[CPU] Extend AMX deconv int8 to support oscale+eltwise+eltwise post ops pattern (#14026)

* Extend AMX deconv int8 to support oscale+eltwise+eltwise post ops pattern.

* Add deconv AMX test cases.

* Add multi-eltwise test on deconv AMX int8.

* Applied review comments.

* Update onednn submodule.

src/plugins/intel_cpu/tests/functional/single_layer_tests/convolution_backprop_data.cpp

@@ -297,12 +297,20 @@ const std::vector<std::vector<ptrdiff_t>> padBegins2d = { {0, 0} };
 const std::vector<std::vector<ptrdiff_t>> padEnds2d = { {0, 0} };
 const std::vector<InferenceEngine::SizeVector> dilations2d = { {1, 1} };
 
+
+const std::vector<InferenceEngine::SizeVector> deconvAmxKernels2d = { {3, 3}, {2, 2}};
+const std::vector<InferenceEngine::SizeVector> deconvAmxStrides2d = { {2, 2}};
+
 /* ============= Deconvolution params (3D) ============= */
 const std::vector<InferenceEngine::SizeVector> kernels3d = { {3, 3, 3}, {1, 1, 1} };
 const std::vector<InferenceEngine::SizeVector> strides3d = { {1, 1, 1}, {2, 2, 2} };
 const std::vector<std::vector<ptrdiff_t>> padBegins3d = { {0, 0, 0} };
 const std::vector<std::vector<ptrdiff_t>> padEnds3d = { {0, 0, 0} };
 const std::vector<InferenceEngine::SizeVector> dilations3d = { {1, 1, 1} };
+
+const std::vector<InferenceEngine::SizeVector> deconvAmxKernels3d = { {3, 3, 3}, {2, 2, 2} };
+const std::vector<InferenceEngine::SizeVector> deconvAmxStrides3d = {  {2, 2, 2} };
+
 /* ============= */
 
 /* INSTANCES */
@@ -488,6 +496,7 @@ const std::vector<DeconvInputData> Blocked_2D_inputs_smoke = {
     }
 };
 
+
 const auto convParams_ExplicitPadding_Blocked_2D_nightly = ::testing::Combine(
     ::testing::ValuesIn(kernels2d),
     // Use 7x7 with stride 1 is too small to generate 15x15 output. It needs a big negative pad which will result
@@ -530,6 +539,17 @@ const auto convParams_ExplicitPadding_Blocked_2D = ::testing::Combine(
     ::testing::ValuesIn(emptyOutputPadding)
 );
 
+const auto convParams_ExplicitPadding_AMX_2D = ::testing::Combine(
+    ::testing::ValuesIn(deconvAmxKernels2d),
+    ::testing::ValuesIn(deconvAmxStrides2d),
+    ::testing::ValuesIn(padBegins2d),
+    ::testing::ValuesIn(padEnds2d),
+    ::testing::ValuesIn(dilations2d),
+    ::testing::ValuesIn(numOutChannels_Blocked),
+    ::testing::Values(ngraph::op::PadType::EXPLICIT),
+    ::testing::ValuesIn(emptyOutputPadding)
+);
+
 INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_Blocked_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         convParams_ExplicitPadding_Blocked_2D,
@@ -550,6 +570,26 @@ INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_Blocked_BF16, DeconvolutionLayerCPUTest
         ::testing::Values(cpuBF16PluginConfig)),
     DeconvolutionLayerCPUTest::getTestCaseName);
 
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_NSPC_BF16_AMX_NO_FUSING, DeconvolutionLayerCPUTest,
+    ::testing::Combine(
+        convParams_ExplicitPadding_AMX_2D,
+        ::testing::ValuesIn(Blocked_2D_inputs_smoke),
+        ::testing::Values(ElementType::f32),
+        ::testing::ValuesIn({emptyFusingSpec}),
+        ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_2D_nspc_amx})),
+        ::testing::Values(cpuBF16PluginConfig)),
+    DeconvolutionLayerCPUTest::getTestCaseName);
+
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_2D_NSPC_INT8_AMX, DeconvolutionLayerCPUTest,
+    ::testing::Combine(
+        convParams_ExplicitPadding_AMX_2D,
+        ::testing::ValuesIn(Blocked_2D_inputs_smoke),
+        ::testing::Values(ElementType::i8),
+        ::testing::ValuesIn({emptyFusingSpec, fusingClampRoundAddRelu}),
+        ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_2D_nspc_amx})),
+        ::testing::Values(cpuEmptyPluginConfig)),
+    DeconvolutionLayerCPUTest::getTestCaseName);
+
 INSTANTIATE_TEST_SUITE_P(nightly_Deconv_2D_Blocked_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         convParams_ExplicitPadding_Blocked_2D_nightly,
@@ -624,6 +664,17 @@ const auto convParams_ExplicitPadding_Blocked_3D = ::testing::Combine(
     ::testing::ValuesIn(emptyOutputPadding)
 );
 
+const auto convParams_ExplicitPadding_AMX_3D = ::testing::Combine(
+    ::testing::ValuesIn(deconvAmxKernels3d),
+    ::testing::ValuesIn(deconvAmxStrides3d),
+    ::testing::ValuesIn(padBegins3d),
+    ::testing::ValuesIn(padEnds3d),
+    ::testing::ValuesIn(dilations3d),
+    ::testing::Values(32),
+    ::testing::Values(ngraph::op::PadType::EXPLICIT),
+    ::testing::ValuesIn(emptyOutputPadding)
+);
+
 INSTANTIATE_TEST_SUITE_P(smoke_Deconv_3D_Blocked_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         convParams_ExplicitPadding_Blocked_3D,
@@ -644,6 +695,27 @@ INSTANTIATE_TEST_SUITE_P(smoke_Deconv_3D_Blocked_BF16, DeconvolutionLayerCPUTest
         ::testing::Values(cpuBF16PluginConfig)),
     DeconvolutionLayerCPUTest::getTestCaseName);
 
+
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_3D_NSPC_BF16_AMX_NO_FUSING, DeconvolutionLayerCPUTest,
+    ::testing::Combine(
+        convParams_ExplicitPadding_AMX_3D,
+        ::testing::ValuesIn(Blocked_3D_inputs_smoke),
+        ::testing::Values(ElementType::f32),
+        ::testing::ValuesIn({emptyFusingSpec}),
+        ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_3D_nspc_amx})),
+        ::testing::Values(cpuBF16PluginConfig)),
+    DeconvolutionLayerCPUTest::getTestCaseName);
+
+INSTANTIATE_TEST_SUITE_P(smoke_Deconv_3D_NSPC_INT8_AMX, DeconvolutionLayerCPUTest,
+    ::testing::Combine(
+        convParams_ExplicitPadding_AMX_3D,
+        ::testing::ValuesIn(Blocked_3D_inputs_smoke),
+        ::testing::Values(ElementType::i8),
+        ::testing::ValuesIn({emptyFusingSpec, fusingClampRoundAddRelu}),
+        ::testing::ValuesIn(filterCPUInfoForDevice({conv_avx512_3D_nspc_amx})),
+        ::testing::Values(cpuEmptyPluginConfig)),
+    DeconvolutionLayerCPUTest::getTestCaseName);
+
 INSTANTIATE_TEST_SUITE_P(nightly_Deconv_3D_Blocked_FP32, DeconvolutionLayerCPUTest,
     ::testing::Combine(
         convParams_ExplicitPadding_Blocked_3D_nightly,

src/plugins/intel_cpu/tests/functional/test_utils/fusing_test_utils.hpp

@@ -244,6 +244,22 @@ const auto fusingScaleShift = fusingSpecificParams{ std::make_shared<postNodesMg
                 return std::make_shared<ngraph::opset1::Add>(cfg.input, constNode);
             }, "Add(PerChannel)"}}), {"Add"} };
 
+const auto fusingClampRoundAddRelu = fusingSpecificParams{ std::make_shared<postNodesMgr>(std::vector<postNodeBuilder>{
+            {[](postNodeConfig& cfg){
+                return ngraph::builder::makeActivation(cfg.input, cfg.type, ngraph::helpers::Clamp, {}, {3.0f, 6.0f});
+            }, "Clamp"},
+            {[](postNodeConfig& cfg){
+                return ngraph::builder::makeActivation(cfg.input, cfg.type, ngraph::helpers::RoundHalfToEven);
+            }, "RoundHalfToEven"},
+            {[](postNodeConfig& cfg){
+                ngraph::Shape secondMultInShape(1, 1);
+                auto secondMultInput = ngraph::builder::makeConstant(cfg.type, secondMultInShape, std::vector<float>{}, true);
+                return std::make_shared<ngraph::opset1::Add>(cfg.input, secondMultInput);
+            }, "AddPerTensor"},
+            {[](postNodeConfig& cfg){
+                return ngraph::builder::makeActivation(cfg.input, cfg.type, ngraph::helpers::Relu);
+            }, "Relu"}}), {"Clamp", "Round", "Add", "Relu"}};
+
 const auto fusingScaleShiftAndFakeQuantizePerChannel = fusingSpecificParams{ std::make_shared<postNodesMgr>(std::vector<postNodeBuilder>{
             {[](postNodeConfig& cfg) {
                 ngraph::Shape newShape = generatePerChannelShape(cfg.target);