| | 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 | * $FreeBSD$ |
| | 27 | */ |
| | 28 | |
| | 29 | #ifndef _FENV_H_ |
| | 30 | #define _FENV_H_ |
| | 31 | |
| | 32 | #include <sys/cdefs.h> |
| | 33 | #include <sys/_types.h> |
| | 34 | |
| | 35 | typedef struct { |
| | 36 | struct { |
| | 37 | __uint32_t __control; |
| | 38 | __uint32_t __status; |
| | 39 | __uint32_t __tag; |
| | 40 | char __other[16]; |
| | 41 | } __x87; |
| | 42 | __uint32_t __mxcsr; |
| | 43 | } fenv_t; |
| | 44 | |
| | 45 | typedef __uint16_t fexcept_t; |
| | 46 | |
| | 47 | /* Exception flags */ |
| | 48 | #define FE_INVALID 0x01 |
| | 49 | #define FE_DENORMAL 0x02 |
| | 50 | #define FE_DIVBYZERO 0x04 |
| | 51 | #define FE_OVERFLOW 0x08 |
| | 52 | #define FE_UNDERFLOW 0x10 |
| | 53 | #define FE_INEXACT 0x20 |
| | 54 | #define FE_ALL_EXCEPT (FE_DIVBYZERO | FE_DENORMAL | FE_INEXACT | \ |
| | 55 | FE_INVALID | FE_OVERFLOW | FE_UNDERFLOW) |
| | 56 | |
| | 57 | /* Rounding modes */ |
| | 58 | #define FE_TONEAREST 0x0000 |
| | 59 | #define FE_DOWNWARD 0x0400 |
| | 60 | #define FE_UPWARD 0x0800 |
| | 61 | #define FE_TOWARDZERO 0x0c00 |
| | 62 | #define _ROUND_MASK (FE_TONEAREST | FE_DOWNWARD | \ |
| | 63 | FE_UPWARD | FE_TOWARDZERO) |
| | 64 | |
| | 65 | /* |
| | 66 | * As compared to the x87 control word, the SSE unit's control word |
| | 67 | * has the rounding control bits offset by 3 and the exception mask |
| | 68 | * bits offset by 7. |
| | 69 | */ |
| | 70 | #define _SSE_ROUND_SHIFT 3 |
| | 71 | #define _SSE_EMASK_SHIFT 7 |
| | 72 | |
| | 73 | __BEGIN_DECLS |
| | 74 | |
| | 75 | /* Default floating-point environment */ |
| | 76 | extern const fenv_t __fe_dfl_env; |
| | 77 | #define FE_DFL_ENV (&__fe_dfl_env) |
| | 78 | |
| | 79 | #define __fldcw(__cw) __asm __volatile("fldcw %0" : : "m" (__cw)) |
| | 80 | #define __fldenv(__env) __asm __volatile("fldenv %0" : : "m" (__env)) |
| | 81 | #define __fldenvx(__env) __asm __volatile("fldenv %0" : : "m" (__env) \ |
| | 82 | : "st", "st(1)", "st(2)", "st(3)", "st(4)", \ |
| | 83 | "st(5)", "st(6)", "st(7)") |
| | 84 | #define __fnclex() __asm __volatile("fnclex") |
| | 85 | #define __fnstenv(__env) __asm __volatile("fnstenv %0" : "=m" (*(__env))) |
| | 86 | #define __fnstcw(__cw) __asm __volatile("fnstcw %0" : "=m" (*(__cw))) |
| | 87 | #define __fnstsw(__sw) __asm __volatile("fnstsw %0" : "=am" (*(__sw))) |
| | 88 | #define __fwait() __asm __volatile("fwait") |
| | 89 | #define __ldmxcsr(__csr) __asm __volatile("ldmxcsr %0" : : "m" (__csr)) |
| | 90 | #define __stmxcsr(__csr) __asm __volatile("stmxcsr %0" : "=m" (*(__csr))) |
| | 91 | |
| | 92 | static __inline int |
| | 93 | feclearexcept(int __excepts) |
| | 94 | { |
| | 95 | fenv_t __env; |
| | 96 | |
| | 97 | if (__excepts == FE_ALL_EXCEPT) { |
| | 98 | __fnclex(); |
| | 99 | } else { |
| | 100 | __fnstenv(&__env.__x87); |
| | 101 | __env.__x87.__status &= ~__excepts; |
| | 102 | __fldenv(__env.__x87); |
| | 103 | } |
| | 104 | __stmxcsr(&__env.__mxcsr); |
| | 105 | __env.__mxcsr &= ~__excepts; |
| | 106 | __ldmxcsr(__env.__mxcsr); |
| | 107 | return (0); |
| | 108 | } |
| | 109 | |
| | 110 | static __inline int |
| | 111 | fegetexceptflag(fexcept_t *__flagp, int __excepts) |
| | 112 | { |
| | 113 | int __mxcsr, __status; |
| | 114 | |
| | 115 | __stmxcsr(&__mxcsr); |
| | 116 | __fnstsw(&__status); |
| | 117 | *__flagp = (__mxcsr | __status) & __excepts; |
| | 118 | return (0); |
| | 119 | } |
| | 120 | |
| | 121 | int fesetexceptflag(const fexcept_t *__flagp, int __excepts); |
| | 122 | int feraiseexcept(int __excepts); |
| | 123 | |
| | 124 | static __inline int |
| | 125 | fetestexcept(int __excepts) |
| | 126 | { |
| | 127 | int __mxcsr, __status; |
| | 128 | |
| | 129 | __stmxcsr(&__mxcsr); |
| | 130 | __fnstsw(&__status); |
| | 131 | return ((__status | __mxcsr) & __excepts); |
| | 132 | } |
| | 133 | |
| | 134 | static __inline int |
| | 135 | fegetround(void) |
| | 136 | { |
| | 137 | int __control; |
| | 138 | |
| | 139 | /* |
| | 140 | * We assume that the x87 and the SSE unit agree on the |
| | 141 | * rounding mode. Reading the control word on the x87 turns |
| | 142 | * out to be about 5 times faster than reading it on the SSE |
| | 143 | * unit on an Opteron 244. |
| | 144 | */ |
| | 145 | __fnstcw(&__control); |
| | 146 | return (__control & _ROUND_MASK); |
| | 147 | } |
| | 148 | |
| | 149 | static __inline int |
| | 150 | fesetround(int __round) |
| | 151 | { |
| | 152 | int __mxcsr, __control; |
| | 153 | |
| | 154 | if (__round & ~_ROUND_MASK) |
| | 155 | return (-1); |
| | 156 | |
| | 157 | __fnstcw(&__control); |
| | 158 | __control &= ~_ROUND_MASK; |
| | 159 | __control |= __round; |
| | 160 | __fldcw(__control); |
| | 161 | |
| | 162 | __stmxcsr(&__mxcsr); |
| | 163 | __mxcsr &= ~(_ROUND_MASK << _SSE_ROUND_SHIFT); |
| | 164 | __mxcsr |= __round << _SSE_ROUND_SHIFT; |
| | 165 | __ldmxcsr(__mxcsr); |
| | 166 | |
| | 167 | return (0); |
| | 168 | } |
| | 169 | |
| | 170 | int fegetenv(fenv_t *__envp); |
| | 171 | int feholdexcept(fenv_t *__envp); |
| | 172 | |
| | 173 | static __inline int |
| | 174 | fesetenv(const fenv_t *__envp) |
| | 175 | { |
| | 176 | |
| | 177 | /* |
| | 178 | * XXX Using fldenvx() instead of fldenv() tells the compiler that this |
| | 179 | * instruction clobbers the i387 register stack. This happens because |
| | 180 | * we restore the tag word from the saved environment. Normally, this |
| | 181 | * would happen anyway and we wouldn't care, because the ABI allows |
| | 182 | * function calls to clobber the i387 regs. However, fesetenv() is |
| | 183 | * inlined, so we need to be more careful. |
| | 184 | */ |
| | 185 | __fldenvx(__envp->__x87); |
| | 186 | __ldmxcsr(__envp->__mxcsr); |
| | 187 | return (0); |
| | 188 | } |
| | 189 | |
| | 190 | int feupdateenv(const fenv_t *__envp); |
| | 191 | |
| | 192 | #if __BSD_VISIBLE |
| | 193 | |
| | 194 | int feenableexcept(int __mask); |
| | 195 | int fedisableexcept(int __mask); |
| | 196 | |
| | 197 | static __inline int |
| | 198 | fegetexcept(void) |
| | 199 | { |
| | 200 | int __control; |
| | 201 | |
| | 202 | /* |
| | 203 | * We assume that the masks for the x87 and the SSE unit are |
| | 204 | * the same. |
| | 205 | */ |
| | 206 | __fnstcw(&__control); |
| | 207 | return (~__control & FE_ALL_EXCEPT); |
| | 208 | } |
| | 209 | |
| | 210 | #endif /* __BSD_VISIBLE */ |
| | 211 | |
| | 212 | __END_DECLS |
| | 213 | |
| | 214 | #endif /* !_FENV_H_ */ |
