Do not allow in place slice if the arg is in CONSTANT buffer set. (#2625)

* Do not allow in place slice if the arg is in CONSTANT buffer set.

* Add a unit test.

* Address PR feedback.

src/ngraph/runtime/cpu/pass/cpu_memory_assignment.cpp

@@ -518,10 +518,21 @@ void runtime::cpu::pass::CPUMemoryAssignment::build_buffer_sets_maps(list<shared
                                 }
                                 if (!no_in_place)
                                 {
+                                    auto bufferID = get_bufferID(input_tensor);
+                                    auto input_buffer_it = m_bufferID_to_tensorSets.find(bufferID);
+                                    NGRAPH_ASSERT(input_buffer_it !=
+                                                  m_bufferID_to_tensorSets.end());
+
                                     if (node->description() == "Slice")
+                                    {
+                                        if (input_buffer_it->second.first !=
+                                            CPUTensorRole::CONSTANT)
                                         {
                                             // build in place slice chain
                                             in_place_slice_chain.insert(output_tensor);
+                                            input_buffer_it->second.second.insert(output_tensor);
+                                            m_tensor_to_bufferID[output_tensor] = bufferID;
+                                        }
                                     }
                                     else
                                     {
@@ -531,11 +542,6 @@ void runtime::cpu::pass::CPUMemoryAssignment::build_buffer_sets_maps(list<shared
                                         {
                                             in_place_slice_chain.insert(output_tensor);
                                         }
-                                    }
-                                    auto bufferID = get_bufferID(input_tensor);
-                                    auto input_buffer_it = m_bufferID_to_tensorSets.find(bufferID);
-                                    NGRAPH_ASSERT(input_buffer_it !=
-                                                  m_bufferID_to_tensorSets.end());
                                         input_buffer_it->second.second.insert(output_tensor);
                                         m_tensor_to_bufferID[output_tensor] = bufferID;
                                     }
@@ -544,6 +550,7 @@ void runtime::cpu::pass::CPUMemoryAssignment::build_buffer_sets_maps(list<shared
                         }
                     }
                 }
+            }
             // process output tensors
             for (auto i = 0; i < node->get_output_size(); i++)
             {

test/cpu_test.cpp

@@ -873,6 +873,30 @@ TEST(cpu_test, memory_reuse_in_place_slice_after_in_place_concat)
     EXPECT_TRUE(test::all_close_f((vector<float>{3, 7}), read_vector<float>(result)));
 }
 
+TEST(cpu_test, memory_reuse_in_place_slice_after_in_place_reshape_from_constant)
+{
+    Shape shape_a{2, 1, 2, 2};
+    Shape shape_r{2, 1, 2, 2};
+    vector<float> a_data(shape_size(shape_a));
+    iota(a_data.begin(), a_data.end(), 1);
+
+    auto A = op::Constant::create(element::f32, shape_a, a_data);
+    auto reshape = make_shared<op::Reshape>(A, AxisVector{0, 1, 2, 3}, shape_r);
+    Shape shape{1, 1, 2, 2};
+    auto slice = make_shared<op::Slice>(reshape, Coordinate{1, 0, 0, 0}, Coordinate{2, 1, 2, 2});
+    auto neg = make_shared<op::Negative>(slice);
+    auto f = make_shared<Function>(neg, ParameterVector{});
+
+    auto backend = runtime::Backend::create("CPU");
+
+    auto result = backend->create_tensor(element::f32, shape);
+
+    auto handle = backend->compile(f);
+    handle->call_with_validate({result}, {});
+    EXPECT_TRUE(test::all_close_f(
+        vector<float>{-5., -6., -7., -8.}, read_vector<float>(result), MIN_FLOAT_TOLERANCE_BITS));
+}
+
 TEST(cpu_test, convert_inplace)
 {
     Shape shape{2, 2};