diff --git a/src/cpu/rnn/rnn_reorders.hpp b/src/cpu/rnn/rnn_reorders.hpp
index fbfedd01..28a6dc1d 100644
--- a/src/cpu/rnn/rnn_reorders.hpp
+++ b/src/cpu/rnn/rnn_reorders.hpp
@@ -128,7 +128,7 @@ struct rnn_weights_reorder_t : public cpu_primitive_t {
             return status::success;
         }
 
-        format_tag_t itag_;
+        format_tag_t itag_ = mkldnn_format_tag_undef;
 
     private:
         void init_scratchpad() {
diff --git a/src/cpu/simple_concat.hpp b/src/cpu/simple_concat.hpp
index fe84eb14..f1b2bd17 100644
--- a/src/cpu/simple_concat.hpp
+++ b/src/cpu/simple_concat.hpp
@@ -98,9 +98,9 @@ struct simple_concat_t: public cpu_primitive_t {
             return status::success;
         }
 
-        int perm_[MKLDNN_MAX_NDIMS];
-        int iperm_[MKLDNN_MAX_NDIMS];
-        dims_t blocks_;
+        int perm_[MKLDNN_MAX_NDIMS] {};
+        int iperm_[MKLDNN_MAX_NDIMS] {};
+        dims_t blocks_ {};
 
         dim_t nelems_to_concat(const memory_desc_wrapper &data_d) const {
             const int ndims = data_d.ndims();
diff --git a/src/cpu/simple_sum.hpp b/src/cpu/simple_sum.hpp
index 46a24810..c47c4a64 100644
--- a/src/cpu/simple_sum.hpp
+++ b/src/cpu/simple_sum.hpp
@@ -71,7 +71,7 @@ struct simple_sum_t: public cpu_primitive_t {
         }
 
         sum_bf16_params_t bf16_p_;
-        dim_t block_size_, nelems_, blocks_number_, tail_;
+        dim_t block_size_ = 0, nelems_= 0, blocks_number_= 0, tail_= 0;
 
     private:
         const dim_t cacheline_size_ = 64; // bytes