diff options
Diffstat (limited to 'src/crypto/slow-hash.c')
| -rw-r--r-- | src/crypto/slow-hash.c | 60 |
1 files changed, 26 insertions, 34 deletions
diff --git a/src/crypto/slow-hash.c b/src/crypto/slow-hash.c index 43b9619f3..6afa28934 100644 --- a/src/crypto/slow-hash.c +++ b/src/crypto/slow-hash.c @@ -1,4 +1,4 @@ -// Copyright (c) 2014-2016, The Monero Project +// Copyright (c) 2014-2017, The Monero Project // // All rights reserved. // @@ -44,6 +44,9 @@ #define INIT_SIZE_BLK 8 #define INIT_SIZE_BYTE (INIT_SIZE_BLK * AES_BLOCK_SIZE) +extern int aesb_single_round(const uint8_t *in, uint8_t*out, const uint8_t *expandedKey); +extern int aesb_pseudo_round(const uint8_t *in, uint8_t *out, const uint8_t *expandedKey); + #if defined(__x86_64__) || (defined(_MSC_VER) && defined(_WIN64)) // Optimised code below, uses x86-specific intrinsics, SSE2, AES-NI // Fall back to more portable code is down at the bottom @@ -138,9 +141,6 @@ #define THREADV __thread #endif -extern int aesb_single_round(const uint8_t *in, uint8_t*out, const uint8_t *expandedKey); -extern int aesb_pseudo_round(const uint8_t *in, uint8_t *out, const uint8_t *expandedKey); - #pragma pack(push, 1) union cn_slow_hash_state { @@ -494,7 +494,7 @@ void slow_hash_free_state(void) * buffer of pseudorandom data by hashing the supplied data. It then uses this * random data to fill a large 2MB buffer with pseudorandom data by iteratively * encrypting it using 10 rounds of AES per entry. After this initialization, - * it executes 500,000 rounds of mixing through the random 2MB buffer using + * it executes 524,288 rounds of mixing through the random 2MB buffer using * AES (typically provided in hardware on modern CPUs) and a 64 bit multiply. * Finally, it re-mixes this large buffer back into * the 200 byte "text" buffer, and then hashes this buffer using one of four @@ -530,7 +530,7 @@ void cn_slow_hash(const void *data, size_t length, char *hash) size_t i, j; uint64_t *p = NULL; - oaes_ctx *aes_ctx; + oaes_ctx *aes_ctx = NULL; int useAes = !force_software_aes() && check_aes_hw(); static void (*const extra_hashes[4])(const void *, size_t, char *) = @@ -578,8 +578,8 @@ void cn_slow_hash(const void *data, size_t length, char *hash) U64(b)[0] = U64(&state.k[16])[0] ^ U64(&state.k[48])[0]; U64(b)[1] = U64(&state.k[16])[1] ^ U64(&state.k[48])[1]; - /* CryptoNight Step 3: Bounce randomly 1 million times through the mixing buffer, - * using 500,000 iterations of the following mixing function. Each execution + /* CryptoNight Step 3: Bounce randomly 1,048,576 times (1<<20) through the mixing buffer, + * using 524,288 iterations of the following mixing function. Each execution * performs two reads and writes from the mixing buffer. */ @@ -722,32 +722,24 @@ union cn_slow_hash_state * key schedule. Don't try to use this for vanilla AES. */ static void aes_expand_key(const uint8_t *key, uint8_t *expandedKey) { -__asm__("mov x2, %1\n\t" : : "r"(key), "r"(expandedKey)); +static const int rcon[] = { + 0x01,0x01,0x01,0x01, + 0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d, // rotate-n-splat + 0x1b,0x1b,0x1b,0x1b }; __asm__( -" adr x3,Lrcon\n" -"\n" " eor v0.16b,v0.16b,v0.16b\n" -" ld1 {v3.16b},[x0],#16\n" -" ld1 {v1.4s,v2.4s},[x3],#32\n" -" b L256\n" -".align 5\n" -"Lrcon:\n" -".long 0x01,0x01,0x01,0x01\n" -".long 0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d,0x0c0f0e0d // rotate-n-splat\n" -".long 0x1b,0x1b,0x1b,0x1b\n" -"\n" -".align 4\n" -"L256:\n" -" ld1 {v4.16b},[x0]\n" -" mov w1,#5\n" -" st1 {v3.4s},[x2],#16\n" +" ld1 {v3.16b},[%0],#16\n" +" ld1 {v1.4s,v2.4s},[%2],#32\n" +" ld1 {v4.16b},[%0]\n" +" mov w2,#5\n" +" st1 {v3.4s},[%1],#16\n" "\n" -"Loop256:\n" +"1:\n" " tbl v6.16b,{v4.16b},v2.16b\n" " ext v5.16b,v0.16b,v3.16b,#12\n" -" st1 {v4.4s},[x2],#16\n" +" st1 {v4.4s},[%1],#16\n" " aese v6.16b,v0.16b\n" -" subs w1,w1,#1\n" +" subs w2,w2,#1\n" "\n" " eor v3.16b,v3.16b,v5.16b\n" " ext v5.16b,v0.16b,v5.16b,#12\n" @@ -757,8 +749,8 @@ __asm__( " eor v3.16b,v3.16b,v5.16b\n" " shl v1.16b,v1.16b,#1\n" " eor v3.16b,v3.16b,v6.16b\n" -" st1 {v3.4s},[x2],#16\n" -" b.eq Ldone\n" +" st1 {v3.4s},[%1],#16\n" +" b.eq 2f\n" "\n" " dup v6.4s,v3.s[3] // just splat\n" " ext v5.16b,v0.16b,v4.16b,#12\n" @@ -771,9 +763,9 @@ __asm__( " eor v4.16b,v4.16b,v5.16b\n" "\n" " eor v4.16b,v4.16b,v6.16b\n" -" b Loop256\n" +" b 1b\n" "\n" -"Ldone:\n"); +"2:\n" : : "r"(key), "r"(expandedKey), "r"(rcon)); } /* An ordinary AES round is a sequence of SubBytes, ShiftRows, MixColumns, AddRoundKey. There @@ -895,8 +887,8 @@ void cn_slow_hash(const void *data, size_t length, char *hash) U64(b)[0] = U64(&state.k[16])[0] ^ U64(&state.k[48])[0]; U64(b)[1] = U64(&state.k[16])[1] ^ U64(&state.k[48])[1]; - /* CryptoNight Step 3: Bounce randomly 1 million times through the mixing buffer, - * using 500,000 iterations of the following mixing function. Each execution + /* CryptoNight Step 3: Bounce randomly 1,048,576 times (1<<20) through the mixing buffer, + * using 524,288 iterations of the following mixing function. Each execution * performs two reads and writes from the mixing buffer. */ |