diff options
Diffstat (limited to 'contrib/netbsd-tests/lib/libm/t_exp.c')
-rw-r--r-- | contrib/netbsd-tests/lib/libm/t_exp.c | 567 |
1 files changed, 567 insertions, 0 deletions
diff --git a/contrib/netbsd-tests/lib/libm/t_exp.c b/contrib/netbsd-tests/lib/libm/t_exp.c new file mode 100644 index 0000000..6d41097 --- /dev/null +++ b/contrib/netbsd-tests/lib/libm/t_exp.c @@ -0,0 +1,567 @@ +/* $NetBSD: t_exp.c,v 1.7 2014/03/17 11:08:11 martin Exp $ */ + +/*- + * Copyright (c) 2011 The NetBSD Foundation, Inc. + * All rights reserved. + * + * This code is derived from software contributed to The NetBSD Foundation + * by Jukka Ruohonen. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. 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. + * + * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS + * ``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 FOUNDATION OR CONTRIBUTORS + * 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 <atf-c.h> +#include <math.h> +#include "t_libm.h" + +/* y = exp(x) */ +static const struct { + double x; + double y; + double e; +} exp_values[] = { + { -10, 0.4539992976248485e-4, 1e-4, }, + { -5, 0.6737946999085467e-2, 1e-2, }, + { -1, 0.3678794411714423, 1e-1, }, + { -0.1, 0.9048374180359595, 1e-1, }, + { 0, 1.0000000000000000, 1, }, + { 0.1, 1.1051709180756477, 1, }, + { 1, 2.7182818284590452, 1, }, + { 5, 148.41315910257660, 1e2, }, + { 10, 22026.465794806718, 1e4, }, +}; + +/* + * exp2/exp2f(3) + */ +ATF_LIBM_TEST(exp2_is_nan, "Test exp2(x) == NaN") +{ +#ifdef T_LIBM_NAN + T_LIBM_CHECK_NAN(0, exp2, T_LIBM_NAN); + T_LIBM_CHECK_NAN(0, exp2f, T_LIBM_NAN); +#else + atf_tc_skip("no NaN on this machine"); +#endif +} + +ATF_LIBM_TEST(exp2_is_plus_zero, "Test exp2(x) == +0.0") +{ +#ifdef T_LIBM_MINUS_INF + T_LIBM_CHECK_PLUS_ZERO(0, exp2, T_LIBM_MINUS_INF); + T_LIBM_CHECK_PLUS_ZERO(0, exp2f, T_LIBM_MINUS_INF); +#else + atf_tc_skip("no +/-Inf on this machine"); +#endif +} + +ATF_LIBM_TEST(exp2_powers, "Test exp2(x) is correct for some integer x") +{ + static const struct { + double x; + double d_y; + double f_y; + } v[] = { + { +0.0, 1.0, 1.0 }, + { -0.0, 1.0, 1.0 }, + { 1, 0x1p1, 0x1p1 }, + { 2, 0x1p2, 0x1p2 }, + { 100, 0x1p100, 0x1p100 }, + { 125, 0x1p125, 0x1p125 }, + { 126, 0x1p126, 0x1p126 }, +#if __DBL_MAX_EXP__ > 129 + { 127, 0x1p127, 0x1p127 }, +#endif +#ifdef T_LIBM_PLUS_INF + { 128, 0x1p128, T_LIBM_PLUS_INF }, + { 129, 0x1p129, T_LIBM_PLUS_INF }, + { 1000, 0x1p1000, T_LIBM_PLUS_INF }, + { 1020, 0x1p1020, T_LIBM_PLUS_INF }, + { 1023, 0x1p1023, T_LIBM_PLUS_INF }, + { 1024, T_LIBM_PLUS_INF, T_LIBM_PLUS_INF }, + { 1030, T_LIBM_PLUS_INF, T_LIBM_PLUS_INF }, + { 1050, T_LIBM_PLUS_INF, T_LIBM_PLUS_INF }, + { 2000, T_LIBM_PLUS_INF, T_LIBM_PLUS_INF }, + { 16383, T_LIBM_PLUS_INF, T_LIBM_PLUS_INF }, + { 16384, T_LIBM_PLUS_INF, T_LIBM_PLUS_INF }, + { 16385, T_LIBM_PLUS_INF, T_LIBM_PLUS_INF }, +#endif + { -1, 0x1p-1, 0x1p-1 }, + { -2, 0x1p-2, 0x1p-2 }, + { -100, 0x1p-100, 0x1p-100 }, + { -127, 0x1p-127, 0x1p-127 }, + { -128, 0x1p-128, 0x1p-128 }, +#if __LDBL_MIN_EXP__ < -129 + { -300, 0x1p-300, 0.0}, + { -400, 0x1p-400, 0.0}, + {-1000, 0x1p-1000, 0.0}, + {-1022, 0x1p-1022, 0.0}, + /* These should be denormal numbers */ + {-1023, 0x1p-1023, 0.0}, + {-1024, 0x1p-1024, 0.0}, + {-1040, 0x1p-1040, 0.0}, + {-1060, 0x1p-1060, 0.0}, + /* This is the smallest result gcc will allow */ + {-1074, 0x1p-1074, 0.0}, +#endif + {-1075, 0x0, 0.0}, + {-1080, 0x0, 0.0}, + {-2000, 0x0, 0.0}, + {-16382, 0x0, 0.0}, + {-16383, 0x0, 0.0}, + {-16384, 0x0, 0.0}, + }; + unsigned int i; + + for (i = 0; i < __arraycount(v); i++) { + T_LIBM_CHECK(i, exp2, v[i].x, v[i].d_y, 0.0); + T_LIBM_CHECK(i, exp2f, v[i].x, v[i].f_y, 0.0); + } +} + +ATF_LIBM_TEST(exp2_values, "Test exp2(x) is correct for some x") +{ + static const struct { + double x; + double y; + double d_eps; + double f_eps; + } v[] = { +#if __DBL_MAX_EXP__ > 128 + /* The largest double constant */ + { 0x1.fffffffffffffp9, 0x1.ffffffffffd3ap1023, + 0x1p969, 0.0 }, + /* The largest float constant */ + { 0x1.fffffep6, 0x1.ffff4ep+127, 6e30, 0.0 }, +#endif +#ifdef T_LIBM_PLUS_INF + { T_LIBM_PLUS_INF, T_LIBM_PLUS_INF, 0.0, 0.0 }, +#endif + + /* The few values from the old tests */ + /* Results from i386/amd64, d_eps needed on i386 */ + { 1.1, 0x1.125fbee250664p+1, 0x1p-52, 0x1.8p-22 }, + { 2.2, 0x1.2611186bae675p+2, 0x1p-51, 0x1.8p-21 }, + { 3.3, 0x1.3b2c47bff8328p+3, 0x1p-50, 0x1.8p-20 }, + { 4.4, 0x1.51cb453b9536ep+4, 0x1p-49, 0x1.8p-19 }, + { 5.5, 0x1.6a09e667f3bcdp+5, 0x1p-48, 0x1.8p-18 }, + { 6.6, 0x1.8406003b2ae5bp+6, 0x1p-47, 0x1.8p-17 }, + /* + * These two currently fail for 'float'. + * 8.8 is definitely out by more than it should be. + */ + { 7.7, 0x1.9fdf8bcce533ep+7, 0x1p-46, 0x1.8p-16 }, + { 8.8, 0x1.bdb8cdadbe124p+8, 0x1p-45, 0x1.8p-15 }, + }; + unsigned int i; + + for (i = 0; i < __arraycount(v); i++) { + T_LIBM_CHECK(i, exp2, v[i].x, v[i].y, v[i].d_eps); + if (i > 1) + T_LIBM_CHECK(i, exp2f, v[i].x, v[i].y, v[i].f_eps); + } +} + + +/* + * exp(3) + */ +ATF_TC(exp_nan); +ATF_TC_HEAD(exp_nan, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test exp(NaN) == NaN"); +} + +ATF_TC_BODY(exp_nan, tc) +{ + const double x = 0.0L / 0.0L; + + if (isnan(exp(x)) == 0) + atf_tc_fail_nonfatal("exp(NaN) != NaN"); +} + +ATF_TC(exp_inf_neg); +ATF_TC_HEAD(exp_inf_neg, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test exp(-Inf) == +0.0"); +} + +ATF_TC_BODY(exp_inf_neg, tc) +{ + const double x = -1.0L / 0.0L; + double y = exp(x); + + if (fabs(y) > 0.0 || signbit(y) != 0) + atf_tc_fail_nonfatal("exp(-Inf) != +0.0"); +} + +ATF_TC(exp_inf_pos); +ATF_TC_HEAD(exp_inf_pos, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test exp(+Inf) == +Inf"); +} + +ATF_TC_BODY(exp_inf_pos, tc) +{ + const double x = 1.0L / 0.0L; + double y = exp(x); + + if (isinf(y) == 0 || signbit(y) != 0) + atf_tc_fail_nonfatal("exp(+Inf) != +Inf"); +} + +ATF_TC(exp_product); +ATF_TC_HEAD(exp_product, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test some selected exp(x)"); +} + +ATF_TC_BODY(exp_product, tc) +{ + double eps; + double x; + double y; + size_t i; + + for (i = 0; i < __arraycount(exp_values); i++) { + x = exp_values[i].x; + y = exp_values[i].y; + eps = 1e-15 * exp_values[i].e; + + if (fabs(exp(x) - y) > eps) + atf_tc_fail_nonfatal("exp(%0.01f) != %18.18e", x, y); + } +} + +ATF_TC(exp_zero_neg); +ATF_TC_HEAD(exp_zero_neg, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test exp(-0.0) == 1.0"); +} + +ATF_TC_BODY(exp_zero_neg, tc) +{ + const double x = -0.0L; + + if (fabs(exp(x) - 1.0) > 0.0) + atf_tc_fail_nonfatal("exp(-0.0) != 1.0"); +} + +ATF_TC(exp_zero_pos); +ATF_TC_HEAD(exp_zero_pos, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test exp(+0.0) == 1.0"); +} + +ATF_TC_BODY(exp_zero_pos, tc) +{ + const double x = 0.0L; + + if (fabs(exp(x) - 1.0) > 0.0) + atf_tc_fail_nonfatal("exp(+0.0) != 1.0"); +} + +/* + * expf(3) + */ +ATF_TC(expf_nan); +ATF_TC_HEAD(expf_nan, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expf(NaN) == NaN"); +} + +ATF_TC_BODY(expf_nan, tc) +{ + const float x = 0.0L / 0.0L; + + if (isnan(expf(x)) == 0) + atf_tc_fail_nonfatal("expf(NaN) != NaN"); +} + +ATF_TC(expf_inf_neg); +ATF_TC_HEAD(expf_inf_neg, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expf(-Inf) == +0.0"); +} + +ATF_TC_BODY(expf_inf_neg, tc) +{ + const float x = -1.0L / 0.0L; + float y = expf(x); + + if (fabsf(y) > 0.0 || signbit(y) != 0) + atf_tc_fail_nonfatal("expf(-Inf) != +0.0"); +} + +ATF_TC(expf_inf_pos); +ATF_TC_HEAD(expf_inf_pos, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expf(+Inf) == +Inf"); +} + +ATF_TC_BODY(expf_inf_pos, tc) +{ + const float x = 1.0L / 0.0L; + float y = expf(x); + + if (isinf(y) == 0 || signbit(y) != 0) + atf_tc_fail_nonfatal("expf(+Inf) != +Inf"); +} + +ATF_TC(expf_product); +ATF_TC_HEAD(expf_product, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test some selected expf(x)"); +} + +ATF_TC_BODY(expf_product, tc) +{ + float eps; + float x; + float y; + size_t i; + + for (i = 0; i < __arraycount(exp_values); i++) { + x = exp_values[i].x; + y = exp_values[i].y; + eps = 1e-6 * exp_values[i].e; + + if (fabsf(expf(x) - y) > eps) + atf_tc_fail_nonfatal("expf(%0.01f) != %18.18e", x, y); + } +} + +ATF_TC(expf_zero_neg); +ATF_TC_HEAD(expf_zero_neg, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expf(-0.0) == 1.0"); +} + +ATF_TC_BODY(expf_zero_neg, tc) +{ + const float x = -0.0L; + + if (fabsf(expf(x) - 1.0f) > 0.0) + atf_tc_fail_nonfatal("expf(-0.0) != 1.0"); +} + +ATF_TC(expf_zero_pos); +ATF_TC_HEAD(expf_zero_pos, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expf(+0.0) == 1.0"); +} + +ATF_TC_BODY(expf_zero_pos, tc) +{ + const float x = 0.0L; + + if (fabsf(expf(x) - 1.0f) > 0.0) + atf_tc_fail_nonfatal("expf(+0.0) != 1.0"); +} + +/* + * expm1(3) + */ +ATF_TC(expm1_nan); +ATF_TC_HEAD(expm1_nan, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1(NaN) == NaN"); +} + +ATF_TC_BODY(expm1_nan, tc) +{ + const double x = 0.0L / 0.0L; + + if (isnan(expm1(x)) == 0) + atf_tc_fail_nonfatal("expm1(NaN) != NaN"); +} + +ATF_TC(expm1_inf_neg); +ATF_TC_HEAD(expm1_inf_neg, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1(-Inf) == -1"); +} + +ATF_TC_BODY(expm1_inf_neg, tc) +{ + const double x = -1.0L / 0.0L; + + if (expm1(x) != -1.0) + atf_tc_fail_nonfatal("expm1(-Inf) != -1.0"); +} + +ATF_TC(expm1_inf_pos); +ATF_TC_HEAD(expm1_inf_pos, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1(+Inf) == +Inf"); +} + +ATF_TC_BODY(expm1_inf_pos, tc) +{ + const double x = 1.0L / 0.0L; + double y = expm1(x); + + if (isinf(y) == 0 || signbit(y) != 0) + atf_tc_fail_nonfatal("expm1(+Inf) != +Inf"); +} + +ATF_TC(expm1_zero_neg); +ATF_TC_HEAD(expm1_zero_neg, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1(-0.0) == -0.0"); +} + +ATF_TC_BODY(expm1_zero_neg, tc) +{ + const double x = -0.0L; + double y = expm1(x); + + if (fabs(y) > 0.0 || signbit(y) == 0) + atf_tc_fail_nonfatal("expm1(-0.0) != -0.0"); +} + +ATF_TC(expm1_zero_pos); +ATF_TC_HEAD(expm1_zero_pos, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1(+0.0) == 1.0"); +} + +ATF_TC_BODY(expm1_zero_pos, tc) +{ + const double x = 0.0L; + double y = expm1(x); + + if (fabs(y) > 0.0 || signbit(y) != 0) + atf_tc_fail_nonfatal("expm1(+0.0) != +0.0"); +} + +/* + * expm1f(3) + */ +ATF_TC(expm1f_nan); +ATF_TC_HEAD(expm1f_nan, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1f(NaN) == NaN"); +} + +ATF_TC_BODY(expm1f_nan, tc) +{ + const float x = 0.0L / 0.0L; + + if (isnan(expm1f(x)) == 0) + atf_tc_fail_nonfatal("expm1f(NaN) != NaN"); +} + +ATF_TC(expm1f_inf_neg); +ATF_TC_HEAD(expm1f_inf_neg, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1f(-Inf) == -1"); +} + +ATF_TC_BODY(expm1f_inf_neg, tc) +{ + const float x = -1.0L / 0.0L; + + if (expm1f(x) != -1.0) + atf_tc_fail_nonfatal("expm1f(-Inf) != -1.0"); +} + +ATF_TC(expm1f_inf_pos); +ATF_TC_HEAD(expm1f_inf_pos, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1f(+Inf) == +Inf"); +} + +ATF_TC_BODY(expm1f_inf_pos, tc) +{ + const float x = 1.0L / 0.0L; + float y = expm1f(x); + + if (isinf(y) == 0 || signbit(y) != 0) + atf_tc_fail_nonfatal("expm1f(+Inf) != +Inf"); +} + +ATF_TC(expm1f_zero_neg); +ATF_TC_HEAD(expm1f_zero_neg, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1f(-0.0) == -0.0"); +} + +ATF_TC_BODY(expm1f_zero_neg, tc) +{ + const float x = -0.0L; + float y = expm1f(x); + + if (fabsf(y) > 0.0 || signbit(y) == 0) + atf_tc_fail_nonfatal("expm1f(-0.0) != -0.0"); +} + +ATF_TC(expm1f_zero_pos); +ATF_TC_HEAD(expm1f_zero_pos, tc) +{ + atf_tc_set_md_var(tc, "descr", "Test expm1f(+0.0) == 1.0"); +} + +ATF_TC_BODY(expm1f_zero_pos, tc) +{ + const float x = 0.0L; + float y = expm1f(x); + + if (fabsf(y) > 0.0 || signbit(y) != 0) + atf_tc_fail_nonfatal("expm1f(+0.0) != +0.0"); +} + +ATF_TP_ADD_TCS(tp) +{ + + ATF_TP_ADD_TC(tp, exp2_is_nan); + ATF_TP_ADD_TC(tp, exp2_is_plus_zero); + ATF_TP_ADD_TC(tp, exp2_values); + ATF_TP_ADD_TC(tp, exp2_powers); + + ATF_TP_ADD_TC(tp, exp_nan); + ATF_TP_ADD_TC(tp, exp_inf_neg); + ATF_TP_ADD_TC(tp, exp_inf_pos); + ATF_TP_ADD_TC(tp, exp_product); + ATF_TP_ADD_TC(tp, exp_zero_neg); + ATF_TP_ADD_TC(tp, exp_zero_pos); + + ATF_TP_ADD_TC(tp, expf_nan); + ATF_TP_ADD_TC(tp, expf_inf_neg); + ATF_TP_ADD_TC(tp, expf_inf_pos); + ATF_TP_ADD_TC(tp, expf_product); + ATF_TP_ADD_TC(tp, expf_zero_neg); + ATF_TP_ADD_TC(tp, expf_zero_pos); + + ATF_TP_ADD_TC(tp, expm1_nan); + ATF_TP_ADD_TC(tp, expm1_inf_neg); + ATF_TP_ADD_TC(tp, expm1_inf_pos); + ATF_TP_ADD_TC(tp, expm1_zero_neg); + ATF_TP_ADD_TC(tp, expm1_zero_pos); + + ATF_TP_ADD_TC(tp, expm1f_nan); + ATF_TP_ADD_TC(tp, expm1f_inf_neg); + ATF_TP_ADD_TC(tp, expm1f_inf_pos); + ATF_TP_ADD_TC(tp, expm1f_zero_neg); + ATF_TP_ADD_TC(tp, expm1f_zero_pos); + + return atf_no_error(); +} |