1 /*- 2 * SPDX-License-Identifier: BSD-4-Clause 3 * 4 * Copyright (c) 2003 Peter Wemm. 5 * Copyright (c) 1990 Andrew Moore, Talke Studio 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 */ 36 37 #ifndef _MACHINE_IEEEFP_H_ 38 #define _MACHINE_IEEEFP_H_ 39 40 /* Deprecated historical FPU control interface */ 41 42 #include <sys/cdefs.h> 43 #include <x86/x86_ieeefp.h> 44 45 /* 46 * IEEE floating point type, constant and function definitions. 47 * XXX: {FP,SSE}*FLD and {FP,SSE}*OFF are undocumented pollution. 48 */ 49 50 /* 51 * SSE mxcsr register bit-field masks. 52 */ 53 #define SSE_STKY_FLD 0x3f /* exception flags */ 54 #define SSE_DAZ_FLD 0x40 /* Denormals are zero */ 55 #define SSE_MSKS_FLD 0x1f80 /* exception masks field */ 56 #define SSE_RND_FLD 0x6000 /* rounding control */ 57 #define SSE_FZ_FLD 0x8000 /* flush to zero on underflow */ 58 59 /* 60 * SSE mxcsr register bit-field offsets (shift counts). 61 */ 62 #define SSE_STKY_OFF 0 /* exception flags offset */ 63 #define SSE_DAZ_OFF 6 /* DAZ exception mask offset */ 64 #define SSE_MSKS_OFF 7 /* other exception masks offset */ 65 #define SSE_RND_OFF 13 /* rounding control offset */ 66 #define SSE_FZ_OFF 15 /* flush to zero offset */ 67 68 /* 69 * General notes about conflicting SSE vs FP status bits. 70 * This code assumes that software will not fiddle with the control 71 * bits of the SSE and x87 in such a way to get them out of sync and 72 * still expect this to work. Break this at your peril. 73 * Because I based this on the i386 port, the x87 state is used for 74 * the fpget*() functions, and is shadowed into the SSE state for 75 * the fpset*() functions. For dual source fpget*() functions, I 76 * merge the two together. I think. 77 */ 78 79 static __inline fp_rnd_t 80 __fpgetround(void) 81 { 82 unsigned short _cw; 83 84 __fnstcw(&_cw); 85 return ((fp_rnd_t)((_cw & FP_RND_FLD) >> FP_RND_OFF)); 86 } 87 88 static __inline fp_rnd_t 89 __fpsetround(fp_rnd_t _m) 90 { 91 fp_rnd_t _p; 92 unsigned _mxcsr; 93 unsigned short _cw, _newcw; 94 95 __fnstcw(&_cw); 96 _p = (fp_rnd_t)((_cw & FP_RND_FLD) >> FP_RND_OFF); 97 _newcw = _cw & ~FP_RND_FLD; 98 _newcw |= (_m << FP_RND_OFF) & FP_RND_FLD; 99 __fnldcw(_cw, _newcw); 100 __stmxcsr(&_mxcsr); 101 _mxcsr &= ~SSE_RND_FLD; 102 _mxcsr |= (_m << SSE_RND_OFF) & SSE_RND_FLD; 103 __ldmxcsr(&_mxcsr); 104 return (_p); 105 } 106 107 /* 108 * Get or set the rounding precision for x87 arithmetic operations. 109 * There is no equivalent SSE mode or control. 110 */ 111 112 static __inline fp_prec_t 113 __fpgetprec(void) 114 { 115 unsigned short _cw; 116 117 __fnstcw(&_cw); 118 return ((fp_prec_t)((_cw & FP_PRC_FLD) >> FP_PRC_OFF)); 119 } 120 121 static __inline fp_prec_t 122 __fpsetprec(fp_prec_t _m) 123 { 124 fp_prec_t _p; 125 unsigned short _cw, _newcw; 126 127 __fnstcw(&_cw); 128 _p = (fp_prec_t)((_cw & FP_PRC_FLD) >> FP_PRC_OFF); 129 _newcw = _cw & ~FP_PRC_FLD; 130 _newcw |= (_m << FP_PRC_OFF) & FP_PRC_FLD; 131 __fnldcw(_cw, _newcw); 132 return (_p); 133 } 134 135 /* 136 * Get or set the exception mask. 137 * Note that the x87 mask bits are inverted by the API -- a mask bit of 1 138 * means disable for x87 and SSE, but for fp*mask() it means enable. 139 */ 140 141 static __inline fp_except_t 142 __fpgetmask(void) 143 { 144 unsigned short _cw; 145 146 __fnstcw(&_cw); 147 return ((~_cw & FP_MSKS_FLD) >> FP_MSKS_OFF); 148 } 149 150 static __inline fp_except_t 151 __fpsetmask(fp_except_t _m) 152 { 153 fp_except_t _p; 154 unsigned _mxcsr; 155 unsigned short _cw, _newcw; 156 157 __fnstcw(&_cw); 158 _p = (~_cw & FP_MSKS_FLD) >> FP_MSKS_OFF; 159 _newcw = _cw & ~FP_MSKS_FLD; 160 _newcw |= (~_m << FP_MSKS_OFF) & FP_MSKS_FLD; 161 __fnldcw(_cw, _newcw); 162 __stmxcsr(&_mxcsr); 163 /* XXX should we clear non-ieee SSE_DAZ_FLD and SSE_FZ_FLD ? */ 164 _mxcsr &= ~SSE_MSKS_FLD; 165 _mxcsr |= (~_m << SSE_MSKS_OFF) & SSE_MSKS_FLD; 166 __ldmxcsr(&_mxcsr); 167 return (_p); 168 } 169 170 static __inline fp_except_t 171 __fpgetsticky(void) 172 { 173 unsigned _ex, _mxcsr; 174 unsigned short _sw; 175 176 __fnstsw(&_sw); 177 _ex = (_sw & FP_STKY_FLD) >> FP_STKY_OFF; 178 __stmxcsr(&_mxcsr); 179 _ex |= (_mxcsr & SSE_STKY_FLD) >> SSE_STKY_OFF; 180 return ((fp_except_t)_ex); 181 } 182 183 #if !defined(__IEEEFP_NOINLINES__) 184 185 #define fpgetmask() __fpgetmask() 186 #define fpgetprec() __fpgetprec() 187 #define fpgetround() __fpgetround() 188 #define fpgetsticky() __fpgetsticky() 189 #define fpsetmask(m) __fpsetmask(m) 190 #define fpsetprec(m) __fpsetprec(m) 191 #define fpsetround(m) __fpsetround(m) 192 193 #else /* __IEEEFP_NOINLINES__ */ 194 195 /* Augment the userland declarations. */ 196 __BEGIN_DECLS 197 extern fp_rnd_t fpgetround(void); 198 extern fp_rnd_t fpsetround(fp_rnd_t); 199 extern fp_except_t fpgetmask(void); 200 extern fp_except_t fpsetmask(fp_except_t); 201 extern fp_except_t fpgetsticky(void); 202 extern fp_except_t fpsetsticky(fp_except_t); 203 fp_prec_t fpgetprec(void); 204 fp_prec_t fpsetprec(fp_prec_t); 205 __END_DECLS 206 207 #endif /* !__IEEEFP_NOINLINES__ */ 208 209 #endif /* !_MACHINE_IEEEFP_H_ */ 210