CPU: Implement op::NotEqual

src/ngraph/runtime/cpu/emitter.cpp

@@ -492,3 +492,22 @@ void Emitter::EMITTER_DECL(EmitNegative)
                    EIGEN_VECTOR_FORMAT(inputs[0].get_layout<DenseTensorViewLayout>()->get_size()) ");\n"
           "    }\n";
 }
+
+void Emitter::EMITTER_DECL(EmitNotEqual)
+{
+    const element::Type& et = (dynamic_pointer_cast<const TensorViewType>(
+                                       n->get_arguments().at(0)->get_value_type()))
+                                      ->get_element_type();
+
+    TU += "    {\n"
+          "        auto arg0 = call_frame->get_tensor_view_data<" + element_type_names[TI(et)] + ">(" + to_string(inputs[0].get_index()) + ");\n"
+          "        auto arg1 = call_frame->get_tensor_view_data<" + element_type_names[TI(et)] + ">(" + to_string(inputs[1].get_index()) + ");\n"
+          "        auto out  = call_frame->get_tensor_view_data<Bool>(" + to_string(outputs[0].get_index()) + ");\n"
+          "        EigenArray1d<Bool>(out, "
+                   EIGEN_VECTOR_FORMAT(outputs[0].get_layout<DenseTensorViewLayout>()->get_size()) ") =\n"
+          "        (EigenArray1d<" + element_type_names[TI(et)] + ">(arg0, "
+                   EIGEN_VECTOR_FORMAT(inputs[0].get_layout<DenseTensorViewLayout>()->get_size()) ") !=\n"
+          "        EigenArray1d<" + element_type_names[TI(et)] + ">(arg1, "
+                   EIGEN_VECTOR_FORMAT(inputs[1].get_layout<DenseTensorViewLayout>()->get_size()) ")).template cast<char>();\n"
+          "    }\n";
+}

src/ngraph/runtime/cpu/emitter.hpp

@@ -60,6 +60,7 @@ namespace ngraph
                 void EMITTER_DECL(EmitLog);
                 void EMITTER_DECL(EmitMaximum);
                 void EMITTER_DECL(EmitNegative);
+                void EMITTER_DECL(EmitNotEqual);
 
             };
         }

src/ngraph/runtime/cpu/external_function.cpp

@@ -86,7 +86,8 @@ static const OpMap dispatcher{{TI(ngraph::op::Add), &Emitter::EmitAdd},
                               {TI(ngraph::op::LessEq), &Emitter::EmitLessEq},
                               {TI(ngraph::op::Log), &Emitter::EmitLog},
                               {TI(ngraph::op::Maximum), &Emitter::EmitMaximum},
-                              {TI(ngraph::op::Negative), &Emitter::EmitNegative}
+                              {TI(ngraph::op::Negative), &Emitter::EmitNegative},
+                              {TI(ngraph::op::NotEqual), &Emitter::EmitNotEqual}
 
                              };
 

test/cpu.cpp

@@ -947,8 +947,7 @@ TEST(cpu, negative)
     ASSERT_EQ((vector<float>{-1, 2, 0, 4.8f, -8.6f, 8.6f}), result->get_vector());
 }
 
-/*
-TEST(execute, notequal)
+TEST(cpu, notequal)
 {
     auto shape = Shape{2, 2, 2};
     auto A = make_shared<op::Parameter>(element::Float32::element_type(), shape);
@@ -956,7 +955,7 @@ TEST(execute, notequal)
     auto rt = make_shared<TensorViewType>(element::Bool::element_type(), shape);
     auto f = make_shared<Function>(make_shared<op::NotEqual>(A, B), rt, op::Parameters{A, B});
 
-    auto manager = runtime::Manager::get("NGVM");
+    auto manager = runtime::Manager::get("CPU");
     auto external = manager->compile(f);
     auto backend = manager->allocate_backend();
     auto cf = backend->make_call_frame(external);
@@ -972,6 +971,7 @@ TEST(execute, notequal)
     ASSERT_EQ((vector<char>{0, 0, 1, 1, 1, 0, 0, 1}), result->get_vector());
 }
 
+/*
 TEST(execute, select)
 {
     auto shape = Shape{2, 2, 2};