Added ARMv8 (aarch64) support.

This commit is contained in:
XMRig 2017-11-06 03:11:35 +03:00
parent d403dcf95c
commit 6cc152e26f
11 changed files with 2188 additions and 60 deletions

View file

@ -6,6 +6,7 @@ option(WITH_AEON "CryptoNight-Lite support" ON)
option(WITH_HTTPD "HTTP REST API" ON)
include (CheckIncludeFile)
include (cmake/cpu.cmake)
set(HEADERS
@ -56,7 +57,6 @@ set(HEADERS_CRYPTO
src/crypto/c_keccak.h
src/crypto/c_skein.h
src/crypto/CryptoNight.h
src/crypto/CryptoNight_p.h
src/crypto/CryptoNight_test.h
src/crypto/groestl_tables.h
src/crypto/hash.h
@ -64,6 +64,12 @@ set(HEADERS_CRYPTO
src/crypto/soft_aes.h
)
if (XMRIG_ARM64)
set(HEADERS_CRYPTO "${HEADERS_CRYPTO}" src/crypto/CryptoNight_arm64.h)
else()
set(HEADERS_CRYPTO "${HEADERS_CRYPTO}" src/crypto/CryptoNight_x86.h)
endif()
set(SOURCES
src/api/Api.cpp
src/api/ApiState.cpp
@ -138,63 +144,13 @@ endif()
add_definitions(/D__STDC_FORMAT_MACROS)
add_definitions(/DUNICODE)
add_definitions(/DRAPIDJSON_SSE2)
#add_definitions(/DAPP_DEBUG)
set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/cmake")
find_package(UV REQUIRED)
if ("${CMAKE_BUILD_TYPE}" STREQUAL "")
set(CMAKE_BUILD_TYPE Release)
endif()
set(CMAKE_CXX_STANDARD_REQUIRED ON)
set(CMAKE_CXX_EXTENSIONS OFF)
set(CMAKE_CXX_STANDARD 11)
# https://cmake.org/cmake/help/latest/variable/CMAKE_LANG_COMPILER_ID.html
if (CMAKE_CXX_COMPILER_ID MATCHES GNU)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -maes -Wall -Wno-strict-aliasing")
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Ofast -funroll-loops -fvariable-expansion-in-unroller -ftree-loop-if-convert-stores -fmerge-all-constants -fbranch-target-load-optimize2")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maes -Wall -fno-exceptions -fno-rtti")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -s -funroll-loops -fvariable-expansion-in-unroller -ftree-loop-if-convert-stores -fmerge-all-constants -fbranch-target-load-optimize2")
if (WIN32)
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static")
else()
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static-libgcc -static-libstdc++")
endif()
add_definitions(/D_GNU_SOURCE)
if (${CMAKE_VERSION} VERSION_LESS "3.1.0")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
endif()
#set(CMAKE_C_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -gdwarf-2")
elseif (CMAKE_CXX_COMPILER_ID MATCHES MSVC)
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} /Ox /Ot /Oi /MT /GL")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /Ox /Ot /Oi /MT /GL")
add_definitions(/D_CRT_SECURE_NO_WARNINGS)
add_definitions(/D_CRT_NONSTDC_NO_WARNINGS)
add_definitions(/DNOMINMAX)
elseif (CMAKE_CXX_COMPILER_ID MATCHES Clang)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -maes -Wall")
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Ofast -funroll-loops -fmerge-all-constants")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maes -Wall -fno-exceptions -fno-rtti -Wno-missing-braces")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -funroll-loops -fmerge-all-constants")
endif()
include(cmake/flags.cmake)
if (WITH_LIBCPUID)
add_subdirectory(src/3rdparty/libcpuid)
@ -204,7 +160,12 @@ if (WITH_LIBCPUID)
set(SOURCES_CPUID src/Cpu.cpp)
else()
add_definitions(/DXMRIG_NO_LIBCPUID)
set(SOURCES_CPUID src/Cpu_stub.cpp)
if (XMRIG_ARM64)
set(SOURCES_CPUID src/Cpu_arm.cpp)
else()
set(SOURCES_CPUID src/Cpu_stub.cpp)
endif()
endif()
CHECK_INCLUDE_FILE (syslog.h HAVE_SYSLOG_H)

16
cmake/cpu.cmake Normal file
View file

@ -0,0 +1,16 @@
if (NOT CMAKE_SYSTEM_PROCESSOR)
message(WARNING "CMAKE_SYSTEM_PROCESSOR not defined")
endif()
if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|AMD64)$")
add_definitions(/DRAPIDJSON_SSE2)
endif()
if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64)$")
set(XMRIG_ARM64 ON)
set(WITH_LIBCPUID OFF)
add_definitions(/DXMRIG_ARM)
endif()

63
cmake/flags.cmake Normal file
View file

@ -0,0 +1,63 @@
set(CMAKE_CXX_STANDARD_REQUIRED ON)
set(CMAKE_CXX_EXTENSIONS OFF)
set(CMAKE_CXX_STANDARD 11)
if ("${CMAKE_BUILD_TYPE}" STREQUAL "")
set(CMAKE_BUILD_TYPE Release)
endif()
if (CMAKE_CXX_COMPILER_ID MATCHES GNU)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wno-strict-aliasing")
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Ofast -funroll-loops -fvariable-expansion-in-unroller -ftree-loop-if-convert-stores -fmerge-all-constants -fbranch-target-load-optimize2")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -fno-exceptions -fno-rtti")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -s -funroll-loops -fvariable-expansion-in-unroller -ftree-loop-if-convert-stores -fmerge-all-constants -fbranch-target-load-optimize2")
if (CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv8-a+crypto")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=armv8-a+crypto -flax-vector-conversions")
else()
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -maes")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maes")
endif()
if (WIN32)
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static")
else()
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static-libgcc -static-libstdc++")
endif()
add_definitions(/D_GNU_SOURCE)
if (${CMAKE_VERSION} VERSION_LESS "3.1.0")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
endif()
#set(CMAKE_C_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -gdwarf-2")
elseif (CMAKE_CXX_COMPILER_ID MATCHES MSVC)
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} /Ox /Ot /Oi /MT /GL")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /Ox /Ot /Oi /MT /GL")
add_definitions(/D_CRT_SECURE_NO_WARNINGS)
add_definitions(/D_CRT_NONSTDC_NO_WARNINGS)
add_definitions(/DNOMINMAX)
elseif (CMAKE_CXX_COMPILER_ID MATCHES Clang)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall")
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Ofast -funroll-loops -fmerge-all-constants")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -fno-exceptions -fno-rtti -Wno-missing-braces")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -funroll-loops -fmerge-all-constants")
if (CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv8-a+crypto")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=armv8-a+crypto")
else()
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -maes")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maes")
endif()
endif()

65
src/3rdparty/aligned_malloc.h vendored Normal file
View file

@ -0,0 +1,65 @@
/* XMRig
* Copyright 2010 Jeff Garzik <jgarzik@pobox.com>
* Copyright 2012-2014 pooler <pooler@litecoinpool.org>
* Copyright 2014 Lucas Jones <https://github.com/lucasjones>
* Copyright 2014-2016 Wolf9466 <https://github.com/OhGodAPet>
* Copyright 2016 Jay D Dee <jayddee246@gmail.com>
* Copyright 2016-2017 XMRig <support@xmrig.com>
*
*
* 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
* (at your option) 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/>.
*/
#ifndef __ALIGNED_MALLOC_H__
#define __ALIGNED_MALLOC_H__
#include <stdlib.h>
#ifndef __cplusplus
extern int posix_memalign(void **__memptr, size_t __alignment, size_t __size);
#else
// Some systems (e.g. those with GNU libc) declare posix_memalign with an
// exception specifier. Via an "egregious workaround" in
// Sema::CheckEquivalentExceptionSpec, Clang accepts the following as a valid
// redeclaration of glibc's declaration.
extern "C" int posix_memalign(void **__memptr, size_t __alignment, size_t __size);
#endif
static __inline__ void *__attribute__((__always_inline__, __malloc__)) _mm_malloc(size_t __size, size_t __align)
{
if (__align == 1) {
return malloc(__size);
}
if (!(__align & (__align - 1)) && __align < sizeof(void *))
__align = sizeof(void *);
void *__mallocedMemory;
if (posix_memalign(&__mallocedMemory, __align, __size)) {
return 0;
}
return __mallocedMemory;
}
static __inline__ void __attribute__((__always_inline__)) _mm_free(void *__p)
{
free(__p);
}
#endif /* __ALIGNED_MALLOC_H__ */

52
src/Cpu_arm.cpp Normal file
View file

@ -0,0 +1,52 @@
/* XMRig
* Copyright 2010 Jeff Garzik <jgarzik@pobox.com>
* Copyright 2012-2014 pooler <pooler@litecoinpool.org>
* Copyright 2014 Lucas Jones <https://github.com/lucasjones>
* Copyright 2014-2016 Wolf9466 <https://github.com/OhGodAPet>
* Copyright 2016 Jay D Dee <jayddee246@gmail.com>
* Copyright 2016-2017 XMRig <support@xmrig.com>
*
*
* 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
* (at your option) 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/>.
*/
#include <string.h>
#include "Cpu.h"
char Cpu::m_brand[64] = { 0 };
int Cpu::m_flags = 0;
int Cpu::m_l2_cache = 0;
int Cpu::m_l3_cache = 0;
int Cpu::m_sockets = 1;
int Cpu::m_totalCores = 0;
int Cpu::m_totalThreads = 0;
int Cpu::optimalThreadsCount(int algo, bool doubleHash, int maxCpuUsage)
{
return m_totalThreads;
}
void Cpu::initCommon()
{
memcpy(m_brand, "Unknown", 7);
m_flags |= X86_64;
m_flags |= AES;
}

View file

@ -23,10 +23,16 @@
#include <stdlib.h>
#include <mm_malloc.h>
#include <sys/mman.h>
#if defined(XMRIG_ARM) && !defined(__clang__)
# include "aligned_malloc.h"
#else
# include <mm_malloc.h>
#endif
#include "crypto/CryptoNight.h"
#include "log/Log.h"
#include "Mem.h"

View file

@ -23,7 +23,13 @@
#include "crypto/CryptoNight.h"
#include "crypto/CryptoNight_p.h"
#if defined(__aarch64__)
# include "crypto/CryptoNight_arm64.h"
#else
# include "crypto/CryptoNight_x86.h"
#endif
#include "crypto/CryptoNight_test.h"
#include "net/Job.h"
#include "net/JobResult.h"

View file

@ -0,0 +1,458 @@
/* XMRig
* Copyright 2010 Jeff Garzik <jgarzik@pobox.com>
* Copyright 2012-2014 pooler <pooler@litecoinpool.org>
* Copyright 2014 Lucas Jones <https://github.com/lucasjones>
* Copyright 2014-2016 Wolf9466 <https://github.com/OhGodAPet>
* Copyright 2016 Jay D Dee <jayddee246@gmail.com>
* Copyright 2016 Imran Yusuff <https://github.com/imranyusuff>
* Copyright 2016-2017 XMRig <support@xmrig.com>
*
*
* 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
* (at your option) 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/>.
*/
#ifndef __CRYPTONIGHT_ARM64_H__
#define __CRYPTONIGHT_ARM64_H__
#if defined(XMRIG_ARM) && !defined(__clang__)
# include "aligned_malloc.h"
#else
# include <mm_malloc.h>
#endif
#include "crypto/CryptoNight.h"
#include "crypto/soft_aes.h"
extern "C"
{
#include "crypto/c_keccak.h"
#include "crypto/c_groestl.h"
#include "crypto/c_blake256.h"
#include "crypto/c_jh.h"
#include "crypto/c_skein.h"
}
static inline void do_blake_hash(const void* input, size_t len, char* output) {
blake256_hash(reinterpret_cast<uint8_t*>(output), static_cast<const uint8_t*>(input), len);
}
static inline void do_groestl_hash(const void* input, size_t len, char* output) {
groestl(static_cast<const uint8_t*>(input), len * 8, reinterpret_cast<uint8_t*>(output));
}
static inline void do_jh_hash(const void* input, size_t len, char* output) {
jh_hash(32 * 8, static_cast<const uint8_t*>(input), 8 * len, reinterpret_cast<uint8_t*>(output));
}
static inline void do_skein_hash(const void* input, size_t len, char* output) {
xmr_skein(static_cast<const uint8_t*>(input), reinterpret_cast<uint8_t*>(output));
}
void (* const extra_hashes[4])(const void *, size_t, char *) = {do_blake_hash, do_groestl_hash, do_jh_hash, do_skein_hash};
static inline __attribute__((always_inline)) __m128i _mm_set_epi64x(const uint64_t a, const uint64_t b)
{
return vcombine_u64(vcreate_u64(b), vcreate_u64(a));
}
/* this one was not implemented yet so here it is */
static inline __attribute__((always_inline)) uint64_t _mm_cvtsi128_si64(__m128i a)
{
return vgetq_lane_u64(a, 0);
}
#define EXTRACT64(X) _mm_cvtsi128_si64(X)
static inline uint64_t __umul128(uint64_t a, uint64_t b, uint64_t* hi)
{
unsigned __int128 r = (unsigned __int128) a * (unsigned __int128) b;
*hi = r >> 64;
return (uint64_t) r;
}
// This will shift and xor tmp1 into itself as 4 32-bit vals such as
// sl_xor(a1 a2 a3 a4) = a1 (a2^a1) (a3^a2^a1) (a4^a3^a2^a1)
static inline __m128i sl_xor(__m128i tmp1)
{
__m128i tmp4;
tmp4 = _mm_slli_si128(tmp1, 0x04);
tmp1 = _mm_xor_si128(tmp1, tmp4);
tmp4 = _mm_slli_si128(tmp4, 0x04);
tmp1 = _mm_xor_si128(tmp1, tmp4);
tmp4 = _mm_slli_si128(tmp4, 0x04);
tmp1 = _mm_xor_si128(tmp1, tmp4);
return tmp1;
}
template<uint8_t rcon>
static inline void aes_genkey_sub(__m128i* xout0, __m128i* xout2)
{
// __m128i xout1 = _mm_aeskeygenassist_si128(*xout2, rcon);
// xout1 = _mm_shuffle_epi32(xout1, 0xFF); // see PSHUFD, set all elems to 4th elem
// *xout0 = sl_xor(*xout0);
// *xout0 = _mm_xor_si128(*xout0, xout1);
// xout1 = _mm_aeskeygenassist_si128(*xout0, 0x00);
// xout1 = _mm_shuffle_epi32(xout1, 0xAA); // see PSHUFD, set all elems to 3rd elem
// *xout2 = sl_xor(*xout2);
// *xout2 = _mm_xor_si128(*xout2, xout1);
}
template<uint8_t rcon>
static inline void soft_aes_genkey_sub(__m128i* xout0, __m128i* xout2)
{
__m128i xout1 = soft_aeskeygenassist<rcon>(*xout2);
xout1 = _mm_shuffle_epi32(xout1, 0xFF); // see PSHUFD, set all elems to 4th elem
*xout0 = sl_xor(*xout0);
*xout0 = _mm_xor_si128(*xout0, xout1);
xout1 = soft_aeskeygenassist<0x00>(*xout0);
xout1 = _mm_shuffle_epi32(xout1, 0xAA); // see PSHUFD, set all elems to 3rd elem
*xout2 = sl_xor(*xout2);
*xout2 = _mm_xor_si128(*xout2, xout1);
}
template<bool SOFT_AES>
static inline void aes_genkey(const __m128i* memory, __m128i* k0, __m128i* k1, __m128i* k2, __m128i* k3, __m128i* k4, __m128i* k5, __m128i* k6, __m128i* k7, __m128i* k8, __m128i* k9)
{
__m128i xout0 = _mm_load_si128(memory);
__m128i xout2 = _mm_load_si128(memory + 1);
*k0 = xout0;
*k1 = xout2;
SOFT_AES ? soft_aes_genkey_sub<0x01>(&xout0, &xout2) : soft_aes_genkey_sub<0x01>(&xout0, &xout2);
*k2 = xout0;
*k3 = xout2;
SOFT_AES ? soft_aes_genkey_sub<0x02>(&xout0, &xout2) : soft_aes_genkey_sub<0x02>(&xout0, &xout2);
*k4 = xout0;
*k5 = xout2;
SOFT_AES ? soft_aes_genkey_sub<0x04>(&xout0, &xout2) : soft_aes_genkey_sub<0x04>(&xout0, &xout2);
*k6 = xout0;
*k7 = xout2;
SOFT_AES ? soft_aes_genkey_sub<0x08>(&xout0, &xout2) : soft_aes_genkey_sub<0x08>(&xout0, &xout2);
*k8 = xout0;
*k9 = xout2;
}
template<bool SOFT_AES>
static inline void aes_round(__m128i key, __m128i* x0, __m128i* x1, __m128i* x2, __m128i* x3, __m128i* x4, __m128i* x5, __m128i* x6, __m128i* x7)
{
if (SOFT_AES) {
*x0 = soft_aesenc(*x0, key);
*x1 = soft_aesenc(*x1, key);
*x2 = soft_aesenc(*x2, key);
*x3 = soft_aesenc(*x3, key);
*x4 = soft_aesenc(*x4, key);
*x5 = soft_aesenc(*x5, key);
*x6 = soft_aesenc(*x6, key);
*x7 = soft_aesenc(*x7, key);
}
else {
*x0 = vaesmcq_u8(vaeseq_u8(*((uint8x16_t *) x0), key));
*x1 = vaesmcq_u8(vaeseq_u8(*((uint8x16_t *) x1), key));
*x2 = vaesmcq_u8(vaeseq_u8(*((uint8x16_t *) x2), key));
*x3 = vaesmcq_u8(vaeseq_u8(*((uint8x16_t *) x3), key));
*x4 = vaesmcq_u8(vaeseq_u8(*((uint8x16_t *) x4), key));
*x5 = vaesmcq_u8(vaeseq_u8(*((uint8x16_t *) x5), key));
*x6 = vaesmcq_u8(vaeseq_u8(*((uint8x16_t *) x6), key));
*x7 = vaesmcq_u8(vaeseq_u8(*((uint8x16_t *) x7), key));
}
}
template<size_t MEM, bool SOFT_AES>
static inline void cn_explode_scratchpad(const __m128i *input, __m128i *output)
{
__m128i xin0, xin1, xin2, xin3, xin4, xin5, xin6, xin7;
__m128i k0, k1, k2, k3, k4, k5, k6, k7, k8, k9;
aes_genkey<SOFT_AES>(input, &k0, &k1, &k2, &k3, &k4, &k5, &k6, &k7, &k8, &k9);
xin0 = _mm_load_si128(input + 4);
xin1 = _mm_load_si128(input + 5);
xin2 = _mm_load_si128(input + 6);
xin3 = _mm_load_si128(input + 7);
xin4 = _mm_load_si128(input + 8);
xin5 = _mm_load_si128(input + 9);
xin6 = _mm_load_si128(input + 10);
xin7 = _mm_load_si128(input + 11);
for (size_t i = 0; i < MEM / sizeof(__m128i); i += 8) {
if (!SOFT_AES) {
aes_round<SOFT_AES>(_mm_setzero_si128(), &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
}
aes_round<SOFT_AES>(k0, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
aes_round<SOFT_AES>(k1, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
aes_round<SOFT_AES>(k2, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
aes_round<SOFT_AES>(k3, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
aes_round<SOFT_AES>(k4, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
aes_round<SOFT_AES>(k5, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
aes_round<SOFT_AES>(k6, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
aes_round<SOFT_AES>(k7, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
aes_round<SOFT_AES>(k8, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
if (!SOFT_AES) {
xin0 ^= k9;
xin1 ^= k9;
xin2 ^= k9;
xin3 ^= k9;
xin4 ^= k9;
xin5 ^= k9;
xin6 ^= k9;
xin7 ^= k9;
}
else {
aes_round<SOFT_AES>(k9, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
}
_mm_store_si128(output + i + 0, xin0);
_mm_store_si128(output + i + 1, xin1);
_mm_store_si128(output + i + 2, xin2);
_mm_store_si128(output + i + 3, xin3);
_mm_store_si128(output + i + 4, xin4);
_mm_store_si128(output + i + 5, xin5);
_mm_store_si128(output + i + 6, xin6);
_mm_store_si128(output + i + 7, xin7);
}
}
template<size_t MEM, bool SOFT_AES>
static inline void cn_implode_scratchpad(const __m128i *input, __m128i *output)
{
__m128i xout0, xout1, xout2, xout3, xout4, xout5, xout6, xout7;
__m128i k0, k1, k2, k3, k4, k5, k6, k7, k8, k9;
aes_genkey<SOFT_AES>(output + 2, &k0, &k1, &k2, &k3, &k4, &k5, &k6, &k7, &k8, &k9);
xout0 = _mm_load_si128(output + 4);
xout1 = _mm_load_si128(output + 5);
xout2 = _mm_load_si128(output + 6);
xout3 = _mm_load_si128(output + 7);
xout4 = _mm_load_si128(output + 8);
xout5 = _mm_load_si128(output + 9);
xout6 = _mm_load_si128(output + 10);
xout7 = _mm_load_si128(output + 11);
for (size_t i = 0; i < MEM / sizeof(__m128i); i += 8)
{
xout0 = _mm_xor_si128(_mm_load_si128(input + i + 0), xout0);
xout1 = _mm_xor_si128(_mm_load_si128(input + i + 1), xout1);
xout2 = _mm_xor_si128(_mm_load_si128(input + i + 2), xout2);
xout3 = _mm_xor_si128(_mm_load_si128(input + i + 3), xout3);
xout4 = _mm_xor_si128(_mm_load_si128(input + i + 4), xout4);
xout5 = _mm_xor_si128(_mm_load_si128(input + i + 5), xout5);
xout6 = _mm_xor_si128(_mm_load_si128(input + i + 6), xout6);
xout7 = _mm_xor_si128(_mm_load_si128(input + i + 7), xout7);
if (!SOFT_AES) {
aes_round<SOFT_AES>(_mm_setzero_si128(), &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
}
aes_round<SOFT_AES>(k0, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
aes_round<SOFT_AES>(k1, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
aes_round<SOFT_AES>(k2, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
aes_round<SOFT_AES>(k3, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
aes_round<SOFT_AES>(k4, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
aes_round<SOFT_AES>(k5, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
aes_round<SOFT_AES>(k6, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
aes_round<SOFT_AES>(k7, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
aes_round<SOFT_AES>(k8, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
if (!SOFT_AES) {
xout0 ^= k9;
xout1 ^= k9;
xout2 ^= k9;
xout3 ^= k9;
xout4 ^= k9;
xout5 ^= k9;
xout6 ^= k9;
xout7 ^= k9;
}
else {
aes_round<SOFT_AES>(k9, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
}
}
_mm_store_si128(output + 4, xout0);
_mm_store_si128(output + 5, xout1);
_mm_store_si128(output + 6, xout2);
_mm_store_si128(output + 7, xout3);
_mm_store_si128(output + 8, xout4);
_mm_store_si128(output + 9, xout5);
_mm_store_si128(output + 10, xout6);
_mm_store_si128(output + 11, xout7);
}
template<size_t ITERATIONS, size_t MEM, size_t MASK, bool SOFT_AES>
inline void cryptonight_hash(const void *__restrict__ input, size_t size, void *__restrict__ output, cryptonight_ctx *__restrict__ ctx)
{
keccak(static_cast<const uint8_t*>(input), (int) size, ctx->state0, 200);
cn_explode_scratchpad<MEM, SOFT_AES>((__m128i*) ctx->state0, (__m128i*) ctx->memory);
const uint8_t* l0 = ctx->memory;
uint64_t* h0 = reinterpret_cast<uint64_t*>(ctx->state0);
uint64_t al0 = h0[0] ^ h0[4];
uint64_t ah0 = h0[1] ^ h0[5];
__m128i bx0 = _mm_set_epi64x(h0[3] ^ h0[7], h0[2] ^ h0[6]);
uint64_t idx0 = h0[0] ^ h0[4];
for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
if (SOFT_AES) {
cx = soft_aesenc(cx, _mm_set_epi64x(ah0, al0));
}
else {
cx = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah0, al0);
}
_mm_store_si128((__m128i *) &l0[idx0 & MASK], _mm_xor_si128(bx0, cx));
idx0 = EXTRACT64(cx);
bx0 = cx;
uint64_t hi, lo, cl, ch;
cl = ((uint64_t*) &l0[idx0 & MASK])[0];
ch = ((uint64_t*) &l0[idx0 & MASK])[1];
lo = __umul128(idx0, cl, &hi);
al0 += hi;
ah0 += lo;
((uint64_t*)&l0[idx0 & MASK])[0] = al0;
((uint64_t*)&l0[idx0 & MASK])[1] = ah0;
ah0 ^= ch;
al0 ^= cl;
idx0 = al0;
}
cn_implode_scratchpad<MEM, SOFT_AES>((__m128i*) ctx->memory, (__m128i*) ctx->state0);
keccakf(h0, 24);
extra_hashes[ctx->state0[0] & 3](ctx->state0, 200, static_cast<char*>(output));
}
template<size_t ITERATIONS, size_t MEM, size_t MASK, bool SOFT_AES>
inline void cryptonight_double_hash(const void *__restrict__ input, size_t size, void *__restrict__ output, struct cryptonight_ctx *__restrict__ ctx)
{
keccak((const uint8_t *) input, (int) size, ctx->state0, 200);
keccak((const uint8_t *) input + size, (int) size, ctx->state1, 200);
const uint8_t* l0 = ctx->memory;
const uint8_t* l1 = ctx->memory + MEM;
uint64_t* h0 = reinterpret_cast<uint64_t*>(ctx->state0);
uint64_t* h1 = reinterpret_cast<uint64_t*>(ctx->state1);
cn_explode_scratchpad<MEM, SOFT_AES>((__m128i*) h0, (__m128i*) l0);
cn_explode_scratchpad<MEM, SOFT_AES>((__m128i*) h1, (__m128i*) l1);
uint64_t al0 = h0[0] ^ h0[4];
uint64_t al1 = h1[0] ^ h1[4];
uint64_t ah0 = h0[1] ^ h0[5];
uint64_t ah1 = h1[1] ^ h1[5];
__m128i bx0 = _mm_set_epi64x(h0[3] ^ h0[7], h0[2] ^ h0[6]);
__m128i bx1 = _mm_set_epi64x(h1[3] ^ h1[7], h1[2] ^ h1[6]);
uint64_t idx0 = h0[0] ^ h0[4];
uint64_t idx1 = h1[0] ^ h1[4];
for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
__m128i cx1 = _mm_load_si128((__m128i *) &l1[idx1 & MASK]);
if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1));
}
else {
cx0 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx0, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah0, al0);
cx1 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx1, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah1, al1);
}
_mm_store_si128((__m128i *) &l0[idx0 & MASK], _mm_xor_si128(bx0, cx0));
_mm_store_si128((__m128i *) &l1[idx1 & MASK], _mm_xor_si128(bx1, cx1));
idx0 = EXTRACT64(cx0);
idx1 = EXTRACT64(cx1);
bx0 = cx0;
bx1 = cx1;
uint64_t hi, lo, cl, ch;
cl = ((uint64_t*) &l0[idx0 & MASK])[0];
ch = ((uint64_t*) &l0[idx0 & MASK])[1];
lo = __umul128(idx0, cl, &hi);
al0 += hi;
ah0 += lo;
((uint64_t*) &l0[idx0 & MASK])[0] = al0;
((uint64_t*) &l0[idx0 & MASK])[1] = ah0;
ah0 ^= ch;
al0 ^= cl;
idx0 = al0;
cl = ((uint64_t*) &l1[idx1 & MASK])[0];
ch = ((uint64_t*) &l1[idx1 & MASK])[1];
lo = __umul128(idx1, cl, &hi);
al1 += hi;
ah1 += lo;
((uint64_t*) &l1[idx1 & MASK])[0] = al1;
((uint64_t*) &l1[idx1 & MASK])[1] = ah1;
ah1 ^= ch;
al1 ^= cl;
idx1 = al1;
}
cn_implode_scratchpad<MEM, SOFT_AES>((__m128i*) l0, (__m128i*) h0);
cn_implode_scratchpad<MEM, SOFT_AES>((__m128i*) l1, (__m128i*) h1);
keccakf(h0, 24);
keccakf(h1, 24);
extra_hashes[ctx->state0[0] & 3](ctx->state0, 200, static_cast<char*>(output));
extra_hashes[ctx->state1[0] & 3](ctx->state1, 200, static_cast<char*>(output) + 32);
}
#endif /* __CRYPTONIGHT_ARM64_H__ */

1497
src/crypto/SSE2NEON.h Normal file

File diff suppressed because it is too large Load diff

View file

@ -26,14 +26,18 @@
*/
#pragma once
#ifdef __GNUC__
#include <x86intrin.h>
#if defined(__aarch64__)
# include "crypto/SSE2NEON.h"
#elif defined(__GNUC__)
# include <x86intrin.h>
#else
#include <intrin.h>
#endif // __GNUC__
# 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),\
@ -109,7 +113,7 @@ static inline uint32_t sub_word(uint32_t key)
saes_sbox[key & 0xff];
}
#ifdef __clang__
#if defined(__clang__) || defined(XMRIG_ARM)
static inline uint32_t _rotr(uint32_t value, uint32_t amount)
{
return (value >> amount) | (value << ((32 - amount) & 31));