::cuda_memory_manager.report();
   }
-#ifdef USE_NVTX
+#ifdef USE_NVTX_API
   nvtxRangePop();
 #endif
   return finalize(block);
@@ -340,7 +340,7 @@ void VMCcuda::advanceWalkersWithDrift()
 
 bool VMCcuda::runWithDrift()
 {
-#ifdef USE_NVTX
+#ifdef USE_NVTX_API
   nvtxRangePushA("VMC:runWithDrift");
 #endif
   resetRun();
@@ -439,7 +439,7 @@ bool VMCcuda::runWithDrift()
     std::cerr << "At the end of VMC with drift" << std::endl;
     gpu::cuda_memory_manager.report();
   }
-#ifdef USE_NVTX
+#ifdef USE_NVTX_API
   nvtxRangePop();
 #endif
   return finalize(block);

src/config.h.cmake.in

@@ -194,9 +194,6 @@
 /* Use SOA version of AA distance table */
 #cmakedefine ENABLE_SOA  @ENABLE_SOA@
 
-/* Enable NVTX regions in CUDA code. */
-#cmakedefine USE_NVTX @USE_NVTX@
-
 #if (__cplusplus >= 201103L)
   #if defined(__INTEL_COMPILER)
     #if defined(__KNC__) || defined(__AVX512F__)
@@ -248,5 +245,8 @@
 /* Use VTune Task API with timers */
 #cmakedefine USE_VTUNE_TASKS @USE_VTUNE_TASKS@
 
+/* Enable NVTX regions in CUDA code. */
+#cmakedefine USE_NVTX_API @USE_NVTX_API@
+
 #endif // QMCPLUSPLUS_CONFIGURATION_H