diff --git a/poseidon2/src/external.rs b/poseidon2/src/external.rs
index c94700cd9..6a61abc55 100644
--- a/poseidon2/src/external.rs
+++ b/poseidon2/src/external.rs
@@ -2,7 +2,7 @@
 
 use ark_ff::PrimeField;
 
-use crate::add_rc_and_sbox;
+use crate::{add_rcs, s_box};
 
 /// The fastest 4x4 MDS matrix.
 /// [ 2 3 1 1 ]
@@ -88,8 +88,9 @@ pub(crate) fn permute_state<F: PrimeField, const T: usize>(
     rc: &'static [F; T],
     d: usize,
 ) {
-    for i in 0..T {
-        add_rc_and_sbox(&mut state[i], rc[i], d)
+    add_rcs(state, rc);
+    for s in state.iter_mut() {
+        s_box(s, d);
     }
     matmul_external(state);
 }
diff --git a/poseidon2/src/internal.rs b/poseidon2/src/internal.rs
index 4b001f882..91f0b087a 100644
--- a/poseidon2/src/internal.rs
+++ b/poseidon2/src/internal.rs
@@ -2,7 +2,7 @@
 
 use ark_ff::PrimeField;
 
-use crate::add_rc_and_sbox;
+use crate::s_box;
 
 /// Matrix multiplication in the internal layers
 /// Given a vector v compute the matrix vector product (1 + diag(v))*state
@@ -55,6 +55,7 @@ pub(crate) fn permute_state<F: PrimeField, const T: usize>(
     d: usize,
     mat_diag_minus_1: &'static [F; T],
 ) {
-    add_rc_and_sbox(&mut state[0], rc, d);
+    state[0] += rc;
+    s_box(&mut state[0], d);
     matmul_internal(state, mat_diag_minus_1);
 }
diff --git a/poseidon2/src/lib.rs b/poseidon2/src/lib.rs
index 41bc77dda..8fe452675 100644
--- a/poseidon2/src/lib.rs
+++ b/poseidon2/src/lib.rs
@@ -122,14 +122,18 @@ impl<F: PrimeField> Poseidon2<F> {
     }
 }
 
-/// A generic method performing the transformation, used both in external and
-/// internal layers:
-///
-/// `s -> (s + rc)^d`
 // @credit: `add_rc_and_sbox_generic()` in plonky3
+/// add RCs to the entire state
+#[inline(always)]
+pub(crate) fn add_rcs<F: PrimeField, const T: usize>(state: &mut [F; T], rc: &[F; T]) {
+    for i in 0..T {
+        state[i] += rc[i];
+    }
+}
+
+/// `s -> s^d`
 #[inline(always)]
-pub(crate) fn add_rc_and_sbox<F: PrimeField>(val: &mut F, rc: F, d: usize) {
-    *val += rc;
+pub(crate) fn s_box<F: PrimeField>(val: &mut F, d: usize) {
     if d == 5 {
         // Perform unrolled computation for val^5, faster
         let original = *val;