SWDEV-510319 - Fixed random segfaults in graph tests

This change fixes random segfaults in graph tests that
are seen after the change make internal callbacks non-blocking.
The callback thread that decreases the GraphExec ref count
may now run after the runtime shutdown. This can cause a segfault
because the hip::device that is accessed in GraphExec destructor
is already destroyed during runtime shutdown. This patch ensures
that the hip::device object  stays alive until after the
callback thread completes.

Change-Id: I75a6ac01f27a0b2250bbd10ed389ebfb322927af

hipamd/src/hip_graph_internal.cpp

@@ -363,6 +363,7 @@ hipError_t GraphExec::Init() {
     status = CaptureAQLPackets();
   }
   instantiateDeviceId_ = hip::getCurrentDevice()->deviceId();
+  static_cast<ReferenceCountedObject*>( hip::getCurrentDevice())->retain();
   return status;
 }
 

hipamd/src/hip_graph_internal.hpp

@@ -505,7 +505,6 @@ struct Graph {
     amd::ScopedLock lock(graphSetLock_);
     graphSet_.insert(this);
     mem_pool_ = device->GetGraphMemoryPool();
-    mem_pool_->retain();
     graphInstantiated_ = false;
     roots_.resize(DEBUG_HIP_FORCE_GRAPH_QUEUES);
     leafs_.resize(DEBUG_HIP_FORCE_GRAPH_QUEUES);
@@ -543,9 +542,6 @@ struct Graph {
       }
     }
     graphUserObj_.clear();
-    if (mem_pool_ != nullptr) {
-      mem_pool_->release();
-    }
     memAllocNodePtrs_.clear();
   }
 
@@ -764,6 +760,9 @@ struct GraphExec : public amd::ReferenceCountedObject, public Graph {
         kernArgManager_->release();
       }
     }
+    if (instantiateDeviceId_ != -1) {
+      static_cast<ReferenceCountedObject*>(g_devices[instantiateDeviceId_])->release();
+    }
   }
 
   Node GetClonedNode(Node node) {