[Fix] [CPU Backend] : Prepack weights for w8a8 oneDNN matmul (#33901)

Signed-off-by: nikhil-arm <nikhil.gupta2@arm.com>

csrc/cpu/dnnl_helper.cpp

@@ -237,12 +237,20 @@ W8A8MatMulPrimitiveHandler::W8A8MatMulPrimitiveHandler(const Args& args)
   };
   dnnl::memory::desc original_b_md({b_k_size_, b_n_size_}, b_type_,
                                    {b_k_stride_, b_n_stride_});
+#ifdef __aarch64__
+  // dummy M size for prepacking weights
+  // Prepacking weights improves performance and avoid runtime reorders
+  constexpr dnnl_dim_t kProbeM = 128;
+#else
+  constexpr dnnl_dim_t kProbeM = DNNL_RUNTIME_DIM_VAL;
+#endif
+
   prepack_weight(args.b_ptr, original_b_md,
                  create_primitive_desc(
-                     MSizeCacheKey{.a_m_size = DNNL_RUNTIME_DIM_VAL,
+                     MSizeCacheKey{.a_m_size = kProbeM,
                                    .use_bias = false,
                                    .bias_type = dnnl::memory::data_type::undef},
-                     true)
+                     /*first_time=*/true)
                      .weights_desc());
   init_runtime_memory_cache(args);
 }