mirror of
https://github.com/xmrig/xmrig.git
synced 2024-11-19 10:31:05 +00:00
9a28ad590c
This time, the performance increase is got with MSVC and GCC. On non-AES CPU, there were an useless load/store SSE2 register. The last MSVC "hack" is replaced by a portable code and he's more complete (a load is saved). On my C2Q6600, with 3 thread, I have +16% with MSVC2015 and +20% with GCC 7.3, compared to official 2.4.4 version.
129 lines
5.9 KiB
C++
129 lines
5.9 KiB
C++
/*
|
|
* This program is free software: you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation, either version 3 of the License, or
|
|
* any later version.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
*
|
|
* Additional permission under GNU GPL version 3 section 7
|
|
*
|
|
* If you modify this Program, or any covered work, by linking or combining
|
|
* it with OpenSSL (or a modified version of that library), containing parts
|
|
* covered by the terms of OpenSSL License and SSLeay License, the licensors
|
|
* of this Program grant you additional permission to convey the resulting work.
|
|
*
|
|
*/
|
|
|
|
/*
|
|
* Parts of this file are originally copyright (c) 2014-2017, The Monero Project
|
|
*/
|
|
#pragma once
|
|
|
|
|
|
#if defined(XMRIG_ARM)
|
|
# include "crypto/SSE2NEON.h"
|
|
#elif defined(__GNUC__)
|
|
# include <x86intrin.h>
|
|
#else
|
|
# include <intrin.h>
|
|
#endif
|
|
|
|
#include <inttypes.h>
|
|
|
|
|
|
#define saes_data(w) {\
|
|
w(0x63), w(0x7c), w(0x77), w(0x7b), w(0xf2), w(0x6b), w(0x6f), w(0xc5),\
|
|
w(0x30), w(0x01), w(0x67), w(0x2b), w(0xfe), w(0xd7), w(0xab), w(0x76),\
|
|
w(0xca), w(0x82), w(0xc9), w(0x7d), w(0xfa), w(0x59), w(0x47), w(0xf0),\
|
|
w(0xad), w(0xd4), w(0xa2), w(0xaf), w(0x9c), w(0xa4), w(0x72), w(0xc0),\
|
|
w(0xb7), w(0xfd), w(0x93), w(0x26), w(0x36), w(0x3f), w(0xf7), w(0xcc),\
|
|
w(0x34), w(0xa5), w(0xe5), w(0xf1), w(0x71), w(0xd8), w(0x31), w(0x15),\
|
|
w(0x04), w(0xc7), w(0x23), w(0xc3), w(0x18), w(0x96), w(0x05), w(0x9a),\
|
|
w(0x07), w(0x12), w(0x80), w(0xe2), w(0xeb), w(0x27), w(0xb2), w(0x75),\
|
|
w(0x09), w(0x83), w(0x2c), w(0x1a), w(0x1b), w(0x6e), w(0x5a), w(0xa0),\
|
|
w(0x52), w(0x3b), w(0xd6), w(0xb3), w(0x29), w(0xe3), w(0x2f), w(0x84),\
|
|
w(0x53), w(0xd1), w(0x00), w(0xed), w(0x20), w(0xfc), w(0xb1), w(0x5b),\
|
|
w(0x6a), w(0xcb), w(0xbe), w(0x39), w(0x4a), w(0x4c), w(0x58), w(0xcf),\
|
|
w(0xd0), w(0xef), w(0xaa), w(0xfb), w(0x43), w(0x4d), w(0x33), w(0x85),\
|
|
w(0x45), w(0xf9), w(0x02), w(0x7f), w(0x50), w(0x3c), w(0x9f), w(0xa8),\
|
|
w(0x51), w(0xa3), w(0x40), w(0x8f), w(0x92), w(0x9d), w(0x38), w(0xf5),\
|
|
w(0xbc), w(0xb6), w(0xda), w(0x21), w(0x10), w(0xff), w(0xf3), w(0xd2),\
|
|
w(0xcd), w(0x0c), w(0x13), w(0xec), w(0x5f), w(0x97), w(0x44), w(0x17),\
|
|
w(0xc4), w(0xa7), w(0x7e), w(0x3d), w(0x64), w(0x5d), w(0x19), w(0x73),\
|
|
w(0x60), w(0x81), w(0x4f), w(0xdc), w(0x22), w(0x2a), w(0x90), w(0x88),\
|
|
w(0x46), w(0xee), w(0xb8), w(0x14), w(0xde), w(0x5e), w(0x0b), w(0xdb),\
|
|
w(0xe0), w(0x32), w(0x3a), w(0x0a), w(0x49), w(0x06), w(0x24), w(0x5c),\
|
|
w(0xc2), w(0xd3), w(0xac), w(0x62), w(0x91), w(0x95), w(0xe4), w(0x79),\
|
|
w(0xe7), w(0xc8), w(0x37), w(0x6d), w(0x8d), w(0xd5), w(0x4e), w(0xa9),\
|
|
w(0x6c), w(0x56), w(0xf4), w(0xea), w(0x65), w(0x7a), w(0xae), w(0x08),\
|
|
w(0xba), w(0x78), w(0x25), w(0x2e), w(0x1c), w(0xa6), w(0xb4), w(0xc6),\
|
|
w(0xe8), w(0xdd), w(0x74), w(0x1f), w(0x4b), w(0xbd), w(0x8b), w(0x8a),\
|
|
w(0x70), w(0x3e), w(0xb5), w(0x66), w(0x48), w(0x03), w(0xf6), w(0x0e),\
|
|
w(0x61), w(0x35), w(0x57), w(0xb9), w(0x86), w(0xc1), w(0x1d), w(0x9e),\
|
|
w(0xe1), w(0xf8), w(0x98), w(0x11), w(0x69), w(0xd9), w(0x8e), w(0x94),\
|
|
w(0x9b), w(0x1e), w(0x87), w(0xe9), w(0xce), w(0x55), w(0x28), w(0xdf),\
|
|
w(0x8c), w(0xa1), w(0x89), w(0x0d), w(0xbf), w(0xe6), w(0x42), w(0x68),\
|
|
w(0x41), w(0x99), w(0x2d), w(0x0f), w(0xb0), w(0x54), w(0xbb), w(0x16) }
|
|
|
|
#define SAES_WPOLY 0x011b
|
|
|
|
#define saes_b2w(b0, b1, b2, b3) (((uint32_t)(b3) << 24) | \
|
|
((uint32_t)(b2) << 16) | ((uint32_t)(b1) << 8) | (b0))
|
|
|
|
#define saes_f2(x) ((x<<1) ^ (((x>>7) & 1) * SAES_WPOLY))
|
|
#define saes_f3(x) (saes_f2(x) ^ x)
|
|
#define saes_h0(x) (x)
|
|
|
|
#define saes_u0(p) saes_b2w(saes_f2(p), p, p, saes_f3(p))
|
|
#define saes_u1(p) saes_b2w(saes_f3(p), saes_f2(p), p, p)
|
|
#define saes_u2(p) saes_b2w( p, saes_f3(p), saes_f2(p), p)
|
|
#define saes_u3(p) saes_b2w( p, p, saes_f3(p), saes_f2(p))
|
|
|
|
alignas(16) const uint32_t saes_table[4][256] = { saes_data(saes_u0), saes_data(saes_u1), saes_data(saes_u2), saes_data(saes_u3) };
|
|
alignas(16) const uint8_t saes_sbox[256] = saes_data(saes_h0);
|
|
|
|
static inline __m128i soft_aesenc(const uint32_t* in, __m128i key)
|
|
{
|
|
const uint32_t x0 = in[0];
|
|
const uint32_t x1 = in[1];
|
|
const uint32_t x2 = in[2];
|
|
const uint32_t x3 = in[3];
|
|
|
|
__m128i out = _mm_set_epi32(
|
|
(saes_table[0][x3 & 0xff] ^ saes_table[1][(x0 >> 8) & 0xff] ^ saes_table[2][(x1 >> 16) & 0xff] ^ saes_table[3][x2 >> 24]),
|
|
(saes_table[0][x2 & 0xff] ^ saes_table[1][(x3 >> 8) & 0xff] ^ saes_table[2][(x0 >> 16) & 0xff] ^ saes_table[3][x1 >> 24]),
|
|
(saes_table[0][x1 & 0xff] ^ saes_table[1][(x2 >> 8) & 0xff] ^ saes_table[2][(x3 >> 16) & 0xff] ^ saes_table[3][x0 >> 24]),
|
|
(saes_table[0][x0 & 0xff] ^ saes_table[1][(x1 >> 8) & 0xff] ^ saes_table[2][(x2 >> 16) & 0xff] ^ saes_table[3][x3 >> 24]));
|
|
|
|
return _mm_xor_si128(out, key);
|
|
}
|
|
|
|
static inline uint32_t sub_word(uint32_t key)
|
|
{
|
|
return (saes_sbox[key >> 24 ] << 24) |
|
|
(saes_sbox[(key >> 16) & 0xff] << 16 ) |
|
|
(saes_sbox[(key >> 8) & 0xff] << 8 ) |
|
|
saes_sbox[key & 0xff];
|
|
}
|
|
|
|
#if defined(__clang__) || defined(XMRIG_ARM)
|
|
static inline uint32_t _rotr(uint32_t value, uint32_t amount)
|
|
{
|
|
return (value >> amount) | (value << ((32 - amount) & 31));
|
|
}
|
|
#endif
|
|
|
|
template<uint8_t rcon>
|
|
static inline __m128i soft_aeskeygenassist(__m128i key)
|
|
{
|
|
const uint32_t X1 = sub_word(_mm_cvtsi128_si32(_mm_shuffle_epi32(key, 0x55)));
|
|
const uint32_t X3 = sub_word(_mm_cvtsi128_si32(_mm_shuffle_epi32(key, 0xFF)));
|
|
return _mm_set_epi32(_rotr(X3, 8) ^ rcon, X3, _rotr(X1, 8) ^ rcon, X1);
|
|
}
|