tools/bench/mpfrbench.c

/* mpfrbench.c  -- compute the timings for the MPFRbench benchmark

Copyright 1999, 2001-2023 Free Software Foundation, Inc.
Contributed by the AriC and Caramba projects, INRIA.

This file is part of the GNU MPFR Library.

The GNU MPFR Library is free software; you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation; either version 3 of the License, or (at your
option) any later version.

The GNU MPFR Library is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
License for more details.

You should have received a copy of the GNU Lesser General Public License
along with the GNU MPFR Library; see the file COPYING.LESSER.  If not, see
https://www.gnu.org/licenses/ or write to the Free Software Foundation, Inc.,
51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA. */

#include <stdlib.h>
#include <stdio.h>
#ifdef HAVE_GETRUSAGE
#include <sys/time.h>
#include <sys/resource.h>
#else
#include <time.h>
#endif
#include "mpfr.h"
#include "benchtime.h"

static unsigned long get_cputime (void);

/* enumeration of the group of functions */
enum egroupfunc
{
  egroup_arith = 0,             /* e.g., arith ... */
  egroup_special,               /* e.g., cos, ... */
  egroup_last                   /* to get the number of enum */
};

/* name of the group of functions */
const char *groupname [] = {
  "Arith  ",
  "Special"
};


struct benchfunc
{
  const char *name;             /* name of the function */
  double (*func_init) (int n, mpfr_t * z, mpfr_t * x, mpfr_t * y); /* compute the time for one call (not accurate) */
  unsigned long int (*func_accurate) (unsigned long int niter, int n, mpfr_t * z, mpfr_t * x, mpfr_t * y, int nop); /* compute the time for "niter" calls (accurate) */
  enum egroupfunc group;        /* group of the function */
  int  noperands;               /* number of operands */
};


/* declare the function to compute the cost for one call of the function */
DECLARE_TIME_2OP (mpfr_mul)
DECLARE_TIME_2OP (mpfr_add)
DECLARE_TIME_2OP (mpfr_sub)
DECLARE_TIME_2OP (mpfr_div)
DECLARE_TIME_1OP (mpfr_sqrt)
DECLARE_TIME_1OP (mpfr_exp)
DECLARE_TIME_1OP (mpfr_log)
DECLARE_TIME_1OP (mpfr_sin)
DECLARE_TIME_1OP (mpfr_cos)
DECLARE_TIME_1OP (mpfr_asin)
DECLARE_TIME_1OP (mpfr_acos)

/* number of operations to score */
#define NB_BENCH_OP 11
/* number of random numbers */
#define NB_RAND_FLOAT 10000

/* list of functions to compute the score */
const struct benchfunc arrayfunc[NB_BENCH_OP] = {
  {"mul", ADDR_TIME_NOP (mpfr_mul), ADDR_ACCURATE_TIME_NOP (mpfr_mul), egroup_arith, 2},
  {"add", ADDR_TIME_NOP (mpfr_add), ADDR_ACCURATE_TIME_NOP (mpfr_add), egroup_arith, 2},
  {"sub", ADDR_TIME_NOP (mpfr_sub), ADDR_ACCURATE_TIME_NOP (mpfr_sub), egroup_arith, 2},
  {"div", ADDR_TIME_NOP (mpfr_div), ADDR_ACCURATE_TIME_NOP (mpfr_div), egroup_arith, 2},
  {"sqrt", ADDR_TIME_NOP (mpfr_sqrt), ADDR_ACCURATE_TIME_NOP (mpfr_sqrt), egroup_special, 1},
  {"exp", ADDR_TIME_NOP (mpfr_exp), ADDR_ACCURATE_TIME_NOP (mpfr_exp), egroup_special, 1},
  {"log", ADDR_TIME_NOP (mpfr_log), ADDR_ACCURATE_TIME_NOP (mpfr_log), egroup_special, 1},
  {"cos", ADDR_TIME_NOP (mpfr_cos), ADDR_ACCURATE_TIME_NOP (mpfr_cos), egroup_special, 1},
  {"sin", ADDR_TIME_NOP (mpfr_sin), ADDR_ACCURATE_TIME_NOP (mpfr_sin), egroup_special, 1},
  {"acos", ADDR_TIME_NOP (mpfr_acos), ADDR_ACCURATE_TIME_NOP (mpfr_acos), egroup_special, 1},
  {"asin", ADDR_TIME_NOP (mpfr_asin), ADDR_ACCURATE_TIME_NOP (mpfr_asin), egroup_special, 1}
};

/* the following arrays must have the same number of elements */

/* list of precisions to test for the first operand */
const int arrayprecision_op1[] =
  { 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384,
    50, 100, 200, 350, 700, 1500, 3000, 6000, 10000, 1500, 3000, 5000,
  };

/* list of precisions to test for the second operand */
const int arrayprecision_op2[] =
  { 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384,
    50, 100, 200, 350, 700, 1500, 3000, 6000, 10000, 3000, 6000, 10000
  };

/* get the time in microseconds */
static unsigned long
get_cputime (void)
{
#ifdef HAVE_GETRUSAGE
  struct rusage ru;

  getrusage (RUSAGE_SELF, &ru);
  return ru.ru_utime.tv_sec * 1000000 + ru.ru_utime.tv_usec
       + ru.ru_stime.tv_sec * 1000000 + ru.ru_stime.tv_usec;
#else
  return (unsigned long) ((double) clock () / ((double) CLOCKS_PER_SEC / 1e6));
#endif
}

/* initialize an array of n random numbers */
static mpfr_t *
bench_random_array (int n, mpfr_prec_t precision, gmp_randstate_t randstate)
{
  int j;
  mpfr_t *ptr;

  ptr = (mpfr_t *) malloc (n * sizeof (mpfr_t));
  if (ptr == NULL)
    {
      printf ("Can't allocate memory for %d numbers\n", n);
      exit (1);
      return NULL;
    }
  for (j = 0; j < n; j++)
    {
      mpfr_init2 (ptr[j], precision);
      mpfr_urandomb (ptr[j], randstate);
    }
  return ptr;
}

/* compute the score for the operation arrayfunc[op] */
static void
compute_score (mpz_t zscore, int op, gmp_randstate_t randstate)
{
  mpfr_t *xptr, *yptr, *zptr;
  int i, j;
  size_t k;
  unsigned long niter, ti;
  double t;
  unsigned long ops_per_sec;
  int countprec = 0;

  mpz_init_set_si (zscore, 1);

  i = op;
  for (k = 0; k < (int) sizeof (arrayprecision_op1) / sizeof (arrayprecision_op1[0]);
       k++, countprec++)
    {
      mpfr_prec_t precision1 = arrayprecision_op1[k];
      mpfr_prec_t precision2 = arrayprecision_op2[k];
      mpfr_prec_t precision3 = arrayprecision_op2[k];

      /* allocate array of random numbers */
      xptr = bench_random_array (NB_RAND_FLOAT, precision1, randstate);
      yptr = bench_random_array (NB_RAND_FLOAT, precision2, randstate);
      zptr = bench_random_array (NB_RAND_FLOAT, precision3, randstate);

      /* compute the number of operations per second */
      if (arrayfunc[i].noperands==2)
        {
          printf ("operation %5s, precision : %5lux%5lu to %5lu bits ... ", arrayfunc[i].name, precision1, precision2, precision3);
        }
      else
        {
          printf ("operation %5s, precision :       %5lu to %5lu bits ... ", arrayfunc[i].name, precision1, precision3);
        }
      fflush (stdout);

      t = arrayfunc[i].func_init (NB_RAND_FLOAT, zptr, xptr, yptr);
      niter = 1 + (unsigned long) (1e6 / t);

      printf (" %10lu iterations ...", niter);
      fflush (stdout);

      /* ti expressed in microseconds */
      ti = arrayfunc[i].func_accurate (niter, NB_RAND_FLOAT, zptr, xptr, yptr, arrayfunc[i].noperands);

      ops_per_sec = (unsigned long) (1000000E0 * niter / (double) ti);

      printf (" %10lu operations per second\n", ops_per_sec);

      mpz_mul_ui (zscore, zscore, ops_per_sec);

      /* free memory */
      for (j = 0; j < NB_RAND_FLOAT; j++)
        {
          mpfr_clear (xptr[j]);
          mpfr_clear (yptr[j]);
          mpfr_clear (zptr[j]);
        }
      free (xptr);
      free (yptr);
      free (zptr);
    }

  mpz_root (zscore, zscore, countprec);
}

/* compute the score for all groups */
static void
compute_groupscore (mpz_t groupscore[], int countop, mpz_t zscore[])
{
  int op;
  enum egroupfunc group;
  int countgroupop;

  for (group = (enum egroupfunc)0; group != egroup_last; group++)
    {
      mpz_init_set_si (groupscore[group], 1);
      for (op = 0, countgroupop = 0; op < countop; op++)
        {
          if (group == arrayfunc[op].group)
            {
              mpz_mul (groupscore[group], groupscore[group], zscore[op]);
              countgroupop++;
            }
        }
      mpz_root (groupscore[group], groupscore[group], countgroupop);
    }
}


/* compute the global score */
static void
compute_globalscore (mpz_t globalscore, int countop, mpz_t zscore[])
{
  int op;

  mpz_init_set_si (globalscore, 1);
  for (op = 0; op < countop; op++)
    {
      mpz_mul (globalscore, globalscore, zscore[op]);
    }
  mpz_root (globalscore, globalscore, countop);
}

int
main (void)
{
  int i;
  enum egroupfunc group;
  mpz_t score[NB_BENCH_OP];
  mpz_t globalscore, groupscore[egroup_last];
  gmp_randstate_t randstate;

  gmp_randinit_default (randstate);

  for (i = 0; i < NB_BENCH_OP; i++)
    {
      compute_score (score[i], i, randstate);
    }
  compute_globalscore (globalscore, NB_BENCH_OP, score);
  compute_groupscore (groupscore, NB_BENCH_OP, score);

  printf ("\n=================================================================\n\n");
  printf ("GMP : %s  MPFR : %s \n", gmp_version, mpfr_get_version ());
#ifdef __GMP_CC
  printf ("GMP compiler : %s\n", __GMP_CC);
#endif
#ifdef __GMP_CFLAGS
  printf ("GMP flags    : %s\n", __GMP_CFLAGS);
#endif
  printf ("\n\n");

  for (i = 0; i < NB_BENCH_OP; i++)
    {
      gmp_printf ("\tscore for %5s : %12Zd\n", arrayfunc[i].name, score[i]);
      if (i == NB_BENCH_OP-1 || arrayfunc[i+1].group != arrayfunc[i].group)
        {
          enum egroupfunc g = arrayfunc[i].group;
          gmp_printf ("group score %s : %12Zd\n\n", groupname[g], groupscore[g]);
        }
    }
  /* divide by 132 the global score to get about 10^3 on a
     Intel(R) Core(TM)2 Quad CPU    Q9550  @ 2.83GHz
     with GMP : 5.1.3  MPFR : 3.1.2
     GMP compiler: gcc -std=gnu99, GMP flags: -O2 -pedantic
     -fomit-frame-pointer -m64 -mtune=core2 -march=core2 */
  mpz_div_ui (globalscore, globalscore, 132);
  gmp_printf ("global score : %12Zd\n\n", globalscore);

  for (i = 0; i < NB_BENCH_OP; i++)
    {
      mpz_clear (score[i]);
    }

  for (group = (enum egroupfunc)0; group != egroup_last; group++)
    {
      mpz_clear (groupscore[group]);
    }
  mpz_clear (globalscore);
  gmp_randclear (randstate);
  return 0;
}