Add very basic feature flagging for NEON

This commit is contained in:
Doridian 2022-03-27 10:19:34 -07:00
commit 62704db5ed
8 changed files with 210 additions and 26 deletions

View file

@ -10,10 +10,13 @@ target_include_directories(pm3rrg_rdv4_hardnested_nosimd PRIVATE
../../include ../../include
../src) ../src)
target_compile_definitions(pm3rrg_rdv4_hardnested_nosimd NOSIMD_BUILD)
## CPU-specific code ## CPU-specific code
## These are mostly for x86-based architectures, which is not useful for many Android devices. ## These are mostly for x86-based architectures, which is not useful for many Android devices.
## Mingw platforms: AMD64 ## Mingw platforms: AMD64
set(X86_CPUS x86 x86_64 i686 AMD64) set(X86_CPUS x86 x86_64 i686 AMD64)
set(ARM64_CPUS arm64 aarch64)
message(STATUS "CMAKE_SYSTEM_PROCESSOR := ${CMAKE_SYSTEM_PROCESSOR}") message(STATUS "CMAKE_SYSTEM_PROCESSOR := ${CMAKE_SYSTEM_PROCESSOR}")
@ -104,6 +107,24 @@ if ("${CMAKE_SYSTEM_PROCESSOR}" IN_LIST X86_CPUS)
$<TARGET_OBJECTS:pm3rrg_rdv4_hardnested_avx> $<TARGET_OBJECTS:pm3rrg_rdv4_hardnested_avx>
$<TARGET_OBJECTS:pm3rrg_rdv4_hardnested_avx2> $<TARGET_OBJECTS:pm3rrg_rdv4_hardnested_avx2>
$<TARGET_OBJECTS:pm3rrg_rdv4_hardnested_avx512>) $<TARGET_OBJECTS:pm3rrg_rdv4_hardnested_avx512>)
elseif ("${CMAKE_SYSTEM_PROCESSOR}" IN_LIST ARM64_CPUS)
message(STATUS "Building optimised arm64 binaries")
## arm64 / NEON
add_library(pm3rrg_rdv4_hardnested_neon OBJECT
hardnested/hardnested_bf_core.c
hardnested/hardnested_bitarray_core.c)
target_compile_options(pm3rrg_rdv4_hardnested_neon PRIVATE -Wall -Werror -O3)
set_property(TARGET pm3rrg_rdv4_hardnested_neon PROPERTY POSITION_INDEPENDENT_CODE ON)
target_include_directories(pm3rrg_rdv4_hardnested_neon PRIVATE
../../common
../../include
../src)
set(SIMD_TARGETS
$<TARGET_OBJECTS:pm3rrg_rdv4_hardnested_neon>)
else () else ()
message(STATUS "Not building optimised targets") message(STATUS "Not building optimised targets")
set(SIMD_TARGETS) set(SIMD_TARGETS)

View file

@ -11,6 +11,9 @@ endif
ifneq ($(findstring amd64, $(cpu_arch)), ) ifneq ($(findstring amd64, $(cpu_arch)), )
MULTIARCHSRCS = hardnested_bf_core.c hardnested_bitarray_core.c MULTIARCHSRCS = hardnested_bf_core.c hardnested_bitarray_core.c
endif endif
ifneq ($(findstring arm64, $(cpu_arch)), )
MULTIARCHSRCS = hardnested_bf_core.c hardnested_bitarray_core.c
endif
ifeq ($(MULTIARCHSRCS), ) ifeq ($(MULTIARCHSRCS), )
MYSRCS += hardnested_bf_core.c hardnested_bitarray_core.c MYSRCS += hardnested_bf_core.c hardnested_bitarray_core.c
endif endif
@ -18,20 +21,30 @@ endif
LIB_A = libhardnested.a LIB_A = libhardnested.a
MYOBJS = $(MYSRCS:%.c=$(OBJDIR)/%.o) MYOBJS = $(MYSRCS:%.c=$(OBJDIR)/%.o)
MYOBJS += $(MULTIARCHSRCS:%.c=$(OBJDIR)/%_NOSIMD.o) \ ifneq ($(findstring arm64, $(cpu_arch)), )
$(MULTIARCHSRCS:%.c=$(OBJDIR)/%_MMX.o) \ MYOBJS += $(MULTIARCHSRCS:%.c=$(OBJDIR)/%_NOSIMD.o) \
$(MULTIARCHSRCS:%.c=$(OBJDIR)/%_SSE2.o) \ $(MULTIARCHSRCS:%.c=$(OBJDIR)/%_NEON.o)
$(MULTIARCHSRCS:%.c=$(OBJDIR)/%_AVX.o) \ else
$(MULTIARCHSRCS:%.c=$(OBJDIR)/%_AVX2.o) MYOBJS += $(MULTIARCHSRCS:%.c=$(OBJDIR)/%_NOSIMD.o) \
$(MULTIARCHSRCS:%.c=$(OBJDIR)/%_MMX.o) \
$(MULTIARCHSRCS:%.c=$(OBJDIR)/%_SSE2.o) \
$(MULTIARCHSRCS:%.c=$(OBJDIR)/%_AVX.o) \
$(MULTIARCHSRCS:%.c=$(OBJDIR)/%_AVX2.o)
endif
SUPPORTS_AVX512 := $(shell echo | $(CC) -E -mavx512f - > /dev/null 2>&1 && echo "True" ) SUPPORTS_AVX512 := $(shell echo | $(CC) -E -mavx512f - > /dev/null 2>&1 && echo "True" )
HARD_SWITCH_NOSIMD = -mno-mmx -mno-sse2 -mno-avx -mno-avx2 HARD_SWITCH_NOSIMD = -mno-mmx -mno-sse2 -mno-avx -mno-avx2 -DNOSIMD_BUILD
HARD_SWITCH_NEON =
HARD_SWITCH_MMX = -mmmx -mno-sse2 -mno-avx -mno-avx2 HARD_SWITCH_MMX = -mmmx -mno-sse2 -mno-avx -mno-avx2
HARD_SWITCH_SSE2 = -mmmx -msse2 -mno-avx -mno-avx2 HARD_SWITCH_SSE2 = -mmmx -msse2 -mno-avx -mno-avx2
HARD_SWITCH_AVX = -mmmx -msse2 -mavx -mno-avx2 HARD_SWITCH_AVX = -mmmx -msse2 -mavx -mno-avx2
HARD_SWITCH_AVX2 = -mmmx -msse2 -mavx -mavx2 HARD_SWITCH_AVX2 = -mmmx -msse2 -mavx -mavx2
HARD_SWITCH_AVX512 = -mmmx -msse2 -mavx -mavx2 -mavx512f HARD_SWITCH_AVX512 = -mmmx -msse2 -mavx -mavx2 -mavx512f
ifneq ($(findstring arm64, $(cpu_arch)), )
SUPPORTS_AVX512=0
HARD_SWITCH_NOSIMD = -DNOSIMD_BUILD
endif
ifeq "$(SUPPORTS_AVX512)" "True" ifeq "$(SUPPORTS_AVX512)" "True"
HARD_SWITCH_NOSIMD += -mno-avx512f HARD_SWITCH_NOSIMD += -mno-avx512f
HARD_SWITCH_MMX += -mno-avx512f HARD_SWITCH_MMX += -mno-avx512f
@ -52,6 +65,12 @@ $(OBJDIR)/%_NOSIMD.o : %.c $(OBJDIR)/%_NOSIMD.d
$(Q)$(CC) $(DEPFLAGS:%.Td=%_NOSIMD.Td) $(CFLAGS) $(HARD_SWITCH_NOSIMD) -c -o $@ $< $(Q)$(CC) $(DEPFLAGS:%.Td=%_NOSIMD.Td) $(CFLAGS) $(HARD_SWITCH_NOSIMD) -c -o $@ $<
$(Q)$(MV) -f $(OBJDIR)/$*_NOSIMD.Td $(OBJDIR)/$*_NOSIMD.d && $(TOUCH) $@ $(Q)$(MV) -f $(OBJDIR)/$*_NOSIMD.Td $(OBJDIR)/$*_NOSIMD.d && $(TOUCH) $@
$(OBJDIR)/%_NEON.o : %.c $(OBJDIR)/%_NEON.d
$(info [-] CC(NEON) $<)
$(Q)$(MKDIR) $(dir $@)
$(Q)$(CC) $(DEPFLAGS:%.Td=%_NEON.Td) $(CFLAGS) $(HARD_SWITCH_NEON) -c -o $@ $<
$(Q)$(MV) -f $(OBJDIR)/$*_NEON.Td $(OBJDIR)/$*_NEON.d && $(TOUCH) $@
$(OBJDIR)/%_MMX.o : %.c $(OBJDIR)/%_MMX.d $(OBJDIR)/%_MMX.o : %.c $(OBJDIR)/%_MMX.d
$(info [-] CC(MMX) $<) $(info [-] CC(MMX) $<)
$(Q)$(MKDIR) $(dir $@) $(Q)$(MKDIR) $(dir $@)

View file

@ -74,7 +74,7 @@ THE SOFTWARE.
#define MAX_BITSLICES 128 #define MAX_BITSLICES 128
#elif defined(__SSE2__) #elif defined(__SSE2__)
#define MAX_BITSLICES 128 #define MAX_BITSLICES 128
#elif defined(__ARM_NEON) #elif defined(__ARM_NEON) && !defined(NOSIMD_BUILD)
#define MAX_BITSLICES 128 #define MAX_BITSLICES 128
#else // MMX or SSE or NOSIMD #else // MMX or SSE or NOSIMD
#define MAX_BITSLICES 64 #define MAX_BITSLICES 64
@ -120,6 +120,9 @@ typedef union {
#elif defined (__MMX__) #elif defined (__MMX__)
#define BITSLICE_TEST_NONCES bitslice_test_nonces_MMX #define BITSLICE_TEST_NONCES bitslice_test_nonces_MMX
#define CRACK_STATES_BITSLICED crack_states_bitsliced_MMX #define CRACK_STATES_BITSLICED crack_states_bitsliced_MMX
#elif defined (__ARM_NEON) && !defined(NOSIMD_BUILD)
#define BITSLICE_TEST_NONCES bitslice_test_nonces_NEON
#define CRACK_STATES_BITSLICED crack_states_bitsliced_NEON
#else #else
#define BITSLICE_TEST_NONCES bitslice_test_nonces_NOSIMD #define BITSLICE_TEST_NONCES bitslice_test_nonces_NOSIMD
#define CRACK_STATES_BITSLICED crack_states_bitsliced_NOSIMD #define CRACK_STATES_BITSLICED crack_states_bitsliced_NOSIMD
@ -132,6 +135,7 @@ crack_states_bitsliced_t crack_states_bitsliced_AVX2;
crack_states_bitsliced_t crack_states_bitsliced_AVX; crack_states_bitsliced_t crack_states_bitsliced_AVX;
crack_states_bitsliced_t crack_states_bitsliced_SSE2; crack_states_bitsliced_t crack_states_bitsliced_SSE2;
crack_states_bitsliced_t crack_states_bitsliced_MMX; crack_states_bitsliced_t crack_states_bitsliced_MMX;
crack_states_bitsliced_t crack_states_bitsliced_NEON;
crack_states_bitsliced_t crack_states_bitsliced_NOSIMD; crack_states_bitsliced_t crack_states_bitsliced_NOSIMD;
crack_states_bitsliced_t crack_states_bitsliced_dispatch; crack_states_bitsliced_t crack_states_bitsliced_dispatch;
@ -141,6 +145,7 @@ bitslice_test_nonces_t bitslice_test_nonces_AVX2;
bitslice_test_nonces_t bitslice_test_nonces_AVX; bitslice_test_nonces_t bitslice_test_nonces_AVX;
bitslice_test_nonces_t bitslice_test_nonces_SSE2; bitslice_test_nonces_t bitslice_test_nonces_SSE2;
bitslice_test_nonces_t bitslice_test_nonces_MMX; bitslice_test_nonces_t bitslice_test_nonces_MMX;
bitslice_test_nonces_t bitslice_test_nonces_NEON;
bitslice_test_nonces_t bitslice_test_nonces_NOSIMD; bitslice_test_nonces_t bitslice_test_nonces_NOSIMD;
bitslice_test_nonces_t bitslice_test_nonces_dispatch; bitslice_test_nonces_t bitslice_test_nonces_dispatch;
@ -545,7 +550,7 @@ out:
#ifndef __MMX__ #ifdef NOSIMD_BUILD
// pointers to functions: // pointers to functions:
crack_states_bitsliced_t *crack_states_bitsliced_function_p = &crack_states_bitsliced_dispatch; crack_states_bitsliced_t *crack_states_bitsliced_function_p = &crack_states_bitsliced_dispatch;
@ -582,6 +587,11 @@ static SIMDExecInstr GetSIMDInstr(void) {
else if (__builtin_cpu_supports("mmx")) else if (__builtin_cpu_supports("mmx"))
instr = SIMD_MMX; instr = SIMD_MMX;
else else
#endif
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon())
instr = SIMD_NEON;
else
#endif #endif
instr = SIMD_NONE; instr = SIMD_NONE;
@ -620,6 +630,11 @@ uint64_t crack_states_bitsliced_dispatch(uint32_t cuid, uint8_t *best_first_byte
case SIMD_MMX: case SIMD_MMX:
crack_states_bitsliced_function_p = &crack_states_bitsliced_MMX; crack_states_bitsliced_function_p = &crack_states_bitsliced_MMX;
break; break;
#endif
#if defined(COMPILER_HAS_SIMD_NEON)
case SIMD_NEON:
crack_states_bitsliced_function_p = &crack_states_bitsliced_NEON;
break;
#endif #endif
case SIMD_AUTO: case SIMD_AUTO:
case SIMD_NONE: case SIMD_NONE:
@ -651,6 +666,11 @@ void bitslice_test_nonces_dispatch(uint32_t nonces_to_bruteforce, const uint32_t
case SIMD_MMX: case SIMD_MMX:
bitslice_test_nonces_function_p = &bitslice_test_nonces_MMX; bitslice_test_nonces_function_p = &bitslice_test_nonces_MMX;
break; break;
#endif
#if defined(COMPILER_HAS_SIMD_NEON)
case SIMD_NEON:
bitslice_test_nonces_function_p = &bitslice_test_nonces_NEON;
break;
#endif #endif
case SIMD_AUTO: case SIMD_AUTO:
case SIMD_NONE: case SIMD_NONE:

View file

@ -61,6 +61,16 @@ THE SOFTWARE.
# endif # endif
#endif #endif
// ARM64 mandates implementation of NEON
#if defined(__arm64__)
#define COMPILER_HAS_SIMD_NEON
#define arm_has_neon() (true)
// ARMv7 or older, NEON is optional and autodetection is difficult
#elif defined(__ARM_NEON)
#define COMPILER_HAS_SIMD_NEON
#define arm_has_neon() (false)
#endif
typedef enum { typedef enum {
SIMD_AUTO, SIMD_AUTO,
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
@ -71,6 +81,9 @@ typedef enum {
SIMD_AVX, SIMD_AVX,
SIMD_SSE2, SIMD_SSE2,
SIMD_MMX, SIMD_MMX,
#endif
#if defined(COMPILER_HAS_SIMD_NEON)
SIMD_NEON,
#endif #endif
SIMD_NONE, SIMD_NONE,
} SIMDExecInstr; } SIMDExecInstr;

View file

@ -98,6 +98,20 @@
#define COUNT_BITARRAY_AND2 count_bitarray_AND2_MMX #define COUNT_BITARRAY_AND2 count_bitarray_AND2_MMX
#define COUNT_BITARRAY_AND3 count_bitarray_AND3_MMX #define COUNT_BITARRAY_AND3 count_bitarray_AND3_MMX
#define COUNT_BITARRAY_AND4 count_bitarray_AND4_MMX #define COUNT_BITARRAY_AND4 count_bitarray_AND4_MMX
#elif defined (__ARM_NEON) && !defined (NOSIMD_BUILD)
#define MALLOC_BITARRAY malloc_bitarray_NEON
#define FREE_BITARRAY free_bitarray_NEON
#define BITCOUNT bitcount_NEON
#define COUNT_STATES count_states_NEON
#define BITARRAY_AND bitarray_AND_NEON
#define BITARRAY_LOW20_AND bitarray_low20_AND_NEON
#define COUNT_BITARRAY_AND count_bitarray_AND_NEON
#define COUNT_BITARRAY_LOW20_AND count_bitarray_low20_AND_NEON
#define BITARRAY_AND4 bitarray_AND4_NEON
#define BITARRAY_OR bitarray_OR_NEON
#define COUNT_BITARRAY_AND2 count_bitarray_AND2_NEON
#define COUNT_BITARRAY_AND3 count_bitarray_AND3_NEON
#define COUNT_BITARRAY_AND4 count_bitarray_AND4_NEON
#else #else
#define MALLOC_BITARRAY malloc_bitarray_NOSIMD #define MALLOC_BITARRAY malloc_bitarray_NOSIMD
#define FREE_BITARRAY free_bitarray_NOSIMD #define FREE_BITARRAY free_bitarray_NOSIMD
@ -117,31 +131,31 @@
// typedefs and declaration of functions: // typedefs and declaration of functions:
typedef uint32_t *malloc_bitarray_t(uint32_t); typedef uint32_t *malloc_bitarray_t(uint32_t);
malloc_bitarray_t malloc_bitarray_AVX512, malloc_bitarray_AVX2, malloc_bitarray_AVX, malloc_bitarray_SSE2, malloc_bitarray_MMX, malloc_bitarray_NOSIMD, malloc_bitarray_dispatch; malloc_bitarray_t malloc_bitarray_AVX512, malloc_bitarray_AVX2, malloc_bitarray_AVX, malloc_bitarray_SSE2, malloc_bitarray_MMX, malloc_bitarray_NOSIMD, malloc_bitarray_NEON, malloc_bitarray_dispatch;
typedef void free_bitarray_t(uint32_t *); typedef void free_bitarray_t(uint32_t *);
free_bitarray_t free_bitarray_AVX512, free_bitarray_AVX2, free_bitarray_AVX, free_bitarray_SSE2, free_bitarray_MMX, free_bitarray_NOSIMD, free_bitarray_dispatch; free_bitarray_t free_bitarray_AVX512, free_bitarray_AVX2, free_bitarray_AVX, free_bitarray_SSE2, free_bitarray_MMX, free_bitarray_NOSIMD, free_bitarray_NEON, free_bitarray_dispatch;
typedef uint32_t bitcount_t(uint32_t); typedef uint32_t bitcount_t(uint32_t);
bitcount_t bitcount_AVX512, bitcount_AVX2, bitcount_AVX, bitcount_SSE2, bitcount_MMX, bitcount_NOSIMD, bitcount_dispatch; bitcount_t bitcount_AVX512, bitcount_AVX2, bitcount_AVX, bitcount_SSE2, bitcount_MMX, bitcount_NOSIMD, bitcount_NEON, bitcount_dispatch;
typedef uint32_t count_states_t(uint32_t *); typedef uint32_t count_states_t(uint32_t *);
count_states_t count_states_AVX512, count_states_AVX2, count_states_AVX, count_states_SSE2, count_states_MMX, count_states_NOSIMD, count_states_dispatch; count_states_t count_states_AVX512, count_states_AVX2, count_states_AVX, count_states_SSE2, count_states_MMX, count_states_NOSIMD, count_states_NEON, count_states_dispatch;
typedef void bitarray_AND_t(uint32_t[], uint32_t[]); typedef void bitarray_AND_t(uint32_t[], uint32_t[]);
bitarray_AND_t bitarray_AND_AVX512, bitarray_AND_AVX2, bitarray_AND_AVX, bitarray_AND_SSE2, bitarray_AND_MMX, bitarray_AND_NOSIMD, bitarray_AND_dispatch; bitarray_AND_t bitarray_AND_AVX512, bitarray_AND_AVX2, bitarray_AND_AVX, bitarray_AND_SSE2, bitarray_AND_MMX, bitarray_AND_NOSIMD, bitarray_AND_NEON, bitarray_AND_dispatch;
typedef void bitarray_low20_AND_t(uint32_t *, uint32_t *); typedef void bitarray_low20_AND_t(uint32_t *, uint32_t *);
bitarray_low20_AND_t bitarray_low20_AND_AVX512, bitarray_low20_AND_AVX2, bitarray_low20_AND_AVX, bitarray_low20_AND_SSE2, bitarray_low20_AND_MMX, bitarray_low20_AND_NOSIMD, bitarray_low20_AND_dispatch; bitarray_low20_AND_t bitarray_low20_AND_AVX512, bitarray_low20_AND_AVX2, bitarray_low20_AND_AVX, bitarray_low20_AND_SSE2, bitarray_low20_AND_MMX, bitarray_low20_AND_NOSIMD, bitarray_low20_AND_NEON, bitarray_low20_AND_dispatch;
typedef uint32_t count_bitarray_AND_t(uint32_t *, uint32_t *); typedef uint32_t count_bitarray_AND_t(uint32_t *, uint32_t *);
count_bitarray_AND_t count_bitarray_AND_AVX512, count_bitarray_AND_AVX2, count_bitarray_AND_AVX, count_bitarray_AND_SSE2, count_bitarray_AND_MMX, count_bitarray_AND_NOSIMD, count_bitarray_AND_dispatch; count_bitarray_AND_t count_bitarray_AND_AVX512, count_bitarray_AND_AVX2, count_bitarray_AND_AVX, count_bitarray_AND_SSE2, count_bitarray_AND_MMX, count_bitarray_AND_NOSIMD, count_bitarray_AND_NEON, count_bitarray_AND_dispatch;
typedef uint32_t count_bitarray_low20_AND_t(uint32_t *, uint32_t *); typedef uint32_t count_bitarray_low20_AND_t(uint32_t *, uint32_t *);
count_bitarray_low20_AND_t count_bitarray_low20_AND_AVX512, count_bitarray_low20_AND_AVX2, count_bitarray_low20_AND_AVX, count_bitarray_low20_AND_SSE2, count_bitarray_low20_AND_MMX, count_bitarray_low20_AND_NOSIMD, count_bitarray_low20_AND_dispatch; count_bitarray_low20_AND_t count_bitarray_low20_AND_AVX512, count_bitarray_low20_AND_AVX2, count_bitarray_low20_AND_AVX, count_bitarray_low20_AND_SSE2, count_bitarray_low20_AND_MMX, count_bitarray_low20_AND_NOSIMD, count_bitarray_low20_AND_NEON, count_bitarray_low20_AND_dispatch;
typedef void bitarray_AND4_t(uint32_t *, uint32_t *, uint32_t *, uint32_t *); typedef void bitarray_AND4_t(uint32_t *, uint32_t *, uint32_t *, uint32_t *);
bitarray_AND4_t bitarray_AND4_AVX512, bitarray_AND4_AVX2, bitarray_AND4_AVX, bitarray_AND4_SSE2, bitarray_AND4_MMX, bitarray_AND4_NOSIMD, bitarray_AND4_dispatch; bitarray_AND4_t bitarray_AND4_AVX512, bitarray_AND4_AVX2, bitarray_AND4_AVX, bitarray_AND4_SSE2, bitarray_AND4_MMX, bitarray_AND4_NOSIMD, bitarray_AND4_NEON, bitarray_AND4_dispatch;
typedef void bitarray_OR_t(uint32_t[], uint32_t[]); typedef void bitarray_OR_t(uint32_t[], uint32_t[]);
bitarray_OR_t bitarray_OR_AVX512, bitarray_OR_AVX2, bitarray_OR_AVX, bitarray_OR_SSE2, bitarray_OR_MMX, bitarray_OR_NOSIMD, bitarray_OR_dispatch; bitarray_OR_t bitarray_OR_AVX512, bitarray_OR_AVX2, bitarray_OR_AVX, bitarray_OR_SSE2, bitarray_OR_MMX, bitarray_OR_NOSIMD, bitarray_OR_NEON, bitarray_OR_dispatch;
typedef uint32_t count_bitarray_AND2_t(uint32_t *, uint32_t *); typedef uint32_t count_bitarray_AND2_t(uint32_t *, uint32_t *);
count_bitarray_AND2_t count_bitarray_AND2_AVX512, count_bitarray_AND2_AVX2, count_bitarray_AND2_AVX, count_bitarray_AND2_SSE2, count_bitarray_AND2_MMX, count_bitarray_AND2_NOSIMD, count_bitarray_AND2_dispatch; count_bitarray_AND2_t count_bitarray_AND2_AVX512, count_bitarray_AND2_AVX2, count_bitarray_AND2_AVX, count_bitarray_AND2_SSE2, count_bitarray_AND2_MMX, count_bitarray_AND2_NOSIMD, count_bitarray_AND2_NEON, count_bitarray_AND2_dispatch;
typedef uint32_t count_bitarray_AND3_t(uint32_t *, uint32_t *, uint32_t *); typedef uint32_t count_bitarray_AND3_t(uint32_t *, uint32_t *, uint32_t *);
count_bitarray_AND3_t count_bitarray_AND3_AVX512, count_bitarray_AND3_AVX2, count_bitarray_AND3_AVX, count_bitarray_AND3_SSE2, count_bitarray_AND3_MMX, count_bitarray_AND3_NOSIMD, count_bitarray_AND3_dispatch; count_bitarray_AND3_t count_bitarray_AND3_AVX512, count_bitarray_AND3_AVX2, count_bitarray_AND3_AVX, count_bitarray_AND3_SSE2, count_bitarray_AND3_MMX, count_bitarray_AND3_NOSIMD, count_bitarray_AND3_NEON, count_bitarray_AND3_dispatch;
typedef uint32_t count_bitarray_AND4_t(uint32_t *, uint32_t *, uint32_t *, uint32_t *); typedef uint32_t count_bitarray_AND4_t(uint32_t *, uint32_t *, uint32_t *, uint32_t *);
count_bitarray_AND4_t count_bitarray_AND4_AVX512, count_bitarray_AND4_AVX2, count_bitarray_AND4_AVX, count_bitarray_AND4_SSE2, count_bitarray_AND4_MMX, count_bitarray_AND4_NOSIMD, count_bitarray_AND4_dispatch; count_bitarray_AND4_t count_bitarray_AND4_AVX512, count_bitarray_AND4_AVX2, count_bitarray_AND4_AVX, count_bitarray_AND4_SSE2, count_bitarray_AND4_MMX, count_bitarray_AND4_NOSIMD, count_bitarray_AND4_NEON, count_bitarray_AND4_dispatch;
inline uint32_t *MALLOC_BITARRAY(uint32_t x) { inline uint32_t *MALLOC_BITARRAY(uint32_t x) {
@ -287,7 +301,7 @@ inline uint32_t COUNT_BITARRAY_AND4(uint32_t *restrict A, uint32_t *restrict B,
} }
#ifndef __MMX__ #ifdef NOSIMD_BUILD
// pointers to functions: // pointers to functions:
malloc_bitarray_t *malloc_bitarray_function_p = &malloc_bitarray_dispatch; malloc_bitarray_t *malloc_bitarray_function_p = &malloc_bitarray_dispatch;
@ -306,6 +320,11 @@ count_bitarray_AND4_t *count_bitarray_AND4_function_p = &count_bitarray_AND4_dis
// determine the available instruction set at runtime and call the correct function // determine the available instruction set at runtime and call the correct function
uint32_t *malloc_bitarray_dispatch(uint32_t x) { uint32_t *malloc_bitarray_dispatch(uint32_t x) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) malloc_bitarray_function_p = &malloc_bitarray_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) malloc_bitarray_function_p = &malloc_bitarray_AVX512; if (__builtin_cpu_supports("avx512f")) malloc_bitarray_function_p = &malloc_bitarray_AVX512;
else else
@ -324,6 +343,11 @@ uint32_t *malloc_bitarray_dispatch(uint32_t x) {
} }
void free_bitarray_dispatch(uint32_t *x) { void free_bitarray_dispatch(uint32_t *x) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) free_bitarray_function_p = &free_bitarray_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) free_bitarray_function_p = &free_bitarray_AVX512; if (__builtin_cpu_supports("avx512f")) free_bitarray_function_p = &free_bitarray_AVX512;
else else
@ -342,6 +366,11 @@ void free_bitarray_dispatch(uint32_t *x) {
} }
uint32_t bitcount_dispatch(uint32_t a) { uint32_t bitcount_dispatch(uint32_t a) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) bitcount_function_p = &bitcount_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) bitcount_function_p = &bitcount_AVX512; if (__builtin_cpu_supports("avx512f")) bitcount_function_p = &bitcount_AVX512;
else else
@ -360,6 +389,11 @@ uint32_t bitcount_dispatch(uint32_t a) {
} }
uint32_t count_states_dispatch(uint32_t *bitarray) { uint32_t count_states_dispatch(uint32_t *bitarray) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) count_states_function_p = &count_states_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) count_states_function_p = &count_states_AVX512; if (__builtin_cpu_supports("avx512f")) count_states_function_p = &count_states_AVX512;
else else
@ -378,6 +412,11 @@ uint32_t count_states_dispatch(uint32_t *bitarray) {
} }
void bitarray_AND_dispatch(uint32_t *A, uint32_t *B) { void bitarray_AND_dispatch(uint32_t *A, uint32_t *B) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) bitarray_AND_function_p = &bitarray_AND_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) bitarray_AND_function_p = &bitarray_AND_AVX512; if (__builtin_cpu_supports("avx512f")) bitarray_AND_function_p = &bitarray_AND_AVX512;
else else
@ -396,6 +435,11 @@ void bitarray_AND_dispatch(uint32_t *A, uint32_t *B) {
} }
void bitarray_low20_AND_dispatch(uint32_t *A, uint32_t *B) { void bitarray_low20_AND_dispatch(uint32_t *A, uint32_t *B) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) bitarray_low20_AND_function_p = &bitarray_low20_AND_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) bitarray_low20_AND_function_p = &bitarray_low20_AND_AVX512; if (__builtin_cpu_supports("avx512f")) bitarray_low20_AND_function_p = &bitarray_low20_AND_AVX512;
else else
@ -414,6 +458,11 @@ void bitarray_low20_AND_dispatch(uint32_t *A, uint32_t *B) {
} }
uint32_t count_bitarray_AND_dispatch(uint32_t *A, uint32_t *B) { uint32_t count_bitarray_AND_dispatch(uint32_t *A, uint32_t *B) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) count_bitarray_AND_function_p = &count_bitarray_AND_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) count_bitarray_AND_function_p = &count_bitarray_AND_AVX512; if (__builtin_cpu_supports("avx512f")) count_bitarray_AND_function_p = &count_bitarray_AND_AVX512;
else else
@ -432,6 +481,11 @@ uint32_t count_bitarray_AND_dispatch(uint32_t *A, uint32_t *B) {
} }
uint32_t count_bitarray_low20_AND_dispatch(uint32_t *A, uint32_t *B) { uint32_t count_bitarray_low20_AND_dispatch(uint32_t *A, uint32_t *B) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_AVX512; if (__builtin_cpu_supports("avx512f")) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_AVX512;
else else
@ -450,6 +504,11 @@ uint32_t count_bitarray_low20_AND_dispatch(uint32_t *A, uint32_t *B) {
} }
void bitarray_AND4_dispatch(uint32_t *A, uint32_t *B, uint32_t *C, uint32_t *D) { void bitarray_AND4_dispatch(uint32_t *A, uint32_t *B, uint32_t *C, uint32_t *D) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) bitarray_AND4_function_p = &bitarray_AND4_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) bitarray_AND4_function_p = &bitarray_AND4_AVX512; if (__builtin_cpu_supports("avx512f")) bitarray_AND4_function_p = &bitarray_AND4_AVX512;
else else
@ -468,6 +527,11 @@ void bitarray_AND4_dispatch(uint32_t *A, uint32_t *B, uint32_t *C, uint32_t *D)
} }
void bitarray_OR_dispatch(uint32_t *A, uint32_t *B) { void bitarray_OR_dispatch(uint32_t *A, uint32_t *B) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) bitarray_OR_function_p = &bitarray_OR_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) bitarray_OR_function_p = &bitarray_OR_AVX512; if (__builtin_cpu_supports("avx512f")) bitarray_OR_function_p = &bitarray_OR_AVX512;
else else
@ -486,6 +550,11 @@ void bitarray_OR_dispatch(uint32_t *A, uint32_t *B) {
} }
uint32_t count_bitarray_AND2_dispatch(uint32_t *A, uint32_t *B) { uint32_t count_bitarray_AND2_dispatch(uint32_t *A, uint32_t *B) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) count_bitarray_AND2_function_p = &count_bitarray_AND2_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) count_bitarray_AND2_function_p = &count_bitarray_AND2_AVX512; if (__builtin_cpu_supports("avx512f")) count_bitarray_AND2_function_p = &count_bitarray_AND2_AVX512;
else else
@ -504,6 +573,11 @@ uint32_t count_bitarray_AND2_dispatch(uint32_t *A, uint32_t *B) {
} }
uint32_t count_bitarray_AND3_dispatch(uint32_t *A, uint32_t *B, uint32_t *C) { uint32_t count_bitarray_AND3_dispatch(uint32_t *A, uint32_t *B, uint32_t *C) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) count_bitarray_AND3_function_p = &count_bitarray_AND3_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) count_bitarray_AND3_function_p = &count_bitarray_AND3_AVX512; if (__builtin_cpu_supports("avx512f")) count_bitarray_AND3_function_p = &count_bitarray_AND3_AVX512;
else else
@ -522,6 +596,11 @@ uint32_t count_bitarray_AND3_dispatch(uint32_t *A, uint32_t *B, uint32_t *C) {
} }
uint32_t count_bitarray_AND4_dispatch(uint32_t *A, uint32_t *B, uint32_t *C, uint32_t *D) { uint32_t count_bitarray_AND4_dispatch(uint32_t *A, uint32_t *B, uint32_t *C, uint32_t *D) {
#if defined(COMPILER_HAS_SIMD_NEON)
if (arm_has_neon()) count_bitarray_AND4_function_p = &count_bitarray_AND4_NEON;
else
#endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
if (__builtin_cpu_supports("avx512f")) count_bitarray_AND4_function_p = &count_bitarray_AND4_AVX512; if (__builtin_cpu_supports("avx512f")) count_bitarray_AND4_function_p = &count_bitarray_AND4_AVX512;
else else

View file

@ -1873,6 +1873,9 @@ static int CmdHF14AMfNestedHard(const char *Cmd) {
#endif #endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
arg_lit0(NULL, "i5", "AVX512"), arg_lit0(NULL, "i5", "AVX512"),
#endif
#if defined(COMPILER_HAS_SIMD_NEON)
arg_lit0(NULL, "ie", "NEON"),
#endif #endif
arg_param_end arg_param_end
}; };
@ -1930,6 +1933,9 @@ static int CmdHF14AMfNestedHard(const char *Cmd) {
#endif #endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
bool i5 = arg_get_lit(ctx, 20); bool i5 = arg_get_lit(ctx, 20);
#endif
#if defined(COMPILER_HAS_SIMD_NEON)
bool ie = arg_get_lit(ctx, 16);
#endif #endif
CLIParserFree(ctx); CLIParserFree(ctx);
@ -1951,6 +1957,12 @@ static int CmdHF14AMfNestedHard(const char *Cmd) {
if (im) if (im)
SetSIMDInstr(SIMD_MMX); SetSIMDInstr(SIMD_MMX);
#endif #endif
#if defined(COMPILER_HAS_SIMD_NEON)
if (ie)
SetSIMDInstr(SIMD_NEON);
#endif
if (in) if (in)
SetSIMDInstr(SIMD_NONE); SetSIMDInstr(SIMD_NONE);
@ -2067,6 +2079,9 @@ static int CmdHF14AMfAutoPWN(const char *Cmd) {
#endif #endif
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
arg_lit0(NULL, "i5", "AVX512"), arg_lit0(NULL, "i5", "AVX512"),
#endif
#if defined(COMPILER_HAS_SIMD_NEON)
arg_lit0(NULL, "ie", "NEON"),
#endif #endif
arg_param_end arg_param_end
}; };
@ -2118,6 +2133,10 @@ static int CmdHF14AMfAutoPWN(const char *Cmd) {
#if defined(COMPILER_HAS_SIMD_AVX512) #if defined(COMPILER_HAS_SIMD_AVX512)
bool i5 = arg_get_lit(ctx, 18); bool i5 = arg_get_lit(ctx, 18);
#endif #endif
#if defined(COMPILER_HAS_SIMD_NEON)
bool ie = arg_get_lit(ctx, 14);
#endif
CLIParserFree(ctx); CLIParserFree(ctx);
//validations //validations
@ -2167,6 +2186,12 @@ static int CmdHF14AMfAutoPWN(const char *Cmd) {
if (im) if (im)
SetSIMDInstr(SIMD_MMX); SetSIMDInstr(SIMD_MMX);
#endif #endif
#if defined(COMPILER_HAS_SIMD_NEON)
if (ie)
SetSIMDInstr(SIMD_NEON);
#endif
if (in) if (in)
SetSIMDInstr(SIMD_NONE); SetSIMDInstr(SIMD_NONE);

View file

@ -101,6 +101,11 @@ static void get_SIMD_instruction_set(char *instruction_set) {
case SIMD_MMX: case SIMD_MMX:
strcpy(instruction_set, "MMX"); strcpy(instruction_set, "MMX");
break; break;
#endif
#if defined(COMPILER_HAS_SIMD_NEON)
case SIMD_NEON:
strcpy(instruction_set, "NEON");
break;
#endif #endif
case SIMD_AUTO: case SIMD_AUTO:
case SIMD_NONE: case SIMD_NONE:

View file

@ -3643,9 +3643,10 @@
"--is sse2", "--is sse2",
"--ia avx", "--ia avx",
"--i2 avx2", "--i2 avx2",
"--i5 avx512" "--i5 avx512",
"--ie neon"
], ],
"usage": "hf mf autopwn [-habslv] [-k <hex>] [-s <dec>] [-f <fn>] [--mini] [--1k] [--2k] [--4k] [--in] [--im] [--is] [--ia] [--i2] [--i5]" "usage": "hf mf autopwn [-habslv] [-k <hex>] [-s <dec>] [-f <fn>] [--mini] [--1k] [--2k] [--4k] [--in] [--im] [--is] [--ia] [--i2] [--i5] [--ie]"
}, },
"hf mf cgetblk": { "hf mf cgetblk": {
"command": "hf mf cgetblk", "command": "hf mf cgetblk",
@ -4122,9 +4123,10 @@
"--is sse2", "--is sse2",
"--ia avx", "--ia avx",
"--i2 avx2", "--i2 avx2",
"--i5 avx512" "--i5 avx512",
"--ie neon"
], ],
"usage": "hf mf hardnested [-habrstw] [-k <hex>] [--blk <dec>] [--tblk <dec>] [--ta] [--tb] [--tk <hex>] [-u <hex>] [-f <fn>] [--in] [--im] [--is] [--ia] [--i2] [--i5]" "usage": "hf mf hardnested [-habrstw] [-k <hex>] [--blk <dec>] [--tblk <dec>] [--ta] [--tb] [--tk <hex>] [-u <hex>] [-f <fn>] [--in] [--im] [--is] [--ia] [--i2] [--i5] [--ie]"
}, },
"hf mf help": { "hf mf help": {
"command": "hf mf help", "command": "hf mf help",