feat(sha256): add sha256d80 backend API and ARM64 kernel entry

sha256/sha256_backend.h

@@ -0,0 +1,65 @@
+#ifndef SHA256_BACKEND_H
+#define SHA256_BACKEND_H
+
+#include <stdint.h>
+
+/*
+ * Compact SHA256 state: only the 8 chaining words (32 bytes).
+ * Avoids copying the bloated OpenSSL SHA256_CTX (~112 bytes) in the hot loop.
+ */
+typedef struct {
+    uint32_t h[8];
+} sha256_state_t;
+
+typedef struct {
+    sha256_state_t init_state;
+    sha256_state_t first_chunk_state;
+    uint8_t block1_template[64];
+    uint8_t block2_template[64];
+} sha256d80_midstate_t;
+
+/* Set state to SHA256 initial values (IV). */
+void sha256_state_init(sha256_state_t *state);
+
+/* Serialize the 8 state words into a 32-byte big-endian digest. */
+void sha256_state_to_digest(const sha256_state_t *state, uint8_t out[32]);
+
+/* Single SHA256 block compression (64-byte block). */
+void sha256_transform_fast(sha256_state_t *state, const uint8_t block[64]);
+
+/*
+ * 2-way interleaved SHA256 block compression.
+ * Processes two independent (state, block) pairs so the CPU can overlap both
+ * instruction chains. On non-ARM builds falls back to two sequential calls.
+ */
+void sha256_transform_fast_2way(
+    sha256_state_t *stA, const uint8_t blkA[64],
+    sha256_state_t *stB, const uint8_t blkB[64]
+);
+
+/* Prepare SHA256d(80-byte header) midstate and constant blocks from header[0..75]. */
+void sha256d80_midstate_init(sha256d80_midstate_t *mid, const uint8_t header_76[76]);
+
+/*
+ * Hash 4 consecutive nonces with SHA256d(header80).
+ * start_nonce lane order: [n, n+1, n+2, n+3].
+ */
+void sha256d80_hash_4way(
+    const sha256d80_midstate_t *mid,
+    uint32_t start_nonce,
+    sha256_state_t out_states[4]
+);
+
+/*
+ * Hash 4 consecutive nonces and return hit mask against target words.
+ * target_words are big-endian words target[0..7].
+ * bit i set => lane i meets target.
+ */
+uint32_t sha256d80_scan_4way(
+    const sha256d80_midstate_t *mid,
+    uint32_t start_nonce,
+    const uint32_t target_words[8],
+    sha256_state_t out_states[4]
+);
+
+#endif