fpu.h revision 279211
1/*- 2 * Copyright (c) 1990 The Regents of the University of California. 3 * All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * William Jolitz. 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 * 4. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * from: @(#)npx.h 5.3 (Berkeley) 1/18/91 33 * $FreeBSD: stable/10/sys/x86/include/fpu.h 279211 2015-02-23 18:38:41Z jhb $ 34 */ 35 36/* 37 * Floating Point Data Structures and Constants 38 * W. Jolitz 1/90 39 */ 40 41#ifndef _X86_FPU_H_ 42#define _X86_FPU_H_ 43 44/* Environment information of floating point unit. */ 45struct env87 { 46 int32_t en_cw; /* control word (16bits) */ 47 int32_t en_sw; /* status word (16bits) */ 48 int32_t en_tw; /* tag word (16bits) */ 49 int32_t en_fip; /* fp instruction pointer */ 50 uint16_t en_fcs; /* fp code segment selector */ 51 uint16_t en_opcode; /* opcode last executed (11 bits) */ 52 int32_t en_foo; /* fp operand offset */ 53 int32_t en_fos; /* fp operand segment selector */ 54}; 55 56/* Contents of each x87 floating point accumulator. */ 57struct fpacc87 { 58 uint8_t fp_bytes[10]; 59}; 60 61/* Floating point context. (i386 fnsave/frstor) */ 62struct save87 { 63 struct env87 sv_env; /* floating point control/status */ 64 struct fpacc87 sv_ac[8]; /* accumulator contents, 0-7 */ 65 uint8_t sv_pad0[4]; /* saved status word (now unused) */ 66 uint8_t sv_pad[64]; 67}; 68 69/* Contents of each SSE extended accumulator. */ 70struct xmmacc { 71 uint8_t xmm_bytes[16]; 72}; 73 74/* Contents of the upper 16 bytes of each AVX extended accumulator. */ 75struct ymmacc { 76 uint8_t ymm_bytes[16]; 77}; 78 79/* Rename structs below depending on machine architecture. */ 80#ifdef __i386__ 81#define __envxmm32 envxmm 82#else 83#define __envxmm32 envxmm32 84#define __envxmm64 envxmm 85#endif 86 87struct __envxmm32 { 88 uint16_t en_cw; /* control word (16bits) */ 89 uint16_t en_sw; /* status word (16bits) */ 90 uint16_t en_tw; /* tag word (16bits) */ 91 uint16_t en_opcode; /* opcode last executed (11 bits) */ 92 uint32_t en_fip; /* fp instruction pointer */ 93 uint16_t en_fcs; /* fp code segment selector */ 94 uint16_t en_pad0; /* padding */ 95 uint32_t en_foo; /* fp operand offset */ 96 uint16_t en_fos; /* fp operand segment selector */ 97 uint16_t en_pad1; /* padding */ 98 uint32_t en_mxcsr; /* SSE control/status register */ 99 uint32_t en_mxcsr_mask; /* valid bits in mxcsr */ 100}; 101 102struct __envxmm64 { 103 uint16_t en_cw; /* control word (16bits) */ 104 uint16_t en_sw; /* status word (16bits) */ 105 uint8_t en_tw; /* tag word (8bits) */ 106 uint8_t en_zero; 107 uint16_t en_opcode; /* opcode last executed (11 bits ) */ 108 uint64_t en_rip; /* fp instruction pointer */ 109 uint64_t en_rdp; /* fp operand pointer */ 110 uint32_t en_mxcsr; /* SSE control/status register */ 111 uint32_t en_mxcsr_mask; /* valid bits in mxcsr */ 112}; 113 114/* Floating point context. (i386 fxsave/fxrstor) */ 115struct savexmm { 116 struct __envxmm32 sv_env; 117 struct { 118 struct fpacc87 fp_acc; 119 uint8_t fp_pad[6]; /* padding */ 120 } sv_fp[8]; 121 struct xmmacc sv_xmm[8]; 122 uint8_t sv_pad[224]; 123} __aligned(16); 124 125#ifdef __i386__ 126union savefpu { 127 struct save87 sv_87; 128 struct savexmm sv_xmm; 129}; 130#else 131/* Floating point context. (amd64 fxsave/fxrstor) */ 132struct savefpu { 133 struct __envxmm64 sv_env; 134 struct { 135 struct fpacc87 fp_acc; 136 uint8_t fp_pad[6]; /* padding */ 137 } sv_fp[8]; 138 struct xmmacc sv_xmm[16]; 139 uint8_t sv_pad[96]; 140} __aligned(16); 141#endif 142 143struct xstate_hdr { 144 uint64_t xstate_bv; 145 uint64_t xstate_xcomp_bv; 146 uint8_t xstate_rsrv0[8]; 147 uint8_t xstate_rsrv[40]; 148}; 149#define XSTATE_XCOMP_BV_COMPACT (1ULL << 63) 150 151struct savexmm_xstate { 152 struct xstate_hdr sx_hd; 153 struct ymmacc sx_ymm[16]; 154}; 155 156struct savexmm_ymm { 157 struct __envxmm32 sv_env; 158 struct { 159 struct fpacc87 fp_acc; 160 int8_t fp_pad[6]; /* padding */ 161 } sv_fp[8]; 162 struct xmmacc sv_xmm[16]; 163 uint8_t sv_pad[96]; 164 struct savexmm_xstate sv_xstate; 165} __aligned(64); 166 167struct savefpu_xstate { 168 struct xstate_hdr sx_hd; 169 struct ymmacc sx_ymm[16]; 170}; 171 172struct savefpu_ymm { 173 struct __envxmm64 sv_env; 174 struct { 175 struct fpacc87 fp_acc; 176 int8_t fp_pad[6]; /* padding */ 177 } sv_fp[8]; 178 struct xmmacc sv_xmm[16]; 179 uint8_t sv_pad[96]; 180 struct savefpu_xstate sv_xstate; 181} __aligned(64); 182 183#undef __envxmm32 184#undef __envxmm64 185 186/* 187 * The hardware default control word for i387's and later coprocessors is 188 * 0x37F, giving: 189 * 190 * round to nearest 191 * 64-bit precision 192 * all exceptions masked. 193 * 194 * FreeBSD/i386 uses 53 bit precision for things like fadd/fsub/fsqrt etc 195 * because of the difference between memory and fpu register stack arguments. 196 * If its using an intermediate fpu register, it has 80/64 bits to work 197 * with. If it uses memory, it has 64/53 bits to work with. However, 198 * gcc is aware of this and goes to a fair bit of trouble to make the 199 * best use of it. 200 * 201 * This is mostly academic for AMD64, because the ABI prefers the use 202 * SSE2 based math. For FreeBSD/amd64, we go with the default settings. 203 */ 204#define __INITIAL_FPUCW__ 0x037F 205#define __INITIAL_FPUCW_I386__ 0x127F 206#define __INITIAL_NPXCW__ __INITIAL_FPUCW_I386__ 207#define __INITIAL_MXCSR__ 0x1F80 208#define __INITIAL_MXCSR_MASK__ 0xFFBF 209 210/* 211 * The current value of %xcr0 is saved in the sv_pad[] field of the FPU 212 * state in the NT_X86_XSTATE note in core dumps. This offset is chosen 213 * to match the offset used by NT_X86_XSTATE in other systems. 214 */ 215#define X86_XSTATE_XCR0_OFFSET 464 216 217#endif /* !_X86_FPU_H_ */ 218