1 files changed, 33 insertions, 14 deletions

diff --git a/offload/liboffload/src/OffloadImpl.cpp b/offload/liboffload/src/OffloadImpl.cpp
index 6486b2b..f5365ca 100644
--- a/offload/liboffload/src/OffloadImpl.cpp
+++ b/offload/liboffload/src/OffloadImpl.cpp
@@ -72,6 +72,8 @@ struct ol_queue_impl_t {
 struct ol_event_impl_t {
   ol_event_impl_t(void *EventInfo, ol_queue_handle_t Queue)
       : EventInfo(EventInfo), Queue(Queue) {}
+  // EventInfo may be null, in which case the event should be considered always
+  // complete
   void *EventInfo;
   ol_queue_handle_t Queue;
 };
@@ -206,7 +208,7 @@ Error initPlugins(OffloadContext &Context) {
 }
 
 Error olInit_impl() {
-  std::lock_guard<std::mutex> Lock{OffloadContextValMutex};
+  std::lock_guard<std::mutex> Lock(OffloadContextValMutex);
 
   if (isOffloadInitialized()) {
     OffloadContext::get().RefCount++;
@@ -224,7 +226,7 @@ Error olInit_impl() {
 }
 
 Error olShutDown_impl() {
-  std::lock_guard<std::mutex> Lock{OffloadContextValMutex};
+  std::lock_guard<std::mutex> Lock(OffloadContextValMutex);
 
   if (--OffloadContext::get().RefCount != 0)
     return Error::success();
@@ -485,16 +487,13 @@ Error olSyncQueue_impl(ol_queue_handle_t Queue) {
   // Host plugin doesn't have a queue set so it's not safe to call synchronize
   // on it, but we have nothing to synchronize in that situation anyway.
   if (Queue->AsyncInfo->Queue) {
-    if (auto Err = Queue->Device->Device->synchronize(Queue->AsyncInfo))
+    // We don't need to release the queue and we would like the ability for
+    // other offload threads to submit work concurrently, so pass "false" here
+    // so we don't release the underlying queue object.
+    if (auto Err = Queue->Device->Device->synchronize(Queue->AsyncInfo, false))
       return Err;
   }
 
-  // Recreate the stream resource so the queue can be reused
-  // TODO: Would be easier for the synchronization to (optionally) not release
-  // it to begin with.
-  if (auto Res = Queue->Device->Device->initAsyncInfo(&Queue->AsyncInfo))
-    return Res;
-
   return Error::success();
 }
 
@@ -509,8 +508,8 @@ Error olWaitEvents_impl(ol_queue_handle_t Queue, ol_event_handle_t *Events,
       return Plugin::error(ErrorCode::INVALID_NULL_HANDLE,
                            "olWaitEvents asked to wait on a NULL event");
 
-    // Do nothing if the event is for this queue
-    if (Event->Queue == Queue)
+    // Do nothing if the event is for this queue or the event is always complete
+    if (Event->Queue == Queue || !Event->EventInfo)
       continue;
 
     if (auto Err = Device->waitEvent(Event->EventInfo, Queue->AsyncInfo))
@@ -528,6 +527,12 @@ Error olGetQueueInfoImplDetail(ol_queue_handle_t Queue,
   switch (PropName) {
   case OL_QUEUE_INFO_DEVICE:
     return Info.write<ol_device_handle_t>(Queue->Device);
+  case OL_QUEUE_INFO_EMPTY: {
+    auto Pending = Queue->Device->Device->hasPendingWork(Queue->AsyncInfo);
+    if (auto Err = Pending.takeError())
+      return Err;
+    return Info.write<bool>(!*Pending);
+  }
   default:
     return createOffloadError(ErrorCode::INVALID_ENUMERATION,
                               "olGetQueueInfo enum '%i' is invalid", PropName);
@@ -548,6 +553,10 @@ Error olGetQueueInfoSize_impl(ol_queue_handle_t Queue, ol_queue_info_t PropName,
 }
 
 Error olSyncEvent_impl(ol_event_handle_t Event) {
+  if (!Event->EventInfo)
+    // Event always complete
+    return Plugin::success();
+
   if (auto Res = Event->Queue->Device->Device->syncEvent(Event->EventInfo))
     return Res;
 
@@ -555,8 +564,9 @@ Error olSyncEvent_impl(ol_event_handle_t Event) {
 }
 
 Error olDestroyEvent_impl(ol_event_handle_t Event) {
-  if (auto Res = Event->Queue->Device->Device->destroyEvent(Event->EventInfo))
-    return Res;
+  if (Event->EventInfo)
+    if (auto Res = Event->Queue->Device->Device->destroyEvent(Event->EventInfo))
+      return Res;
 
   return olDestroy(Event);
 }
@@ -590,7 +600,16 @@ Error olGetEventInfoSize_impl(ol_event_handle_t Event, ol_event_info_t PropName,
 }
 
 Error olCreateEvent_impl(ol_queue_handle_t Queue, ol_event_handle_t *EventOut) {
+  auto Pending = Queue->Device->Device->hasPendingWork(Queue->AsyncInfo);
+  if (auto Err = Pending.takeError())
+    return Err;
+
   *EventOut = new ol_event_impl_t(nullptr, Queue);
+  if (!*Pending)
+    // Queue is empty, don't record an event and consider the event always
+    // complete
+    return Plugin::success();
+
   if (auto Res = Queue->Device->Device->createEvent(&(*EventOut)->EventInfo))
     return Res;
 
@@ -725,7 +744,7 @@ Error olGetSymbol_impl(ol_program_handle_t Program, const char *Name,
                        ol_symbol_kind_t Kind, ol_symbol_handle_t *Symbol) {
   auto &Device = Program->Image->getDevice();
 
-  std::lock_guard<std::mutex> Lock{Program->SymbolListMutex};
+  std::lock_guard<std::mutex> Lock(Program->SymbolListMutex);
 
   switch (Kind) {
   case OL_SYMBOL_KIND_KERNEL: {