diff --git a/CMakeLists.txt b/CMakeLists.txt
index fee7e0e7..c2d3262e 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -230,8 +230,11 @@ enable_cxx_compiler_flag_if_supported("-Wall")
 enable_cxx_compiler_flag_if_supported("-Wextra")
 enable_cxx_compiler_flag_if_supported("-Wpedantic")
 enable_cxx_compiler_flag_if_supported("-pedantic")
+# PGI and Intel don't like the concurrentqueue code.
 if (NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "PGI")
-    enable_cxx_compiler_flag_if_supported("-Werror")
+    if (NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "Intel")
+        enable_cxx_compiler_flag_if_supported("-Werror")
+    endif (NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "Intel")
 endif (NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "PGI")
 
 # ---------------------------------------------------
@@ -279,6 +282,7 @@ if ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
         set(CMAKE_CXX_SUPPORT_FLAG "-std=c++14" CACHE STRING "CXX Support Flag" FORCE)
     endif()
 endif()
+
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${CMAKE_CXX_SUPPORT_FLAG}")
 
 # Check if architecture is x86 or not
diff --git a/cmake/Modules/FindOMPT.cmake b/cmake/Modules/FindOMPT.cmake
index 84f1a7c0..0fd448e6 100644
--- a/cmake/Modules/FindOMPT.cmake
+++ b/cmake/Modules/FindOMPT.cmake
@@ -32,7 +32,7 @@ endif()
 pkg_check_modules(PC_OMPT QUIET OMPT)
 set(OMPT_DEFINITIONS ${PC_OMPT_CFLAGS_OTHER})
 
-find_path(OMPT_INCLUDE_DIR ompt.h
+find_path(OMPT_INCLUDE_DIR omp-tools.h
           HINTS ${PC_OMPT_INCLUDEDIR} ${PC_OMPT_INCLUDE_DIRS} ${OMPT_ROOT}/include)
 
 find_library(OMPT_LIBRARY NAMES omp iomp5 gomp
diff --git a/cmake/tests/ompt_test/ompt_test.cpp b/cmake/tests/ompt_test/ompt_test.cpp
index 424da4d2..35f57d65 100644
--- a/cmake/tests/ompt_test/ompt_test.cpp
+++ b/cmake/tests/ompt_test/ompt_test.cpp
@@ -2,7 +2,7 @@
 
 #include "stdio.h"
 #include <omp.h>
-#include <ompt.h>
+#include <omp-tools.h>
 #include <iostream>
 
 #define cb_t(name) (ompt_callback_t)&name
@@ -49,11 +49,11 @@ on_ompt_callback_parallel_end(
 inline static void register_callback(ompt_callbacks_t name, ompt_callback_t cb) {
   int ret = ompt_set_callback(name, cb);
 
-  switch(ret) { 
+  switch(ret) {
     case ompt_set_never:
-      fprintf(stderr, "TAU: WARNING: Callback for event %d could not be registered\n", name); 
-      break; 
-    case ompt_set_sometimes: 
+      fprintf(stderr, "TAU: WARNING: Callback for event %d could not be registered\n", name);
+      break;
+    case ompt_set_sometimes:
       printf("TAU: Callback for event %d registered with return value %s\n", name, "ompt_set_sometimes");
       break;
     case ompt_set_sometimes_paired:
@@ -95,7 +95,7 @@ extern "C" int ompt_initialize(
   register_callback(ompt_callback_parallel_end, cb_t(on_ompt_callback_parallel_end));
   //register_callback(ompt_callback_task_create, cb_t(on_ompt_callback_task_create));
   //register_callback(ompt_callback_task_schedule, cb_t(on_ompt_callback_task_schedule));
-  //register_callback(ompt_callback_implicit_task, cb_t(on_ompt_callback_implicit_task)); //Sometimes high-overhead, but unfortunately we cannot avoid this as it is a required event 
+  //register_callback(ompt_callback_implicit_task, cb_t(on_ompt_callback_implicit_task)); //Sometimes high-overhead, but unfortunately we cannot avoid this as it is a required event
   //register_callback(ompt_callback_thread_begin, cb_t(on_ompt_callback_thread_begin));
   //register_callback(ompt_callback_thread_end, cb_t(on_ompt_callback_thread_end));
 
diff --git a/src/apex/activity_trace_async.cpp b/src/apex/activity_trace_async.cpp
index 83f71a4b..292c9fb4 100644
--- a/src/apex/activity_trace_async.cpp
+++ b/src/apex/activity_trace_async.cpp
@@ -18,7 +18,9 @@
 #include "thread_instance.hpp"
 #include "apex_options.hpp"
 #include "trace_event_listener.hpp"
+#ifdef APEX_HAVE_OTF2
 #include "otf2_listener.hpp"
+#endif
 
 static void __attribute__((constructor)) initTrace(void);
 //static void __attribute__((destructor)) flushTrace(void);
@@ -109,12 +111,13 @@ void store_profiler_data(const std::string &name, uint32_t correlationId,
             (apex::trace_event_listener*)instance->the_trace_event_listener;
         tel->on_async_event(device, context, stream, prof);
     }
+#ifdef APEX_HAVE_OTF2
     if (apex::apex_options::use_otf2()) {
         apex::otf2_listener * tol =
             (apex::otf2_listener*)instance->the_otf2_listener;
         tol->on_async_event(device, context, stream, prof);
     }
-
+#endif
     // have the listeners handle the end of this task
     instance->complete_task(tt);
 }
diff --git a/src/apex/apex_ompt.cpp b/src/apex/apex_ompt.cpp
index 4dfca81e..e81195ea 100644
--- a/src/apex/apex_ompt.cpp
+++ b/src/apex/apex_ompt.cpp
@@ -3,7 +3,7 @@
 //  Distributed under the Boost Software License, Version 1.0. (See accompanying
 //  file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
 
-#include <ompt.h>
+#include <omp-tools.h>
 #include <unordered_map>
 #include "string.h"
 #include "stdio.h"
@@ -25,8 +25,9 @@ fprintf( stderr, __VA_ARGS__ ); fflush(stderr); \
 #endif
 
 std::mutex apex_apex_threadid_mutex;
-std::atomic<uint64_t> apex_numthreads(0);
-APEX_NATIVE_TLS uint64_t apex_threadid(-1);
+std::atomic<uint64_t> apex_numthreads{0};
+APEX_NATIVE_TLS int64_t apex_threadid{-1};
+static std::atomic<bool> enabled{false};
 
 class linked_timer {
     public:
@@ -90,7 +91,10 @@ namespace apex {
  */
 void ompt_force_shutdown(void) {
     DEBUG_PRINT("Forcing shutdown of OpenMP Tools API\n");
+    /* The Intel generated code has some odd destructor race conditions, so
+     * don't force the runtime to shut down. */
     /* OpenMP might not have been used... */
+    enabled = false;
     if (ompt_finalize_tool) {
         ompt_finalize_tool();
     }
@@ -102,6 +106,7 @@ void ompt_force_shutdown(void) {
 
 void apex_ompt_start(const char * state, ompt_data_t * ompt_data,
         ompt_data_t * region_data, bool auto_start) {
+    if (!enabled) { return; }
     static std::shared_ptr<apex::task_wrapper> nothing(nullptr);
     linked_timer* tmp;
     /* First check if there's a parent "region" - could be a task */
@@ -135,6 +140,7 @@ void apex_ompt_start(const char * state, ompt_data_t * ompt_data,
 }
 
 void apex_ompt_stop(ompt_data_t * ompt_data) {
+    if (!enabled) { return; }
     APEX_ASSERT(ompt_data->ptr);
     void* tmp = ((linked_timer*)(ompt_data->ptr))->prev;
     delete((linked_timer*)(ompt_data->ptr));
@@ -152,6 +158,7 @@ extern "C" void apex_thread_begin(
     ompt_thread_t thread_type,   /* type of thread */
     ompt_data_t *thread_data     /* data of thread */)
 {
+    if (!enabled) { return; }
     APEX_UNUSED(thread_data);
     {
         std::unique_lock<std::mutex> l(apex_apex_threadid_mutex);
@@ -182,6 +189,7 @@ extern "C" void apex_thread_begin(
 extern "C" void apex_thread_end(
     ompt_data_t *thread_data              /* data of thread                      */
 ) {
+    if (!enabled) { return; }
     APEX_UNUSED(thread_data);
     apex::exit_thread();
 }
@@ -195,6 +203,7 @@ static void apex_parallel_region_begin (
     int flags,                                   /* flags */
     const void *codeptr_ra                       /* return address of runtime call      */
 ) {
+    if (!enabled) { return; }
     APEX_UNUSED(encountering_task_data);
     APEX_UNUSED(encountering_task_frame);
     APEX_UNUSED(requested_team_size);
@@ -212,6 +221,7 @@ static void apex_parallel_region_end (
     int flags,                            /* flags              */
     const void *codeptr_ra                /* return address of runtime call      */
 ) {
+    if (!enabled) { return; }
     APEX_UNUSED(encountering_task_data);
     APEX_UNUSED(flags);
     APEX_UNUSED(codeptr_ra);
@@ -228,6 +238,7 @@ extern "C" void apex_task_create (
     int has_dependences,                        /* created task has dependences   */
     const void *codeptr_ra                      /* return address of runtime call */
 ) {
+    if (!enabled) { return; }
     APEX_UNUSED(encountering_task_frame);
     APEX_UNUSED(has_dependences);
     APEX_UNUSED(codeptr_ra);
@@ -289,6 +300,7 @@ extern "C" void apex_task_schedule(
     ompt_task_status_t prior_task_status, /* status of prior task */
     ompt_data_t *next_task_data           /* data of next task    */
     ) {
+    if (!enabled) { return; }
     DEBUG_PRINT("%lu: Task Schedule prior: %p, status: %d, next: %p\n", apex_threadid, (void*)prior_task_data, prior_task_status, (void*)next_task_data);
     if (prior_task_data != nullptr) {
         linked_timer* prior = (linked_timer*)(prior_task_data->ptr);
@@ -327,6 +339,7 @@ extern "C" void apex_implicit_task(
     unsigned int thread_num,        /* thread number of calling thread */
     int flags
   ) {
+    if (!enabled) { return; }
     APEX_UNUSED(team_size);
     APEX_UNUSED(thread_num);
     APEX_UNUSED(flags);
@@ -358,6 +371,7 @@ extern "C" void apex_target (
     ompt_id_t target_id,
     const void *codeptr_ra
 ) {
+    if (!enabled) { return; }
 }
 
 /* Event #9, target data */
@@ -369,6 +383,7 @@ extern "C" void apex_target_data_op (
     void *device_addr,
     size_t bytes
 ) {
+    if (!enabled) { return; }
 }
 
 /* Event #10, target submit */
@@ -376,6 +391,7 @@ extern "C" void apex_target_submit (
     ompt_id_t target_id,
     ompt_id_t host_op_id
 ) {
+    if (!enabled) { return; }
 }
 
 /* Event #11, tool control */
@@ -385,6 +401,7 @@ extern "C" void apex_control(
     void *arg,             /* argument of control call            */
     const void *codeptr_ra /* return address of runtime call      */
     ) {
+    if (!enabled) { return; }
 }
 
 /* Event #12, device initialize */
@@ -395,12 +412,14 @@ extern "C" void apex_device_initialize (
     ompt_function_lookup_t lookup,
     const char *documentation
 ) {
+    if (!enabled) { return; }
 }
 
 /* Event #13, device finalize */
 extern "C" void apex_device_finalize (
     uint64_t device_num
 ) {
+    if (!enabled) { return; }
 }
 
 /* Event #14, device load */
@@ -414,6 +433,7 @@ extern "C" void apex_device_load_t (
     void * device_addr,
     uint64_t module_id
 ) {
+    if (!enabled) { return; }
 }
 
 /* Event #15, device load */
@@ -421,6 +441,7 @@ extern "C" void apex_device_unload (
     uint64_t device_num,
     uint64_t module_id
 ) {
+    if (!enabled) { return; }
 }
 
 #endif // placeholder functions
@@ -441,6 +462,7 @@ extern "C" void apex_sync_region_wait (
     ompt_data_t *task_data,         /* data of task                   */
     const void *codeptr_ra          /* return address of runtime call */
 ) {
+    if (!enabled) { return; }
     char * tmp_str;
     static const char * barrier_str = "Barrier Wait";
     static const char * barrier_i_str = "Implicit Barrier Wait";
@@ -504,6 +526,7 @@ extern "C" void apex_ompt_work (
     uint64_t count,                 /* quantity of work               */
     const void *codeptr_ra          /* return address of runtime call */
     ) {
+    if (!enabled) { return; }
     APEX_UNUSED(count); // unused on end
 
     char * tmp_str;
@@ -584,6 +607,7 @@ extern "C" void apex_ompt_master (
     ompt_data_t *task_data,         /* data of task                        */
     const void *codeptr_ra          /* return address of runtime call      */
 ) {
+    if (!enabled) { return; }
     if (endpoint == ompt_scope_begin) {
         if (codeptr_ra != nullptr) {
             char regionIDstr[128] = {0};
@@ -660,6 +684,7 @@ extern "C" void apex_ompt_flush (
     ompt_data_t *thread_data, /* data of thread                      */
     const void *codeptr_ra    /* return address of runtime call      */
 ) {
+    if (!enabled) { return; }
     APEX_UNUSED(thread_data);
     if (codeptr_ra != nullptr) {
         char regionIDstr[128] = {0};
@@ -676,6 +701,7 @@ extern "C" void apex_ompt_cancel (
     int flags,                /* cancel flags                        */
     const void *codeptr_ra    /* return address of runtime call      */
 ) {
+    if (!enabled) { return; }
     char regionIDstr[128] = {0};
     if (flags & ompt_cancel_parallel) {
         if (codeptr_ra != nullptr) {
@@ -747,6 +773,7 @@ extern "C" void apex_ompt_cancel (
 extern "C" void apex_ompt_idle (
     ompt_scope_endpoint_t endpoint /* endpoint of idle time               */
 ) {
+    if (!enabled) { return; }
     static APEX_NATIVE_TLS apex::profiler* p = nullptr;
     if (endpoint == ompt_scope_begin) {
         p = apex::start("OpenMP Idle");
@@ -935,6 +962,7 @@ int ompt_initialize(ompt_function_lookup_t lookup, int initial_device_num,
         }
 
     }
+    enabled = true;
 
     DEBUG_PRINT("done.\n"); fflush(stderr);
     return 1;
@@ -944,6 +972,7 @@ void ompt_finalize(ompt_data_t* tool_data)
 {
     APEX_UNUSED(tool_data);
     DEBUG_PRINT("OpenMP runtime is shutting down...\n");
+    enabled = false;
     apex::finalize();
 }
 
diff --git a/src/openmp/ompt_master.c b/src/openmp/ompt_master.c
index aba78a00..a778cdc2 100644
--- a/src/openmp/ompt_master.c
+++ b/src/openmp/ompt_master.c
@@ -4,7 +4,6 @@
 
 int main (void) {
 	int a, i;
- 	apex_init(__func__, 0, 1);
     apex_set_use_screen_output(1);
 #pragma omp parallel shared(a) private(i)
 	{
@@ -16,5 +15,4 @@ int main (void) {
 #pragma omp master
   		printf ("Sum is %d\n", a);
 	}
-    apex_finalize();
 }
\ No newline at end of file
diff --git a/src/openmp/ompt_parallel_region.c b/src/openmp/ompt_parallel_region.c
index de46437e..9f8d313e 100644
--- a/src/openmp/ompt_parallel_region.c
+++ b/src/openmp/ompt_parallel_region.c
@@ -4,7 +4,6 @@
 #include "apex.h"
 
 int main (int argc, char** argv) {
- 	apex_init(__func__, 0, 1);
     apex_set_use_screen_output(1);
 #pragma omp parallel
     {
diff --git a/src/openmp/ompt_reduction.c b/src/openmp/ompt_reduction.c
index bc20303b..31ef6ceb 100644
--- a/src/openmp/ompt_reduction.c
+++ b/src/openmp/ompt_reduction.c
@@ -11,7 +11,6 @@
 int main (int argc, char** argv) {
 	int   i, n, chunk;
  	double a[100], b[100], result;
- 	apex_init(__func__, 0, 1);
     apex_set_use_screen_output(1);
 
  	/* Some initializations */
@@ -30,7 +29,6 @@ int main (int argc, char** argv) {
 
  	printf("Final result= %f\n",result);
 
-    apex_finalize();
     return 0;
 }
 
diff --git a/src/openmp/ompt_sections.c b/src/openmp/ompt_sections.c
index c6e5afeb..0f56bb47 100644
--- a/src/openmp/ompt_sections.c
+++ b/src/openmp/ompt_sections.c
@@ -13,7 +13,6 @@ void c() {
 }
 
 int main (void) {
- 	apex_init(__func__, 0, 1);
     apex_set_use_screen_output(1);
 #pragma omp parallel sections
     {
@@ -24,5 +23,4 @@ int main (void) {
     #pragma omp section
         c();
     }
-    apex_finalize();
 }
\ No newline at end of file
diff --git a/src/openmp/ompt_single.c b/src/openmp/ompt_single.c
index 6d8eef61..eb46a578 100644
--- a/src/openmp/ompt_single.c
+++ b/src/openmp/ompt_single.c
@@ -4,7 +4,6 @@
 
 int main (void) {
 	int a, i;
- 	apex_init(__func__, 0, 1);
     apex_set_use_screen_output(1);
 #pragma omp parallel shared(a) private(i)
 	{
@@ -16,5 +15,4 @@ int main (void) {
 #pragma omp single
   		printf ("Sum is %d\n", a);
 	}
-    apex_finalize();
 }
\ No newline at end of file
diff --git a/src/openmp/ompt_sync_region_wait.c b/src/openmp/ompt_sync_region_wait.c
index 97e3d1cb..bb438ffd 100644
--- a/src/openmp/ompt_sync_region_wait.c
+++ b/src/openmp/ompt_sync_region_wait.c
@@ -10,7 +10,6 @@
 int main (int argc, char** argv) {
 	int   i, n, chunk;
  	double a[100], b[100], result;
- 	apex_init(__func__, 0, 1);
     apex_set_use_screen_output(1);
 
  	/* Some initializations */
@@ -33,7 +32,6 @@ int main (int argc, char** argv) {
 
  	printf("Final result= %f\n",result);
 
-    apex_finalize();
     return 0;
 }
 
diff --git a/src/openmp/ompt_target.c b/src/openmp/ompt_target.c
index 44a9804a..69254122 100644
--- a/src/openmp/ompt_target.c
+++ b/src/openmp/ompt_target.c
@@ -8,7 +8,6 @@ double start;
 double end;
 int main (void) {
   int sum = 0;
- 	apex_init(__func__, 0, 1);
     apex_set_use_screen_output(1);
   start = omp_get_wtime();
   printf("Num devices available: %d\n",omp_get_num_devices() );
@@ -21,6 +20,5 @@ int main (void) {
   end = omp_get_wtime();
   printf ("time %f\n",(end-start));
   printf("sum = %d\n",sum);
-    apex_finalize();
   return 0;
 }
diff --git a/src/openmp/ompt_thread.c b/src/openmp/ompt_thread.c
index 7b02580b..b3c4990b 100644
--- a/src/openmp/ompt_thread.c
+++ b/src/openmp/ompt_thread.c
@@ -4,7 +4,6 @@
 #include "apex.h"
 
 int main (int argc, char** argv) {
- 	apex_init(__func__, 0, 1);
     apex_set_use_screen_output(1);
 #pragma omp parallel
     {
@@ -13,7 +12,6 @@ int main (int argc, char** argv) {
             omp_get_num_threads());
         fflush(stdout);
     }
-    apex_finalize();
     return 0;
 }
 
diff --git a/src/unit_tests/C++/CMakeLists.txt b/src/unit_tests/C++/CMakeLists.txt
index 19b099de..b12bc81d 100644
--- a/src/unit_tests/C++/CMakeLists.txt
+++ b/src/unit_tests/C++/CMakeLists.txt
@@ -57,6 +57,8 @@ endif()
 
 # std::threads crash when linked statically. :(
 if (NOT BUILD_STATIC_EXECUTABLES)
+# Intel can't do std::futures (std::__once_callable)
+if (NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "Intel")
   if ((NOT OTF2_FOUND))
     set(example_programs "${example_programs};apex_fibonacci_std_async;apex_fibonacci_std_async2")
   endif()
@@ -66,6 +68,7 @@ if (NOT BUILD_STATIC_EXECUTABLES)
       set(example_programs "${example_programs};apex_throttle_event")
     endif (APEX_THROTTLE)
   endif ((NOT DEFINED TAU_ROOT) AND (NOT USE_TAU) AND (NOT TAU_FOUND))
+endif (NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "Intel")
 endif()
 
 foreach(example_program ${example_programs})