feat: optimized hash_elements for blake3 hasher

benches/README.md

@@ -28,7 +28,7 @@ The second scenario is that of sequential hashing where we take a sequence of le
 
 | Function            | BLAKE3 | SHA3    | Poseidon  | Rp64_256  | RPO_256 |
 | ------------------- | -------| ------- | --------- | --------- | ------- |
-| Apple M1 Pro        | 1.1 us | 1.5 us  |  19.4 us  |   118 us  | 70 us   |
+| Apple M1 Pro        | 1.0 us | 1.5 us  |  19.4 us  |   118 us  | 70 us   |
 | Apple M2            | 1.0 us | 1.5 us  |  17.4 us  |   103 us  | 65 us   |
 | Amazon Graviton 3   | 1.4 us |         |           |           | 114 us  |
 | AMD Ryzen 9 5950X   | 0.8 us | 1.7 us  |  15.7 us  |   120 us  | 72 us   |

src/hash/blake/mod.rs

@@ -1,7 +1,5 @@
 use super::{Digest, ElementHasher, Felt, FieldElement, Hasher, StarkField};
-use crate::utils::{
-    uninit_vector, ByteReader, ByteWriter, Deserializable, DeserializationError, Serializable,
-};
+use crate::utils::{ByteReader, ByteWriter, Deserializable, DeserializationError, Serializable};
 use core::{
     mem::{size_of, transmute, transmute_copy},
     ops::Deref,
@@ -290,15 +288,25 @@ where
     let digest = if Felt::IS_CANONICAL {
         blake3::hash(E::elements_as_bytes(elements))
     } else {
-        let base_elements = E::slice_as_base_elements(elements);
-        let blen = base_elements.len() << 3;
+        let mut hasher = blake3::Hasher::new();
+
+        // BLAKE3 state is 64 bytes - so, we can absorb 64 bytes into the state in a single
+        // permutation. we move the elements into the hasher via the buffer to give the CPU
+        // a chance to process multiple element-to-byte conversions in parallel
+        let mut buf = [0_u8; 64];
+        let mut chunk_iter = E::slice_as_base_elements(elements).chunks_exact(8);
+        for chunk in chunk_iter.by_ref() {
+            for i in 0..8 {
+                buf[i * 8..(i + 1) * 8].copy_from_slice(&chunk[i].as_int().to_le_bytes());
+            }
+            hasher.update(&buf);
+        }
 
-        let mut bytes = unsafe { uninit_vector(blen) };
-        for (idx, element) in base_elements.iter().enumerate() {
-            bytes[idx * 8..(idx + 1) * 8].copy_from_slice(&element.as_int().to_le_bytes());
+        for element in chunk_iter.remainder() {
+            hasher.update(&element.as_int().to_le_bytes());
         }
 
-        blake3::hash(&bytes)
+        hasher.finalize()
     };
     *shrink_bytes(&digest.into())
 }

src/hash/blake/tests.rs

@@ -1,6 +1,22 @@
 use super::*;
 use crate::utils::collections::Vec;
 use proptest::prelude::*;
+use rand_utils::rand_vector;
+
+#[test]
+fn blake3_hash_elements() {
+    // test multiple of 8
+    let elements = rand_vector::<Felt>(16);
+    let expected = compute_expected_element_hash(&elements);
+    let actual: [u8; 32] = hash_elements(&elements);
+    assert_eq!(&expected, &actual);
+
+    // test not multiple of 8
+    let elements = rand_vector::<Felt>(17);
+    let expected = compute_expected_element_hash(&elements);
+    let actual: [u8; 32] = hash_elements(&elements);
+    assert_eq!(&expected, &actual);
+}
 
 proptest! {
     #[test]
@@ -18,3 +34,14 @@ proptest! {
         Blake3_256::hash(vec);
     }
 }
+
+// HELPER FUNCTIONS
+// ================================================================================================
+
+fn compute_expected_element_hash(elements: &[Felt]) -> blake3::Hash {
+    let mut bytes = Vec::new();
+    for element in elements.iter() {
+        bytes.extend_from_slice(&element.as_int().to_le_bytes());
+    }
+    blake3::hash(&bytes)
+}