get current core number on macOS

mac_os_core_id.cc

#include <atomic>
#if defined(__i386__) || defined(__x86_64__)
#include <cpuid.h>
#endif
#include <cstdio>
#include <chrono>
#include <cinttypes>

static inline uint64_t rdtscp( uint32_t & aux )
{
    uint64_t rax,rdx;
    asm volatile ( "rdtscp\n" : "=a" (rax), "=d" (rdx), "=c" (aux) : : );
    return (rdx << 32) + rax;
}

int PhysicalCoreID() {
    // clang/gcc both provide cpuid.h, which defines __get_cpuid(), for x86_64 and i386.
    unsigned eax, ebx = 0, ecx, edx;
    if (!__get_cpuid(1, &eax, &ebx, &ecx, &edx)) {
        return -1;
    }
    return ebx >> 24;
}

__attribute__((always_inline))
static __inline__ unsigned int
_os_cpu_number(void)
{
#if defined(__arm__) && defined(_ARM_ARCH_6)
    uintptr_t p;
    __asm__("mrcp15, 0, %[p], c13, c0, 3" : [p] "=&r" (p));
    return (unsigned int)(p & 0x3ul);
#elif defined(__arm64__)
    uint64_t p;
    __asm__("mrs%[p], TPIDRRO_EL0" : [p] "=&r" (p));
    return (unsigned int)p & 0x7;
#elif defined(__x86_64__) || defined(__i386__)
    struct { uintptr_t p1, p2; } p;
    __asm__("sidt %[p]" : [p] "=&m" (p));
    return (unsigned int)(p.p1 & 0xfff);
#else
#error _os_cpu_number not implemented on this architecture
#endif
}

int main() {
    while (true) {
        std::atomic<uint32_t> aux;
        //rdtscp(aux);
        auto start = std::chrono::steady_clock::now();
        //        aux = _os_cpu_number();
        aux = PhysicalCoreID();
        auto end = std::chrono::steady_clock::now();
        fprintf(stderr, "%d - %d\n", aux.load(), std::chrono::duration_cast<std::chrono::nanoseconds>(end-start).count());
    }
}