1 files changed, 198 insertions, 0 deletions
diff --git a/thirdparty/BLAKE3/src/portable.rs b/thirdparty/BLAKE3/src/portable.rs
new file mode 100644
index 000000000..0a569cec7
--- /dev/null
+++ b/thirdparty/BLAKE3/src/portable.rs
@@ -0,0 +1,198 @@
+use crate::{
+    counter_high, counter_low, CVBytes, CVWords, IncrementCounter, BLOCK_LEN, IV, MSG_SCHEDULE,
+    OUT_LEN,
+};
+use arrayref::{array_mut_ref, array_ref};
+
+#[inline(always)]
+fn g(state: &mut [u32; 16], a: usize, b: usize, c: usize, d: usize, x: u32, y: u32) {
+    state[a] = state[a].wrapping_add(state[b]).wrapping_add(x);
+    state[d] = (state[d] ^ state[a]).rotate_right(16);
+    state[c] = state[c].wrapping_add(state[d]);
+    state[b] = (state[b] ^ state[c]).rotate_right(12);
+    state[a] = state[a].wrapping_add(state[b]).wrapping_add(y);
+    state[d] = (state[d] ^ state[a]).rotate_right(8);
+    state[c] = state[c].wrapping_add(state[d]);
+    state[b] = (state[b] ^ state[c]).rotate_right(7);
+}
+
+#[inline(always)]
+fn round(state: &mut [u32; 16], msg: &[u32; 16], round: usize) {
+    // Select the message schedule based on the round.
+    let schedule = MSG_SCHEDULE[round];
+
+    // Mix the columns.
+    g(state, 0, 4, 8, 12, msg[schedule[0]], msg[schedule[1]]);
+    g(state, 1, 5, 9, 13, msg[schedule[2]], msg[schedule[3]]);
+    g(state, 2, 6, 10, 14, msg[schedule[4]], msg[schedule[5]]);
+    g(state, 3, 7, 11, 15, msg[schedule[6]], msg[schedule[7]]);
+
+    // Mix the diagonals.
+    g(state, 0, 5, 10, 15, msg[schedule[8]], msg[schedule[9]]);
+    g(state, 1, 6, 11, 12, msg[schedule[10]], msg[schedule[11]]);
+    g(state, 2, 7, 8, 13, msg[schedule[12]], msg[schedule[13]]);
+    g(state, 3, 4, 9, 14, msg[schedule[14]], msg[schedule[15]]);
+}
+
+#[inline(always)]
+fn compress_pre(
+    cv: &CVWords,
+    block: &[u8; BLOCK_LEN],
+    block_len: u8,
+    counter: u64,
+    flags: u8,
+) -> [u32; 16] {
+    let block_words = crate::platform::words_from_le_bytes_64(block);
+
+    let mut state = [
+        cv[0],
+        cv[1],
+        cv[2],
+        cv[3],
+        cv[4],
+        cv[5],
+        cv[6],
+        cv[7],
+        IV[0],
+        IV[1],
+        IV[2],
+        IV[3],
+        counter_low(counter),
+        counter_high(counter),
+        block_len as u32,
+        flags as u32,
+    ];
+
+    round(&mut state, &block_words, 0);
+    round(&mut state, &block_words, 1);
+    round(&mut state, &block_words, 2);
+    round(&mut state, &block_words, 3);
+    round(&mut state, &block_words, 4);
+    round(&mut state, &block_words, 5);
+    round(&mut state, &block_words, 6);
+
+    state
+}
+
+pub fn compress_in_place(
+    cv: &mut CVWords,
+    block: &[u8; BLOCK_LEN],
+    block_len: u8,
+    counter: u64,
+    flags: u8,
+) {
+    let state = compress_pre(cv, block, block_len, counter, flags);
+
+    cv[0] = state[0] ^ state[8];
+    cv[1] = state[1] ^ state[9];
+    cv[2] = state[2] ^ state[10];
+    cv[3] = state[3] ^ state[11];
+    cv[4] = state[4] ^ state[12];
+    cv[5] = state[5] ^ state[13];
+    cv[6] = state[6] ^ state[14];
+    cv[7] = state[7] ^ state[15];
+}
+
+pub fn compress_xof(
+    cv: &CVWords,
+    block: &[u8; BLOCK_LEN],
+    block_len: u8,
+    counter: u64,
+    flags: u8,
+) -> [u8; 64] {
+    let mut state = compress_pre(cv, block, block_len, counter, flags);
+    state[0] ^= state[8];
+    state[1] ^= state[9];
+    state[2] ^= state[10];
+    state[3] ^= state[11];
+    state[4] ^= state[12];
+    state[5] ^= state[13];
+    state[6] ^= state[14];
+    state[7] ^= state[15];
+    state[8] ^= cv[0];
+    state[9] ^= cv[1];
+    state[10] ^= cv[2];
+    state[11] ^= cv[3];
+    state[12] ^= cv[4];
+    state[13] ^= cv[5];
+    state[14] ^= cv[6];
+    state[15] ^= cv[7];
+    crate::platform::le_bytes_from_words_64(&state)
+}
+
+pub fn hash1<A: arrayvec::Array<Item = u8>>(
+    input: &A,
+    key: &CVWords,
+    counter: u64,
+    flags: u8,
+    flags_start: u8,
+    flags_end: u8,
+    out: &mut CVBytes,
+) {
+    debug_assert_eq!(A::CAPACITY % BLOCK_LEN, 0, "uneven blocks");
+    let mut cv = *key;
+    let mut block_flags = flags | flags_start;
+    let mut slice = input.as_slice();
+    while slice.len() >= BLOCK_LEN {
+        if slice.len() == BLOCK_LEN {
+            block_flags |= flags_end;
+        }
+        compress_in_place(
+            &mut cv,
+            array_ref!(slice, 0, BLOCK_LEN),
+            BLOCK_LEN as u8,
+            counter,
+            block_flags,
+        );
+        block_flags = flags;
+        slice = &slice[BLOCK_LEN..];
+    }
+    *out = crate::platform::le_bytes_from_words_32(&cv);
+}
+
+pub fn hash_many<A: arrayvec::Array<Item = u8>>(
+    inputs: &[&A],
+    key: &CVWords,
+    mut counter: u64,
+    increment_counter: IncrementCounter,
+    flags: u8,
+    flags_start: u8,
+    flags_end: u8,
+    out: &mut [u8],
+) {
+    debug_assert!(out.len() >= inputs.len() * OUT_LEN, "out too short");
+    for (&input, output) in inputs.iter().zip(out.chunks_exact_mut(OUT_LEN)) {
+        hash1(
+            input,
+            key,
+            counter,
+            flags,
+            flags_start,
+            flags_end,
+            array_mut_ref!(output, 0, OUT_LEN),
+        );
+        if increment_counter.yes() {
+            counter += 1;
+        }
+    }
+}
+
+#[cfg(test)]
+pub mod test {
+    use super::*;
+
+    // This is basically testing the portable implementation against itself,
+    // but it also checks that compress_in_place and compress_xof are
+    // consistent. And there are tests against the reference implementation and
+    // against hardcoded test vectors elsewhere.
+    #[test]
+    fn test_compress() {
+        crate::test::test_compress_fn(compress_in_place, compress_xof);
+    }
+
+    // Ditto.
+    #[test]
+    fn test_hash_many() {
+        crate::test::test_hash_many_fn(hash_many, hash_many);
+    }
+}