[GPU] Fix concat_input_order pass

Author: sshlyapn

src/plugins/intel_gpu/src/graph/graph_optimizer/concat_input_order.cpp

@@ -48,7 +48,7 @@ bool can_shuffle_features(program_node& node, program_node& concat_node, stream&
     if (pass_through) {
         // Primitives that are feature order invariant, pass-through shuffled features to users
         for (auto& user : node.get_users()) {
-            if (!can_shuffle_features(*user, concat_node, stream))
+            if (!can_shuffle_features(*user, node, stream))
                 return false;
         }
         return true;

src/plugins/intel_gpu/tests/unit/test_cases/concatenation_gpu_test.cpp

@@ -723,28 +723,24 @@ TEST(concat_gpu, no_exception_in_input_order_opt_b_fs_yx_fsv16_with_conv_port2)
     auto concat_input1 = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { 1, 48, 6, 6 }});
     auto concat_input2 = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { 1, 96, 6, 6 }});
     auto concat_input3 = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { 1, 128, 6, 6 }});
-    auto conv_input = engine.allocate_memory({ data_types::f32, format::b_fs_yx_fsv16, { 1, 192, 6, 6 } });
-    auto weights0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 296, 192, 1, 1 } });
+    auto weights0 = engine.allocate_memory({ data_types::f32, format::bfyx, { 296, 296, 1, 1 } });
 
     std::vector<float> concat_input0_data(concat_input0->get_layout().count());
     std::vector<float> concat_input1_data(concat_input1->get_layout().count());
     std::vector<float> concat_input2_data(concat_input2->get_layout().count());
     std::vector<float> concat_input3_data(concat_input3->get_layout().count());
-    std::vector<float> conv_input_data(conv_input->get_layout().count());
     std::vector<float> weights0_data(weights0->get_layout().count());
 
     std::iota(concat_input0_data.begin(), concat_input0_data.end(), 0.f);
     std::iota(concat_input1_data.begin(), concat_input1_data.end(), 0.f);
     std::iota(concat_input2_data.begin(), concat_input2_data.end(), 0.f);
     std::iota(concat_input3_data.begin(), concat_input3_data.end(), 0.f);
-    std::iota(conv_input_data.begin(), conv_input_data.end(), 0.f);
     std::iota(weights0_data.begin(), weights0_data.end(), 0.f);
 
     set_values(concat_input0, concat_input0_data);
     set_values(concat_input1, concat_input1_data);
     set_values(concat_input2, concat_input2_data);
     set_values(concat_input3, concat_input3_data);
-    set_values(conv_input, conv_input_data);
     set_values(weights0, weights0_data);
 
     layout reorder_layout(data_types::f32, format::b_fs_yx_fsv16, {1, 296, 6, 6});
@@ -753,16 +749,15 @@ TEST(concat_gpu, no_exception_in_input_order_opt_b_fs_yx_fsv16_with_conv_port2)
                       input_layout("concat_input1", concat_input1->get_layout()),
                       input_layout("concat_input2", concat_input2->get_layout()),
                       input_layout("concat_input3", concat_input3->get_layout()),
-                      input_layout("conv_input", conv_input->get_layout()),
                       concatenation("concat",
                                     { input_info("concat_input0"), input_info("concat_input1"), input_info("concat_input2"), input_info("concat_input3")  },
                                     1,
                                     data_types::f32,
                                     padding{{0, 0, 0, 0}, 0}),
+                      pooling("pooling", input_info("concat"), pooling_mode::max, { 2, 2 }, { 1, 1 }),
                       data("weights0", weights0),
-                      convolution("conv0", input_info("conv_input"), "weights0", "", 1, { 1, 1 }, {1, 1}, {0, 0}, {0, 0}, false),
-                      eltwise("eltwise", input_info("conv0"), input_info("concat"), eltwise_mode::sum),
-                      permute("permute", input_info("eltwise"), {0, 1, 2, 3}));
+                      convolution("conv0", input_info("pooling"), "weights0", "", 1, { 1, 1 }, {1, 1}, {0, 0}, {0, 0}, false),
+                      permute("permute", input_info("conv0"), {0, 1, 2, 3}));
 
     ov::intel_gpu::ExecutionConfig config = get_test_default_config(engine);
     config.set_property(ov::intel_gpu::optimize_data(true));
@@ -772,7 +767,6 @@ TEST(concat_gpu, no_exception_in_input_order_opt_b_fs_yx_fsv16_with_conv_port2)
     network.set_input_data("concat_input1", concat_input1);
     network.set_input_data("concat_input2", concat_input2);
     network.set_input_data("concat_input3", concat_input3);
-    network.set_input_data("conv_input", conv_input);
 
     ASSERT_NO_FATAL_FAILURE(network.execute());
 }