[Snippets][CPU] Enabled N, K blocking for MatMul FP32 with transposed_b=true

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

@@ -29,8 +29,7 @@ bool pass::AdjustBrgemmCopyBLoopPorts::update_loop_info(
                  *  1) VNNI format is applied: KN4k for I8/U8, or KN2k for BF16
                  *  2) Zero padding is applied if N4k < 256 or N2k < 64
                  */
-                if (brgemm_utils::with_repacking(brg->get_type()) && precision != element::f32 &&
-                    loop_port.is_incremented()) {
+                if (brgemm_utils::with_repacking(brg->get_type()) && loop_port.is_incremented()) {
                     // K blocking loop: account for zero padding
                     if (loop_port.get_dim_idx() == 1) {
                         const auto ptr_incr = loop_desc.ptr_increment;

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

@@ -54,9 +54,11 @@ std::tuple<size_t, size_t, size_t> BrgemmCPUBlocking::get_blocking_params(
 
     size_t m_blk, n_blk, k_blk;
     std::tie(m_blk, n_blk, k_blk) = BrgemmBlockingBase::get_blocking_params(brgemm_expr);
-    // Note: K,N blocking is functionally enabled, need to turn it on after blocking heuristic is updated to cover
-    // the low precision cases (ticket: 156014)
-    if (with_repacking(brgemm->get_type())) {
+    // [TODO]: K,N blocking is functionally enabled, need to turn it on after blocking heuristic is updated to cover
+    //         the low precision cases (ticket: 156014)
+    //         Please note that FP32 MatMul with `transposed_b=true` has type `with_repacking` despite the precision.
+    const auto precision = brgemm_expr->get_node()->get_input_element_type(1);
+    if (with_repacking(brgemm->get_type()) && precision != element::f32) {
         n_blk = get_full_dim_value();
         k_blk = get_full_dim_value();
     }

src/plugins/intel_cpu/tests/unit/snippets_transformations/x64/lowered/brgemm_blocking.cpp

@@ -193,6 +193,42 @@ TEST_F(BrgemmCPUBlockingTest, BlockingIsNotNeeded) {
     }
 }
 
+TEST_F(BrgemmCPUBlockingTest, WithTransposeB) {
+    const ov::Dimension::value_type m = 384;
+    const ov::Dimension::value_type k = 1024;
+    const ov::Dimension::value_type n = 384;
+    const ov::PartialShape input_shape_a{1, 16, m, k};
+    const ov::PartialShape input_shape_b{1, 16, n, k};
+    const auto precision_a = ov::element::f32;
+    const auto precision_b = ov::element::f32;
+    const std::vector<size_t> layout_input{0, 1, 3, 2};
+
+    {
+        auto data_a = linear_ir->push_node<ov::opset10::Parameter>(precision_a, input_shape_a);
+        auto data_b = linear_ir->push_node<ov::opset10::Parameter>(precision_b, input_shape_b);
+        auto copy_b = linear_ir->push_node<BrgemmCopyB>(data_b.second, precision_a, BRGEMM_TYPE::REPACKING_ONLY, 0, 0, 0, layout_input);
+        init_expr_descriptors(*copy_b.first);
+
+        auto brgemm = linear_ir->push_node<BrgemmCPU>(data_a.second, copy_b.second, BRGEMM_TYPE::REPACKING_ONLY);
+        init_expr_descriptors(*brgemm.first);
+        auto result = linear_ir->push_node<ov::opset10::Result>(brgemm.second);
+    }
+    {
+        auto data_a = linear_ir_ref->push_node<ov::opset10::Parameter>(precision_a, input_shape_a);
+        auto data_b = linear_ir_ref->push_node<ov::opset10::Parameter>(precision_b, input_shape_b);
+        auto copy_b = linear_ir_ref->push_node<BrgemmCopyB>(data_b.second, precision_a, BRGEMM_TYPE::REPACKING_ONLY, 0, 0, 0, layout_input);
+        const auto copy_b_expr = *copy_b.first;
+        init_expr_descriptors(copy_b_expr, {{full_dim, full_dim}, {full_dim, full_dim}});
+
+        auto brgemm = linear_ir_ref->push_node<BrgemmCPU>(data_a.second, copy_b.second, BRGEMM_TYPE::REPACKING_ONLY);
+        const auto& brgemm_expr = *brgemm.first;
+        init_expr_descriptors(brgemm_expr, {{m_blk, k_blk}, {k_blk, n_blk}, {m_blk, n_blk}});
+        create_brgemm_loop_infos(linear_ir_ref, brgemm_expr, m, m_blk, k, k_blk, n, n_blk);
+        brgemm_expr->set_loop_ids({2, 1, 0});
+        auto result = linear_ir_ref->push_node<ov::opset10::Result>(brgemm.second);
+    }
+}
+
 TEST_F(BrgemmCPUBlockingTest, WithDataRepacking) {
     // Skipped because K,N blocking is disabled until heuristic is updated (ticket: 156014)
     GTEST_SKIP();