← Watch feed
hashes: Add optimized x86 SHA256d for 64-byte input
Public commit record
73/100 · Adequate
What the developer wrote
Authored by jrakibi
hashes: Add optimized x86 SHA256d for 64-byte input
Add the three transforms for sha256(sha256(64_bytes))
Transforms 1 & 2 compute the inner `sha256(64_bytes)`:
- Transform 1: 64 bytes already fill the first block (no
padding needed), so we apply 64 rounds of compression
normally.
After T1: state = state + initial_state
- Transform 2: the message schedule at this step is constant
and known in advance (64 bytes of padding), so
we precompute W[i]+K[i] into MIDS and apply 64 rounds (no
schedule expansion needed).
After T2: state = state + saved_T1_state
(state now contains sha256(64_bytes))
- Transform 3 computes sha256(output_of_T1_and_T2). The output
is 32 bytes, so the block contains 32 bytes and the rest is
known padding. we precompute the message
schedule for the known words (w8-w15) into FINS.
After T3: state = state + initial_state
(state now contains the sha256d result)
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context
What changed, and why it matters
This commit adds a new, highly optimized x86 implementation for computing a double SHA-256 hash (SHA256d) on exactly 64 bytes of input. It uses Intel SHA-NI CPU instructions and precomputed constants to speed up Bitcoin-style hashing. There is no indication in the commit that this fixes a security bug; it appears to be a pure performance optimization.
Recommended action
Treat as a routine performance optimization, not a security patch. If this function is later exposed, ensure it is only invoked after runtime CPU feature detection (e.g., `is_x86_feature_detected!("sha")`) and add unit tests comparing its output against the generic SHA256d implementation to guard against correctness bugs that could affect consensus-critical code.
Security signals we found
01
New unsafe code using x86 SHA-NI intrinsics
02
Manual SHA-256 round unrolling with precomputed constants
03
No input validation beyond fixed-size array references
04
No memory safety issues evident in diff (fixed-size stack/output buffers, no user-controlled lengths)
05
No mention of vulnerability fix or security relevance in commit message
Evidence from the diff
The patch introduces sha256d_64_x86, an unsafe Rust function gated by target_feature(enable = "sha,sse2,ssse3,sse4.1"). It manually unrolls the SHA-256 compression rounds for SHA256d on a 64-byte input using x86 SHA-NI intrinsics (_mm_sha256rnds2_epu32, _mm_sha256msg1_epu32, _mm_sha256msg2_epu32). Transform 1 hashes the 64-byte input block. Transform 2 hashes the fixed padding block (precomputed W[i]+K[i] in MIDS). Transform 3 hashes the 32-byte inner digest with fixed padding (precomputed in FINS/FINAL). The function is not yet wired into any public API in this commit; it is simply added alongside the existing process_block function.
Changed components
hashes/src/sha256/crypto/x86_shani.rsInspect captured patch +517 / −0
diff --git a/hashes/src/sha256/crypto/x86_shani.rs b/hashes/src/sha256/crypto/x86_shani.rs
index 87e0d250..05c6c29f 100644
--- a/hashes/src/sha256/crypto/x86_shani.rs
+++ b/hashes/src/sha256/crypto/x86_shani.rs
@@ -276,3 +276,520 @@ pub(super) unsafe fn process_block(state: &mut [u32; 8], block: &[u8]) {
_mm_storeu_si128(state.as_mut_ptr().add(4).cast::<__m128i>(), state1);
}
+/// Computes `SHA256d` of a single 64-byte input using x86 SHA-NI.
+#[target_feature(enable = "sha,sse2,ssse3,sse4.1")]
+unsafe fn sha256d_64_x86(output: &mut [u8; 32], input: &[u8; 64]) {
+ // SHA256 round constants
+ #[rustfmt::skip]
+ const K: [u32; 64] = [
+ 0x428A2F98, 0x71374491, 0xB5C0FBCF, 0xE9B5DBA5,
+ 0x3956C25B, 0x59F111F1, 0x923F82A4, 0xAB1C5ED5,
+ 0xD807AA98, 0x12835B01, 0x243185BE, 0x550C7DC3,
+ 0x72BE5D74, 0x80DEB1FE, 0x9BDC06A7, 0xC19BF174,
+ 0xE49B69C1, 0xEFBE4786, 0x0FC19DC6, 0x240CA1CC,
+ 0x2DE92C6F, 0x4A7484AA, 0x5CB0A9DC, 0x76F988DA,
+ 0x983E5152, 0xA831C66D, 0xB00327C8, 0xBF597FC7,
+ 0xC6E00BF3, 0xD5A79147, 0x06CA6351, 0x14292967,
+ 0x27B70A85, 0x2E1B2138, 0x4D2C6DFC, 0x53380D13,
+ 0x650A7354, 0x766A0ABB, 0x81C2C92E, 0x92722C85,
+ 0xA2BFE8A1, 0xA81A664B, 0xC24B8B70, 0xC76C51A3,
+ 0xD192E819, 0xD6990624, 0xF40E3585, 0x106AA070,
+ 0x19A4C116, 0x1E376C08, 0x2748774C, 0x34B0BCB5,
+ 0x391C0CB3, 0x4ED8AA4A, 0x5B9CCA4F, 0x682E6FF3,
+ 0x748F82EE, 0x78A5636F, 0x84C87814, 0x8CC70208,
+ 0x90BEFFFA, 0xA4506CEB, 0xBEF9A3F7, 0xC67178F2,
+ ];
+
+ // Precomputed W[i] + K[i] for the 2nd transform (padding block).
+ #[rustfmt::skip]
+ const MIDS: [u32; 64] = [
+ 0xc28a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
+ 0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
+ 0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
+ 0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf374,
+ 0x649b69c1, 0xf0fe4786, 0x0fe1edc6, 0x240cf254,
+ 0x4fe9346f, 0x6cc984be, 0x61b9411e, 0x16f988fa,
+ 0xf2c65152, 0xa88e5a6d, 0xb019fc65, 0xb9d99ec7,
+ 0x9a1231c3, 0xe70eeaa0, 0xfdb1232b, 0xc7353eb0,
+ 0x3069bad5, 0xcb976d5f, 0x5a0f118f, 0xdc1eeefd,
+ 0x0a35b689, 0xde0b7a04, 0x58f4ca9d, 0xe15d5b16,
+ 0x007f3e86, 0x37088980, 0xa507ea32, 0x6fab9537,
+ 0x17406110, 0x0d8cd6f1, 0xcdaa3b6d, 0xc0bbbe37,
+ 0x83613bda, 0xdb48a363, 0x0b02e931, 0x6fd15ca7,
+ 0x521afaca, 0x31338431, 0x6ed41a95, 0x6d437890,
+ 0xc39c91f2, 0x9eccabbd, 0xb5c9a0e6, 0x532fb63c,
+ 0xd2c741c6, 0x07237ea3, 0xa4954b68, 0x4c191d76
+ ];
+
+ // Precomputed values for Transform 3 rounds 9-16.
+ // FINS[0..3]: msg2 + K[8..11]
+ // FINS[4..7]: _mm_sha256msg1_epu32(msg2, msg3)
+ // FINS[8..11]: msg2 + K[12..15]
+ #[rustfmt::skip]
+ const FINS: [u32; 12] = [
+ 0x5807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
+ 0x80000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf274,
+ ];
+
+ // Padding processed in the 3rd transform (byteswapped).
+ const FINAL: [u32; 8] = [0x80000000, 0, 0, 0, 0, 0, 0, 0x100];
+
+ #[allow(non_snake_case)]
+ let MASK: __m128i = _mm_set_epi64x(
+ 0x0c0d_0e0f_0809_0a0bu64 as i64,
+ 0x0405_0607_0001_0203u64 as i64,
+ );
+
+ // Preshuffled SHA256 initial hash values for x86 SHA-NI.
+ let init0: __m128i = _mm_set_epi64x(0x6a09e667bb67ae85u64 as i64, 0x510e527f9b05688cu64 as i64);
+ let init1: __m128i = _mm_set_epi64x(0x3c6ef372a54ff53au64 as i64, 0x1f83d9ab5be0cd19u64 as i64);
+
+ let (mut state0, mut state1);
+ let (abef_save, cdgh_save);
+ let (mut msg, mut tmp);
+ let (mut msg0, mut msg1, mut msg2, mut msg3);
+
+ // ------------------ Transform 1 -------------------
+
+ // Load state
+ state0 = init0;
+ state1 = init1;
+
+ // Rounds 0-3
+ msg = _mm_loadu_si128(input.as_ptr().add(0).cast::<__m128i>());
+ msg0 = _mm_shuffle_epi8(msg, MASK);
+ msg = _mm_add_epi32(msg0, _mm_loadu_si128(K.as_ptr().add(0).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 4-7
+ msg = _mm_loadu_si128(input.as_ptr().add(16).cast::<__m128i>());
+ msg1 = _mm_shuffle_epi8(msg, MASK);
+ msg = _mm_add_epi32(msg1, _mm_loadu_si128(K.as_ptr().add(4).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg0 = _mm_sha256msg1_epu32(msg0, msg1);
+
+ // Rounds 8-11
+ msg = _mm_loadu_si128(input.as_ptr().add(32).cast::<__m128i>());
+ msg2 = _mm_shuffle_epi8(msg, MASK);
+ msg = _mm_add_epi32(msg2, _mm_loadu_si128(K.as_ptr().add(8).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg1 = _mm_sha256msg1_epu32(msg1, msg2);
+
+ // Rounds 12-15
+ msg = _mm_loadu_si128(input.as_ptr().add(48).cast::<__m128i>());
+ msg3 = _mm_shuffle_epi8(msg, MASK);
+ msg = _mm_add_epi32(msg3, _mm_loadu_si128(K.as_ptr().add(12).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg3, msg2, 4);
+ msg0 = _mm_add_epi32(msg0, tmp);
+ msg0 = _mm_sha256msg2_epu32(msg0, msg3);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg2 = _mm_sha256msg1_epu32(msg2, msg3);
+
+ // Rounds 16-19
+ msg = _mm_add_epi32(msg0, _mm_loadu_si128(K.as_ptr().add(16).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg0, msg3, 4);
+ msg1 = _mm_add_epi32(msg1, tmp);
+ msg1 = _mm_sha256msg2_epu32(msg1, msg0);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg3 = _mm_sha256msg1_epu32(msg3, msg0);
+
+ // Rounds 20-23
+ msg = _mm_add_epi32(msg1, _mm_loadu_si128(K.as_ptr().add(20).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg1, msg0, 4);
+ msg2 = _mm_add_epi32(msg2, tmp);
+ msg2 = _mm_sha256msg2_epu32(msg2, msg1);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg0 = _mm_sha256msg1_epu32(msg0, msg1);
+
+ // Rounds 24-27
+ msg = _mm_add_epi32(msg2, _mm_loadu_si128(K.as_ptr().add(24).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg2, msg1, 4);
+ msg3 = _mm_add_epi32(msg3, tmp);
+ msg3 = _mm_sha256msg2_epu32(msg3, msg2);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg1 = _mm_sha256msg1_epu32(msg1, msg2);
+
+ // Rounds 28-31
+ msg = _mm_add_epi32(msg3, _mm_loadu_si128(K.as_ptr().add(28).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg3, msg2, 4);
+ msg0 = _mm_add_epi32(msg0, tmp);
+ msg0 = _mm_sha256msg2_epu32(msg0, msg3);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg2 = _mm_sha256msg1_epu32(msg2, msg3);
+
+ // Rounds 32-35
+ msg = _mm_add_epi32(msg0, _mm_loadu_si128(K.as_ptr().add(32).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg0, msg3, 4);
+ msg1 = _mm_add_epi32(msg1, tmp);
+ msg1 = _mm_sha256msg2_epu32(msg1, msg0);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg3 = _mm_sha256msg1_epu32(msg3, msg0);
+
+ // Rounds 36-39
+ msg = _mm_add_epi32(msg1, _mm_loadu_si128(K.as_ptr().add(36).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg1, msg0, 4);
+ msg2 = _mm_add_epi32(msg2, tmp);
+ msg2 = _mm_sha256msg2_epu32(msg2, msg1);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg0 = _mm_sha256msg1_epu32(msg0, msg1);
+
+ // Rounds 40-43
+ msg = _mm_add_epi32(msg2, _mm_loadu_si128(K.as_ptr().add(40).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg2, msg1, 4);
+ msg3 = _mm_add_epi32(msg3, tmp);
+ msg3 = _mm_sha256msg2_epu32(msg3, msg2);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg1 = _mm_sha256msg1_epu32(msg1, msg2);
+
+ // Rounds 44-47
+ msg = _mm_add_epi32(msg3, _mm_loadu_si128(K.as_ptr().add(44).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg3, msg2, 4);
+ msg0 = _mm_add_epi32(msg0, tmp);
+ msg0 = _mm_sha256msg2_epu32(msg0, msg3);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg2 = _mm_sha256msg1_epu32(msg2, msg3);
+
+ // Rounds 48-51
+ msg = _mm_add_epi32(msg0, _mm_loadu_si128(K.as_ptr().add(48).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg0, msg3, 4);
+ msg1 = _mm_add_epi32(msg1, tmp);
+ msg1 = _mm_sha256msg2_epu32(msg1, msg0);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg3 = _mm_sha256msg1_epu32(msg3, msg0);
+
+ // Rounds 52-55
+ msg = _mm_add_epi32(msg1, _mm_loadu_si128(K.as_ptr().add(52).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg1, msg0, 4);
+ msg2 = _mm_add_epi32(msg2, tmp);
+ msg2 = _mm_sha256msg2_epu32(msg2, msg1);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 56-59
+ msg = _mm_add_epi32(msg2, _mm_loadu_si128(K.as_ptr().add(56).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg2, msg1, 4);
+ msg3 = _mm_add_epi32(msg3, tmp);
+ msg3 = _mm_sha256msg2_epu32(msg3, msg2);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 60-63
+ msg = _mm_add_epi32(msg3, _mm_loadu_si128(K.as_ptr().add(60).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Transform 1: Update state
+ state0 = _mm_add_epi32(state0, init0);
+ state1 = _mm_add_epi32(state1, init1);
+
+ // ------------------ Transform 2 -------------------
+
+ // Save state
+ abef_save = state0;
+ cdgh_save = state1;
+
+ // Rounds 0-3
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(0).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 4-7
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(4).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 8-11
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(8).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 12-15
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(12).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 16-19
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(16).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 20-23
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(20).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 24-27
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(24).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 28-31
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(28).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 32-35
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(32).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 36-39
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(36).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 40-43
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(40).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 44-47
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(44).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 48-51
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(48).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 52-55
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(52).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 56-59
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(56).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 60-63
+ msg = _mm_loadu_si128(MIDS.as_ptr().add(60).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Transform 2: Update state
+ state0 = _mm_add_epi32(state0, abef_save);
+ state1 = _mm_add_epi32(state1, cdgh_save);
+
+ // Unshuffle to extract hash
+ tmp = _mm_shuffle_epi32(state0, 0x1B);
+ state1 = _mm_shuffle_epi32(state1, 0xB1);
+ msg0 = _mm_blend_epi16(tmp, state1, 0xF0);
+ msg1 = _mm_alignr_epi8(state1, tmp, 8);
+
+ // ------------------ Transform 3 -------------------
+
+ // Load state
+ state0 = init0;
+ state1 = init1;
+
+ // Rounds 0-3
+ msg = _mm_add_epi32(msg0, _mm_loadu_si128(K.as_ptr().add(0).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 4-7
+ msg = _mm_add_epi32(msg1, _mm_loadu_si128(K.as_ptr().add(4).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg0 = _mm_sha256msg1_epu32(msg0, msg1);
+
+ // Rounds 8-11
+ msg2 = _mm_loadu_si128(FINS.as_ptr().add(4).cast::<__m128i>());
+ msg = _mm_loadu_si128(FINS.as_ptr().add(0).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg1 = _mm_sha256msg1_epu32(msg1, msg2);
+
+ // Rounds 12-15
+ msg3 = _mm_loadu_si128(FINAL.as_ptr().add(4).cast::<__m128i>());
+ msg = _mm_loadu_si128(FINS.as_ptr().add(8).cast::<__m128i>());
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg3, msg2, 4);
+ msg0 = _mm_add_epi32(msg0, tmp);
+ msg0 = _mm_sha256msg2_epu32(msg0, msg3);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg2 = _mm_sha256msg1_epu32(msg2, msg3);
+
+ // Rounds 16-19
+ msg = _mm_add_epi32(msg0, _mm_loadu_si128(K.as_ptr().add(16).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg0, msg3, 4);
+ msg1 = _mm_add_epi32(msg1, tmp);
+ msg1 = _mm_sha256msg2_epu32(msg1, msg0);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg3 = _mm_sha256msg1_epu32(msg3, msg0);
+
+ // Rounds 20-23
+ msg = _mm_add_epi32(msg1, _mm_loadu_si128(K.as_ptr().add(20).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg1, msg0, 4);
+ msg2 = _mm_add_epi32(msg2, tmp);
+ msg2 = _mm_sha256msg2_epu32(msg2, msg1);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg0 = _mm_sha256msg1_epu32(msg0, msg1);
+
+ // Rounds 24-27
+ msg = _mm_add_epi32(msg2, _mm_loadu_si128(K.as_ptr().add(24).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg2, msg1, 4);
+ msg3 = _mm_add_epi32(msg3, tmp);
+ msg3 = _mm_sha256msg2_epu32(msg3, msg2);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg1 = _mm_sha256msg1_epu32(msg1, msg2);
+
+ // Rounds 28-31
+ msg = _mm_add_epi32(msg3, _mm_loadu_si128(K.as_ptr().add(28).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg3, msg2, 4);
+ msg0 = _mm_add_epi32(msg0, tmp);
+ msg0 = _mm_sha256msg2_epu32(msg0, msg3);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg2 = _mm_sha256msg1_epu32(msg2, msg3);
+
+ // Rounds 32-35
+ msg = _mm_add_epi32(msg0, _mm_loadu_si128(K.as_ptr().add(32).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg0, msg3, 4);
+ msg1 = _mm_add_epi32(msg1, tmp);
+ msg1 = _mm_sha256msg2_epu32(msg1, msg0);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg3 = _mm_sha256msg1_epu32(msg3, msg0);
+
+ // Rounds 36-39
+ msg = _mm_add_epi32(msg1, _mm_loadu_si128(K.as_ptr().add(36).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg1, msg0, 4);
+ msg2 = _mm_add_epi32(msg2, tmp);
+ msg2 = _mm_sha256msg2_epu32(msg2, msg1);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg0 = _mm_sha256msg1_epu32(msg0, msg1);
+
+ // Rounds 40-43
+ msg = _mm_add_epi32(msg2, _mm_loadu_si128(K.as_ptr().add(40).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg2, msg1, 4);
+ msg3 = _mm_add_epi32(msg3, tmp);
+ msg3 = _mm_sha256msg2_epu32(msg3, msg2);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg1 = _mm_sha256msg1_epu32(msg1, msg2);
+
+ // Rounds 44-47
+ msg = _mm_add_epi32(msg3, _mm_loadu_si128(K.as_ptr().add(44).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg3, msg2, 4);
+ msg0 = _mm_add_epi32(msg0, tmp);
+ msg0 = _mm_sha256msg2_epu32(msg0, msg3);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg2 = _mm_sha256msg1_epu32(msg2, msg3);
+
+ // Rounds 48-51
+ msg = _mm_add_epi32(msg0, _mm_loadu_si128(K.as_ptr().add(48).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg0, msg3, 4);
+ msg1 = _mm_add_epi32(msg1, tmp);
+ msg1 = _mm_sha256msg2_epu32(msg1, msg0);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+ msg3 = _mm_sha256msg1_epu32(msg3, msg0);
+
+ // Rounds 52-55
+ msg = _mm_add_epi32(msg1, _mm_loadu_si128(K.as_ptr().add(52).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg1, msg0, 4);
+ msg2 = _mm_add_epi32(msg2, tmp);
+ msg2 = _mm_sha256msg2_epu32(msg2, msg1);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 56-59
+ msg = _mm_add_epi32(msg2, _mm_loadu_si128(K.as_ptr().add(56).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ tmp = _mm_alignr_epi8(msg2, msg1, 4);
+ msg3 = _mm_add_epi32(msg3, tmp);
+ msg3 = _mm_sha256msg2_epu32(msg3, msg2);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+ // Rounds 60-63
+ msg = _mm_add_epi32(msg3, _mm_loadu_si128(K.as_ptr().add(60).cast::<__m128i>()));
+ state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
+ msg = _mm_shuffle_epi32(msg, 0x0E);
+ state0 = _mm_sha256rnds2_epu32(state0, state1, msg);
+
+
+ // Transform 3: Update state
+ state0 = _mm_add_epi32(state0, init0);
+ state1 = _mm_add_epi32(state1, init1);
+
+ // Unshuffle
+ tmp = _mm_shuffle_epi32(state0, 0x1B);
+ state1 = _mm_shuffle_epi32(state1, 0xB1);
+ state0 = _mm_blend_epi16(tmp, state1, 0xF0);
+ state1 = _mm_alignr_epi8(state1, tmp, 8);
+
+ // Store result (byte-swap to big-endian)
+ // CAST SAFETY: storeu_si128 does not require alignment.
+ _mm_storeu_si128(output.as_mut_ptr().add(0).cast::<__m128i>(), _mm_shuffle_epi8(state0, MASK));
+ _mm_storeu_si128(output.as_mut_ptr().add(16).cast::<__m128i>(), _mm_shuffle_epi8(state1, MASK));
+}
+
Risk score
Our methodology →Why this scored 16/100
Human-validated context
Community notes
Notes can correct, qualify, or add evidence to the AI analysis. Every note shown here has been validated by a human moderator.
No validated notes yet.
The AI analysis stands alone for now. Submit a note if you can add evidence or important context.