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 #include <sys/cdefs.h> 30 #include <sys/types.h> 31 #include <machine/npx.h> 32 #include "fenv.h" 33 34 const fenv_t __fe_dfl_env = { 35 __INITIAL_NPXCW__, 36 0x0000, 37 0x0000, 38 0x1f80, 39 0xffffffff, 40 { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 41 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xff, 0xff } 42 }; 43 44 enum __sse_support __has_sse = 45 #ifdef __SSE__ 46 __SSE_YES; 47 #else 48 __SSE_UNK; 49 #endif 50 51 #define getfl(x) __asm __volatile("pushfl\n\tpopl %0" : "=mr" (*(x))) 52 #define setfl(x) __asm __volatile("pushl %0\n\tpopfl" : : "g" (x)) 53 #define cpuid_dx(x) __asm __volatile("pushl %%ebx\n\tmovl $1, %%eax\n\t" \ 54 "cpuid\n\tpopl %%ebx" \ 55 : "=d" (*(x)) : : "eax", "ecx") 56 57 /* 58 * Test for SSE support on this processor. We need to do this because 59 * we need to use ldmxcsr/stmxcsr to get correct results if any part 60 * of the program was compiled to use SSE floating-point, but we can't 61 * use SSE on older processors. 62 */ 63 int 64 __test_sse(void) 65 { 66 int flag, nflag; 67 int dx_features; 68 69 /* Am I a 486? */ 70 getfl(&flag); 71 nflag = flag ^ 0x200000; 72 setfl(nflag); 73 getfl(&nflag); 74 if (flag != nflag) { 75 /* Not a 486, so CPUID should work. */ 76 cpuid_dx(&dx_features); 77 if (dx_features & 0x2000000) { 78 __has_sse = __SSE_YES; 79 return (1); 80 } 81 } 82 __has_sse = __SSE_NO; 83 return (0); 84 } 85 86 int 87 fesetexceptflag(const fexcept_t *flagp, int excepts) 88 { 89 fenv_t env; 90 __uint32_t mxcsr; 91 92 __fnstenv(&env); 93 env.__status &= ~excepts; 94 env.__status |= *flagp & excepts; 95 __fldenv(env); 96 97 if (__HAS_SSE()) { 98 __stmxcsr(&mxcsr); 99 mxcsr &= ~excepts; 100 mxcsr |= *flagp & excepts; 101 __ldmxcsr(mxcsr); 102 } 103 104 return (0); 105 } 106 107 int 108 feraiseexcept(int excepts) 109 { 110 fexcept_t ex = excepts; 111 112 fesetexceptflag(&ex, excepts); 113 __fwait(); 114 return (0); 115 } 116 117 int 118 fegetenv(fenv_t *envp) 119 { 120 __uint32_t mxcsr; 121 122 __fnstenv(envp); 123 /* 124 * fnstenv masks all exceptions, so we need to restore 125 * the old control word to avoid this side effect. 126 */ 127 __fldcw(envp->__control); 128 if (__HAS_SSE()) { 129 __stmxcsr(&mxcsr); 130 __set_mxcsr(*envp, mxcsr); 131 } 132 return (0); 133 } 134 135 int 136 feholdexcept(fenv_t *envp) 137 { 138 __uint32_t mxcsr; 139 140 __fnstenv(envp); 141 __fnclex(); 142 if (__HAS_SSE()) { 143 __stmxcsr(&mxcsr); 144 __set_mxcsr(*envp, mxcsr); 145 mxcsr &= ~FE_ALL_EXCEPT; 146 mxcsr |= FE_ALL_EXCEPT << _SSE_EMASK_SHIFT; 147 __ldmxcsr(mxcsr); 148 } 149 return (0); 150 } 151 152 int 153 feupdateenv(const fenv_t *envp) 154 { 155 __uint32_t mxcsr; 156 __uint16_t status; 157 158 __fnstsw(&status); 159 if (__HAS_SSE()) 160 __stmxcsr(&mxcsr); 161 else 162 mxcsr = 0; 163 fesetenv(envp); 164 feraiseexcept((mxcsr | status) & FE_ALL_EXCEPT); 165 return (0); 166 } 167 168 int 169 __feenableexcept(int mask) 170 { 171 __uint32_t mxcsr, omask; 172 __uint16_t control; 173 174 mask &= FE_ALL_EXCEPT; 175 __fnstcw(&control); 176 if (__HAS_SSE()) 177 __stmxcsr(&mxcsr); 178 else 179 mxcsr = 0; 180 omask = (control | mxcsr >> _SSE_EMASK_SHIFT) & FE_ALL_EXCEPT; 181 control &= ~mask; 182 __fldcw(control); 183 if (__HAS_SSE()) { 184 mxcsr &= ~(mask << _SSE_EMASK_SHIFT); 185 __ldmxcsr(mxcsr); 186 } 187 return (~omask); 188 } 189 190 int 191 __fedisableexcept(int mask) 192 { 193 __uint32_t mxcsr, omask; 194 __uint16_t control; 195 196 mask &= FE_ALL_EXCEPT; 197 __fnstcw(&control); 198 if (__HAS_SSE()) 199 __stmxcsr(&mxcsr); 200 else 201 mxcsr = 0; 202 omask = (control | mxcsr >> _SSE_EMASK_SHIFT) & FE_ALL_EXCEPT; 203 control |= mask; 204 __fldcw(control); 205 if (__HAS_SSE()) { 206 mxcsr |= mask << _SSE_EMASK_SHIFT; 207 __ldmxcsr(mxcsr); 208 } 209 return (~omask); 210 } 211 212 __weak_reference(__feenableexcept, feenableexcept); 213 __weak_reference(__fedisableexcept, fedisableexcept); 214