diff --git a/llvm/include/llvm/Support/TimeProfiler.h b/llvm/include/llvm/Support/TimeProfiler.h
--- a/llvm/include/llvm/Support/TimeProfiler.h
+++ b/llvm/include/llvm/Support/TimeProfiler.h
@@ -98,6 +98,9 @@
 /// Finish a time trace profiler running on a worker thread.
 void timeTraceProfilerFinishThread();
 
+/// Finish a time trace profiler running on a different thread.
+void timeTraceProfilerFinishThreadExternal(llvm::TimeTraceProfiler *prof);
+
 /// Is the time trace profiler enabled, i.e. initialized?
 inline bool timeTraceProfilerEnabled() {
   return getTimeTraceProfilerInstance() != nullptr;
diff --git a/llvm/lib/Support/TimeProfiler.cpp b/llvm/lib/Support/TimeProfiler.cpp
--- a/llvm/lib/Support/TimeProfiler.cpp
+++ b/llvm/lib/Support/TimeProfiler.cpp
@@ -315,6 +315,14 @@
   TimeTraceProfilerInstance = nullptr;
 }
 
+// Finish TimeTraceProfilerInstance on a different thread
+void llvm::timeTraceProfilerFinishThreadExternal(
+    llvm::TimeTraceProfiler *prof) {
+  auto &Instances = getTimeTraceProfilerInstances();
+  std::lock_guard<std::mutex> Lock(Instances.Lock);
+  Instances.List.push_back(prof);
+}
+
 void llvm::timeTraceProfilerWrite(raw_pwrite_stream &OS) {
   assert(TimeTraceProfilerInstance != nullptr &&
          "Profiler object can't be null");
diff --git a/openmp/libomptarget/src/private.h b/openmp/libomptarget/src/private.h
--- a/openmp/libomptarget/src/private.h
+++ b/openmp/libomptarget/src/private.h
@@ -189,12 +189,18 @@
   }
 }
 
+void checkTimeTraceInitThisThread();
+
 #include "llvm/Support/TimeProfiler.h"
-#define TIMESCOPE() llvm::TimeTraceScope TimeScope(__FUNCTION__)
+#define TIMESCOPE()                                                            \
+  checkTimeTraceInitThisThread();                                              \
+  llvm::TimeTraceScope TimeScope(__FUNCTION__)
 #define TIMESCOPE_WITH_IDENT(IDENT)                                            \
+  checkTimeTraceInitThisThread();                                              \
   SourceInfo SI(IDENT);                                                        \
   llvm::TimeTraceScope TimeScope(__FUNCTION__, SI.getProfileLocation())
 #define TIMESCOPE_WITH_NAME_AND_IDENT(NAME, IDENT)                             \
+  checkTimeTraceInitThisThread();                                              \
   SourceInfo SI(IDENT);                                                        \
   llvm::TimeTraceScope TimeScope(NAME, SI.getProfileLocation())
 #else
diff --git a/openmp/libomptarget/src/rtl.cpp b/openmp/libomptarget/src/rtl.cpp
--- a/openmp/libomptarget/src/rtl.cpp
+++ b/openmp/libomptarget/src/rtl.cpp
@@ -43,6 +43,10 @@
 
 static char *ProfileTraceFile = nullptr;
 
+// List of TimeTraceProfiler instances on other threads
+std::vector<llvm::TimeTraceProfiler *> *profileInstancesOtherThreads = nullptr;
+std::mutex profLock;
+
 __attribute__((constructor(101))) void init() {
   DP("Init target library!\n");
 
@@ -62,8 +66,10 @@
 
   ProfileTraceFile = getenv("LIBOMPTARGET_PROFILE");
   // TODO: add a configuration option for time granularity
-  if (ProfileTraceFile)
+  if (ProfileTraceFile) {
+    profileInstancesOtherThreads = new std::vector<llvm::TimeTraceProfiler *>;
     timeTraceProfilerInitialize(500 /* us */, "libomptarget");
+  }
 }
 
 __attribute__((destructor(101))) void deinit() {
@@ -71,11 +77,29 @@
   delete PM;
 
   if (ProfileTraceFile) {
+    for (auto prof : *profileInstancesOtherThreads)
+      timeTraceProfilerFinishThreadExternal(prof);
     // TODO: add env var for file output
     if (auto E = timeTraceProfilerWrite(ProfileTraceFile, "-"))
       fprintf(stderr, "Error writing out the time trace\n");
 
     timeTraceProfilerCleanup();
+    delete profileInstancesOtherThreads;
+  }
+}
+
+// Check if time trace profiler is already initialized on this thread.
+// If not, initialize it and save a pointer to call finish before
+// writing out the file.
+void checkTimeTraceInitThisThread() {
+  if (ProfileTraceFile) {
+    if (llvm::getTimeTraceProfilerInstance() == nullptr) {
+      DP("initializing profiler on thread");
+      llvm::timeTraceProfilerInitialize(500, "libomptarget");
+      llvm::TimeTraceProfiler *prof = llvm::getTimeTraceProfilerInstance();
+      std::lock_guard<std::mutex> Lock(profLock);
+      profileInstancesOtherThreads->push_back(prof);
+    }
   }
 }