Fixes and cleanup

src/common/snippets/src/runtime_configurator.cpp

@@ -261,7 +261,6 @@ void RuntimeConfigurator::update_data_offsets(const std::vector<VectorDims>& sha
                                               const std::vector<std::vector<size_t>>& layouts) const {
     OPENVINO_ASSERT(shapes.size() == m_io_num, "Number of custom shapes must be 0 or be equal to m_io_num");
     OPENVINO_ASSERT(layouts.size() == m_io_num, "Number of custom layouts must be 0 or be equal to m_io_num");
-    std::cout << "[ INFO ] Offsets\n";
     for (size_t i = 0; i <  m_io_num; ++i) {
         // offsets represent distance between consecutive elements of corresponding dimension.
         // If a dim size == 1, then the next dim starts immediately and the stride is 0
@@ -282,7 +281,6 @@ void RuntimeConfigurator::update_data_offsets(const std::vector<VectorDims>& sha
         const auto idx_stride = m_config->tensor_rank - shape.size();
         compute_offsets(shape, offsets, m_config->tensor_rank, m_io_data_sizes[i], idx_stride);
 
-        std::cout << "\t offsets[" << i << "] = " << ov::PartialShape(offsets) << std::endl;
         const auto& layout = layouts[i];
         if (!layout.empty()) {
             std::vector<size_t> reordered_offsets(offsets.size());

src/plugins/intel_cpu/src/emitters/snippets/cpu_runtime_configurator.cpp

@@ -144,7 +144,7 @@ void CPURuntimeConfigurator::update_requested_descs(const ov::snippets::lowered:
         const bool brgemm_with_extracted_repacking =
             std::any_of(consumers.begin(), consumers.end(), [](const ov::snippets::lowered::ExpressionPort& port) {
                 auto brgemm = ov::as_type_ptr<ov::intel_cpu::BrgemmCPU>(port.get_expr()->get_node());
-                return brgemm && brgemm_utils::with_repacking(brgemm->get_type());
+                return port.get_index() == 1 && brgemm && brgemm_utils::with_repacking(brgemm->get_type());
             });
         if (brgemm_with_extracted_repacking) {
             const auto& desc = param->get_output_port_descriptor(0);
@@ -183,15 +183,9 @@ void CPURuntimeConfigurator::adjust_offsets_from_descs(const ov::snippets::lower
             const auto& blocked_shape = optimal_desc->as<DnnlBlockedMemoryDesc>()->getBlockDims();
 
             ov::snippets::VectorDims shape_for_offset(m_config->tensor_rank - original_shape.size(), 1);
-            // Parallel work amount is copied from original shape
-            shape_for_offset.insert(shape_for_offset.end(), original_shape.begin(), original_shape.end() - m_config->tile_rank);
-            // Only first dim is batch, the rest are repacked KN
-            shape_for_offset.insert(shape_for_offset.end(), blocked_shape.begin() + 1, blocked_shape.end());
-            std::cout << "[ INFO ] shape_for_offset = " << ov::PartialShape(shape_for_offset) << std::endl;
-
+            shape_for_offset.insert(shape_for_offset.end(), blocked_shape.begin(), blocked_shape.end());
             auto& offsets = m_config->io_data_offsets[i];
             compute_offsets(shape_for_offset, offsets, shape_for_offset.size(), m_io_data_sizes[i], 0);
-            std::cout << "[ INFO ] offsets[*] = " << ov::PartialShape(offsets) << std::endl;
             OPENVINO_ASSERT(ov::snippets::utils::is_planar_layout(parameter->get_output_port_descriptor(0)->get_layout()));
         }
     }

src/plugins/intel_cpu/src/transformations/snippets/x64/pass/lowered/adjust_brgemm_copy_b_loop_ports.cpp

@@ -86,7 +86,6 @@ bool pass::AdjustBrgemmCopyBLoopPorts::run(const snippets::lowered::LinearIR& li
             }
         }
         for (const auto& target_port : ports) {
-            std::cout << "[ INFO ] AdjustBrgemmCopyBLoopPorts works\n";
             const auto &target_loop_ids = target_port.get_expr()->get_loop_ids();
 
             // If loop ids match, it means there is no blocking loop

src/plugins/intel_cpu/src/transformations/snippets/x64/pass/move_brgemm_repacking_out.cpp

@@ -50,7 +50,9 @@ pass::MoveBrgemmRepackingOut::MoveBrgemmRepackingOut() {
         if (!ov::snippets::utils::is_planar_layout(layout) || copy_b_node->get_output_size() != 1 ||
             transformation_callback(copy_b_node))
             return false;
-        return ov::replace_output_update_name(copy_b_out, copy_b_in);
+        return false;
+        std::cout << "[ INFO ] MoveBrgemmRepackingOut is finished\n";
+        // return ov::replace_output_update_name(copy_b_out, copy_b_in);
     };
 
     auto m = std::make_shared<ov::pass::pattern::Matcher>(m_copy_b, matcher_name);