mpf_t init benchmark

mpf_vector_bench.cpp

#include <iostream>
#include <gmpxx.h>
#include <cstdint>
#include <chrono>
#include <vector>
#include <stdlib.h>

#define PREC 64*4
#define PREC_LIMB_SIZE 4

const size_t N = 100;
const size_t sample = 156250;

void bench1() {
    //mpf_t a[sample], b[sample], c[sample];
    //mp_limb_t a_mp_d[(PREC_LIMB_SIZE + 1) * sample];
    //mp_limb_t b_mp_d[(PREC_LIMB_SIZE + 1) * sample];
    //mp_limb_t c_mp_d[(PREC_LIMB_SIZE + 1) * sample];

    mpf_t *a = (mpf_t *)malloc(sizeof(mpf_t) * sample);
    mpf_t *b = (mpf_t *)malloc(sizeof(mpf_t) * sample);
    mpf_t *c = (mpf_t *)malloc(sizeof(mpf_t) * sample);

    mp_limb_t *a_mp_d = (mp_limb_t *)malloc(sizeof(mp_limb_t)*
            (PREC_LIMB_SIZE + 1) * sample);
    mp_limb_t *b_mp_d = (mp_limb_t *)malloc(sizeof(mp_limb_t)*
            (PREC_LIMB_SIZE + 1) * sample);
    mp_limb_t *c_mp_d = (mp_limb_t *)malloc(sizeof(mp_limb_t)*
            (PREC_LIMB_SIZE + 1) * sample);

    for (size_t i = 0; i < sample; i++) {
        a[i]->_mp_d = a_mp_d + (PREC_LIMB_SIZE + 1) * i;
        b[i]->_mp_d = b_mp_d + (PREC_LIMB_SIZE + 1) * i;
        c[i]->_mp_d = c_mp_d + (PREC_LIMB_SIZE + 1) * i;

        mpf_set_d(a[i], i * 81.01919);
        mpf_set_d(b[i], i * 11.4514);
        //mpf_set_d(c[i], 0);

        a[i]->_mp_prec = PREC_LIMB_SIZE;
        b[i]->_mp_prec = PREC_LIMB_SIZE;
        c[i]->_mp_prec = PREC_LIMB_SIZE;
    }

    for (size_t i = 0; i < sample; i++) {
        mpf_mul(c[i], a[i], b[i]);
    }
    //gmp_printf("%.16Ff\n", a[1]);
    //gmp_printf("%.16Ff\n", b[1]);
    //gmp_printf("%.16Ff\n", c[1]);

    free(a);
    free(b);
    free(c);
    free(a_mp_d);
    free(b_mp_d);
    free(c_mp_d);
}

void bench2() {
    //mpf_t a[sample], b[sample], c[sample];
    mpf_t *a = (mpf_t *)malloc(sizeof(mpf_t) * sample);
    mpf_t *b = (mpf_t *)malloc(sizeof(mpf_t) * sample);
    mpf_t *c = (mpf_t *)malloc(sizeof(mpf_t) * sample);

    for (size_t i = 0; i < sample; i++) {
        mpf_init2(a[i], PREC);
        mpf_init2(b[i], PREC);
        mpf_init2(c[i], PREC);

        mpf_set_d(a[i], i * 81.01919);
        mpf_set_d(b[i], i * 11.4514);
    }

    for (size_t i = 0; i < sample; i++) {
        mpf_mul(c[i], a[i], b[i]);
    }

    //gmp_printf("%.16Ff\n", a[1]);
    //gmp_printf("%.16Ff\n", b[1]);
    //gmp_printf("%.16Ff\n", c[1]);
    for (size_t i = 0; i < sample; i++) {
        mpf_clear(a[i]);
        mpf_clear(b[i]);
        mpf_clear(c[i]);
    }

    free(a);
    free(b);
    free(c);
}

void bench3(size_t n) {
    //mpf_t a[sample], b[sample], c[sample];
    //mp_limb_t a_mp_d[(PREC_LIMB_SIZE + 1) * sample];
    //mp_limb_t b_mp_d[(PREC_LIMB_SIZE + 1) * sample];
    //mp_limb_t c_mp_d[(PREC_LIMB_SIZE + 1) * sample];

    mpf_t *a = (mpf_t *)malloc(sizeof(mpf_t) * n);
    mpf_t *b = (mpf_t *)malloc(sizeof(mpf_t) * n);
    mpf_t *c = (mpf_t *)malloc(sizeof(mpf_t) * n);

    mp_limb_t *a_mp_d = (mp_limb_t *)malloc(sizeof(mp_limb_t)*
            (PREC_LIMB_SIZE + 1) * n);
    mp_limb_t *b_mp_d = (mp_limb_t *)malloc(sizeof(mp_limb_t)*
            (PREC_LIMB_SIZE + 1) * n);
    mp_limb_t *c_mp_d = (mp_limb_t *)malloc(sizeof(mp_limb_t)*
            (PREC_LIMB_SIZE + 1) * n);

    for (size_t i = 0; i < sample; i++) {
        a[i]->_mp_d = a_mp_d + (PREC_LIMB_SIZE + 1) * i;
        b[i]->_mp_d = b_mp_d + (PREC_LIMB_SIZE + 1) * i;
        c[i]->_mp_d = c_mp_d + (PREC_LIMB_SIZE + 1) * i;

        mpf_set_d(a[i], i * 81.01919);
        mpf_set_d(b[i], i * 11.4514);
        //mpf_set_d(c[i], 0);

        a[i]->_mp_prec = PREC_LIMB_SIZE;
        b[i]->_mp_prec = PREC_LIMB_SIZE;
        c[i]->_mp_prec = PREC_LIMB_SIZE;
    }

    for (size_t i = 0; i < sample; i++) {
        mpf_mul(c[i], a[i], b[i]);
    }
    //gmp_printf("%.16Ff\n", a[1]);
    //gmp_printf("%.16Ff\n", b[1]);
    //gmp_printf("%.16Ff\n", c[1]);

    free(a);
    free(b);
    free(c);
    free(a_mp_d);
    free(b_mp_d);
    free(c_mp_d);
}

void bench4() {
    //mpf_class a[sample], b[sample], c[sample];
    std::vector<mpf_class> a(sample);
    std::vector<mpf_class> b(sample);
    std::vector<mpf_class> c(sample);

    for (size_t i = 0; i < sample; i++) {
        a[i] = i * 81.01919;
        b[i] = i * 11.4514;
        //mpf_set_d(a[i], i * 81.01919);
        //mpf_set_d(b[i], i * 11.4514);
    }

    for (size_t i = 0; i < sample; i++) {
        mpf_mul(c[i].get_mpf_t(), a[i].get_mpf_t(), b[i].get_mpf_t());
    }

    //gmp_printf("%.16Ff\n", a[1]);
    //gmp_printf("%.16Ff\n", b[1]);
    //gmp_printf("%.16Ff\n", c[1]);
}


int main() {
    auto start_time = std::chrono::high_resolution_clock::now();
    for (size_t i = 0; i < N; i++) {
        bench1();
    }
    auto end_time = std::chrono::high_resolution_clock::now();
    auto duration = std::chrono::duration_cast<std::chrono::microseconds>(end_time - start_time);
    std::cerr << "\tExecution Time: " << (duration.count() / double(N) / 1.0e6)
<< " seconds" << std::endl;

    start_time = std::chrono::high_resolution_clock::now();
    for (size_t i = 0; i < N; i++) {
        bench2();
    }
    end_time = std::chrono::high_resolution_clock::now();
    duration = std::chrono::duration_cast<std::chrono::microseconds>(end_time -
start_time);
    std::cerr << "\tExecution Time: " << (duration.count() / double(N) / 1.0e6)
<< " seconds" << std::endl;


    start_time = std::chrono::high_resolution_clock::now();
    for (size_t i = 0; i < N; i++) {
        bench3(sample);
    }
    end_time = std::chrono::high_resolution_clock::now();
    duration = std::chrono::duration_cast<std::chrono::microseconds>(end_time -
start_time);
    std::cerr << "\tExecution Time: " << (duration.count() / double(N) / 1.0e6)
<< " seconds" << std::endl;

    mpf_set_default_prec(PREC);
    start_time = std::chrono::high_resolution_clock::now();
    for (size_t i = 0; i < N; i++) {
        bench4();
    }
    end_time = std::chrono::high_resolution_clock::now();
    duration = std::chrono::duration_cast<std::chrono::microseconds>(end_time -
start_time);
    std::cerr << "\tExecution Time: " << (duration.count() / double(N) / 1.0e6)
<< " seconds" << std::endl;

}

result.log

ykm11: $ g++ piyo.cpp -lgmpxx -lgmp -O3
ykm11: $ ./a.out
        Execution Time: 0.0101858 seconds
        Execution Time: 0.0294115 seconds
        Execution Time: 0.00777957 seconds
        Execution Time: 0.0221703 seconds