Merge pull request #1560 from evoskuil/master

Single block shani.

Author: evoskuil

include/bitcoin/system/hash/sha/algorithm.hpp

@@ -166,6 +166,9 @@ class algorithm
     template <typename xWord, if_extended<xWord> = true>
     using wstate_t = std_array<xWord, sizeof(state_t) / sizeof(xWord)>;
 
+    ////template <typename xWord, if_extended<xWord> = true>
+    ////using wblock_t = std_array<xWord, sizeof(block_t) / sizeof(xWord)>;
+
     /// Other types.
     /// -----------------------------------------------------------------------
 
@@ -236,17 +239,23 @@ class algorithm
     /// Padding.
     /// -----------------------------------------------------------------------
 
+    /// Scheduled padding (new and existing buffer objects).
     template <size_t Blocks>
     static CONSTEVAL buffer_t scheduled_pad() NOEXCEPT;
-    static CONSTEVAL chunk_t chunk_pad() NOEXCEPT;
-    static CONSTEVAL pad_t stream_pad() NOEXCEPT;
-
     template <size_t Blocks>
     static constexpr void schedule_n(buffer_t& buffer) NOEXCEPT;
     static constexpr void schedule_n(buffer_t& buffer, size_t blocks) NOEXCEPT;
     static constexpr void schedule_1(buffer_t& buffer) NOEXCEPT;
-    static constexpr void pad_half(buffer_t& buffer) NOEXCEPT;
-    static constexpr void pad_n(buffer_t& buffer, count_t blocks) NOEXCEPT;
+
+    /// Unscheduled padding (new objects).
+    static words_t pad_block() NOEXCEPT;
+    static words_t pad_blocks(count_t blocks) NOEXCEPT;
+    static CONSTEVAL chunk_t chunk_pad() NOEXCEPT;
+    static CONSTEVAL pad_t stream_pad() NOEXCEPT;
+
+    /// Unscheduled padding (update block or buffer object).
+    static constexpr void pad_half(auto& buffer) NOEXCEPT;
+    static constexpr void pad_n(auto& buffer, count_t blocks) NOEXCEPT;
 
     /// Double hashing.
     /// -----------------------------------------------------------------------
@@ -363,6 +372,8 @@ class algorithm
     /// Native SHA optimizations (single blocks).
     /// -----------------------------------------------------------------------
 
+    template <bool Swap>
+    INLINE static xint128_t bytes(xint128_t message) NOEXCEPT;
     INLINE static void shuffle(xint128_t& state0, xint128_t& state1) NOEXCEPT;
     INLINE static void unshuffle(xint128_t& state0, xint128_t& state1) NOEXCEPT;
     INLINE static void prepare(xint128_t& message0, xint128_t message1) NOEXCEPT;
@@ -373,7 +384,14 @@ class algorithm
     INLINE static void round_4(xint128_t& state0, xint128_t& state1,
         xint128_t message) NOEXCEPT;
 
-    static void native_rounds(state_t& state, iblocks_t& blocks) NOEXCEPT;
+    template <bool Swap>
+    INLINE static void native_rounds(xint128_t& lo, xint128_t& hi,
+        const block_t& block) NOEXCEPT;
+
+    static void native_(state_t& state, iblocks_t& blocks) NOEXCEPT;
+    static void native_(state_t& state, const block_t& block) NOEXCEPT;
+    INLINE static void native_preswapped(state_t& state,
+        const words_t& block) NOEXCEPT;
 
 public:
     /// Summary public values.

include/bitcoin/system/impl/hash/sha/algorithm_double.ipp

@@ -65,9 +65,10 @@ double_hash(const ablocks_t<Size>& blocks) NOEXCEPT
 {
     static_assert(is_same_type<state_t, chunk_t>);
 
-    buffer_t buffer{};
     auto state = H::get;
     iterate(state, blocks);
+
+    buffer_t buffer{};
     schedule_n<Size>(buffer);
     compress(state, buffer);
 
@@ -77,6 +78,7 @@ double_hash(const ablocks_t<Size>& blocks) NOEXCEPT
     schedule(buffer);
     state = H::get;
     compress(state, buffer);
+
     return output(state);
 }
 
@@ -89,9 +91,10 @@ double_hash(iblocks_t&& blocks) NOEXCEPT
     // Save block count, as iterable decrements.
     const auto count = blocks.size();
 
-    buffer_t buffer{};
     auto state = H::get;
     iterate(state, blocks);
+
+    buffer_t buffer{};
     schedule_n(buffer, count);
     compress(state, buffer);
 
@@ -101,6 +104,7 @@ double_hash(iblocks_t&& blocks) NOEXCEPT
     schedule(buffer);
     state = H::get;
     compress(state, buffer);
+
     return output(state);
 }
 
@@ -110,9 +114,9 @@ double_hash(const block_t& block) NOEXCEPT
 {
     static_assert(is_same_type<state_t, chunk_t>);
 
-    buffer_t buffer{};
-
     auto state = H::get;
+
+    buffer_t buffer{};
     input(buffer, block);
     schedule(buffer);
     compress(state, buffer);
@@ -125,6 +129,7 @@ double_hash(const block_t& block) NOEXCEPT
     schedule(buffer);
     state = H::get;
     compress(state, buffer);
+
     return output(state);
 }
 
@@ -134,8 +139,9 @@ double_hash(const half_t& half) NOEXCEPT
 {
     static_assert(is_same_type<state_t, chunk_t>);
 
-    buffer_t buffer{};
     auto state = H::get;
+
+    buffer_t buffer{};
     input_left(buffer, half);
     pad_half(buffer);
     schedule(buffer);
@@ -147,6 +153,7 @@ double_hash(const half_t& half) NOEXCEPT
     schedule(buffer);
     state = H::get;
     compress(state, buffer);
+
     return output(state);
 }
 
@@ -156,8 +163,9 @@ double_hash(const half_t& left, const half_t& right) NOEXCEPT
 {
     static_assert(is_same_type<state_t, chunk_t>);
 
-    buffer_t buffer{};
     auto state = H::get;
+
+    buffer_t buffer{};
     input_left(buffer, left);
     input_right(buffer, right);
     schedule(buffer);
@@ -171,6 +179,7 @@ double_hash(const half_t& left, const half_t& right) NOEXCEPT
     schedule(buffer);
     state = H::get;
     compress(state, buffer);
+
     return output(state);
 }
 

include/bitcoin/system/impl/hash/sha/algorithm_iterate.ipp

@@ -243,19 +243,19 @@ iterate_vector(state_t& state, const ablocks_t<Size>& blocks) NOEXCEPT
 // intel-sha-extensions-white-paper-402097.pdf
 
 TEMPLATE
+template <size_t Size>
 INLINE void CLASS::
-iterate_native(state_t& state, iblocks_t& blocks) NOEXCEPT
+iterate_native(state_t& state, const ablocks_t<Size>& blocks) NOEXCEPT
 {
-    native_rounds(state, blocks);
+    iblocks_t iblocks{ array_cast<byte_t>(blocks) };
+    native_(state, iblocks);
 }
 
 TEMPLATE
-template <size_t Size>
 INLINE void CLASS::
-iterate_native(state_t& state, const ablocks_t<Size>& blocks) NOEXCEPT
+iterate_native(state_t& state, iblocks_t& blocks) NOEXCEPT
 {
-    iblocks_t iblocks{ array_cast<byte_t>(blocks) };
-    native_rounds(state, iblocks);
+    native_(state, blocks);
 }
 
 // Dispatch and normal forms.