HelenOS sources

root/uspace/app/tester/float/float2.c

/* [<][>][^][v][top][bottom][index][help] */

DEFINITIONS

This source file includes following definitions.
  1. cmp_float
  2. cmp_double
  3. test_float2

/*
 * Copyright (c) 2014 Martin Decky
 * Copyright (c) 2015 Jiri Svoboda
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * - Redistributions of source code must retain the above copyright
 *   notice, this list of conditions and the following disclaimer.
 * - Redistributions in binary form must reproduce the above copyright
 *   notice, this list of conditions and the following disclaimer in the
 *   documentation and/or other materials provided with the distribution.
 * - The name of the author may not be used to endorse or promote products
 *   derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <stddef.h>
#include <math.h>
#include "../tester.h"

#define OPERANDS         10
#define PRECISIONF    10000
#define PRECISION 100000000

static double arguments[OPERANDS] = {
        3.5, -2.1, 100.0, 50.0, -1024.0, 0.0, 768.3156, 1080.499999, -600.0, 1.0
};

static double results_cos[OPERANDS] = {
        -0.936456687291, -0.504846104600, 0.862318872288, 0.964966028492,
        0.987353618220, 1.0, -0.194939922623, 0.978471923925, -0.999023478833,
        0.540302305868
};

static double results_sin[OPERANDS] = {
        -0.350783227690, -0.863209366649, -0.506365641110, -0.262374853704,
        0.158533380044, 0.0, 0.980815184715, -0.206379975025, -0.044182448332,
        0.841470984808
};

static double results_trunc[OPERANDS] = {
        3.0, -2.0, 100.0, 50.0, -1024.0, 0.0, 768.0, 1080.0, -600.0, 1.0
};

static bool cmp_float(float a, float b)
{
        float r;

        /* XXX Need fabsf() */
        if (b < 1.0 / PRECISIONF && b > -1.0 / PRECISIONF)
                r = a;
        else
                r = a / b - 1.0;

        /* XXX Need fabsf() */
        if (r < 0.0)
                r = -r;

        return r < 1.0 / PRECISIONF;
}

static bool cmp_double(double a, double b)
{
        double r;

        /* XXX Need fabs() */
        if (b < 1.0 / PRECISION && b > -1.0 / PRECISION)
                r = a;
        else
                r = a / b - 1.0;

        /* XXX Need fabs() */
        if (r < 0.0)
                r = -r;

        return r < 1.0 / PRECISION;
}

const char *test_float2(void)
{
        bool fail = false;

        for (unsigned int i = 0; i < OPERANDS; i++) {
                double res = cos(arguments[i]);

                if (!cmp_double(res, results_cos[i])) {
                        TPRINTF("Double precision cos failed "
                            "(%lf != %lf, arg %u)\n", res, results_cos[i], i);
                        fail = true;
                }
        }

        for (unsigned int i = 0; i < OPERANDS; i++) {
                float res = cosf(arguments[i]);

                if (!cmp_float(res, results_cos[i])) {
                        TPRINTF("Single precision cos failed "
                            "(%f != %lf, arg %u)\n", res, results_cos[i], i);
                        fail = true;
                }
        }

        for (unsigned int i = 0; i < OPERANDS; i++) {
                double res = sin(arguments[i]);

                if (!cmp_double(res, results_sin[i])) {
                        TPRINTF("Double precision sin failed "
                            "(%lf != %lf, arg %u)\n", res, results_sin[i], i);
                        fail = true;
                }
        }

        for (unsigned int i = 0; i < OPERANDS; i++) {
                float res = sinf(arguments[i]);

                if (!cmp_float(res, results_sin[i])) {
                        TPRINTF("Single precision sin failed "
                            "(%f != %lf, arg %u)\n", res, results_sin[i], i);
                        fail = true;
                }
        }

        for (unsigned int i = 0; i < OPERANDS; i++) {
                double res = trunc(arguments[i]);

                if (!cmp_double(res, results_trunc[i])) {
                        TPRINTF("Double precision trunc failed "
                            "(%lf != %lf, arg %u)\n", res, results_trunc[i], i);
                        fail = true;
                }
        }

        for (unsigned int i = 0; i < OPERANDS; i++) {
                float res = truncf(arguments[i]);

                if (!cmp_float(res, results_trunc[i])) {
                        TPRINTF("Single precision trunc failed "
                            "(%f != %lf, arg %u)\n", res, results_trunc[i], i);
                        fail = true;
                }
        }

        if (fail)
                return "Floating point imprecision";

        return NULL;
}

/* [<][>][^][v][top][bottom][index][help] */
HelenOS homepage, sources at GitHub