|
| 1 | +/*- |
| 2 | + * Copyright (c) 2004-2005 David Schultz <das@FreeBSD.ORG> |
| 3 | + * All rights reserved. |
| 4 | + * |
| 5 | + * Redistribution and use in source and binary forms, with or without |
| 6 | + * modification, are permitted provided that the following conditions |
| 7 | + * are met: |
| 8 | + * 1. Redistributions of source code must retain the above copyright |
| 9 | + * notice, this list of conditions and the following disclaimer. |
| 10 | + * 2. Redistributions in binary form must reproduce the above copyright |
| 11 | + * notice, this list of conditions and the following disclaimer in the |
| 12 | + * documentation and/or other materials provided with the distribution. |
| 13 | + * |
| 14 | + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND |
| 15 | + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 16 | + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 17 | + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
| 18 | + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 19 | + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 20 | + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 21 | + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 22 | + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 23 | + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 24 | + * SUCH DAMAGE. |
| 25 | + * |
| 26 | + * $FreeBSD: src/lib/msun/amd64/fenv.h,v 1.8 2011/10/10 15:43:09 das Exp $ |
| 27 | + */ |
| 28 | + |
| 29 | +#ifndef _FENV_H_ |
| 30 | +#define _FENV_H_ |
| 31 | + |
| 32 | +#include "cdefs-compat.h" |
| 33 | +#include "types-compat.h" |
| 34 | + |
| 35 | +#include "math_private.h" |
| 36 | + |
| 37 | +#ifndef __fenv_static |
| 38 | +#define __fenv_static static |
| 39 | +#endif |
| 40 | + |
| 41 | +typedef struct { |
| 42 | + struct { |
| 43 | + uint32_t __control; |
| 44 | + uint32_t __status; |
| 45 | + uint32_t __tag; |
| 46 | + char __other[16]; |
| 47 | + } __x87; |
| 48 | + uint32_t __mxcsr; |
| 49 | +} fenv_t; |
| 50 | + |
| 51 | +typedef uint16_t fexcept_t; |
| 52 | + |
| 53 | +/* Exception flags */ |
| 54 | +#define FE_INVALID 0x01 |
| 55 | +#define FE_DENORMAL 0x02 |
| 56 | +#define FE_DIVBYZERO 0x04 |
| 57 | +#define FE_OVERFLOW 0x08 |
| 58 | +#define FE_UNDERFLOW 0x10 |
| 59 | +#define FE_INEXACT 0x20 |
| 60 | +#define FE_ALL_EXCEPT (FE_DIVBYZERO | FE_DENORMAL | FE_INEXACT | \ |
| 61 | + FE_INVALID | FE_OVERFLOW | FE_UNDERFLOW) |
| 62 | + |
| 63 | +/* Rounding modes */ |
| 64 | +#define FE_TONEAREST 0x0000 |
| 65 | +#define FE_DOWNWARD 0x0400 |
| 66 | +#define FE_UPWARD 0x0800 |
| 67 | +#define FE_TOWARDZERO 0x0c00 |
| 68 | +#define _ROUND_MASK (FE_TONEAREST | FE_DOWNWARD | \ |
| 69 | + FE_UPWARD | FE_TOWARDZERO) |
| 70 | + |
| 71 | +/* |
| 72 | + * As compared to the x87 control word, the SSE unit's control word |
| 73 | + * has the rounding control bits offset by 3 and the exception mask |
| 74 | + * bits offset by 7. |
| 75 | + */ |
| 76 | +#define _SSE_ROUND_SHIFT 3 |
| 77 | +#define _SSE_EMASK_SHIFT 7 |
| 78 | + |
| 79 | +__BEGIN_DECLS |
| 80 | + |
| 81 | +/* Default floating-point environment */ |
| 82 | +extern const fenv_t __fe_dfl_env; |
| 83 | +#define FE_DFL_ENV (&__fe_dfl_env) |
| 84 | + |
| 85 | +#define __fldcw(__cw) __asm __volatile("fldcw %0" : : "m" (__cw)) |
| 86 | +#define __fldenv(__env) __asm __volatile("fldenv %0" : : "m" (__env)) |
| 87 | +#define __fldenvx(__env) __asm __volatile("fldenv %0" : : "m" (__env) \ |
| 88 | + : "st", "st(1)", "st(2)", "st(3)", "st(4)", \ |
| 89 | + "st(5)", "st(6)", "st(7)") |
| 90 | +#define __fnclex() __asm __volatile("fnclex") |
| 91 | +#define __fnstenv(__env) __asm __volatile("fnstenv %0" : "=m" (*(__env))) |
| 92 | +#define __fnstcw(__cw) __asm __volatile("fnstcw %0" : "=m" (*(__cw))) |
| 93 | +#define __fnstsw(__sw) __asm __volatile("fnstsw %0" : "=am" (*(__sw))) |
| 94 | +#define __fwait() __asm __volatile("fwait") |
| 95 | +#define __ldmxcsr(__csr) __asm __volatile("ldmxcsr %0" : : "m" (__csr)) |
| 96 | +#define __stmxcsr(__csr) __asm __volatile("stmxcsr %0" : "=m" (*(__csr))) |
| 97 | + |
| 98 | +__fenv_static __attribute__((always_inline)) DLLEXPORT inline int |
| 99 | +feclearexcept(int __excepts) |
| 100 | +{ |
| 101 | + fenv_t __env; |
| 102 | + |
| 103 | + if (__excepts == FE_ALL_EXCEPT) { |
| 104 | + __fnclex(); |
| 105 | + } else { |
| 106 | + __fnstenv(&__env.__x87); |
| 107 | + __env.__x87.__status &= ~__excepts; |
| 108 | + __fldenv(__env.__x87); |
| 109 | + } |
| 110 | + __stmxcsr(&__env.__mxcsr); |
| 111 | + __env.__mxcsr &= ~__excepts; |
| 112 | + __ldmxcsr(__env.__mxcsr); |
| 113 | + return (0); |
| 114 | +} |
| 115 | + |
| 116 | +__fenv_static DLLEXPORT inline int |
| 117 | +fegetexceptflag(fexcept_t *__flagp, int __excepts) |
| 118 | +{ |
| 119 | + uint32_t __mxcsr; |
| 120 | + uint16_t __status; |
| 121 | + |
| 122 | + __stmxcsr(&__mxcsr); |
| 123 | + __fnstsw(&__status); |
| 124 | + *__flagp = (__mxcsr | __status) & __excepts; |
| 125 | + return (0); |
| 126 | +} |
| 127 | + |
| 128 | +DLLEXPORT int fesetexceptflag(const fexcept_t *__flagp, int __excepts); |
| 129 | +DLLEXPORT int feraiseexcept(int __excepts); |
| 130 | + |
| 131 | +__fenv_static __attribute__((always_inline)) DLLEXPORT inline int |
| 132 | +fetestexcept(int __excepts) |
| 133 | +{ |
| 134 | + uint32_t __mxcsr; |
| 135 | + uint16_t __status; |
| 136 | + |
| 137 | + __stmxcsr(&__mxcsr); |
| 138 | + __fnstsw(&__status); |
| 139 | + return ((__status | __mxcsr) & __excepts); |
| 140 | +} |
| 141 | + |
| 142 | +__fenv_static DLLEXPORT inline int |
| 143 | +fegetround(void) |
| 144 | +{ |
| 145 | + uint16_t __control; |
| 146 | + |
| 147 | + /* |
| 148 | + * We assume that the x87 and the SSE unit agree on the |
| 149 | + * rounding mode. Reading the control word on the x87 turns |
| 150 | + * out to be about 5 times faster than reading it on the SSE |
| 151 | + * unit on an Opteron 244. |
| 152 | + */ |
| 153 | + __fnstcw(&__control); |
| 154 | + return (__control & _ROUND_MASK); |
| 155 | +} |
| 156 | + |
| 157 | +__fenv_static DLLEXPORT inline int |
| 158 | +fesetround(int __round) |
| 159 | +{ |
| 160 | + uint32_t __mxcsr; |
| 161 | + uint16_t __control; |
| 162 | + |
| 163 | + if (__round & ~_ROUND_MASK) |
| 164 | + return (-1); |
| 165 | + |
| 166 | + __fnstcw(&__control); |
| 167 | + __control &= ~_ROUND_MASK; |
| 168 | + __control |= __round; |
| 169 | + __fldcw(__control); |
| 170 | + |
| 171 | + __stmxcsr(&__mxcsr); |
| 172 | + __mxcsr &= ~(_ROUND_MASK << _SSE_ROUND_SHIFT); |
| 173 | + __mxcsr |= __round << _SSE_ROUND_SHIFT; |
| 174 | + __ldmxcsr(__mxcsr); |
| 175 | + |
| 176 | + return (0); |
| 177 | +} |
| 178 | + |
| 179 | +DLLEXPORT int fegetenv(fenv_t *__envp); |
| 180 | +DLLEXPORT int feholdexcept(fenv_t *__envp); |
| 181 | + |
| 182 | +__fenv_static DLLEXPORT inline int |
| 183 | +fesetenv(const fenv_t *__envp) |
| 184 | +{ |
| 185 | + |
| 186 | + /* |
| 187 | + * XXX Using fldenvx() instead of fldenv() tells the compiler that this |
| 188 | + * instruction clobbers the i387 register stack. This happens because |
| 189 | + * we restore the tag word from the saved environment. Normally, this |
| 190 | + * would happen anyway and we wouldn't care, because the ABI allows |
| 191 | + * function calls to clobber the i387 regs. However, fesetenv() is |
| 192 | + * inlined, so we need to be more careful. |
| 193 | + */ |
| 194 | + __fldenvx(__envp->__x87); |
| 195 | + __ldmxcsr(__envp->__mxcsr); |
| 196 | + return (0); |
| 197 | +} |
| 198 | + |
| 199 | +DLLEXPORT int feupdateenv(const fenv_t *__envp); |
| 200 | + |
| 201 | +#if __BSD_VISIBLE |
| 202 | + |
| 203 | +DLLEXPORT int feenableexcept(int __mask); |
| 204 | +DLLEXPORT int fedisableexcept(int __mask); |
| 205 | + |
| 206 | +/* We currently provide no external definition of fegetexcept(). */ |
| 207 | +static inline DLLEXPORT int |
| 208 | +fegetexcept(void) |
| 209 | +{ |
| 210 | + uint16_t __control; |
| 211 | + |
| 212 | + /* |
| 213 | + * We assume that the masks for the x87 and the SSE unit are |
| 214 | + * the same. |
| 215 | + */ |
| 216 | + __fnstcw(&__control); |
| 217 | + return (~__control & FE_ALL_EXCEPT); |
| 218 | +} |
| 219 | + |
| 220 | +#endif /* __BSD_VISIBLE */ |
| 221 | + |
| 222 | +__END_DECLS |
| 223 | + |
| 224 | +#endif /* !_FENV_H_ */ |
0 commit comments