IntelGPU backend: Profile data improved (#1932)

* IntelGPU backend: Profile data improved

* PR1932. Comments addressed

src/ngraph/runtime/intelgpu/intelgpu_backend.cpp

@@ -1665,17 +1665,17 @@ vector<runtime::PerformanceCounter>
     return rc;
 }
 
-static Shape get_shape_by_name(const shared_ptr<Function> func, const string& name)
+static Node* get_node_by_name(const shared_ptr<Function> func, const string& name)
 {
     for (shared_ptr<Node> node : func->get_ops())
     {
         if (node->get_name() == name)
         {
-            return node->get_output_shape(0);
+            return node.get();
         }
     }
 
-    return Shape();
+    return nullptr;
 }
 
 void runtime::intelgpu::IntelGPUBackend::print_call_performance(
@@ -1748,7 +1748,7 @@ void runtime::intelgpu::IntelGPUBackend::print_call_performance(
         for (auto it = data.rbegin(); (it != data.rend()) && (limit_count > 0); ++it, --limit_count)
         {
             const string ngraph_node_name = convert_cldnn_names(func, it->second.item_name);
-            const Shape ngraph_node_shape = get_shape_by_name(func, ngraph_node_name);
+            const Node* ngraph_node = get_node_by_name(func, ngraph_node_name);
 
             cout << func_name << delim << setw(max_item_name_size) << it->second.item_name << delim
                  << "time(ms)" << delim << scientific << setprecision(2) << it->first;
@@ -1756,7 +1756,30 @@ void runtime::intelgpu::IntelGPUBackend::print_call_performance(
             {
                 cout << delim << item.first << "(ms)" << delim << item.second;
             }
-            cout << delim << ngraph_node_name << delim << ngraph_node_shape << "\n";
+            cout << delim << ngraph_node_name;
+
+            if (ngraph_node) // it might be initialized by nullptr
+            {
+                // print all input shapes for the Node
+                size_t arg_idx = 0;
+                for (const descriptor::Input& op_input : ngraph_node->get_inputs())
+                {
+                    cout << delim << op_input.get_element_type().c_type_string() << " input"
+                         << arg_idx << vector_to_string(op_input.get_shape());
+                    ++arg_idx;
+                }
+
+                // print all output shapes for the Node
+                arg_idx = 0;
+                for (const descriptor::Output& op_output : ngraph_node->get_outputs())
+                {
+                    cout << delim << op_output.get_element_type().c_type_string() << " output"
+                         << arg_idx << vector_to_string(op_output.get_shape());
+                    ++arg_idx;
+                }
+            }
+
+            cout << "\n";
         }
 
         // Print bottom line summary