#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
+#ifndef __APPLE__
#include <malloc.h>
+#endif
// this needs to be compiled several times for each instruction set.
// For each instruction set, define a dedicated function name:
inline uint32_t *MALLOC_BITARRAY(uint32_t x)
{
-#ifdef _WIN32
+#if defined (_WIN32)
return __builtin_assume_aligned(_aligned_malloc((x), __BIGGEST_ALIGNMENT__), __BIGGEST_ALIGNMENT__);
+#elif defined (__APPLE__)
+ uint32_t *allocated_memory;
+ if (posix_memalign((void**)&allocated_memory, __BIGGEST_ALIGNMENT__, x)) {
+ return NULL;
+ } else {
+ return __builtin_assume_aligned(allocated_memory, __BIGGEST_ALIGNMENT__);
+ }
#else
return __builtin_assume_aligned(memalign(__BIGGEST_ALIGNMENT__, (x)), __BIGGEST_ALIGNMENT__);
#endif
// determine the available instruction set at runtime and call the correct function
uint32_t *malloc_bitarray_dispatch(uint32_t x) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) malloc_bitarray_function_p = &malloc_bitarray_AVX512;
else if (__builtin_cpu_supports("avx2")) malloc_bitarray_function_p = &malloc_bitarray_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) malloc_bitarray_function_p = &malloc_bitarray_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) malloc_bitarray_function_p = &malloc_bitarray_AVX;
else if (__builtin_cpu_supports("sse2")) malloc_bitarray_function_p = &malloc_bitarray_SSE2;
else if (__builtin_cpu_supports("mmx")) malloc_bitarray_function_p = &malloc_bitarray_MMX;
- else
+ else
+ #endif
#endif
malloc_bitarray_function_p = &malloc_bitarray_NOSIMD;
}
void free_bitarray_dispatch(uint32_t *x) {
-#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+#if defined (__i386__) || defined (__x86_64__)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) free_bitarray_function_p = &free_bitarray_AVX512;
else if (__builtin_cpu_supports("avx2")) free_bitarray_function_p = &free_bitarray_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) free_bitarray_function_p = &free_bitarray_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) free_bitarray_function_p = &free_bitarray_AVX;
else if (__builtin_cpu_supports("sse2")) free_bitarray_function_p = &free_bitarray_SSE2;
else if (__builtin_cpu_supports("mmx")) free_bitarray_function_p = &free_bitarray_MMX;
- else
+ else
+ #endif
#endif
free_bitarray_function_p = &free_bitarray_NOSIMD;
uint32_t bitcount_dispatch(uint32_t a) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) bitcount_function_p = &bitcount_AVX512;
else if (__builtin_cpu_supports("avx2")) bitcount_function_p = &bitcount_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) bitcount_function_p = &bitcount_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) bitcount_function_p = &bitcount_AVX;
else if (__builtin_cpu_supports("sse2")) bitcount_function_p = &bitcount_SSE2;
else if (__builtin_cpu_supports("mmx")) bitcount_function_p = &bitcount_MMX;
- else
+ else
+ #endif
#endif
bitcount_function_p = &bitcount_NOSIMD;
uint32_t count_states_dispatch(uint32_t *bitarray) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) count_states_function_p = &count_states_AVX512;
else if (__builtin_cpu_supports("avx2")) count_states_function_p = &count_states_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) count_states_function_p = &count_states_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) count_states_function_p = &count_states_AVX;
else if (__builtin_cpu_supports("sse2")) count_states_function_p = &count_states_SSE2;
else if (__builtin_cpu_supports("mmx")) count_states_function_p = &count_states_MMX;
- else
+ else
+ #endif
#endif
count_states_function_p = &count_states_NOSIMD;
void bitarray_AND_dispatch(uint32_t *A, uint32_t *B) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) bitarray_AND_function_p = &bitarray_AND_AVX512;
else if (__builtin_cpu_supports("avx2")) bitarray_AND_function_p = &bitarray_AND_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) bitarray_AND_function_p = &bitarray_AND_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) bitarray_AND_function_p = &bitarray_AND_AVX;
else if (__builtin_cpu_supports("sse2")) bitarray_AND_function_p = &bitarray_AND_SSE2;
else if (__builtin_cpu_supports("mmx")) bitarray_AND_function_p = &bitarray_AND_MMX;
- else
+ else
+ #endif
#endif
bitarray_AND_function_p = &bitarray_AND_NOSIMD;
}
void bitarray_low20_AND_dispatch(uint32_t *A, uint32_t *B) {
-#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+#if defined (__i386__) || defined (__x86_64__)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) bitarray_low20_AND_function_p = &bitarray_low20_AND_AVX512;
else if (__builtin_cpu_supports("avx2")) bitarray_low20_AND_function_p = &bitarray_low20_AND_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) bitarray_low20_AND_function_p = &bitarray_low20_AND_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) bitarray_low20_AND_function_p = &bitarray_low20_AND_AVX;
else if (__builtin_cpu_supports("sse2")) bitarray_low20_AND_function_p = &bitarray_low20_AND_SSE2;
else if (__builtin_cpu_supports("mmx")) bitarray_low20_AND_function_p = &bitarray_low20_AND_MMX;
- else
+ else
+ #endif
#endif
bitarray_low20_AND_function_p = &bitarray_low20_AND_NOSIMD;
uint32_t count_bitarray_AND_dispatch(uint32_t *A, uint32_t *B) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) count_bitarray_AND_function_p = &count_bitarray_AND_AVX512;
else if (__builtin_cpu_supports("avx2")) count_bitarray_AND_function_p = &count_bitarray_AND_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) count_bitarray_AND_function_p = &count_bitarray_AND_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) count_bitarray_AND_function_p = &count_bitarray_AND_AVX;
else if (__builtin_cpu_supports("sse2")) count_bitarray_AND_function_p = &count_bitarray_AND_SSE2;
else if (__builtin_cpu_supports("mmx")) count_bitarray_AND_function_p = &count_bitarray_AND_MMX;
- else
+ else
+ #endif
#endif
count_bitarray_AND_function_p = &count_bitarray_AND_NOSIMD;
uint32_t count_bitarray_low20_AND_dispatch(uint32_t *A, uint32_t *B) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_AVX512;
else if (__builtin_cpu_supports("avx2")) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_AVX;
else if (__builtin_cpu_supports("sse2")) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_SSE2;
else if (__builtin_cpu_supports("mmx")) count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_MMX;
- else
+ else
+ #endif
#endif
count_bitarray_low20_AND_function_p = &count_bitarray_low20_AND_NOSIMD;
void bitarray_AND4_dispatch(uint32_t *A, uint32_t *B, uint32_t *C, uint32_t *D) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) bitarray_AND4_function_p = &bitarray_AND4_AVX512;
else if (__builtin_cpu_supports("avx2")) bitarray_AND4_function_p = &bitarray_AND4_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) bitarray_AND4_function_p = &bitarray_AND4_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) bitarray_AND4_function_p = &bitarray_AND4_AVX;
else if (__builtin_cpu_supports("sse2")) bitarray_AND4_function_p = &bitarray_AND4_SSE2;
else if (__builtin_cpu_supports("mmx")) bitarray_AND4_function_p = &bitarray_AND4_MMX;
- else
+ else
+ #endif
#endif
bitarray_AND4_function_p = &bitarray_AND4_NOSIMD;
void bitarray_OR_dispatch(uint32_t *A, uint32_t *B) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) bitarray_OR_function_p = &bitarray_OR_AVX512;
else if (__builtin_cpu_supports("avx2")) bitarray_OR_function_p = &bitarray_OR_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) bitarray_OR_function_p = &bitarray_OR_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) bitarray_OR_function_p = &bitarray_OR_AVX;
else if (__builtin_cpu_supports("sse2")) bitarray_OR_function_p = &bitarray_OR_SSE2;
else if (__builtin_cpu_supports("mmx")) bitarray_OR_function_p = &bitarray_OR_MMX;
- else
+ else
+ #endif
#endif
bitarray_OR_function_p = &bitarray_OR_NOSIMD;
uint32_t count_bitarray_AND2_dispatch(uint32_t *A, uint32_t *B) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) count_bitarray_AND2_function_p = &count_bitarray_AND2_AVX512;
else if (__builtin_cpu_supports("avx2")) count_bitarray_AND2_function_p = &count_bitarray_AND2_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) count_bitarray_AND2_function_p = &count_bitarray_AND2_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) count_bitarray_AND2_function_p = &count_bitarray_AND2_AVX;
else if (__builtin_cpu_supports("sse2")) count_bitarray_AND2_function_p = &count_bitarray_AND2_SSE2;
else if (__builtin_cpu_supports("mmx")) count_bitarray_AND2_function_p = &count_bitarray_AND2_MMX;
- else
+ else
+ #endif
#endif
count_bitarray_AND2_function_p = &count_bitarray_AND2_NOSIMD;
uint32_t count_bitarray_AND3_dispatch(uint32_t *A, uint32_t *B, uint32_t *C) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) count_bitarray_AND3_function_p = &count_bitarray_AND3_AVX512;
else if (__builtin_cpu_supports("avx2")) count_bitarray_AND3_function_p = &count_bitarray_AND3_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) count_bitarray_AND3_function_p = &count_bitarray_AND3_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) count_bitarray_AND3_function_p = &count_bitarray_AND3_AVX;
else if (__builtin_cpu_supports("sse2")) count_bitarray_AND3_function_p = &count_bitarray_AND3_SSE2;
else if (__builtin_cpu_supports("mmx")) count_bitarray_AND3_function_p = &count_bitarray_AND3_MMX;
- else
+ else
+ #endif
#endif
count_bitarray_AND3_function_p = &count_bitarray_AND3_NOSIMD;
uint32_t count_bitarray_AND4_dispatch(uint32_t *A, uint32_t *B, uint32_t *C, uint32_t *D) {
#if defined (__i386__) || defined (__x86_64__)
- #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
+ #if !defined(__APPLE__) || (defined(__APPLE__) && (__clang_major__ > 8))
+ #if (__GNUC__ >= 5) && (__GNUC__ > 5 || __GNUC_MINOR__ > 2)
if (__builtin_cpu_supports("avx512f")) count_bitarray_AND4_function_p = &count_bitarray_AND4_AVX512;
else if (__builtin_cpu_supports("avx2")) count_bitarray_AND4_function_p = &count_bitarray_AND4_AVX2;
- #else
+ #else
if (__builtin_cpu_supports("avx2")) count_bitarray_AND4_function_p = &count_bitarray_AND4_AVX2;
- #endif
+ #endif
else if (__builtin_cpu_supports("avx")) count_bitarray_AND4_function_p = &count_bitarray_AND4_AVX;
else if (__builtin_cpu_supports("sse2")) count_bitarray_AND4_function_p = &count_bitarray_AND4_SSE2;
else if (__builtin_cpu_supports("mmx")) count_bitarray_AND4_function_p = &count_bitarray_AND4_MMX;
- else
+ else
+ #endif
#endif
count_bitarray_AND4_function_p = &count_bitarray_AND4_NOSIMD;