Current Path : /usr/src/tools/regression/include/tgmath/ |
FreeBSD hs32.drive.ne.jp 9.1-RELEASE FreeBSD 9.1-RELEASE #1: Wed Jan 14 12:18:08 JST 2015 root@hs32.drive.ne.jp:/sys/amd64/compile/hs32 amd64 |
Current File : //usr/src/tools/regression/include/tgmath/tgmath.c |
/*- * Copyright (c) 2004 Stefan Farfeleder <stefanf@FreeBSD.org> * All rights reserved. * * 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 AUTHOR 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 AUTHOR 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. * * $FreeBSD: release/9.1.0/tools/regression/include/tgmath/tgmath.c 137587 2004-11-11 19:47:55Z nik $ */ #include <sys/cdefs.h> /* All of our functions have side effects, __pure2 causes functions calls to * be optimised away. Stop that. */ #undef __pure2 #define __pure2 #include <assert.h> #include <stdio.h> #include <tgmath.h> int n_float, n_double, n_long_double; int n_float_complex, n_double_complex, n_long_double_complex; int currtest = 0; #define TGMACRO(FNC) \ TGMACRO_REAL(FNC) \ TGMACRO_COMPLEX(c ## FNC) #define TGMACRO_REAL(FNC) \ float (FNC ## f)(float x) { n_float++; } \ double (FNC)(double x) { n_double++; } \ long double (FNC ## l)(long double x) { n_long_double++; } #define TGMACRO_REAL_REAL(FNC) \ float (FNC ## f)(float x, float y) { n_float++; } \ double (FNC)(double x, double y) { n_double++; } \ long double \ (FNC ## l)(long double x, long double y) { n_long_double++; } #define TGMACRO_REAL_FIXED_RET(FNC, TYPE) \ TYPE (FNC ## f)(float x) { n_float++; } \ TYPE (FNC)(double x) { n_double++; } \ TYPE (FNC ## l)(long double x) { n_long_double++; } #define TGMACRO_COMPLEX(FNC) \ float complex (FNC ## f)(float complex x) { n_float_complex++; }\ double complex (FNC)(double complex x) { n_double_complex++; } \ long double complex \ (FNC ## l)(long double complex x) { n_long_double_complex++; } #define TGMACRO_COMPLEX_REAL_RET(FNC) \ float (FNC ## f)(float complex x) { n_float_complex++; } \ double (FNC)(double complex x) { n_double_complex++; } \ long double \ (FNC ## l)(long double complex x) { n_long_double_complex++; } /* 7.22#4 */ TGMACRO(acos) TGMACRO(asin) TGMACRO(atan) TGMACRO(acosh) TGMACRO(asinh) TGMACRO(atanh) TGMACRO(cos) TGMACRO(sin) TGMACRO(tan) TGMACRO(cosh) TGMACRO(sinh) TGMACRO(tanh) TGMACRO(exp) TGMACRO(log) TGMACRO_REAL_REAL(pow) float complex (cpowf)(float complex x, float complex y) { n_float_complex++; } double complex (cpow)(double complex x, double complex y) { n_double_complex++; } long double complex (cpowl)(long double complex x, long double complex y) { n_long_double_complex++; } TGMACRO(sqrt) TGMACRO_REAL(fabs) TGMACRO_COMPLEX_REAL_RET(cabs) /* 7.22#5 */ TGMACRO_REAL_REAL(atan2) TGMACRO_REAL(cbrt) TGMACRO_REAL(ceil) TGMACRO_REAL_REAL(copysign) TGMACRO_REAL(erf) TGMACRO_REAL(erfc) TGMACRO_REAL(exp2) TGMACRO_REAL(expm1) TGMACRO_REAL_REAL(fdim) TGMACRO_REAL(floor) float (fmaf)(float x, float y, float z) { n_float++; } double (fma)(double x, double y, double z) { n_double++; } long double (fmal)(long double x, long double y, long double z) { n_long_double++; } TGMACRO_REAL_REAL(fmax) TGMACRO_REAL_REAL(fmin) TGMACRO_REAL_REAL(fmod) float (frexpf)(float x, int *e) { n_float++; } double (frexp)(double x, int *e) { n_double++; } long double (frexpl)(long double x, int *e) { n_long_double++; } TGMACRO_REAL_REAL(hypot) TGMACRO_REAL_FIXED_RET(ilogb, int) float (ldexpf)(float x, int e) { n_float++; } double (ldexp)(double x, int e) { n_double++; } long double (ldexpl)(long double x, int e) { n_long_double++; } TGMACRO_REAL(lgamma) TGMACRO_REAL_FIXED_RET(llrint, long long) TGMACRO_REAL_FIXED_RET(llround, long long) TGMACRO_REAL(log10) TGMACRO_REAL(log1p) TGMACRO_REAL(log2) TGMACRO_REAL(logb) TGMACRO_REAL_FIXED_RET(lrint, long) TGMACRO_REAL_FIXED_RET(lround, long) TGMACRO_REAL(nearbyint) TGMACRO_REAL_REAL(nextafter) float (nexttowardf)(float x, long double y) { n_float++; } double (nexttoward)(double x, long double y) { n_double++; } long double (nexttowardl)(long double x, long double y) { n_long_double++; } TGMACRO_REAL_REAL(remainder) float (remquof)(float x, float y, int *q) { n_float++; } double (remquo)(double x, double y, int *q) { n_double++; } long double (remquol)(long double x, long double y, int *q) { n_long_double++; } TGMACRO_REAL(rint) TGMACRO_REAL(round) float (scalbnf)(float x, int n) { n_float++; } double (scalbn)(double x, int n) { n_double++; } long double (scalbnl)(long double x, int n) { n_long_double++; } float (scalblnf)(float x, long n) { n_float++; } double (scalbln)(double x, long n) { n_double++; } long double (scalblnl)(long double x, long n) { n_long_double++; } TGMACRO_REAL(tgamma) TGMACRO_REAL(trunc) /* 7.22#6 */ TGMACRO_COMPLEX_REAL_RET(carg) TGMACRO_COMPLEX_REAL_RET(cimag) TGMACRO_COMPLEX(conj) TGMACRO_COMPLEX(cproj) TGMACRO_COMPLEX_REAL_RET(creal) long double ld; double d; float f; long double complex ldc; double complex dc; float complex fc; unsigned long long ull; int i; _Bool b; #define SAMETYPE(EXP, TYPE) \ __builtin_types_compatible_p(__typeof__(EXP), TYPE) #define CLEAR_COUNTERS \ (n_float = n_double = n_long_double = 0, \ n_float_complex = n_double_complex = n_long_double_complex = 0, 1) #define RUN_TEST(EXP, TYPE) (EXP, SAMETYPE(EXP, TYPE)) #define PASS_REAL_ARG_REAL_RET(FNC) PASS_REAL_ARG_REAL_RET_(FNC,) #define PASS_REAL_ARG_REAL_RET_(FNC, SUFFIX) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(1.l), long double) && \ RUN_TEST(FNC(ld), long double) && \ n_long_double ## SUFFIX == 2 && \ RUN_TEST(FNC(1.), double) && \ RUN_TEST(FNC(d), double) && \ RUN_TEST(FNC(1ull), double) && \ RUN_TEST(FNC(ull), double) && \ RUN_TEST(FNC(1), double) && \ RUN_TEST(FNC(i), double) && \ RUN_TEST(FNC((_Bool)0), double) && \ RUN_TEST(FNC(b), double) && \ n_double ## SUFFIX == 8 && \ RUN_TEST(FNC(1.f), float) && \ RUN_TEST(FNC(f), float) && \ n_float ## SUFFIX == 2 #define PASS_REAL_ARG_FIXED_RET(FNC, RET) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(1.l), RET) && \ RUN_TEST(FNC(ld), RET) && \ n_long_double == 2 && \ RUN_TEST(FNC(1.), RET) && \ RUN_TEST(FNC(d), RET) && \ RUN_TEST(FNC(1ull), RET) && \ RUN_TEST(FNC(ull), RET) && \ RUN_TEST(FNC(1), RET) && \ RUN_TEST(FNC(i), RET) && \ RUN_TEST(FNC((_Bool)0), RET) && \ RUN_TEST(FNC(b), RET) && \ n_double == 8 && \ RUN_TEST(FNC(1.f), RET) && \ RUN_TEST(FNC(f), RET) && \ n_float == 2 #define PASS_REAL_FIXED_ARG_REAL_RET(FNC, ARG2) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(1.l, ARG2), long double) && \ RUN_TEST(FNC(ld, ARG2), long double) && \ n_long_double == 2 && \ RUN_TEST(FNC(1., ARG2), double) && \ RUN_TEST(FNC(d, ARG2), double) && \ RUN_TEST(FNC(1ull, ARG2), double) && \ RUN_TEST(FNC(ull, ARG2), double) && \ RUN_TEST(FNC(1, ARG2), double) && \ RUN_TEST(FNC(i, ARG2), double) && \ RUN_TEST(FNC((_Bool)0, ARG2), double) && \ RUN_TEST(FNC(b, ARG2), double) && \ n_double == 8 && \ RUN_TEST(FNC(1.f, ARG2), float) && \ RUN_TEST(FNC(f, ARG2), float) && \ n_float == 2 #define PASS_REAL_REAL_ARG_REAL_RET(FNC) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(1.l, 1.l), long double) && \ RUN_TEST(FNC(1.l, 1.), long double) && \ RUN_TEST(FNC(1.l, 1.f), long double) && \ RUN_TEST(FNC(1.l, 1), long double) && \ RUN_TEST(FNC(1.l, (_Bool)0), long double) && \ RUN_TEST(FNC(1.l, ld), long double) && \ RUN_TEST(FNC(1., ld), long double) && \ RUN_TEST(FNC(1.f, ld), long double) && \ RUN_TEST(FNC(1, ld), long double) && \ RUN_TEST(FNC((_Bool)0, ld), long double) && \ n_long_double == 10 && \ RUN_TEST(FNC(d, 1.), double) && \ RUN_TEST(FNC(d, 1.f), double) && \ RUN_TEST(FNC(d, 1l), double) && \ RUN_TEST(FNC(d, (_Bool)0), double) && \ RUN_TEST(FNC(1., 1.), double) && \ RUN_TEST(FNC(1.f, 1.), double) && \ RUN_TEST(FNC(1l, 1.), double) && \ RUN_TEST(FNC((_Bool)0, 1.), double) && \ RUN_TEST(FNC(1ull, f), double) && \ RUN_TEST(FNC(1.f, ull), double) && \ RUN_TEST(FNC(1, 1l), double) && \ RUN_TEST(FNC(1u, i), double) && \ RUN_TEST(FNC((_Bool)0, 1.f), double) && \ RUN_TEST(FNC(1.f, b), double) && \ n_double == 14 && \ RUN_TEST(FNC(1.f, 1.f), float) && \ RUN_TEST(FNC(1.f, 1.f), float) && \ RUN_TEST(FNC(f, 1.f), float) && \ RUN_TEST(FNC(f, f), float) && \ n_float == 4 #define PASS_REAL_REAL_FIXED_ARG_REAL_RET(FNC, ARG3) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(1.l, 1.l, ARG3), long double) && \ RUN_TEST(FNC(1.l, 1., ARG3), long double) && \ RUN_TEST(FNC(1.l, 1.f, ARG3), long double) && \ RUN_TEST(FNC(1.l, 1, ARG3), long double) && \ RUN_TEST(FNC(1.l, (_Bool)0, ARG3), long double) && \ RUN_TEST(FNC(1.l, ld, ARG3), long double) && \ RUN_TEST(FNC(1., ld, ARG3), long double) && \ RUN_TEST(FNC(1.f, ld, ARG3), long double) && \ RUN_TEST(FNC(1, ld, ARG3), long double) && \ RUN_TEST(FNC((_Bool)0, ld, ARG3), long double) && \ n_long_double == 10 && \ RUN_TEST(FNC(d, 1., ARG3), double) && \ RUN_TEST(FNC(d, 1.f, ARG3), double) && \ RUN_TEST(FNC(d, 1l, ARG3), double) && \ RUN_TEST(FNC(d, (_Bool)0, ARG3), double) && \ RUN_TEST(FNC(1., 1., ARG3), double) && \ RUN_TEST(FNC(1.f, 1., ARG3), double) && \ RUN_TEST(FNC(1l, 1., ARG3), double) && \ RUN_TEST(FNC((_Bool)0, 1., ARG3), double) && \ RUN_TEST(FNC(1ull, f, ARG3), double) && \ RUN_TEST(FNC(1.f, ull, ARG3), double) && \ RUN_TEST(FNC(1, 1l, ARG3), double) && \ RUN_TEST(FNC(1u, i, ARG3), double) && \ RUN_TEST(FNC((_Bool)0, 1.f, ARG3), double) && \ RUN_TEST(FNC(1.f, b, ARG3), double) && \ n_double == 14 && \ RUN_TEST(FNC(1.f, 1.f, ARG3), float) && \ RUN_TEST(FNC(1.f, 1.f, ARG3), float) && \ RUN_TEST(FNC(f, 1.f, ARG3), float) && \ RUN_TEST(FNC(f, f, ARG3), float) && \ n_float == 4 #define PASS_REAL_REAL_REAL_ARG_REAL_RET(FNC) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(ld, d, f), long double) && \ RUN_TEST(FNC(1, ld, ld), long double) && \ RUN_TEST(FNC(1, d, ld), long double) && \ n_long_double == 3 && \ RUN_TEST(FNC(1, f, 1.f), double) && \ RUN_TEST(FNC(f, d, 1.f), double) && \ RUN_TEST(FNC(f, 1.f, 1.), double) && \ n_double == 3 && \ RUN_TEST(FNC(f, 1.f, f), float) && \ n_float == 1 #define PASS_REAL_ARG_COMPLEX_RET(FNC) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(1.l), long double complex) && \ RUN_TEST(FNC(ld), long double complex) && \ n_long_double_complex == 2 && \ RUN_TEST(FNC(1.), double complex) && \ RUN_TEST(FNC(d), double complex) && \ RUN_TEST(FNC(1l), double complex) && \ RUN_TEST(FNC(i), double complex) && \ RUN_TEST(FNC(b), double complex) && \ n_double_complex == 5 && \ RUN_TEST(FNC(1.f), float complex) && \ RUN_TEST(FNC(f), float complex) && \ n_float_complex == 2 #define PASS_COMPLEX_ARG_COMPLEX_RET(FNC) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(ldc), long double complex) && \ n_long_double_complex == 1 && \ RUN_TEST(FNC(dc), double complex) && \ n_double_complex == 1 && \ RUN_TEST(FNC(fc), float complex) && \ RUN_TEST(FNC(I), float complex) && \ n_float_complex == 2 #define PASS_COMPLEX_ARG_REAL_RET(FNC) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(ldc), long double) && \ n_long_double_complex == 1 && \ RUN_TEST(FNC(dc), double) && \ n_double_complex == 1 && \ RUN_TEST(FNC(fc), float) && \ RUN_TEST(FNC(I), float) && \ n_float_complex == 2 #define PASS_COMPLEX_COMPLEX_ARG_COMPLEX_RET(FNC) \ CLEAR_COUNTERS && \ RUN_TEST(FNC(ldc, ldc), long double complex) && \ RUN_TEST(FNC(ldc, dc), long double complex) && \ RUN_TEST(FNC(ldc, fc), long double complex) && \ RUN_TEST(FNC(ldc, ld), long double complex) && \ RUN_TEST(FNC(ldc, d), long double complex) && \ RUN_TEST(FNC(ldc, f), long double complex) && \ RUN_TEST(FNC(ldc, i), long double complex) && \ RUN_TEST(FNC(dc, ldc), long double complex) && \ RUN_TEST(FNC(I, ldc), long double complex) && \ RUN_TEST(FNC(1.l, ldc), long double complex) && \ RUN_TEST(FNC(1., ldc), long double complex) && \ RUN_TEST(FNC(1.f, ldc), long double complex) && \ RUN_TEST(FNC(1, ldc), long double complex) && \ RUN_TEST(FNC(ld, dc), long double complex) && \ RUN_TEST(FNC(ld, fc), long double complex) && \ RUN_TEST(FNC(I, 1.l), long double complex) && \ RUN_TEST(FNC(dc, 1.l), long double complex) && \ n_long_double_complex == 17 && \ RUN_TEST(FNC(dc, dc), double complex) && \ RUN_TEST(FNC(dc, fc), double complex) && \ RUN_TEST(FNC(dc, d), double complex) && \ RUN_TEST(FNC(dc, f), double complex) && \ RUN_TEST(FNC(dc, ull), double complex) && \ RUN_TEST(FNC(I, dc), double complex) && \ RUN_TEST(FNC(1., dc), double complex) && \ RUN_TEST(FNC(1, dc), double complex) && \ RUN_TEST(FNC(fc, d), double complex) && \ RUN_TEST(FNC(1, I), double complex) && \ n_double_complex == 10 && \ RUN_TEST(FNC(fc, fc), float complex) && \ RUN_TEST(FNC(fc, I), float complex) && \ RUN_TEST(FNC(1.f, fc), float complex) && \ n_float_complex == 3 int failed = 0; #define PRINT(STR, X) do { \ currtest++; \ int result = (X); \ if (!result) \ failed = 1; \ printf("%s %d - %s\n", result ? "ok" : "not ok", currtest, (STR)); \ fflush(stdout); \ } while (0) int main(void) { printf("1..60\n"); /* 7.22#4 */ PRINT("acos", PASS_REAL_ARG_REAL_RET(acos) && PASS_COMPLEX_ARG_COMPLEX_RET(acos)); PRINT("asin", PASS_REAL_ARG_REAL_RET(asin) && PASS_COMPLEX_ARG_COMPLEX_RET(asin)); PRINT("atan", PASS_REAL_ARG_REAL_RET(atan) && PASS_COMPLEX_ARG_COMPLEX_RET(atan)); PRINT("acosh", PASS_REAL_ARG_REAL_RET(acosh) && PASS_COMPLEX_ARG_COMPLEX_RET(acosh)); PRINT("asinh", PASS_REAL_ARG_REAL_RET(asinh) && PASS_COMPLEX_ARG_COMPLEX_RET(asinh)); PRINT("atanh", PASS_REAL_ARG_REAL_RET(atanh) && PASS_COMPLEX_ARG_COMPLEX_RET(atanh)); PRINT("cos", PASS_REAL_ARG_REAL_RET(cos) && PASS_COMPLEX_ARG_COMPLEX_RET(cos)); PRINT("sin", PASS_REAL_ARG_REAL_RET(sin) && PASS_COMPLEX_ARG_COMPLEX_RET(sin)); PRINT("tan", PASS_REAL_ARG_REAL_RET(tan) && PASS_COMPLEX_ARG_COMPLEX_RET(tan)); PRINT("cosh", PASS_REAL_ARG_REAL_RET(cosh) && PASS_COMPLEX_ARG_COMPLEX_RET(cosh)); PRINT("sinh", PASS_REAL_ARG_REAL_RET(sinh) && PASS_COMPLEX_ARG_COMPLEX_RET(sinh)); PRINT("tanh", PASS_REAL_ARG_REAL_RET(tanh) && PASS_COMPLEX_ARG_COMPLEX_RET(tanh)); PRINT("exp", PASS_REAL_ARG_REAL_RET(exp) && PASS_COMPLEX_ARG_COMPLEX_RET(exp)); PRINT("log", PASS_REAL_ARG_REAL_RET(log) && PASS_COMPLEX_ARG_COMPLEX_RET(log)); PRINT("pow", PASS_REAL_REAL_ARG_REAL_RET(pow) && PASS_COMPLEX_COMPLEX_ARG_COMPLEX_RET(pow)); PRINT("sqrt", PASS_REAL_ARG_REAL_RET(sqrt) && PASS_COMPLEX_ARG_COMPLEX_RET(sqrt)); PRINT("fabs", PASS_REAL_ARG_REAL_RET(fabs) && PASS_COMPLEX_ARG_REAL_RET(fabs)); /* 7.22#5 */ PRINT("atan2", PASS_REAL_REAL_ARG_REAL_RET(atan2)); PRINT("cbrt", PASS_REAL_ARG_REAL_RET(cbrt)); PRINT("ceil", PASS_REAL_ARG_REAL_RET(ceil)); PRINT("copysign", PASS_REAL_REAL_ARG_REAL_RET(copysign)); PRINT("erf", PASS_REAL_ARG_REAL_RET(erf)); PRINT("erfc", PASS_REAL_ARG_REAL_RET(erfc)); PRINT("exp2", PASS_REAL_ARG_REAL_RET(exp2)); PRINT("expm1", PASS_REAL_ARG_REAL_RET(expm1)); PRINT("fdim", PASS_REAL_REAL_ARG_REAL_RET(fdim)); PRINT("floor", PASS_REAL_ARG_REAL_RET(floor)); PRINT("fma", PASS_REAL_REAL_REAL_ARG_REAL_RET(fma)); PRINT("fmax", PASS_REAL_REAL_ARG_REAL_RET(fmax)); PRINT("fmin", PASS_REAL_REAL_ARG_REAL_RET(fmin)); PRINT("fmod", PASS_REAL_REAL_ARG_REAL_RET(fmod)); PRINT("frexp", PASS_REAL_FIXED_ARG_REAL_RET(frexp, &i)); PRINT("hypot", PASS_REAL_REAL_ARG_REAL_RET(hypot)); PRINT("ilogb", PASS_REAL_ARG_FIXED_RET(ilogb, int)); PRINT("ldexp", PASS_REAL_FIXED_ARG_REAL_RET(ldexp, 1) && PASS_REAL_FIXED_ARG_REAL_RET(ldexp, ld) && PASS_REAL_FIXED_ARG_REAL_RET(ldexp, ldc)); PRINT("lgamma", PASS_REAL_ARG_REAL_RET(lgamma)); PRINT("llrint", PASS_REAL_ARG_FIXED_RET(llrint, long long)); PRINT("llround", PASS_REAL_ARG_FIXED_RET(llround, long long)); PRINT("log10", PASS_REAL_ARG_REAL_RET(log10)); PRINT("log1p", PASS_REAL_ARG_REAL_RET(log1p)); PRINT("log2", PASS_REAL_ARG_REAL_RET(log2)); PRINT("logb", PASS_REAL_ARG_REAL_RET(logb)); PRINT("lrint", PASS_REAL_ARG_FIXED_RET(lrint, long)); PRINT("lround", PASS_REAL_ARG_FIXED_RET(lround, long)); PRINT("nearbyint", PASS_REAL_ARG_REAL_RET(nearbyint)); PRINT("nextafter", PASS_REAL_REAL_ARG_REAL_RET(nextafter)); PRINT("nexttoward", PASS_REAL_FIXED_ARG_REAL_RET(nexttoward, 1) && PASS_REAL_FIXED_ARG_REAL_RET(nexttoward, ull) && PASS_REAL_FIXED_ARG_REAL_RET(nexttoward, d) && PASS_REAL_FIXED_ARG_REAL_RET(nexttoward, fc)); PRINT("remainder", PASS_REAL_REAL_ARG_REAL_RET(remainder)); PRINT("remquo", PASS_REAL_REAL_FIXED_ARG_REAL_RET(remquo, &i)); PRINT("rint", PASS_REAL_ARG_REAL_RET(rint)); PRINT("round", PASS_REAL_ARG_REAL_RET(round)); PRINT("scalbn", PASS_REAL_FIXED_ARG_REAL_RET(scalbn, 1) && PASS_REAL_FIXED_ARG_REAL_RET(scalbn, b) && PASS_REAL_FIXED_ARG_REAL_RET(scalbn, I)); PRINT("scalbln", PASS_REAL_FIXED_ARG_REAL_RET(scalbln, i) && PASS_REAL_FIXED_ARG_REAL_RET(scalbln, 1.l) && PASS_REAL_FIXED_ARG_REAL_RET(scalbln, dc)); PRINT("tgamma", PASS_REAL_ARG_REAL_RET(tgamma)); PRINT("trunc", PASS_REAL_ARG_REAL_RET(trunc)); /* 7.22#6 */ PRINT("carg", PASS_REAL_ARG_REAL_RET_(carg, _complex) && PASS_COMPLEX_ARG_REAL_RET(carg)); PRINT("cimag", PASS_REAL_ARG_REAL_RET_(cimag, _complex) && PASS_COMPLEX_ARG_REAL_RET(cimag)); PRINT("conj", PASS_REAL_ARG_COMPLEX_RET(conj) && PASS_COMPLEX_ARG_COMPLEX_RET(conj)); PRINT("cproj", PASS_REAL_ARG_COMPLEX_RET(cproj) && PASS_COMPLEX_ARG_COMPLEX_RET(cproj)); PRINT("creal", PASS_REAL_ARG_REAL_RET_(creal, _complex) && PASS_COMPLEX_ARG_REAL_RET(creal)); }