nihilist/monero
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Update slow-hash.c

Browse Source 
1. Added AES-NI support for modern processors.
This commit is contained in:
NoodleDoodleNoodleDoodleNoodleDoodleNoo 2014-05-21 09:29:34 -07:00
parent 3b887decca
commit 44f61c3965
1 changed files with 234 additions and 128 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

264
src/crypto/slow-hash.c
View File

@ -11,52 +11,93 @@
#include "hash-ops.h" #include "hash-ops.h"
#include "oaes_lib.h" #include "oaes_lib.h"
static void (*const extra_hashes[4])(const void *, size_t, char *) = { #include <emmintrin.h>
hash_extra_blake, hash_extra_groestl, hash_extra_jh, hash_extra_skein
};
#define MEMORY (1 << 21) /* 2 MiB */ #if defined(_MSC_VER) || defined(__INTEL_COMPILER)
#include <intrin.h>
#define STATIC
#define INLINE __inline
#if !defined(RDATA_ALIGN16)
#define RDATA_ALIGN16 __declspec(align(16))
#endif
#else
#include <wmmintrin.h>
#define STATIC static
#define INLINE inline
#if !defined(RDATA_ALIGN16)
#define RDATA_ALIGN16 __attribute__ ((aligned(16)))
#endif
#endif
#define MEMORY (1 << 21) // 2MB scratchpad
#define ITER (1 << 20) #define ITER (1 << 20)
#define AES_BLOCK_SIZE 16 #define AES_BLOCK_SIZE 16
#define AES_KEY_SIZE 32 /*16*/ #define AES_KEY_SIZE 32
#define INIT_SIZE_BLK 8 #define INIT_SIZE_BLK 8
#define INIT_SIZE_BYTE (INIT_SIZE_BLK * AES_BLOCK_SIZE) #define INIT_SIZE_BYTE (INIT_SIZE_BLK * AES_BLOCK_SIZE)
#define U64(x) ((uint64_t *) (x))
#define R128(x) ((__m128i *) (x))
extern int aesb_single_round(const uint8_t *in, uint8_t*out, const uint8_t *expandedKey); 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); extern int aesb_pseudo_round(const uint8_t *in, uint8_t *out, const uint8_t *expandedKey);
static size_t e2i(const uint8_t* a, size_t count) { return (*((uint64_t*)a) / AES_BLOCK_SIZE) & (count - 1); } #pragma pack(push, 1)
union cn_slow_hash_state
{
union hash_state hs;
struct
{
uint8_t k[64];
uint8_t init[INIT_SIZE_BYTE];
};
};
#pragma pack(pop)
static void mul(const uint8_t* a, const uint8_t* b, uint8_t* res) { #if defined(_MSC_VER) || defined(__INTEL_COMPILER)
#define cpuid(info,x) __cpuidex(info,x,0)
#else
void cpuid(int CPUInfo[4], int InfoType)
{
__asm__ __volatile__
(
"cpuid":
"=a" (CPUInfo[0]),
"=b" (CPUInfo[1]),
"=c" (CPUInfo[2]),
"=d" (CPUInfo[3]) :
"a" (InfoType), "c" (0)
);
}
#endif
STATIC INLINE void mul(const uint8_t *a, const uint8_t *b, uint8_t *res)
{
uint64_t a0, b0; uint64_t a0, b0;
uint64_t hi, lo; uint64_t hi, lo;
a0 = SWAP64LE(((uint64_t*)a)[0]); a0 = U64(a)[0];
b0 = SWAP64LE(((uint64_t*)b)[0]); b0 = U64(b)[0];
lo = mul128(a0, b0, &hi); lo = mul128(a0, b0, &hi);
((uint64_t*)res)[0] = SWAP64LE(hi); U64(res)[0] = hi;
((uint64_t*)res)[1] = SWAP64LE(lo); U64(res)[1] = lo;
} }
static void sum_half_blocks(uint8_t* a, const uint8_t* b) { STATIC INLINE void sum_half_blocks(uint8_t *a, const uint8_t *b)
{
uint64_t a0, a1, b0, b1; uint64_t a0, a1, b0, b1;
a0 = U64(a)[0];
a0 = SWAP64LE(((uint64_t*)a)[0]); a1 = U64(a)[1];
a1 = SWAP64LE(((uint64_t*)a)[1]); b0 = U64(b)[0];
b0 = SWAP64LE(((uint64_t*)b)[0]); b1 = U64(b)[1];
b1 = SWAP64LE(((uint64_t*)b)[1]);
a0 += b0; a0 += b0;
a1 += b1; a1 += b1;
((uint64_t*)a)[0] = SWAP64LE(a0); U64(a)[0] = a0;
((uint64_t*)a)[1] = SWAP64LE(a1); U64(a)[1] = a1;
}
#define U64(x) ((uint64_t *) (x))
static void copy_block(uint8_t* dst, const uint8_t* src) {
memcpy(dst, src, AES_BLOCK_SIZE);
} }
static void swap_blocks(uint8_t *a, uint8_t *b){ STATIC INLINE void swap_blocks(uint8_t *a, uint8_t *b)
{
uint64_t t[2]; uint64_t t[2];
U64(t)[0] = U64(a)[0]; U64(t)[0] = U64(a)[0];
U64(t)[1] = U64(a)[1]; U64(t)[1] = U64(a)[1];
@ -66,90 +107,155 @@ static void swap_blocks(uint8_t *a, uint8_t *b){
U64(b)[1] = U64(t)[1]; U64(b)[1] = U64(t)[1];
} }
static void xor_blocks(uint8_t* a, const uint8_t* b) { STATIC INLINE void xor_blocks(uint8_t *a, const uint8_t *b)
size_t i; {
for (i = 0; i < AES_BLOCK_SIZE; i++) { U64(a)[0] ^= U64(b)[0];
a[i] ^= b[i]; U64(a)[1] ^= U64(b)[1];
}
} }
#pragma pack(push, 1) STATIC INLINE int check_aes_hw(void)
union cn_slow_hash_state { {
union hash_state hs; int cpuid_results[4];
struct { static int supported = -1;
uint8_t k[64];
uint8_t init[INIT_SIZE_BYTE];
};
};
#pragma pack(pop)
void cn_slow_hash(const void *data, size_t length, char *hash) { if(supported >= 0)
return supported;
cpuid(cpuid_results,1);
return supported = cpuid_results[2] & (1 << 25);
}
STATIC INLINE void aesni_pseudo_round(const uint8_t *in, uint8_t *out,
const uint8_t *expandedKey)
{
__m128i *k = R128(expandedKey);
__m128i d;
d = _mm_loadu_si128(R128(in));
d = _mm_aesenc_si128(d, *R128(&k[0]));
d = _mm_aesenc_si128(d, *R128(&k[1]));
d = _mm_aesenc_si128(d, *R128(&k[2]));
d = _mm_aesenc_si128(d, *R128(&k[3]));
d = _mm_aesenc_si128(d, *R128(&k[4]));
d = _mm_aesenc_si128(d, *R128(&k[5]));
d = _mm_aesenc_si128(d, *R128(&k[6]));
d = _mm_aesenc_si128(d, *R128(&k[7]));
d = _mm_aesenc_si128(d, *R128(&k[8]));
d = _mm_aesenc_si128(d, *R128(&k[9]));
_mm_storeu_si128((R128(out)), d);
}
void cn_slow_hash(const void *data, size_t length, char *hash)
{
uint8_t long_state[MEMORY]; uint8_t long_state[MEMORY];
union cn_slow_hash_state state;
uint8_t text[INIT_SIZE_BYTE]; uint8_t text[INIT_SIZE_BYTE];
uint8_t a[AES_BLOCK_SIZE]; uint8_t a[AES_BLOCK_SIZE];
uint8_t b[AES_BLOCK_SIZE]; uint8_t b[AES_BLOCK_SIZE];
uint8_t c[AES_BLOCK_SIZE];
uint8_t d[AES_BLOCK_SIZE]; uint8_t d[AES_BLOCK_SIZE];
size_t i, j;
uint8_t aes_key[AES_KEY_SIZE]; uint8_t aes_key[AES_KEY_SIZE];
RDATA_ALIGN16 uint8_t expandedKey[256];
union cn_slow_hash_state state;
size_t i, j;
uint8_t *p = NULL;
oaes_ctx *aes_ctx; oaes_ctx *aes_ctx;
int useAes = check_aes_hw();
static void (*const extra_hashes[4])(const void *, size_t, char *) =
{
hash_extra_blake, hash_extra_groestl, hash_extra_jh, hash_extra_skein
};
hash_process(&state.hs, data, length); hash_process(&state.hs, data, length);
memcpy(text, state.init, INIT_SIZE_BYTE); memcpy(text, state.init, INIT_SIZE_BYTE);
memcpy(aes_key, state.hs.b, AES_KEY_SIZE);
aes_ctx = (oaes_ctx *) oaes_alloc();
oaes_key_import_data(aes_ctx, aes_key, AES_KEY_SIZE); aes_ctx = (oaes_ctx *) oaes_alloc();
for (i = 0; i < MEMORY / INIT_SIZE_BYTE; i++) { oaes_key_import_data(aes_ctx, state.hs.b, AES_KEY_SIZE);
for (j = 0; j < INIT_SIZE_BLK; j++) {
aesb_pseudo_round(&text[AES_BLOCK_SIZE * j], &text[AES_BLOCK_SIZE * j], aes_ctx->key->exp_data); // use aligned data
} memcpy(expandedKey, aes_ctx->key->exp_data, aes_ctx->key->exp_data_len);
if(useAes)
{
for(i = 0; i < MEMORY / INIT_SIZE_BYTE; i++)
{
for(j = 0; j < INIT_SIZE_BLK; j++)
aesni_pseudo_round(&text[AES_BLOCK_SIZE * j], &text[AES_BLOCK_SIZE * j], expandedKey);
memcpy(&long_state[i * INIT_SIZE_BYTE], text, INIT_SIZE_BYTE); memcpy(&long_state[i * INIT_SIZE_BYTE], text, INIT_SIZE_BYTE);
} }
}
else
{
for(i = 0; i < MEMORY / INIT_SIZE_BYTE; i++)
{
for(j = 0; j < INIT_SIZE_BLK; j++)
aesb_pseudo_round(&text[AES_BLOCK_SIZE * j], &text[AES_BLOCK_SIZE * j], expandedKey);
for (i = 0; i < 16; i++) { memcpy(&long_state[i * INIT_SIZE_BYTE], text, INIT_SIZE_BYTE);
a[i] = state.k[ i] ^ state.k[32 + i]; }
b[i] = state.k[16 + i] ^ state.k[48 + i];
} }
for (i = 0; i < ITER / 2; i++) { U64(a)[0] = U64(&state.k[0])[0] ^ U64(&state.k[32])[0];
/* Dependency chain: address -> read value ------+ U64(a)[1] = U64(&state.k[0])[1] ^ U64(&state.k[32])[1];
* written value <-+ hard function (AES or MUL) <+ U64(b)[0] = U64(&state.k[16])[0] ^ U64(&state.k[48])[0];
* next address <-+ U64(b)[1] = U64(&state.k[16])[1] ^ U64(&state.k[48])[1];
*/
/* Iteration 1 */ for(i = 0; i < ITER / 2; i++)
j = e2i(a, MEMORY / AES_BLOCK_SIZE); {
copy_block(c, &long_state[j * AES_BLOCK_SIZE]); #define TOTALBLOCKS (MEMORY / AES_BLOCK_SIZE)
aesb_single_round(c, c, a); #define state_index(x) (((*((uint64_t *)x) >> 4) & (TOTALBLOCKS - 1)) << 4)
xor_blocks(b, c);
swap_blocks(b, c); // Iteration 1
copy_block(&long_state[j * AES_BLOCK_SIZE], c); p = &long_state[state_index(a)];
assert(j == e2i(a, MEMORY / AES_BLOCK_SIZE));
if(useAes)
_mm_storeu_si128(R128(p), _mm_aesenc_si128(_mm_loadu_si128(R128(p)), _mm_loadu_si128(R128(a))));
else
aesb_single_round(p, p, a);
xor_blocks(b, p);
swap_blocks(b, p);
swap_blocks(a, b); swap_blocks(a, b);
/* Iteration 2 */
j = e2i(a, MEMORY / AES_BLOCK_SIZE); // Iteration 2
copy_block(c, &long_state[j * AES_BLOCK_SIZE]); p = &long_state[state_index(a)];
mul(a, c, d);
mul(a, p, d);
sum_half_blocks(b, d); sum_half_blocks(b, d);
swap_blocks(b, c); swap_blocks(b, p);
xor_blocks(b, c); xor_blocks(b, p);
copy_block(&long_state[j * AES_BLOCK_SIZE], c);
assert(j == e2i(a, MEMORY / AES_BLOCK_SIZE));
swap_blocks(a, b); swap_blocks(a, b);
} }
memcpy(text, state.init, INIT_SIZE_BYTE); memcpy(text, state.init, INIT_SIZE_BYTE);
oaes_key_import_data(aes_ctx, &state.hs.b[32], AES_KEY_SIZE); oaes_key_import_data(aes_ctx, &state.hs.b[32], AES_KEY_SIZE);
for (i = 0; i < MEMORY / INIT_SIZE_BYTE; i++) { memcpy(expandedKey, aes_ctx->key->exp_data, aes_ctx->key->exp_data_len);
for (j = 0; j < INIT_SIZE_BLK; j++) { if(useAes)
{
for(i = 0; i < MEMORY / INIT_SIZE_BYTE; i++)
{
for(j = 0; j < INIT_SIZE_BLK; j++)
{
xor_blocks(&text[j * AES_BLOCK_SIZE], &long_state[i * INIT_SIZE_BYTE + j * AES_BLOCK_SIZE]); xor_blocks(&text[j * AES_BLOCK_SIZE], &long_state[i * INIT_SIZE_BYTE + j * AES_BLOCK_SIZE]);
aesb_pseudo_round(&text[AES_BLOCK_SIZE * j], &text[AES_BLOCK_SIZE * j], aes_ctx->key->exp_data); aesni_pseudo_round(&text[j * AES_BLOCK_SIZE], &text[j * AES_BLOCK_SIZE], expandedKey);
} }
} }
}
else
{
for(i = 0; i < MEMORY / INIT_SIZE_BYTE; i++)
{
for(j = 0; j < INIT_SIZE_BLK; j++)
{
xor_blocks(&text[j * AES_BLOCK_SIZE], &long_state[i * INIT_SIZE_BYTE + j * AES_BLOCK_SIZE]);
aesb_pseudo_round(&text[AES_BLOCK_SIZE * j], &text[AES_BLOCK_SIZE * j], expandedKey);
}
}
}
oaes_free((OAES_CTX **) &aes_ctx);
memcpy(state.init, text, INIT_SIZE_BYTE); memcpy(state.init, text, INIT_SIZE_BYTE);
hash_permutation(&state.hs); hash_permutation(&state.hs);
/*memcpy(hash, &state, 32);*/
extra_hashes[state.hs.b[0] & 3](&state, 200, hash); extra_hashes[state.hs.b[0] & 3](&state, 200, hash);
oaes_free((OAES_CTX **) &aes_ctx);
} }