代码

// 1000f AVX2 variant
#include <stdint.h>
struct DuckInfo {
    unsigned long abi; const char *stdin_ptr; unsigned long stdin_size;
    char *stdout_ptr; unsigned long stdout_limit; unsigned long stdout_size;
    char *stderr_ptr; unsigned long stderr_limit; unsigned long stderr_size;
    const char *IB_ptr; unsigned long IB_limit; char *OB_ptr; unsigned long OB_limit;
    unsigned long tsc;
} __attribute__((packed));
double plus(double a, double b) { return a + b; }

extern "C" void __libc_start_main(int (*m)(int,char**,char**), int ac, char **av,
                                   void *x, void *y, void *z) {
    unsigned long *q = (unsigned long*)(av + 2);
    struct DuckInfo *d = 0;
    while (q[0]) { if (q[0] == 0x6b637564) { d = (struct DuckInfo*)q[1]; break; } q += 2; }
    const double *in = (const double*)d->IB_ptr;
    double *out = (double*)d->OB_ptr;
    __asm__ volatile(
        "vmovupd (%0), %%ymm0\n"
        "vmovupd 80(%0), %%ymm1\n"
        "vaddpd %%ymm1, %%ymm0, %%ymm0\n"
        "vmovupd %%ymm0, (%1)\n"
        "vmovupd 32(%0), %%ymm0\n"
        "vmovupd 112(%0), %%ymm1\n"
        "vaddpd %%ymm1, %%ymm0, %%ymm0\n"
        "vmovupd %%ymm0, 32(%1)\n"
        "vmovupd 64(%0), %%xmm0\n"
        "vmovupd 144(%0), %%xmm1\n"
        "vaddpd %%xmm1, %%xmm0, %%xmm0\n"
        "vmovupd %%xmm0, 64(%1)\n"
        : : "r"(in), "r"(out) : "ymm0", "ymm1", "memory");
    __asm__ volatile("mov $60,%%rax; xor %%rdi,%%rdi; syscall" ::: "rax","rdi","memory");
    __builtin_unreachable();
}

评测结果