sha2: make compress consume blocks

sha2/src/core_api.rs

@@ -1,7 +1,6 @@
 use crate::{consts, sha256::compress256, sha512::compress512};
 use core::{convert::TryInto, fmt, slice::from_ref};
 use digest::{
-    array::Array,
     block_buffer::Eager,
     core_api::{
         AlgorithmName, Block, BlockSizeUser, Buffer, BufferKindUser, OutputSizeUser, TruncSide,
@@ -39,7 +38,6 @@ impl UpdateCore for Sha256VarCore {
     #[inline]
     fn update_blocks(&mut self, blocks: &[Block<Self>]) {
         self.block_len += blocks.len() as u64;
-        let blocks = Array::cast_slice_to_core(blocks);
         compress256(&mut self.state, blocks);
     }
 }
@@ -66,7 +64,7 @@ impl VariableOutputCore for Sha256VarCore {
     fn finalize_variable_core(&mut self, buffer: &mut Buffer<Self>, out: &mut Output<Self>) {
         let bs = Self::BlockSize::U64;
         let bit_len = 8 * (buffer.get_pos() as u64 + bs * self.block_len);
-        buffer.len64_padding_be(bit_len, |b| compress256(&mut self.state, from_ref(&b.0)));
+        buffer.len64_padding_be(bit_len, |b| compress256(&mut self.state, from_ref(b)));
 
         for (chunk, v) in out.chunks_exact_mut(4).zip(self.state.iter()) {
             chunk.copy_from_slice(&v.to_be_bytes());
@@ -155,7 +153,6 @@ impl UpdateCore for Sha512VarCore {
     #[inline]
     fn update_blocks(&mut self, blocks: &[Block<Self>]) {
         self.block_len += blocks.len() as u128;
-        let blocks = Array::cast_slice_to_core(blocks);
         compress512(&mut self.state, blocks);
     }
 }
@@ -184,7 +181,7 @@ impl VariableOutputCore for Sha512VarCore {
     fn finalize_variable_core(&mut self, buffer: &mut Buffer<Self>, out: &mut Output<Self>) {
         let bs = Self::BlockSize::U64 as u128;
         let bit_len = 8 * (buffer.get_pos() as u128 + bs * self.block_len);
-        buffer.len128_padding_be(bit_len, |b| compress512(&mut self.state, from_ref(&b.0)));
+        buffer.len128_padding_be(bit_len, |b| compress512(&mut self.state, from_ref(b)));
 
         for (chunk, v) in out.chunks_exact_mut(8).zip(self.state.iter()) {
             chunk.copy_from_slice(&v.to_be_bytes());

sha2/src/sha256.rs

@@ -1,3 +1,7 @@
+use crate::Sha256VarCore;
+
+type Block = digest::core_api::Block<Sha256VarCore>;
+
 cfg_if::cfg_if! {
     if #[cfg(feature = "force-soft")] {
         mod soft;
@@ -24,6 +28,6 @@ cfg_if::cfg_if! {
 /// This is a low-level "hazmat" API which provides direct access to the core
 /// functionality of SHA-256.
 #[cfg_attr(docsrs, doc(cfg(feature = "compress")))]
-pub fn compress256(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
+pub fn compress256(state: &mut [u32; 8], blocks: &[Block]) {
     compress(state, blocks)
 }

sha2/src/sha256/aarch64.rs

@@ -6,11 +6,12 @@
 
 use core::arch::{aarch64::*, asm};
 
+use super::Block;
 use crate::consts::K32;
 
 cpufeatures::new!(sha2_hwcap, "sha2");
 
-pub fn compress(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
+pub fn compress(state: &mut [u32; 8], blocks: &[Block]) {
     // TODO: Replace with https://github.com/rust-lang/rfcs/pull/2725
     // after stabilization
     if sha2_hwcap::get() {
@@ -21,7 +22,7 @@ pub fn compress(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
 }
 
 #[target_feature(enable = "sha2")]
-unsafe fn sha256_compress(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
+unsafe fn sha256_compress(state: &mut [u32; 8], blocks: &[Block]) {
     // SAFETY: Requires the sha2 feature.
 
     // Load state into vectors.

sha2/src/sha256/loongarch64_asm.rs

@@ -1,5 +1,7 @@
 //! LoongArch64 assembly backend
 
+use super::Block;
+
 macro_rules! c {
     ($($l:expr)*) => {
         concat!($($l ,)*)
@@ -78,7 +80,7 @@ macro_rules! roundtail {
     };
 }
 
-pub fn compress(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
+pub fn compress(state: &mut [u32; 8], blocks: &[Block]) {
     if blocks.is_empty() {
         return;
     }

sha2/src/sha256/soft.rs

@@ -1,6 +1,8 @@
 #![allow(clippy::many_single_char_names)]
 use crate::consts::K32;
 
+use super::Block;
+
 #[inline(always)]
 fn shr(v: [u32; 4], o: u32) -> [u32; 4] {
     [v[0] >> o, v[1] >> o, v[2] >> o, v[3] >> o]
@@ -227,7 +229,7 @@ fn sha256_digest_block_u32(state: &mut [u32; 8], block: &[u32; 16]) {
     state[7] = state[7].wrapping_add(h);
 }
 
-pub fn compress(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
+pub fn compress(state: &mut [u32; 8], blocks: &[Block]) {
     for block in blocks {
         let mut block_u32 = [0u32; 16];
         for (o, chunk) in block_u32.iter_mut().zip(block.chunks_exact(4)) {

sha2/src/sha256/x86.rs

@@ -7,6 +7,8 @@ use core::arch::x86::*;
 #[cfg(target_arch = "x86_64")]
 use core::arch::x86_64::*;
 
+use super::Block;
+
 unsafe fn schedule(v0: __m128i, v1: __m128i, v2: __m128i, v3: __m128i) -> __m128i {
     let t1 = _mm_sha256msg1_epu32(v0, v1);
     let t2 = _mm_alignr_epi8(v3, v2, 4);
@@ -39,7 +41,7 @@ macro_rules! schedule_rounds4 {
 // we use unaligned loads with `__m128i` pointers
 #[allow(clippy::cast_ptr_alignment)]
 #[target_feature(enable = "sha,sse2,ssse3,sse4.1")]
-unsafe fn digest_blocks(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
+unsafe fn digest_blocks(state: &mut [u32; 8], blocks: &[Block]) {
     #[allow(non_snake_case)]
     let MASK: __m128i = _mm_set_epi64x(
         0x0C0D_0E0F_0809_0A0Bu64 as i64,
@@ -99,7 +101,7 @@ unsafe fn digest_blocks(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
 
 cpufeatures::new!(shani_cpuid, "sha", "sse2", "ssse3", "sse4.1");
 
-pub fn compress(state: &mut [u32; 8], blocks: &[[u8; 64]]) {
+pub fn compress(state: &mut [u32; 8], blocks: &[Block]) {
     // TODO: Replace with https://github.com/rust-lang/rfcs/pull/2725
     // after stabilization
     if shani_cpuid::get() {

sha2/src/sha512.rs

@@ -1,3 +1,7 @@
+use crate::Sha512VarCore;
+
+type Block = digest::core_api::Block<Sha512VarCore>;
+
 cfg_if::cfg_if! {
     if #[cfg(feature = "force-soft")] {
         mod soft;
@@ -24,6 +28,6 @@ cfg_if::cfg_if! {
 /// This is a low-level "hazmat" API which provides direct access to the core
 /// functionality of SHA-512.
 #[cfg_attr(docsrs, doc(cfg(feature = "compress")))]
-pub fn compress512(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
+pub fn compress512(state: &mut [u64; 8], blocks: &[Block]) {
     compress(state, blocks)
 }

sha2/src/sha512/aarch64.rs

@@ -2,11 +2,12 @@
 
 use core::arch::{aarch64::*, asm};
 
+use super::Block;
 use crate::consts::K64;
 
 cpufeatures::new!(sha3_hwcap, "sha3");
 
-pub fn compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
+pub fn compress(state: &mut [u64; 8], blocks: &[Block]) {
     // TODO: Replace with https://github.com/rust-lang/rfcs/pull/2725
     // after stabilization
     if sha3_hwcap::get() {
@@ -17,7 +18,7 @@ pub fn compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
 }
 
 #[target_feature(enable = "sha3")]
-unsafe fn sha512_compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
+unsafe fn sha512_compress(state: &mut [u64; 8], blocks: &[Block]) {
     // SAFETY: Requires the sha3 feature.
 
     // Load state into vectors.

sha2/src/sha512/loongarch64_asm.rs

@@ -1,5 +1,7 @@
 //! LoongArch64 assembly backend
 
+use super::Block;
+
 macro_rules! c {
     ($($l:expr)*) => {
         concat!($($l ,)*)
@@ -77,7 +79,7 @@ macro_rules! roundtail {
     };
 }
 
-pub fn compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
+pub fn compress(state: &mut [u64; 8], blocks: &[Block]) {
     if blocks.is_empty() {
         return;
     }

sha2/src/sha512/soft.rs

@@ -1,6 +1,8 @@
 #![allow(clippy::many_single_char_names)]
 use crate::consts::K64;
 
+use super::Block;
+
 /// Not an intrinsic, but works like an unaligned load.
 fn sha512load(v0: [u64; 2], v1: [u64; 2]) -> [u64; 2] {
     [v1[1], v0[0]]
@@ -208,7 +210,7 @@ pub fn sha512_digest_block_u64(state: &mut [u64; 8], block: &[u64; 16]) {
     state[7] = state[7].wrapping_add(h);
 }
 
-pub fn compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
+pub fn compress(state: &mut [u64; 8], blocks: &[Block]) {
     for block in blocks {
         let mut block_u32 = [0u64; 16];
         for (o, chunk) in block_u32.iter_mut().zip(block.chunks_exact(8)) {

sha2/src/sha512/x86.rs

@@ -9,11 +9,12 @@ use core::arch::x86::*;
 #[cfg(target_arch = "x86_64")]
 use core::arch::x86_64::*;
 
+use super::Block;
 use crate::consts::K64;
 
 cpufeatures::new!(avx2_cpuid, "avx2");
 
-pub fn compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
+pub fn compress(state: &mut [u64; 8], blocks: &[Block]) {
     // TODO: Replace with https://github.com/rust-lang/rfcs/pull/2725
     // after stabilization
     if avx2_cpuid::get() {
@@ -26,7 +27,7 @@ pub fn compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
 }
 
 #[target_feature(enable = "avx2")]
-unsafe fn sha512_compress_x86_64_avx2(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
+unsafe fn sha512_compress_x86_64_avx2(state: &mut [u64; 8], blocks: &[Block]) {
     let mut start_block = 0;
 
     if blocks.len() & 0b1 != 0 {
@@ -55,7 +56,7 @@ unsafe fn sha512_compress_x86_64_avx2(state: &mut [u64; 8], blocks: &[[u8; 128]]
 }
 
 #[inline(always)]
-unsafe fn sha512_compress_x86_64_avx(state: &mut [u64; 8], block: &[u8; 128]) {
+unsafe fn sha512_compress_x86_64_avx(state: &mut [u64; 8], block: &Block) {
     let mut ms = [_mm_setzero_si128(); 8];
     let mut x = [_mm_setzero_si128(); 8];