Report internal queue creation to tools.

Debug agent requires handles to internal queues for single step debugging.
Added tools only API hsa_amd_runtime_queue_create_register for reporting.

hsa_amd_runtime_queue_create_register sets a callback which is invoked
when internal queues are created.

Change-Id: Ia5190ae724fadba686c15f25b2cd085350eeff0e


[ROCm/ROCR-Runtime commit: 757502ccd6]

projects/rocr-runtime/runtime/hsa-runtime/core/inc/runtime.h

@@ -328,6 +328,11 @@ class Runtime {
     return system_event_handler_;
   }
 
+  hsa_status_t SetInternalQueueCreateNotifier(hsa_amd_runtime_queue_notifier callback,
+                                              void* user_data);
+
+  void InternalQueueCreateNotify(const hsa_queue_t* queue, hsa_agent_t agent);
+
  protected:
   static void AsyncEventsLoop(void*);
 
@@ -503,6 +508,11 @@ class Runtime {
 
   void* system_event_handler_user_data_;
 
+  // Internal queue creation notifier
+  AMD::callback_t<hsa_amd_runtime_queue_notifier> internal_queue_create_notifier_;
+
+  void* internal_queue_create_notifier_user_data_;
+
   // Holds reference count to runtime object.
   std::atomic<uint32_t> ref_count_;
 

projects/rocr-runtime/runtime/hsa-runtime/core/runtime/amd_gpu_agent.cpp

@@ -513,6 +513,9 @@ core::Queue* GpuAgent::CreateInterceptibleQueue() {
   // Disabled intercept of internal queues pending tools updates.
   core::Queue* queue = nullptr;
   QueueCreate(minAqlSize_, HSA_QUEUE_TYPE_MULTI, NULL, NULL, 0, 0, &queue);
+  if (queue != nullptr)
+    core::Runtime::runtime_singleton_->InternalQueueCreateNotify(core::Queue::Convert(queue),
+                                                                 this->public_handle());
   return queue;
 }
 

projects/rocr-runtime/runtime/hsa-runtime/core/runtime/hsa_api_trace.cpp

@@ -56,6 +56,9 @@ hsa_status_t hsa_amd_queue_intercept_create(
     hsa_agent_t agent_handle, uint32_t size, hsa_queue_type32_t type,
     void (*callback)(hsa_status_t status, hsa_queue_t* source, void* data), void* data,
     uint32_t private_segment_size, uint32_t group_segment_size, hsa_queue_t** queue);
+
+hsa_status_t hsa_amd_runtime_queue_create_register(hsa_amd_runtime_queue_notifier callback,
+                                                   void* user_data);
 }
 
 namespace core {
@@ -382,6 +385,7 @@ void HsaApiTable::UpdateAmdExts() {
   amd_ext_api.hsa_amd_queue_intercept_register_fn = AMD::hsa_amd_queue_intercept_register;
   amd_ext_api.hsa_amd_queue_set_priority_fn = AMD::hsa_amd_queue_set_priority;
   amd_ext_api.hsa_amd_memory_async_copy_rect_fn = AMD::hsa_amd_memory_async_copy_rect;
+  amd_ext_api.hsa_amd_runtime_queue_create_register_fn = AMD::hsa_amd_runtime_queue_create_register;
 }
 
 class Init {

projects/rocr-runtime/runtime/hsa-runtime/core/runtime/hsa_ext_amd.cpp

@@ -924,4 +924,12 @@ hsa_status_t HSA_API hsa_amd_queue_set_priority(hsa_queue_t* queue,
   CATCH;
 }
 
+hsa_status_t hsa_amd_runtime_queue_create_register(hsa_amd_runtime_queue_notifier callback,
+                                                   void* user_data) {
+  TRY;
+  IS_OPEN();
+  return core::Runtime::runtime_singleton_->SetInternalQueueCreateNotifier(callback, user_data);
+  CATCH;
+}
+
 } // end of AMD namespace

projects/rocr-runtime/runtime/hsa-runtime/core/runtime/runtime.cpp

@@ -1476,4 +1476,20 @@ hsa_status_t Runtime::SetCustomSystemEventHandler(hsa_amd_system_event_callback_
   }
 }
 
+hsa_status_t Runtime::SetInternalQueueCreateNotifier(hsa_amd_runtime_queue_notifier callback,
+                                                     void* user_data) {
+  if (internal_queue_create_notifier_) {
+    return HSA_STATUS_ERROR;
+  } else {
+    internal_queue_create_notifier_ = callback;
+    internal_queue_create_notifier_user_data_ = user_data;
+    return HSA_STATUS_SUCCESS;
+  }
+}
+
+void Runtime::InternalQueueCreateNotify(const hsa_queue_t* queue, hsa_agent_t agent) {
+  if (internal_queue_create_notifier_)
+    internal_queue_create_notifier_(queue, agent, internal_queue_create_notifier_user_data_);
+}
+
 }  // namespace core

projects/rocr-runtime/runtime/hsa-runtime/inc/hsa_api_trace.h

@@ -92,6 +92,11 @@ hsa_status_t hsa_amd_queue_intercept_create(
     void (*callback)(hsa_status_t status, hsa_queue_t* source, void* data), void* data,
     uint32_t private_segment_size, uint32_t group_segment_size, hsa_queue_t** queue);
 
+typedef void (*hsa_amd_runtime_queue_notifier)(const hsa_queue_t* queue, hsa_agent_t agent,
+                                               void* data);
+hsa_status_t hsa_amd_runtime_queue_create_register(hsa_amd_runtime_queue_notifier callback,
+                                                   void* user_data);
+
 // Structure of Version used to identify an instance of Api table
 // Must be the first member (offsetof == 0) of all API tables.
 // This is the root of the table passing ABI.
@@ -173,6 +178,7 @@ struct AmdExtTable {
   decltype(hsa_amd_queue_intercept_register)* hsa_amd_queue_intercept_register_fn;
   decltype(hsa_amd_queue_set_priority)* hsa_amd_queue_set_priority_fn;
   decltype(hsa_amd_memory_async_copy_rect)* hsa_amd_memory_async_copy_rect_fn;
+  decltype(hsa_amd_runtime_queue_create_register)* hsa_amd_runtime_queue_create_register_fn;
 };
 
 // Table to export HSA Core Runtime Apis