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 * $FreeBSD$ 34 */ 35 36 /* 37 * 287/387 NPX Coprocessor Data Structures and Constants 38 * W. Jolitz 1/90 39 */ 40 41 #ifndef _MACHINE_NPX_H_ 42 #define _MACHINE_NPX_H_ 43 44 /* Environment information of floating point unit */ 45 struct env87 { 46 long en_cw; /* control word (16bits) */ 47 long en_sw; /* status word (16bits) */ 48 long en_tw; /* tag word (16bits) */ 49 long en_fip; /* floating point instruction pointer */ 50 u_short en_fcs; /* floating code segment selector */ 51 u_short en_opcode; /* opcode last executed (11 bits ) */ 52 long en_foo; /* floating operand offset */ 53 long en_fos; /* floating operand segment selector */ 54 }; 55 56 /* Contents of each floating point accumulator */ 57 struct fpacc87 { 58 #ifdef dontdef /* too unportable */ 59 u_long fp_mantlo; /* mantissa low (31:0) */ 60 u_long fp_manthi; /* mantissa high (63:32) */ 61 int fp_exp:15; /* exponent */ 62 int fp_sgn:1; /* mantissa sign */ 63 #else 64 u_char fp_bytes[10]; 65 #endif 66 }; 67 68 /* Floating point context */ 69 struct save87 { 70 struct env87 sv_env; /* floating point control/status */ 71 struct fpacc87 sv_ac[8]; /* accumulator contents, 0-7 */ 72 u_char sv_pad0[4]; /* padding for (now unused) saved status word */ 73 /* 74 * Bogus padding for emulators. Emulators should use their own 75 * struct and arrange to store into this struct (ending here) 76 * before it is inspected for ptracing or for core dumps. Some 77 * emulators overwrite the whole struct. We have no good way of 78 * knowing how much padding to leave. Leave just enough for the 79 * GPL emulator's i387_union (176 bytes total). 80 */ 81 u_char sv_pad[64]; /* padding; used by emulators */ 82 }; 83 84 struct envxmm { 85 u_int16_t en_cw; /* control word (16bits) */ 86 u_int16_t en_sw; /* status word (16bits) */ 87 u_int16_t en_tw; /* tag word (16bits) */ 88 u_int16_t en_opcode; /* opcode last executed (11 bits ) */ 89 u_int32_t en_fip; /* floating point instruction pointer */ 90 u_int16_t en_fcs; /* floating code segment selector */ 91 u_int16_t en_pad0; /* padding */ 92 u_int32_t en_foo; /* floating operand offset */ 93 u_int16_t en_fos; /* floating operand segment selector */ 94 u_int16_t en_pad1; /* padding */ 95 u_int32_t en_mxcsr; /* SSE control/status register */ 96 u_int32_t en_mxcsr_mask; /* valid bits in mxcsr */ 97 }; 98 99 /* Contents of each SSE extended accumulator */ 100 struct xmmacc { 101 u_char xmm_bytes[16]; 102 }; 103 104 /* Contents of the upper 16 bytes of each AVX extended accumulator */ 105 struct ymmacc { 106 uint8_t ymm_bytes[16]; 107 }; 108 109 struct savexmm { 110 struct envxmm sv_env; 111 struct { 112 struct fpacc87 fp_acc; 113 u_char fp_pad[6]; /* padding */ 114 } sv_fp[8]; 115 struct xmmacc sv_xmm[8]; 116 u_char sv_pad[224]; 117 } __aligned(16); 118 119 union savefpu { 120 struct save87 sv_87; 121 struct savexmm sv_xmm; 122 }; 123 124 struct xstate_hdr { 125 uint64_t xstate_bv; 126 uint8_t xstate_rsrv0[16]; 127 uint8_t xstate_rsrv[40]; 128 }; 129 130 struct savexmm_xstate { 131 struct xstate_hdr sx_hd; 132 struct ymmacc sx_ymm[16]; 133 }; 134 135 struct savexmm_ymm { 136 struct envxmm sv_env; 137 struct { 138 struct fpacc87 fp_acc; 139 int8_t fp_pad[6]; /* padding */ 140 } sv_fp[8]; 141 struct xmmacc sv_xmm[16]; 142 uint8_t sv_pad[96]; 143 struct savexmm_xstate sv_xstate; 144 } __aligned(64); 145 146 /* 147 * The hardware default control word for i387's and later coprocessors is 148 * 0x37F, giving: 149 * 150 * round to nearest 151 * 64-bit precision 152 * all exceptions masked. 153 * 154 * We modify the affine mode bit and precision bits in this to give: 155 * 156 * affine mode for 287's (if they work at all) (1 in bitfield 1<<12) 157 * 53-bit precision (2 in bitfield 3<<8) 158 * 159 * 64-bit precision often gives bad results with high level languages 160 * because it makes the results of calculations depend on whether 161 * intermediate values are stored in memory or in FPU registers. 162 */ 163 #define __INITIAL_NPXCW__ 0x127F 164 #define __INITIAL_MXCSR__ 0x1F80 165 166 #ifdef _KERNEL 167 168 #define PCB_USER_FPU(pcb) (((pcb)->pcb_flags & PCB_KERNNPX) == 0) 169 170 int npxdna(void); 171 void npxdrop(void); 172 void npxexit(struct thread *td); 173 int npxformat(void); 174 int npxgetregs(struct thread *td); 175 void npxinit(void); 176 void npxsave(union savefpu *addr); 177 void npxsetregs(struct thread *td, union savefpu *addr); 178 int npxtrap(void); 179 void npxuserinited(struct thread *); 180 struct fpu_kern_ctx *fpu_kern_alloc_ctx(u_int flags); 181 void fpu_kern_free_ctx(struct fpu_kern_ctx *ctx); 182 int fpu_kern_enter(struct thread *td, struct fpu_kern_ctx *ctx, 183 u_int flags); 184 int fpu_kern_leave(struct thread *td, struct fpu_kern_ctx *ctx); 185 int fpu_kern_thread(u_int flags); 186 int is_fpu_kern_thread(u_int flags); 187 188 /* 189 * Flags for fpu_kern_enter() and fpu_kern_thread(). 190 */ 191 #define FPU_KERN_NORMAL 0x0000 192 #define FPU_KERN_NOWAIT 0x0001 193 194 #endif 195 196 #endif /* !_MACHINE_NPX_H_ */ 197