stm32/hash: Improve performance when hashing

chunks_exact produces better code

2 files changed, 10 insertions, 5 deletions

diff --git a/embassy-stm32/CHANGELOG.md b/embassy-stm32/CHANGELOG.md
index 0f044419e..d2dccaea5 100644
--- a/embassy-stm32/CHANGELOG.md
+++ b/embassy-stm32/CHANGELOG.md
@@ -8,6 +8,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 <!-- next-header -->
 ## Unreleased - ReleaseDate
 
+- feat: Improve blocking hash speed
 - fix: Fix vrefbuf building with log feature
 - fix: Fix performing a hash after performing a hmac
 - chore: Updated stm32-metapac and stm32-data dependencies
diff --git a/embassy-stm32/src/hash/mod.rs b/embassy-stm32/src/hash/mod.rs
index a4c275242..90c06c0d8 100644
--- a/embassy-stm32/src/hash/mod.rs
+++ b/embassy-stm32/src/hash/mod.rs
@@ -353,13 +353,17 @@ impl<'d, T: Instance, M: Mode> Hash<'d, T, M> {
         let num_valid_bits: u8 = (8 * (input.len() % 4)) as u8;
         T::regs().str().modify(|w| w.set_nblw(num_valid_bits));
 
-        let mut i = 0;
-        while i < input.len() {
+        let mut chunks = input.chunks_exact(4);
+        for chunk in &mut chunks {
+            T::regs()
+                .din()
+                .write_value(u32::from_ne_bytes(chunk.try_into().unwrap()));
+        }
+        let rem = chunks.remainder();
+        if !rem.is_empty() {
             let mut word: [u8; 4] = [0; 4];
-            let copy_idx = min(i + 4, input.len());
-            word[0..copy_idx - i].copy_from_slice(&input[i..copy_idx]);
+            word[0..rem.len()].copy_from_slice(rem);
             T::regs().din().write_value(u32::from_ne_bytes(word));
-            i += 4;
         }
     }