From:	<ams AT ludd DOT luth DOT se>
Message-Id:	<200308171216.h7HCGL2T021232@speedy.ludd.luth.se>
Subject:	Re: tests/libc/c99/math/t-fpclas.c
In-Reply-To:	<200308171115.h7HBFAri017492@speedy.ludd.luth.se>
	"from ams AT ludd DOT luth DOT se at Aug 17, 2003 01:15:10 pm"
To:	djgpp-workers AT delorie DOT com
Date:	Sun, 17 Aug 2003 14:16:21 +0200 (CEST)
X-Mailer:	ELM [version 2.4ME+ PL78 (25)]
MIME-Version:	1.0
X-MailScanner:	Found to be clean
Reply-To:	djgpp-workers AT delorie DOT com
Errors-To:	nobody AT delorie DOT com
X-Mailing-List:	djgpp-workers AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

HELO

Ok. Hopefully a working t-fpclas.c now. It did find bugs in
fpclassifyld(). Worth to note that printf() and fpclassifyld() have
different ideas about entries 6-9 for long doubles.

/*
 * File t-fpclas.c (8.3 of t-fpclassify.c).
 *
 * Copyright (C) 2003 Martin Str@"omberg <ams AT ludd DOT luth DOT se>.
 *
 * This software may be used freely so long as this copyright notice is
 * left intact. There is no warranty on this software.
 *
 */

#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <libc/ieee.h>

typedef struct { 
  float_t number; 	/* Number to test. */
  int class; 		/* Which class. */
} test_float_t;

static const test_float_t tests_float[] = 
{ 
  /* Zeros. */
  { { 0, 0, 0 }, FP_ZERO }, /* 0.0 */
  { { 0, 0, 1 }, FP_ZERO }, /* -0.0 */

  /* Subnormals aka denormals. */
  { { 1, 0, 0 }, FP_SUBNORMAL }, /* Very small number. */
  { { 1, 0, 1 }, FP_SUBNORMAL }, /* Very small -number. */

  /* Normals. */
  { { 1, 1, 0 }, FP_NORMAL }, /* Small number. */
  { { 1, 1, 1 }, FP_NORMAL }, /* Small -number. */
  { { 0xff, 0xfe, 0 }, FP_NORMAL }, /* Big number. */
  { { 0xff, 0xfe, 1 }, FP_NORMAL }, /* Big -number. */

  /* Infs. */
  { { 0, 0xff, 0 }, FP_INFINITE }, /* Inf */
  { { 0, 0xff, 1 }, FP_INFINITE }, /* -Inf */

  /* NaNs. */
  { { 1, 0xff, 0 }, FP_NAN }, /* SNaN */
  { { 1, 0xff, 1 }, FP_NAN }, /* -SNaN */
  { { 0x7fffff, 0xff, 0 }, FP_NAN }, /* QNaN */
  { { 0x7fffff, 0xff, 1 }, FP_NAN }, /* -QNaN */
};

static const size_t n_tests_float = sizeof(tests_float) / sizeof(tests_float[0]);

typedef struct { 
  double_t number; 	/* Number to test. */
  int class; 		/* Which class. */
} test_double_t;

static const test_double_t tests_double[] =
{ 
  /* Zeros. */
  { { 0, 0, 0, 0 }, FP_ZERO }, /* 0.0 */
  { { 0, 0, 0, 1 }, FP_ZERO }, /* -0.0 */

  /* Subnormals aka denormals. */
  { { 1, 0, 0, 0 }, FP_SUBNORMAL }, /* Very small number. */
  { { 1, 0, 0, 1 }, FP_SUBNORMAL }, /* Very small -number. */

  /* Normals. */
  { { 1, 0, 1, 0 }, FP_NORMAL }, /* Small number. */
  { { 1, 0, 1, 1 }, FP_NORMAL }, /* Small -number. */
  { { 0xffffffff, 0x7ffff, 0x7fe, 0 }, FP_NORMAL }, /* Big number. */
  { { 0xffffffff, 0x7ffff, 0x7fe, 1 }, FP_NORMAL }, /* Big -number. */

  /* Infs. */
  { { 0, 0, 0x7ff, 0 }, FP_INFINITE }, /* Inf */
  { { 0, 0, 0x7ff, 1 }, FP_INFINITE }, /* -Inf */

  /* NaNs. */
  { { 1, 0, 0x7ff, 0 }, FP_NAN }, /* SNaN */
  { { 1, 0, 0x7ff, 1 }, FP_NAN }, /* -SNaN */
  { { 0, 0xfffff, 0x7ff, 0 }, FP_NAN }, /* QNaN */
  { { 0, 0xfffff, 0x7ff, 1 }, FP_NAN }, /* -QNaN */
};

static const size_t n_tests_double = sizeof(tests_double) / sizeof(tests_double[0]);

typedef struct { 
  long_double_t number; /* Number to test. */
  int class; 		/* Which class. */
} test_long_double_t;

static const test_long_double_t tests_long_double[] =
{ 
  /* Zeros. */
  { { 0, 0, 0, 0 }, FP_ZERO }, /* 0.0. */
  { { 0, 0, 0, 1 }, FP_ZERO }, /* -0.0. */

  /* Subnormals aka denormals. */
  { { 1, 0, 0, 0 }, FP_SUBNORMAL }, /* Subnormal, smallest possible > 0. */
  { { 1, 0, 0, 1 }, FP_SUBNORMAL }, /* -Subnormal, biggest possible < 0. */
  { { 0, 1, 0, 0 }, FP_SUBNORMAL }, /* Subnormal. */
  { { 0, 1, 0, 1 }, FP_SUBNORMAL }, /* -Subnormal. */


  /* Unnormals. */
  { { 0, 0x80000000, 0, 0 }, FP_UNNORMAL }, /* Unnormal, 0 with integer bit set. */
  { { 0, 0x80000000, 0, 1 }, FP_UNNORMAL }, /* Unnormal, -0 with integer bit set. */
  { { 1, 0x80000000, 0, 0 }, FP_UNNORMAL }, /* Unnormal, smallest possible > 0 with integer bit set. */
  { { 1, 0x80000000, 0, 1 }, FP_UNNORMAL }, /* Unnormal, biggest possible < 0 with integer bit set. */
  { { 1, 0, 1, 0 }, FP_UNNORMAL }, /* Small number with missing 1. */
  { { 1, 0, 1, 1 }, FP_UNNORMAL }, /* Small -number with missing 1. */
  { { 0xffffffff, 0x7fffffff, 0x7ffe, 0 }, FP_UNNORMAL }, /* Big number with missing 1. */
  { { 0xffffffff, 0x7fffffff, 0x7ffe, 1 }, FP_UNNORMAL }, /* Big -number with missing 1. */
  { { 0, 0, 0x7fff, 0 }, FP_UNNORMAL }, /* Inf with missing 1. */
  { { 0, 0, 0x7fff, 1 }, FP_UNNORMAL }, /* -Inf with missing 1. */
  { { 1, 0, 0x7fff, 0 }, FP_UNNORMAL }, /* SNaN with missing 1 */
  { { 1, 0, 0x7fff, 1 }, FP_UNNORMAL }, /* -SNaN with missing 1 */
  { { 0, 0x7fffffff, 0x7fff, 0 }, FP_UNNORMAL }, /* QNaN with missing 1 */
  { { 0, 0x7fffffff, 0x7fff, 1 }, FP_UNNORMAL }, /* -QNaN with missing 1 */

  /* Normals. */
  { { 1, 0x80000000, 1, 0 }, FP_NORMAL }, /* Normal, smallest possible > 0. */
  { { 1, 0x80000000, 1, 1 }, FP_NORMAL }, /* Normal, biggest possible < 0. */
  { { 0xffffffff, 0xffffffff, 0x7ffe, 0 }, FP_NORMAL }, /* Big number. */
  { { 0xffffffff, 0xffffffff, 0x7ffe, 1 }, FP_NORMAL }, /* Big -number. */

  /* Infs. */
  { { 0, 0x80000000, 0x7fff, 0 }, FP_INFINITE }, /* Inf. */
  { { 0, 0x80000000, 0x7fff, 1 }, FP_INFINITE }, /* -Inf. */

  /* NaNs. */
  { { 1, 0x80000000, 0x7fff, 0 }, FP_NAN }, /* SNaN */
  { { 1, 0x80000000, 0x7fff, 1 }, FP_NAN }, /* -SNaN */
  { { 0, 0xffffffff, 0x7fff, 0 }, FP_NAN }, /* QNaN */
  { { 0, 0xffffffff, 0x7fff, 1 }, FP_NAN }, /* -QNaN */
};

static const size_t n_tests_long_double = sizeof(tests_long_double) / sizeof(tests_long_double[0]);

int
main(void)
{
  int result;
  size_t i;

  printf("FP_ZERO=0x%x, FP_SUBNORMAL=0x%x, FP_NORMAL=0x%x, FP_INFINITE=0x%d,\n"
	 "FP_NAN=0x%x, FP_UNNORMAL=0x%x\n\n", 
	 FP_ZERO, FP_SUBNORMAL, FP_NORMAL, FP_INFINITE, 
	 FP_NAN, FP_UNNORMAL);

  puts("Float tests:");
  for(i = 0; i < n_tests_float; i++)
  {
    result = fpclassify(*(const float *)(&(tests_float[i].number)));
    printf("\t%G -> 0x%x: ", *(const float *)(&(tests_float[i].number)), result);
    if( result == tests_float[i].class )
    {
      puts("Ok.");
    }
    else
    {
      puts("Fail.");
    }
  }

  puts("Double tests:");
  for(i = 0; i < n_tests_double; i++)
  {
    result = fpclassify(*(const double *)(&(tests_double[i].number)));
    printf("\t%G -> 0x%x: ", *(const double *)(&(tests_double[i].number)), result);
    if( result == tests_double[i].class )
    {
      puts("Ok.");
    }
    else
    {
      puts("Fail.");
    }
  }

  puts("Long double tests:");
  for(i = 0; i < n_tests_long_double; i++)
  {
    result = fpclassify(*(const long double *)(&(tests_long_double[i].number)));
    printf("\t%LG -> 0x%x: ", *(const long double *)(&(tests_long_double[i].number)), result);
    if( result == tests_long_double[i].class )
    {
      puts("Ok.");
    }
    else
    {
      puts("Fail.");
    }
  }

  return(EXIT_SUCCESS);
}


Right,

						MartinS

- Raw text -