Merge pull request #5231

108c625b CryptonightR: define out i386/x86_64 specific code on other archs (moneromooo-monero)
1b8757dd slow-hash: fix build on arm (moneromooo-monero)
5057eb11 cmake: ARCH_ID fixes for cross compilation (TheCharlatan)
pull/200/head
Riccardo Spagni 5 years ago
commit 39d7d3113b
No known key found for this signature in database
GPG Key ID: 55432DF31CCD4FCD

@ -115,6 +115,9 @@ string(TOLOWER ${CMAKE_BUILD_TYPE} CMAKE_BUILD_TYPE_LOWER)
# to identify the target architecture, to direct logic in this cmake script.
# Since ARCH is a cached variable, it will not be set on first cmake invocation.
if (NOT ARCH OR ARCH STREQUAL "" OR ARCH STREQUAL "native" OR ARCH STREQUAL "default")
if(CMAKE_SYSTEM_PROCESSOR STREQUAL "")
set(CMAKE_SYSTEM_PROCESSOR ${CMAKE_HOST_SYSTEM_PROCESSOR})
endif()
set(ARCH_ID "${CMAKE_SYSTEM_PROCESSOR}")
else()
set(ARCH_ID "${ARCH}")

@ -47,6 +47,8 @@ set (CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER) # Find programs on host
set (CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY) # Find libs in target
set (CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY) # Find includes in target
set(CMAKE_SYSTEM_PROCESSOR ${CMAKE_HOST_SYSTEM_PROCESSOR} CACHE STRING "" FORCE)
# specify the cross compiler to be used. Darwin uses clang provided by the SDK.
if(CMAKE_SYSTEM_NAME STREQUAL "Darwin")
SET(CMAKE_C_COMPILER @prefix@/native/bin/clang)
@ -88,6 +90,11 @@ endif()
if(ARCHITECTURE STREQUAL "i686" AND CMAKE_SYSTEM_NAME STREQUAL "Linux")
SET(LINUX_32 ON)
SET(ARCH_ID "i386")
endif()
if(ARCHITECTURE STREQUAL "x86_64" AND CMAKE_SYSTEM_NAME STREQUAL "Linux")
SET(ARCH_ID "x86_64")
endif()
#Create a new global cmake flag that indicates building with depends

@ -46,9 +46,12 @@ set(crypto_sources
skein.c
slow-hash.c
CryptonightR_JIT.c
CryptonightR_template.S
tree-hash.c)
if(ARCH_ID STREQUAL "i386" OR ARCH_ID STREQUAL "x86_64" OR ARCH_ID STREQUAL "x86-64")
list(APPEND crypto_sources CryptonightR_template.S)
endif()
set(crypto_headers)
set(crypto_private_headers

@ -12,6 +12,7 @@
#include "CryptonightR_template.h"
static const uint8_t prologue[] = {
#if defined __i386 || defined __x86_64__
0x4C, 0x8B, 0xD7, // mov r10, rdi
0x53, // push rbx
0x55, // push rbp
@ -26,9 +27,11 @@ static const uint8_t prologue[] = {
0x41, 0x8B, 0x42, 0x18, // mov eax, DWORD PTR [r10+24]
0x41, 0x8B, 0x52, 0x1C, // mov edx, DWORD PTR [r10+28]
0x45, 0x8B, 0x4A, 0x20, // mov r9d, DWORD PTR [r10+32]
#endif
};
static const uint8_t epilogue[] = {
#if defined __i386 || defined __x86_64__
0x49, 0x8B, 0xE3, // mov rsp, r11
0x41, 0x89, 0x1A, // mov DWORD PTR [r10], ebx
0x41, 0x89, 0x72, 0x04, // mov DWORD PTR [r10+4], esi
@ -38,6 +41,7 @@ static const uint8_t epilogue[] = {
0x5D, // pop rbp
0x5B, // pop rbx
0xC3, // ret
#endif
};
#define APPEND_CODE(src, size) \
@ -50,6 +54,7 @@ static const uint8_t epilogue[] = {
int v4_generate_JIT_code(const struct V4_Instruction* code, v4_random_math_JIT_func buf, const size_t buf_size)
{
#if defined __i386 || defined __x86_64__
uint8_t* JIT_code = (uint8_t*) buf;
const uint8_t* JIT_code_end = JIT_code + buf_size;
@ -99,4 +104,7 @@ int v4_generate_JIT_code(const struct V4_Instruction* code, v4_random_math_JIT_f
__builtin___clear_cache((char*)buf, (char*)JIT_code);
return 0;
#else
return 1;
#endif
}

@ -8,7 +8,11 @@
// - Call v4_generate_JIT_code with "buf" pointed to memory allocated on previous step
// - Call the generated code instead of "v4_random_math(code, r)", omit the "code" parameter
typedef void (*v4_random_math_JIT_func)(uint32_t* r) __attribute__((sysv_abi));
typedef void (*v4_random_math_JIT_func)(uint32_t* r)
#if defined __i386 || defined __x86_64__
__attribute__((sysv_abi))
#endif
;
// Given the random math sequence, generates machine code (x86-64) for it
// Returns 0 if code was generated successfully

@ -1,6 +1,8 @@
#ifndef CRYPTONIGHTR_TEMPLATE_H
#define CRYPTONIGHTR_TEMPLATE_H
#if defined __i386 || defined __x86_64__
void CryptonightR_instruction0(void);
void CryptonightR_instruction1(void);
void CryptonightR_instruction2(void);
@ -1036,4 +1038,6 @@ const void* instructions_mov[257] = {
CryptonightR_instruction_mov256,
};
#endif
#endif // CRYPTONIGHTR_TEMPLATE_H

@ -65,6 +65,31 @@ static void local_abort(const char *msg)
#endif
}
volatile int use_v4_jit_flag = -1;
static inline int use_v4_jit(void)
{
#if defined(__x86_64__)
if (use_v4_jit_flag != -1)
return use_v4_jit_flag;
const char *env = getenv("MONERO_USE_CNV4_JIT");
if (!env) {
use_v4_jit_flag = 0;
}
else if (!strcmp(env, "0") || !strcmp(env, "no")) {
use_v4_jit_flag = 0;
}
else {
use_v4_jit_flag = 1;
}
return use_v4_jit_flag;
#else
return 0;
#endif
}
#define VARIANT1_1(p) \
do if (variant == 1) \
{ \
@ -494,31 +519,6 @@ STATIC INLINE int force_software_aes(void)
return use;
}
volatile int use_v4_jit_flag = -1;
STATIC INLINE int use_v4_jit(void)
{
#if defined(__x86_64__)
if (use_v4_jit_flag != -1)
return use_v4_jit_flag;
const char *env = getenv("MONERO_USE_CNV4_JIT");
if (!env) {
use_v4_jit_flag = 0;
}
else if (!strcmp(env, "0") || !strcmp(env, "no")) {
use_v4_jit_flag = 0;
}
else {
use_v4_jit_flag = 1;
}
return use_v4_jit_flag;
#else
return 0;
#endif
}
STATIC INLINE int check_aes_hw(void)
{
int cpuid_results[4];
@ -1029,6 +1029,8 @@ void slow_hash_free_state(void)
#define U64(x) ((uint64_t *) (x))
#define hp_jitfunc ((v4_random_math_JIT_func)NULL)
STATIC INLINE void xor64(uint64_t *a, const uint64_t b)
{
*a ^= b;
@ -1574,6 +1576,8 @@ void cn_slow_hash(const void *data, size_t length, char *hash, int variant, int
#else
// Portable implementation as a fallback
#define hp_jitfunc ((v4_random_math_JIT_func)NULL)
void slow_hash_allocate_state(void)
{
// Do nothing, this is just to maintain compatibility with the upgraded slow-hash.c

Loading…
Cancel
Save