From bc9f6000f6752153e5e1902259d5f491a88a1ae5 Mon Sep 17 00:00:00 2001 From: Ulrich Drepper Date: Sun, 23 Mar 1997 01:54:07 +0000 Subject: Update. 1997-03-23 02:11 Ulrich Drepper * time/sys/time.h: Make values ITIMER_* also available as macros. * elf/dl-support.c (_dl_sysdep_read_whole_file): Don't call __fstat but instead __fxstat directly to avoid dependency on libc.a when inline failed. * sysdeps/generic/dl-sysdep.c (_dl_sysdep_read_whole_file): Likewise. * math/Makefile (libm-routines): Add s_remquo. * sysdeps/libm-i387/s_remquo.S: New file. * sysdeps/libm-i387/s_remquof.S: New file. * sysdeps/libm-i387/s_remquol.S: New file. * sysdeps/libm-ieee754/s_remquo.c: New file. * sysdeps/libm-ieee754/s_remquof.c: New file. * sysdeps/libm-ieee754/s_remquol.c: New file. * math/libm-test.c (remquo_test): New function. * sysdeps/libm-ieee754/s_rintl.c: Handle implicit leading one correctly. 1997-03-22 14:06 Ulrich Drepper * math/Makefile (libm-calls): Add s_nan. * sysdeps/libm-ieee754/s_nan.c: New file. Implement `nan' function. * sysdeps/libm-ieee754/s_nanf.c: New file. Implement `nanf' function. * sysdeps/libm-ieee754/s_nanl.c: New file. Implement `nanl' function. * math/libm-test.c (basic_tests): Add tests for `nan' function. * math/libm-test.c (copysign_test): New function. (main): Call copysign_test. 1997-03-22 06:28 Ulrich Drepper * sysdeps/libm-ieee754/s_nextafter.c: Return y if x == y. * sysdeps/libm-ieee754/s_nextafterf.c: Likewise. * sysdeps/libm-ieee754/s_nextafterl.c: Likewise. * sysdeps/libm-i387/s_nextafterl.c: Likewise. * math/libm-test.c (fdim_test, fmin_test, fmax_test, nextafter_test): New functions. Test these functions. (main): Call above new functions. 1997-03-22 04:53 Ulrich Drepper * Net release 2.0.2. 1997-03-22 04:37 Ulrich Drepper * sysdeps/unix/sysv/linux/termbits.h (OXTABS): Define as alias for XTABS. * sysdeps/unix/sysv/linux/alpha/termbits.h (OXTABS): Likewise. * termios/sys/ttydefaults.h: Partly revert patch by Andreas Schwab of Sun Dec 15 16:33:44 1996. Proposed by Thomas Bushnell . 1997-03-21 13:41 Roland McGrath * sysdeps/sparc/dl-machine.h (elf_machine_rel): Rewritten as for i386. Check here for non-SHN_UNDEF STB_LOCAL symbols don't do any lookup or consult their values. (elf_machine_lookup_noexec_p, elf_machine_lookup_noplt_p, ELF_MACHINE_RELOC_NOPLT): New macros. * elf/rtld.c (dl_main): Pass ELF_MACHINE_RELOC_NOPLT to _dl_lookup_symbol in place of DL_LOOKUP_NOPLT. * sysdeps/i386/dl-machine.h (ELF_MACHINE_RELOC_NOPLT): New macro. * sysdeps/i386/dl-machine.h (elf_machine_rel): Rewritten to do the symbol lookup before checking reloc type except for R_386_RELATIVE. (elf_machine_lookup_noexec_p, elf_machine_lookup_noplt_p): New macros. * elf/dl-reloc.c (RESOLVE): Remove STB_LOCAL check; let it be per-machine since it supposedly can't happen on i386. * elf/dl-lookup.c (do_lookup): Change arg FLAGS to RELOC_TYPE. Use elf_machine_lookup_{noexec,noplt}_p macros on it. Remove gratuitous indirection from REF arg; change callers. (_dl_lookup_symbol, _dl_lookup_versioned_symbol): Change arg name. (_dl_lookup_symbol_skip, _dl_lookup_versioned_symbol_skip): Remove FLAGS arg altogether. * elf/dlsym.c: Remove argument of FLAGS parameter. * elf/dlvsym.c: Likewise. * elf/link.h: Update decls. * Makefile (distribute): Add BUGS. --- math/libm-test.c | 267 +++++++++++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 251 insertions(+), 16 deletions(-) (limited to 'math/libm-test.c') diff --git a/math/libm-test.c b/math/libm-test.c index a0482df758..8995f85475 100644 --- a/math/libm-test.c +++ b/math/libm-test.c @@ -102,10 +102,10 @@ static MATHTYPE plus_infty, minus_infty, nan_value; typedef MATHTYPE (*mathfunc) (MATHTYPE); -#define ISINF(x) \ -(sizeof (x) == sizeof (float) ? \ - isinff (x) \ - : sizeof (x) == sizeof (double) ? \ +#define ISINF(x) \ +(sizeof (x) == sizeof (float) ? \ + isinff (x) \ + : sizeof (x) == sizeof (double) ? \ isinf (x) : isinfl (x)) @@ -1222,6 +1222,222 @@ pow_test (void) } +static void +fdim_test (void) +{ + check ("fdim (+0, +0) = +0", FUNC(fdim) (0, 0), 0); + check ("fdim (9, 0) = 9", FUNC(fdim) (9, 0), 9); + check ("fdim (0, 9) = 0", FUNC(fdim) (0, 9), 0); + check ("fdim (-9, 0) = 9", FUNC(fdim) (-9, 0), 0); + check ("fdim (0, -9) = 9", FUNC(fdim) (0, -9), 9); + + check_isinfp ("fdim (+inf, 9) = +inf", FUNC(fdim) (plus_infty, 9)); + check_isinfp ("fdim (+inf, -9) = +inf", FUNC(fdim) (plus_infty, -9)); + check ("fdim (-inf, 9) = 0", FUNC(fdim) (minus_infty, 9), 0); + check ("fdim (-inf, -9) = 0", FUNC(fdim) (minus_infty, -9), 0); + check_isinfp ("fdim (+9, -inf) = +inf", FUNC(fdim) (9, minus_infty)); + check_isinfp ("fdim (-9, -inf) = +inf", FUNC(fdim) (-9, minus_infty)); + check ("fdim (9, inf) = 0", FUNC(fdim) (9, plus_infty), 0); + check ("fdim (-9, inf) = 0", FUNC(fdim) (-9, plus_infty), 0); + + check_isnan ("fdim (0, NaN) = NaN", FUNC(fdim) (0, nan_value)); + check_isnan ("fdim (9, NaN) = NaN", FUNC(fdim) (9, nan_value)); + check_isnan ("fdim (-9, NaN) = NaN", FUNC(fdim) (-9, nan_value)); + check_isnan ("fdim (NaN, 9) = NaN", FUNC(fdim) (nan_value, 9)); + check_isnan ("fdim (NaN, -9) = NaN", FUNC(fdim) (nan_value, -9)); + check_isnan ("fdim (+inf, NaN) = NaN", FUNC(fdim) (plus_infty, nan_value)); + check_isnan ("fdim (-inf, NaN) = NaN", FUNC(fdim) (minus_infty, nan_value)); + check_isnan ("fdim (NaN, +inf) = NaN", FUNC(fdim) (nan_value, plus_infty)); + check_isnan ("fdim (NaN, -inf) = NaN", FUNC(fdim) (nan_value, minus_infty)); + check_isnan ("fdim (NaN, NaN) = NaN", FUNC(fdim) (nan_value, nan_value)); +} + + +static void +fmin_test (void) +{ + check ("fmin (+0, +0) = +0", FUNC(fmin) (0, 0), 0); + check ("fmin (9, 0) = 0", FUNC(fmin) (9, 0), 0); + check ("fmin (0, 9) = 0", FUNC(fmin) (0, 9), 0); + check ("fmin (-9, 0) = -9", FUNC(fmin) (-9, 0), -9); + check ("fmin (0, -9) = -9", FUNC(fmin) (0, -9), -9); + + check ("fmin (+inf, 9) = 9", FUNC(fmin) (plus_infty, 9), 9); + check ("fmin (9, +inf) = 9", FUNC(fmin) (9, plus_infty), 9); + check ("fmin (+inf, -9) = -9", FUNC(fmin) (plus_infty, -9), -9); + check ("fmin (-9, +inf) = -9", FUNC(fmin) (-9, plus_infty), -9); + check_isinfn ("fmin (-inf, 9) = -inf", FUNC(fmin) (minus_infty, 9)); + check_isinfn ("fmin (-inf, -9) = -inf", FUNC(fmin) (minus_infty, -9)); + check_isinfn ("fmin (+9, -inf) = -inf", FUNC(fmin) (9, minus_infty)); + check_isinfn ("fmin (-9, -inf) = -inf", FUNC(fmin) (-9, minus_infty)); + + check ("fmin (0, NaN) = 0", FUNC(fmin) (0, nan_value), 0); + check ("fmin (9, NaN) = 9", FUNC(fmin) (9, nan_value), 9); + check ("fmin (-9, NaN) = 9", FUNC(fmin) (-9, nan_value), -9); + check ("fmin (NaN, 0) = 0", FUNC(fmin) (nan_value, 0), 0); + check ("fmin (NaN, 9) = NaN", FUNC(fmin) (nan_value, 9), 9); + check ("fmin (NaN, -9) = NaN", FUNC(fmin) (nan_value, -9), -9); + check_isinfp ("fmin (+inf, NaN) = +inf", FUNC(fmin) (plus_infty, nan_value)); + check_isinfn ("fmin (-inf, NaN) = -inf", FUNC(fmin) (minus_infty, nan_value)); + check_isinfp ("fmin (NaN, +inf) = +inf", FUNC(fmin) (nan_value, plus_infty)); + check_isinfn ("fmin (NaN, -inf) = -inf", FUNC(fmin) (nan_value, minus_infty)); + check_isnan ("fmin (NaN, NaN) = NaN", FUNC(fmin) (nan_value, nan_value)); +} + + +static void +fmax_test (void) +{ + check ("fmax (+0, +0) = +0", FUNC(fmax) (0, 0), 0); + check ("fmax (9, 0) = 9", FUNC(fmax) (9, 0), 9); + check ("fmax (0, 9) = 9", FUNC(fmax) (0, 9), 9); + check ("fmax (-9, 0) = 0", FUNC(fmax) (-9, 0), 0); + check ("fmax (0, -9) = 0", FUNC(fmax) (0, -9), 0); + + check_isinfp ("fmax (+inf, 9) = +inf", FUNC(fmax) (plus_infty, 9)); + check_isinfp ("fmax (9, +inf) = +inf", FUNC(fmax) (0, plus_infty)); + check_isinfp ("fmax (-9, +inf) = +inf", FUNC(fmax) (-9, plus_infty)); + check_isinfp ("fmax (+inf, -9) = +inf", FUNC(fmax) (plus_infty, -9)); + check ("fmax (-inf, 9) = 9", FUNC(fmax) (minus_infty, 9), 9); + check ("fmax (-inf, -9) = -9", FUNC(fmax) (minus_infty, -9), -9); + check ("fmax (+9, -inf) = 9", FUNC(fmax) (9, minus_infty), 9); + check ("fmax (-9, -inf) = -9", FUNC(fmax) (-9, minus_infty), -9); + + check ("fmax (0, NaN) = 0", FUNC(fmax) (0, nan_value), 0); + check ("fmax (9, NaN) = 9", FUNC(fmax) (9, nan_value), 9); + check ("fmax (-9, NaN) = 9", FUNC(fmax) (-9, nan_value), -9); + check ("fmax (NaN, 0) = 0", FUNC(fmax) (nan_value, 0), 0); + check ("fmax (NaN, 9) = NaN", FUNC(fmax) (nan_value, 9), 9); + check ("fmax (NaN, -9) = NaN", FUNC(fmax) (nan_value, -9), -9); + check_isinfp ("fmax (+inf, NaN) = +inf", FUNC(fmax) (plus_infty, nan_value)); + check_isinfn ("fmax (-inf, NaN) = -inf", FUNC(fmax) (minus_infty, nan_value)); + check_isinfp ("fmax (NaN, +inf) = +inf", FUNC(fmax) (nan_value, plus_infty)); + check_isinfn ("fmax (NaN, -inf) = -inf", FUNC(fmax) (nan_value, minus_infty)); + check_isnan ("fmax (NaN, NaN) = NaN", FUNC(fmax) (nan_value, nan_value)); +} + + +static void +nextafter_test (void) +{ + MATHTYPE x; + + check ("nextafter (+0, +0) = +0", FUNC(nextafter) (0, 0), 0); + check ("nextafter (-0, +0) = +0", FUNC(nextafter) (minus_zero, 0), 0); + check ("nextafter (+0, -0) = -0", FUNC(nextafter) (0, minus_zero), + minus_zero); + check ("nextafter (-0, -0) = -0", FUNC(nextafter) (minus_zero, minus_zero), + minus_zero); + + check ("nextafter (9, 9) = 9", FUNC(nextafter) (9, 9), 9); + check ("nextafter (-9, -9) = -9", FUNC(nextafter) (-9, -9), -9); + check_isinfp ("nextafter (+inf, +inf) = +inf", + FUNC(nextafter) (plus_infty, plus_infty)); + check_isinfn ("nextafter (-inf, -inf) = -inf", + FUNC(nextafter) (minus_infty, minus_infty)); + + x = rand () * 1.1; + check_isnan ("nextafter (NaN, x) = NaN", FUNC(nextafter) (nan_value, x)); + check_isnan ("nextafter (x, NaN) = NaN", FUNC(nextafter) (x, nan_value)); + check_isnan ("nextafter (NaN, NaN) = NaN", FUNC(nextafter) (nan_value, + nan_value)); + + /* XXX We need the hexadecimal FP number representation here for further + tests. */ +} + + +static void +copysign_test (void) +{ + check ("copysign (0, 4) = 0", FUNC(copysign) (0, 4), 0); + check ("copysign (0, -4) = -0", FUNC(copysign) (0, -4), minus_zero); + check ("copysign (-0, 4) = 0", FUNC(copysign) (minus_zero, 4), 0); + check ("copysign (-0, -4) = -0", FUNC(copysign) (minus_zero, -4), + minus_zero); + + check_isinfp ("copysign (+inf, 0) = +inf", FUNC(copysign) (plus_infty, 0)); + check_isinfn ("copysign (+inf, -0) = -inf", FUNC(copysign) (plus_infty, + minus_zero)); + check_isinfp ("copysign (-inf, 0) = +inf", FUNC(copysign) (minus_infty, 0)); + check_isinfn ("copysign (-inf, -0) = -inf", FUNC(copysign) (minus_infty, + minus_zero)); + + check ("copysign (0, +inf) = 0", FUNC(copysign) (0, plus_infty), 0); + check ("copysign (0, -inf) = -0", FUNC(copysign) (0, minus_zero), + minus_zero); + check ("copysign (-0, +inf) = 0", FUNC(copysign) (minus_zero, plus_infty), + 0); + check ("copysign (-0, -inf) = -0", FUNC(copysign) (minus_zero, minus_zero), + minus_zero); + + /* XXX More correctly we would have to check the sign of the NaN. */ + check_isnan ("copysign (+NaN, 0) = +inf", FUNC(copysign) (nan_value, 0)); + check_isnan ("copysign (+NaN, -0) = -inf", FUNC(copysign) (nan_value, + minus_zero)); + check_isnan ("copysign (-NaN, 0) = +inf", FUNC(copysign) (-nan_value, 0)); + check_isnan ("copysign (-NaN, -0) = -inf", FUNC(copysign) (-nan_value, + minus_zero)); +} + + +static void +trunc_test (void) +{ + check ("trunc(0) = 0", FUNC(trunc) (0), 0); + check ("trunc(-0) = -0", FUNC(trunc) (minus_zero), minus_zero); + check ("trunc(0.625) = 0", FUNC(trunc) (0.625), 0); + check ("trunc(-0.625) = -0", FUNC(trunc) (-0.625), minus_zero); + check ("trunc(1) = 1", FUNC(trunc) (1), 1); + check ("trunc(-1) = -1", FUNC(trunc) (-1), -1); + check ("trunc(1.625) = 1", FUNC(trunc) (1.625), 1); + check ("trunc(-1.625) = -1", FUNC(trunc) (-1.625), -1); + + check ("trunc(1048580.625) = 1048580", FUNC(trunc) (1048580.625L), + 1048580L); + check ("trunc(-1048580.625) = -1048580", FUNC(trunc) (-1048580.625L), + -1048580L); + + check ("trunc(8388610.125) = 8388610", FUNC(trunc) (8388610.125L), + 8388610.0L); + check ("trunc(-8388610.125) = -8388610", FUNC(trunc) (-8388610.125L), + -8388610.0L); + + check ("trunc(4294967296.625) = 4294967296", FUNC(trunc) (4294967296.625L), + 4294967296.0L); + check ("trunc(-4294967296.625) = -4294967296", + FUNC(trunc) (-4294967296.625L), -4294967296.0L); + + check_isinfp ("trunc(+inf) = +inf", FUNC(trunc) (plus_infty)); + check_isinfn ("trunc(-inf) = -inf", FUNC(trunc) (minus_infty)); + check_isnan ("trunc(NaN) = NaN", FUNC(trunc) (nan_value)); +} + + +static void +remquo_test (void) +{ + int quo; + MATHTYPE result; + + result = FUNC(remquo) (1.625, 1.0, &quo); + check ("remquo(1.625, 1.0, &x) == -0.375", result, -0.375); + check ("remquo(1.625, 1.0, &x) puts 1 in x", quo, 1); + + result = FUNC(remquo) (-1.625, 1.0, &quo); + check ("remquo(-1.625, 1.0, &x) == 0.375", result, 0.375); + check ("remquo(-1.625, 1.0, &x) puts 1 in x", quo, -1); + + result = FUNC(remquo) (1.625, -1.0, &quo); + check ("remquo(1.125, -1.0, &x) == 0.125", result, 0.125); + check ("remquo(1.125, -1.0, &x) puts 1 in x", quo, -1); + + result = FUNC(remquo) (-1.625, -1.0, &quo); + check ("remquo(-1.125, -1.0, &x) == 0.125", result, 0.125); + check ("remquo(-1.125, -1.0, &x) puts 1 in x", quo, 1); +} + + static void inverse_func_pair_test (const char *test_name, mathfunc f1, mathfunc inverse, @@ -1244,33 +1460,33 @@ inverse_func_pair_test (const char *test_name, static void inverse_functions (void) { - inverse_func_pair_test ("(asin(sin(x)) == x", + inverse_func_pair_test ("asin(sin(x)) == x", FUNC(sin), FUNC(asin), 1.0, CHOOSE (0, 0, 1e-7L)); - inverse_func_pair_test ("(sin(asin(x)) == x", + inverse_func_pair_test ("sin(asin(x)) == x", FUNC(asin), FUNC(sin), 1.0, 0.0); - inverse_func_pair_test ("(acos(cos(x)) == x", + inverse_func_pair_test ("acos(cos(x)) == x", FUNC(cos), FUNC(acos), 1.0, CHOOSE (0, 1e-15L, 0)); - inverse_func_pair_test ("(cos(acos(x)) == x", + inverse_func_pair_test ("cos(acos(x)) == x", FUNC(acos), FUNC(cos), 1.0, 0.0); - inverse_func_pair_test ("(atan(tan(x)) == x", + inverse_func_pair_test ("atan(tan(x)) == x", FUNC(tan), FUNC(atan), 1.0, 0.0); - inverse_func_pair_test ("(tan(atan(x)) == x", + inverse_func_pair_test ("tan(atan(x)) == x", FUNC(atan), FUNC(tan), 1.0, CHOOSE (0, 1e-15L, 0)); - inverse_func_pair_test ("(asinh(sinh(x)) == x", + inverse_func_pair_test ("asinh(sinh(x)) == x", FUNC(sinh), FUNC(asinh), 1.0, CHOOSE (1e-18L, 0, 0)); - inverse_func_pair_test ("(sinh(asinh(x)) == x", + inverse_func_pair_test ("sinh(asinh(x)) == x", FUNC(asinh), FUNC(sinh), 1.0, 0.0); - inverse_func_pair_test ("(acosh(cosh(x)) == x", + inverse_func_pair_test ("acosh(cosh(x)) == x", FUNC(cosh), FUNC(acosh), 1.0, CHOOSE (1e-18L, 1e-15L, 0)); - inverse_func_pair_test ("(cosh(acosh(x)) == x", + inverse_func_pair_test ("cosh(acosh(x)) == x", FUNC(acosh), FUNC(cosh), 1.0, 0.0); - inverse_func_pair_test ("(atanh(tanh(x)) == x", + inverse_func_pair_test ("atanh(tanh(x)) == x", FUNC(tanh), FUNC(atanh), 1.0, CHOOSE (0, 1e-15L, 0)); - inverse_func_pair_test ("(tanh(atanh(x)) == x", + inverse_func_pair_test ("tanh(atanh(x)) == x", FUNC(atanh), FUNC(tanh), 1.0, 0.0); } @@ -1401,6 +1617,15 @@ basic_tests (void) check_bool ("!isinf (-NAN)", !(isinf (-NAN))); check_bool ("NAN != NAN", NAN != NAN); + /* + And again with the value returned by the `nan' function. + */ + check_bool ("isnan (NAN)", isnan (FUNC(nan) (""))); + check_bool ("isnan (-NAN)", isnan (-FUNC(nan) (""))); + check_bool ("!isinf (NAN)", !(isinf (FUNC(nan) ("")))); + check_bool ("!isinf (-NAN)", !(isinf (-FUNC(nan) ("")))); + check_bool ("NAN != NAN", FUNC(nan) ("") != FUNC(nan) ("")); + /* test if EPSILON is ok */ x1 = MATHCONST (1.0); x2 = x1 + CHOOSE (LDBL_EPSILON, DBL_EPSILON, FLT_EPSILON); @@ -1520,6 +1745,16 @@ main (int argc, char *argv[]) fpclassify_test (); hypot_test (); pow_test (); + fdim_test (); + fmin_test (); + fmax_test (); + nextafter_test (); + copysign_test (); + trunc_test (); +#if 0 + /* XXX I'm not sure what is the correct result. */ + remquo_test (); +#endif identities (); inverse_functions (); -- cgit v1.2.3