Merge pull request #1050 from SChernykh/dev

Integrated RandomX, added RandomXL (Loki)
This commit is contained in:
xmrig 2019-07-02 22:33:35 +07:00 committed by GitHub
commit 02410fa084
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
79 changed files with 9877 additions and 37 deletions

View file

@ -161,13 +161,48 @@ set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/cmake")
find_package(UV REQUIRED) find_package(UV REQUIRED)
if (WITH_RANDOMX) if (WITH_RANDOMX)
find_package(RandomX REQUIRED) include_directories(src/crypto/randomx)
include_directories(${RANDOMX_INCLUDE_DIR})
add_definitions(/DXMRIG_ALGO_RANDOMX) add_definitions(/DXMRIG_ALGO_RANDOMX)
set(SOURCES_CRYPTO
"${SOURCES_CRYPTO}"
src/crypto/randomx/aes_hash.cpp
src/crypto/randomx/argon2_ref.c
src/crypto/randomx/bytecode_machine.cpp
src/crypto/randomx/dataset.cpp
src/crypto/randomx/soft_aes.cpp
src/crypto/randomx/virtual_memory.cpp
src/crypto/randomx/vm_interpreted.cpp
src/crypto/randomx/allocator.cpp
src/crypto/randomx/randomx.cpp
src/crypto/randomx/superscalar.cpp
src/crypto/randomx/vm_compiled.cpp
src/crypto/randomx/vm_interpreted_light.cpp
src/crypto/randomx/argon2_core.c
src/crypto/randomx/blake2_generator.cpp
src/crypto/randomx/instructions_portable.cpp
src/crypto/randomx/reciprocal.c
src/crypto/randomx/virtual_machine.cpp
src/crypto/randomx/vm_compiled_light.cpp
src/crypto/randomx/blake2/blake2b.c
)
if (NOT ARCH_ID)
set(ARCH_ID ${CMAKE_HOST_SYSTEM_PROCESSOR})
endif()
if (CMAKE_C_COMPILER_ID MATCHES MSVC)
enable_language(ASM_MASM)
list(APPEND SOURCES_CRYPTO
src/crypto/randomx/jit_compiler_x86_static.asm
src/crypto/randomx/jit_compiler_x86.cpp
)
elseif (ARCH_ID STREQUAL "x86_64" OR ARCH_ID STREQUAL "x86-64" OR ARCH_ID STREQUAL "amd64" OR ARCH_ID STREQUAL "AMD64")
list(APPEND SOURCES_CRYPTO
src/crypto/randomx/jit_compiler_x86_static.S
src/crypto/randomx/jit_compiler_x86.cpp
)
# cheat because cmake and ccache hate each other
set_property(SOURCE src/crypto/randomx/jit_compiler_x86_static.S PROPERTY LANGUAGE C)
endif()
else() else()
set(RANDOMX_LIBRARIES "")
remove_definitions(/DXMRIG_ALGO_RANDOMX) remove_definitions(/DXMRIG_ALGO_RANDOMX)
endif() endif()
@ -241,4 +276,4 @@ if (WITH_DEBUG_LOG)
endif() endif()
add_executable(${CMAKE_PROJECT_NAME} ${HEADERS} ${SOURCES} ${SOURCES_OS} ${SOURCES_CPUID} ${HEADERS_CRYPTO} ${SOURCES_CRYPTO} ${SOURCES_SYSLOG} ${HTTP_SOURCES} ${TLS_SOURCES} ${XMRIG_ASM_SOURCES} ${CN_GPU_SOURCES}) add_executable(${CMAKE_PROJECT_NAME} ${HEADERS} ${SOURCES} ${SOURCES_OS} ${SOURCES_CPUID} ${HEADERS_CRYPTO} ${SOURCES_CRYPTO} ${SOURCES_SYSLOG} ${HTTP_SOURCES} ${TLS_SOURCES} ${XMRIG_ASM_SOURCES} ${CN_GPU_SOURCES})
target_link_libraries(${CMAKE_PROJECT_NAME} ${XMRIG_ASM_LIBRARY} ${OPENSSL_LIBRARIES} ${UV_LIBRARIES} ${RANDOMX_LIBRARIES} ${EXTRA_LIBS} ${CPUID_LIB}) target_link_libraries(${CMAKE_PROJECT_NAME} ${XMRIG_ASM_LIBRARY} ${OPENSSL_LIBRARIES} ${UV_LIBRARIES} ${EXTRA_LIBS} ${CPUID_LIB})

View file

@ -1,25 +0,0 @@
find_path(
RANDOMX_INCLUDE_DIR
NAMES randomx.h
PATHS "${XMRIG_DEPS}" ENV "XMRIG_DEPS"
PATH_SUFFIXES "include"
NO_DEFAULT_PATH
)
find_path(RANDOMX_INCLUDE_DIR NAMES randomx.h)
find_library(
RANDOMX_LIBRARY
NAMES librandomx.a randomx librandomx
PATHS "${XMRIG_DEPS}" ENV "XMRIG_DEPS"
PATH_SUFFIXES "lib"
NO_DEFAULT_PATH
)
find_library(RANDOMX_LIBRARY NAMES librandomx.a randomx librandomx)
set(RANDOMX_LIBRARIES ${RANDOMX_LIBRARY})
set(RANDOMX_INCLUDE_DIRS ${RANDOMX_INCLUDE_DIR})
include(FindPackageHandleStandardArgs)
find_package_handle_standard_args(RANDOMX DEFAULT_MSG RANDOMX_LIBRARY RANDOMX_INCLUDE_DIR)

View file

@ -17,7 +17,7 @@ if (CMAKE_CXX_COMPILER_ID MATCHES GNU)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wno-strict-aliasing") set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wno-strict-aliasing")
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Ofast") set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Ofast")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -fno-exceptions -fno-rtti -Wno-class-memaccess") set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -fexceptions -fno-rtti -Wno-class-memaccess")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -s") set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -s")
if (XMRIG_ARMv8) if (XMRIG_ARMv8)

View file

@ -518,6 +518,7 @@ void xmrig::Pool::rebuild()
addVariant(VARIANT_ZLS); addVariant(VARIANT_ZLS);
addVariant(VARIANT_DOUBLE); addVariant(VARIANT_DOUBLE);
addVariant(VARIANT_RX_WOW); addVariant(VARIANT_RX_WOW);
addVariant(VARIANT_RX_LOKI);
addVariant(VARIANT_AUTO); addVariant(VARIANT_AUTO);
# endif # endif
} }

View file

@ -81,6 +81,7 @@ enum Variant {
VARIANT_ZLS = 15, // CryptoNight variant 2 with 3/4 iterations (Zelerius) VARIANT_ZLS = 15, // CryptoNight variant 2 with 3/4 iterations (Zelerius)
VARIANT_DOUBLE = 16, // CryptoNight variant 2 with double iterations (X-CASH) VARIANT_DOUBLE = 16, // CryptoNight variant 2 with double iterations (X-CASH)
VARIANT_RX_WOW = 17, // RandomX (Wownero) VARIANT_RX_WOW = 17, // RandomX (Wownero)
VARIANT_RX_LOKI = 18, // RandomX (Loki)
VARIANT_MAX VARIANT_MAX
}; };

View file

@ -72,6 +72,7 @@ static AlgoData const algorithms[] = {
# ifdef XMRIG_ALGO_RANDOMX # ifdef XMRIG_ALGO_RANDOMX
{ "randomx/wow", "rx/wow", xmrig::RANDOM_X, xmrig::VARIANT_RX_WOW }, { "randomx/wow", "rx/wow", xmrig::RANDOM_X, xmrig::VARIANT_RX_WOW },
{ "randomx/loki", "rx/loki", xmrig::RANDOM_X, xmrig::VARIANT_RX_LOKI },
{ "randomx", "rx", xmrig::RANDOM_X, xmrig::VARIANT_RX_WOW }, { "randomx", "rx", xmrig::RANDOM_X, xmrig::VARIANT_RX_WOW },
# endif # endif
@ -145,6 +146,7 @@ static const char *variants[] = {
"zls", "zls",
"double", "double",
"rx/wow", "rx/wow",
"rx/loki",
}; };

View file

@ -0,0 +1,214 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "soft_aes.h"
#include "randomx.h"
#define AES_HASH_1R_STATE0 0xd7983aad, 0xcc82db47, 0x9fa856de, 0x92b52c0d
#define AES_HASH_1R_STATE1 0xace78057, 0xf59e125a, 0x15c7b798, 0x338d996e
#define AES_HASH_1R_STATE2 0xe8a07ce4, 0x5079506b, 0xae62c7d0, 0x6a770017
#define AES_HASH_1R_STATE3 0x7e994948, 0x79a10005, 0x07ad828d, 0x630a240c
#define AES_HASH_1R_XKEY0 0x06890201, 0x90dc56bf, 0x8b24949f, 0xf6fa8389
#define AES_HASH_1R_XKEY1 0xed18f99b, 0xee1043c6, 0x51f4e03c, 0x61b263d1
/*
Calculate a 512-bit hash of 'input' using 4 lanes of AES.
The input is treated as a set of round keys for the encryption
of the initial state.
'inputSize' must be a multiple of 64.
For a 2 MiB input, this has the same security as 32768-round
AES encryption.
Hashing throughput: >20 GiB/s per CPU core with hardware AES
*/
template<bool softAes>
void hashAes1Rx4(const void *input, size_t inputSize, void *hash) {
const uint8_t* inptr = (uint8_t*)input;
const uint8_t* inputEnd = inptr + inputSize;
rx_vec_i128 state0, state1, state2, state3;
rx_vec_i128 in0, in1, in2, in3;
//intial state
state0 = rx_set_int_vec_i128(AES_HASH_1R_STATE0);
state1 = rx_set_int_vec_i128(AES_HASH_1R_STATE1);
state2 = rx_set_int_vec_i128(AES_HASH_1R_STATE2);
state3 = rx_set_int_vec_i128(AES_HASH_1R_STATE3);
//process 64 bytes at a time in 4 lanes
while (inptr < inputEnd) {
in0 = rx_load_vec_i128((rx_vec_i128*)inptr + 0);
in1 = rx_load_vec_i128((rx_vec_i128*)inptr + 1);
in2 = rx_load_vec_i128((rx_vec_i128*)inptr + 2);
in3 = rx_load_vec_i128((rx_vec_i128*)inptr + 3);
state0 = aesenc<softAes>(state0, in0);
state1 = aesdec<softAes>(state1, in1);
state2 = aesenc<softAes>(state2, in2);
state3 = aesdec<softAes>(state3, in3);
inptr += 64;
}
//two extra rounds to achieve full diffusion
rx_vec_i128 xkey0 = rx_set_int_vec_i128(AES_HASH_1R_XKEY0);
rx_vec_i128 xkey1 = rx_set_int_vec_i128(AES_HASH_1R_XKEY1);
state0 = aesenc<softAes>(state0, xkey0);
state1 = aesdec<softAes>(state1, xkey0);
state2 = aesenc<softAes>(state2, xkey0);
state3 = aesdec<softAes>(state3, xkey0);
state0 = aesenc<softAes>(state0, xkey1);
state1 = aesdec<softAes>(state1, xkey1);
state2 = aesenc<softAes>(state2, xkey1);
state3 = aesdec<softAes>(state3, xkey1);
//output hash
rx_store_vec_i128((rx_vec_i128*)hash + 0, state0);
rx_store_vec_i128((rx_vec_i128*)hash + 1, state1);
rx_store_vec_i128((rx_vec_i128*)hash + 2, state2);
rx_store_vec_i128((rx_vec_i128*)hash + 3, state3);
}
template void hashAes1Rx4<false>(const void *input, size_t inputSize, void *hash);
template void hashAes1Rx4<true>(const void *input, size_t inputSize, void *hash);
#define AES_GEN_1R_KEY0 0xb4f44917, 0xdbb5552b, 0x62716609, 0x6daca553
#define AES_GEN_1R_KEY1 0x0da1dc4e, 0x1725d378, 0x846a710d, 0x6d7caf07
#define AES_GEN_1R_KEY2 0x3e20e345, 0xf4c0794f, 0x9f947ec6, 0x3f1262f1
#define AES_GEN_1R_KEY3 0x49169154, 0x16314c88, 0xb1ba317c, 0x6aef8135
/*
Fill 'buffer' with pseudorandom data based on 512-bit 'state'.
The state is encrypted using a single AES round per 16 bytes of output
in 4 lanes.
'outputSize' must be a multiple of 64.
The modified state is written back to 'state' to allow multiple
calls to this function.
*/
template<bool softAes>
void fillAes1Rx4(void *state, size_t outputSize, void *buffer) {
const uint8_t* outptr = (uint8_t*)buffer;
const uint8_t* outputEnd = outptr + outputSize;
rx_vec_i128 state0, state1, state2, state3;
rx_vec_i128 key0, key1, key2, key3;
key0 = rx_set_int_vec_i128(AES_GEN_1R_KEY0);
key1 = rx_set_int_vec_i128(AES_GEN_1R_KEY1);
key2 = rx_set_int_vec_i128(AES_GEN_1R_KEY2);
key3 = rx_set_int_vec_i128(AES_GEN_1R_KEY3);
state0 = rx_load_vec_i128((rx_vec_i128*)state + 0);
state1 = rx_load_vec_i128((rx_vec_i128*)state + 1);
state2 = rx_load_vec_i128((rx_vec_i128*)state + 2);
state3 = rx_load_vec_i128((rx_vec_i128*)state + 3);
while (outptr < outputEnd) {
state0 = aesdec<softAes>(state0, key0);
state1 = aesenc<softAes>(state1, key1);
state2 = aesdec<softAes>(state2, key2);
state3 = aesenc<softAes>(state3, key3);
rx_store_vec_i128((rx_vec_i128*)outptr + 0, state0);
rx_store_vec_i128((rx_vec_i128*)outptr + 1, state1);
rx_store_vec_i128((rx_vec_i128*)outptr + 2, state2);
rx_store_vec_i128((rx_vec_i128*)outptr + 3, state3);
outptr += 64;
}
rx_store_vec_i128((rx_vec_i128*)state + 0, state0);
rx_store_vec_i128((rx_vec_i128*)state + 1, state1);
rx_store_vec_i128((rx_vec_i128*)state + 2, state2);
rx_store_vec_i128((rx_vec_i128*)state + 3, state3);
}
template void fillAes1Rx4<true>(void *state, size_t outputSize, void *buffer);
template void fillAes1Rx4<false>(void *state, size_t outputSize, void *buffer);
template<bool softAes>
void fillAes4Rx4(void *state, size_t outputSize, void *buffer) {
const uint8_t* outptr = (uint8_t*)buffer;
const uint8_t* outputEnd = outptr + outputSize;
rx_vec_i128 state0, state1, state2, state3;
rx_vec_i128 key0, key1, key2, key3, key4, key5, key6, key7;
key0 = RandomX_CurrentConfig.fillAes4Rx4_Key[0];
key1 = RandomX_CurrentConfig.fillAes4Rx4_Key[1];
key2 = RandomX_CurrentConfig.fillAes4Rx4_Key[2];
key3 = RandomX_CurrentConfig.fillAes4Rx4_Key[3];
key4 = RandomX_CurrentConfig.fillAes4Rx4_Key[4];
key5 = RandomX_CurrentConfig.fillAes4Rx4_Key[5];
key6 = RandomX_CurrentConfig.fillAes4Rx4_Key[6];
key7 = RandomX_CurrentConfig.fillAes4Rx4_Key[7];
state0 = rx_load_vec_i128((rx_vec_i128*)state + 0);
state1 = rx_load_vec_i128((rx_vec_i128*)state + 1);
state2 = rx_load_vec_i128((rx_vec_i128*)state + 2);
state3 = rx_load_vec_i128((rx_vec_i128*)state + 3);
while (outptr < outputEnd) {
state0 = aesdec<softAes>(state0, key0);
state1 = aesenc<softAes>(state1, key0);
state2 = aesdec<softAes>(state2, key4);
state3 = aesenc<softAes>(state3, key4);
state0 = aesdec<softAes>(state0, key1);
state1 = aesenc<softAes>(state1, key1);
state2 = aesdec<softAes>(state2, key5);
state3 = aesenc<softAes>(state3, key5);
state0 = aesdec<softAes>(state0, key2);
state1 = aesenc<softAes>(state1, key2);
state2 = aesdec<softAes>(state2, key6);
state3 = aesenc<softAes>(state3, key6);
state0 = aesdec<softAes>(state0, key3);
state1 = aesenc<softAes>(state1, key3);
state2 = aesdec<softAes>(state2, key7);
state3 = aesenc<softAes>(state3, key7);
rx_store_vec_i128((rx_vec_i128*)outptr + 0, state0);
rx_store_vec_i128((rx_vec_i128*)outptr + 1, state1);
rx_store_vec_i128((rx_vec_i128*)outptr + 2, state2);
rx_store_vec_i128((rx_vec_i128*)outptr + 3, state3);
outptr += 64;
}
}
template void fillAes4Rx4<true>(void *state, size_t outputSize, void *buffer);
template void fillAes4Rx4<false>(void *state, size_t outputSize, void *buffer);

View file

@ -0,0 +1,40 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstddef>
template<bool softAes>
void hashAes1Rx4(const void *input, size_t inputSize, void *hash);
template<bool softAes>
void fillAes1Rx4(void *state, size_t outputSize, void *buffer);
template<bool softAes>
void fillAes4Rx4(void *state, size_t outputSize, void *buffer);

View file

@ -0,0 +1,60 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <new>
#include "allocator.hpp"
#include "intrin_portable.h"
#include "virtual_memory.hpp"
#include "common.hpp"
namespace randomx {
template<size_t alignment>
void* AlignedAllocator<alignment>::allocMemory(size_t count) {
void *mem = rx_aligned_alloc(count, alignment);
if (mem == nullptr)
throw std::bad_alloc();
return mem;
}
template<size_t alignment>
void AlignedAllocator<alignment>::freeMemory(void* ptr, size_t count) {
rx_aligned_free(ptr);
}
template class AlignedAllocator<CacheLineSize>;
void* LargePageAllocator::allocMemory(size_t count) {
return allocLargePagesMemory(count);
}
void LargePageAllocator::freeMemory(void* ptr, size_t count) {
freePagedMemory(ptr, count);
};
}

View file

@ -0,0 +1,46 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstddef>
namespace randomx {
template<size_t alignment>
struct AlignedAllocator {
static void* allocMemory(size_t);
static void freeMemory(void*, size_t);
};
struct LargePageAllocator {
static void* allocMemory(size_t);
static void freeMemory(void*, size_t);
};
}

229
src/crypto/randomx/argon2.h Normal file
View file

@ -0,0 +1,229 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
#pragma once
#include <stdint.h>
#include <stddef.h>
#include <limits.h>
/*
* Argon2 input parameter restrictions
*/
/* Minimum and maximum number of lanes (degree of parallelism) */
#define ARGON2_MIN_LANES UINT32_C(1)
#define ARGON2_MAX_LANES UINT32_C(0xFFFFFF)
/* Minimum and maximum number of threads */
#define ARGON2_MIN_THREADS UINT32_C(1)
#define ARGON2_MAX_THREADS UINT32_C(0xFFFFFF)
/* Number of synchronization points between lanes per pass */
#define ARGON2_SYNC_POINTS UINT32_C(4)
/* Minimum and maximum digest size in bytes */
#define ARGON2_MIN_OUTLEN UINT32_C(4)
#define ARGON2_MAX_OUTLEN UINT32_C(0xFFFFFFFF)
/* Minimum and maximum number of memory blocks (each of BLOCK_SIZE bytes) */
#define ARGON2_MIN_MEMORY (2 * ARGON2_SYNC_POINTS) /* 2 blocks per slice */
#define ARGON2_MIN(a, b) ((a) < (b) ? (a) : (b))
/* Max memory size is addressing-space/2, topping at 2^32 blocks (4 TB) */
#define ARGON2_MAX_MEMORY_BITS \
ARGON2_MIN(UINT32_C(32), (sizeof(void *) * CHAR_BIT - 10 - 1))
#define ARGON2_MAX_MEMORY \
ARGON2_MIN(UINT32_C(0xFFFFFFFF), UINT64_C(1) << ARGON2_MAX_MEMORY_BITS)
/* Minimum and maximum number of passes */
#define ARGON2_MIN_TIME UINT32_C(1)
#define ARGON2_MAX_TIME UINT32_C(0xFFFFFFFF)
/* Minimum and maximum password length in bytes */
#define ARGON2_MIN_PWD_LENGTH UINT32_C(0)
#define ARGON2_MAX_PWD_LENGTH UINT32_C(0xFFFFFFFF)
/* Minimum and maximum associated data length in bytes */
#define ARGON2_MIN_AD_LENGTH UINT32_C(0)
#define ARGON2_MAX_AD_LENGTH UINT32_C(0xFFFFFFFF)
/* Minimum and maximum salt length in bytes */
#define ARGON2_MIN_SALT_LENGTH UINT32_C(8)
#define ARGON2_MAX_SALT_LENGTH UINT32_C(0xFFFFFFFF)
/* Minimum and maximum key length in bytes */
#define ARGON2_MIN_SECRET UINT32_C(0)
#define ARGON2_MAX_SECRET UINT32_C(0xFFFFFFFF)
/* Flags to determine which fields are securely wiped (default = no wipe). */
#define ARGON2_DEFAULT_FLAGS UINT32_C(0)
#define ARGON2_FLAG_CLEAR_PASSWORD (UINT32_C(1) << 0)
#define ARGON2_FLAG_CLEAR_SECRET (UINT32_C(1) << 1)
/* Error codes */
typedef enum Argon2_ErrorCodes {
ARGON2_OK = 0,
ARGON2_OUTPUT_PTR_NULL = -1,
ARGON2_OUTPUT_TOO_SHORT = -2,
ARGON2_OUTPUT_TOO_LONG = -3,
ARGON2_PWD_TOO_SHORT = -4,
ARGON2_PWD_TOO_LONG = -5,
ARGON2_SALT_TOO_SHORT = -6,
ARGON2_SALT_TOO_LONG = -7,
ARGON2_AD_TOO_SHORT = -8,
ARGON2_AD_TOO_LONG = -9,
ARGON2_SECRET_TOO_SHORT = -10,
ARGON2_SECRET_TOO_LONG = -11,
ARGON2_TIME_TOO_SMALL = -12,
ARGON2_TIME_TOO_LARGE = -13,
ARGON2_MEMORY_TOO_LITTLE = -14,
ARGON2_MEMORY_TOO_MUCH = -15,
ARGON2_LANES_TOO_FEW = -16,
ARGON2_LANES_TOO_MANY = -17,
ARGON2_PWD_PTR_MISMATCH = -18, /* NULL ptr with non-zero length */
ARGON2_SALT_PTR_MISMATCH = -19, /* NULL ptr with non-zero length */
ARGON2_SECRET_PTR_MISMATCH = -20, /* NULL ptr with non-zero length */
ARGON2_AD_PTR_MISMATCH = -21, /* NULL ptr with non-zero length */
ARGON2_MEMORY_ALLOCATION_ERROR = -22,
ARGON2_FREE_MEMORY_CBK_NULL = -23,
ARGON2_ALLOCATE_MEMORY_CBK_NULL = -24,
ARGON2_INCORRECT_PARAMETER = -25,
ARGON2_INCORRECT_TYPE = -26,
ARGON2_OUT_PTR_MISMATCH = -27,
ARGON2_THREADS_TOO_FEW = -28,
ARGON2_THREADS_TOO_MANY = -29,
ARGON2_MISSING_ARGS = -30,
ARGON2_ENCODING_FAIL = -31,
ARGON2_DECODING_FAIL = -32,
ARGON2_THREAD_FAIL = -33,
ARGON2_DECODING_LENGTH_FAIL = -34,
ARGON2_VERIFY_MISMATCH = -35
} argon2_error_codes;
/* Memory allocator types --- for external allocation */
typedef int(*allocate_fptr)(uint8_t **memory, size_t bytes_to_allocate);
typedef void(*deallocate_fptr)(uint8_t *memory, size_t bytes_to_allocate);
/* Argon2 external data structures */
/*
*****
* Context: structure to hold Argon2 inputs:
* output array and its length,
* password and its length,
* salt and its length,
* secret and its length,
* associated data and its length,
* number of passes, amount of used memory (in KBytes, can be rounded up a bit)
* number of parallel threads that will be run.
* All the parameters above affect the output hash value.
* Additionally, two function pointers can be provided to allocate and
* deallocate the memory (if NULL, memory will be allocated internally).
* Also, three flags indicate whether to erase password, secret as soon as they
* are pre-hashed (and thus not needed anymore), and the entire memory
*****
* Simplest situation: you have output array out[8], password is stored in
* pwd[32], salt is stored in salt[16], you do not have keys nor associated
* data. You need to spend 1 GB of RAM and you run 5 passes of Argon2d with
* 4 parallel lanes.
* You want to erase the password, but you're OK with last pass not being
* erased. You want to use the default memory allocator.
* Then you initialize:
Argon2_Context(out,8,pwd,32,salt,16,NULL,0,NULL,0,5,1<<20,4,4,NULL,NULL,true,false,false,false)
*/
typedef struct Argon2_Context {
uint8_t *out; /* output array */
uint32_t outlen; /* digest length */
uint8_t *pwd; /* password array */
uint32_t pwdlen; /* password length */
uint8_t *salt; /* salt array */
uint32_t saltlen; /* salt length */
uint8_t *secret; /* key array */
uint32_t secretlen; /* key length */
uint8_t *ad; /* associated data array */
uint32_t adlen; /* associated data length */
uint32_t t_cost; /* number of passes */
uint32_t m_cost; /* amount of memory requested (KB) */
uint32_t lanes; /* number of lanes */
uint32_t threads; /* maximum number of threads */
uint32_t version; /* version number */
allocate_fptr allocate_cbk; /* pointer to memory allocator */
deallocate_fptr free_cbk; /* pointer to memory deallocator */
uint32_t flags; /* array of bool options */
} argon2_context;
/* Argon2 primitive type */
typedef enum Argon2_type {
Argon2_d = 0,
Argon2_i = 1,
Argon2_id = 2
} argon2_type;
/* Version of the algorithm */
typedef enum Argon2_version {
ARGON2_VERSION_10 = 0x10,
ARGON2_VERSION_13 = 0x13,
ARGON2_VERSION_NUMBER = ARGON2_VERSION_13
} argon2_version;

View file

@ -0,0 +1,516 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
/*For memory wiping*/
#ifdef _MSC_VER
#include <windows.h>
#include <winbase.h> /* For SecureZeroMemory */
#endif
#if defined __STDC_LIB_EXT1__
#define __STDC_WANT_LIB_EXT1__ 1
#endif
#define VC_GE_2005(version) (version >= 1400)
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "argon2_core.h"
#include "blake2/blake2.h"
#include "blake2/blake2-impl.h"
#ifdef GENKAT
#include "genkat.h"
#endif
#if defined(__clang__)
#if __has_attribute(optnone)
#define NOT_OPTIMIZED __attribute__((optnone))
#endif
#elif defined(__GNUC__)
#define GCC_VERSION \
(__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)
#if GCC_VERSION >= 40400
#define NOT_OPTIMIZED __attribute__((optimize("O0")))
#endif
#endif
#ifndef NOT_OPTIMIZED
#define NOT_OPTIMIZED
#endif
/***************Instance and Position constructors**********/
void rxa2_init_block_value(block *b, uint8_t in) { memset(b->v, in, sizeof(b->v)); }
void rxa2_copy_block(block *dst, const block *src) {
memcpy(dst->v, src->v, sizeof(uint64_t) * ARGON2_QWORDS_IN_BLOCK);
}
void rxa2_xor_block(block *dst, const block *src) {
int i;
for (i = 0; i < ARGON2_QWORDS_IN_BLOCK; ++i) {
dst->v[i] ^= src->v[i];
}
}
static void load_block(block *dst, const void *input) {
unsigned i;
for (i = 0; i < ARGON2_QWORDS_IN_BLOCK; ++i) {
dst->v[i] = load64((const uint8_t *)input + i * sizeof(dst->v[i]));
}
}
static void store_block(void *output, const block *src) {
unsigned i;
for (i = 0; i < ARGON2_QWORDS_IN_BLOCK; ++i) {
store64((uint8_t *)output + i * sizeof(src->v[i]), src->v[i]);
}
}
/***************Memory functions*****************/
int rxa2_allocate_memory(const argon2_context *context, uint8_t **memory,
size_t num, size_t size) {
size_t memory_size = num * size;
if (memory == NULL) {
return ARGON2_MEMORY_ALLOCATION_ERROR;
}
/* 1. Check for multiplication overflow */
if (size != 0 && memory_size / size != num) {
return ARGON2_MEMORY_ALLOCATION_ERROR;
}
/* 2. Try to allocate with appropriate allocator */
if (context->allocate_cbk) {
(context->allocate_cbk)(memory, memory_size);
}
else {
*memory = (uint8_t*)malloc(memory_size);
}
if (*memory == NULL) {
return ARGON2_MEMORY_ALLOCATION_ERROR;
}
return ARGON2_OK;
}
void rxa2_free_memory(const argon2_context *context, uint8_t *memory,
size_t num, size_t size) {
size_t memory_size = num * size;
rxa2_clear_internal_memory(memory, memory_size);
if (context->free_cbk) {
(context->free_cbk)(memory, memory_size);
}
else {
free(memory);
}
}
void NOT_OPTIMIZED rxa2_secure_wipe_memory(void *v, size_t n) {
#if defined(_MSC_VER) && VC_GE_2005(_MSC_VER)
SecureZeroMemory(v, n);
#elif defined memset_s
memset_s(v, n, 0, n);
#elif defined(__OpenBSD__)
explicit_bzero(v, n);
#else
static void *(*const volatile memset_sec)(void *, int, size_t) = &memset;
memset_sec(v, 0, n);
#endif
}
/* Memory clear flag defaults to true. */
#define FLAG_clear_internal_memory 0
void rxa2_clear_internal_memory(void *v, size_t n) {
if (FLAG_clear_internal_memory && v) {
rxa2_secure_wipe_memory(v, n);
}
}
uint32_t rxa2_index_alpha(const argon2_instance_t *instance,
const argon2_position_t *position, uint32_t pseudo_rand,
int same_lane) {
/*
* Pass 0:
* This lane : all already finished segments plus already constructed
* blocks in this segment
* Other lanes : all already finished segments
* Pass 1+:
* This lane : (SYNC_POINTS - 1) last segments plus already constructed
* blocks in this segment
* Other lanes : (SYNC_POINTS - 1) last segments
*/
uint32_t reference_area_size;
uint64_t relative_position;
uint32_t start_position, absolute_position;
if (0 == position->pass) {
/* First pass */
if (0 == position->slice) {
/* First slice */
reference_area_size =
position->index - 1; /* all but the previous */
}
else {
if (same_lane) {
/* The same lane => add current segment */
reference_area_size =
position->slice * instance->segment_length +
position->index - 1;
}
else {
reference_area_size =
position->slice * instance->segment_length +
((position->index == 0) ? (-1) : 0);
}
}
}
else {
/* Second pass */
if (same_lane) {
reference_area_size = instance->lane_length -
instance->segment_length + position->index -
1;
}
else {
reference_area_size = instance->lane_length -
instance->segment_length +
((position->index == 0) ? (-1) : 0);
}
}
/* 1.2.4. Mapping pseudo_rand to 0..<reference_area_size-1> and produce
* relative position */
relative_position = pseudo_rand;
relative_position = relative_position * relative_position >> 32;
relative_position = reference_area_size - 1 -
(reference_area_size * relative_position >> 32);
/* 1.2.5 Computing starting position */
start_position = 0;
if (0 != position->pass) {
start_position = (position->slice == ARGON2_SYNC_POINTS - 1)
? 0
: (position->slice + 1) * instance->segment_length;
}
/* 1.2.6. Computing absolute position */
absolute_position = (start_position + relative_position) %
instance->lane_length; /* absolute position */
return absolute_position;
}
/* Single-threaded version for p=1 case */
static int fill_memory_blocks_st(argon2_instance_t *instance) {
uint32_t r, s, l;
for (r = 0; r < instance->passes; ++r) {
for (s = 0; s < ARGON2_SYNC_POINTS; ++s) {
for (l = 0; l < instance->lanes; ++l) {
argon2_position_t position = { r, l, (uint8_t)s, 0 };
rxa2_fill_segment(instance, position);
}
}
#ifdef GENKAT
internal_kat(instance, r); /* Print all memory blocks */
#endif
}
return ARGON2_OK;
}
int rxa2_fill_memory_blocks(argon2_instance_t *instance) {
if (instance == NULL || instance->lanes == 0) {
return ARGON2_INCORRECT_PARAMETER;
}
return fill_memory_blocks_st(instance);
}
int rxa2_validate_inputs(const argon2_context *context) {
if (NULL == context) {
return ARGON2_INCORRECT_PARAMETER;
}
if (NULL == context->out) {
return ARGON2_OUTPUT_PTR_NULL;
}
/* Validate output length */
if (ARGON2_MIN_OUTLEN > context->outlen) {
return ARGON2_OUTPUT_TOO_SHORT;
}
if (ARGON2_MAX_OUTLEN < context->outlen) {
return ARGON2_OUTPUT_TOO_LONG;
}
/* Validate password (required param) */
if (NULL == context->pwd) {
if (0 != context->pwdlen) {
return ARGON2_PWD_PTR_MISMATCH;
}
}
if (ARGON2_MIN_PWD_LENGTH > context->pwdlen) {
return ARGON2_PWD_TOO_SHORT;
}
if (ARGON2_MAX_PWD_LENGTH < context->pwdlen) {
return ARGON2_PWD_TOO_LONG;
}
/* Validate salt (required param) */
if (NULL == context->salt) {
if (0 != context->saltlen) {
return ARGON2_SALT_PTR_MISMATCH;
}
}
if (ARGON2_MIN_SALT_LENGTH > context->saltlen) {
return ARGON2_SALT_TOO_SHORT;
}
if (ARGON2_MAX_SALT_LENGTH < context->saltlen) {
return ARGON2_SALT_TOO_LONG;
}
/* Validate secret (optional param) */
if (NULL == context->secret) {
if (0 != context->secretlen) {
return ARGON2_SECRET_PTR_MISMATCH;
}
}
else {
if (ARGON2_MIN_SECRET > context->secretlen) {
return ARGON2_SECRET_TOO_SHORT;
}
if (ARGON2_MAX_SECRET < context->secretlen) {
return ARGON2_SECRET_TOO_LONG;
}
}
/* Validate associated data (optional param) */
if (NULL == context->ad) {
if (0 != context->adlen) {
return ARGON2_AD_PTR_MISMATCH;
}
}
else {
if (ARGON2_MIN_AD_LENGTH > context->adlen) {
return ARGON2_AD_TOO_SHORT;
}
if (ARGON2_MAX_AD_LENGTH < context->adlen) {
return ARGON2_AD_TOO_LONG;
}
}
/* Validate memory cost */
if (ARGON2_MIN_MEMORY > context->m_cost) {
return ARGON2_MEMORY_TOO_LITTLE;
}
if (ARGON2_MAX_MEMORY < context->m_cost) {
return ARGON2_MEMORY_TOO_MUCH;
}
if (context->m_cost < 8 * context->lanes) {
return ARGON2_MEMORY_TOO_LITTLE;
}
/* Validate time cost */
if (ARGON2_MIN_TIME > context->t_cost) {
return ARGON2_TIME_TOO_SMALL;
}
if (ARGON2_MAX_TIME < context->t_cost) {
return ARGON2_TIME_TOO_LARGE;
}
/* Validate lanes */
if (ARGON2_MIN_LANES > context->lanes) {
return ARGON2_LANES_TOO_FEW;
}
if (ARGON2_MAX_LANES < context->lanes) {
return ARGON2_LANES_TOO_MANY;
}
/* Validate threads */
if (ARGON2_MIN_THREADS > context->threads) {
return ARGON2_THREADS_TOO_FEW;
}
if (ARGON2_MAX_THREADS < context->threads) {
return ARGON2_THREADS_TOO_MANY;
}
if (NULL != context->allocate_cbk && NULL == context->free_cbk) {
return ARGON2_FREE_MEMORY_CBK_NULL;
}
if (NULL == context->allocate_cbk && NULL != context->free_cbk) {
return ARGON2_ALLOCATE_MEMORY_CBK_NULL;
}
return ARGON2_OK;
}
void rxa2_fill_first_blocks(uint8_t *blockhash, const argon2_instance_t *instance) {
uint32_t l;
/* Make the first and second block in each lane as G(H0||0||i) or
G(H0||1||i) */
uint8_t blockhash_bytes[ARGON2_BLOCK_SIZE];
for (l = 0; l < instance->lanes; ++l) {
store32(blockhash + ARGON2_PREHASH_DIGEST_LENGTH, 0);
store32(blockhash + ARGON2_PREHASH_DIGEST_LENGTH + 4, l);
rxa2_blake2b_long(blockhash_bytes, ARGON2_BLOCK_SIZE, blockhash,
ARGON2_PREHASH_SEED_LENGTH);
load_block(&instance->memory[l * instance->lane_length + 0],
blockhash_bytes);
store32(blockhash + ARGON2_PREHASH_DIGEST_LENGTH, 1);
rxa2_blake2b_long(blockhash_bytes, ARGON2_BLOCK_SIZE, blockhash,
ARGON2_PREHASH_SEED_LENGTH);
load_block(&instance->memory[l * instance->lane_length + 1],
blockhash_bytes);
}
rxa2_clear_internal_memory(blockhash_bytes, ARGON2_BLOCK_SIZE);
}
void rxa2_initial_hash(uint8_t *blockhash, argon2_context *context, argon2_type type) {
blake2b_state BlakeHash;
uint8_t value[sizeof(uint32_t)];
if (NULL == context || NULL == blockhash) {
return;
}
blake2b_init(&BlakeHash, ARGON2_PREHASH_DIGEST_LENGTH);
store32(&value, context->lanes);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
store32(&value, context->outlen);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
store32(&value, context->m_cost);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
store32(&value, context->t_cost);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
store32(&value, context->version);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
store32(&value, (uint32_t)type);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
store32(&value, context->pwdlen);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
if (context->pwd != NULL) {
blake2b_update(&BlakeHash, (const uint8_t *)context->pwd,
context->pwdlen);
if (context->flags & ARGON2_FLAG_CLEAR_PASSWORD) {
rxa2_secure_wipe_memory(context->pwd, context->pwdlen);
context->pwdlen = 0;
}
}
store32(&value, context->saltlen);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
if (context->salt != NULL) {
blake2b_update(&BlakeHash, (const uint8_t *)context->salt, context->saltlen);
}
store32(&value, context->secretlen);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
if (context->secret != NULL) {
blake2b_update(&BlakeHash, (const uint8_t *)context->secret,
context->secretlen);
if (context->flags & ARGON2_FLAG_CLEAR_SECRET) {
rxa2_secure_wipe_memory(context->secret, context->secretlen);
context->secretlen = 0;
}
}
store32(&value, context->adlen);
blake2b_update(&BlakeHash, (const uint8_t *)&value, sizeof(value));
if (context->ad != NULL) {
blake2b_update(&BlakeHash, (const uint8_t *)context->ad,
context->adlen);
}
blake2b_final(&BlakeHash, blockhash, ARGON2_PREHASH_DIGEST_LENGTH);
}
int rxa2_argon_initialize(argon2_instance_t *instance, argon2_context *context) {
uint8_t blockhash[ARGON2_PREHASH_SEED_LENGTH];
int result = ARGON2_OK;
if (instance == NULL || context == NULL)
return ARGON2_INCORRECT_PARAMETER;
instance->context_ptr = context;
/* 1. Memory allocation */
/*result = allocate_memory(context, (uint8_t **)&(instance->memory), instance->memory_blocks, sizeof(block));
if (result != ARGON2_OK) {
return result;
}*/
/* 2. Initial hashing */
/* H_0 + 8 extra bytes to produce the first blocks */
/* uint8_t blockhash[ARGON2_PREHASH_SEED_LENGTH]; */
/* Hashing all inputs */
rxa2_initial_hash(blockhash, context, instance->type);
/* Zeroing 8 extra bytes */
rxa2_clear_internal_memory(blockhash + ARGON2_PREHASH_DIGEST_LENGTH,
ARGON2_PREHASH_SEED_LENGTH -
ARGON2_PREHASH_DIGEST_LENGTH);
/* 3. Creating first blocks, we always have at least two blocks in a slice
*/
rxa2_fill_first_blocks(blockhash, instance);
/* Clearing the hash */
rxa2_clear_internal_memory(blockhash, ARGON2_PREHASH_SEED_LENGTH);
return ARGON2_OK;
}

View file

@ -0,0 +1,254 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
#ifndef ARGON2_CORE_H
#define ARGON2_CORE_H
#include <stdint.h>
#include "argon2.h"
#if defined(__cplusplus)
extern "C" {
#endif
#define CONST_CAST(x) (x)(uintptr_t)
/**********************Argon2 internal constants*******************************/
enum argon2_core_constants {
/* Memory block size in bytes */
ARGON2_BLOCK_SIZE = 1024,
ARGON2_QWORDS_IN_BLOCK = ARGON2_BLOCK_SIZE / 8,
ARGON2_OWORDS_IN_BLOCK = ARGON2_BLOCK_SIZE / 16,
ARGON2_HWORDS_IN_BLOCK = ARGON2_BLOCK_SIZE / 32,
ARGON2_512BIT_WORDS_IN_BLOCK = ARGON2_BLOCK_SIZE / 64,
/* Number of pseudo-random values generated by one call to Blake in Argon2i
to
generate reference block positions */
ARGON2_ADDRESSES_IN_BLOCK = 128,
/* Pre-hashing digest length and its extension*/
ARGON2_PREHASH_DIGEST_LENGTH = 64,
ARGON2_PREHASH_SEED_LENGTH = 72
};
/*************************Argon2 internal data types***********************/
/*
* Structure for the (1KB) memory block implemented as 128 64-bit words.
* Memory blocks can be copied, XORed. Internal words can be accessed by [] (no
* bounds checking).
*/
typedef struct block_ { uint64_t v[ARGON2_QWORDS_IN_BLOCK]; } block;
/*****************Functions that work with the block******************/
/* Initialize each byte of the block with @in */
void rxa2_init_block_value(block *b, uint8_t in);
/* Copy block @src to block @dst */
void rxa2_copy_block(block *dst, const block *src);
/* XOR @src onto @dst bytewise */
void rxa2_xor_block(block *dst, const block *src);
/*
* Argon2 instance: memory pointer, number of passes, amount of memory, type,
* and derived values.
* Used to evaluate the number and location of blocks to construct in each
* thread
*/
typedef struct Argon2_instance_t {
block *memory; /* Memory pointer */
uint32_t version;
uint32_t passes; /* Number of passes */
uint32_t memory_blocks; /* Number of blocks in memory */
uint32_t segment_length;
uint32_t lane_length;
uint32_t lanes;
uint32_t threads;
argon2_type type;
int print_internals; /* whether to print the memory blocks */
argon2_context *context_ptr; /* points back to original context */
} argon2_instance_t;
/*
* Argon2 position: where we construct the block right now. Used to distribute
* work between threads.
*/
typedef struct Argon2_position_t {
uint32_t pass;
uint32_t lane;
uint8_t slice;
uint32_t index;
} argon2_position_t;
/*Struct that holds the inputs for thread handling FillSegment*/
typedef struct Argon2_thread_data {
argon2_instance_t *instance_ptr;
argon2_position_t pos;
} argon2_thread_data;
/*************************Argon2 core functions********************************/
/* Allocates memory to the given pointer, uses the appropriate allocator as
* specified in the context. Total allocated memory is num*size.
* @param context argon2_context which specifies the allocator
* @param memory pointer to the pointer to the memory
* @param size the size in bytes for each element to be allocated
* @param num the number of elements to be allocated
* @return ARGON2_OK if @memory is a valid pointer and memory is allocated
*/
int rxa2_allocate_memory(const argon2_context *context, uint8_t **memory,
size_t num, size_t size);
/*
* Frees memory at the given pointer, uses the appropriate deallocator as
* specified in the context. Also cleans the memory using clear_internal_memory.
* @param context argon2_context which specifies the deallocator
* @param memory pointer to buffer to be freed
* @param size the size in bytes for each element to be deallocated
* @param num the number of elements to be deallocated
*/
void rxa2_free_memory(const argon2_context *context, uint8_t *memory,
size_t num, size_t size);
/* Function that securely cleans the memory. This ignores any flags set
* regarding clearing memory. Usually one just calls clear_internal_memory.
* @param mem Pointer to the memory
* @param s Memory size in bytes
*/
void rxa2_secure_wipe_memory(void *v, size_t n);
/* Function that securely clears the memory if FLAG_clear_internal_memory is
* set. If the flag isn't set, this function does nothing.
* @param mem Pointer to the memory
* @param s Memory size in bytes
*/
void rxa2_clear_internal_memory(void *v, size_t n);
/*
* Computes absolute position of reference block in the lane following a skewed
* distribution and using a pseudo-random value as input
* @param instance Pointer to the current instance
* @param position Pointer to the current position
* @param pseudo_rand 32-bit pseudo-random value used to determine the position
* @param same_lane Indicates if the block will be taken from the current lane.
* If so we can reference the current segment
* @pre All pointers must be valid
*/
uint32_t rxa2_index_alpha(const argon2_instance_t *instance,
const argon2_position_t *position, uint32_t pseudo_rand,
int same_lane);
/*
* Function that validates all inputs against predefined restrictions and return
* an error code
* @param context Pointer to current Argon2 context
* @return ARGON2_OK if everything is all right, otherwise one of error codes
* (all defined in <argon2.h>
*/
int rxa2_validate_inputs(const argon2_context *context);
/*
* Hashes all the inputs into @a blockhash[PREHASH_DIGEST_LENGTH], clears
* password and secret if needed
* @param context Pointer to the Argon2 internal structure containing memory
* pointer, and parameters for time and space requirements.
* @param blockhash Buffer for pre-hashing digest
* @param type Argon2 type
* @pre @a blockhash must have at least @a PREHASH_DIGEST_LENGTH bytes
* allocated
*/
void rxa2_initial_hash(uint8_t *blockhash, argon2_context *context,
argon2_type type);
/*
* Function creates first 2 blocks per lane
* @param instance Pointer to the current instance
* @param blockhash Pointer to the pre-hashing digest
* @pre blockhash must point to @a PREHASH_SEED_LENGTH allocated values
*/
void rxa2_fill_first_blocks(uint8_t *blockhash, const argon2_instance_t *instance);
/*
* Function allocates memory, hashes the inputs with Blake, and creates first
* two blocks. Returns the pointer to the main memory with 2 blocks per lane
* initialized
* @param context Pointer to the Argon2 internal structure containing memory
* pointer, and parameters for time and space requirements.
* @param instance Current Argon2 instance
* @return Zero if successful, -1 if memory failed to allocate. @context->state
* will be modified if successful.
*/
int rxa2_argon_initialize(argon2_instance_t *instance, argon2_context *context);
/*
* XORing the last block of each lane, hashing it, making the tag. Deallocates
* the memory.
* @param context Pointer to current Argon2 context (use only the out parameters
* from it)
* @param instance Pointer to current instance of Argon2
* @pre instance->state must point to necessary amount of memory
* @pre context->out must point to outlen bytes of memory
* @pre if context->free_cbk is not NULL, it should point to a function that
* deallocates memory
*/
void rxa2_finalize(const argon2_context *context, argon2_instance_t *instance);
/*
* Function that fills the segment using previous segments also from other
* threads
* @param context current context
* @param instance Pointer to the current instance
* @param position Current position
* @pre all block pointers must be valid
*/
void rxa2_fill_segment(const argon2_instance_t *instance,
argon2_position_t position);
/*
* Function that fills the entire memory t_cost times based on the first two
* blocks in each lane
* @param instance Pointer to the current instance
* @return ARGON2_OK if successful, @context->state
*/
int rxa2_fill_memory_blocks(argon2_instance_t *instance);
#if defined(__cplusplus)
}
#endif
#endif

View file

@ -0,0 +1,214 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
#include <stdint.h>
#include <string.h>
#include <stdlib.h>
#include "argon2.h"
#include "argon2_core.h"
#include "blake2/blamka-round-ref.h"
#include "blake2/blake2-impl.h"
#include "blake2/blake2.h"
/*
* Function fills a new memory block and optionally XORs the old block over the new one.
* @next_block must be initialized.
* @param prev_block Pointer to the previous block
* @param ref_block Pointer to the reference block
* @param next_block Pointer to the block to be constructed
* @param with_xor Whether to XOR into the new block (1) or just overwrite (0)
* @pre all block pointers must be valid
*/
static void fill_block(const block *prev_block, const block *ref_block,
block *next_block, int with_xor) {
block blockR, block_tmp;
unsigned i;
rxa2_copy_block(&blockR, ref_block);
rxa2_xor_block(&blockR, prev_block);
rxa2_copy_block(&block_tmp, &blockR);
/* Now blockR = ref_block + prev_block and block_tmp = ref_block + prev_block */
if (with_xor) {
/* Saving the next block contents for XOR over: */
rxa2_xor_block(&block_tmp, next_block);
/* Now blockR = ref_block + prev_block and
block_tmp = ref_block + prev_block + next_block */
}
/* Apply Blake2 on columns of 64-bit words: (0,1,...,15) , then
(16,17,..31)... finally (112,113,...127) */
for (i = 0; i < 8; ++i) {
BLAKE2_ROUND_NOMSG(
blockR.v[16 * i], blockR.v[16 * i + 1], blockR.v[16 * i + 2],
blockR.v[16 * i + 3], blockR.v[16 * i + 4], blockR.v[16 * i + 5],
blockR.v[16 * i + 6], blockR.v[16 * i + 7], blockR.v[16 * i + 8],
blockR.v[16 * i + 9], blockR.v[16 * i + 10], blockR.v[16 * i + 11],
blockR.v[16 * i + 12], blockR.v[16 * i + 13], blockR.v[16 * i + 14],
blockR.v[16 * i + 15]);
}
/* Apply Blake2 on rows of 64-bit words: (0,1,16,17,...112,113), then
(2,3,18,19,...,114,115).. finally (14,15,30,31,...,126,127) */
for (i = 0; i < 8; i++) {
BLAKE2_ROUND_NOMSG(
blockR.v[2 * i], blockR.v[2 * i + 1], blockR.v[2 * i + 16],
blockR.v[2 * i + 17], blockR.v[2 * i + 32], blockR.v[2 * i + 33],
blockR.v[2 * i + 48], blockR.v[2 * i + 49], blockR.v[2 * i + 64],
blockR.v[2 * i + 65], blockR.v[2 * i + 80], blockR.v[2 * i + 81],
blockR.v[2 * i + 96], blockR.v[2 * i + 97], blockR.v[2 * i + 112],
blockR.v[2 * i + 113]);
}
rxa2_copy_block(next_block, &block_tmp);
rxa2_xor_block(next_block, &blockR);
}
static void next_addresses(block *address_block, block *input_block,
const block *zero_block) {
input_block->v[6]++;
fill_block(zero_block, input_block, address_block, 0);
fill_block(zero_block, address_block, address_block, 0);
}
void rxa2_fill_segment(const argon2_instance_t *instance,
argon2_position_t position) {
block *ref_block = NULL, *curr_block = NULL;
block address_block, input_block, zero_block;
uint64_t pseudo_rand, ref_index, ref_lane;
uint32_t prev_offset, curr_offset;
uint32_t starting_index;
uint32_t i;
int data_independent_addressing;
if (instance == NULL) {
return;
}
data_independent_addressing =
(instance->type == Argon2_i) ||
(instance->type == Argon2_id && (position.pass == 0) &&
(position.slice < ARGON2_SYNC_POINTS / 2));
if (data_independent_addressing) {
rxa2_init_block_value(&zero_block, 0);
rxa2_init_block_value(&input_block, 0);
input_block.v[0] = position.pass;
input_block.v[1] = position.lane;
input_block.v[2] = position.slice;
input_block.v[3] = instance->memory_blocks;
input_block.v[4] = instance->passes;
input_block.v[5] = instance->type;
}
starting_index = 0;
if ((0 == position.pass) && (0 == position.slice)) {
starting_index = 2; /* we have already generated the first two blocks */
/* Don't forget to generate the first block of addresses: */
if (data_independent_addressing) {
next_addresses(&address_block, &input_block, &zero_block);
}
}
/* Offset of the current block */
curr_offset = position.lane * instance->lane_length +
position.slice * instance->segment_length + starting_index;
if (0 == curr_offset % instance->lane_length) {
/* Last block in this lane */
prev_offset = curr_offset + instance->lane_length - 1;
}
else {
/* Previous block */
prev_offset = curr_offset - 1;
}
for (i = starting_index; i < instance->segment_length;
++i, ++curr_offset, ++prev_offset) {
/*1.1 Rotating prev_offset if needed */
if (curr_offset % instance->lane_length == 1) {
prev_offset = curr_offset - 1;
}
/* 1.2 Computing the index of the reference block */
/* 1.2.1 Taking pseudo-random value from the previous block */
if (data_independent_addressing) {
if (i % ARGON2_ADDRESSES_IN_BLOCK == 0) {
next_addresses(&address_block, &input_block, &zero_block);
}
pseudo_rand = address_block.v[i % ARGON2_ADDRESSES_IN_BLOCK];
}
else {
pseudo_rand = instance->memory[prev_offset].v[0];
}
/* 1.2.2 Computing the lane of the reference block */
ref_lane = ((pseudo_rand >> 32)) % instance->lanes;
if ((position.pass == 0) && (position.slice == 0)) {
/* Can not reference other lanes yet */
ref_lane = position.lane;
}
/* 1.2.3 Computing the number of possible reference block within the
* lane.
*/
position.index = i;
ref_index = rxa2_index_alpha(instance, &position, pseudo_rand & 0xFFFFFFFF,
ref_lane == position.lane);
/* 2 Creating a new block */
ref_block =
instance->memory + instance->lane_length * ref_lane + ref_index;
curr_block = instance->memory + curr_offset;
if (ARGON2_VERSION_10 == instance->version) {
/* version 1.2.1 and earlier: overwrite, not XOR */
fill_block(instance->memory + prev_offset, ref_block, curr_block, 0);
}
else {
if (0 == position.pass) {
fill_block(instance->memory + prev_offset, ref_block,
curr_block, 0);
}
else {
fill_block(instance->memory + prev_offset, ref_block,
curr_block, 1);
}
}
}
}

View file

@ -0,0 +1,10 @@
;# restore callee-saved registers - System V AMD64 ABI
pop r15
pop r14
pop r13
pop r12
pop rbp
pop rbx
;# program finished
ret 0

View file

@ -0,0 +1,19 @@
;# save VM register values
pop rcx
mov qword ptr [rcx+0], r8
mov qword ptr [rcx+8], r9
mov qword ptr [rcx+16], r10
mov qword ptr [rcx+24], r11
mov qword ptr [rcx+32], r12
mov qword ptr [rcx+40], r13
mov qword ptr [rcx+48], r14
mov qword ptr [rcx+56], r15
movdqa xmmword ptr [rcx+64], xmm0
movdqa xmmword ptr [rcx+80], xmm1
movdqa xmmword ptr [rcx+96], xmm2
movdqa xmmword ptr [rcx+112], xmm3
lea rcx, [rcx+64]
movdqa xmmword ptr [rcx+64], xmm4
movdqa xmmword ptr [rcx+80], xmm5
movdqa xmmword ptr [rcx+96], xmm6
movdqa xmmword ptr [rcx+112], xmm7

View file

@ -0,0 +1,24 @@
;# restore callee-saved registers - Microsoft x64 calling convention
movdqu xmm15, xmmword ptr [rsp]
movdqu xmm14, xmmword ptr [rsp+16]
movdqu xmm13, xmmword ptr [rsp+32]
movdqu xmm12, xmmword ptr [rsp+48]
movdqu xmm11, xmmword ptr [rsp+64]
add rsp, 80
movdqu xmm10, xmmword ptr [rsp]
movdqu xmm9, xmmword ptr [rsp+16]
movdqu xmm8, xmmword ptr [rsp+32]
movdqu xmm7, xmmword ptr [rsp+48]
movdqu xmm6, xmmword ptr [rsp+64]
add rsp, 80
pop r15
pop r14
pop r13
pop r12
pop rsi
pop rdi
pop rbp
pop rbx
;# program finished
ret

View file

@ -0,0 +1,32 @@
mov rdx, rax
and eax, RANDOMX_SCRATCHPAD_MASK
lea rcx, [rsi+rax]
push rcx
xor r8, qword ptr [rcx+0]
xor r9, qword ptr [rcx+8]
xor r10, qword ptr [rcx+16]
xor r11, qword ptr [rcx+24]
xor r12, qword ptr [rcx+32]
xor r13, qword ptr [rcx+40]
xor r14, qword ptr [rcx+48]
xor r15, qword ptr [rcx+56]
ror rdx, 32
and edx, RANDOMX_SCRATCHPAD_MASK
lea rcx, [rsi+rdx]
push rcx
cvtdq2pd xmm0, qword ptr [rcx+0]
cvtdq2pd xmm1, qword ptr [rcx+8]
cvtdq2pd xmm2, qword ptr [rcx+16]
cvtdq2pd xmm3, qword ptr [rcx+24]
cvtdq2pd xmm4, qword ptr [rcx+32]
cvtdq2pd xmm5, qword ptr [rcx+40]
cvtdq2pd xmm6, qword ptr [rcx+48]
cvtdq2pd xmm7, qword ptr [rcx+56]
andps xmm4, xmm13
andps xmm5, xmm13
andps xmm6, xmm13
andps xmm7, xmm13
orps xmm4, xmm14
orps xmm5, xmm14
orps xmm6, xmm14
orps xmm7, xmm14

View file

@ -0,0 +1,19 @@
xor eax, eax
pop rcx
mov qword ptr [rcx+0], r8
mov qword ptr [rcx+8], r9
mov qword ptr [rcx+16], r10
mov qword ptr [rcx+24], r11
mov qword ptr [rcx+32], r12
mov qword ptr [rcx+40], r13
mov qword ptr [rcx+48], r14
mov qword ptr [rcx+56], r15
pop rcx
xorpd xmm0, xmm4
xorpd xmm1, xmm5
xorpd xmm2, xmm6
xorpd xmm3, xmm7
movapd xmmword ptr [rcx+0], xmm0
movapd xmmword ptr [rcx+16], xmm1
movapd xmmword ptr [rcx+32], xmm2
movapd xmmword ptr [rcx+48], xmm3

View file

@ -0,0 +1,34 @@
;# callee-saved registers - System V AMD64 ABI
push rbx
push rbp
push r12
push r13
push r14
push r15
;# function arguments
mov rbx, rcx ;# loop counter
push rdi ;# RegisterFile& registerFile
mov rcx, rdi
mov rbp, qword ptr [rsi] ;# "mx", "ma"
mov rdi, qword ptr [rsi+8] ;# uint8_t* dataset
mov rsi, rdx ;# uint8_t* scratchpad
mov rax, rbp
;# zero integer registers
xor r8, r8
xor r9, r9
xor r10, r10
xor r11, r11
xor r12, r12
xor r13, r13
xor r14, r14
xor r15, r15
;# load constant registers
lea rcx, [rcx+120]
movapd xmm8, xmmword ptr [rcx+72]
movapd xmm9, xmmword ptr [rcx+88]
movapd xmm10, xmmword ptr [rcx+104]
movapd xmm11, xmmword ptr [rcx+120]

View file

@ -0,0 +1,47 @@
;# callee-saved registers - Microsoft x64 calling convention
push rbx
push rbp
push rdi
push rsi
push r12
push r13
push r14
push r15
sub rsp, 80
movdqu xmmword ptr [rsp+64], xmm6
movdqu xmmword ptr [rsp+48], xmm7
movdqu xmmword ptr [rsp+32], xmm8
movdqu xmmword ptr [rsp+16], xmm9
movdqu xmmword ptr [rsp+0], xmm10
sub rsp, 80
movdqu xmmword ptr [rsp+64], xmm11
movdqu xmmword ptr [rsp+48], xmm12
movdqu xmmword ptr [rsp+32], xmm13
movdqu xmmword ptr [rsp+16], xmm14
movdqu xmmword ptr [rsp+0], xmm15
;# function arguments
push rcx ;# RegisterFile& registerFile
mov rbp, qword ptr [rdx] ;# "mx", "ma"
mov rdi, qword ptr [rdx+8] ;# uint8_t* dataset
mov rsi, r8 ;# uint8_t* scratchpad
mov rbx, r9 ;# loop counter
mov rax, rbp
;# zero integer registers
xor r8, r8
xor r9, r9
xor r10, r10
xor r11, r11
xor r12, r12
xor r13, r13
xor r14, r14
xor r15, r15
;# load constant registers
lea rcx, [rcx+120]
movapd xmm8, xmmword ptr [rcx+72]
movapd xmm9, xmmword ptr [rcx+88]
movapd xmm10, xmmword ptr [rcx+104]
movapd xmm11, xmmword ptr [rcx+120]

View file

@ -0,0 +1,17 @@
xor rbp, rax ;# modify "mx"
mov edx, ebp ;# edx = mx
and edx, RANDOMX_DATASET_BASE_MASK
prefetchnta byte ptr [rdi+rdx]
ror rbp, 32 ;# swap "ma" and "mx"
mov edx, ebp ;# edx = ma
and edx, RANDOMX_DATASET_BASE_MASK
lea rcx, [rdi+rdx] ;# dataset cache line
xor r8, qword ptr [rcx+0]
xor r9, qword ptr [rcx+8]
xor r10, qword ptr [rcx+16]
xor r11, qword ptr [rcx+24]
xor r12, qword ptr [rcx+32]
xor r13, qword ptr [rcx+40]
xor r14, qword ptr [rcx+48]
xor r15, qword ptr [rcx+56]

View file

@ -0,0 +1,10 @@
mov rbx, qword ptr [rsp+64]
xor r8, qword ptr [rsp+56]
xor r9, qword ptr [rsp+48]
xor r10, qword ptr [rsp+40]
xor r11, qword ptr [rsp+32]
xor r12, qword ptr [rsp+24]
xor r13, qword ptr [rsp+16]
xor r14, qword ptr [rsp+8]
xor r15, qword ptr [rsp+0]
add rsp, 72

View file

@ -0,0 +1,17 @@
sub rsp, 72
mov qword ptr [rsp+64], rbx
mov qword ptr [rsp+56], r8
mov qword ptr [rsp+48], r9
mov qword ptr [rsp+40], r10
mov qword ptr [rsp+32], r11
mov qword ptr [rsp+24], r12
mov qword ptr [rsp+16], r13
mov qword ptr [rsp+8], r14
mov qword ptr [rsp+0], r15
xor rbp, rax ;# modify "mx"
ror rbp, 32 ;# swap "ma" and "mx"
mov ebx, ebp ;# ecx = ma
and ebx, RANDOMX_DATASET_BASE_MASK
shr ebx, 6 ;# ebx = Dataset block number
;# add ebx, datasetOffset / 64
;# call 32768

View file

@ -0,0 +1,24 @@
r0_mul:
;#/ 6364136223846793005
db 45, 127, 149, 76, 45, 244, 81, 88
r1_add:
;#/ 9298411001130361340
db 252, 161, 245, 89, 138, 151, 10, 129
r2_add:
;#/ 12065312585734608966
db 70, 216, 194, 56, 223, 153, 112, 167
r3_add:
;#/ 9306329213124626780
db 92, 73, 34, 191, 28, 185, 38, 129
r4_add:
;#/ 5281919268842080866
db 98, 138, 159, 23, 151, 37, 77, 73
r5_add:
;#/ 10536153434571861004
db 12, 236, 170, 206, 185, 239, 55, 146
r6_add:
;#/ 3398623926847679864
db 120, 45, 230, 108, 116, 86, 42, 47
r7_add:
;#/ 9549104520008361294
db 78, 229, 44, 182, 247, 59, 133, 132

View file

@ -0,0 +1,8 @@
xor r8, qword ptr [rbx+0]
xor r9, qword ptr [rbx+8]
xor r10, qword ptr [rbx+16]
xor r11, qword ptr [rbx+24]
xor r12, qword ptr [rbx+32]
xor r13, qword ptr [rbx+40]
xor r14, qword ptr [rbx+48]
xor r15, qword ptr [rbx+56]

View file

@ -0,0 +1,4 @@
and rbx, RANDOMX_CACHE_MASK
shl rbx, 6
add rbx, rdi
prefetchnta byte ptr [rbx]

View file

@ -0,0 +1,6 @@
mantissaMask:
db 255, 255, 255, 255, 255, 255, 255, 0, 255, 255, 255, 255, 255, 255, 255, 0
exp240:
db 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
scaleMask:
db 0, 0, 0, 0, 0, 0, 240, 128, 0, 0, 0, 0, 0, 0, 240, 128

View file

@ -0,0 +1,7 @@
mov edx, 1
mov r8, rcx
xor eax, eax
bsr rcx, rcx
shl rdx, cl
div r8
ret

View file

@ -0,0 +1,76 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
#ifndef PORTABLE_BLAKE2_IMPL_H
#define PORTABLE_BLAKE2_IMPL_H
#include <stdint.h>
#include "endian.h"
static FORCE_INLINE uint64_t load48(const void *src) {
const uint8_t *p = (const uint8_t *)src;
uint64_t w = *p++;
w |= (uint64_t)(*p++) << 8;
w |= (uint64_t)(*p++) << 16;
w |= (uint64_t)(*p++) << 24;
w |= (uint64_t)(*p++) << 32;
w |= (uint64_t)(*p++) << 40;
return w;
}
static FORCE_INLINE void store48(void *dst, uint64_t w) {
uint8_t *p = (uint8_t *)dst;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
}
static FORCE_INLINE uint32_t rotr32(const uint32_t w, const unsigned c) {
return (w >> c) | (w << (32 - c));
}
static FORCE_INLINE uint64_t rotr64(const uint64_t w, const unsigned c) {
return (w >> c) | (w << (64 - c));
}
#endif

View file

@ -0,0 +1,107 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
#ifndef PORTABLE_BLAKE2_H
#define PORTABLE_BLAKE2_H
#include <stdint.h>
#include <limits.h>
#if defined(__cplusplus)
extern "C" {
#endif
enum blake2b_constant {
BLAKE2B_BLOCKBYTES = 128,
BLAKE2B_OUTBYTES = 64,
BLAKE2B_KEYBYTES = 64,
BLAKE2B_SALTBYTES = 16,
BLAKE2B_PERSONALBYTES = 16
};
#pragma pack(push, 1)
typedef struct __blake2b_param {
uint8_t digest_length; /* 1 */
uint8_t key_length; /* 2 */
uint8_t fanout; /* 3 */
uint8_t depth; /* 4 */
uint32_t leaf_length; /* 8 */
uint64_t node_offset; /* 16 */
uint8_t node_depth; /* 17 */
uint8_t inner_length; /* 18 */
uint8_t reserved[14]; /* 32 */
uint8_t salt[BLAKE2B_SALTBYTES]; /* 48 */
uint8_t personal[BLAKE2B_PERSONALBYTES]; /* 64 */
} blake2b_param;
#pragma pack(pop)
typedef struct __blake2b_state {
uint64_t h[8];
uint64_t t[2];
uint64_t f[2];
uint8_t buf[BLAKE2B_BLOCKBYTES];
unsigned buflen;
unsigned outlen;
uint8_t last_node;
} blake2b_state;
/* Ensure param structs have not been wrongly padded */
/* Poor man's static_assert */
enum {
blake2_size_check_0 = 1 / !!(CHAR_BIT == 8),
blake2_size_check_2 =
1 / !!(sizeof(blake2b_param) == sizeof(uint64_t) * CHAR_BIT)
};
/* Streaming API */
int blake2b_init(blake2b_state *S, size_t outlen);
int blake2b_init_key(blake2b_state *S, size_t outlen, const void *key,
size_t keylen);
int blake2b_init_param(blake2b_state *S, const blake2b_param *P);
int blake2b_update(blake2b_state *S, const void *in, size_t inlen);
int blake2b_final(blake2b_state *S, void *out, size_t outlen);
/* Simple API */
int blake2b(void *out, size_t outlen, const void *in, size_t inlen,
const void *key, size_t keylen);
/* Argon2 Team - Begin Code */
int rxa2_blake2b_long(void *out, size_t outlen, const void *in, size_t inlen);
/* Argon2 Team - End Code */
#if defined(__cplusplus)
}
#endif
#endif

View file

@ -0,0 +1,409 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "blake2.h"
#include "blake2-impl.h"
static const uint64_t blake2b_IV[8] = {
UINT64_C(0x6a09e667f3bcc908), UINT64_C(0xbb67ae8584caa73b),
UINT64_C(0x3c6ef372fe94f82b), UINT64_C(0xa54ff53a5f1d36f1),
UINT64_C(0x510e527fade682d1), UINT64_C(0x9b05688c2b3e6c1f),
UINT64_C(0x1f83d9abfb41bd6b), UINT64_C(0x5be0cd19137e2179) };
static const unsigned int blake2b_sigma[12][16] = {
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3},
{11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4},
{7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8},
{9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13},
{2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9},
{12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11},
{13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10},
{6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5},
{10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0},
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3},
};
static FORCE_INLINE void blake2b_set_lastnode(blake2b_state *S) {
S->f[1] = (uint64_t)-1;
}
static FORCE_INLINE void blake2b_set_lastblock(blake2b_state *S) {
if (S->last_node) {
blake2b_set_lastnode(S);
}
S->f[0] = (uint64_t)-1;
}
static FORCE_INLINE void blake2b_increment_counter(blake2b_state *S,
uint64_t inc) {
S->t[0] += inc;
S->t[1] += (S->t[0] < inc);
}
static FORCE_INLINE void blake2b_invalidate_state(blake2b_state *S) {
//clear_internal_memory(S, sizeof(*S)); /* wipe */
blake2b_set_lastblock(S); /* invalidate for further use */
}
static FORCE_INLINE void blake2b_init0(blake2b_state *S) {
memset(S, 0, sizeof(*S));
memcpy(S->h, blake2b_IV, sizeof(S->h));
}
int blake2b_init_param(blake2b_state *S, const blake2b_param *P) {
const unsigned char *p = (const unsigned char *)P;
unsigned int i;
if (NULL == P || NULL == S) {
return -1;
}
blake2b_init0(S);
/* IV XOR Parameter Block */
for (i = 0; i < 8; ++i) {
S->h[i] ^= load64(&p[i * sizeof(S->h[i])]);
}
S->outlen = P->digest_length;
return 0;
}
/* Sequential blake2b initialization */
int blake2b_init(blake2b_state *S, size_t outlen) {
blake2b_param P;
if (S == NULL) {
return -1;
}
if ((outlen == 0) || (outlen > BLAKE2B_OUTBYTES)) {
blake2b_invalidate_state(S);
return -1;
}
/* Setup Parameter Block for unkeyed BLAKE2 */
P.digest_length = (uint8_t)outlen;
P.key_length = 0;
P.fanout = 1;
P.depth = 1;
P.leaf_length = 0;
P.node_offset = 0;
P.node_depth = 0;
P.inner_length = 0;
memset(P.reserved, 0, sizeof(P.reserved));
memset(P.salt, 0, sizeof(P.salt));
memset(P.personal, 0, sizeof(P.personal));
return blake2b_init_param(S, &P);
}
int blake2b_init_key(blake2b_state *S, size_t outlen, const void *key, size_t keylen) {
blake2b_param P;
if (S == NULL) {
return -1;
}
if ((outlen == 0) || (outlen > BLAKE2B_OUTBYTES)) {
blake2b_invalidate_state(S);
return -1;
}
if ((key == 0) || (keylen == 0) || (keylen > BLAKE2B_KEYBYTES)) {
blake2b_invalidate_state(S);
return -1;
}
/* Setup Parameter Block for keyed BLAKE2 */
P.digest_length = (uint8_t)outlen;
P.key_length = (uint8_t)keylen;
P.fanout = 1;
P.depth = 1;
P.leaf_length = 0;
P.node_offset = 0;
P.node_depth = 0;
P.inner_length = 0;
memset(P.reserved, 0, sizeof(P.reserved));
memset(P.salt, 0, sizeof(P.salt));
memset(P.personal, 0, sizeof(P.personal));
if (blake2b_init_param(S, &P) < 0) {
blake2b_invalidate_state(S);
return -1;
}
{
uint8_t block[BLAKE2B_BLOCKBYTES];
memset(block, 0, BLAKE2B_BLOCKBYTES);
memcpy(block, key, keylen);
blake2b_update(S, block, BLAKE2B_BLOCKBYTES);
/* Burn the key from stack */
//clear_internal_memory(block, BLAKE2B_BLOCKBYTES);
}
return 0;
}
static void blake2b_compress(blake2b_state *S, const uint8_t *block) {
uint64_t m[16];
uint64_t v[16];
unsigned int i, r;
for (i = 0; i < 16; ++i) {
m[i] = load64(block + i * sizeof(m[i]));
}
for (i = 0; i < 8; ++i) {
v[i] = S->h[i];
}
v[8] = blake2b_IV[0];
v[9] = blake2b_IV[1];
v[10] = blake2b_IV[2];
v[11] = blake2b_IV[3];
v[12] = blake2b_IV[4] ^ S->t[0];
v[13] = blake2b_IV[5] ^ S->t[1];
v[14] = blake2b_IV[6] ^ S->f[0];
v[15] = blake2b_IV[7] ^ S->f[1];
#define G(r, i, a, b, c, d) \
do { \
a = a + b + m[blake2b_sigma[r][2 * i + 0]]; \
d = rotr64(d ^ a, 32); \
c = c + d; \
b = rotr64(b ^ c, 24); \
a = a + b + m[blake2b_sigma[r][2 * i + 1]]; \
d = rotr64(d ^ a, 16); \
c = c + d; \
b = rotr64(b ^ c, 63); \
} while ((void)0, 0)
#define ROUND(r) \
do { \
G(r, 0, v[0], v[4], v[8], v[12]); \
G(r, 1, v[1], v[5], v[9], v[13]); \
G(r, 2, v[2], v[6], v[10], v[14]); \
G(r, 3, v[3], v[7], v[11], v[15]); \
G(r, 4, v[0], v[5], v[10], v[15]); \
G(r, 5, v[1], v[6], v[11], v[12]); \
G(r, 6, v[2], v[7], v[8], v[13]); \
G(r, 7, v[3], v[4], v[9], v[14]); \
} while ((void)0, 0)
for (r = 0; r < 12; ++r) {
ROUND(r);
}
for (i = 0; i < 8; ++i) {
S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
}
#undef G
#undef ROUND
}
int blake2b_update(blake2b_state *S, const void *in, size_t inlen) {
const uint8_t *pin = (const uint8_t *)in;
if (inlen == 0) {
return 0;
}
/* Sanity check */
if (S == NULL || in == NULL) {
return -1;
}
/* Is this a reused state? */
if (S->f[0] != 0) {
return -1;
}
if (S->buflen + inlen > BLAKE2B_BLOCKBYTES) {
/* Complete current block */
size_t left = S->buflen;
size_t fill = BLAKE2B_BLOCKBYTES - left;
memcpy(&S->buf[left], pin, fill);
blake2b_increment_counter(S, BLAKE2B_BLOCKBYTES);
blake2b_compress(S, S->buf);
S->buflen = 0;
inlen -= fill;
pin += fill;
/* Avoid buffer copies when possible */
while (inlen > BLAKE2B_BLOCKBYTES) {
blake2b_increment_counter(S, BLAKE2B_BLOCKBYTES);
blake2b_compress(S, pin);
inlen -= BLAKE2B_BLOCKBYTES;
pin += BLAKE2B_BLOCKBYTES;
}
}
memcpy(&S->buf[S->buflen], pin, inlen);
S->buflen += (unsigned int)inlen;
return 0;
}
int blake2b_final(blake2b_state *S, void *out, size_t outlen) {
uint8_t buffer[BLAKE2B_OUTBYTES] = { 0 };
unsigned int i;
/* Sanity checks */
if (S == NULL || out == NULL || outlen < S->outlen) {
return -1;
}
/* Is this a reused state? */
if (S->f[0] != 0) {
return -1;
}
blake2b_increment_counter(S, S->buflen);
blake2b_set_lastblock(S);
memset(&S->buf[S->buflen], 0, BLAKE2B_BLOCKBYTES - S->buflen); /* Padding */
blake2b_compress(S, S->buf);
for (i = 0; i < 8; ++i) { /* Output full hash to temp buffer */
store64(buffer + sizeof(S->h[i]) * i, S->h[i]);
}
memcpy(out, buffer, S->outlen);
//clear_internal_memory(buffer, sizeof(buffer));
//clear_internal_memory(S->buf, sizeof(S->buf));
//clear_internal_memory(S->h, sizeof(S->h));
return 0;
}
int blake2b(void *out, size_t outlen, const void *in, size_t inlen,
const void *key, size_t keylen) {
blake2b_state S;
int ret = -1;
/* Verify parameters */
if (NULL == in && inlen > 0) {
goto fail;
}
if (NULL == out || outlen == 0 || outlen > BLAKE2B_OUTBYTES) {
goto fail;
}
if ((NULL == key && keylen > 0) || keylen > BLAKE2B_KEYBYTES) {
goto fail;
}
if (keylen > 0) {
if (blake2b_init_key(&S, outlen, key, keylen) < 0) {
goto fail;
}
}
else {
if (blake2b_init(&S, outlen) < 0) {
goto fail;
}
}
if (blake2b_update(&S, in, inlen) < 0) {
goto fail;
}
ret = blake2b_final(&S, out, outlen);
fail:
//clear_internal_memory(&S, sizeof(S));
return ret;
}
/* Argon2 Team - Begin Code */
int rxa2_blake2b_long(void *pout, size_t outlen, const void *in, size_t inlen) {
uint8_t *out = (uint8_t *)pout;
blake2b_state blake_state;
uint8_t outlen_bytes[sizeof(uint32_t)] = { 0 };
int ret = -1;
if (outlen > UINT32_MAX) {
goto fail;
}
/* Ensure little-endian byte order! */
store32(outlen_bytes, (uint32_t)outlen);
#define TRY(statement) \
do { \
ret = statement; \
if (ret < 0) { \
goto fail; \
} \
} while ((void)0, 0)
if (outlen <= BLAKE2B_OUTBYTES) {
TRY(blake2b_init(&blake_state, outlen));
TRY(blake2b_update(&blake_state, outlen_bytes, sizeof(outlen_bytes)));
TRY(blake2b_update(&blake_state, in, inlen));
TRY(blake2b_final(&blake_state, out, outlen));
}
else {
uint32_t toproduce;
uint8_t out_buffer[BLAKE2B_OUTBYTES];
uint8_t in_buffer[BLAKE2B_OUTBYTES];
TRY(blake2b_init(&blake_state, BLAKE2B_OUTBYTES));
TRY(blake2b_update(&blake_state, outlen_bytes, sizeof(outlen_bytes)));
TRY(blake2b_update(&blake_state, in, inlen));
TRY(blake2b_final(&blake_state, out_buffer, BLAKE2B_OUTBYTES));
memcpy(out, out_buffer, BLAKE2B_OUTBYTES / 2);
out += BLAKE2B_OUTBYTES / 2;
toproduce = (uint32_t)outlen - BLAKE2B_OUTBYTES / 2;
while (toproduce > BLAKE2B_OUTBYTES) {
memcpy(in_buffer, out_buffer, BLAKE2B_OUTBYTES);
TRY(blake2b(out_buffer, BLAKE2B_OUTBYTES, in_buffer,
BLAKE2B_OUTBYTES, NULL, 0));
memcpy(out, out_buffer, BLAKE2B_OUTBYTES / 2);
out += BLAKE2B_OUTBYTES / 2;
toproduce -= BLAKE2B_OUTBYTES / 2;
}
memcpy(in_buffer, out_buffer, BLAKE2B_OUTBYTES);
TRY(blake2b(out_buffer, toproduce, in_buffer, BLAKE2B_OUTBYTES, NULL,
0));
memcpy(out, out_buffer, toproduce);
}
fail:
//clear_internal_memory(&blake_state, sizeof(blake_state));
return ret;
#undef TRY
}
/* Argon2 Team - End Code */

View file

@ -0,0 +1,73 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
#ifndef BLAKE_ROUND_MKA_H
#define BLAKE_ROUND_MKA_H
#include "blake2.h"
#include "blake2-impl.h"
/* designed by the Lyra PHC team */
static FORCE_INLINE uint64_t fBlaMka(uint64_t x, uint64_t y) {
const uint64_t m = UINT64_C(0xFFFFFFFF);
const uint64_t xy = (x & m) * (y & m);
return x + y + 2 * xy;
}
#define G(a, b, c, d) \
do { \
a = fBlaMka(a, b); \
d = rotr64(d ^ a, 32); \
c = fBlaMka(c, d); \
b = rotr64(b ^ c, 24); \
a = fBlaMka(a, b); \
d = rotr64(d ^ a, 16); \
c = fBlaMka(c, d); \
b = rotr64(b ^ c, 63); \
} while ((void)0, 0)
#define BLAKE2_ROUND_NOMSG(v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, \
v12, v13, v14, v15) \
do { \
G(v0, v4, v8, v12); \
G(v1, v5, v9, v13); \
G(v2, v6, v10, v14); \
G(v3, v7, v11, v15); \
G(v0, v5, v10, v15); \
G(v1, v6, v11, v12); \
G(v2, v7, v8, v13); \
G(v3, v4, v9, v14); \
} while ((void)0, 0)
#endif

View file

@ -0,0 +1,107 @@
#pragma once
#include <stdint.h>
#include <string.h>
#if defined(_MSC_VER)
#define FORCE_INLINE __inline
#elif defined(__GNUC__) || defined(__clang__)
#define FORCE_INLINE __inline__
#else
#define FORCE_INLINE
#endif
/* Argon2 Team - Begin Code */
/*
Not an exhaustive list, but should cover the majority of modern platforms
Additionally, the code will always be correct---this is only a performance
tweak.
*/
#if (defined(__BYTE_ORDER__) && \
(__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)) || \
defined(__LITTLE_ENDIAN__) || defined(__ARMEL__) || defined(__MIPSEL__) || \
defined(__AARCH64EL__) || defined(__amd64__) || defined(__i386__) || \
defined(_M_IX86) || defined(_M_X64) || defined(_M_AMD64) || \
defined(_M_ARM)
#define NATIVE_LITTLE_ENDIAN
#endif
/* Argon2 Team - End Code */
static FORCE_INLINE uint32_t load32(const void *src) {
#if defined(NATIVE_LITTLE_ENDIAN)
uint32_t w;
memcpy(&w, src, sizeof w);
return w;
#else
const uint8_t *p = (const uint8_t *)src;
uint32_t w = *p++;
w |= (uint32_t)(*p++) << 8;
w |= (uint32_t)(*p++) << 16;
w |= (uint32_t)(*p++) << 24;
return w;
#endif
}
static FORCE_INLINE uint64_t load64_native(const void *src) {
uint64_t w;
memcpy(&w, src, sizeof w);
return w;
}
static FORCE_INLINE uint64_t load64(const void *src) {
#if defined(NATIVE_LITTLE_ENDIAN)
return load64_native(src);
#else
const uint8_t *p = (const uint8_t *)src;
uint64_t w = *p++;
w |= (uint64_t)(*p++) << 8;
w |= (uint64_t)(*p++) << 16;
w |= (uint64_t)(*p++) << 24;
w |= (uint64_t)(*p++) << 32;
w |= (uint64_t)(*p++) << 40;
w |= (uint64_t)(*p++) << 48;
w |= (uint64_t)(*p++) << 56;
return w;
#endif
}
static FORCE_INLINE void store32(void *dst, uint32_t w) {
#if defined(NATIVE_LITTLE_ENDIAN)
memcpy(dst, &w, sizeof w);
#else
uint8_t *p = (uint8_t *)dst;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
#endif
}
static FORCE_INLINE void store64_native(void *dst, uint64_t w) {
memcpy(dst, &w, sizeof w);
}
static FORCE_INLINE void store64(void *dst, uint64_t w) {
#if defined(NATIVE_LITTLE_ENDIAN)
store64_native(dst, w);
#else
uint8_t *p = (uint8_t *)dst;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
w >>= 8;
*p++ = (uint8_t)w;
#endif
}

View file

@ -0,0 +1,62 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <stddef.h>
#include "blake2/blake2.h"
#include "blake2/endian.h"
#include "blake2_generator.hpp"
namespace randomx {
constexpr int maxSeedSize = 60;
Blake2Generator::Blake2Generator(const void* seed, size_t seedSize, int nonce) : dataIndex(sizeof(data)) {
memset(data, 0, sizeof(data));
memcpy(data, seed, seedSize > maxSeedSize ? maxSeedSize : seedSize);
store32(&data[maxSeedSize], nonce);
}
uint8_t Blake2Generator::getByte() {
checkData(1);
return data[dataIndex++];
}
uint32_t Blake2Generator::getInt32() {
checkData(4);
auto ret = load32(&data[dataIndex]);
dataIndex += 4;
return ret;
}
void Blake2Generator::checkData(const size_t bytesNeeded) {
if (dataIndex + bytesNeeded > sizeof(data)) {
blake2b(data, sizeof(data), data, sizeof(data), nullptr, 0);
dataIndex = 0;
}
}
}

View file

@ -0,0 +1,46 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
namespace randomx {
class Blake2Generator {
public:
Blake2Generator(const void* seed, size_t seedSize, int nonce = 0);
uint8_t getByte();
uint32_t getInt32();
private:
void checkData(const size_t);
uint8_t data[64];
size_t dataIndex;
};
}

View file

@ -0,0 +1,482 @@
/*
Copyright (c) 2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "bytecode_machine.hpp"
#include "reciprocal.h"
namespace randomx {
const int_reg_t BytecodeMachine::zero = 0;
#define INSTR_CASE(x) case InstructionType::x: \
exe_ ## x(ibc, pc, scratchpad, config); \
break;
void BytecodeMachine::executeInstruction(RANDOMX_EXE_ARGS) {
switch (ibc.type)
{
INSTR_CASE(IADD_RS)
INSTR_CASE(IADD_M)
INSTR_CASE(ISUB_R)
INSTR_CASE(ISUB_M)
INSTR_CASE(IMUL_R)
INSTR_CASE(IMUL_M)
INSTR_CASE(IMULH_R)
INSTR_CASE(IMULH_M)
INSTR_CASE(ISMULH_R)
INSTR_CASE(ISMULH_M)
INSTR_CASE(INEG_R)
INSTR_CASE(IXOR_R)
INSTR_CASE(IXOR_M)
INSTR_CASE(IROR_R)
INSTR_CASE(IROL_R)
INSTR_CASE(ISWAP_R)
INSTR_CASE(FSWAP_R)
INSTR_CASE(FADD_R)
INSTR_CASE(FADD_M)
INSTR_CASE(FSUB_R)
INSTR_CASE(FSUB_M)
INSTR_CASE(FSCAL_R)
INSTR_CASE(FMUL_R)
INSTR_CASE(FDIV_M)
INSTR_CASE(FSQRT_R)
INSTR_CASE(CBRANCH)
INSTR_CASE(CFROUND)
INSTR_CASE(ISTORE)
case InstructionType::NOP:
break;
case InstructionType::IMUL_RCP: //executed as IMUL_R
default:
UNREACHABLE;
}
}
void BytecodeMachine::compileInstruction(RANDOMX_GEN_ARGS) {
int opcode = instr.opcode;
if (opcode < RandomX_CurrentConfig.CEIL_IADD_RS) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IADD_RS;
ibc.idst = &nreg->r[dst];
if (dst != RegisterNeedsDisplacement) {
ibc.isrc = &nreg->r[src];
ibc.shift = instr.getModShift();
ibc.imm = 0;
}
else {
ibc.isrc = &nreg->r[src];
ibc.shift = instr.getModShift();
ibc.imm = signExtend2sCompl(instr.getImm32());
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IADD_M) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IADD_M;
ibc.idst = &nreg->r[dst];
ibc.imm = signExtend2sCompl(instr.getImm32());
if (src != dst) {
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
}
else {
ibc.isrc = &zero;
ibc.memMask = ScratchpadL3Mask;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_ISUB_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::ISUB_R;
ibc.idst = &nreg->r[dst];
if (src != dst) {
ibc.isrc = &nreg->r[src];
}
else {
ibc.imm = signExtend2sCompl(instr.getImm32());
ibc.isrc = &ibc.imm;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_ISUB_M) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::ISUB_M;
ibc.idst = &nreg->r[dst];
ibc.imm = signExtend2sCompl(instr.getImm32());
if (src != dst) {
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
}
else {
ibc.isrc = &zero;
ibc.memMask = ScratchpadL3Mask;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IMUL_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IMUL_R;
ibc.idst = &nreg->r[dst];
if (src != dst) {
ibc.isrc = &nreg->r[src];
}
else {
ibc.imm = signExtend2sCompl(instr.getImm32());
ibc.isrc = &ibc.imm;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IMUL_M) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IMUL_M;
ibc.idst = &nreg->r[dst];
ibc.imm = signExtend2sCompl(instr.getImm32());
if (src != dst) {
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
}
else {
ibc.isrc = &zero;
ibc.memMask = ScratchpadL3Mask;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IMULH_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IMULH_R;
ibc.idst = &nreg->r[dst];
ibc.isrc = &nreg->r[src];
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IMULH_M) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IMULH_M;
ibc.idst = &nreg->r[dst];
ibc.imm = signExtend2sCompl(instr.getImm32());
if (src != dst) {
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
}
else {
ibc.isrc = &zero;
ibc.memMask = ScratchpadL3Mask;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_ISMULH_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::ISMULH_R;
ibc.idst = &nreg->r[dst];
ibc.isrc = &nreg->r[src];
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_ISMULH_M) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::ISMULH_M;
ibc.idst = &nreg->r[dst];
ibc.imm = signExtend2sCompl(instr.getImm32());
if (src != dst) {
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
}
else {
ibc.isrc = &zero;
ibc.memMask = ScratchpadL3Mask;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IMUL_RCP) {
uint64_t divisor = instr.getImm32();
if (!isPowerOf2(divisor)) {
auto dst = instr.dst % RegistersCount;
ibc.type = InstructionType::IMUL_R;
ibc.idst = &nreg->r[dst];
ibc.imm = randomx_reciprocal(divisor);
ibc.isrc = &ibc.imm;
registerUsage[dst] = i;
}
else {
ibc.type = InstructionType::NOP;
}
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_INEG_R) {
auto dst = instr.dst % RegistersCount;
ibc.type = InstructionType::INEG_R;
ibc.idst = &nreg->r[dst];
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IXOR_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IXOR_R;
ibc.idst = &nreg->r[dst];
if (src != dst) {
ibc.isrc = &nreg->r[src];
}
else {
ibc.imm = signExtend2sCompl(instr.getImm32());
ibc.isrc = &ibc.imm;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IXOR_M) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IXOR_M;
ibc.idst = &nreg->r[dst];
ibc.imm = signExtend2sCompl(instr.getImm32());
if (src != dst) {
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
}
else {
ibc.isrc = &zero;
ibc.memMask = ScratchpadL3Mask;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IROR_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IROR_R;
ibc.idst = &nreg->r[dst];
if (src != dst) {
ibc.isrc = &nreg->r[src];
}
else {
ibc.imm = instr.getImm32();
ibc.isrc = &ibc.imm;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_IROL_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::IROL_R;
ibc.idst = &nreg->r[dst];
if (src != dst) {
ibc.isrc = &nreg->r[src];
}
else {
ibc.imm = instr.getImm32();
ibc.isrc = &ibc.imm;
}
registerUsage[dst] = i;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_ISWAP_R) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
if (src != dst) {
ibc.idst = &nreg->r[dst];
ibc.isrc = &nreg->r[src];
ibc.type = InstructionType::ISWAP_R;
registerUsage[dst] = i;
registerUsage[src] = i;
}
else {
ibc.type = InstructionType::NOP;
}
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FSWAP_R) {
auto dst = instr.dst % RegistersCount;
ibc.type = InstructionType::FSWAP_R;
if (dst < RegisterCountFlt)
ibc.fdst = &nreg->f[dst];
else
ibc.fdst = &nreg->e[dst - RegisterCountFlt];
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FADD_R) {
auto dst = instr.dst % RegisterCountFlt;
auto src = instr.src % RegisterCountFlt;
ibc.type = InstructionType::FADD_R;
ibc.fdst = &nreg->f[dst];
ibc.fsrc = &nreg->a[src];
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FADD_M) {
auto dst = instr.dst % RegisterCountFlt;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::FADD_M;
ibc.fdst = &nreg->f[dst];
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
ibc.imm = signExtend2sCompl(instr.getImm32());
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FSUB_R) {
auto dst = instr.dst % RegisterCountFlt;
auto src = instr.src % RegisterCountFlt;
ibc.type = InstructionType::FSUB_R;
ibc.fdst = &nreg->f[dst];
ibc.fsrc = &nreg->a[src];
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FSUB_M) {
auto dst = instr.dst % RegisterCountFlt;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::FSUB_M;
ibc.fdst = &nreg->f[dst];
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
ibc.imm = signExtend2sCompl(instr.getImm32());
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FSCAL_R) {
auto dst = instr.dst % RegisterCountFlt;
ibc.fdst = &nreg->f[dst];
ibc.type = InstructionType::FSCAL_R;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FMUL_R) {
auto dst = instr.dst % RegisterCountFlt;
auto src = instr.src % RegisterCountFlt;
ibc.type = InstructionType::FMUL_R;
ibc.fdst = &nreg->e[dst];
ibc.fsrc = &nreg->a[src];
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FDIV_M) {
auto dst = instr.dst % RegisterCountFlt;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::FDIV_M;
ibc.fdst = &nreg->e[dst];
ibc.isrc = &nreg->r[src];
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
ibc.imm = signExtend2sCompl(instr.getImm32());
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_FSQRT_R) {
auto dst = instr.dst % RegisterCountFlt;
ibc.type = InstructionType::FSQRT_R;
ibc.fdst = &nreg->e[dst];
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_CBRANCH) {
ibc.type = InstructionType::CBRANCH;
//jump condition
int creg = instr.dst % RegistersCount;
ibc.idst = &nreg->r[creg];
ibc.target = registerUsage[creg];
int shift = instr.getModCond() + RandomX_CurrentConfig.JumpOffset;
ibc.imm = signExtend2sCompl(instr.getImm32()) | (1ULL << shift);
if (RandomX_CurrentConfig.JumpOffset > 0 || shift > 0) //clear the bit below the condition mask - this limits the number of successive jumps to 2
ibc.imm &= ~(1ULL << (shift - 1));
ibc.memMask = RandomX_CurrentConfig.ConditionMask_Calculated << shift;
//mark all registers as used
for (unsigned j = 0; j < RegistersCount; ++j) {
registerUsage[j] = i;
}
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_CFROUND) {
auto src = instr.src % RegistersCount;
ibc.isrc = &nreg->r[src];
ibc.type = InstructionType::CFROUND;
ibc.imm = instr.getImm32() & 63;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_ISTORE) {
auto dst = instr.dst % RegistersCount;
auto src = instr.src % RegistersCount;
ibc.type = InstructionType::ISTORE;
ibc.idst = &nreg->r[dst];
ibc.isrc = &nreg->r[src];
ibc.imm = signExtend2sCompl(instr.getImm32());
if (instr.getModCond() < StoreL3Condition)
ibc.memMask = (instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
else
ibc.memMask = ScratchpadL3Mask;
return;
}
if (opcode < RandomX_CurrentConfig.CEIL_NOP) {
ibc.type = InstructionType::NOP;
return;
}
UNREACHABLE;
}
}

View file

@ -0,0 +1,288 @@
/*
Copyright (c) 2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include "common.hpp"
#include "intrin_portable.h"
#include "instruction.hpp"
#include "program.hpp"
namespace randomx {
//register file in machine byte order
struct NativeRegisterFile {
int_reg_t r[RegistersCount] = { 0 };
rx_vec_f128 f[RegisterCountFlt];
rx_vec_f128 e[RegisterCountFlt];
rx_vec_f128 a[RegisterCountFlt];
};
struct InstructionByteCode {
union {
int_reg_t* idst;
rx_vec_f128* fdst;
};
union {
const int_reg_t* isrc;
const rx_vec_f128* fsrc;
};
union {
uint64_t imm;
int64_t simm;
};
InstructionType type;
union {
int16_t target;
uint16_t shift;
};
uint32_t memMask;
};
#define RANDOMX_EXE_ARGS InstructionByteCode& ibc, int& pc, uint8_t* scratchpad, ProgramConfiguration& config
#define RANDOMX_GEN_ARGS Instruction& instr, int i, InstructionByteCode& ibc
class BytecodeMachine;
typedef void(BytecodeMachine::*InstructionGenBytecode)(RANDOMX_GEN_ARGS);
class BytecodeMachine {
public:
void beginCompilation(NativeRegisterFile& regFile) {
for (unsigned i = 0; i < RegistersCount; ++i) {
registerUsage[i] = -1;
}
nreg = &regFile;
}
void compileProgram(Program& program, InstructionByteCode* bytecode, NativeRegisterFile& regFile) {
beginCompilation(regFile);
for (unsigned i = 0; i < RandomX_CurrentConfig.ProgramSize; ++i) {
auto& instr = program(i);
auto& ibc = bytecode[i];
compileInstruction(instr, i, ibc);
}
}
static void executeBytecode(InstructionByteCode* bytecode, uint8_t* scratchpad, ProgramConfiguration& config) {
for (int pc = 0; pc < RandomX_CurrentConfig.ProgramSize; ++pc) {
auto& ibc = bytecode[pc];
executeInstruction(ibc, pc, scratchpad, config);
}
}
void compileInstruction(RANDOMX_GEN_ARGS)
#ifdef RANDOMX_GEN_TABLE
{
auto generator = genTable[instr.opcode];
(this->*generator)(instr, i, ibc);
}
#else
;
#endif
static void executeInstruction(RANDOMX_EXE_ARGS);
static void exe_IADD_RS(RANDOMX_EXE_ARGS) {
*ibc.idst += (*ibc.isrc << ibc.shift) + ibc.imm;
}
static void exe_IADD_M(RANDOMX_EXE_ARGS) {
*ibc.idst += load64(getScratchpadAddress(ibc, scratchpad));
}
static void exe_ISUB_R(RANDOMX_EXE_ARGS) {
*ibc.idst -= *ibc.isrc;
}
static void exe_ISUB_M(RANDOMX_EXE_ARGS) {
*ibc.idst -= load64(getScratchpadAddress(ibc, scratchpad));
}
static void exe_IMUL_R(RANDOMX_EXE_ARGS) {
*ibc.idst *= *ibc.isrc;
}
static void exe_IMUL_M(RANDOMX_EXE_ARGS) {
*ibc.idst *= load64(getScratchpadAddress(ibc, scratchpad));
}
static void exe_IMULH_R(RANDOMX_EXE_ARGS) {
*ibc.idst = mulh(*ibc.idst, *ibc.isrc);
}
static void exe_IMULH_M(RANDOMX_EXE_ARGS) {
*ibc.idst = mulh(*ibc.idst, load64(getScratchpadAddress(ibc, scratchpad)));
}
static void exe_ISMULH_R(RANDOMX_EXE_ARGS) {
*ibc.idst = smulh(unsigned64ToSigned2sCompl(*ibc.idst), unsigned64ToSigned2sCompl(*ibc.isrc));
}
static void exe_ISMULH_M(RANDOMX_EXE_ARGS) {
*ibc.idst = smulh(unsigned64ToSigned2sCompl(*ibc.idst), unsigned64ToSigned2sCompl(load64(getScratchpadAddress(ibc, scratchpad))));
}
static void exe_INEG_R(RANDOMX_EXE_ARGS) {
*ibc.idst = ~(*ibc.idst) + 1; //two's complement negative
}
static void exe_IXOR_R(RANDOMX_EXE_ARGS) {
*ibc.idst ^= *ibc.isrc;
}
static void exe_IXOR_M(RANDOMX_EXE_ARGS) {
*ibc.idst ^= load64(getScratchpadAddress(ibc, scratchpad));
}
static void exe_IROR_R(RANDOMX_EXE_ARGS) {
*ibc.idst = rotr64(*ibc.idst, *ibc.isrc & 63);
}
static void exe_IROL_R(RANDOMX_EXE_ARGS) {
*ibc.idst = rotl64(*ibc.idst, *ibc.isrc & 63);
}
static void exe_ISWAP_R(RANDOMX_EXE_ARGS) {
int_reg_t temp = *ibc.isrc;
*(int_reg_t*)ibc.isrc = *ibc.idst;
*ibc.idst = temp;
}
static void exe_FSWAP_R(RANDOMX_EXE_ARGS) {
*ibc.fdst = rx_swap_vec_f128(*ibc.fdst);
}
static void exe_FADD_R(RANDOMX_EXE_ARGS) {
*ibc.fdst = rx_add_vec_f128(*ibc.fdst, *ibc.fsrc);
}
static void exe_FADD_M(RANDOMX_EXE_ARGS) {
rx_vec_f128 fsrc = rx_cvt_packed_int_vec_f128(getScratchpadAddress(ibc, scratchpad));
*ibc.fdst = rx_add_vec_f128(*ibc.fdst, fsrc);
}
static void exe_FSUB_R(RANDOMX_EXE_ARGS) {
*ibc.fdst = rx_sub_vec_f128(*ibc.fdst, *ibc.fsrc);
}
static void exe_FSUB_M(RANDOMX_EXE_ARGS) {
rx_vec_f128 fsrc = rx_cvt_packed_int_vec_f128(getScratchpadAddress(ibc, scratchpad));
*ibc.fdst = rx_sub_vec_f128(*ibc.fdst, fsrc);
}
static void exe_FSCAL_R(RANDOMX_EXE_ARGS) {
const rx_vec_f128 mask = rx_set1_vec_f128(0x80F0000000000000);
*ibc.fdst = rx_xor_vec_f128(*ibc.fdst, mask);
}
static void exe_FMUL_R(RANDOMX_EXE_ARGS) {
*ibc.fdst = rx_mul_vec_f128(*ibc.fdst, *ibc.fsrc);
}
static void exe_FDIV_M(RANDOMX_EXE_ARGS) {
rx_vec_f128 fsrc = maskRegisterExponentMantissa(
config,
rx_cvt_packed_int_vec_f128(getScratchpadAddress(ibc, scratchpad))
);
*ibc.fdst = rx_div_vec_f128(*ibc.fdst, fsrc);
}
static void exe_FSQRT_R(RANDOMX_EXE_ARGS) {
*ibc.fdst = rx_sqrt_vec_f128(*ibc.fdst);
}
static void exe_CBRANCH(RANDOMX_EXE_ARGS) {
*ibc.idst += ibc.imm;
if ((*ibc.idst & ibc.memMask) == 0) {
pc = ibc.target;
}
}
static void exe_CFROUND(RANDOMX_EXE_ARGS) {
rx_set_rounding_mode(rotr64(*ibc.isrc, ibc.imm) % 4);
}
static void exe_ISTORE(RANDOMX_EXE_ARGS) {
store64(scratchpad + ((*ibc.idst + ibc.imm) & ibc.memMask), *ibc.isrc);
}
protected:
static rx_vec_f128 maskRegisterExponentMantissa(ProgramConfiguration& config, rx_vec_f128 x) {
const rx_vec_f128 xmantissaMask = rx_set_vec_f128(dynamicMantissaMask, dynamicMantissaMask);
const rx_vec_f128 xexponentMask = rx_load_vec_f128((const double*)&config.eMask);
x = rx_and_vec_f128(x, xmantissaMask);
x = rx_or_vec_f128(x, xexponentMask);
return x;
}
private:
static const int_reg_t zero;
int registerUsage[RegistersCount];
NativeRegisterFile* nreg;
static void* getScratchpadAddress(InstructionByteCode& ibc, uint8_t* scratchpad) {
uint32_t addr = (*ibc.isrc + ibc.imm) & ibc.memMask;
return scratchpad + addr;
}
#ifdef RANDOMX_GEN_TABLE
static InstructionGenBytecode genTable[256];
void gen_IADD_RS(RANDOMX_GEN_ARGS);
void gen_IADD_M(RANDOMX_GEN_ARGS);
void gen_ISUB_R(RANDOMX_GEN_ARGS);
void gen_ISUB_M(RANDOMX_GEN_ARGS);
void gen_IMUL_R(RANDOMX_GEN_ARGS);
void gen_IMUL_M(RANDOMX_GEN_ARGS);
void gen_IMULH_R(RANDOMX_GEN_ARGS);
void gen_IMULH_M(RANDOMX_GEN_ARGS);
void gen_ISMULH_R(RANDOMX_GEN_ARGS);
void gen_ISMULH_M(RANDOMX_GEN_ARGS);
void gen_IMUL_RCP(RANDOMX_GEN_ARGS);
void gen_INEG_R(RANDOMX_GEN_ARGS);
void gen_IXOR_R(RANDOMX_GEN_ARGS);
void gen_IXOR_M(RANDOMX_GEN_ARGS);
void gen_IROR_R(RANDOMX_GEN_ARGS);
void gen_IROL_R(RANDOMX_GEN_ARGS);
void gen_ISWAP_R(RANDOMX_GEN_ARGS);
void gen_FSWAP_R(RANDOMX_GEN_ARGS);
void gen_FADD_R(RANDOMX_GEN_ARGS);
void gen_FADD_M(RANDOMX_GEN_ARGS);
void gen_FSUB_R(RANDOMX_GEN_ARGS);
void gen_FSUB_M(RANDOMX_GEN_ARGS);
void gen_FSCAL_R(RANDOMX_GEN_ARGS);
void gen_FMUL_R(RANDOMX_GEN_ARGS);
void gen_FDIV_M(RANDOMX_GEN_ARGS);
void gen_FSQRT_R(RANDOMX_GEN_ARGS);
void gen_CBRANCH(RANDOMX_GEN_ARGS);
void gen_CFROUND(RANDOMX_GEN_ARGS);
void gen_ISTORE(RANDOMX_GEN_ARGS);
void gen_NOP(RANDOMX_GEN_ARGS);
#endif
};
}

View file

@ -0,0 +1,173 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include <iostream>
#include <climits>
#include "blake2/endian.h"
#include "configuration.h"
#include "randomx.h"
namespace randomx {
//static_assert(RANDOMX_ARGON_MEMORY > 0, "RANDOMX_ARGON_MEMORY must be greater than 0.");
//static_assert((RANDOMX_ARGON_MEMORY & (RANDOMX_ARGON_MEMORY - 1)) == 0, "RANDOMX_ARGON_MEMORY must be a power of 2.");
//static_assert(RANDOMX_DATASET_BASE_SIZE >= 64, "RANDOMX_DATASET_BASE_SIZE must be at least 64.");
//static_assert((RANDOMX_DATASET_BASE_SIZE & (RANDOMX_DATASET_BASE_SIZE - 1)) == 0, "RANDOMX_DATASET_BASE_SIZE must be a power of 2.");
//static_assert(RANDOMX_DATASET_BASE_SIZE <= 4294967296ULL, "RANDOMX_DATASET_BASE_SIZE must not exceed 4294967296.");
//static_assert(RANDOMX_DATASET_EXTRA_SIZE % 64 == 0, "RANDOMX_DATASET_EXTRA_SIZE must be divisible by 64.");
//static_assert((uint64_t)RANDOMX_DATASET_BASE_SIZE + RANDOMX_DATASET_EXTRA_SIZE <= 17179869184, "Dataset size must not exceed 16 GiB.");
//static_assert(RANDOMX_PROGRAM_SIZE > 0, "RANDOMX_PROGRAM_SIZE must be greater than 0");
//static_assert(RANDOMX_PROGRAM_ITERATIONS > 0, "RANDOMX_PROGRAM_ITERATIONS must be greater than 0");
//static_assert(RANDOMX_PROGRAM_COUNT > 0, "RANDOMX_PROGRAM_COUNT must be greater than 0");
//static_assert((RANDOMX_SCRATCHPAD_L3 & (RANDOMX_SCRATCHPAD_L3 - 1)) == 0, "RANDOMX_SCRATCHPAD_L3 must be a power of 2.");
//static_assert(RANDOMX_SCRATCHPAD_L3 >= RANDOMX_SCRATCHPAD_L2, "RANDOMX_SCRATCHPAD_L3 must be greater than or equal to RANDOMX_SCRATCHPAD_L2.");
//static_assert((RANDOMX_SCRATCHPAD_L2 & (RANDOMX_SCRATCHPAD_L2 - 1)) == 0, "RANDOMX_SCRATCHPAD_L2 must be a power of 2.");
//static_assert(RANDOMX_SCRATCHPAD_L2 >= RANDOMX_SCRATCHPAD_L1, "RANDOMX_SCRATCHPAD_L2 must be greater than or equal to RANDOMX_SCRATCHPAD_L1.");
//static_assert(RANDOMX_SCRATCHPAD_L1 >= 64, "RANDOMX_SCRATCHPAD_L1 must be at least 64.");
//static_assert((RANDOMX_SCRATCHPAD_L1 & (RANDOMX_SCRATCHPAD_L1 - 1)) == 0, "RANDOMX_SCRATCHPAD_L1 must be a power of 2.");
//static_assert(RANDOMX_CACHE_ACCESSES > 1, "RANDOMX_CACHE_ACCESSES must be greater than 1");
//static_assert(RANDOMX_SUPERSCALAR_LATENCY > 0, "RANDOMX_SUPERSCALAR_LATENCY must be greater than 0");
//static_assert(RANDOMX_JUMP_BITS > 0, "RANDOMX_JUMP_BITS must be greater than 0.");
//static_assert(RANDOMX_JUMP_OFFSET >= 0, "RANDOMX_JUMP_OFFSET must be greater than or equal to 0.");
//static_assert(RANDOMX_JUMP_BITS + RANDOMX_JUMP_OFFSET <= 16, "RANDOMX_JUMP_BITS + RANDOMX_JUMP_OFFSET must not exceed 16.");
/*constexpr int wtSum = RANDOMX_FREQ_IADD_RS + RANDOMX_FREQ_IADD_M + RANDOMX_FREQ_ISUB_R + \
RANDOMX_FREQ_ISUB_M + RANDOMX_FREQ_IMUL_R + RANDOMX_FREQ_IMUL_M + RANDOMX_FREQ_IMULH_R + \
RANDOMX_FREQ_IMULH_M + RANDOMX_FREQ_ISMULH_R + RANDOMX_FREQ_ISMULH_M + RANDOMX_FREQ_IMUL_RCP + \
RANDOMX_FREQ_INEG_R + RANDOMX_FREQ_IXOR_R + RANDOMX_FREQ_IXOR_M + RANDOMX_FREQ_IROR_R + RANDOMX_FREQ_IROL_R + RANDOMX_FREQ_ISWAP_R + \
RANDOMX_FREQ_FSWAP_R + RANDOMX_FREQ_FADD_R + RANDOMX_FREQ_FADD_M + RANDOMX_FREQ_FSUB_R + RANDOMX_FREQ_FSUB_M + \
RANDOMX_FREQ_FSCAL_R + RANDOMX_FREQ_FMUL_R + RANDOMX_FREQ_FDIV_M + RANDOMX_FREQ_FSQRT_R + RANDOMX_FREQ_CBRANCH + \
RANDOMX_FREQ_CFROUND + RANDOMX_FREQ_ISTORE + RANDOMX_FREQ_NOP;*/
//static_assert(wtSum == 256, "Sum of instruction frequencies must be 256.");
constexpr uint32_t ArgonBlockSize = 1024;
constexpr int SuperscalarMaxSize = 3 * RANDOMX_SUPERSCALAR_MAX_LATENCY + 2;
constexpr size_t CacheLineSize = RANDOMX_DATASET_ITEM_SIZE;
#define ScratchpadSize RandomX_CurrentConfig.ScratchpadL3_Size
#define CacheLineAlignMask RandomX_CurrentConfig.CacheLineAlignMask_Calculated
#define DatasetExtraItems RandomX_CurrentConfig.DatasetExtraItems_Calculated
constexpr int StoreL3Condition = 14;
//Prevent some unsafe configurations.
#ifndef RANDOMX_UNSAFE
//static_assert((uint64_t)ArgonBlockSize * RANDOMX_CACHE_ACCESSES * RANDOMX_ARGON_MEMORY + 33554432 >= (uint64_t)RANDOMX_DATASET_BASE_SIZE + RANDOMX_DATASET_EXTRA_SIZE, "Unsafe configuration: Memory-time tradeoffs");
//static_assert((128 + RANDOMX_PROGRAM_SIZE * RANDOMX_FREQ_ISTORE / 256) * (RANDOMX_PROGRAM_COUNT * RANDOMX_PROGRAM_ITERATIONS) >= RANDOMX_SCRATCHPAD_L3, "Unsafe configuration: Insufficient Scratchpad writes");
//static_assert(RANDOMX_PROGRAM_COUNT > 1, "Unsafe configuration: Program filtering strategies");
//static_assert(RANDOMX_PROGRAM_SIZE >= 64, "Unsafe configuration: Low program entropy");
//static_assert(RANDOMX_PROGRAM_ITERATIONS >= 400, "Unsafe configuration: High compilation overhead");
#endif
#ifdef TRACE
constexpr bool trace = true;
#else
constexpr bool trace = false;
#endif
#ifndef UNREACHABLE
#ifdef __GNUC__
#define UNREACHABLE __builtin_unreachable()
#elif _MSC_VER
#define UNREACHABLE __assume(false)
#else
#define UNREACHABLE
#endif
#endif
#if defined(_M_X64) || defined(__x86_64__)
#define RANDOMX_HAVE_COMPILER 1
class JitCompilerX86;
using JitCompiler = JitCompilerX86;
#elif defined(__aarch64__)
#define RANDOMX_HAVE_COMPILER 0
class JitCompilerA64;
using JitCompiler = JitCompilerA64;
#else
#define RANDOMX_HAVE_COMPILER 0
class JitCompilerFallback;
using JitCompiler = JitCompilerFallback;
#endif
using addr_t = uint32_t;
using int_reg_t = uint64_t;
struct fpu_reg_t {
double lo;
double hi;
};
#define ScratchpadL1Mask RandomX_CurrentConfig.ScratchpadL1Mask_Calculated
#define ScratchpadL1Mask16 RandomX_CurrentConfig.ScratchpadL1Mask16_Calculated
#define ScratchpadL2Mask RandomX_CurrentConfig.ScratchpadL2Mask_Calculated
#define ScratchpadL2Mask16 RandomX_CurrentConfig.ScratchpadL2Mask16_Calculated
#define ScratchpadL3Mask RandomX_CurrentConfig.ScratchpadL3Mask_Calculated
#define ScratchpadL3Mask64 RandomX_CurrentConfig.ScratchpadL3Mask64_Calculated
constexpr int RegistersCount = 8;
constexpr int RegisterCountFlt = RegistersCount / 2;
constexpr int RegisterNeedsDisplacement = 5; //x86 r13 register
constexpr int RegisterNeedsSib = 4; //x86 r12 register
inline bool isPowerOf2(uint64_t x) {
return (x & (x - 1)) == 0;
}
constexpr int mantissaSize = 52;
constexpr int exponentSize = 11;
constexpr uint64_t mantissaMask = (1ULL << mantissaSize) - 1;
constexpr uint64_t exponentMask = (1ULL << exponentSize) - 1;
constexpr int exponentBias = 1023;
constexpr int dynamicExponentBits = 4;
constexpr int staticExponentBits = 4;
constexpr uint64_t constExponentBits = 0x300;
constexpr uint64_t dynamicMantissaMask = (1ULL << (mantissaSize + dynamicExponentBits)) - 1;
struct MemoryRegisters {
addr_t mx, ma;
uint8_t* memory = nullptr;
};
//register file in little-endian byte order
struct RegisterFile {
int_reg_t r[RegistersCount];
fpu_reg_t f[RegisterCountFlt];
fpu_reg_t e[RegisterCountFlt];
fpu_reg_t a[RegisterCountFlt];
};
typedef void(ProgramFunc)(RegisterFile&, MemoryRegisters&, uint8_t* /* scratchpad */, uint64_t);
typedef void(DatasetInitFunc)(randomx_cache* cache, uint8_t* dataset, uint32_t startBlock, uint32_t endBlock);
typedef void(DatasetDeallocFunc)(randomx_dataset*);
typedef void(CacheDeallocFunc)(randomx_cache*);
typedef void(CacheInitializeFunc)(randomx_cache*, const void*, size_t);
}

View file

@ -0,0 +1,47 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
// Increase it if some configs use more cache accesses
#define RANDOMX_CACHE_MAX_ACCESSES 16
// Increase it if some configs use larger superscalar latency
#define RANDOMX_SUPERSCALAR_MAX_LATENCY 256
// Increase it if some configs use larger cache
#define RANDOMX_CACHE_MAX_SIZE 268435456
// Increase it if some configs use larger dataset
#define RANDOMX_DATASET_MAX_SIZE 2181038080
// Increase it if some configs use larger programs
#define RANDOMX_PROGRAM_MAX_SIZE 512
// Increase it if some configs use larger scratchpad
#define RANDOMX_SCRATCHPAD_L3_MAX_SIZE 2097152

View file

@ -0,0 +1,189 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/* Original code from Argon2 reference source code package used under CC0 Licence
* https://github.com/P-H-C/phc-winner-argon2
* Copyright 2015
* Daniel Dinu, Dmitry Khovratovich, Jean-Philippe Aumasson, and Samuel Neves
*/
#include <new>
#include <algorithm>
#include <stdexcept>
#include <cstring>
#include <limits>
#include <cstring>
#include "common.hpp"
#include "dataset.hpp"
#include "virtual_memory.hpp"
#include "superscalar.hpp"
#include "blake2_generator.hpp"
#include "reciprocal.h"
#include "blake2/endian.h"
#include "argon2.h"
#include "argon2_core.h"
#include "jit_compiler.hpp"
#include "intrin_portable.h"
//static_assert(RANDOMX_ARGON_MEMORY % (RANDOMX_ARGON_LANES * ARGON2_SYNC_POINTS) == 0, "RANDOMX_ARGON_MEMORY - invalid value");
static_assert(ARGON2_BLOCK_SIZE == randomx::ArgonBlockSize, "Unpexpected value of ARGON2_BLOCK_SIZE");
namespace randomx {
template<class Allocator>
void deallocCache(randomx_cache* cache) {
if (cache->memory != nullptr)
Allocator::freeMemory(cache->memory, RANDOMX_CACHE_MAX_SIZE);
if (cache->jit != nullptr)
delete cache->jit;
}
template void deallocCache<DefaultAllocator>(randomx_cache* cache);
template void deallocCache<LargePageAllocator>(randomx_cache* cache);
void initCache(randomx_cache* cache, const void* key, size_t keySize) {
uint32_t memory_blocks, segment_length;
argon2_instance_t instance;
argon2_context context;
context.out = nullptr;
context.outlen = 0;
context.pwd = CONST_CAST(uint8_t *)key;
context.pwdlen = (uint32_t)keySize;
context.salt = CONST_CAST(uint8_t *)RandomX_CurrentConfig.ArgonSalt;
context.saltlen = (uint32_t)strlen(RandomX_CurrentConfig.ArgonSalt);
context.secret = NULL;
context.secretlen = 0;
context.ad = NULL;
context.adlen = 0;
context.t_cost = RandomX_CurrentConfig.ArgonIterations;
context.m_cost = RandomX_CurrentConfig.ArgonMemory;
context.lanes = RandomX_CurrentConfig.ArgonLanes;
context.threads = 1;
context.allocate_cbk = NULL;
context.free_cbk = NULL;
context.flags = ARGON2_DEFAULT_FLAGS;
context.version = ARGON2_VERSION_NUMBER;
/* 2. Align memory size */
/* Minimum memory_blocks = 8L blocks, where L is the number of lanes */
memory_blocks = context.m_cost;
segment_length = memory_blocks / (context.lanes * ARGON2_SYNC_POINTS);
instance.version = context.version;
instance.memory = NULL;
instance.passes = context.t_cost;
instance.memory_blocks = memory_blocks;
instance.segment_length = segment_length;
instance.lane_length = segment_length * ARGON2_SYNC_POINTS;
instance.lanes = context.lanes;
instance.threads = context.threads;
instance.type = Argon2_d;
instance.memory = (block*)cache->memory;
if (instance.threads > instance.lanes) {
instance.threads = instance.lanes;
}
/* 3. Initialization: Hashing inputs, allocating memory, filling first
* blocks
*/
rxa2_argon_initialize(&instance, &context);
rxa2_fill_memory_blocks(&instance);
cache->reciprocalCache.clear();
randomx::Blake2Generator gen(key, keySize);
for (int i = 0; i < RandomX_CurrentConfig.CacheAccesses; ++i) {
randomx::generateSuperscalar(cache->programs[i], gen);
for (unsigned j = 0; j < cache->programs[i].getSize(); ++j) {
auto& instr = cache->programs[i](j);
if ((SuperscalarInstructionType)instr.opcode == SuperscalarInstructionType::IMUL_RCP) {
auto rcp = randomx_reciprocal(instr.getImm32());
instr.setImm32(cache->reciprocalCache.size());
cache->reciprocalCache.push_back(rcp);
}
}
}
}
void initCacheCompile(randomx_cache* cache, const void* key, size_t keySize) {
initCache(cache, key, keySize);
cache->jit->generateSuperscalarHash(cache->programs, cache->reciprocalCache);
cache->jit->generateDatasetInitCode();
}
constexpr uint64_t superscalarMul0 = 6364136223846793005ULL;
constexpr uint64_t superscalarAdd1 = 9298411001130361340ULL;
constexpr uint64_t superscalarAdd2 = 12065312585734608966ULL;
constexpr uint64_t superscalarAdd3 = 9306329213124626780ULL;
constexpr uint64_t superscalarAdd4 = 5281919268842080866ULL;
constexpr uint64_t superscalarAdd5 = 10536153434571861004ULL;
constexpr uint64_t superscalarAdd6 = 3398623926847679864ULL;
constexpr uint64_t superscalarAdd7 = 9549104520008361294ULL;
static inline uint8_t* getMixBlock(uint64_t registerValue, uint8_t *memory) {
const uint32_t mask = (RandomX_CurrentConfig.ArgonMemory * randomx::ArgonBlockSize) / CacheLineSize - 1;
return memory + (registerValue & mask) * CacheLineSize;
}
void initDatasetItem(randomx_cache* cache, uint8_t* out, uint64_t itemNumber) {
int_reg_t rl[8];
uint8_t* mixBlock;
uint64_t registerValue = itemNumber;
rl[0] = (itemNumber + 1) * superscalarMul0;
rl[1] = rl[0] ^ superscalarAdd1;
rl[2] = rl[0] ^ superscalarAdd2;
rl[3] = rl[0] ^ superscalarAdd3;
rl[4] = rl[0] ^ superscalarAdd4;
rl[5] = rl[0] ^ superscalarAdd5;
rl[6] = rl[0] ^ superscalarAdd6;
rl[7] = rl[0] ^ superscalarAdd7;
for (unsigned i = 0; i < RandomX_CurrentConfig.CacheAccesses; ++i) {
mixBlock = getMixBlock(registerValue, cache->memory);
rx_prefetch_nta(mixBlock);
SuperscalarProgram& prog = cache->programs[i];
executeSuperscalar(rl, prog, &cache->reciprocalCache);
for (unsigned q = 0; q < 8; ++q)
rl[q] ^= load64_native(mixBlock + 8 * q);
registerValue = rl[prog.getAddressRegister()];
}
memcpy(out, &rl, CacheLineSize);
}
void initDataset(randomx_cache* cache, uint8_t* dataset, uint32_t startItem, uint32_t endItem) {
for (uint32_t itemNumber = startItem; itemNumber < endItem; ++itemNumber, dataset += CacheLineSize)
initDatasetItem(cache, dataset, itemNumber);
}
}

View file

@ -0,0 +1,80 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include <vector>
#include <type_traits>
#include "common.hpp"
#include "superscalar_program.hpp"
#include "allocator.hpp"
/* Global scope for C binding */
struct randomx_dataset {
uint8_t* memory = nullptr;
randomx::DatasetDeallocFunc* dealloc;
};
/* Global scope for C binding */
struct randomx_cache {
uint8_t* memory = nullptr;
randomx::CacheDeallocFunc* dealloc;
randomx::JitCompiler* jit;
randomx::CacheInitializeFunc* initialize;
randomx::DatasetInitFunc* datasetInit;
randomx::SuperscalarProgram programs[RANDOMX_CACHE_MAX_ACCESSES];
std::vector<uint64_t> reciprocalCache;
bool isInitialized() {
return programs[0].getSize() != 0;
}
};
//A pointer to a standard-layout struct object points to its initial member
static_assert(std::is_standard_layout<randomx_dataset>(), "randomx_dataset must be a standard-layout struct");
static_assert(std::is_standard_layout<randomx_cache>(), "randomx_cache must be a standard-layout struct");
namespace randomx {
using DefaultAllocator = AlignedAllocator<CacheLineSize>;
template<class Allocator>
void deallocDataset(randomx_dataset* dataset) {
if (dataset->memory != nullptr)
Allocator::freeMemory(dataset->memory, RANDOMX_DATASET_MAX_SIZE);
}
template<class Allocator>
void deallocCache(randomx_cache* cache);
void initCache(randomx_cache*, const void*, size_t);
void initCacheCompile(randomx_cache*, const void*, size_t);
void initDatasetItem(randomx_cache* cache, uint8_t* out, uint64_t blockNumber);
void initDataset(randomx_cache* cache, uint8_t* dataset, uint32_t startBlock, uint32_t endBlock);
}

View file

@ -0,0 +1,102 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include <type_traits>
#include "blake2/endian.h"
namespace randomx {
class Instruction;
enum class InstructionType : uint16_t {
IADD_RS = 0,
IADD_M = 1,
ISUB_R = 2,
ISUB_M = 3,
IMUL_R = 4,
IMUL_M = 5,
IMULH_R = 6,
IMULH_M = 7,
ISMULH_R = 8,
ISMULH_M = 9,
IMUL_RCP = 10,
INEG_R = 11,
IXOR_R = 12,
IXOR_M = 13,
IROR_R = 14,
IROL_R = 15,
ISWAP_R = 16,
FSWAP_R = 17,
FADD_R = 18,
FADD_M = 19,
FSUB_R = 20,
FSUB_M = 21,
FSCAL_R = 22,
FMUL_R = 23,
FDIV_M = 24,
FSQRT_R = 25,
CBRANCH = 26,
CFROUND = 27,
ISTORE = 28,
NOP = 29,
};
class Instruction {
public:
uint32_t getImm32() const {
return load32(&imm32);
}
void setImm32(uint32_t val) {
return store32(&imm32, val);
}
int getModMem() const {
return mod % 4; //bits 0-1
}
int getModShift() const {
return (mod >> 2) % 4; //bits 2-3
}
int getModCond() const {
return mod >> 4; //bits 4-7
}
void setMod(uint8_t val) {
mod = val;
}
uint8_t opcode;
uint8_t dst;
uint8_t src;
uint8_t mod;
uint32_t imm32;
};
static_assert(sizeof(Instruction) == 8, "Invalid size of struct randomx::Instruction");
static_assert(std::is_standard_layout<Instruction>(), "randomx::Instruction must be a standard-layout struct");
}

View file

@ -0,0 +1,193 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <cfenv>
#include <cmath>
#include "common.hpp"
#include "intrin_portable.h"
#include "blake2/endian.h"
#if defined(__SIZEOF_INT128__)
typedef unsigned __int128 uint128_t;
typedef __int128 int128_t;
uint64_t mulh(uint64_t a, uint64_t b) {
return ((uint128_t)a * b) >> 64;
}
int64_t smulh(int64_t a, int64_t b) {
return ((int128_t)a * b) >> 64;
}
#define HAVE_MULH
#define HAVE_SMULH
#endif
#if defined(_MSC_VER)
#define HAS_VALUE(X) X ## 0
#define EVAL_DEFINE(X) HAS_VALUE(X)
#include <intrin.h>
#include <stdlib.h>
uint64_t rotl64(uint64_t x, unsigned int c) {
return _rotl64(x, c);
}
uint64_t rotr64(uint64_t x, unsigned int c) {
return _rotr64(x, c);
}
#define HAVE_ROTL64
#define HAVE_ROTR64
#if EVAL_DEFINE(__MACHINEARM64_X64(1))
uint64_t mulh(uint64_t a, uint64_t b) {
return __umulh(a, b);
}
#define HAVE_MULH
#endif
#if EVAL_DEFINE(__MACHINEX64(1))
int64_t smulh(int64_t a, int64_t b) {
int64_t hi;
_mul128(a, b, &hi);
return hi;
}
#define HAVE_SMULH
#endif
static void setRoundMode_(uint32_t mode) {
_controlfp(mode, _MCW_RC);
}
#define HAVE_SETROUNDMODE_IMPL
#endif
#ifndef HAVE_SETROUNDMODE_IMPL
static void setRoundMode_(uint32_t mode) {
fesetround(mode);
}
#endif
#ifndef HAVE_ROTR64
uint64_t rotr64(uint64_t a, unsigned int b) {
return (a >> b) | (a << (-b & 63));
}
#define HAVE_ROTR64
#endif
#ifndef HAVE_ROTL64
uint64_t rotl64(uint64_t a, unsigned int b) {
return (a << b) | (a >> (-b & 63));
}
#define HAVE_ROTL64
#endif
#ifndef HAVE_MULH
#define LO(x) ((x)&0xffffffff)
#define HI(x) ((x)>>32)
uint64_t mulh(uint64_t a, uint64_t b) {
uint64_t ah = HI(a), al = LO(a);
uint64_t bh = HI(b), bl = LO(b);
uint64_t x00 = al * bl;
uint64_t x01 = al * bh;
uint64_t x10 = ah * bl;
uint64_t x11 = ah * bh;
uint64_t m1 = LO(x10) + LO(x01) + HI(x00);
uint64_t m2 = HI(x10) + HI(x01) + LO(x11) + HI(m1);
uint64_t m3 = HI(x11) + HI(m2);
return (m3 << 32) + LO(m2);
}
#define HAVE_MULH
#endif
#ifndef HAVE_SMULH
int64_t smulh(int64_t a, int64_t b) {
int64_t hi = mulh(a, b);
if (a < 0LL) hi -= b;
if (b < 0LL) hi -= a;
return hi;
}
#define HAVE_SMULH
#endif
#ifdef RANDOMX_DEFAULT_FENV
void rx_reset_float_state() {
setRoundMode_(FE_TONEAREST);
rx_set_double_precision(); //set precision to 53 bits if needed by the platform
}
void rx_set_rounding_mode(uint32_t mode) {
switch (mode & 3) {
case RoundDown:
setRoundMode_(FE_DOWNWARD);
break;
case RoundUp:
setRoundMode_(FE_UPWARD);
break;
case RoundToZero:
setRoundMode_(FE_TOWARDZERO);
break;
case RoundToNearest:
setRoundMode_(FE_TONEAREST);
break;
default:
UNREACHABLE;
}
}
#endif
#ifdef RANDOMX_USE_X87
#ifdef _M_IX86
void rx_set_double_precision() {
_control87(_PC_53, _MCW_PC);
}
#elif defined(__i386)
void rx_set_double_precision() {
uint16_t volatile x87cw;
asm volatile("fstcw %0" : "=m" (x87cw));
x87cw &= ~0x300;
x87cw |= 0x200;
asm volatile("fldcw %0" : : "m" (x87cw));
}
#endif
#endif //RANDOMX_USE_X87
union double_ser_t {
double f;
uint64_t i;
};
double loadDoublePortable(const void* addr) {
double_ser_t ds;
ds.i = load64(addr);
return ds.f;
}

View file

@ -0,0 +1,605 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include "blake2/endian.h"
constexpr int32_t unsigned32ToSigned2sCompl(uint32_t x) {
return (-1 == ~0) ? (int32_t)x : (x > INT32_MAX ? (-(int32_t)(UINT32_MAX - x) - 1) : (int32_t)x);
}
constexpr int64_t unsigned64ToSigned2sCompl(uint64_t x) {
return (-1 == ~0) ? (int64_t)x : (x > INT64_MAX ? (-(int64_t)(UINT64_MAX - x) - 1) : (int64_t)x);
}
constexpr uint64_t signExtend2sCompl(uint32_t x) {
return (-1 == ~0) ? (int64_t)(int32_t)(x) : (x > INT32_MAX ? (x | 0xffffffff00000000ULL) : (uint64_t)x);
}
constexpr int RoundToNearest = 0;
constexpr int RoundDown = 1;
constexpr int RoundUp = 2;
constexpr int RoundToZero = 3;
//MSVC doesn't define __SSE2__, so we have to define it manually if SSE2 is available
#if !defined(__SSE2__) && (defined(_M_X64) || (defined(_M_IX86_FP) && _M_IX86_FP == 2))
#define __SSE2__ 1
#endif
//MSVC doesn't define __AES__
#if defined(_MSC_VER) && defined(__SSE2__)
#define __AES__
#endif
//the library "sqrt" function provided by MSVC for x86 targets doesn't give
//the correct results, so we have to use inline assembly to call x87 fsqrt directly
#if !defined(__SSE2__)
#if defined(_M_IX86)
inline double __cdecl rx_sqrt(double x) {
__asm {
fld x
fsqrt
}
}
#define rx_sqrt rx_sqrt
void rx_set_double_precision();
#define RANDOMX_USE_X87
#elif defined(__i386)
void rx_set_double_precision();
#define RANDOMX_USE_X87
#endif
#endif //__SSE2__
#if !defined(rx_sqrt)
#define rx_sqrt sqrt
#endif
#if !defined(RANDOMX_USE_X87)
#define rx_set_double_precision(x)
#endif
#ifdef __SSE2__
#ifdef __GNUC__
#include <x86intrin.h>
#else
#include <intrin.h>
#endif
typedef __m128i rx_vec_i128;
typedef __m128d rx_vec_f128;
#define rx_aligned_alloc(a, b) _mm_malloc(a,b)
#define rx_aligned_free(a) _mm_free(a)
#define rx_prefetch_nta(x) _mm_prefetch((const char *)(x), _MM_HINT_NTA)
#define rx_load_vec_f128 _mm_load_pd
#define rx_store_vec_f128 _mm_store_pd
#define rx_add_vec_f128 _mm_add_pd
#define rx_sub_vec_f128 _mm_sub_pd
#define rx_mul_vec_f128 _mm_mul_pd
#define rx_div_vec_f128 _mm_div_pd
#define rx_sqrt_vec_f128 _mm_sqrt_pd
FORCE_INLINE rx_vec_f128 rx_swap_vec_f128(rx_vec_f128 a) {
return _mm_shuffle_pd(a, a, 1);
}
FORCE_INLINE rx_vec_f128 rx_set_vec_f128(uint64_t x1, uint64_t x0) {
return _mm_castsi128_pd(_mm_set_epi64x(x1, x0));
}
FORCE_INLINE rx_vec_f128 rx_set1_vec_f128(uint64_t x) {
return _mm_castsi128_pd(_mm_set1_epi64x(x));
}
#define rx_xor_vec_f128 _mm_xor_pd
#define rx_and_vec_f128 _mm_and_pd
#define rx_or_vec_f128 _mm_or_pd
#ifdef __AES__
#define rx_aesenc_vec_i128 _mm_aesenc_si128
#define rx_aesdec_vec_i128 _mm_aesdec_si128
#define HAVE_AES
#endif //__AES__
FORCE_INLINE int rx_vec_i128_x(rx_vec_i128 a) {
return _mm_cvtsi128_si32(a);
}
FORCE_INLINE int rx_vec_i128_y(rx_vec_i128 a) {
return _mm_cvtsi128_si32(_mm_shuffle_epi32(a, 0x55));
}
FORCE_INLINE int rx_vec_i128_z(rx_vec_i128 a) {
return _mm_cvtsi128_si32(_mm_shuffle_epi32(a, 0xaa));
}
FORCE_INLINE int rx_vec_i128_w(rx_vec_i128 a) {
return _mm_cvtsi128_si32(_mm_shuffle_epi32(a, 0xff));
}
#define rx_set_int_vec_i128 _mm_set_epi32
#define rx_xor_vec_i128 _mm_xor_si128
#define rx_load_vec_i128 _mm_load_si128
#define rx_store_vec_i128 _mm_store_si128
FORCE_INLINE rx_vec_f128 rx_cvt_packed_int_vec_f128(const void* addr) {
__m128i ix = _mm_loadl_epi64((const __m128i*)addr);
return _mm_cvtepi32_pd(ix);
}
constexpr uint32_t rx_mxcsr_default = 0x9FC0; //Flush to zero, denormals are zero, default rounding mode, all exceptions disabled
FORCE_INLINE void rx_reset_float_state() {
_mm_setcsr(rx_mxcsr_default);
}
FORCE_INLINE void rx_set_rounding_mode(uint32_t mode) {
_mm_setcsr(rx_mxcsr_default | (mode << 13));
}
#elif defined(__PPC64__) && defined(__ALTIVEC__) && defined(__VSX__) //sadly only POWER7 and newer will be able to use SIMD acceleration. Earlier processors cant use doubles or 64 bit integers with SIMD
#include <cstdint>
#include <stdexcept>
#include <cstdlib>
#include <altivec.h>
#undef vector
#undef pixel
#undef bool
typedef __vector uint8_t __m128i;
typedef __vector uint32_t __m128l;
typedef __vector int __m128li;
typedef __vector uint64_t __m128ll;
typedef __vector double __m128d;
typedef __m128i rx_vec_i128;
typedef __m128d rx_vec_f128;
typedef union{
rx_vec_i128 i;
rx_vec_f128 d;
uint64_t u64[2];
double d64[2];
uint32_t u32[4];
int i32[4];
} vec_u;
#define rx_aligned_alloc(a, b) malloc(a)
#define rx_aligned_free(a) free(a)
#define rx_prefetch_nta(x)
/* Splat 64-bit long long to 2 64-bit long longs */
FORCE_INLINE __m128i vec_splat2sd (int64_t scalar)
{ return (__m128i) vec_splats (scalar); }
FORCE_INLINE rx_vec_f128 rx_load_vec_f128(const double* pd) {
#if defined(NATIVE_LITTLE_ENDIAN)
return (rx_vec_f128)vec_vsx_ld(0,pd);
#else
vec_u t;
t.u64[0] = load64(pd + 0);
t.u64[1] = load64(pd + 1);
return (rx_vec_f128)t.d;
#endif
}
FORCE_INLINE void rx_store_vec_f128(double* mem_addr, rx_vec_f128 a) {
#if defined(NATIVE_LITTLE_ENDIAN)
vec_vsx_st(a,0,(rx_vec_f128*)mem_addr);
#else
vec_u _a;
_a.d = a;
store64(mem_addr + 0, _a.u64[0]);
store64(mem_addr + 1, _a.u64[1]);
#endif
}
FORCE_INLINE rx_vec_f128 rx_swap_vec_f128(rx_vec_f128 a) {
return (rx_vec_f128)vec_perm((__m128i)a,(__m128i)a,(__m128i){8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7});
}
FORCE_INLINE rx_vec_f128 rx_add_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
return (rx_vec_f128)vec_add(a,b);
}
FORCE_INLINE rx_vec_f128 rx_sub_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
return (rx_vec_f128)vec_sub(a,b);
}
FORCE_INLINE rx_vec_f128 rx_mul_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
return (rx_vec_f128)vec_mul(a,b);
}
FORCE_INLINE rx_vec_f128 rx_div_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
return (rx_vec_f128)vec_div(a,b);
}
FORCE_INLINE rx_vec_f128 rx_sqrt_vec_f128(rx_vec_f128 a) {
return (rx_vec_f128)vec_sqrt(a);
}
FORCE_INLINE rx_vec_i128 rx_set1_long_vec_i128(uint64_t a) {
return (rx_vec_i128)vec_splat2sd(a);
}
FORCE_INLINE rx_vec_f128 rx_vec_i128_vec_f128(rx_vec_i128 a) {
return (rx_vec_f128)a;
}
FORCE_INLINE rx_vec_f128 rx_set_vec_f128(uint64_t x1, uint64_t x0) {
return (rx_vec_f128)(__m128ll){x0,x1};
}
FORCE_INLINE rx_vec_f128 rx_set1_vec_f128(uint64_t x) {
return (rx_vec_f128)vec_splat2sd(x);
}
FORCE_INLINE rx_vec_f128 rx_xor_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
return (rx_vec_f128)vec_xor(a,b);
}
FORCE_INLINE rx_vec_f128 rx_and_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
return (rx_vec_f128)vec_and(a,b);
}
FORCE_INLINE rx_vec_f128 rx_or_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
return (rx_vec_f128)vec_or(a,b);
}
#if defined(__CRYPTO__)
FORCE_INLINE __m128ll vrev(__m128i v){
#if defined(NATIVE_LITTLE_ENDIAN)
return (__m128ll)vec_perm((__m128i)v,(__m128i){0},(__m128i){15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0});
#else
return (__m128ll)vec_perm((__m128i)v,(__m128i){0},(__m128i){3,2,1,0, 7,6,5,4, 11,10,9,8, 15,14,13,12});
#endif
}
FORCE_INLINE rx_vec_i128 rx_aesenc_vec_i128(rx_vec_i128 v, rx_vec_i128 rkey) {
__m128ll _v = vrev(v);
__m128ll _rkey = vrev(rkey);
__m128ll result = vrev((__m128i)__builtin_crypto_vcipher(_v,_rkey));
return (rx_vec_i128)result;
}
FORCE_INLINE rx_vec_i128 rx_aesdec_vec_i128(rx_vec_i128 v, rx_vec_i128 rkey) {
__m128ll _v = vrev(v);
__m128ll zero = (__m128ll){0};
__m128ll out = vrev((__m128i)__builtin_crypto_vncipher(_v,zero));
return (rx_vec_i128)vec_xor((__m128i)out,rkey);
}
#define HAVE_AES
#endif //__CRYPTO__
FORCE_INLINE int rx_vec_i128_x(rx_vec_i128 a) {
vec_u _a;
_a.i = a;
return _a.i32[0];
}
FORCE_INLINE int rx_vec_i128_y(rx_vec_i128 a) {
vec_u _a;
_a.i = a;
return _a.i32[1];
}
FORCE_INLINE int rx_vec_i128_z(rx_vec_i128 a) {
vec_u _a;
_a.i = a;
return _a.i32[2];
}
FORCE_INLINE int rx_vec_i128_w(rx_vec_i128 a) {
vec_u _a;
_a.i = a;
return _a.i32[3];
}
FORCE_INLINE rx_vec_i128 rx_set_int_vec_i128(int _I3, int _I2, int _I1, int _I0) {
return (rx_vec_i128)((__m128li){_I0,_I1,_I2,_I3});
};
FORCE_INLINE rx_vec_i128 rx_xor_vec_i128(rx_vec_i128 _A, rx_vec_i128 _B) {
return (rx_vec_i128)vec_xor(_A,_B);
}
FORCE_INLINE rx_vec_i128 rx_load_vec_i128(rx_vec_i128 const *_P) {
#if defined(NATIVE_LITTLE_ENDIAN)
return *_P;
#else
uint32_t* ptr = (uint32_t*)_P;
vec_u c;
c.u32[0] = load32(ptr + 0);
c.u32[1] = load32(ptr + 1);
c.u32[2] = load32(ptr + 2);
c.u32[3] = load32(ptr + 3);
return (rx_vec_i128)c.i;
#endif
}
FORCE_INLINE void rx_store_vec_i128(rx_vec_i128 *_P, rx_vec_i128 _B) {
#if defined(NATIVE_LITTLE_ENDIAN)
*_P = _B;
#else
uint32_t* ptr = (uint32_t*)_P;
vec_u B;
B.i = _B;
store32(ptr + 0, B.u32[0]);
store32(ptr + 1, B.u32[1]);
store32(ptr + 2, B.u32[2]);
store32(ptr + 3, B.u32[3]);
#endif
}
FORCE_INLINE rx_vec_f128 rx_cvt_packed_int_vec_f128(const void* addr) {
vec_u x;
x.d64[0] = (double)unsigned32ToSigned2sCompl(load32((uint8_t*)addr + 0));
x.d64[1] = (double)unsigned32ToSigned2sCompl(load32((uint8_t*)addr + 4));
return (rx_vec_f128)x.d;
}
#define RANDOMX_DEFAULT_FENV
void rx_reset_float_state();
void rx_set_rounding_mode(uint32_t mode);
#else //end altivec
#include <cstdint>
#include <stdexcept>
#include <cstdlib>
#include <cmath>
typedef union {
uint64_t u64[2];
uint32_t u32[4];
uint16_t u16[8];
uint8_t u8[16];
} rx_vec_i128;
typedef union {
struct {
double lo;
double hi;
};
rx_vec_i128 i;
} rx_vec_f128;
#define rx_aligned_alloc(a, b) malloc(a)
#define rx_aligned_free(a) free(a)
#define rx_prefetch_nta(x)
FORCE_INLINE rx_vec_f128 rx_load_vec_f128(const double* pd) {
rx_vec_f128 x;
x.i.u64[0] = load64(pd + 0);
x.i.u64[1] = load64(pd + 1);
return x;
}
FORCE_INLINE void rx_store_vec_f128(double* mem_addr, rx_vec_f128 a) {
store64(mem_addr + 0, a.i.u64[0]);
store64(mem_addr + 1, a.i.u64[1]);
}
FORCE_INLINE rx_vec_f128 rx_swap_vec_f128(rx_vec_f128 a) {
double temp = a.hi;
a.hi = a.lo;
a.lo = temp;
return a;
}
FORCE_INLINE rx_vec_f128 rx_add_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
rx_vec_f128 x;
x.lo = a.lo + b.lo;
x.hi = a.hi + b.hi;
return x;
}
FORCE_INLINE rx_vec_f128 rx_sub_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
rx_vec_f128 x;
x.lo = a.lo - b.lo;
x.hi = a.hi - b.hi;
return x;
}
FORCE_INLINE rx_vec_f128 rx_mul_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
rx_vec_f128 x;
x.lo = a.lo * b.lo;
x.hi = a.hi * b.hi;
return x;
}
FORCE_INLINE rx_vec_f128 rx_div_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
rx_vec_f128 x;
x.lo = a.lo / b.lo;
x.hi = a.hi / b.hi;
return x;
}
FORCE_INLINE rx_vec_f128 rx_sqrt_vec_f128(rx_vec_f128 a) {
rx_vec_f128 x;
x.lo = rx_sqrt(a.lo);
x.hi = rx_sqrt(a.hi);
return x;
}
FORCE_INLINE rx_vec_i128 rx_set1_long_vec_i128(uint64_t a) {
rx_vec_i128 x;
x.u64[0] = a;
x.u64[1] = a;
return x;
}
FORCE_INLINE rx_vec_f128 rx_vec_i128_vec_f128(rx_vec_i128 a) {
rx_vec_f128 x;
x.i = a;
return x;
}
FORCE_INLINE rx_vec_f128 rx_set_vec_f128(uint64_t x1, uint64_t x0) {
rx_vec_f128 v;
v.i.u64[0] = x0;
v.i.u64[1] = x1;
return v;
}
FORCE_INLINE rx_vec_f128 rx_set1_vec_f128(uint64_t x) {
rx_vec_f128 v;
v.i.u64[0] = x;
v.i.u64[1] = x;
return v;
}
FORCE_INLINE rx_vec_f128 rx_xor_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
rx_vec_f128 x;
x.i.u64[0] = a.i.u64[0] ^ b.i.u64[0];
x.i.u64[1] = a.i.u64[1] ^ b.i.u64[1];
return x;
}
FORCE_INLINE rx_vec_f128 rx_and_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
rx_vec_f128 x;
x.i.u64[0] = a.i.u64[0] & b.i.u64[0];
x.i.u64[1] = a.i.u64[1] & b.i.u64[1];
return x;
}
FORCE_INLINE rx_vec_f128 rx_or_vec_f128(rx_vec_f128 a, rx_vec_f128 b) {
rx_vec_f128 x;
x.i.u64[0] = a.i.u64[0] | b.i.u64[0];
x.i.u64[1] = a.i.u64[1] | b.i.u64[1];
return x;
}
FORCE_INLINE int rx_vec_i128_x(rx_vec_i128 a) {
return a.u32[0];
}
FORCE_INLINE int rx_vec_i128_y(rx_vec_i128 a) {
return a.u32[1];
}
FORCE_INLINE int rx_vec_i128_z(rx_vec_i128 a) {
return a.u32[2];
}
FORCE_INLINE int rx_vec_i128_w(rx_vec_i128 a) {
return a.u32[3];
}
FORCE_INLINE rx_vec_i128 rx_set_int_vec_i128(int _I3, int _I2, int _I1, int _I0) {
rx_vec_i128 v;
v.u32[0] = _I0;
v.u32[1] = _I1;
v.u32[2] = _I2;
v.u32[3] = _I3;
return v;
};
FORCE_INLINE rx_vec_i128 rx_xor_vec_i128(rx_vec_i128 _A, rx_vec_i128 _B) {
rx_vec_i128 c;
c.u32[0] = _A.u32[0] ^ _B.u32[0];
c.u32[1] = _A.u32[1] ^ _B.u32[1];
c.u32[2] = _A.u32[2] ^ _B.u32[2];
c.u32[3] = _A.u32[3] ^ _B.u32[3];
return c;
}
FORCE_INLINE rx_vec_i128 rx_load_vec_i128(rx_vec_i128 const*_P) {
#if defined(NATIVE_LITTLE_ENDIAN)
return *_P;
#else
uint32_t* ptr = (uint32_t*)_P;
rx_vec_i128 c;
c.u32[0] = load32(ptr + 0);
c.u32[1] = load32(ptr + 1);
c.u32[2] = load32(ptr + 2);
c.u32[3] = load32(ptr + 3);
return c;
#endif
}
FORCE_INLINE void rx_store_vec_i128(rx_vec_i128 *_P, rx_vec_i128 _B) {
#if defined(NATIVE_LITTLE_ENDIAN)
*_P = _B;
#else
uint32_t* ptr = (uint32_t*)_P;
store32(ptr + 0, _B.u32[0]);
store32(ptr + 1, _B.u32[1]);
store32(ptr + 2, _B.u32[2]);
store32(ptr + 3, _B.u32[3]);
#endif
}
FORCE_INLINE rx_vec_f128 rx_cvt_packed_int_vec_f128(const void* addr) {
rx_vec_f128 x;
x.lo = (double)unsigned32ToSigned2sCompl(load32((uint8_t*)addr + 0));
x.hi = (double)unsigned32ToSigned2sCompl(load32((uint8_t*)addr + 4));
return x;
}
#define RANDOMX_DEFAULT_FENV
void rx_reset_float_state();
void rx_set_rounding_mode(uint32_t mode);
#endif
#ifndef HAVE_AES
static const char* platformError = "Platform doesn't support hardware AES";
#include <stdexcept>
FORCE_INLINE rx_vec_i128 rx_aesenc_vec_i128(rx_vec_i128 v, rx_vec_i128 rkey) {
throw std::runtime_error(platformError);
}
FORCE_INLINE rx_vec_i128 rx_aesdec_vec_i128(rx_vec_i128 v, rx_vec_i128 rkey) {
throw std::runtime_error(platformError);
}
#endif
double loadDoublePortable(const void* addr);
uint64_t mulh(uint64_t, uint64_t);
int64_t smulh(int64_t, int64_t);
uint64_t rotl64(uint64_t, unsigned int);
uint64_t rotr64(uint64_t, unsigned int);

View file

@ -0,0 +1,37 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#if defined(_M_X64) || defined(__x86_64__)
#include "jit_compiler_x86.hpp"
#elif defined(__aarch64__)
#include "jit_compiler_a64.hpp"
#else
#include "jit_compiler_fallback.hpp"
#endif

View file

@ -0,0 +1,73 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include <vector>
#include <stdexcept>
#include "common.hpp"
namespace randomx {
class Program;
class ProgramConfiguration;
class SuperscalarProgram;
class JitCompilerA64 {
public:
JitCompilerA64() {
throw std::runtime_error("ARM64 JIT compiler is not implemented yet.");
}
void generateProgram(Program&, ProgramConfiguration&) {
}
void generateProgramLight(Program&, ProgramConfiguration&, uint32_t) {
}
template<size_t N>
void generateSuperscalarHash(SuperscalarProgram(&programs)[N], std::vector<uint64_t> &) {
}
void generateDatasetInitCode() {
}
ProgramFunc* getProgramFunc() {
return nullptr;
}
DatasetInitFunc* getDatasetInitFunc() {
return nullptr;
}
uint8_t* getCode() {
return nullptr;
}
size_t getCodeSize() {
return 0;
}
};
}

View file

@ -0,0 +1,73 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include <vector>
#include <stdexcept>
#include "common.hpp"
namespace randomx {
class Program;
class ProgramConfiguration;
class SuperscalarProgram;
class JitCompilerFallback {
public:
JitCompilerFallback() {
throw std::runtime_error("JIT compilation is not supported on this platform");
}
void generateProgram(Program&, ProgramConfiguration&) {
}
void generateProgramLight(Program&, ProgramConfiguration&, uint32_t) {
}
template<size_t N>
void generateSuperscalarHash(SuperscalarProgram(&programs)[N], std::vector<uint64_t> &) {
}
void generateDatasetInitCode() {
}
ProgramFunc* getProgramFunc() {
return nullptr;
}
DatasetInitFunc* getDatasetInitFunc() {
return nullptr;
}
uint8_t* getCode() {
return nullptr;
}
size_t getCodeSize() {
return 0;
}
};
}

View file

@ -0,0 +1,774 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <stdexcept>
#include <cstring>
#include <climits>
#include "jit_compiler_x86.hpp"
#include "jit_compiler_x86_static.hpp"
#include "superscalar.hpp"
#include "program.hpp"
#include "reciprocal.h"
#include "virtual_memory.hpp"
namespace randomx {
/*
REGISTER ALLOCATION:
; rax -> temporary
; rbx -> iteration counter "ic"
; rcx -> temporary
; rdx -> temporary
; rsi -> scratchpad pointer
; rdi -> dataset pointer
; rbp -> memory registers "ma" (high 32 bits), "mx" (low 32 bits)
; rsp -> stack pointer
; r8 -> "r0"
; r9 -> "r1"
; r10 -> "r2"
; r11 -> "r3"
; r12 -> "r4"
; r13 -> "r5"
; r14 -> "r6"
; r15 -> "r7"
; xmm0 -> "f0"
; xmm1 -> "f1"
; xmm2 -> "f2"
; xmm3 -> "f3"
; xmm4 -> "e0"
; xmm5 -> "e1"
; xmm6 -> "e2"
; xmm7 -> "e3"
; xmm8 -> "a0"
; xmm9 -> "a1"
; xmm10 -> "a2"
; xmm11 -> "a3"
; xmm12 -> temporary
; xmm13 -> E 'and' mask = 0x00ffffffffffffff00ffffffffffffff
; xmm14 -> E 'or' mask = 0x3*00000000******3*00000000******
; xmm15 -> scale mask = 0x81f000000000000081f0000000000000
*/
const uint8_t* codePrologue = (uint8_t*)&randomx_program_prologue;
const uint8_t* codeLoopBegin = (uint8_t*)&randomx_program_loop_begin;
const uint8_t* codeLoopLoad = (uint8_t*)&randomx_program_loop_load;
const uint8_t* codeProgamStart = (uint8_t*)&randomx_program_start;
const uint8_t* codeReadDataset = (uint8_t*)&randomx_program_read_dataset;
const uint8_t* codeReadDatasetLightSshInit = (uint8_t*)&randomx_program_read_dataset_sshash_init;
const uint8_t* codeReadDatasetLightSshFin = (uint8_t*)&randomx_program_read_dataset_sshash_fin;
const uint8_t* codeDatasetInit = (uint8_t*)&randomx_dataset_init;
const uint8_t* codeLoopStore = (uint8_t*)&randomx_program_loop_store;
const uint8_t* codeLoopEnd = (uint8_t*)&randomx_program_loop_end;
const uint8_t* codeEpilogue = (uint8_t*)&randomx_program_epilogue;
const uint8_t* codeProgramEnd = (uint8_t*)&randomx_program_end;
const uint8_t* codeShhLoad = (uint8_t*)&randomx_sshash_load;
const uint8_t* codeShhPrefetch = (uint8_t*)&randomx_sshash_prefetch;
const uint8_t* codeShhEnd = (uint8_t*)&randomx_sshash_end;
const uint8_t* codeShhInit = (uint8_t*)&randomx_sshash_init;
const int32_t prologueSize = codeLoopBegin - codePrologue;
const int32_t loopLoadSize = codeProgamStart - codeLoopLoad;
const int32_t readDatasetSize = codeReadDatasetLightSshInit - codeReadDataset;
const int32_t readDatasetLightInitSize = codeReadDatasetLightSshFin - codeReadDatasetLightSshInit;
const int32_t readDatasetLightFinSize = codeLoopStore - codeReadDatasetLightSshFin;
const int32_t loopStoreSize = codeLoopEnd - codeLoopStore;
const int32_t datasetInitSize = codeEpilogue - codeDatasetInit;
const int32_t epilogueSize = codeShhLoad - codeEpilogue;
const int32_t codeSshLoadSize = codeShhPrefetch - codeShhLoad;
const int32_t codeSshPrefetchSize = codeShhEnd - codeShhPrefetch;
const int32_t codeSshInitSize = codeProgramEnd - codeShhInit;
const int32_t epilogueOffset = CodeSize - epilogueSize;
constexpr int32_t superScalarHashOffset = 32768;
static const uint8_t REX_ADD_RR[] = { 0x4d, 0x03 };
static const uint8_t REX_ADD_RM[] = { 0x4c, 0x03 };
static const uint8_t REX_SUB_RR[] = { 0x4d, 0x2b };
static const uint8_t REX_SUB_RM[] = { 0x4c, 0x2b };
static const uint8_t REX_MOV_RR[] = { 0x41, 0x8b };
static const uint8_t REX_MOV_RR64[] = { 0x49, 0x8b };
static const uint8_t REX_MOV_R64R[] = { 0x4c, 0x8b };
static const uint8_t REX_IMUL_RR[] = { 0x4d, 0x0f, 0xaf };
static const uint8_t REX_IMUL_RRI[] = { 0x4d, 0x69 };
static const uint8_t REX_IMUL_RM[] = { 0x4c, 0x0f, 0xaf };
static const uint8_t REX_MUL_R[] = { 0x49, 0xf7 };
static const uint8_t REX_MUL_M[] = { 0x48, 0xf7 };
static const uint8_t REX_81[] = { 0x49, 0x81 };
static const uint8_t AND_EAX_I = 0x25;
static const uint8_t MOV_EAX_I = 0xb8;
static const uint8_t MOV_RAX_I[] = { 0x48, 0xb8 };
static const uint8_t MOV_RCX_I[] = { 0x48, 0xb9 };
static const uint8_t REX_LEA[] = { 0x4f, 0x8d };
static const uint8_t REX_MUL_MEM[] = { 0x48, 0xf7, 0x24, 0x0e };
static const uint8_t REX_IMUL_MEM[] = { 0x48, 0xf7, 0x2c, 0x0e };
static const uint8_t REX_SHR_RAX[] = { 0x48, 0xc1, 0xe8 };
static const uint8_t RAX_ADD_SBB_1[] = { 0x48, 0x83, 0xC0, 0x01, 0x48, 0x83, 0xD8, 0x00 };
static const uint8_t MUL_RCX[] = { 0x48, 0xf7, 0xe1 };
static const uint8_t REX_SHR_RDX[] = { 0x48, 0xc1, 0xea };
static const uint8_t REX_SH[] = { 0x49, 0xc1 };
static const uint8_t MOV_RCX_RAX_SAR_RCX_63[] = { 0x48, 0x89, 0xc1, 0x48, 0xc1, 0xf9, 0x3f };
static const uint8_t AND_ECX_I[] = { 0x81, 0xe1 };
static const uint8_t ADD_RAX_RCX[] = { 0x48, 0x01, 0xC8 };
static const uint8_t SAR_RAX_I8[] = { 0x48, 0xC1, 0xF8 };
static const uint8_t NEG_RAX[] = { 0x48, 0xF7, 0xD8 };
static const uint8_t ADD_R_RAX[] = { 0x4C, 0x03 };
static const uint8_t XOR_EAX_EAX[] = { 0x33, 0xC0 };
static const uint8_t ADD_RDX_R[] = { 0x4c, 0x01 };
static const uint8_t SUB_RDX_R[] = { 0x4c, 0x29 };
static const uint8_t SAR_RDX_I8[] = { 0x48, 0xC1, 0xFA };
static const uint8_t TEST_RDX_RDX[] = { 0x48, 0x85, 0xD2 };
static const uint8_t SETS_AL_ADD_RDX_RAX[] = { 0x0F, 0x98, 0xC0, 0x48, 0x03, 0xD0 };
static const uint8_t REX_NEG[] = { 0x49, 0xF7 };
static const uint8_t REX_XOR_RR[] = { 0x4D, 0x33 };
static const uint8_t REX_XOR_RI[] = { 0x49, 0x81 };
static const uint8_t REX_XOR_RM[] = { 0x4c, 0x33 };
static const uint8_t REX_ROT_CL[] = { 0x49, 0xd3 };
static const uint8_t REX_ROT_I8[] = { 0x49, 0xc1 };
static const uint8_t SHUFPD[] = { 0x66, 0x0f, 0xc6 };
static const uint8_t REX_ADDPD[] = { 0x66, 0x41, 0x0f, 0x58 };
static const uint8_t REX_CVTDQ2PD_XMM12[] = { 0xf3, 0x44, 0x0f, 0xe6, 0x24, 0x06 };
static const uint8_t REX_SUBPD[] = { 0x66, 0x41, 0x0f, 0x5c };
static const uint8_t REX_XORPS[] = { 0x41, 0x0f, 0x57 };
static const uint8_t REX_MULPD[] = { 0x66, 0x41, 0x0f, 0x59 };
static const uint8_t REX_MAXPD[] = { 0x66, 0x41, 0x0f, 0x5f };
static const uint8_t REX_DIVPD[] = { 0x66, 0x41, 0x0f, 0x5e };
static const uint8_t SQRTPD[] = { 0x66, 0x0f, 0x51 };
static const uint8_t AND_OR_MOV_LDMXCSR[] = { 0x25, 0x00, 0x60, 0x00, 0x00, 0x0D, 0xC0, 0x9F, 0x00, 0x00, 0x50, 0x0F, 0xAE, 0x14, 0x24, 0x58 };
static const uint8_t ROL_RAX[] = { 0x48, 0xc1, 0xc0 };
static const uint8_t XOR_ECX_ECX[] = { 0x33, 0xC9 };
static const uint8_t REX_CMP_R32I[] = { 0x41, 0x81 };
static const uint8_t REX_CMP_M32I[] = { 0x81, 0x3c, 0x06 };
static const uint8_t MOVAPD[] = { 0x66, 0x0f, 0x29 };
static const uint8_t REX_MOV_MR[] = { 0x4c, 0x89 };
static const uint8_t REX_XOR_EAX[] = { 0x41, 0x33 };
static const uint8_t SUB_EBX[] = { 0x83, 0xEB, 0x01 };
static const uint8_t JNZ[] = { 0x0f, 0x85 };
static const uint8_t JMP = 0xe9;
static const uint8_t REX_XOR_RAX_R64[] = { 0x49, 0x33 };
static const uint8_t REX_XCHG[] = { 0x4d, 0x87 };
static const uint8_t REX_ANDPS_XMM12[] = { 0x45, 0x0F, 0x54, 0xE5, 0x45, 0x0F, 0x56, 0xE6 };
static const uint8_t REX_PADD[] = { 0x66, 0x44, 0x0f };
static const uint8_t PADD_OPCODES[] = { 0xfc, 0xfd, 0xfe, 0xd4 };
static const uint8_t CALL = 0xe8;
static const uint8_t REX_ADD_I[] = { 0x49, 0x81 };
static const uint8_t REX_TEST[] = { 0x49, 0xF7 };
static const uint8_t JZ[] = { 0x0f, 0x84 };
static const uint8_t RET = 0xc3;
static const uint8_t LEA_32[] = { 0x67, 0x41, 0x8d };
static const uint8_t MOVNTI[] = { 0x4c, 0x0f, 0xc3 };
static const uint8_t ADD_EBX_I[] = { 0x81, 0xc3 };
static const uint8_t NOP1[] = { 0x90 };
static const uint8_t NOP2[] = { 0x66, 0x90 };
static const uint8_t NOP3[] = { 0x66, 0x66, 0x90 };
static const uint8_t NOP4[] = { 0x0F, 0x1F, 0x40, 0x00 };
static const uint8_t NOP5[] = { 0x0F, 0x1F, 0x44, 0x00, 0x00 };
static const uint8_t NOP6[] = { 0x66, 0x0F, 0x1F, 0x44, 0x00, 0x00 };
static const uint8_t NOP7[] = { 0x0F, 0x1F, 0x80, 0x00, 0x00, 0x00, 0x00 };
static const uint8_t NOP8[] = { 0x0F, 0x1F, 0x84, 0x00, 0x00, 0x00, 0x00, 0x00 };
static const uint8_t* NOPX[] = { NOP1, NOP2, NOP3, NOP4, NOP5, NOP6, NOP7, NOP8 };
size_t JitCompilerX86::getCodeSize() {
return codePos - prologueSize;
}
JitCompilerX86::JitCompilerX86() {
code = (uint8_t*)allocExecutableMemory(CodeSize);
memcpy(code, codePrologue, prologueSize);
memcpy(code + epilogueOffset, codeEpilogue, epilogueSize);
}
JitCompilerX86::~JitCompilerX86() {
freePagedMemory(code, CodeSize);
}
void JitCompilerX86::generateProgram(Program& prog, ProgramConfiguration& pcfg) {
generateProgramPrologue(prog, pcfg);
memcpy(code + codePos, RandomX_CurrentConfig.codeReadDatasetTweaked, readDatasetSize);
codePos += readDatasetSize;
generateProgramEpilogue(prog);
}
void JitCompilerX86::generateProgramLight(Program& prog, ProgramConfiguration& pcfg, uint32_t datasetOffset) {
generateProgramPrologue(prog, pcfg);
emit(RandomX_CurrentConfig.codeReadDatasetLightSshInitTweaked, readDatasetLightInitSize);
emit(ADD_EBX_I);
emit32(datasetOffset / CacheLineSize);
emitByte(CALL);
emit32(superScalarHashOffset - (codePos + 4));
emit(codeReadDatasetLightSshFin, readDatasetLightFinSize);
generateProgramEpilogue(prog);
}
template<size_t N>
void JitCompilerX86::generateSuperscalarHash(SuperscalarProgram(&programs)[N], std::vector<uint64_t> &reciprocalCache) {
memcpy(code + superScalarHashOffset, codeShhInit, codeSshInitSize);
codePos = superScalarHashOffset + codeSshInitSize;
for (unsigned j = 0; j < RandomX_CurrentConfig.CacheAccesses; ++j) {
SuperscalarProgram& prog = programs[j];
for (unsigned i = 0; i < prog.getSize(); ++i) {
Instruction& instr = prog(i);
generateSuperscalarCode(instr, reciprocalCache);
}
emit(codeShhLoad, codeSshLoadSize);
if (j < RandomX_CurrentConfig.CacheAccesses - 1) {
emit(REX_MOV_RR64);
emitByte(0xd8 + prog.getAddressRegister());
emit(RandomX_CurrentConfig.codeShhPrefetchTweaked, codeSshPrefetchSize);
#ifdef RANDOMX_ALIGN
int align = (codePos % 16);
while (align != 0) {
int nopSize = 16 - align;
if (nopSize > 8) nopSize = 8;
emit(NOPX[nopSize - 1], nopSize);
align = (codePos % 16);
}
#endif
}
}
emitByte(RET);
}
template
void JitCompilerX86::generateSuperscalarHash(SuperscalarProgram(&programs)[RANDOMX_CACHE_MAX_ACCESSES], std::vector<uint64_t> &reciprocalCache);
void JitCompilerX86::generateDatasetInitCode() {
memcpy(code, codeDatasetInit, datasetInitSize);
}
void JitCompilerX86::generateProgramPrologue(Program& prog, ProgramConfiguration& pcfg) {
instructionOffsets.clear();
for (unsigned i = 0; i < 8; ++i) {
registerUsage[i] = -1;
}
codePos = prologueSize;
memcpy(code + codePos - 48, &pcfg.eMask, sizeof(pcfg.eMask));
emit(REX_XOR_RAX_R64);
emitByte(0xc0 + pcfg.readReg0);
emit(REX_XOR_RAX_R64);
emitByte(0xc0 + pcfg.readReg1);
memcpy(code + codePos, RandomX_CurrentConfig.codeLoopLoadTweaked, loopLoadSize);
codePos += loopLoadSize;
for (unsigned i = 0; i < prog.getSize(); ++i) {
Instruction& instr = prog(i);
instr.src %= RegistersCount;
instr.dst %= RegistersCount;
generateCode(instr, i);
}
emit(REX_MOV_RR);
emitByte(0xc0 + pcfg.readReg2);
emit(REX_XOR_EAX);
emitByte(0xc0 + pcfg.readReg3);
}
void JitCompilerX86::generateProgramEpilogue(Program& prog) {
memcpy(code + codePos, codeLoopStore, loopStoreSize);
codePos += loopStoreSize;
emit(SUB_EBX);
emit(JNZ);
emit32(prologueSize - codePos - 4);
emitByte(JMP);
emit32(epilogueOffset - codePos - 4);
}
void JitCompilerX86::generateCode(Instruction& instr, int i) {
instructionOffsets.push_back(codePos);
auto generator = engine[instr.opcode];
(this->*generator)(instr, i);
}
void JitCompilerX86::generateSuperscalarCode(Instruction& instr, std::vector<uint64_t> &reciprocalCache) {
switch ((SuperscalarInstructionType)instr.opcode)
{
case randomx::SuperscalarInstructionType::ISUB_R:
emit(REX_SUB_RR);
emitByte(0xc0 + 8 * instr.dst + instr.src);
break;
case randomx::SuperscalarInstructionType::IXOR_R:
emit(REX_XOR_RR);
emitByte(0xc0 + 8 * instr.dst + instr.src);
break;
case randomx::SuperscalarInstructionType::IADD_RS:
emit(REX_LEA);
emitByte(0x04 + 8 * instr.dst);
genSIB(instr.getModShift(), instr.src, instr.dst);
break;
case randomx::SuperscalarInstructionType::IMUL_R:
emit(REX_IMUL_RR);
emitByte(0xc0 + 8 * instr.dst + instr.src);
break;
case randomx::SuperscalarInstructionType::IROR_C:
emit(REX_ROT_I8);
emitByte(0xc8 + instr.dst);
emitByte(instr.getImm32() & 63);
break;
case randomx::SuperscalarInstructionType::IADD_C7:
emit(REX_81);
emitByte(0xc0 + instr.dst);
emit32(instr.getImm32());
break;
case randomx::SuperscalarInstructionType::IXOR_C7:
emit(REX_XOR_RI);
emitByte(0xf0 + instr.dst);
emit32(instr.getImm32());
break;
case randomx::SuperscalarInstructionType::IADD_C8:
emit(REX_81);
emitByte(0xc0 + instr.dst);
emit32(instr.getImm32());
#ifdef RANDOMX_ALIGN
emit(NOP1);
#endif
break;
case randomx::SuperscalarInstructionType::IXOR_C8:
emit(REX_XOR_RI);
emitByte(0xf0 + instr.dst);
emit32(instr.getImm32());
#ifdef RANDOMX_ALIGN
emit(NOP1);
#endif
break;
case randomx::SuperscalarInstructionType::IADD_C9:
emit(REX_81);
emitByte(0xc0 + instr.dst);
emit32(instr.getImm32());
#ifdef RANDOMX_ALIGN
emit(NOP2);
#endif
break;
case randomx::SuperscalarInstructionType::IXOR_C9:
emit(REX_XOR_RI);
emitByte(0xf0 + instr.dst);
emit32(instr.getImm32());
#ifdef RANDOMX_ALIGN
emit(NOP2);
#endif
break;
case randomx::SuperscalarInstructionType::IMULH_R:
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.dst);
emit(REX_MUL_R);
emitByte(0xe0 + instr.src);
emit(REX_MOV_R64R);
emitByte(0xc2 + 8 * instr.dst);
break;
case randomx::SuperscalarInstructionType::ISMULH_R:
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.dst);
emit(REX_MUL_R);
emitByte(0xe8 + instr.src);
emit(REX_MOV_R64R);
emitByte(0xc2 + 8 * instr.dst);
break;
case randomx::SuperscalarInstructionType::IMUL_RCP:
emit(MOV_RAX_I);
emit64(reciprocalCache[instr.getImm32()]);
emit(REX_IMUL_RM);
emitByte(0xc0 + 8 * instr.dst);
break;
default:
UNREACHABLE;
}
}
void JitCompilerX86::genAddressReg(Instruction& instr, bool rax = true) {
emit(LEA_32);
emitByte(0x80 + instr.src + (rax ? 0 : 8));
if (instr.src == RegisterNeedsSib) {
emitByte(0x24);
}
emit32(instr.getImm32());
if (rax)
emitByte(AND_EAX_I);
else
emit(AND_ECX_I);
emit32(instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
}
void JitCompilerX86::genAddressRegDst(Instruction& instr) {
emit(LEA_32);
emitByte(0x80 + instr.dst);
if (instr.dst == RegisterNeedsSib) {
emitByte(0x24);
}
emit32(instr.getImm32());
emitByte(AND_EAX_I);
if (instr.getModCond() < StoreL3Condition) {
emit32(instr.getModMem() ? ScratchpadL1Mask : ScratchpadL2Mask);
}
else {
emit32(ScratchpadL3Mask);
}
}
void JitCompilerX86::genAddressImm(Instruction& instr) {
emit32(instr.getImm32() & ScratchpadL3Mask);
}
void JitCompilerX86::h_IADD_RS(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
emit(REX_LEA);
if (instr.dst == RegisterNeedsDisplacement)
emitByte(0xac);
else
emitByte(0x04 + 8 * instr.dst);
genSIB(instr.getModShift(), instr.src, instr.dst);
if (instr.dst == RegisterNeedsDisplacement)
emit32(instr.getImm32());
}
void JitCompilerX86::h_IADD_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr);
emit(REX_ADD_RM);
emitByte(0x04 + 8 * instr.dst);
emitByte(0x06);
}
else {
emit(REX_ADD_RM);
emitByte(0x86 + 8 * instr.dst);
genAddressImm(instr);
}
}
void JitCompilerX86::genSIB(int scale, int index, int base) {
emitByte((scale << 6) | (index << 3) | base);
}
void JitCompilerX86::h_ISUB_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
emit(REX_SUB_RR);
emitByte(0xc0 + 8 * instr.dst + instr.src);
}
else {
emit(REX_81);
emitByte(0xe8 + instr.dst);
emit32(instr.getImm32());
}
}
void JitCompilerX86::h_ISUB_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr);
emit(REX_SUB_RM);
emitByte(0x04 + 8 * instr.dst);
emitByte(0x06);
}
else {
emit(REX_SUB_RM);
emitByte(0x86 + 8 * instr.dst);
genAddressImm(instr);
}
}
void JitCompilerX86::h_IMUL_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
emit(REX_IMUL_RR);
emitByte(0xc0 + 8 * instr.dst + instr.src);
}
else {
emit(REX_IMUL_RRI);
emitByte(0xc0 + 9 * instr.dst);
emit32(instr.getImm32());
}
}
void JitCompilerX86::h_IMUL_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr);
emit(REX_IMUL_RM);
emitByte(0x04 + 8 * instr.dst);
emitByte(0x06);
}
else {
emit(REX_IMUL_RM);
emitByte(0x86 + 8 * instr.dst);
genAddressImm(instr);
}
}
void JitCompilerX86::h_IMULH_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.dst);
emit(REX_MUL_R);
emitByte(0xe0 + instr.src);
emit(REX_MOV_R64R);
emitByte(0xc2 + 8 * instr.dst);
}
void JitCompilerX86::h_IMULH_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr, false);
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.dst);
emit(REX_MUL_MEM);
}
else {
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.dst);
emit(REX_MUL_M);
emitByte(0xa6);
genAddressImm(instr);
}
emit(REX_MOV_R64R);
emitByte(0xc2 + 8 * instr.dst);
}
void JitCompilerX86::h_ISMULH_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.dst);
emit(REX_MUL_R);
emitByte(0xe8 + instr.src);
emit(REX_MOV_R64R);
emitByte(0xc2 + 8 * instr.dst);
}
void JitCompilerX86::h_ISMULH_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr, false);
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.dst);
emit(REX_IMUL_MEM);
}
else {
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.dst);
emit(REX_MUL_M);
emitByte(0xae);
genAddressImm(instr);
}
emit(REX_MOV_R64R);
emitByte(0xc2 + 8 * instr.dst);
}
void JitCompilerX86::h_IMUL_RCP(Instruction& instr, int i) {
uint64_t divisor = instr.getImm32();
if (!isPowerOf2(divisor)) {
registerUsage[instr.dst] = i;
emit(MOV_RAX_I);
emit64(randomx_reciprocal_fast(divisor));
emit(REX_IMUL_RM);
emitByte(0xc0 + 8 * instr.dst);
}
}
void JitCompilerX86::h_INEG_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
emit(REX_NEG);
emitByte(0xd8 + instr.dst);
}
void JitCompilerX86::h_IXOR_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
emit(REX_XOR_RR);
emitByte(0xc0 + 8 * instr.dst + instr.src);
}
else {
emit(REX_XOR_RI);
emitByte(0xf0 + instr.dst);
emit32(instr.getImm32());
}
}
void JitCompilerX86::h_IXOR_M(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
genAddressReg(instr);
emit(REX_XOR_RM);
emitByte(0x04 + 8 * instr.dst);
emitByte(0x06);
}
else {
emit(REX_XOR_RM);
emitByte(0x86 + 8 * instr.dst);
genAddressImm(instr);
}
}
void JitCompilerX86::h_IROR_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
emit(REX_MOV_RR);
emitByte(0xc8 + instr.src);
emit(REX_ROT_CL);
emitByte(0xc8 + instr.dst);
}
else {
emit(REX_ROT_I8);
emitByte(0xc8 + instr.dst);
emitByte(instr.getImm32() & 63);
}
}
void JitCompilerX86::h_IROL_R(Instruction& instr, int i) {
registerUsage[instr.dst] = i;
if (instr.src != instr.dst) {
emit(REX_MOV_RR);
emitByte(0xc8 + instr.src);
emit(REX_ROT_CL);
emitByte(0xc0 + instr.dst);
}
else {
emit(REX_ROT_I8);
emitByte(0xc0 + instr.dst);
emitByte(instr.getImm32() & 63);
}
}
void JitCompilerX86::h_ISWAP_R(Instruction& instr, int i) {
if (instr.src != instr.dst) {
registerUsage[instr.dst] = i;
registerUsage[instr.src] = i;
emit(REX_XCHG);
emitByte(0xc0 + instr.src + 8 * instr.dst);
}
}
void JitCompilerX86::h_FSWAP_R(Instruction& instr, int i) {
emit(SHUFPD);
emitByte(0xc0 + 9 * instr.dst);
emitByte(1);
}
void JitCompilerX86::h_FADD_R(Instruction& instr, int i) {
instr.dst %= RegisterCountFlt;
instr.src %= RegisterCountFlt;
emit(REX_ADDPD);
emitByte(0xc0 + instr.src + 8 * instr.dst);
}
void JitCompilerX86::h_FADD_M(Instruction& instr, int i) {
instr.dst %= RegisterCountFlt;
genAddressReg(instr);
emit(REX_CVTDQ2PD_XMM12);
emit(REX_ADDPD);
emitByte(0xc4 + 8 * instr.dst);
}
void JitCompilerX86::h_FSUB_R(Instruction& instr, int i) {
instr.dst %= RegisterCountFlt;
instr.src %= RegisterCountFlt;
emit(REX_SUBPD);
emitByte(0xc0 + instr.src + 8 * instr.dst);
}
void JitCompilerX86::h_FSUB_M(Instruction& instr, int i) {
instr.dst %= RegisterCountFlt;
genAddressReg(instr);
emit(REX_CVTDQ2PD_XMM12);
emit(REX_SUBPD);
emitByte(0xc4 + 8 * instr.dst);
}
void JitCompilerX86::h_FSCAL_R(Instruction& instr, int i) {
instr.dst %= RegisterCountFlt;
emit(REX_XORPS);
emitByte(0xc7 + 8 * instr.dst);
}
void JitCompilerX86::h_FMUL_R(Instruction& instr, int i) {
instr.dst %= RegisterCountFlt;
instr.src %= RegisterCountFlt;
emit(REX_MULPD);
emitByte(0xe0 + instr.src + 8 * instr.dst);
}
void JitCompilerX86::h_FDIV_M(Instruction& instr, int i) {
instr.dst %= RegisterCountFlt;
genAddressReg(instr);
emit(REX_CVTDQ2PD_XMM12);
emit(REX_ANDPS_XMM12);
emit(REX_DIVPD);
emitByte(0xe4 + 8 * instr.dst);
}
void JitCompilerX86::h_FSQRT_R(Instruction& instr, int i) {
instr.dst %= RegisterCountFlt;
emit(SQRTPD);
emitByte(0xe4 + 9 * instr.dst);
}
void JitCompilerX86::h_CFROUND(Instruction& instr, int i) {
emit(REX_MOV_RR64);
emitByte(0xc0 + instr.src);
int rotate = (13 - (instr.getImm32() & 63)) & 63;
if (rotate != 0) {
emit(ROL_RAX);
emitByte(rotate);
}
emit(AND_OR_MOV_LDMXCSR);
}
void JitCompilerX86::h_CBRANCH(Instruction& instr, int i) {
int reg = instr.dst;
int target = registerUsage[reg] + 1;
emit(REX_ADD_I);
emitByte(0xc0 + reg);
int shift = instr.getModCond() + RandomX_CurrentConfig.JumpOffset;
uint32_t imm = instr.getImm32() | (1UL << shift);
if (RandomX_CurrentConfig.JumpOffset > 0 || shift > 0)
imm &= ~(1UL << (shift - 1));
emit32(imm);
emit(REX_TEST);
emitByte(0xc0 + reg);
emit32(RandomX_CurrentConfig.ConditionMask_Calculated << shift);
emit(JZ);
emit32(instructionOffsets[target] - (codePos + 4));
//mark all registers as used
for (unsigned j = 0; j < RegistersCount; ++j) {
registerUsage[j] = i;
}
}
void JitCompilerX86::h_ISTORE(Instruction& instr, int i) {
genAddressRegDst(instr);
emit(REX_MOV_MR);
emitByte(0x04 + 8 * instr.src);
emitByte(0x06);
}
void JitCompilerX86::h_NOP(Instruction& instr, int i) {
emit(NOP1);
}
InstructionGeneratorX86 JitCompilerX86::engine[256] = {};
}

View file

@ -0,0 +1,141 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include <cstring>
#include <vector>
#include "common.hpp"
namespace randomx {
class Program;
class ProgramConfiguration;
class SuperscalarProgram;
class JitCompilerX86;
class Instruction;
typedef void(JitCompilerX86::*InstructionGeneratorX86)(Instruction&, int);
constexpr uint32_t CodeSize = 64 * 1024;
class JitCompilerX86 {
public:
JitCompilerX86();
~JitCompilerX86();
void generateProgram(Program&, ProgramConfiguration&);
void generateProgramLight(Program&, ProgramConfiguration&, uint32_t);
template<size_t N>
void generateSuperscalarHash(SuperscalarProgram (&programs)[N], std::vector<uint64_t> &);
void generateDatasetInitCode();
ProgramFunc* getProgramFunc() {
return (ProgramFunc*)code;
}
DatasetInitFunc* getDatasetInitFunc() {
return (DatasetInitFunc*)code;
}
uint8_t* getCode() {
return code;
}
size_t getCodeSize();
static InstructionGeneratorX86 engine[256];
std::vector<int32_t> instructionOffsets;
int registerUsage[RegistersCount];
uint8_t* code;
int32_t codePos;
void generateProgramPrologue(Program&, ProgramConfiguration&);
void generateProgramEpilogue(Program&);
void genAddressReg(Instruction&, bool);
void genAddressRegDst(Instruction&);
void genAddressImm(Instruction&);
void genSIB(int scale, int index, int base);
void generateCode(Instruction&, int);
void generateSuperscalarCode(Instruction &, std::vector<uint64_t> &);
void emitByte(uint8_t val) {
code[codePos] = val;
codePos++;
}
void emit32(uint32_t val) {
memcpy(code + codePos, &val, sizeof val);
codePos += sizeof val;
}
void emit64(uint64_t val) {
memcpy(code + codePos, &val, sizeof val);
codePos += sizeof val;
}
template<size_t N>
void emit(const uint8_t (&src)[N]) {
emit(src, N);
}
void emit(const uint8_t* src, size_t count) {
memcpy(code + codePos, src, count);
codePos += count;
}
void h_IADD_RS(Instruction&, int);
void h_IADD_M(Instruction&, int);
void h_ISUB_R(Instruction&, int);
void h_ISUB_M(Instruction&, int);
void h_IMUL_R(Instruction&, int);
void h_IMUL_M(Instruction&, int);
void h_IMULH_R(Instruction&, int);
void h_IMULH_M(Instruction&, int);
void h_ISMULH_R(Instruction&, int);
void h_ISMULH_M(Instruction&, int);
void h_IMUL_RCP(Instruction&, int);
void h_INEG_R(Instruction&, int);
void h_IXOR_R(Instruction&, int);
void h_IXOR_M(Instruction&, int);
void h_IROR_R(Instruction&, int);
void h_IROL_R(Instruction&, int);
void h_ISWAP_R(Instruction&, int);
void h_FSWAP_R(Instruction&, int);
void h_FADD_R(Instruction&, int);
void h_FADD_M(Instruction&, int);
void h_FSUB_R(Instruction&, int);
void h_FSUB_M(Instruction&, int);
void h_FSCAL_R(Instruction&, int);
void h_FMUL_R(Instruction&, int);
void h_FDIV_M(Instruction&, int);
void h_FSQRT_R(Instruction&, int);
void h_CBRANCH(Instruction&, int);
void h_CFROUND(Instruction&, int);
void h_ISTORE(Instruction&, int);
void h_NOP(Instruction&, int);
};
}

View file

@ -0,0 +1,208 @@
# Copyright (c) 2018-2019, tevador <tevador@gmail.com>
#
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
# * Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
# * Neither the name of the copyright holder nor the
# names of its contributors may be used to endorse or promote products
# derived from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.intel_syntax noprefix
#if defined(__APPLE__)
.text
#define DECL(x) _##x
#else
.section .text
#define DECL(x) x
#endif
#if defined(__WIN32__) || defined(__CYGWIN__)
#define WINABI
#endif
.global DECL(randomx_program_prologue)
.global DECL(randomx_program_loop_begin)
.global DECL(randomx_program_loop_load)
.global DECL(randomx_program_start)
.global DECL(randomx_program_read_dataset)
.global DECL(randomx_program_read_dataset_sshash_init)
.global DECL(randomx_program_read_dataset_sshash_fin)
.global DECL(randomx_program_loop_store)
.global DECL(randomx_program_loop_end)
.global DECL(randomx_dataset_init)
.global DECL(randomx_program_epilogue)
.global DECL(randomx_sshash_load)
.global DECL(randomx_sshash_prefetch)
.global DECL(randomx_sshash_end)
.global DECL(randomx_sshash_init)
.global DECL(randomx_program_end)
.global DECL(randomx_reciprocal_fast)
#define RANDOMX_SCRATCHPAD_MASK 2097088
#define RANDOMX_DATASET_BASE_MASK 2147483584
#define RANDOMX_CACHE_MASK 4194303
#define db .byte
.balign 64
DECL(randomx_program_prologue):
#if defined(WINABI)
#include "asm/program_prologue_win64.inc"
#else
#include "asm/program_prologue_linux.inc"
#endif
movapd xmm13, xmmword ptr mantissaMask[rip]
movapd xmm14, xmmword ptr exp240[rip]
movapd xmm15, xmmword ptr scaleMask[rip]
jmp DECL(randomx_program_loop_begin)
.balign 64
#include "asm/program_xmm_constants.inc"
.balign 64
DECL(randomx_program_loop_begin):
nop
DECL(randomx_program_loop_load):
#include "asm/program_loop_load.inc"
DECL(randomx_program_start):
nop
DECL(randomx_program_read_dataset):
#include "asm/program_read_dataset.inc"
DECL(randomx_program_read_dataset_sshash_init):
#include "asm/program_read_dataset_sshash_init.inc"
DECL(randomx_program_read_dataset_sshash_fin):
#include "asm/program_read_dataset_sshash_fin.inc"
DECL(randomx_program_loop_store):
#include "asm/program_loop_store.inc"
DECL(randomx_program_loop_end):
nop
.balign 64
DECL(randomx_dataset_init):
push rbx
push rbp
push r12
push r13
push r14
push r15
#if defined(WINABI)
push rdi
push rsi
mov rdi, qword ptr [rcx] ;# cache->memory
mov rsi, rdx ;# dataset
mov rbp, r8 ;# block index
push r9 ;# max. block index
#else
mov rdi, qword ptr [rdi] ;# cache->memory
;# dataset in rsi
mov rbp, rdx ;# block index
push rcx ;# max. block index
#endif
init_block_loop:
prefetchw byte ptr [rsi]
mov rbx, rbp
.byte 232 ;# 0xE8 = call
;# .set CALL_LOC,
.int 32768 - (call_offset - DECL(randomx_dataset_init))
call_offset:
mov qword ptr [rsi+0], r8
mov qword ptr [rsi+8], r9
mov qword ptr [rsi+16], r10
mov qword ptr [rsi+24], r11
mov qword ptr [rsi+32], r12
mov qword ptr [rsi+40], r13
mov qword ptr [rsi+48], r14
mov qword ptr [rsi+56], r15
add rbp, 1
add rsi, 64
cmp rbp, qword ptr [rsp]
jb init_block_loop
pop rax
#if defined(WINABI)
pop rsi
pop rdi
#endif
pop r15
pop r14
pop r13
pop r12
pop rbp
pop rbx
ret
.balign 64
DECL(randomx_program_epilogue):
#include "asm/program_epilogue_store.inc"
#if defined(WINABI)
#include "asm/program_epilogue_win64.inc"
#else
#include "asm/program_epilogue_linux.inc"
#endif
.balign 64
DECL(randomx_sshash_load):
#include "asm/program_sshash_load.inc"
DECL(randomx_sshash_prefetch):
#include "asm/program_sshash_prefetch.inc"
DECL(randomx_sshash_end):
nop
.balign 64
DECL(randomx_sshash_init):
lea r8, [rbx+1]
#include "asm/program_sshash_prefetch.inc"
imul r8, qword ptr r0_mul[rip]
mov r9, qword ptr r1_add[rip]
xor r9, r8
mov r10, qword ptr r2_add[rip]
xor r10, r8
mov r11, qword ptr r3_add[rip]
xor r11, r8
mov r12, qword ptr r4_add[rip]
xor r12, r8
mov r13, qword ptr r5_add[rip]
xor r13, r8
mov r14, qword ptr r6_add[rip]
xor r14, r8
mov r15, qword ptr r7_add[rip]
xor r15, r8
jmp DECL(randomx_program_end)
.balign 64
#include "asm/program_sshash_constants.inc"
.balign 64
DECL(randomx_program_end):
nop
DECL(randomx_reciprocal_fast):
#if !defined(WINABI)
mov rcx, rdi
#endif
#include "asm/randomx_reciprocal.inc"

View file

@ -0,0 +1,199 @@
; Copyright (c) 2018-2019, tevador <tevador@gmail.com>
;
; All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in the
; documentation and/or other materials provided with the distribution.
; * Neither the name of the copyright holder nor the
; names of its contributors may be used to endorse or promote products
; derived from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
; ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
; WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
; DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
; FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
; DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
; SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
; CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
; OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
IFDEF RAX
_RANDOMX_JITX86_STATIC SEGMENT PAGE READ EXECUTE
PUBLIC randomx_program_prologue
PUBLIC randomx_program_loop_begin
PUBLIC randomx_program_loop_load
PUBLIC randomx_program_start
PUBLIC randomx_program_read_dataset
PUBLIC randomx_program_read_dataset_sshash_init
PUBLIC randomx_program_read_dataset_sshash_fin
PUBLIC randomx_dataset_init
PUBLIC randomx_program_loop_store
PUBLIC randomx_program_loop_end
PUBLIC randomx_program_epilogue
PUBLIC randomx_sshash_load
PUBLIC randomx_sshash_prefetch
PUBLIC randomx_sshash_end
PUBLIC randomx_sshash_init
PUBLIC randomx_program_end
PUBLIC randomx_reciprocal_fast
RANDOMX_SCRATCHPAD_MASK EQU 2097088
RANDOMX_DATASET_BASE_MASK EQU 2147483584
RANDOMX_CACHE_MASK EQU 4194303
ALIGN 64
randomx_program_prologue PROC
include asm/program_prologue_win64.inc
movapd xmm13, xmmword ptr [mantissaMask]
movapd xmm14, xmmword ptr [exp240]
movapd xmm15, xmmword ptr [scaleMask]
jmp randomx_program_loop_begin
randomx_program_prologue ENDP
ALIGN 64
include asm/program_xmm_constants.inc
ALIGN 64
randomx_program_loop_begin PROC
nop
randomx_program_loop_begin ENDP
randomx_program_loop_load PROC
include asm/program_loop_load.inc
randomx_program_loop_load ENDP
randomx_program_start PROC
nop
randomx_program_start ENDP
randomx_program_read_dataset PROC
include asm/program_read_dataset.inc
randomx_program_read_dataset ENDP
randomx_program_read_dataset_sshash_init PROC
include asm/program_read_dataset_sshash_init.inc
randomx_program_read_dataset_sshash_init ENDP
randomx_program_read_dataset_sshash_fin PROC
include asm/program_read_dataset_sshash_fin.inc
randomx_program_read_dataset_sshash_fin ENDP
randomx_program_loop_store PROC
include asm/program_loop_store.inc
randomx_program_loop_store ENDP
randomx_program_loop_end PROC
nop
randomx_program_loop_end ENDP
ALIGN 64
randomx_dataset_init PROC
push rbx
push rbp
push rdi
push rsi
push r12
push r13
push r14
push r15
mov rdi, qword ptr [rcx] ;# cache->memory
mov rsi, rdx ;# dataset
mov rbp, r8 ;# block index
push r9 ;# max. block index
init_block_loop:
prefetchw byte ptr [rsi]
mov rbx, rbp
db 232 ;# 0xE8 = call
dd 32768 - distance
distance equ $ - offset randomx_dataset_init
mov qword ptr [rsi+0], r8
mov qword ptr [rsi+8], r9
mov qword ptr [rsi+16], r10
mov qword ptr [rsi+24], r11
mov qword ptr [rsi+32], r12
mov qword ptr [rsi+40], r13
mov qword ptr [rsi+48], r14
mov qword ptr [rsi+56], r15
add rbp, 1
add rsi, 64
cmp rbp, qword ptr [rsp]
jb init_block_loop
pop r9
pop r15
pop r14
pop r13
pop r12
pop rsi
pop rdi
pop rbp
pop rbx
ret
randomx_dataset_init ENDP
ALIGN 64
randomx_program_epilogue PROC
include asm/program_epilogue_store.inc
include asm/program_epilogue_win64.inc
randomx_program_epilogue ENDP
ALIGN 64
randomx_sshash_load PROC
include asm/program_sshash_load.inc
randomx_sshash_load ENDP
randomx_sshash_prefetch PROC
include asm/program_sshash_prefetch.inc
randomx_sshash_prefetch ENDP
randomx_sshash_end PROC
nop
randomx_sshash_end ENDP
ALIGN 64
randomx_sshash_init PROC
lea r8, [rbx+1]
include asm/program_sshash_prefetch.inc
imul r8, qword ptr [r0_mul]
mov r9, qword ptr [r1_add]
xor r9, r8
mov r10, qword ptr [r2_add]
xor r10, r8
mov r11, qword ptr [r3_add]
xor r11, r8
mov r12, qword ptr [r4_add]
xor r12, r8
mov r13, qword ptr [r5_add]
xor r13, r8
mov r14, qword ptr [r6_add]
xor r14, r8
mov r15, qword ptr [r7_add]
xor r15, r8
jmp randomx_program_end
randomx_sshash_init ENDP
ALIGN 64
include asm/program_sshash_constants.inc
ALIGN 64
randomx_program_end PROC
nop
randomx_program_end ENDP
randomx_reciprocal_fast PROC
include asm/randomx_reciprocal.inc
randomx_reciprocal_fast ENDP
_RANDOMX_JITX86_STATIC ENDS
ENDIF
END

View file

@ -0,0 +1,48 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
extern "C" {
void randomx_program_prologue();
void randomx_program_loop_begin();
void randomx_program_loop_load();
void randomx_program_start();
void randomx_program_read_dataset();
void randomx_program_read_dataset_sshash_init();
void randomx_program_read_dataset_sshash_fin();
void randomx_program_loop_store();
void randomx_program_loop_end();
void randomx_dataset_init();
void randomx_program_epilogue();
void randomx_sshash_load();
void randomx_sshash_prefetch();
void randomx_sshash_end();
void randomx_sshash_init();
void randomx_program_end();
}

View file

@ -0,0 +1,60 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include "common.hpp"
#include "instruction.hpp"
#include "blake2/endian.h"
namespace randomx {
struct ProgramConfiguration {
uint64_t eMask[2];
uint32_t readReg0, readReg1, readReg2, readReg3;
};
class Program {
public:
Instruction& operator()(int pc) {
return programBuffer[pc];
}
uint64_t getEntropy(int i) {
return load64(&entropyBuffer[i]);
}
uint32_t getSize() {
return RandomX_CurrentConfig.ProgramSize;
}
private:
uint64_t entropyBuffer[16];
Instruction programBuffer[RANDOMX_PROGRAM_MAX_SIZE];
};
static_assert(sizeof(Program) % 64 == 0, "Invalid size of class randomx::Program");
}

View file

@ -0,0 +1,476 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "randomx.h"
#include "dataset.hpp"
#include "vm_interpreted.hpp"
#include "vm_interpreted_light.hpp"
#include "vm_compiled.hpp"
#include "vm_compiled_light.hpp"
#include "blake2/blake2.h"
#include "jit_compiler_x86_static.hpp"
#include <cassert>
RandomX_ConfigurationWownero::RandomX_ConfigurationWownero()
{
ArgonSalt = "RandomWOW\x01";
ProgramIterations = 1024;
ProgramCount = 16;
ScratchpadL2_Size = 131072;
ScratchpadL3_Size = 1048576;
RANDOMX_FREQ_IROR_R = 10;
RANDOMX_FREQ_IROL_R = 0;
RANDOMX_FREQ_FSWAP_R = 8;
RANDOMX_FREQ_FADD_R = 20;
RANDOMX_FREQ_FSUB_R = 20;
RANDOMX_FREQ_FMUL_R = 20;
fillAes4Rx4_Key[0] = rx_set_int_vec_i128(0xcf359e95, 0x141f82b7, 0x7ffbe4a6, 0xf890465d);
fillAes4Rx4_Key[1] = rx_set_int_vec_i128(0x6741ffdc, 0xbd5c5ac3, 0xfee8278a, 0x6a55c450);
fillAes4Rx4_Key[2] = rx_set_int_vec_i128(0x3d324aac, 0xa7279ad2, 0xd524fde4, 0x114c47a4);
fillAes4Rx4_Key[3] = rx_set_int_vec_i128(0x76f6db08, 0x42d3dbd9, 0x99a9aeff, 0x810c3a2a);
fillAes4Rx4_Key[4] = fillAes4Rx4_Key[0];
fillAes4Rx4_Key[5] = fillAes4Rx4_Key[1];
fillAes4Rx4_Key[6] = fillAes4Rx4_Key[2];
fillAes4Rx4_Key[7] = fillAes4Rx4_Key[3];
}
RandomX_ConfigurationLoki::RandomX_ConfigurationLoki()
{
ArgonIterations = 4;
ArgonLanes = 2;
ArgonSalt = "RandomXL\x12";
ProgramSize = 320;
ProgramCount = 7;
}
RandomX_ConfigurationBase::RandomX_ConfigurationBase()
: ArgonMemory(262144)
, ArgonIterations(3)
, ArgonLanes(1)
, ArgonSalt("RandomX\x03")
, CacheAccesses(8)
, SuperscalarLatency(170)
, DatasetBaseSize(2147483648)
, DatasetExtraSize(33554368)
, ScratchpadL1_Size(16384)
, ScratchpadL2_Size(262144)
, ScratchpadL3_Size(2097152)
, ProgramSize(256)
, ProgramIterations(2048)
, ProgramCount(8)
, JumpBits(8)
, JumpOffset(8)
, RANDOMX_FREQ_IADD_RS(25)
, RANDOMX_FREQ_IADD_M(7)
, RANDOMX_FREQ_ISUB_R(16)
, RANDOMX_FREQ_ISUB_M(7)
, RANDOMX_FREQ_IMUL_R(16)
, RANDOMX_FREQ_IMUL_M(4)
, RANDOMX_FREQ_IMULH_R(4)
, RANDOMX_FREQ_IMULH_M(1)
, RANDOMX_FREQ_ISMULH_R(4)
, RANDOMX_FREQ_ISMULH_M(1)
, RANDOMX_FREQ_IMUL_RCP(8)
, RANDOMX_FREQ_INEG_R(2)
, RANDOMX_FREQ_IXOR_R(15)
, RANDOMX_FREQ_IXOR_M(5)
, RANDOMX_FREQ_IROR_R(8)
, RANDOMX_FREQ_IROL_R(2)
, RANDOMX_FREQ_ISWAP_R(4)
, RANDOMX_FREQ_FSWAP_R(4)
, RANDOMX_FREQ_FADD_R(16)
, RANDOMX_FREQ_FADD_M(5)
, RANDOMX_FREQ_FSUB_R(16)
, RANDOMX_FREQ_FSUB_M(5)
, RANDOMX_FREQ_FSCAL_R(6)
, RANDOMX_FREQ_FMUL_R(32)
, RANDOMX_FREQ_FDIV_M(4)
, RANDOMX_FREQ_FSQRT_R(6)
, RANDOMX_FREQ_CBRANCH(16)
, RANDOMX_FREQ_CFROUND(1)
, RANDOMX_FREQ_ISTORE(16)
, RANDOMX_FREQ_NOP(0)
{
fillAes4Rx4_Key[0] = rx_set_int_vec_i128(0x99e5d23f, 0x2f546d2b, 0xd1833ddb, 0x6421aadd);
fillAes4Rx4_Key[1] = rx_set_int_vec_i128(0xa5dfcde5, 0x06f79d53, 0xb6913f55, 0xb20e3450);
fillAes4Rx4_Key[2] = rx_set_int_vec_i128(0x171c02bf, 0x0aa4679f, 0x515e7baf, 0x5c3ed904);
fillAes4Rx4_Key[3] = rx_set_int_vec_i128(0xd8ded291, 0xcd673785, 0xe78f5d08, 0x85623763);
fillAes4Rx4_Key[4] = rx_set_int_vec_i128(0x229effb4, 0x3d518b6d, 0xe3d6a7a6, 0xb5826f73);
fillAes4Rx4_Key[5] = rx_set_int_vec_i128(0xb272b7d2, 0xe9024d4e, 0x9c10b3d9, 0xc7566bf3);
fillAes4Rx4_Key[6] = rx_set_int_vec_i128(0xf63befa7, 0x2ba9660a, 0xf765a38b, 0xf273c9e7);
fillAes4Rx4_Key[7] = rx_set_int_vec_i128(0xc0b0762d, 0x0c06d1fd, 0x915839de, 0x7a7cd609);
#if defined(_M_X64) || defined(__x86_64__)
{
const uint8_t* a = (const uint8_t*)&randomx_sshash_prefetch;
const uint8_t* b = (const uint8_t*)&randomx_sshash_end;
memcpy(codeShhPrefetchTweaked, a, b - a);
}
{
const uint8_t* a = (const uint8_t*)&randomx_program_read_dataset;
const uint8_t* b = (const uint8_t*)&randomx_program_read_dataset_sshash_init;
memcpy(codeReadDatasetTweaked, a, b - a);
}
{
const uint8_t* a = (const uint8_t*)&randomx_program_read_dataset_sshash_init;
const uint8_t* b = (const uint8_t*)&randomx_program_read_dataset_sshash_fin;
memcpy(codeReadDatasetLightSshInitTweaked, a, b - a);
}
{
const uint8_t* a = (const uint8_t*)&randomx_program_loop_load;
const uint8_t* b = (const uint8_t*)&randomx_program_start;
memcpy(codeLoopLoadTweaked, a, b - a);
}
#endif
}
void RandomX_ConfigurationBase::Apply()
{
#if defined(_M_X64) || defined(__x86_64__)
*(uint32_t*)(codeShhPrefetchTweaked + 3) = ArgonMemory * 16 - 1;
const uint32_t DatasetBaseMask = DatasetBaseSize - RANDOMX_DATASET_ITEM_SIZE;
*(uint32_t*)(codeReadDatasetTweaked + 7) = DatasetBaseMask;
*(uint32_t*)(codeReadDatasetTweaked + 23) = DatasetBaseMask;
*(uint32_t*)(codeReadDatasetLightSshInitTweaked + 59) = DatasetBaseMask;
#endif
CacheLineAlignMask_Calculated = (DatasetBaseSize - 1) & ~(RANDOMX_DATASET_ITEM_SIZE - 1);
DatasetExtraItems_Calculated = DatasetExtraSize / RANDOMX_DATASET_ITEM_SIZE;
ScratchpadL1Mask_Calculated = (ScratchpadL1_Size / sizeof(uint64_t) - 1) * 8;
ScratchpadL1Mask16_Calculated = (ScratchpadL1_Size / sizeof(uint64_t) / 2 - 1) * 16;
ScratchpadL2Mask_Calculated = (ScratchpadL2_Size / sizeof(uint64_t) - 1) * 8;
ScratchpadL2Mask16_Calculated = (ScratchpadL2_Size / sizeof(uint64_t) / 2 - 1) * 16;
ScratchpadL3Mask_Calculated = (((ScratchpadL3_Size / sizeof(uint64_t)) - 1) * 8);
ScratchpadL3Mask64_Calculated = ((ScratchpadL3_Size / sizeof(uint64_t)) / 8 - 1) * 64;
#if defined(_M_X64) || defined(__x86_64__)
*(uint32_t*)(codeLoopLoadTweaked + 4) = ScratchpadL3Mask64_Calculated;
*(uint32_t*)(codeLoopLoadTweaked + 50) = ScratchpadL3Mask64_Calculated;
#endif
ConditionMask_Calculated = (1 << JumpBits) - 1;
constexpr int CEIL_NULL = 0;
int k = 0;
#if defined(_M_X64) || defined(__x86_64__)
#define JIT_HANDLE(x, prev) randomx::JitCompilerX86::engine[k] = &randomx::JitCompilerX86::h_##x
#else
#define JIT_HANDLE(x, prev)
#endif
#define INST_HANDLE(x, prev) \
CEIL_##x = CEIL_##prev + RANDOMX_FREQ_##x; \
for (; k < CEIL_##x; ++k) { JIT_HANDLE(x, prev); }
INST_HANDLE(IADD_RS, NULL);
INST_HANDLE(IADD_M, IADD_RS);
INST_HANDLE(ISUB_R, IADD_M);
INST_HANDLE(ISUB_M, ISUB_R);
INST_HANDLE(IMUL_R, ISUB_M);
INST_HANDLE(IMUL_M, IMUL_R);
INST_HANDLE(IMULH_R, IMUL_M);
INST_HANDLE(IMULH_M, IMULH_R);
INST_HANDLE(ISMULH_R, IMULH_M);
INST_HANDLE(ISMULH_M, ISMULH_R);
INST_HANDLE(IMUL_RCP, ISMULH_M);
INST_HANDLE(INEG_R, IMUL_RCP);
INST_HANDLE(IXOR_R, INEG_R);
INST_HANDLE(IXOR_M, IXOR_R);
INST_HANDLE(IROR_R, IXOR_M);
INST_HANDLE(IROL_R, IROR_R);
INST_HANDLE(ISWAP_R, IROL_R);
INST_HANDLE(FSWAP_R, ISWAP_R);
INST_HANDLE(FADD_R, FSWAP_R);
INST_HANDLE(FADD_M, FADD_R);
INST_HANDLE(FSUB_R, FADD_M);
INST_HANDLE(FSUB_M, FSUB_R);
INST_HANDLE(FSCAL_R, FSUB_M);
INST_HANDLE(FMUL_R, FSCAL_R);
INST_HANDLE(FDIV_M, FMUL_R);
INST_HANDLE(FSQRT_R, FDIV_M);
INST_HANDLE(CBRANCH, FSQRT_R);
INST_HANDLE(CFROUND, CBRANCH);
INST_HANDLE(ISTORE, CFROUND);
INST_HANDLE(NOP, ISTORE);
#undef INST_HANDLE
}
RandomX_ConfigurationMonero RandomX_MoneroConfig;
RandomX_ConfigurationWownero RandomX_WowneroConfig;
RandomX_ConfigurationLoki RandomX_LokiConfig;
RandomX_ConfigurationBase RandomX_CurrentConfig;
extern "C" {
randomx_cache *randomx_alloc_cache(randomx_flags flags) {
randomx_cache *cache;
try {
cache = new randomx_cache();
switch (flags & (RANDOMX_FLAG_JIT | RANDOMX_FLAG_LARGE_PAGES)) {
case RANDOMX_FLAG_DEFAULT:
cache->dealloc = &randomx::deallocCache<randomx::DefaultAllocator>;
cache->jit = nullptr;
cache->initialize = &randomx::initCache;
cache->datasetInit = &randomx::initDataset;
cache->memory = (uint8_t*)randomx::DefaultAllocator::allocMemory(RANDOMX_CACHE_MAX_SIZE);
break;
case RANDOMX_FLAG_JIT:
cache->dealloc = &randomx::deallocCache<randomx::DefaultAllocator>;
cache->jit = new randomx::JitCompiler();
cache->initialize = &randomx::initCacheCompile;
cache->datasetInit = cache->jit->getDatasetInitFunc();
cache->memory = (uint8_t*)randomx::DefaultAllocator::allocMemory(RANDOMX_CACHE_MAX_SIZE);
break;
case RANDOMX_FLAG_LARGE_PAGES:
cache->dealloc = &randomx::deallocCache<randomx::LargePageAllocator>;
cache->jit = nullptr;
cache->initialize = &randomx::initCache;
cache->datasetInit = &randomx::initDataset;
cache->memory = (uint8_t*)randomx::LargePageAllocator::allocMemory(RANDOMX_CACHE_MAX_SIZE);
break;
case RANDOMX_FLAG_JIT | RANDOMX_FLAG_LARGE_PAGES:
cache->dealloc = &randomx::deallocCache<randomx::LargePageAllocator>;
cache->jit = new randomx::JitCompiler();
cache->initialize = &randomx::initCacheCompile;
cache->datasetInit = cache->jit->getDatasetInitFunc();
cache->memory = (uint8_t*)randomx::LargePageAllocator::allocMemory(RANDOMX_CACHE_MAX_SIZE);
break;
default:
UNREACHABLE;
}
}
catch (std::exception &ex) {
if (cache != nullptr) {
randomx_release_cache(cache);
cache = nullptr;
}
}
return cache;
}
void randomx_init_cache(randomx_cache *cache, const void *key, size_t keySize) {
assert(cache != nullptr);
assert(keySize == 0 || key != nullptr);
cache->initialize(cache, key, keySize);
}
void randomx_release_cache(randomx_cache* cache) {
assert(cache != nullptr);
cache->dealloc(cache);
delete cache;
}
randomx_dataset *randomx_alloc_dataset(randomx_flags flags) {
randomx_dataset *dataset;
try {
dataset = new randomx_dataset();
if (flags & RANDOMX_FLAG_LARGE_PAGES) {
dataset->dealloc = &randomx::deallocDataset<randomx::LargePageAllocator>;
dataset->memory = (uint8_t*)randomx::LargePageAllocator::allocMemory(RANDOMX_DATASET_MAX_SIZE);
}
else {
dataset->dealloc = &randomx::deallocDataset<randomx::DefaultAllocator>;
dataset->memory = (uint8_t*)randomx::DefaultAllocator::allocMemory(RANDOMX_DATASET_MAX_SIZE);
}
}
catch (std::exception &ex) {
if (dataset != nullptr) {
randomx_release_dataset(dataset);
dataset = nullptr;
}
}
return dataset;
}
#define DatasetItemCount ((RandomX_CurrentConfig.DatasetBaseSize + RandomX_CurrentConfig.DatasetExtraSize) / RANDOMX_DATASET_ITEM_SIZE)
unsigned long randomx_dataset_item_count() {
return DatasetItemCount;
}
void randomx_init_dataset(randomx_dataset *dataset, randomx_cache *cache, unsigned long startItem, unsigned long itemCount) {
assert(dataset != nullptr);
assert(cache != nullptr);
assert(startItem < DatasetItemCount && itemCount <= DatasetItemCount);
assert(startItem + itemCount <= DatasetItemCount);
cache->datasetInit(cache, dataset->memory + startItem * randomx::CacheLineSize, startItem, startItem + itemCount);
}
void *randomx_get_dataset_memory(randomx_dataset *dataset) {
assert(dataset != nullptr);
return dataset->memory;
}
void randomx_release_dataset(randomx_dataset *dataset) {
assert(dataset != nullptr);
dataset->dealloc(dataset);
delete dataset;
}
randomx_vm *randomx_create_vm(randomx_flags flags, randomx_cache *cache, randomx_dataset *dataset) {
assert(cache != nullptr || (flags & RANDOMX_FLAG_FULL_MEM));
assert(cache == nullptr || cache->isInitialized());
assert(dataset != nullptr || !(flags & RANDOMX_FLAG_FULL_MEM));
randomx_vm *vm = nullptr;
try {
switch (flags & (RANDOMX_FLAG_FULL_MEM | RANDOMX_FLAG_JIT | RANDOMX_FLAG_HARD_AES | RANDOMX_FLAG_LARGE_PAGES)) {
case RANDOMX_FLAG_DEFAULT:
vm = new randomx::InterpretedLightVmDefault();
break;
case RANDOMX_FLAG_FULL_MEM:
vm = new randomx::InterpretedVmDefault();
break;
case RANDOMX_FLAG_JIT:
vm = new randomx::CompiledLightVmDefault();
break;
case RANDOMX_FLAG_FULL_MEM | RANDOMX_FLAG_JIT:
vm = new randomx::CompiledVmDefault();
break;
case RANDOMX_FLAG_HARD_AES:
vm = new randomx::InterpretedLightVmHardAes();
break;
case RANDOMX_FLAG_FULL_MEM | RANDOMX_FLAG_HARD_AES:
vm = new randomx::InterpretedVmHardAes();
break;
case RANDOMX_FLAG_JIT | RANDOMX_FLAG_HARD_AES:
vm = new randomx::CompiledLightVmHardAes();
break;
case RANDOMX_FLAG_FULL_MEM | RANDOMX_FLAG_JIT | RANDOMX_FLAG_HARD_AES:
vm = new randomx::CompiledVmHardAes();
break;
case RANDOMX_FLAG_LARGE_PAGES:
vm = new randomx::InterpretedLightVmLargePage();
break;
case RANDOMX_FLAG_FULL_MEM | RANDOMX_FLAG_LARGE_PAGES:
vm = new randomx::InterpretedVmLargePage();
break;
case RANDOMX_FLAG_JIT | RANDOMX_FLAG_LARGE_PAGES:
vm = new randomx::CompiledLightVmLargePage();
break;
case RANDOMX_FLAG_FULL_MEM | RANDOMX_FLAG_JIT | RANDOMX_FLAG_LARGE_PAGES:
vm = new randomx::CompiledVmLargePage();
break;
case RANDOMX_FLAG_HARD_AES | RANDOMX_FLAG_LARGE_PAGES:
vm = new randomx::InterpretedLightVmLargePageHardAes();
break;
case RANDOMX_FLAG_FULL_MEM | RANDOMX_FLAG_HARD_AES | RANDOMX_FLAG_LARGE_PAGES:
vm = new randomx::InterpretedVmLargePageHardAes();
break;
case RANDOMX_FLAG_JIT | RANDOMX_FLAG_HARD_AES | RANDOMX_FLAG_LARGE_PAGES:
vm = new randomx::CompiledLightVmLargePageHardAes();
break;
case RANDOMX_FLAG_FULL_MEM | RANDOMX_FLAG_JIT | RANDOMX_FLAG_HARD_AES | RANDOMX_FLAG_LARGE_PAGES:
vm = new randomx::CompiledVmLargePageHardAes();
break;
default:
UNREACHABLE;
}
if(cache != nullptr)
vm->setCache(cache);
if(dataset != nullptr)
vm->setDataset(dataset);
vm->allocate();
}
catch (std::exception &ex) {
delete vm;
vm = nullptr;
}
return vm;
}
void randomx_vm_set_cache(randomx_vm *machine, randomx_cache* cache) {
assert(machine != nullptr);
assert(cache != nullptr && cache->isInitialized());
machine->setCache(cache);
}
void randomx_vm_set_dataset(randomx_vm *machine, randomx_dataset *dataset) {
assert(machine != nullptr);
assert(dataset != nullptr);
machine->setDataset(dataset);
}
void randomx_destroy_vm(randomx_vm *machine) {
assert(machine != nullptr);
delete machine;
}
void randomx_calculate_hash(randomx_vm *machine, const void *input, size_t inputSize, void *output) {
assert(machine != nullptr);
assert(inputSize == 0 || input != nullptr);
assert(output != nullptr);
alignas(16) uint64_t tempHash[8];
int blakeResult = blake2b(tempHash, sizeof(tempHash), input, inputSize, nullptr, 0);
assert(blakeResult == 0);
machine->initScratchpad(&tempHash);
machine->resetRoundingMode();
for (int chain = 0; chain < RandomX_CurrentConfig.ProgramCount - 1; ++chain) {
machine->run(&tempHash);
blakeResult = blake2b(tempHash, sizeof(tempHash), machine->getRegisterFile(), sizeof(randomx::RegisterFile), nullptr, 0);
assert(blakeResult == 0);
}
machine->run(&tempHash);
machine->getFinalResult(output, RANDOMX_HASH_SIZE);
}
}

View file

@ -0,0 +1,332 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#ifndef RANDOMX_H
#define RANDOMX_H
#include <stddef.h>
#include <stdint.h>
#include <type_traits>
#include "intrin_portable.h"
#define RANDOMX_HASH_SIZE 32
#define RANDOMX_DATASET_ITEM_SIZE 64
#ifndef RANDOMX_EXPORT
#define RANDOMX_EXPORT
#endif
typedef enum {
RANDOMX_FLAG_DEFAULT = 0,
RANDOMX_FLAG_LARGE_PAGES = 1,
RANDOMX_FLAG_HARD_AES = 2,
RANDOMX_FLAG_FULL_MEM = 4,
RANDOMX_FLAG_JIT = 8,
} randomx_flags;
typedef struct randomx_dataset randomx_dataset;
typedef struct randomx_cache randomx_cache;
typedef struct randomx_vm randomx_vm;
struct RandomX_ConfigurationBase
{
RandomX_ConfigurationBase();
void Apply();
uint32_t ArgonMemory;
uint32_t ArgonIterations;
uint32_t ArgonLanes;
const char* ArgonSalt;
uint32_t CacheAccesses;
uint32_t SuperscalarLatency;
uint32_t DatasetBaseSize;
uint32_t DatasetExtraSize;
uint32_t ScratchpadL1_Size;
uint32_t ScratchpadL2_Size;
uint32_t ScratchpadL3_Size;
uint32_t ProgramSize;
uint32_t ProgramIterations;
uint32_t ProgramCount;
uint32_t JumpBits;
uint32_t JumpOffset;
uint32_t RANDOMX_FREQ_IADD_RS;
uint32_t RANDOMX_FREQ_IADD_M;
uint32_t RANDOMX_FREQ_ISUB_R;
uint32_t RANDOMX_FREQ_ISUB_M;
uint32_t RANDOMX_FREQ_IMUL_R;
uint32_t RANDOMX_FREQ_IMUL_M;
uint32_t RANDOMX_FREQ_IMULH_R;
uint32_t RANDOMX_FREQ_IMULH_M;
uint32_t RANDOMX_FREQ_ISMULH_R;
uint32_t RANDOMX_FREQ_ISMULH_M;
uint32_t RANDOMX_FREQ_IMUL_RCP;
uint32_t RANDOMX_FREQ_INEG_R;
uint32_t RANDOMX_FREQ_IXOR_R;
uint32_t RANDOMX_FREQ_IXOR_M;
uint32_t RANDOMX_FREQ_IROR_R;
uint32_t RANDOMX_FREQ_IROL_R;
uint32_t RANDOMX_FREQ_ISWAP_R;
uint32_t RANDOMX_FREQ_FSWAP_R;
uint32_t RANDOMX_FREQ_FADD_R;
uint32_t RANDOMX_FREQ_FADD_M;
uint32_t RANDOMX_FREQ_FSUB_R;
uint32_t RANDOMX_FREQ_FSUB_M;
uint32_t RANDOMX_FREQ_FSCAL_R;
uint32_t RANDOMX_FREQ_FMUL_R;
uint32_t RANDOMX_FREQ_FDIV_M;
uint32_t RANDOMX_FREQ_FSQRT_R;
uint32_t RANDOMX_FREQ_CBRANCH;
uint32_t RANDOMX_FREQ_CFROUND;
uint32_t RANDOMX_FREQ_ISTORE;
uint32_t RANDOMX_FREQ_NOP;
rx_vec_i128 fillAes4Rx4_Key[8];
uint8_t codeShhPrefetchTweaked[20];
uint8_t codeReadDatasetTweaked[64];
uint8_t codeReadDatasetLightSshInitTweaked[68];
uint8_t codeLoopLoadTweaked[140];
uint32_t CacheLineAlignMask_Calculated;
uint32_t DatasetExtraItems_Calculated;
uint32_t ScratchpadL1Mask_Calculated;
uint32_t ScratchpadL1Mask16_Calculated;
uint32_t ScratchpadL2Mask_Calculated;
uint32_t ScratchpadL2Mask16_Calculated;
uint32_t ScratchpadL3Mask_Calculated;
uint32_t ScratchpadL3Mask64_Calculated;
uint32_t ConditionMask_Calculated;
int CEIL_IADD_RS;
int CEIL_IADD_M;
int CEIL_ISUB_R;
int CEIL_ISUB_M;
int CEIL_IMUL_R;
int CEIL_IMUL_M;
int CEIL_IMULH_R;
int CEIL_IMULH_M;
int CEIL_ISMULH_R;
int CEIL_ISMULH_M;
int CEIL_IMUL_RCP;
int CEIL_INEG_R;
int CEIL_IXOR_R;
int CEIL_IXOR_M;
int CEIL_IROR_R;
int CEIL_IROL_R;
int CEIL_ISWAP_R;
int CEIL_FSWAP_R;
int CEIL_FADD_R;
int CEIL_FADD_M;
int CEIL_FSUB_R;
int CEIL_FSUB_M;
int CEIL_FSCAL_R;
int CEIL_FMUL_R;
int CEIL_FDIV_M;
int CEIL_FSQRT_R;
int CEIL_CBRANCH;
int CEIL_CFROUND;
int CEIL_ISTORE;
int CEIL_NOP;
};
struct RandomX_ConfigurationMonero : public RandomX_ConfigurationBase {};
struct RandomX_ConfigurationWownero : public RandomX_ConfigurationBase { RandomX_ConfigurationWownero(); };
struct RandomX_ConfigurationLoki : public RandomX_ConfigurationBase { RandomX_ConfigurationLoki(); };
extern RandomX_ConfigurationMonero RandomX_MoneroConfig;
extern RandomX_ConfigurationWownero RandomX_WowneroConfig;
extern RandomX_ConfigurationLoki RandomX_LokiConfig;
extern RandomX_ConfigurationBase RandomX_CurrentConfig;
template<typename T>
void randomx_apply_config(const T& config)
{
static_assert(sizeof(T) == sizeof(RandomX_ConfigurationBase), "Invalid RandomX configuration struct size");
static_assert(std::is_base_of<RandomX_ConfigurationBase, T>::value, "Incompatible RandomX configuration struct");
RandomX_CurrentConfig = config;
RandomX_CurrentConfig.Apply();
}
#if defined(__cplusplus)
extern "C" {
#endif
/**
* Creates a randomx_cache structure and allocates memory for RandomX Cache.
*
* @param flags is any combination of these 2 flags (each flag can be set or not set):
* RANDOMX_FLAG_LARGE_PAGES - allocate memory in large pages
* RANDOMX_FLAG_JIT - create cache structure with JIT compilation support; this makes
* subsequent Dataset initialization faster
*
* @return Pointer to an allocated randomx_cache structure.
* NULL is returned if memory allocation fails or if the RANDOMX_FLAG_JIT
* is set and JIT compilation is not supported on the current platform.
*/
RANDOMX_EXPORT randomx_cache *randomx_alloc_cache(randomx_flags flags);
/**
* Initializes the cache memory and SuperscalarHash using the provided key value.
*
* @param cache is a pointer to a previously allocated randomx_cache structure. Must not be NULL.
* @param key is a pointer to memory which contains the key value. Must not be NULL.
* @param keySize is the number of bytes of the key.
*/
RANDOMX_EXPORT void randomx_init_cache(randomx_cache *cache, const void *key, size_t keySize);
/**
* Releases all memory occupied by the randomx_cache structure.
*
* @param cache is a pointer to a previously allocated randomx_cache structure.
*/
RANDOMX_EXPORT void randomx_release_cache(randomx_cache* cache);
/**
* Creates a randomx_dataset structure and allocates memory for RandomX Dataset.
*
* @param flags is the initialization flags. Only one flag is supported (can be set or not set):
* RANDOMX_FLAG_LARGE_PAGES - allocate memory in large pages
*
* @return Pointer to an allocated randomx_dataset structure.
* NULL is returned if memory allocation fails.
*/
RANDOMX_EXPORT randomx_dataset *randomx_alloc_dataset(randomx_flags flags);
/**
* Gets the number of items contained in the dataset.
*
* @return the number of items contained in the dataset.
*/
RANDOMX_EXPORT unsigned long randomx_dataset_item_count(void);
/**
* Initializes dataset items.
*
* Note: In order to use the Dataset, all items from 0 to (randomx_dataset_item_count() - 1) must be initialized.
* This may be done by several calls to this function using non-overlapping item sequences.
*
* @param dataset is a pointer to a previously allocated randomx_dataset structure. Must not be NULL.
* @param cache is a pointer to a previously allocated and initialized randomx_cache structure. Must not be NULL.
* @param startItem is the item number where intialization should start.
* @param itemCount is the number of items that should be initialized.
*/
RANDOMX_EXPORT void randomx_init_dataset(randomx_dataset *dataset, randomx_cache *cache, unsigned long startItem, unsigned long itemCount);
/**
* Returns a pointer to the internal memory buffer of the dataset structure. The size
* of the internal memory buffer is randomx_dataset_item_count() * RANDOMX_DATASET_ITEM_SIZE.
*
* @param dataset is dataset is a pointer to a previously allocated randomx_dataset structure. Must not be NULL.
*
* @return Pointer to the internal memory buffer of the dataset structure.
*/
RANDOMX_EXPORT void *randomx_get_dataset_memory(randomx_dataset *dataset);
/**
* Releases all memory occupied by the randomx_dataset structure.
*
* @param dataset is a pointer to a previously allocated randomx_dataset structure.
*/
RANDOMX_EXPORT void randomx_release_dataset(randomx_dataset *dataset);
/**
* Creates and initializes a RandomX virtual machine.
*
* @param flags is any combination of these 4 flags (each flag can be set or not set):
* RANDOMX_FLAG_LARGE_PAGES - allocate scratchpad memory in large pages
* RANDOMX_FLAG_HARD_AES - virtual machine will use hardware accelerated AES
* RANDOMX_FLAG_FULL_MEM - virtual machine will use the full dataset
* RANDOMX_FLAG_JIT - virtual machine will use a JIT compiler
* The numeric values of the flags are ordered so that a higher value will provide
* faster hash calculation and a lower numeric value will provide higher portability.
* Using RANDOMX_FLAG_DEFAULT (all flags not set) works on all platforms, but is the slowest.
* @param cache is a pointer to an initialized randomx_cache structure. Can be
* NULL if RANDOMX_FLAG_FULL_MEM is set.
* @param dataset is a pointer to a randomx_dataset structure. Can be NULL
* if RANDOMX_FLAG_FULL_MEM is not set.
*
* @return Pointer to an initialized randomx_vm structure.
* Returns NULL if:
* (1) Scratchpad memory allocation fails.
* (2) The requested initialization flags are not supported on the current platform.
* (3) cache parameter is NULL and RANDOMX_FLAG_FULL_MEM is not set
* (4) dataset parameter is NULL and RANDOMX_FLAG_FULL_MEM is set
*/
RANDOMX_EXPORT randomx_vm *randomx_create_vm(randomx_flags flags, randomx_cache *cache, randomx_dataset *dataset);
/**
* Reinitializes a virtual machine with a new Cache. This function should be called anytime
* the Cache is reinitialized with a new key.
*
* @param machine is a pointer to a randomx_vm structure that was initialized
* without RANDOMX_FLAG_FULL_MEM. Must not be NULL.
* @param cache is a pointer to an initialized randomx_cache structure. Must not be NULL.
*/
RANDOMX_EXPORT void randomx_vm_set_cache(randomx_vm *machine, randomx_cache* cache);
/**
* Reinitializes a virtual machine with a new Dataset.
*
* @param machine is a pointer to a randomx_vm structure that was initialized
* with RANDOMX_FLAG_FULL_MEM. Must not be NULL.
* @param dataset is a pointer to an initialized randomx_dataset structure. Must not be NULL.
*/
RANDOMX_EXPORT void randomx_vm_set_dataset(randomx_vm *machine, randomx_dataset *dataset);
/**
* Releases all memory occupied by the randomx_vm structure.
*
* @param machine is a pointer to a previously created randomx_vm structure.
*/
RANDOMX_EXPORT void randomx_destroy_vm(randomx_vm *machine);
/**
* Calculates a RandomX hash value.
*
* @param machine is a pointer to a randomx_vm structure. Must not be NULL.
* @param input is a pointer to memory to be hashed. Must not be NULL.
* @param inputSize is the number of bytes to be hashed.
* @param output is a pointer to memory where the hash will be stored. Must not
* be NULL and at least RANDOMX_HASH_SIZE bytes must be available for writing.
*/
RANDOMX_EXPORT void randomx_calculate_hash(randomx_vm *machine, const void *input, size_t inputSize, void *output);
#if defined(__cplusplus)
}
#endif
#endif

View file

@ -0,0 +1,80 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <assert.h>
#include "reciprocal.h"
/*
Calculates rcp = 2**x / divisor for highest integer x such that rcp < 2**64.
divisor must not be 0 or a power of 2
Equivalent x86 assembly (divisor in rcx):
mov edx, 1
mov r8, rcx
xor eax, eax
bsr rcx, rcx
shl rdx, cl
div r8
ret
*/
uint64_t randomx_reciprocal(uint64_t divisor) {
assert(divisor != 0);
const uint64_t p2exp63 = 1ULL << 63;
uint64_t quotient = p2exp63 / divisor, remainder = p2exp63 % divisor;
unsigned bsr = 0; //highest set bit in divisor
for (uint64_t bit = divisor; bit > 0; bit >>= 1)
bsr++;
for (unsigned shift = 0; shift < bsr; shift++) {
if (remainder >= divisor - remainder) {
quotient = quotient * 2 + 1;
remainder = remainder * 2 - divisor;
}
else {
quotient = quotient * 2;
remainder = remainder * 2;
}
}
return quotient;
}
#if !RANDOMX_HAVE_FAST_RECIPROCAL
uint64_t randomx_reciprocal_fast(uint64_t divisor) {
return randomx_reciprocal(divisor);
}
#endif

View file

@ -0,0 +1,48 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <stdint.h>
#if defined(_M_X64) || defined(__x86_64__)
#define RANDOMX_HAVE_FAST_RECIPROCAL 1
#else
#define RANDOMX_HAVE_FAST_RECIPROCAL 0
#endif
#if defined(__cplusplus)
extern "C" {
#endif
uint64_t randomx_reciprocal(uint64_t);
uint64_t randomx_reciprocal_fast(uint64_t);
#if defined(__cplusplus)
}
#endif

View file

@ -0,0 +1,364 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "soft_aes.h"
alignas(16) const uint8_t sbox[256] = {
0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5, 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0, 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc, 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a, 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0, 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b, 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85, 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5, 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17, 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88, 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c, 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9, 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6, 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e, 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94, 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68, 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16,
};
alignas(16) const uint32_t lutEnc0[256] = {
0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6, 0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591,
0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56, 0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec,
0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa, 0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb,
0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45, 0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b,
0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c, 0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83,
0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9, 0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a,
0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d, 0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f,
0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df, 0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea,
0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34, 0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b,
0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d, 0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413,
0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1, 0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6,
0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972, 0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85,
0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed, 0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511,
0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe, 0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b,
0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05, 0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1,
0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142, 0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf,
0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3, 0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e,
0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a, 0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6,
0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3, 0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b,
0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428, 0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad,
0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14, 0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8,
0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4, 0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2,
0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda, 0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949,
0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf, 0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810,
0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c, 0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697,
0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e, 0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f,
0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc, 0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c,
0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969, 0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27,
0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122, 0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433,
0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9, 0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5,
0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a, 0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0,
0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e, 0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c,
};
alignas(16) const uint32_t lutEnc1[256] = {
0x6363c6a5, 0x7c7cf884, 0x7777ee99, 0x7b7bf68d, 0xf2f2ff0d, 0x6b6bd6bd, 0x6f6fdeb1, 0xc5c59154,
0x30306050, 0x01010203, 0x6767cea9, 0x2b2b567d, 0xfefee719, 0xd7d7b562, 0xabab4de6, 0x7676ec9a,
0xcaca8f45, 0x82821f9d, 0xc9c98940, 0x7d7dfa87, 0xfafaef15, 0x5959b2eb, 0x47478ec9, 0xf0f0fb0b,
0xadad41ec, 0xd4d4b367, 0xa2a25ffd, 0xafaf45ea, 0x9c9c23bf, 0xa4a453f7, 0x7272e496, 0xc0c09b5b,
0xb7b775c2, 0xfdfde11c, 0x93933dae, 0x26264c6a, 0x36366c5a, 0x3f3f7e41, 0xf7f7f502, 0xcccc834f,
0x3434685c, 0xa5a551f4, 0xe5e5d134, 0xf1f1f908, 0x7171e293, 0xd8d8ab73, 0x31316253, 0x15152a3f,
0x0404080c, 0xc7c79552, 0x23234665, 0xc3c39d5e, 0x18183028, 0x969637a1, 0x05050a0f, 0x9a9a2fb5,
0x07070e09, 0x12122436, 0x80801b9b, 0xe2e2df3d, 0xebebcd26, 0x27274e69, 0xb2b27fcd, 0x7575ea9f,
0x0909121b, 0x83831d9e, 0x2c2c5874, 0x1a1a342e, 0x1b1b362d, 0x6e6edcb2, 0x5a5ab4ee, 0xa0a05bfb,
0x5252a4f6, 0x3b3b764d, 0xd6d6b761, 0xb3b37dce, 0x2929527b, 0xe3e3dd3e, 0x2f2f5e71, 0x84841397,
0x5353a6f5, 0xd1d1b968, 0x00000000, 0xededc12c, 0x20204060, 0xfcfce31f, 0xb1b179c8, 0x5b5bb6ed,
0x6a6ad4be, 0xcbcb8d46, 0xbebe67d9, 0x3939724b, 0x4a4a94de, 0x4c4c98d4, 0x5858b0e8, 0xcfcf854a,
0xd0d0bb6b, 0xefefc52a, 0xaaaa4fe5, 0xfbfbed16, 0x434386c5, 0x4d4d9ad7, 0x33336655, 0x85851194,
0x45458acf, 0xf9f9e910, 0x02020406, 0x7f7ffe81, 0x5050a0f0, 0x3c3c7844, 0x9f9f25ba, 0xa8a84be3,
0x5151a2f3, 0xa3a35dfe, 0x404080c0, 0x8f8f058a, 0x92923fad, 0x9d9d21bc, 0x38387048, 0xf5f5f104,
0xbcbc63df, 0xb6b677c1, 0xdadaaf75, 0x21214263, 0x10102030, 0xffffe51a, 0xf3f3fd0e, 0xd2d2bf6d,
0xcdcd814c, 0x0c0c1814, 0x13132635, 0xececc32f, 0x5f5fbee1, 0x979735a2, 0x444488cc, 0x17172e39,
0xc4c49357, 0xa7a755f2, 0x7e7efc82, 0x3d3d7a47, 0x6464c8ac, 0x5d5dbae7, 0x1919322b, 0x7373e695,
0x6060c0a0, 0x81811998, 0x4f4f9ed1, 0xdcdca37f, 0x22224466, 0x2a2a547e, 0x90903bab, 0x88880b83,
0x46468cca, 0xeeeec729, 0xb8b86bd3, 0x1414283c, 0xdedea779, 0x5e5ebce2, 0x0b0b161d, 0xdbdbad76,
0xe0e0db3b, 0x32326456, 0x3a3a744e, 0x0a0a141e, 0x494992db, 0x06060c0a, 0x2424486c, 0x5c5cb8e4,
0xc2c29f5d, 0xd3d3bd6e, 0xacac43ef, 0x6262c4a6, 0x919139a8, 0x959531a4, 0xe4e4d337, 0x7979f28b,
0xe7e7d532, 0xc8c88b43, 0x37376e59, 0x6d6ddab7, 0x8d8d018c, 0xd5d5b164, 0x4e4e9cd2, 0xa9a949e0,
0x6c6cd8b4, 0x5656acfa, 0xf4f4f307, 0xeaeacf25, 0x6565caaf, 0x7a7af48e, 0xaeae47e9, 0x08081018,
0xbaba6fd5, 0x7878f088, 0x25254a6f, 0x2e2e5c72, 0x1c1c3824, 0xa6a657f1, 0xb4b473c7, 0xc6c69751,
0xe8e8cb23, 0xdddda17c, 0x7474e89c, 0x1f1f3e21, 0x4b4b96dd, 0xbdbd61dc, 0x8b8b0d86, 0x8a8a0f85,
0x7070e090, 0x3e3e7c42, 0xb5b571c4, 0x6666ccaa, 0x484890d8, 0x03030605, 0xf6f6f701, 0x0e0e1c12,
0x6161c2a3, 0x35356a5f, 0x5757aef9, 0xb9b969d0, 0x86861791, 0xc1c19958, 0x1d1d3a27, 0x9e9e27b9,
0xe1e1d938, 0xf8f8eb13, 0x98982bb3, 0x11112233, 0x6969d2bb, 0xd9d9a970, 0x8e8e0789, 0x949433a7,
0x9b9b2db6, 0x1e1e3c22, 0x87871592, 0xe9e9c920, 0xcece8749, 0x5555aaff, 0x28285078, 0xdfdfa57a,
0x8c8c038f, 0xa1a159f8, 0x89890980, 0x0d0d1a17, 0xbfbf65da, 0xe6e6d731, 0x424284c6, 0x6868d0b8,
0x414182c3, 0x999929b0, 0x2d2d5a77, 0x0f0f1e11, 0xb0b07bcb, 0x5454a8fc, 0xbbbb6dd6, 0x16162c3a,
};
alignas(16) const uint32_t lutEnc2[256] = {
0x63c6a563, 0x7cf8847c, 0x77ee9977, 0x7bf68d7b, 0xf2ff0df2, 0x6bd6bd6b, 0x6fdeb16f, 0xc59154c5,
0x30605030, 0x01020301, 0x67cea967, 0x2b567d2b, 0xfee719fe, 0xd7b562d7, 0xab4de6ab, 0x76ec9a76,
0xca8f45ca, 0x821f9d82, 0xc98940c9, 0x7dfa877d, 0xfaef15fa, 0x59b2eb59, 0x478ec947, 0xf0fb0bf0,
0xad41ecad, 0xd4b367d4, 0xa25ffda2, 0xaf45eaaf, 0x9c23bf9c, 0xa453f7a4, 0x72e49672, 0xc09b5bc0,
0xb775c2b7, 0xfde11cfd, 0x933dae93, 0x264c6a26, 0x366c5a36, 0x3f7e413f, 0xf7f502f7, 0xcc834fcc,
0x34685c34, 0xa551f4a5, 0xe5d134e5, 0xf1f908f1, 0x71e29371, 0xd8ab73d8, 0x31625331, 0x152a3f15,
0x04080c04, 0xc79552c7, 0x23466523, 0xc39d5ec3, 0x18302818, 0x9637a196, 0x050a0f05, 0x9a2fb59a,
0x070e0907, 0x12243612, 0x801b9b80, 0xe2df3de2, 0xebcd26eb, 0x274e6927, 0xb27fcdb2, 0x75ea9f75,
0x09121b09, 0x831d9e83, 0x2c58742c, 0x1a342e1a, 0x1b362d1b, 0x6edcb26e, 0x5ab4ee5a, 0xa05bfba0,
0x52a4f652, 0x3b764d3b, 0xd6b761d6, 0xb37dceb3, 0x29527b29, 0xe3dd3ee3, 0x2f5e712f, 0x84139784,
0x53a6f553, 0xd1b968d1, 0x00000000, 0xedc12ced, 0x20406020, 0xfce31ffc, 0xb179c8b1, 0x5bb6ed5b,
0x6ad4be6a, 0xcb8d46cb, 0xbe67d9be, 0x39724b39, 0x4a94de4a, 0x4c98d44c, 0x58b0e858, 0xcf854acf,
0xd0bb6bd0, 0xefc52aef, 0xaa4fe5aa, 0xfbed16fb, 0x4386c543, 0x4d9ad74d, 0x33665533, 0x85119485,
0x458acf45, 0xf9e910f9, 0x02040602, 0x7ffe817f, 0x50a0f050, 0x3c78443c, 0x9f25ba9f, 0xa84be3a8,
0x51a2f351, 0xa35dfea3, 0x4080c040, 0x8f058a8f, 0x923fad92, 0x9d21bc9d, 0x38704838, 0xf5f104f5,
0xbc63dfbc, 0xb677c1b6, 0xdaaf75da, 0x21426321, 0x10203010, 0xffe51aff, 0xf3fd0ef3, 0xd2bf6dd2,
0xcd814ccd, 0x0c18140c, 0x13263513, 0xecc32fec, 0x5fbee15f, 0x9735a297, 0x4488cc44, 0x172e3917,
0xc49357c4, 0xa755f2a7, 0x7efc827e, 0x3d7a473d, 0x64c8ac64, 0x5dbae75d, 0x19322b19, 0x73e69573,
0x60c0a060, 0x81199881, 0x4f9ed14f, 0xdca37fdc, 0x22446622, 0x2a547e2a, 0x903bab90, 0x880b8388,
0x468cca46, 0xeec729ee, 0xb86bd3b8, 0x14283c14, 0xdea779de, 0x5ebce25e, 0x0b161d0b, 0xdbad76db,
0xe0db3be0, 0x32645632, 0x3a744e3a, 0x0a141e0a, 0x4992db49, 0x060c0a06, 0x24486c24, 0x5cb8e45c,
0xc29f5dc2, 0xd3bd6ed3, 0xac43efac, 0x62c4a662, 0x9139a891, 0x9531a495, 0xe4d337e4, 0x79f28b79,
0xe7d532e7, 0xc88b43c8, 0x376e5937, 0x6ddab76d, 0x8d018c8d, 0xd5b164d5, 0x4e9cd24e, 0xa949e0a9,
0x6cd8b46c, 0x56acfa56, 0xf4f307f4, 0xeacf25ea, 0x65caaf65, 0x7af48e7a, 0xae47e9ae, 0x08101808,
0xba6fd5ba, 0x78f08878, 0x254a6f25, 0x2e5c722e, 0x1c38241c, 0xa657f1a6, 0xb473c7b4, 0xc69751c6,
0xe8cb23e8, 0xdda17cdd, 0x74e89c74, 0x1f3e211f, 0x4b96dd4b, 0xbd61dcbd, 0x8b0d868b, 0x8a0f858a,
0x70e09070, 0x3e7c423e, 0xb571c4b5, 0x66ccaa66, 0x4890d848, 0x03060503, 0xf6f701f6, 0x0e1c120e,
0x61c2a361, 0x356a5f35, 0x57aef957, 0xb969d0b9, 0x86179186, 0xc19958c1, 0x1d3a271d, 0x9e27b99e,
0xe1d938e1, 0xf8eb13f8, 0x982bb398, 0x11223311, 0x69d2bb69, 0xd9a970d9, 0x8e07898e, 0x9433a794,
0x9b2db69b, 0x1e3c221e, 0x87159287, 0xe9c920e9, 0xce8749ce, 0x55aaff55, 0x28507828, 0xdfa57adf,
0x8c038f8c, 0xa159f8a1, 0x89098089, 0x0d1a170d, 0xbf65dabf, 0xe6d731e6, 0x4284c642, 0x68d0b868,
0x4182c341, 0x9929b099, 0x2d5a772d, 0x0f1e110f, 0xb07bcbb0, 0x54a8fc54, 0xbb6dd6bb, 0x162c3a16,
};
alignas(16) const uint32_t lutEnc3[256] = {
0xc6a56363, 0xf8847c7c, 0xee997777, 0xf68d7b7b, 0xff0df2f2, 0xd6bd6b6b, 0xdeb16f6f, 0x9154c5c5,
0x60503030, 0x02030101, 0xcea96767, 0x567d2b2b, 0xe719fefe, 0xb562d7d7, 0x4de6abab, 0xec9a7676,
0x8f45caca, 0x1f9d8282, 0x8940c9c9, 0xfa877d7d, 0xef15fafa, 0xb2eb5959, 0x8ec94747, 0xfb0bf0f0,
0x41ecadad, 0xb367d4d4, 0x5ffda2a2, 0x45eaafaf, 0x23bf9c9c, 0x53f7a4a4, 0xe4967272, 0x9b5bc0c0,
0x75c2b7b7, 0xe11cfdfd, 0x3dae9393, 0x4c6a2626, 0x6c5a3636, 0x7e413f3f, 0xf502f7f7, 0x834fcccc,
0x685c3434, 0x51f4a5a5, 0xd134e5e5, 0xf908f1f1, 0xe2937171, 0xab73d8d8, 0x62533131, 0x2a3f1515,
0x080c0404, 0x9552c7c7, 0x46652323, 0x9d5ec3c3, 0x30281818, 0x37a19696, 0x0a0f0505, 0x2fb59a9a,
0x0e090707, 0x24361212, 0x1b9b8080, 0xdf3de2e2, 0xcd26ebeb, 0x4e692727, 0x7fcdb2b2, 0xea9f7575,
0x121b0909, 0x1d9e8383, 0x58742c2c, 0x342e1a1a, 0x362d1b1b, 0xdcb26e6e, 0xb4ee5a5a, 0x5bfba0a0,
0xa4f65252, 0x764d3b3b, 0xb761d6d6, 0x7dceb3b3, 0x527b2929, 0xdd3ee3e3, 0x5e712f2f, 0x13978484,
0xa6f55353, 0xb968d1d1, 0x00000000, 0xc12ceded, 0x40602020, 0xe31ffcfc, 0x79c8b1b1, 0xb6ed5b5b,
0xd4be6a6a, 0x8d46cbcb, 0x67d9bebe, 0x724b3939, 0x94de4a4a, 0x98d44c4c, 0xb0e85858, 0x854acfcf,
0xbb6bd0d0, 0xc52aefef, 0x4fe5aaaa, 0xed16fbfb, 0x86c54343, 0x9ad74d4d, 0x66553333, 0x11948585,
0x8acf4545, 0xe910f9f9, 0x04060202, 0xfe817f7f, 0xa0f05050, 0x78443c3c, 0x25ba9f9f, 0x4be3a8a8,
0xa2f35151, 0x5dfea3a3, 0x80c04040, 0x058a8f8f, 0x3fad9292, 0x21bc9d9d, 0x70483838, 0xf104f5f5,
0x63dfbcbc, 0x77c1b6b6, 0xaf75dada, 0x42632121, 0x20301010, 0xe51affff, 0xfd0ef3f3, 0xbf6dd2d2,
0x814ccdcd, 0x18140c0c, 0x26351313, 0xc32fecec, 0xbee15f5f, 0x35a29797, 0x88cc4444, 0x2e391717,
0x9357c4c4, 0x55f2a7a7, 0xfc827e7e, 0x7a473d3d, 0xc8ac6464, 0xbae75d5d, 0x322b1919, 0xe6957373,
0xc0a06060, 0x19988181, 0x9ed14f4f, 0xa37fdcdc, 0x44662222, 0x547e2a2a, 0x3bab9090, 0x0b838888,
0x8cca4646, 0xc729eeee, 0x6bd3b8b8, 0x283c1414, 0xa779dede, 0xbce25e5e, 0x161d0b0b, 0xad76dbdb,
0xdb3be0e0, 0x64563232, 0x744e3a3a, 0x141e0a0a, 0x92db4949, 0x0c0a0606, 0x486c2424, 0xb8e45c5c,
0x9f5dc2c2, 0xbd6ed3d3, 0x43efacac, 0xc4a66262, 0x39a89191, 0x31a49595, 0xd337e4e4, 0xf28b7979,
0xd532e7e7, 0x8b43c8c8, 0x6e593737, 0xdab76d6d, 0x018c8d8d, 0xb164d5d5, 0x9cd24e4e, 0x49e0a9a9,
0xd8b46c6c, 0xacfa5656, 0xf307f4f4, 0xcf25eaea, 0xcaaf6565, 0xf48e7a7a, 0x47e9aeae, 0x10180808,
0x6fd5baba, 0xf0887878, 0x4a6f2525, 0x5c722e2e, 0x38241c1c, 0x57f1a6a6, 0x73c7b4b4, 0x9751c6c6,
0xcb23e8e8, 0xa17cdddd, 0xe89c7474, 0x3e211f1f, 0x96dd4b4b, 0x61dcbdbd, 0x0d868b8b, 0x0f858a8a,
0xe0907070, 0x7c423e3e, 0x71c4b5b5, 0xccaa6666, 0x90d84848, 0x06050303, 0xf701f6f6, 0x1c120e0e,
0xc2a36161, 0x6a5f3535, 0xaef95757, 0x69d0b9b9, 0x17918686, 0x9958c1c1, 0x3a271d1d, 0x27b99e9e,
0xd938e1e1, 0xeb13f8f8, 0x2bb39898, 0x22331111, 0xd2bb6969, 0xa970d9d9, 0x07898e8e, 0x33a79494,
0x2db69b9b, 0x3c221e1e, 0x15928787, 0xc920e9e9, 0x8749cece, 0xaaff5555, 0x50782828, 0xa57adfdf,
0x038f8c8c, 0x59f8a1a1, 0x09808989, 0x1a170d0d, 0x65dabfbf, 0xd731e6e6, 0x84c64242, 0xd0b86868,
0x82c34141, 0x29b09999, 0x5a772d2d, 0x1e110f0f, 0x7bcbb0b0, 0xa8fc5454, 0x6dd6bbbb, 0x2c3a1616,
};
alignas(16) const uint32_t lutDec0[256] = {
0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a, 0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b,
0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5, 0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5,
0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d, 0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b,
0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295, 0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e,
0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927, 0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d,
0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362, 0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9,
0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52, 0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566,
0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3, 0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed,
0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e, 0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4,
0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4, 0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd,
0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d, 0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060,
0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967, 0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879,
0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000, 0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c,
0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36, 0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624,
0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b, 0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c,
0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12, 0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14,
0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3, 0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b,
0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8, 0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684,
0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7, 0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177,
0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947, 0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322,
0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498, 0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f,
0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54, 0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382,
0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf, 0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb,
0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83, 0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef,
0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029, 0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235,
0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733, 0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117,
0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4, 0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546,
0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb, 0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d,
0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb, 0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a,
0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773, 0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478,
0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2, 0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff,
0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664, 0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0,
};
alignas(16) const uint32_t lutDec1[256] = {
0xa7f45150, 0x65417e53, 0xa4171ac3, 0x5e273a96, 0x6bab3bcb, 0x459d1ff1, 0x58faacab, 0x03e34b93,
0xfa302055, 0x6d76adf6, 0x76cc8891, 0x4c02f525, 0xd7e54ffc, 0xcb2ac5d7, 0x44352680, 0xa362b58f,
0x5ab1de49, 0x1bba2567, 0x0eea4598, 0xc0fe5de1, 0x752fc302, 0xf04c8112, 0x97468da3, 0xf9d36bc6,
0x5f8f03e7, 0x9c921595, 0x7a6dbfeb, 0x595295da, 0x83bed42d, 0x217458d3, 0x69e04929, 0xc8c98e44,
0x89c2756a, 0x798ef478, 0x3e58996b, 0x71b927dd, 0x4fe1beb6, 0xad88f017, 0xac20c966, 0x3ace7db4,
0x4adf6318, 0x311ae582, 0x33519760, 0x7f536245, 0x7764b1e0, 0xae6bbb84, 0xa081fe1c, 0x2b08f994,
0x68487058, 0xfd458f19, 0x6cde9487, 0xf87b52b7, 0xd373ab23, 0x024b72e2, 0x8f1fe357, 0xab55662a,
0x28ebb207, 0xc2b52f03, 0x7bc5869a, 0x0837d3a5, 0x872830f2, 0xa5bf23b2, 0x6a0302ba, 0x8216ed5c,
0x1ccf8a2b, 0xb479a792, 0xf207f3f0, 0xe2694ea1, 0xf4da65cd, 0xbe0506d5, 0x6234d11f, 0xfea6c48a,
0x532e349d, 0x55f3a2a0, 0xe18a0532, 0xebf6a475, 0xec830b39, 0xef6040aa, 0x9f715e06, 0x106ebd51,
0x8a213ef9, 0x06dd963d, 0x053eddae, 0xbde64d46, 0x8d5491b5, 0x5dc47105, 0xd406046f, 0x155060ff,
0xfb981924, 0xe9bdd697, 0x434089cc, 0x9ed96777, 0x42e8b0bd, 0x8b890788, 0x5b19e738, 0xeec879db,
0x0a7ca147, 0x0f427ce9, 0x1e84f8c9, 0x00000000, 0x86800983, 0xed2b3248, 0x70111eac, 0x725a6c4e,
0xff0efdfb, 0x38850f56, 0xd5ae3d1e, 0x392d3627, 0xd90f0a64, 0xa65c6821, 0x545b9bd1, 0x2e36243a,
0x670a0cb1, 0xe757930f, 0x96eeb4d2, 0x919b1b9e, 0xc5c0804f, 0x20dc61a2, 0x4b775a69, 0x1a121c16,
0xba93e20a, 0x2aa0c0e5, 0xe0223c43, 0x171b121d, 0x0d090e0b, 0xc78bf2ad, 0xa8b62db9, 0xa91e14c8,
0x19f15785, 0x0775af4c, 0xdd99eebb, 0x607fa3fd, 0x2601f79f, 0xf5725cbc, 0x3b6644c5, 0x7efb5b34,
0x29438b76, 0xc623cbdc, 0xfcedb668, 0xf1e4b863, 0xdc31d7ca, 0x85634210, 0x22971340, 0x11c68420,
0x244a857d, 0x3dbbd2f8, 0x32f9ae11, 0xa129c76d, 0x2f9e1d4b, 0x30b2dcf3, 0x52860dec, 0xe3c177d0,
0x16b32b6c, 0xb970a999, 0x489411fa, 0x64e94722, 0x8cfca8c4, 0x3ff0a01a, 0x2c7d56d8, 0x903322ef,
0x4e4987c7, 0xd138d9c1, 0xa2ca8cfe, 0x0bd49836, 0x81f5a6cf, 0xde7aa528, 0x8eb7da26, 0xbfad3fa4,
0x9d3a2ce4, 0x9278500d, 0xcc5f6a9b, 0x467e5462, 0x138df6c2, 0xb8d890e8, 0xf7392e5e, 0xafc382f5,
0x805d9fbe, 0x93d0697c, 0x2dd56fa9, 0x1225cfb3, 0x99acc83b, 0x7d1810a7, 0x639ce86e, 0xbb3bdb7b,
0x7826cd09, 0x18596ef4, 0xb79aec01, 0x9a4f83a8, 0x6e95e665, 0xe6ffaa7e, 0xcfbc2108, 0xe815efe6,
0x9be7bad9, 0x366f4ace, 0x099fead4, 0x7cb029d6, 0xb2a431af, 0x233f2a31, 0x94a5c630, 0x66a235c0,
0xbc4e7437, 0xca82fca6, 0xd090e0b0, 0xd8a73315, 0x9804f14a, 0xdaec41f7, 0x50cd7f0e, 0xf691172f,
0xd64d768d, 0xb0ef434d, 0x4daacc54, 0x0496e4df, 0xb5d19ee3, 0x886a4c1b, 0x1f2cc1b8, 0x5165467f,
0xea5e9d04, 0x358c015d, 0x7487fa73, 0x410bfb2e, 0x1d67b35a, 0xd2db9252, 0x5610e933, 0x47d66d13,
0x61d79a8c, 0x0ca1377a, 0x14f8598e, 0x3c13eb89, 0x27a9ceee, 0xc961b735, 0xe51ce1ed, 0xb1477a3c,
0xdfd29c59, 0x73f2553f, 0xce141879, 0x37c773bf, 0xcdf753ea, 0xaafd5f5b, 0x6f3ddf14, 0xdb447886,
0xf3afca81, 0xc468b93e, 0x3424382c, 0x40a3c25f, 0xc31d1672, 0x25e2bc0c, 0x493c288b, 0x950dff41,
0x01a83971, 0xb30c08de, 0xe4b4d89c, 0xc1566490, 0x84cb7b61, 0xb632d570, 0x5c6c4874, 0x57b8d042,
};
alignas(16) const uint32_t lutDec2[256] = {
0xf45150a7, 0x417e5365, 0x171ac3a4, 0x273a965e, 0xab3bcb6b, 0x9d1ff145, 0xfaacab58, 0xe34b9303,
0x302055fa, 0x76adf66d, 0xcc889176, 0x02f5254c, 0xe54ffcd7, 0x2ac5d7cb, 0x35268044, 0x62b58fa3,
0xb1de495a, 0xba25671b, 0xea45980e, 0xfe5de1c0, 0x2fc30275, 0x4c8112f0, 0x468da397, 0xd36bc6f9,
0x8f03e75f, 0x9215959c, 0x6dbfeb7a, 0x5295da59, 0xbed42d83, 0x7458d321, 0xe0492969, 0xc98e44c8,
0xc2756a89, 0x8ef47879, 0x58996b3e, 0xb927dd71, 0xe1beb64f, 0x88f017ad, 0x20c966ac, 0xce7db43a,
0xdf63184a, 0x1ae58231, 0x51976033, 0x5362457f, 0x64b1e077, 0x6bbb84ae, 0x81fe1ca0, 0x08f9942b,
0x48705868, 0x458f19fd, 0xde94876c, 0x7b52b7f8, 0x73ab23d3, 0x4b72e202, 0x1fe3578f, 0x55662aab,
0xebb20728, 0xb52f03c2, 0xc5869a7b, 0x37d3a508, 0x2830f287, 0xbf23b2a5, 0x0302ba6a, 0x16ed5c82,
0xcf8a2b1c, 0x79a792b4, 0x07f3f0f2, 0x694ea1e2, 0xda65cdf4, 0x0506d5be, 0x34d11f62, 0xa6c48afe,
0x2e349d53, 0xf3a2a055, 0x8a0532e1, 0xf6a475eb, 0x830b39ec, 0x6040aaef, 0x715e069f, 0x6ebd5110,
0x213ef98a, 0xdd963d06, 0x3eddae05, 0xe64d46bd, 0x5491b58d, 0xc471055d, 0x06046fd4, 0x5060ff15,
0x981924fb, 0xbdd697e9, 0x4089cc43, 0xd967779e, 0xe8b0bd42, 0x8907888b, 0x19e7385b, 0xc879dbee,
0x7ca1470a, 0x427ce90f, 0x84f8c91e, 0x00000000, 0x80098386, 0x2b3248ed, 0x111eac70, 0x5a6c4e72,
0x0efdfbff, 0x850f5638, 0xae3d1ed5, 0x2d362739, 0x0f0a64d9, 0x5c6821a6, 0x5b9bd154, 0x36243a2e,
0x0a0cb167, 0x57930fe7, 0xeeb4d296, 0x9b1b9e91, 0xc0804fc5, 0xdc61a220, 0x775a694b, 0x121c161a,
0x93e20aba, 0xa0c0e52a, 0x223c43e0, 0x1b121d17, 0x090e0b0d, 0x8bf2adc7, 0xb62db9a8, 0x1e14c8a9,
0xf1578519, 0x75af4c07, 0x99eebbdd, 0x7fa3fd60, 0x01f79f26, 0x725cbcf5, 0x6644c53b, 0xfb5b347e,
0x438b7629, 0x23cbdcc6, 0xedb668fc, 0xe4b863f1, 0x31d7cadc, 0x63421085, 0x97134022, 0xc6842011,
0x4a857d24, 0xbbd2f83d, 0xf9ae1132, 0x29c76da1, 0x9e1d4b2f, 0xb2dcf330, 0x860dec52, 0xc177d0e3,
0xb32b6c16, 0x70a999b9, 0x9411fa48, 0xe9472264, 0xfca8c48c, 0xf0a01a3f, 0x7d56d82c, 0x3322ef90,
0x4987c74e, 0x38d9c1d1, 0xca8cfea2, 0xd498360b, 0xf5a6cf81, 0x7aa528de, 0xb7da268e, 0xad3fa4bf,
0x3a2ce49d, 0x78500d92, 0x5f6a9bcc, 0x7e546246, 0x8df6c213, 0xd890e8b8, 0x392e5ef7, 0xc382f5af,
0x5d9fbe80, 0xd0697c93, 0xd56fa92d, 0x25cfb312, 0xacc83b99, 0x1810a77d, 0x9ce86e63, 0x3bdb7bbb,
0x26cd0978, 0x596ef418, 0x9aec01b7, 0x4f83a89a, 0x95e6656e, 0xffaa7ee6, 0xbc2108cf, 0x15efe6e8,
0xe7bad99b, 0x6f4ace36, 0x9fead409, 0xb029d67c, 0xa431afb2, 0x3f2a3123, 0xa5c63094, 0xa235c066,
0x4e7437bc, 0x82fca6ca, 0x90e0b0d0, 0xa73315d8, 0x04f14a98, 0xec41f7da, 0xcd7f0e50, 0x91172ff6,
0x4d768dd6, 0xef434db0, 0xaacc544d, 0x96e4df04, 0xd19ee3b5, 0x6a4c1b88, 0x2cc1b81f, 0x65467f51,
0x5e9d04ea, 0x8c015d35, 0x87fa7374, 0x0bfb2e41, 0x67b35a1d, 0xdb9252d2, 0x10e93356, 0xd66d1347,
0xd79a8c61, 0xa1377a0c, 0xf8598e14, 0x13eb893c, 0xa9ceee27, 0x61b735c9, 0x1ce1ede5, 0x477a3cb1,
0xd29c59df, 0xf2553f73, 0x141879ce, 0xc773bf37, 0xf753eacd, 0xfd5f5baa, 0x3ddf146f, 0x447886db,
0xafca81f3, 0x68b93ec4, 0x24382c34, 0xa3c25f40, 0x1d1672c3, 0xe2bc0c25, 0x3c288b49, 0x0dff4195,
0xa8397101, 0x0c08deb3, 0xb4d89ce4, 0x566490c1, 0xcb7b6184, 0x32d570b6, 0x6c48745c, 0xb8d04257,
};
alignas(16) const uint32_t lutDec3[256] = {
0x5150a7f4, 0x7e536541, 0x1ac3a417, 0x3a965e27, 0x3bcb6bab, 0x1ff1459d, 0xacab58fa, 0x4b9303e3,
0x2055fa30, 0xadf66d76, 0x889176cc, 0xf5254c02, 0x4ffcd7e5, 0xc5d7cb2a, 0x26804435, 0xb58fa362,
0xde495ab1, 0x25671bba, 0x45980eea, 0x5de1c0fe, 0xc302752f, 0x8112f04c, 0x8da39746, 0x6bc6f9d3,
0x03e75f8f, 0x15959c92, 0xbfeb7a6d, 0x95da5952, 0xd42d83be, 0x58d32174, 0x492969e0, 0x8e44c8c9,
0x756a89c2, 0xf478798e, 0x996b3e58, 0x27dd71b9, 0xbeb64fe1, 0xf017ad88, 0xc966ac20, 0x7db43ace,
0x63184adf, 0xe582311a, 0x97603351, 0x62457f53, 0xb1e07764, 0xbb84ae6b, 0xfe1ca081, 0xf9942b08,
0x70586848, 0x8f19fd45, 0x94876cde, 0x52b7f87b, 0xab23d373, 0x72e2024b, 0xe3578f1f, 0x662aab55,
0xb20728eb, 0x2f03c2b5, 0x869a7bc5, 0xd3a50837, 0x30f28728, 0x23b2a5bf, 0x02ba6a03, 0xed5c8216,
0x8a2b1ccf, 0xa792b479, 0xf3f0f207, 0x4ea1e269, 0x65cdf4da, 0x06d5be05, 0xd11f6234, 0xc48afea6,
0x349d532e, 0xa2a055f3, 0x0532e18a, 0xa475ebf6, 0x0b39ec83, 0x40aaef60, 0x5e069f71, 0xbd51106e,
0x3ef98a21, 0x963d06dd, 0xddae053e, 0x4d46bde6, 0x91b58d54, 0x71055dc4, 0x046fd406, 0x60ff1550,
0x1924fb98, 0xd697e9bd, 0x89cc4340, 0x67779ed9, 0xb0bd42e8, 0x07888b89, 0xe7385b19, 0x79dbeec8,
0xa1470a7c, 0x7ce90f42, 0xf8c91e84, 0x00000000, 0x09838680, 0x3248ed2b, 0x1eac7011, 0x6c4e725a,
0xfdfbff0e, 0x0f563885, 0x3d1ed5ae, 0x3627392d, 0x0a64d90f, 0x6821a65c, 0x9bd1545b, 0x243a2e36,
0x0cb1670a, 0x930fe757, 0xb4d296ee, 0x1b9e919b, 0x804fc5c0, 0x61a220dc, 0x5a694b77, 0x1c161a12,
0xe20aba93, 0xc0e52aa0, 0x3c43e022, 0x121d171b, 0x0e0b0d09, 0xf2adc78b, 0x2db9a8b6, 0x14c8a91e,
0x578519f1, 0xaf4c0775, 0xeebbdd99, 0xa3fd607f, 0xf79f2601, 0x5cbcf572, 0x44c53b66, 0x5b347efb,
0x8b762943, 0xcbdcc623, 0xb668fced, 0xb863f1e4, 0xd7cadc31, 0x42108563, 0x13402297, 0x842011c6,
0x857d244a, 0xd2f83dbb, 0xae1132f9, 0xc76da129, 0x1d4b2f9e, 0xdcf330b2, 0x0dec5286, 0x77d0e3c1,
0x2b6c16b3, 0xa999b970, 0x11fa4894, 0x472264e9, 0xa8c48cfc, 0xa01a3ff0, 0x56d82c7d, 0x22ef9033,
0x87c74e49, 0xd9c1d138, 0x8cfea2ca, 0x98360bd4, 0xa6cf81f5, 0xa528de7a, 0xda268eb7, 0x3fa4bfad,
0x2ce49d3a, 0x500d9278, 0x6a9bcc5f, 0x5462467e, 0xf6c2138d, 0x90e8b8d8, 0x2e5ef739, 0x82f5afc3,
0x9fbe805d, 0x697c93d0, 0x6fa92dd5, 0xcfb31225, 0xc83b99ac, 0x10a77d18, 0xe86e639c, 0xdb7bbb3b,
0xcd097826, 0x6ef41859, 0xec01b79a, 0x83a89a4f, 0xe6656e95, 0xaa7ee6ff, 0x2108cfbc, 0xefe6e815,
0xbad99be7, 0x4ace366f, 0xead4099f, 0x29d67cb0, 0x31afb2a4, 0x2a31233f, 0xc63094a5, 0x35c066a2,
0x7437bc4e, 0xfca6ca82, 0xe0b0d090, 0x3315d8a7, 0xf14a9804, 0x41f7daec, 0x7f0e50cd, 0x172ff691,
0x768dd64d, 0x434db0ef, 0xcc544daa, 0xe4df0496, 0x9ee3b5d1, 0x4c1b886a, 0xc1b81f2c, 0x467f5165,
0x9d04ea5e, 0x015d358c, 0xfa737487, 0xfb2e410b, 0xb35a1d67, 0x9252d2db, 0xe9335610, 0x6d1347d6,
0x9a8c61d7, 0x377a0ca1, 0x598e14f8, 0xeb893c13, 0xceee27a9, 0xb735c961, 0xe1ede51c, 0x7a3cb147,
0x9c59dfd2, 0x553f73f2, 0x1879ce14, 0x73bf37c7, 0x53eacdf7, 0x5f5baafd, 0xdf146f3d, 0x7886db44,
0xca81f3af, 0xb93ec468, 0x382c3424, 0xc25f40a3, 0x1672c31d, 0xbc0c25e2, 0x288b493c, 0xff41950d,
0x397101a8, 0x08deb30c, 0xd89ce4b4, 0x6490c156, 0x7b6184cb, 0xd570b632, 0x48745c6c, 0xd04257b8,
};
rx_vec_i128 soft_aesenc(rx_vec_i128 in, rx_vec_i128 key) {
uint32_t s0, s1, s2, s3;
s0 = rx_vec_i128_w(in);
s1 = rx_vec_i128_z(in);
s2 = rx_vec_i128_y(in);
s3 = rx_vec_i128_x(in);
rx_vec_i128 out = rx_set_int_vec_i128(
(lutEnc0[s0 & 0xff] ^ lutEnc1[(s3 >> 8) & 0xff] ^ lutEnc2[(s2 >> 16) & 0xff] ^ lutEnc3[s1 >> 24]),
(lutEnc0[s1 & 0xff] ^ lutEnc1[(s0 >> 8) & 0xff] ^ lutEnc2[(s3 >> 16) & 0xff] ^ lutEnc3[s2 >> 24]),
(lutEnc0[s2 & 0xff] ^ lutEnc1[(s1 >> 8) & 0xff] ^ lutEnc2[(s0 >> 16) & 0xff] ^ lutEnc3[s3 >> 24]),
(lutEnc0[s3 & 0xff] ^ lutEnc1[(s2 >> 8) & 0xff] ^ lutEnc2[(s1 >> 16) & 0xff] ^ lutEnc3[s0 >> 24])
);
return rx_xor_vec_i128(out, key);
}
rx_vec_i128 soft_aesdec(rx_vec_i128 in, rx_vec_i128 key) {
uint32_t s0, s1, s2, s3;
s0 = rx_vec_i128_w(in);
s1 = rx_vec_i128_z(in);
s2 = rx_vec_i128_y(in);
s3 = rx_vec_i128_x(in);
rx_vec_i128 out = rx_set_int_vec_i128(
(lutDec0[s0 & 0xff] ^ lutDec1[(s1 >> 8) & 0xff] ^ lutDec2[(s2 >> 16) & 0xff] ^ lutDec3[s3 >> 24]),
(lutDec0[s1 & 0xff] ^ lutDec1[(s2 >> 8) & 0xff] ^ lutDec2[(s3 >> 16) & 0xff] ^ lutDec3[s0 >> 24]),
(lutDec0[s2 & 0xff] ^ lutDec1[(s3 >> 8) & 0xff] ^ lutDec2[(s0 >> 16) & 0xff] ^ lutDec3[s1 >> 24]),
(lutDec0[s3 & 0xff] ^ lutDec1[(s0 >> 8) & 0xff] ^ lutDec2[(s1 >> 16) & 0xff] ^ lutDec3[s2 >> 24])
);
return rx_xor_vec_i128(out, key);
}

View file

@ -0,0 +1,46 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <stdint.h>
#include "intrin_portable.h"
rx_vec_i128 soft_aesenc(rx_vec_i128 in, rx_vec_i128 key);
rx_vec_i128 soft_aesdec(rx_vec_i128 in, rx_vec_i128 key);
template<bool soft>
inline rx_vec_i128 aesenc(rx_vec_i128 in, rx_vec_i128 key) {
return soft ? soft_aesenc(in, key) : rx_aesenc_vec_i128(in, key);
}
template<bool soft>
inline rx_vec_i128 aesdec(rx_vec_i128 in, rx_vec_i128 key) {
return soft ? soft_aesdec(in, key) : rx_aesdec_vec_i128(in, key);
}

View file

@ -0,0 +1,891 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "configuration.h"
#include "program.hpp"
#include "blake2/endian.h"
#include "superscalar.hpp"
#include "intrin_portable.h"
#include "reciprocal.h"
namespace randomx {
static bool isMultiplication(SuperscalarInstructionType type) {
return type == SuperscalarInstructionType::IMUL_R || type == SuperscalarInstructionType::IMULH_R || type == SuperscalarInstructionType::ISMULH_R || type == SuperscalarInstructionType::IMUL_RCP;
}
//uOPs (micro-ops) are represented only by the execution port they can go to
namespace ExecutionPort {
using type = int;
constexpr type Null = 0;
constexpr type P0 = 1;
constexpr type P1 = 2;
constexpr type P5 = 4;
constexpr type P01 = P0 | P1;
constexpr type P05 = P0 | P5;
constexpr type P015 = P0 | P1 | P5;
}
//Macro-operation as output of the x86 decoder
//Usually one macro-op = one x86 instruction, but 2 instructions are sometimes fused into 1 macro-op
//Macro-op can consist of 1 or 2 uOPs.
class MacroOp {
public:
MacroOp(const char* name, int size)
: name_(name), size_(size), latency_(0), uop1_(ExecutionPort::Null), uop2_(ExecutionPort::Null) {}
MacroOp(const char* name, int size, int latency, ExecutionPort::type uop)
: name_(name), size_(size), latency_(latency), uop1_(uop), uop2_(ExecutionPort::Null) {}
MacroOp(const char* name, int size, int latency, ExecutionPort::type uop1, ExecutionPort::type uop2)
: name_(name), size_(size), latency_(latency), uop1_(uop1), uop2_(uop2) {}
MacroOp(const MacroOp& parent, bool dependent)
: name_(parent.name_), size_(parent.size_), latency_(parent.latency_), uop1_(parent.uop1_), uop2_(parent.uop2_), dependent_(dependent) {}
const char* getName() const {
return name_;
}
int getSize() const {
return size_;
}
int getLatency() const {
return latency_;
}
ExecutionPort::type getUop1() const {
return uop1_;
}
ExecutionPort::type getUop2() const {
return uop2_;
}
bool isSimple() const {
return uop2_ == ExecutionPort::Null;
}
bool isEliminated() const {
return uop1_ == ExecutionPort::Null;
}
bool isDependent() const {
return dependent_;
}
static const MacroOp Add_rr;
static const MacroOp Add_ri;
static const MacroOp Lea_sib;
static const MacroOp Sub_rr;
static const MacroOp Imul_rr;
static const MacroOp Imul_r;
static const MacroOp Mul_r;
static const MacroOp Mov_rr;
static const MacroOp Mov_ri64;
static const MacroOp Xor_rr;
static const MacroOp Xor_ri;
static const MacroOp Ror_rcl;
static const MacroOp Ror_ri;
static const MacroOp TestJz_fused;
static const MacroOp Xor_self;
static const MacroOp Cmp_ri;
static const MacroOp Setcc_r;
private:
const char* name_;
int size_;
int latency_;
ExecutionPort::type uop1_;
ExecutionPort::type uop2_;
bool dependent_ = false;
};
//Size: 3 bytes
const MacroOp MacroOp::Add_rr = MacroOp("add r,r", 3, 1, ExecutionPort::P015);
const MacroOp MacroOp::Sub_rr = MacroOp("sub r,r", 3, 1, ExecutionPort::P015);
const MacroOp MacroOp::Xor_rr = MacroOp("xor r,r", 3, 1, ExecutionPort::P015);
const MacroOp MacroOp::Imul_r = MacroOp("imul r", 3, 4, ExecutionPort::P1, ExecutionPort::P5);
const MacroOp MacroOp::Mul_r = MacroOp("mul r", 3, 4, ExecutionPort::P1, ExecutionPort::P5);
const MacroOp MacroOp::Mov_rr = MacroOp("mov r,r", 3);
//Size: 4 bytes
const MacroOp MacroOp::Lea_sib = MacroOp("lea r,r+r*s", 4, 1, ExecutionPort::P01);
const MacroOp MacroOp::Imul_rr = MacroOp("imul r,r", 4, 3, ExecutionPort::P1);
const MacroOp MacroOp::Ror_ri = MacroOp("ror r,i", 4, 1, ExecutionPort::P05);
//Size: 7 bytes (can be optionally padded with nop to 8 or 9 bytes)
const MacroOp MacroOp::Add_ri = MacroOp("add r,i", 7, 1, ExecutionPort::P015);
const MacroOp MacroOp::Xor_ri = MacroOp("xor r,i", 7, 1, ExecutionPort::P015);
//Size: 10 bytes
const MacroOp MacroOp::Mov_ri64 = MacroOp("mov rax,i64", 10, 1, ExecutionPort::P015);
//Unused:
const MacroOp MacroOp::Ror_rcl = MacroOp("ror r,cl", 3, 1, ExecutionPort::P0, ExecutionPort::P5);
const MacroOp MacroOp::Xor_self = MacroOp("xor rcx,rcx", 3);
const MacroOp MacroOp::Cmp_ri = MacroOp("cmp r,i", 7, 1, ExecutionPort::P015);
const MacroOp MacroOp::Setcc_r = MacroOp("setcc cl", 3, 1, ExecutionPort::P05);
const MacroOp MacroOp::TestJz_fused = MacroOp("testjz r,i", 13, 0, ExecutionPort::P5);
const MacroOp IMULH_R_ops_array[] = { MacroOp::Mov_rr, MacroOp::Mul_r, MacroOp::Mov_rr };
const MacroOp ISMULH_R_ops_array[] = { MacroOp::Mov_rr, MacroOp::Imul_r, MacroOp::Mov_rr };
const MacroOp IMUL_RCP_ops_array[] = { MacroOp::Mov_ri64, MacroOp(MacroOp::Imul_rr, true) };
class SuperscalarInstructionInfo {
public:
const char* getName() const {
return name_;
}
int getSize() const {
return ops_.size();
}
bool isSimple() const {
return getSize() == 1;
}
int getLatency() const {
return latency_;
}
const MacroOp& getOp(int index) const {
return ops_[index];
}
SuperscalarInstructionType getType() const {
return type_;
}
int getResultOp() const {
return resultOp_;
}
int getDstOp() const {
return dstOp_;
}
int getSrcOp() const {
return srcOp_;
}
static const SuperscalarInstructionInfo ISUB_R;
static const SuperscalarInstructionInfo IXOR_R;
static const SuperscalarInstructionInfo IADD_RS;
static const SuperscalarInstructionInfo IMUL_R;
static const SuperscalarInstructionInfo IROR_C;
static const SuperscalarInstructionInfo IADD_C7;
static const SuperscalarInstructionInfo IXOR_C7;
static const SuperscalarInstructionInfo IADD_C8;
static const SuperscalarInstructionInfo IXOR_C8;
static const SuperscalarInstructionInfo IADD_C9;
static const SuperscalarInstructionInfo IXOR_C9;
static const SuperscalarInstructionInfo IMULH_R;
static const SuperscalarInstructionInfo ISMULH_R;
static const SuperscalarInstructionInfo IMUL_RCP;
static const SuperscalarInstructionInfo NOP;
private:
const char* name_;
SuperscalarInstructionType type_;
std::vector<MacroOp> ops_;
int latency_;
int resultOp_ = 0;
int dstOp_ = 0;
int srcOp_;
SuperscalarInstructionInfo(const char* name)
: name_(name), type_(SuperscalarInstructionType::INVALID), latency_(0) {}
SuperscalarInstructionInfo(const char* name, SuperscalarInstructionType type, const MacroOp& op, int srcOp)
: name_(name), type_(type), latency_(op.getLatency()), srcOp_(srcOp) {
ops_.push_back(MacroOp(op));
}
template <size_t N>
SuperscalarInstructionInfo(const char* name, SuperscalarInstructionType type, const MacroOp(&arr)[N], int resultOp, int dstOp, int srcOp)
: name_(name), type_(type), latency_(0), resultOp_(resultOp), dstOp_(dstOp), srcOp_(srcOp) {
for (unsigned i = 0; i < N; ++i) {
ops_.push_back(MacroOp(arr[i]));
latency_ += ops_.back().getLatency();
}
static_assert(N > 1, "Invalid array size");
}
};
const SuperscalarInstructionInfo SuperscalarInstructionInfo::ISUB_R = SuperscalarInstructionInfo("ISUB_R", SuperscalarInstructionType::ISUB_R, MacroOp::Sub_rr, 0);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IXOR_R = SuperscalarInstructionInfo("IXOR_R", SuperscalarInstructionType::IXOR_R, MacroOp::Xor_rr, 0);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IADD_RS = SuperscalarInstructionInfo("IADD_RS", SuperscalarInstructionType::IADD_RS, MacroOp::Lea_sib, 0);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IMUL_R = SuperscalarInstructionInfo("IMUL_R", SuperscalarInstructionType::IMUL_R, MacroOp::Imul_rr, 0);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IROR_C = SuperscalarInstructionInfo("IROR_C", SuperscalarInstructionType::IROR_C, MacroOp::Ror_ri, -1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IADD_C7 = SuperscalarInstructionInfo("IADD_C7", SuperscalarInstructionType::IADD_C7, MacroOp::Add_ri, -1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IXOR_C7 = SuperscalarInstructionInfo("IXOR_C7", SuperscalarInstructionType::IXOR_C7, MacroOp::Xor_ri, -1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IADD_C8 = SuperscalarInstructionInfo("IADD_C8", SuperscalarInstructionType::IADD_C8, MacroOp::Add_ri, -1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IXOR_C8 = SuperscalarInstructionInfo("IXOR_C8", SuperscalarInstructionType::IXOR_C8, MacroOp::Xor_ri, -1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IADD_C9 = SuperscalarInstructionInfo("IADD_C9", SuperscalarInstructionType::IADD_C9, MacroOp::Add_ri, -1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IXOR_C9 = SuperscalarInstructionInfo("IXOR_C9", SuperscalarInstructionType::IXOR_C9, MacroOp::Xor_ri, -1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IMULH_R = SuperscalarInstructionInfo("IMULH_R", SuperscalarInstructionType::IMULH_R, IMULH_R_ops_array, 1, 0, 1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::ISMULH_R = SuperscalarInstructionInfo("ISMULH_R", SuperscalarInstructionType::ISMULH_R, ISMULH_R_ops_array, 1, 0, 1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::IMUL_RCP = SuperscalarInstructionInfo("IMUL_RCP", SuperscalarInstructionType::IMUL_RCP, IMUL_RCP_ops_array, 1, 1, -1);
const SuperscalarInstructionInfo SuperscalarInstructionInfo::NOP = SuperscalarInstructionInfo("NOP");
//these are some of the options how to split a 16-byte window into 3 or 4 x86 instructions.
//RandomX uses instructions with a native size of 3 (sub, xor, mul, mov), 4 (lea, mul), 7 (xor, add immediate) or 10 bytes (mov 64-bit immediate).
//Slots with sizes of 8 or 9 bytes need to be padded with a nop instruction.
const int buffer0[] = { 4, 8, 4 };
const int buffer1[] = { 7, 3, 3, 3 };
const int buffer2[] = { 3, 7, 3, 3 };
const int buffer3[] = { 4, 9, 3 };
const int buffer4[] = { 4, 4, 4, 4 };
const int buffer5[] = { 3, 3, 10 };
class DecoderBuffer {
public:
static const DecoderBuffer Default;
template <size_t N>
DecoderBuffer(const char* name, int index, const int(&arr)[N])
: name_(name), index_(index), counts_(arr), opsCount_(N) {}
const int* getCounts() const {
return counts_;
}
int getSize() const {
return opsCount_;
}
int getIndex() const {
return index_;
}
const char* getName() const {
return name_;
}
const DecoderBuffer* fetchNext(SuperscalarInstructionType instrType, int cycle, int mulCount, Blake2Generator& gen) const {
//If the current RandomX instruction is "IMULH", the next fetch configuration must be 3-3-10
//because the full 128-bit multiplication instruction is 3 bytes long and decodes to 2 uOPs on Intel CPUs.
//Intel CPUs can decode at most 4 uOPs per cycle, so this requires a 2-1-1 configuration for a total of 3 macro ops.
if (instrType == SuperscalarInstructionType::IMULH_R || instrType == SuperscalarInstructionType::ISMULH_R)
return &decodeBuffer3310;
//To make sure that the multiplication port is saturated, a 4-4-4-4 configuration is generated if the number of multiplications
//is lower than the number of cycles.
if (mulCount < cycle + 1)
return &decodeBuffer4444;
//If the current RandomX instruction is "IMUL_RCP", the next buffer must begin with a 4-byte slot for multiplication.
if(instrType == SuperscalarInstructionType::IMUL_RCP)
return (gen.getByte() & 1) ? &decodeBuffer484 : &decodeBuffer493;
//Default: select a random fetch configuration.
return fetchNextDefault(gen);
}
private:
const char* name_;
int index_;
const int* counts_;
int opsCount_;
DecoderBuffer() : index_(-1) {}
static const DecoderBuffer decodeBuffer484;
static const DecoderBuffer decodeBuffer7333;
static const DecoderBuffer decodeBuffer3733;
static const DecoderBuffer decodeBuffer493;
static const DecoderBuffer decodeBuffer4444;
static const DecoderBuffer decodeBuffer3310;
static const DecoderBuffer* decodeBuffers[4];
const DecoderBuffer* fetchNextDefault(Blake2Generator& gen) const {
return decodeBuffers[gen.getByte() & 3];
}
};
const DecoderBuffer DecoderBuffer::decodeBuffer484 = DecoderBuffer("4,8,4", 0, buffer0);
const DecoderBuffer DecoderBuffer::decodeBuffer7333 = DecoderBuffer("7,3,3,3", 1, buffer1);
const DecoderBuffer DecoderBuffer::decodeBuffer3733 = DecoderBuffer("3,7,3,3", 2, buffer2);
const DecoderBuffer DecoderBuffer::decodeBuffer493 = DecoderBuffer("4,9,3", 3, buffer3);
const DecoderBuffer DecoderBuffer::decodeBuffer4444 = DecoderBuffer("4,4,4,4", 4, buffer4);
const DecoderBuffer DecoderBuffer::decodeBuffer3310 = DecoderBuffer("3,3,10", 5, buffer5);
const DecoderBuffer* DecoderBuffer::decodeBuffers[4] = {
&DecoderBuffer::decodeBuffer484,
&DecoderBuffer::decodeBuffer7333,
&DecoderBuffer::decodeBuffer3733,
&DecoderBuffer::decodeBuffer493,
};
const DecoderBuffer DecoderBuffer::Default = DecoderBuffer();
const SuperscalarInstructionInfo* slot_3[] = { &SuperscalarInstructionInfo::ISUB_R, &SuperscalarInstructionInfo::IXOR_R };
const SuperscalarInstructionInfo* slot_3L[] = { &SuperscalarInstructionInfo::ISUB_R, &SuperscalarInstructionInfo::IXOR_R, &SuperscalarInstructionInfo::IMULH_R, &SuperscalarInstructionInfo::ISMULH_R };
const SuperscalarInstructionInfo* slot_4[] = { &SuperscalarInstructionInfo::IROR_C, &SuperscalarInstructionInfo::IADD_RS };
const SuperscalarInstructionInfo* slot_7[] = { &SuperscalarInstructionInfo::IXOR_C7, &SuperscalarInstructionInfo::IADD_C7 };
const SuperscalarInstructionInfo* slot_8[] = { &SuperscalarInstructionInfo::IXOR_C8, &SuperscalarInstructionInfo::IADD_C8 };
const SuperscalarInstructionInfo* slot_9[] = { &SuperscalarInstructionInfo::IXOR_C9, &SuperscalarInstructionInfo::IADD_C9 };
const SuperscalarInstructionInfo* slot_10 = &SuperscalarInstructionInfo::IMUL_RCP;
static bool selectRegister(std::vector<int>& availableRegisters, Blake2Generator& gen, int& reg) {
int index;
if (availableRegisters.size() == 0)
return false;
if (availableRegisters.size() > 1) {
index = gen.getInt32() % availableRegisters.size();
}
else {
index = 0;
}
reg = availableRegisters[index];
return true;
}
class RegisterInfo {
public:
RegisterInfo() : latency(0), lastOpGroup(SuperscalarInstructionType::INVALID), lastOpPar(-1), value(0) {}
int latency;
SuperscalarInstructionType lastOpGroup;
int lastOpPar;
int value;
};
//"SuperscalarInstruction" consists of one or more macro-ops
class SuperscalarInstruction {
public:
void toInstr(Instruction& instr) { //translate to a RandomX instruction format
instr.opcode = (int)getType();
instr.dst = dst_;
instr.src = src_ >= 0 ? src_ : dst_;
instr.setMod(mod_);
instr.setImm32(imm32_);
}
void createForSlot(Blake2Generator& gen, int slotSize, int fetchType, bool isLast, bool isFirst) {
switch (slotSize)
{
case 3:
//if this is the last slot, we can also select "IMULH" instructions
if (isLast) {
create(slot_3L[gen.getByte() & 3], gen);
}
else {
create(slot_3[gen.getByte() & 1], gen);
}
break;
case 4:
//if this is the 4-4-4-4 buffer, issue multiplications as the first 3 instructions
if (fetchType == 4 && !isLast) {
create(&SuperscalarInstructionInfo::IMUL_R, gen);
}
else {
create(slot_4[gen.getByte() & 1], gen);
}
break;
case 7:
create(slot_7[gen.getByte() & 1], gen);
break;
case 8:
create(slot_8[gen.getByte() & 1], gen);
break;
case 9:
create(slot_9[gen.getByte() & 1], gen);
break;
case 10:
create(slot_10, gen);
break;
default:
UNREACHABLE;
}
}
void create(const SuperscalarInstructionInfo* info, Blake2Generator& gen) {
info_ = info;
reset();
switch (info->getType())
{
case SuperscalarInstructionType::ISUB_R: {
mod_ = 0;
imm32_ = 0;
opGroup_ = SuperscalarInstructionType::IADD_RS;
groupParIsSource_ = true;
} break;
case SuperscalarInstructionType::IXOR_R: {
mod_ = 0;
imm32_ = 0;
opGroup_ = SuperscalarInstructionType::IXOR_R;
groupParIsSource_ = true;
} break;
case SuperscalarInstructionType::IADD_RS: {
mod_ = gen.getByte();
imm32_ = 0;
opGroup_ = SuperscalarInstructionType::IADD_RS;
groupParIsSource_ = true;
} break;
case SuperscalarInstructionType::IMUL_R: {
mod_ = 0;
imm32_ = 0;
opGroup_ = SuperscalarInstructionType::IMUL_R;
groupParIsSource_ = true;
} break;
case SuperscalarInstructionType::IROR_C: {
mod_ = 0;
do {
imm32_ = gen.getByte() & 63;
} while (imm32_ == 0);
opGroup_ = SuperscalarInstructionType::IROR_C;
opGroupPar_ = -1;
} break;
case SuperscalarInstructionType::IADD_C7:
case SuperscalarInstructionType::IADD_C8:
case SuperscalarInstructionType::IADD_C9: {
mod_ = 0;
imm32_ = gen.getInt32();
opGroup_ = SuperscalarInstructionType::IADD_C7;
opGroupPar_ = -1;
} break;
case SuperscalarInstructionType::IXOR_C7:
case SuperscalarInstructionType::IXOR_C8:
case SuperscalarInstructionType::IXOR_C9: {
mod_ = 0;
imm32_ = gen.getInt32();
opGroup_ = SuperscalarInstructionType::IXOR_C7;
opGroupPar_ = -1;
} break;
case SuperscalarInstructionType::IMULH_R: {
canReuse_ = true;
mod_ = 0;
imm32_ = 0;
opGroup_ = SuperscalarInstructionType::IMULH_R;
opGroupPar_ = gen.getInt32();
} break;
case SuperscalarInstructionType::ISMULH_R: {
canReuse_ = true;
mod_ = 0;
imm32_ = 0;
opGroup_ = SuperscalarInstructionType::ISMULH_R;
opGroupPar_ = gen.getInt32();
} break;
case SuperscalarInstructionType::IMUL_RCP: {
mod_ = 0;
do {
imm32_ = gen.getInt32();
} while ((imm32_ & (imm32_ - 1)) == 0);
opGroup_ = SuperscalarInstructionType::IMUL_RCP;
opGroupPar_ = -1;
} break;
default:
break;
}
}
bool selectDestination(int cycle, bool allowChainedMul, RegisterInfo (&registers)[8], Blake2Generator& gen) {
/*if (allowChainedMultiplication && opGroup_ == SuperscalarInstructionType::IMUL_R)
std::cout << "Selecting destination with chained MUL enabled" << std::endl;*/
std::vector<int> availableRegisters;
//Conditions for the destination register:
// * value must be ready at the required cycle
// * cannot be the same as the source register unless the instruction allows it
// - this avoids optimizable instructions such as "xor r, r" or "sub r, r"
// * register cannot be multiplied twice in a row unless allowChainedMul is true
// - this avoids accumulation of trailing zeroes in registers due to excessive multiplication
// - allowChainedMul is set to true if an attempt to find source/destination registers failed (this is quite rare, but prevents a catastrophic failure of the generator)
// * either the last instruction applied to the register or its source must be different than this instruction
// - this avoids optimizable instruction sequences such as "xor r1, r2; xor r1, r2" or "ror r, C1; ror r, C2" or "add r, C1; add r, C2"
// * register r5 cannot be the destination of the IADD_RS instruction (limitation of the x86 lea instruction)
for (unsigned i = 0; i < 8; ++i) {
if (registers[i].latency <= cycle && (canReuse_ || i != src_) && (allowChainedMul || opGroup_ != SuperscalarInstructionType::IMUL_R || registers[i].lastOpGroup != SuperscalarInstructionType::IMUL_R) && (registers[i].lastOpGroup != opGroup_ || registers[i].lastOpPar != opGroupPar_) && (info_->getType() != SuperscalarInstructionType::IADD_RS || i != RegisterNeedsDisplacement))
availableRegisters.push_back(i);
}
return selectRegister(availableRegisters, gen, dst_);
}
bool selectSource(int cycle, RegisterInfo(&registers)[8], Blake2Generator& gen) {
std::vector<int> availableRegisters;
//all registers that are ready at the cycle
for (unsigned i = 0; i < 8; ++i) {
if (registers[i].latency <= cycle)
availableRegisters.push_back(i);
}
//if there are only 2 available registers for IADD_RS and one of them is r5, select it as the source because it cannot be the destination
if (availableRegisters.size() == 2 && info_->getType() == SuperscalarInstructionType::IADD_RS) {
if (availableRegisters[0] == RegisterNeedsDisplacement || availableRegisters[1] == RegisterNeedsDisplacement) {
opGroupPar_ = src_ = RegisterNeedsDisplacement;
return true;
}
}
if (selectRegister(availableRegisters, gen, src_)) {
if (groupParIsSource_)
opGroupPar_ = src_;
return true;
}
return false;
}
SuperscalarInstructionType getType() {
return info_->getType();
}
int getSource() {
return src_;
}
int getDestination() {
return dst_;
}
SuperscalarInstructionType getGroup() {
return opGroup_;
}
int getGroupPar() {
return opGroupPar_;
}
const SuperscalarInstructionInfo& getInfo() const {
return *info_;
}
static const SuperscalarInstruction Null;
private:
const SuperscalarInstructionInfo* info_;
int src_ = -1;
int dst_ = -1;
int mod_;
uint32_t imm32_;
SuperscalarInstructionType opGroup_;
int opGroupPar_;
bool canReuse_ = false;
bool groupParIsSource_ = false;
void reset() {
src_ = dst_ = -1;
canReuse_ = groupParIsSource_ = false;
}
SuperscalarInstruction(const SuperscalarInstructionInfo* info) : info_(info) {
}
};
const SuperscalarInstruction SuperscalarInstruction::Null = SuperscalarInstruction(&SuperscalarInstructionInfo::NOP);
constexpr int CYCLE_MAP_SIZE = RANDOMX_SUPERSCALAR_MAX_LATENCY + 4;
constexpr int LOOK_FORWARD_CYCLES = 4;
constexpr int MAX_THROWAWAY_COUNT = 256;
template<bool commit>
static int scheduleUop(ExecutionPort::type uop, ExecutionPort::type(&portBusy)[CYCLE_MAP_SIZE][3], int cycle) {
//The scheduling here is done optimistically by checking port availability in order P5 -> P0 -> P1 to not overload
//port P1 (multiplication) by instructions that can go to any port.
for (; cycle < RandomX_CurrentConfig.SuperscalarLatency + 4; ++cycle) {
if ((uop & ExecutionPort::P5) != 0 && !portBusy[cycle][2]) {
if (commit) {
if (trace) std::cout << "; P5 at cycle " << cycle << std::endl;
portBusy[cycle][2] = uop;
}
return cycle;
}
if ((uop & ExecutionPort::P0) != 0 && !portBusy[cycle][0]) {
if (commit) {
if (trace) std::cout << "; P0 at cycle " << cycle << std::endl;
portBusy[cycle][0] = uop;
}
return cycle;
}
if ((uop & ExecutionPort::P1) != 0 && !portBusy[cycle][1]) {
if (commit) {
if (trace) std::cout << "; P1 at cycle " << cycle << std::endl;
portBusy[cycle][1] = uop;
}
return cycle;
}
}
return -1;
}
template<bool commit>
static int scheduleMop(const MacroOp& mop, ExecutionPort::type(&portBusy)[CYCLE_MAP_SIZE][3], int cycle, int depCycle) {
//if this macro-op depends on the previous one, increase the starting cycle if needed
//this handles an explicit dependency chain in IMUL_RCP
if (mop.isDependent()) {
cycle = (cycle > depCycle) ? cycle : depCycle;
}
//move instructions are eliminated and don't need an execution unit
if (mop.isEliminated()) {
if (commit)
if (trace) std::cout << "; (eliminated)" << std::endl;
return cycle;
}
else if (mop.isSimple()) {
//this macro-op has only one uOP
return scheduleUop<commit>(mop.getUop1(), portBusy, cycle);
}
else {
//macro-ops with 2 uOPs are scheduled conservatively by requiring both uOPs to execute in the same cycle
for (; cycle < RandomX_CurrentConfig.SuperscalarLatency + 4; ++cycle) {
int cycle1 = scheduleUop<false>(mop.getUop1(), portBusy, cycle);
int cycle2 = scheduleUop<false>(mop.getUop2(), portBusy, cycle);
if (cycle1 == cycle2) {
if (commit) {
scheduleUop<true>(mop.getUop1(), portBusy, cycle1);
scheduleUop<true>(mop.getUop2(), portBusy, cycle2);
}
return cycle1;
}
}
}
return -1;
}
void generateSuperscalar(SuperscalarProgram& prog, Blake2Generator& gen) {
ExecutionPort::type portBusy[CYCLE_MAP_SIZE][3];
memset(portBusy, 0, sizeof(portBusy));
RegisterInfo registers[8];
const DecoderBuffer* decodeBuffer = &DecoderBuffer::Default;
SuperscalarInstruction currentInstruction = SuperscalarInstruction::Null;
int macroOpIndex = 0;
int codeSize = 0;
int macroOpCount = 0;
int cycle = 0;
int depCycle = 0;
int retireCycle = 0;
bool portsSaturated = false;
int programSize = 0;
int mulCount = 0;
int decodeCycle;
int throwAwayCount = 0;
//decode instructions for RANDOMX_SUPERSCALAR_LATENCY cycles or until an execution port is saturated.
//Each decode cycle decodes 16 bytes of x86 code.
//Since a decode cycle produces on average 3.45 macro-ops and there are only 3 ALU ports, execution ports are always
//saturated first. The cycle limit is present only to guarantee loop termination.
//Program size is limited to SuperscalarMaxSize instructions.
for (decodeCycle = 0; decodeCycle < RandomX_CurrentConfig.SuperscalarLatency && !portsSaturated && programSize < 3 * RandomX_CurrentConfig.SuperscalarLatency + 2; ++decodeCycle) {
//select a decode configuration
decodeBuffer = decodeBuffer->fetchNext(currentInstruction.getType(), decodeCycle, mulCount, gen);
if (trace) std::cout << "; ------------- fetch cycle " << cycle << " (" << decodeBuffer->getName() << ")" << std::endl;
int bufferIndex = 0;
//fill all instruction slots in the current decode buffer
while (bufferIndex < decodeBuffer->getSize()) {
int topCycle = cycle;
//if we have issued all macro-ops for the current RandomX instruction, create a new instruction
if (macroOpIndex >= currentInstruction.getInfo().getSize()) {
if (portsSaturated || programSize >= 3 * RandomX_CurrentConfig.SuperscalarLatency + 2)
break;
//select an instruction so that the first macro-op fits into the current slot
currentInstruction.createForSlot(gen, decodeBuffer->getCounts()[bufferIndex], decodeBuffer->getIndex(), decodeBuffer->getSize() == bufferIndex + 1, bufferIndex == 0);
macroOpIndex = 0;
if (trace) std::cout << "; " << currentInstruction.getInfo().getName() << std::endl;
}
const MacroOp& mop = currentInstruction.getInfo().getOp(macroOpIndex);
if (trace) std::cout << mop.getName() << " ";
//calculate the earliest cycle when this macro-op (all of its uOPs) can be scheduled for execution
int scheduleCycle = scheduleMop<false>(mop, portBusy, cycle, depCycle);
if (scheduleCycle < 0) {
if (trace) std::cout << "Unable to map operation '" << mop.getName() << "' to execution port (cycle " << cycle << ")" << std::endl;
//__debugbreak();
portsSaturated = true;
break;
}
//find a source register (if applicable) that will be ready when this instruction executes
if (macroOpIndex == currentInstruction.getInfo().getSrcOp()) {
int forward;
//if no suitable operand is ready, look up to LOOK_FORWARD_CYCLES forward
for (forward = 0; forward < LOOK_FORWARD_CYCLES && !currentInstruction.selectSource(scheduleCycle, registers, gen); ++forward) {
if (trace) std::cout << "; src STALL at cycle " << cycle << std::endl;
++scheduleCycle;
++cycle;
}
//if no register was found, throw the instruction away and try another one
if (forward == LOOK_FORWARD_CYCLES) {
if (throwAwayCount < MAX_THROWAWAY_COUNT) {
throwAwayCount++;
macroOpIndex = currentInstruction.getInfo().getSize();
if (trace) std::cout << "; THROW away " << currentInstruction.getInfo().getName() << std::endl;
//cycle = topCycle;
continue;
}
//abort this decode buffer
if (trace) std::cout << "Aborting at cycle " << cycle << " with decode buffer " << decodeBuffer->getName() << " - source registers not available for operation " << currentInstruction.getInfo().getName() << std::endl;
currentInstruction = SuperscalarInstruction::Null;
break;
}
if (trace) std::cout << "; src = r" << currentInstruction.getSource() << std::endl;
}
//find a destination register that will be ready when this instruction executes
if (macroOpIndex == currentInstruction.getInfo().getDstOp()) {
int forward;
for (forward = 0; forward < LOOK_FORWARD_CYCLES && !currentInstruction.selectDestination(scheduleCycle, throwAwayCount > 0, registers, gen); ++forward) {
if (trace) std::cout << "; dst STALL at cycle " << cycle << std::endl;
++scheduleCycle;
++cycle;
}
if (forward == LOOK_FORWARD_CYCLES) { //throw instruction away
if (throwAwayCount < MAX_THROWAWAY_COUNT) {
throwAwayCount++;
macroOpIndex = currentInstruction.getInfo().getSize();
if (trace) std::cout << "; THROW away " << currentInstruction.getInfo().getName() << std::endl;
//cycle = topCycle;
continue;
}
//abort this decode buffer
if (trace) std::cout << "Aborting at cycle " << cycle << " with decode buffer " << decodeBuffer->getName() << " - destination registers not available" << std::endl;
currentInstruction = SuperscalarInstruction::Null;
break;
}
if (trace) std::cout << "; dst = r" << currentInstruction.getDestination() << std::endl;
}
throwAwayCount = 0;
//recalculate when the instruction can be scheduled for execution based on operand availability
scheduleCycle = scheduleMop<true>(mop, portBusy, scheduleCycle, scheduleCycle);
//calculate when the result will be ready
depCycle = scheduleCycle + mop.getLatency();
//if this instruction writes the result, modify register information
// RegisterInfo.latency - which cycle the register will be ready
// RegisterInfo.lastOpGroup - the last operation that was applied to the register
// RegisterInfo.lastOpPar - the last operation source value (-1 = constant, 0-7 = register)
if (macroOpIndex == currentInstruction.getInfo().getResultOp()) {
int dst = currentInstruction.getDestination();
RegisterInfo& ri = registers[dst];
retireCycle = depCycle;
ri.latency = retireCycle;
ri.lastOpGroup = currentInstruction.getGroup();
ri.lastOpPar = currentInstruction.getGroupPar();
if (trace) std::cout << "; RETIRED at cycle " << retireCycle << std::endl;
}
codeSize += mop.getSize();
bufferIndex++;
macroOpIndex++;
macroOpCount++;
//terminating condition
if (scheduleCycle >= RandomX_CurrentConfig.SuperscalarLatency) {
portsSaturated = true;
}
cycle = topCycle;
//when all macro-ops of the current instruction have been issued, add the instruction into the program
if (macroOpIndex >= currentInstruction.getInfo().getSize()) {
currentInstruction.toInstr(prog(programSize++));
mulCount += isMultiplication(currentInstruction.getType());
}
}
++cycle;
}
double ipc = (macroOpCount / (double)retireCycle);
memset(prog.asicLatencies, 0, sizeof(prog.asicLatencies));
//Calculate ASIC latency:
//Assumes 1 cycle latency for all operations and unlimited parallelization.
for (int i = 0; i < programSize; ++i) {
Instruction& instr = prog(i);
int latDst = prog.asicLatencies[instr.dst] + 1;
int latSrc = instr.dst != instr.src ? prog.asicLatencies[instr.src] + 1 : 0;
prog.asicLatencies[instr.dst] = (latDst > latSrc) ? latDst : latSrc;
}
//address register is the register with the highest ASIC latency
int asicLatencyMax = 0;
int addressReg = 0;
for (int i = 0; i < 8; ++i) {
if (prog.asicLatencies[i] > asicLatencyMax) {
asicLatencyMax = prog.asicLatencies[i];
addressReg = i;
}
prog.cpuLatencies[i] = registers[i].latency;
}
prog.setSize(programSize);
prog.setAddressRegister(addressReg);
prog.cpuLatency = retireCycle;
prog.asicLatency = asicLatencyMax;
prog.codeSize = codeSize;
prog.macroOps = macroOpCount;
prog.decodeCycles = decodeCycle;
prog.ipc = ipc;
prog.mulCount = mulCount;
/*if(INFO) std::cout << "; ALU port utilization:" << std::endl;
if (INFO) std::cout << "; (* = in use, _ = idle)" << std::endl;
int portCycles = 0;
for (int i = 0; i < RandomX_Config.SuperscalarLatency + 4; ++i) {
std::cout << "; " << std::setw(3) << i << " ";
for (int j = 0; j < 3; ++j) {
std::cout << (portBusy[i][j] ? '*' : '_');
portCycles += !!portBusy[i][j];
}
std::cout << std::endl;
}*/
}
void executeSuperscalar(int_reg_t(&r)[8], SuperscalarProgram& prog, std::vector<uint64_t> *reciprocals) {
for (unsigned j = 0; j < prog.getSize(); ++j) {
Instruction& instr = prog(j);
switch ((SuperscalarInstructionType)instr.opcode)
{
case SuperscalarInstructionType::ISUB_R:
r[instr.dst] -= r[instr.src];
break;
case SuperscalarInstructionType::IXOR_R:
r[instr.dst] ^= r[instr.src];
break;
case SuperscalarInstructionType::IADD_RS:
r[instr.dst] += r[instr.src] << instr.getModShift();
break;
case SuperscalarInstructionType::IMUL_R:
r[instr.dst] *= r[instr.src];
break;
case SuperscalarInstructionType::IROR_C:
r[instr.dst] = rotr64(r[instr.dst], instr.getImm32());
break;
case SuperscalarInstructionType::IADD_C7:
case SuperscalarInstructionType::IADD_C8:
case SuperscalarInstructionType::IADD_C9:
r[instr.dst] += signExtend2sCompl(instr.getImm32());
break;
case SuperscalarInstructionType::IXOR_C7:
case SuperscalarInstructionType::IXOR_C8:
case SuperscalarInstructionType::IXOR_C9:
r[instr.dst] ^= signExtend2sCompl(instr.getImm32());
break;
case SuperscalarInstructionType::IMULH_R:
r[instr.dst] = mulh(r[instr.dst], r[instr.src]);
break;
case SuperscalarInstructionType::ISMULH_R:
r[instr.dst] = smulh(r[instr.dst], r[instr.src]);
break;
case SuperscalarInstructionType::IMUL_RCP:
if (reciprocals != nullptr)
r[instr.dst] *= (*reciprocals)[instr.getImm32()];
else
r[instr.dst] *= randomx_reciprocal(instr.getImm32());
break;
default:
UNREACHABLE;
}
}
}
}

View file

@ -0,0 +1,60 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include <vector>
#include "superscalar_program.hpp"
#include "blake2_generator.hpp"
namespace randomx {
// Intel Ivy Bridge reference
enum class SuperscalarInstructionType { //uOPs (decode) execution ports latency code size
ISUB_R = 0, //1 p015 1 3 (sub)
IXOR_R = 1, //1 p015 1 3 (xor)
IADD_RS = 2, //1 p01 1 4 (lea)
IMUL_R = 3, //1 p1 3 4 (imul)
IROR_C = 4, //1 p05 1 4 (ror)
IADD_C7 = 5, //1 p015 1 7 (add)
IXOR_C7 = 6, //1 p015 1 7 (xor)
IADD_C8 = 7, //1+0 p015 1 7+1 (add+nop)
IXOR_C8 = 8, //1+0 p015 1 7+1 (xor+nop)
IADD_C9 = 9, //1+0 p015 1 7+2 (add+nop)
IXOR_C9 = 10, //1+0 p015 1 7+2 (xor+nop)
IMULH_R = 11, //1+2+1 0+(p1,p5)+0 3 3+3+3 (mov+mul+mov)
ISMULH_R = 12, //1+2+1 0+(p1,p5)+0 3 3+3+3 (mov+imul+mov)
IMUL_RCP = 13, //1+1 p015+p1 4 10+4 (mov+imul)
COUNT = 14,
INVALID = -1
};
void generateSuperscalar(SuperscalarProgram& prog, Blake2Generator& gen);
void executeSuperscalar(uint64_t(&r)[8], SuperscalarProgram& prog, std::vector<uint64_t> *reciprocals = nullptr);
}

View file

@ -0,0 +1,73 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include "instruction.hpp"
#include "common.hpp"
namespace randomx {
class SuperscalarProgram {
public:
Instruction& operator()(int pc) {
return programBuffer[pc];
}
uint32_t getSize() {
return size;
}
void setSize(uint32_t val) {
size = val;
}
int getAddressRegister() {
return addrReg;
}
void setAddressRegister(int val) {
addrReg = val;
}
Instruction programBuffer[SuperscalarMaxSize];
uint32_t size
#ifndef NDEBUG
= 0
#endif
;
int addrReg;
double ipc;
int codeSize;
int macroOps;
int decodeCycles;
int cpuLatency;
int asicLatency;
int mulCount;
int cpuLatencies[8];
int asicLatencies[8];
};
}

View file

@ -0,0 +1,137 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <cstring>
#include <iomanip>
#include <stdexcept>
#include "virtual_machine.hpp"
#include "common.hpp"
#include "aes_hash.hpp"
#include "blake2/blake2.h"
#include "intrin_portable.h"
#include "allocator.hpp"
randomx_vm::~randomx_vm() {
}
void randomx_vm::resetRoundingMode() {
rx_reset_float_state();
}
namespace randomx {
static inline uint64_t getSmallPositiveFloatBits(uint64_t entropy) {
auto exponent = entropy >> 59; //0..31
auto mantissa = entropy & mantissaMask;
exponent += exponentBias;
exponent &= exponentMask;
exponent <<= mantissaSize;
return exponent | mantissa;
}
static inline uint64_t getStaticExponent(uint64_t entropy) {
auto exponent = constExponentBits;
exponent |= (entropy >> (64 - staticExponentBits)) << dynamicExponentBits;
exponent <<= mantissaSize;
return exponent;
}
static inline uint64_t getFloatMask(uint64_t entropy) {
constexpr uint64_t mask22bit = (1ULL << 22) - 1;
return (entropy & mask22bit) | getStaticExponent(entropy);
}
}
void randomx_vm::initialize() {
store64(&reg.a[0].lo, randomx::getSmallPositiveFloatBits(program.getEntropy(0)));
store64(&reg.a[0].hi, randomx::getSmallPositiveFloatBits(program.getEntropy(1)));
store64(&reg.a[1].lo, randomx::getSmallPositiveFloatBits(program.getEntropy(2)));
store64(&reg.a[1].hi, randomx::getSmallPositiveFloatBits(program.getEntropy(3)));
store64(&reg.a[2].lo, randomx::getSmallPositiveFloatBits(program.getEntropy(4)));
store64(&reg.a[2].hi, randomx::getSmallPositiveFloatBits(program.getEntropy(5)));
store64(&reg.a[3].lo, randomx::getSmallPositiveFloatBits(program.getEntropy(6)));
store64(&reg.a[3].hi, randomx::getSmallPositiveFloatBits(program.getEntropy(7)));
mem.ma = program.getEntropy(8) & CacheLineAlignMask;
mem.mx = program.getEntropy(10);
auto addressRegisters = program.getEntropy(12);
config.readReg0 = 0 + (addressRegisters & 1);
addressRegisters >>= 1;
config.readReg1 = 2 + (addressRegisters & 1);
addressRegisters >>= 1;
config.readReg2 = 4 + (addressRegisters & 1);
addressRegisters >>= 1;
config.readReg3 = 6 + (addressRegisters & 1);
datasetOffset = (program.getEntropy(13) % (DatasetExtraItems + 1)) * randomx::CacheLineSize;
store64(&config.eMask[0], randomx::getFloatMask(program.getEntropy(14)));
store64(&config.eMask[1], randomx::getFloatMask(program.getEntropy(15)));
}
namespace randomx {
alignas(16) volatile static rx_vec_i128 aesDummy;
template<class Allocator, bool softAes>
VmBase<Allocator, softAes>::~VmBase() {
Allocator::freeMemory(scratchpad, RANDOMX_SCRATCHPAD_L3_MAX_SIZE);
}
template<class Allocator, bool softAes>
void VmBase<Allocator, softAes>::allocate() {
if (datasetPtr == nullptr)
throw std::invalid_argument("Cache/Dataset not set");
if (!softAes) { //if hardware AES is not supported, it's better to fail now than to return a ticking bomb
rx_vec_i128 tmp = rx_load_vec_i128((const rx_vec_i128*)&aesDummy);
tmp = rx_aesenc_vec_i128(tmp, tmp);
rx_store_vec_i128((rx_vec_i128*)&aesDummy, tmp);
}
scratchpad = (uint8_t*)Allocator::allocMemory(RANDOMX_SCRATCHPAD_L3_MAX_SIZE);
}
template<class Allocator, bool softAes>
void VmBase<Allocator, softAes>::getFinalResult(void* out, size_t outSize) {
hashAes1Rx4<softAes>(scratchpad, ScratchpadSize, &reg.a);
blake2b(out, outSize, &reg, sizeof(RegisterFile), nullptr, 0);
}
template<class Allocator, bool softAes>
void VmBase<Allocator, softAes>::initScratchpad(void* seed) {
fillAes1Rx4<softAes>(seed, ScratchpadSize, scratchpad);
}
template<class Allocator, bool softAes>
void VmBase<Allocator, softAes>::generateProgram(void* seed) {
fillAes4Rx4<softAes>(seed, sizeof(program), &program);
}
template class VmBase<AlignedAllocator<CacheLineSize>, false>;
template class VmBase<AlignedAllocator<CacheLineSize>, true>;
template class VmBase<LargePageAllocator, false>;
template class VmBase<LargePageAllocator, true>;
}

View file

@ -0,0 +1,83 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstdint>
#include "common.hpp"
#include "program.hpp"
/* Global namespace for C binding */
class randomx_vm {
public:
virtual ~randomx_vm() = 0;
virtual void allocate() = 0;
virtual void getFinalResult(void* out, size_t outSize) = 0;
virtual void setDataset(randomx_dataset* dataset) { }
virtual void setCache(randomx_cache* cache) { }
virtual void initScratchpad(void* seed) = 0;
virtual void run(void* seed) = 0;
void resetRoundingMode();
randomx::RegisterFile *getRegisterFile() {
return &reg;
}
const void* getScratchpad() {
return scratchpad;
}
const randomx::Program& getProgram()
{
return program;
}
protected:
void initialize();
alignas(64) randomx::Program program;
alignas(64) randomx::RegisterFile reg;
alignas(16) randomx::ProgramConfiguration config;
randomx::MemoryRegisters mem;
uint8_t* scratchpad;
union {
randomx_cache* cachePtr = nullptr;
randomx_dataset* datasetPtr;
};
uint64_t datasetOffset;
};
namespace randomx {
template<class Allocator, bool softAes>
class VmBase : public randomx_vm {
public:
~VmBase() override;
void allocate() override;
void initScratchpad(void* seed) override;
void getFinalResult(void* out, size_t outSize) override;
protected:
void generateProgram(void* seed);
};
}

View file

@ -0,0 +1,105 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "virtual_memory.hpp"
#include <stdexcept>
#if defined(_WIN32) || defined(__CYGWIN__)
#include <windows.h>
#else
#ifdef __APPLE__
#include <mach/vm_statistics.h>
#endif
#include <sys/types.h>
#include <sys/mman.h>
#ifndef MAP_ANONYMOUS
#define MAP_ANONYMOUS MAP_ANON
#endif
#endif
#if defined(_WIN32) || defined(__CYGWIN__)
std::string getErrorMessage(const char* function) {
LPSTR messageBuffer = nullptr;
size_t size = FormatMessageA(FORMAT_MESSAGE_ALLOCATE_BUFFER | FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_IGNORE_INSERTS,
NULL, GetLastError(), MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), (LPSTR)&messageBuffer, 0, NULL);
std::string message(messageBuffer, size);
LocalFree(messageBuffer);
return std::string(function) + std::string(": ") + message;
}
#endif
void* allocExecutableMemory(std::size_t bytes) {
void* mem;
#if defined(_WIN32) || defined(__CYGWIN__)
mem = VirtualAlloc(nullptr, bytes, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
if (mem == nullptr)
throw std::runtime_error(getErrorMessage("allocExecutableMemory - VirtualAlloc"));
#else
mem = mmap(nullptr, bytes, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
if (mem == MAP_FAILED)
throw std::runtime_error("allocExecutableMemory - mmap failed");
#endif
return mem;
}
constexpr std::size_t align(std::size_t pos, std::size_t align) {
return ((pos - 1) / align + 1) * align;
}
void* allocLargePagesMemory(std::size_t bytes) {
void* mem;
#if defined(_WIN32) || defined(__CYGWIN__)
auto pageMinimum = GetLargePageMinimum();
if (pageMinimum > 0)
mem = VirtualAlloc(NULL, align(bytes, pageMinimum), MEM_COMMIT | MEM_RESERVE | MEM_LARGE_PAGES, PAGE_READWRITE);
else
throw std::runtime_error("allocLargePagesMemory - Large pages are not supported");
if (mem == nullptr)
throw std::runtime_error(getErrorMessage("allocLargePagesMemory - VirtualAlloc"));
#else
#ifdef __APPLE__
mem = mmap(nullptr, bytes, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, VM_FLAGS_SUPERPAGE_SIZE_2MB, 0);
#elif defined(__FreeBSD__)
mem = mmap(nullptr, bytes, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS | MAP_ALIGNED_SUPER, -1, 0);
#else
mem = mmap(nullptr, bytes, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB | MAP_POPULATE, -1, 0);
#endif
if (mem == MAP_FAILED)
throw std::runtime_error("allocLargePagesMemory - mmap failed");
#endif
return mem;
}
void freePagedMemory(void* ptr, std::size_t bytes) {
#if defined(_WIN32) || defined(__CYGWIN__)
VirtualFree(ptr, 0, MEM_RELEASE);
#else
munmap(ptr, bytes);
#endif
}

View file

@ -0,0 +1,35 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <cstddef>
void* allocExecutableMemory(std::size_t);
void* allocLargePagesMemory(std::size_t);
void freePagedMemory(void*, std::size_t);

View file

@ -0,0 +1,60 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "vm_compiled.hpp"
#include "common.hpp"
namespace randomx {
static_assert(sizeof(MemoryRegisters) == 2 * sizeof(addr_t) + sizeof(uintptr_t), "Invalid alignment of struct randomx::MemoryRegisters");
static_assert(sizeof(RegisterFile) == 256, "Invalid alignment of struct randomx::RegisterFile");
template<class Allocator, bool softAes>
void CompiledVm<Allocator, softAes>::setDataset(randomx_dataset* dataset) {
datasetPtr = dataset;
}
template<class Allocator, bool softAes>
void CompiledVm<Allocator, softAes>::run(void* seed) {
VmBase<Allocator, softAes>::generateProgram(seed);
randomx_vm::initialize();
compiler.generateProgram(program, config);
mem.memory = datasetPtr->memory + datasetOffset;
execute();
}
template<class Allocator, bool softAes>
void CompiledVm<Allocator, softAes>::execute() {
compiler.getProgramFunc()(reg, mem, scratchpad, RandomX_CurrentConfig.ProgramIterations);
}
template class CompiledVm<AlignedAllocator<CacheLineSize>, false>;
template class CompiledVm<AlignedAllocator<CacheLineSize>, true>;
template class CompiledVm<LargePageAllocator, false>;
template class CompiledVm<LargePageAllocator, true>;
}

View file

@ -0,0 +1,72 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <new>
#include <cstdint>
#include "virtual_machine.hpp"
#include "jit_compiler.hpp"
#include "allocator.hpp"
#include "dataset.hpp"
namespace randomx {
template<class Allocator, bool softAes>
class CompiledVm : public VmBase<Allocator, softAes> {
public:
void* operator new(size_t size) {
void* ptr = AlignedAllocator<CacheLineSize>::allocMemory(size);
if (ptr == nullptr)
throw std::bad_alloc();
return ptr;
}
void operator delete(void* ptr) {
AlignedAllocator<CacheLineSize>::freeMemory(ptr, sizeof(CompiledVm));
}
void setDataset(randomx_dataset* dataset) override;
void run(void* seed) override;
using VmBase<Allocator, softAes>::mem;
using VmBase<Allocator, softAes>::program;
using VmBase<Allocator, softAes>::config;
using VmBase<Allocator, softAes>::reg;
using VmBase<Allocator, softAes>::scratchpad;
using VmBase<Allocator, softAes>::datasetPtr;
using VmBase<Allocator, softAes>::datasetOffset;
protected:
void execute();
JitCompiler compiler;
};
using CompiledVmDefault = CompiledVm<AlignedAllocator<CacheLineSize>, true>;
using CompiledVmHardAes = CompiledVm<AlignedAllocator<CacheLineSize>, false>;
using CompiledVmLargePage = CompiledVm<LargePageAllocator, true>;
using CompiledVmLargePageHardAes = CompiledVm<LargePageAllocator, false>;
}

View file

@ -0,0 +1,54 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "vm_compiled_light.hpp"
#include "common.hpp"
#include <stdexcept>
namespace randomx {
template<class Allocator, bool softAes>
void CompiledLightVm<Allocator, softAes>::setCache(randomx_cache* cache) {
cachePtr = cache;
mem.memory = cache->memory;
compiler.generateSuperscalarHash(cache->programs, cache->reciprocalCache);
}
template<class Allocator, bool softAes>
void CompiledLightVm<Allocator, softAes>::run(void* seed) {
VmBase<Allocator, softAes>::generateProgram(seed);
randomx_vm::initialize();
compiler.generateProgramLight(program, config, datasetOffset);
CompiledVm<Allocator, softAes>::execute();
}
template class CompiledLightVm<AlignedAllocator<CacheLineSize>, false>;
template class CompiledLightVm<AlignedAllocator<CacheLineSize>, true>;
template class CompiledLightVm<LargePageAllocator, false>;
template class CompiledLightVm<LargePageAllocator, true>;
}

View file

@ -0,0 +1,64 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <new>
#include "vm_compiled.hpp"
namespace randomx {
template<class Allocator, bool softAes>
class CompiledLightVm : public CompiledVm<Allocator, softAes> {
public:
void* operator new(size_t size) {
void* ptr = AlignedAllocator<CacheLineSize>::allocMemory(size);
if (ptr == nullptr)
throw std::bad_alloc();
return ptr;
}
void operator delete(void* ptr) {
AlignedAllocator<CacheLineSize>::freeMemory(ptr, sizeof(CompiledLightVm));
}
void setCache(randomx_cache* cache) override;
void setDataset(randomx_dataset* dataset) override { }
void run(void* seed) override;
using CompiledVm<Allocator, softAes>::mem;
using CompiledVm<Allocator, softAes>::compiler;
using CompiledVm<Allocator, softAes>::program;
using CompiledVm<Allocator, softAes>::config;
using CompiledVm<Allocator, softAes>::cachePtr;
using CompiledVm<Allocator, softAes>::datasetOffset;
};
using CompiledLightVmDefault = CompiledLightVm<AlignedAllocator<CacheLineSize>, true>;
using CompiledLightVmHardAes = CompiledLightVm<AlignedAllocator<CacheLineSize>, false>;
using CompiledLightVmLargePage = CompiledLightVm<LargePageAllocator, true>;
using CompiledLightVmLargePageHardAes = CompiledLightVm<LargePageAllocator, false>;
}

View file

@ -0,0 +1,125 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "vm_interpreted.hpp"
#include "dataset.hpp"
#include "intrin_portable.h"
#include "reciprocal.h"
namespace randomx {
template<class Allocator, bool softAes>
void InterpretedVm<Allocator, softAes>::setDataset(randomx_dataset* dataset) {
datasetPtr = dataset;
mem.memory = dataset->memory;
}
template<class Allocator, bool softAes>
void InterpretedVm<Allocator, softAes>::run(void* seed) {
VmBase<Allocator, softAes>::generateProgram(seed);
randomx_vm::initialize();
execute();
}
template<class Allocator, bool softAes>
void InterpretedVm<Allocator, softAes>::execute() {
NativeRegisterFile nreg;
for(unsigned i = 0; i < RegisterCountFlt; ++i)
nreg.a[i] = rx_load_vec_f128(&reg.a[i].lo);
compileProgram(program, bytecode, nreg);
uint32_t spAddr0 = mem.mx;
uint32_t spAddr1 = mem.ma;
for(unsigned ic = 0; ic < RandomX_CurrentConfig.ProgramIterations; ++ic) {
uint64_t spMix = nreg.r[config.readReg0] ^ nreg.r[config.readReg1];
spAddr0 ^= spMix;
spAddr0 &= ScratchpadL3Mask64;
spAddr1 ^= spMix >> 32;
spAddr1 &= ScratchpadL3Mask64;
for (unsigned i = 0; i < RegistersCount; ++i)
nreg.r[i] ^= load64(scratchpad + spAddr0 + 8 * i);
for (unsigned i = 0; i < RegisterCountFlt; ++i)
nreg.f[i] = rx_cvt_packed_int_vec_f128(scratchpad + spAddr1 + 8 * i);
for (unsigned i = 0; i < RegisterCountFlt; ++i)
nreg.e[i] = maskRegisterExponentMantissa(config, rx_cvt_packed_int_vec_f128(scratchpad + spAddr1 + 8 * (RegisterCountFlt + i)));
executeBytecode(bytecode, scratchpad, config);
mem.mx ^= nreg.r[config.readReg2] ^ nreg.r[config.readReg3];
mem.mx &= CacheLineAlignMask;
datasetPrefetch(datasetOffset + mem.mx);
datasetRead(datasetOffset + mem.ma, nreg.r);
std::swap(mem.mx, mem.ma);
for (unsigned i = 0; i < RegistersCount; ++i)
store64(scratchpad + spAddr1 + 8 * i, nreg.r[i]);
for (unsigned i = 0; i < RegisterCountFlt; ++i)
nreg.f[i] = rx_xor_vec_f128(nreg.f[i], nreg.e[i]);
for (unsigned i = 0; i < RegisterCountFlt; ++i)
rx_store_vec_f128((double*)(scratchpad + spAddr0 + 16 * i), nreg.f[i]);
spAddr0 = 0;
spAddr1 = 0;
}
for (unsigned i = 0; i < RegistersCount; ++i)
store64(&reg.r[i], nreg.r[i]);
for (unsigned i = 0; i < RegisterCountFlt; ++i)
rx_store_vec_f128(&reg.f[i].lo, nreg.f[i]);
for (unsigned i = 0; i < RegisterCountFlt; ++i)
rx_store_vec_f128(&reg.e[i].lo, nreg.e[i]);
}
template<class Allocator, bool softAes>
void InterpretedVm<Allocator, softAes>::datasetRead(uint64_t address, int_reg_t(&r)[RegistersCount]) {
uint64_t* datasetLine = (uint64_t*)(mem.memory + address);
for (int i = 0; i < RegistersCount; ++i)
r[i] ^= datasetLine[i];
}
template<class Allocator, bool softAes>
void InterpretedVm<Allocator, softAes>::datasetPrefetch(uint64_t address) {
rx_prefetch_nta(mem.memory + address);
}
template class InterpretedVm<AlignedAllocator<CacheLineSize>, false>;
template class InterpretedVm<AlignedAllocator<CacheLineSize>, true>;
template class InterpretedVm<LargePageAllocator, false>;
template class InterpretedVm<LargePageAllocator, true>;
}

View file

@ -0,0 +1,75 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <new>
#include <vector>
#include "common.hpp"
#include "virtual_machine.hpp"
#include "bytecode_machine.hpp"
#include "intrin_portable.h"
#include "allocator.hpp"
namespace randomx {
template<class Allocator, bool softAes>
class InterpretedVm : public VmBase<Allocator, softAes>, public BytecodeMachine {
public:
using VmBase<Allocator, softAes>::mem;
using VmBase<Allocator, softAes>::scratchpad;
using VmBase<Allocator, softAes>::program;
using VmBase<Allocator, softAes>::config;
using VmBase<Allocator, softAes>::reg;
using VmBase<Allocator, softAes>::datasetPtr;
using VmBase<Allocator, softAes>::datasetOffset;
void* operator new(size_t size) {
void* ptr = AlignedAllocator<CacheLineSize>::allocMemory(size);
if (ptr == nullptr)
throw std::bad_alloc();
return ptr;
}
void operator delete(void* ptr) {
AlignedAllocator<CacheLineSize>::freeMemory(ptr, sizeof(InterpretedVm));
}
void run(void* seed) override;
void setDataset(randomx_dataset* dataset) override;
protected:
virtual void datasetRead(uint64_t blockNumber, int_reg_t(&r)[RegistersCount]);
virtual void datasetPrefetch(uint64_t blockNumber);
private:
void execute();
InstructionByteCode bytecode[RANDOMX_PROGRAM_MAX_SIZE];
};
using InterpretedVmDefault = InterpretedVm<AlignedAllocator<CacheLineSize>, true>;
using InterpretedVmHardAes = InterpretedVm<AlignedAllocator<CacheLineSize>, false>;
using InterpretedVmLargePage = InterpretedVm<LargePageAllocator, true>;
using InterpretedVmLargePageHardAes = InterpretedVm<LargePageAllocator, false>;
}

View file

@ -0,0 +1,55 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "vm_interpreted_light.hpp"
#include "dataset.hpp"
namespace randomx {
template<class Allocator, bool softAes>
void InterpretedLightVm<Allocator, softAes>::setCache(randomx_cache* cache) {
cachePtr = cache;
mem.memory = cache->memory;
}
template<class Allocator, bool softAes>
void InterpretedLightVm<Allocator, softAes>::datasetRead(uint64_t address, int_reg_t(&r)[8]) {
uint32_t itemNumber = address / CacheLineSize;
int_reg_t rl[8];
initDatasetItem(cachePtr, (uint8_t*)rl, itemNumber);
for (unsigned q = 0; q < 8; ++q)
r[q] ^= rl[q];
}
template class InterpretedLightVm<AlignedAllocator<CacheLineSize>, false>;
template class InterpretedLightVm<AlignedAllocator<CacheLineSize>, true>;
template class InterpretedLightVm<LargePageAllocator, false>;
template class InterpretedLightVm<LargePageAllocator, true>;
}

View file

@ -0,0 +1,61 @@
/*
Copyright (c) 2018-2019, tevador <tevador@gmail.com>
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include <new>
#include "vm_interpreted.hpp"
namespace randomx {
template<class Allocator, bool softAes>
class InterpretedLightVm : public InterpretedVm<Allocator, softAes> {
public:
using VmBase<Allocator, softAes>::mem;
using VmBase<Allocator, softAes>::cachePtr;
void* operator new(size_t size) {
void* ptr = AlignedAllocator<CacheLineSize>::allocMemory(size);
if (ptr == nullptr)
throw std::bad_alloc();
return ptr;
}
void operator delete(void* ptr) {
AlignedAllocator<CacheLineSize>::freeMemory(ptr, sizeof(InterpretedLightVm));
}
void setDataset(randomx_dataset* dataset) override { }
void setCache(randomx_cache* cache) override;
protected:
void datasetRead(uint64_t address, int_reg_t(&r)[8]) override;
void datasetPrefetch(uint64_t address) override { }
};
using InterpretedLightVmDefault = InterpretedLightVm<AlignedAllocator<CacheLineSize>, true>;
using InterpretedLightVmHardAes = InterpretedLightVm<AlignedAllocator<CacheLineSize>, false>;
using InterpretedLightVmLargePage = InterpretedLightVm<LargePageAllocator, true>;
using InterpretedLightVmLargePageHardAes = InterpretedLightVm<LargePageAllocator, false>;
}

View file

@ -391,6 +391,7 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
cryptonight_quad_hash<CRYPTONIGHT, true, VARIANT_DOUBLE>, cryptonight_quad_hash<CRYPTONIGHT, true, VARIANT_DOUBLE>,
cryptonight_penta_hash<CRYPTONIGHT, true, VARIANT_DOUBLE>, cryptonight_penta_hash<CRYPTONIGHT, true, VARIANT_DOUBLE>,
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_LOKI
# ifdef XMRIG_ALGO_CN_LITE # ifdef XMRIG_ALGO_CN_LITE
cryptonight_single_hash<CRYPTONIGHT_LITE, false, VARIANT_0>, cryptonight_single_hash<CRYPTONIGHT_LITE, false, VARIANT_0>,
@ -431,6 +432,7 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_LOKI
# else # else
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0 nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1 nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1
@ -450,6 +452,7 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_LOKI
# endif # endif
# ifdef XMRIG_ALGO_CN_HEAVY # ifdef XMRIG_ALGO_CN_HEAVY
@ -503,6 +506,7 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_LOKI
# else # else
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0 nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1 nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1
@ -522,6 +526,7 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_LOKI
# endif # endif
# ifdef XMRIG_ALGO_CN_PICO # ifdef XMRIG_ALGO_CN_PICO
@ -554,6 +559,7 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_LOKI
# else # else
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0 nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1 nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1
@ -573,6 +579,7 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_LOKI
# endif # endif
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0 nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1 nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1
@ -592,6 +599,7 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_ZLS
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_DOUBLE
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_WOW
nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RX_LOKI
}; };
static_assert(count == sizeof(func_table) / sizeof(func_table[0]), "func_table size mismatch"); static_assert(count == sizeof(func_table) / sizeof(func_table[0]), "func_table size mismatch");

View file

@ -162,9 +162,9 @@ void MultiWorker<N>::start()
const xmrig::Variant v = m_state.job.algorithm().variant(); const xmrig::Variant v = m_state.job.algorithm().variant();
# ifdef XMRIG_ALGO_RANDOMX # ifdef XMRIG_ALGO_RANDOMX
if (v == xmrig::VARIANT_RX_WOW) { if ((v == xmrig::VARIANT_RX_WOW) || (v == xmrig::VARIANT_RX_LOKI)) {
allocateRandomX_VM(); allocateRandomX_VM();
Workers::updateDataset(m_state.job.seedHash(), m_totalWays); Workers::updateDataset(m_state.job.seedHash(), v, m_totalWays);
randomx_calculate_hash(m_rx_vm, m_state.blob, m_state.job.size(), m_hash); randomx_calculate_hash(m_rx_vm, m_state.blob, m_state.job.size(), m_hash);
} }
else else

View file

@ -65,6 +65,7 @@ uv_rwlock_t Workers::m_rx_dataset_lock;
randomx_cache *Workers::m_rx_cache = nullptr; randomx_cache *Workers::m_rx_cache = nullptr;
randomx_dataset *Workers::m_rx_dataset = nullptr; randomx_dataset *Workers::m_rx_dataset = nullptr;
uint8_t Workers::m_rx_seed_hash[32] = {}; uint8_t Workers::m_rx_seed_hash[32] = {};
xmrig::Variant Workers::m_rx_variant = xmrig::VARIANT_MAX;
std::atomic<uint32_t> Workers::m_rx_dataset_init_thread_counter = {}; std::atomic<uint32_t> Workers::m_rx_dataset_init_thread_counter = {};
#endif #endif
@ -380,10 +381,10 @@ void Workers::start(IWorker *worker)
#ifdef XMRIG_ALGO_RANDOMX #ifdef XMRIG_ALGO_RANDOMX
void Workers::updateDataset(const uint8_t* seed_hash, const uint32_t num_threads) void Workers::updateDataset(const uint8_t* seed_hash, xmrig::Variant variant, const uint32_t num_threads)
{ {
// Check if we need to update cache and dataset // Check if we need to update cache and dataset
if (memcmp(m_rx_seed_hash, seed_hash, sizeof(m_rx_seed_hash)) == 0) if ((memcmp(m_rx_seed_hash, seed_hash, sizeof(m_rx_seed_hash)) == 0) && (m_rx_variant == variant))
return; return;
const uint32_t thread_id = m_rx_dataset_init_thread_counter++; const uint32_t thread_id = m_rx_dataset_init_thread_counter++;
@ -400,10 +401,27 @@ void Workers::updateDataset(const uint8_t* seed_hash, const uint32_t num_threads
// One of the threads updates cache // One of the threads updates cache
uv_rwlock_wrlock(&m_rx_dataset_lock); uv_rwlock_wrlock(&m_rx_dataset_lock);
if (m_rx_variant != variant) {
switch (variant) {
case xmrig::VARIANT_RX_WOW:
randomx_apply_config(RandomX_WowneroConfig);
break;
case xmrig::VARIANT_RX_LOKI:
randomx_apply_config(RandomX_LokiConfig);
break;
default:
randomx_apply_config(RandomX_MoneroConfig);
break;
}
m_rx_variant = variant;
}
if (memcmp(m_rx_seed_hash, seed_hash, sizeof(m_rx_seed_hash)) != 0) { if (memcmp(m_rx_seed_hash, seed_hash, sizeof(m_rx_seed_hash)) != 0) {
memcpy(m_rx_seed_hash, seed_hash, sizeof(m_rx_seed_hash)); memcpy(m_rx_seed_hash, seed_hash, sizeof(m_rx_seed_hash));
randomx_init_cache(m_rx_cache, m_rx_seed_hash, sizeof(m_rx_seed_hash)); randomx_init_cache(m_rx_cache, m_rx_seed_hash, sizeof(m_rx_seed_hash));
} }
uv_rwlock_wrunlock(&m_rx_dataset_lock); uv_rwlock_wrunlock(&m_rx_dataset_lock);
// All threads update dataset // All threads update dataset

View file

@ -77,7 +77,7 @@ public:
# endif # endif
# ifdef XMRIG_ALGO_RANDOMX # ifdef XMRIG_ALGO_RANDOMX
static void updateDataset(const uint8_t* seed_hash, uint32_t num_threads); static void updateDataset(const uint8_t* seed_hash, xmrig::Variant variant, uint32_t num_threads);
static randomx_dataset* getDataset(); static randomx_dataset* getDataset();
# endif # endif
@ -131,6 +131,7 @@ private:
static randomx_cache *m_rx_cache; static randomx_cache *m_rx_cache;
static randomx_dataset *m_rx_dataset; static randomx_dataset *m_rx_dataset;
static uint8_t m_rx_seed_hash[32]; static uint8_t m_rx_seed_hash[32];
static xmrig::Variant m_rx_variant;
static std::atomic<uint32_t> m_rx_dataset_init_thread_counter; static std::atomic<uint32_t> m_rx_dataset_init_thread_counter;
# endif # endif
}; };