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 * Floating Point Data Structures and Constants 38 * W. Jolitz 1/90 39 */ 40 41 #ifndef _MACHINE_FPU_H_ 42 #define _MACHINE_FPU_H_ 43 44 /* Contents of each x87 floating point accumulator */ 45 struct fpacc87 { 46 uint8_t fp_bytes[10]; 47 }; 48 49 /* Contents of each SSE extended accumulator */ 50 struct xmmacc { 51 uint8_t xmm_bytes[16]; 52 }; 53 54 /* Contents of the upper 16 bytes of each AVX extended accumulator */ 55 struct ymmacc { 56 uint8_t ymm_bytes[16]; 57 }; 58 59 struct envxmm { 60 uint16_t en_cw; /* control word (16bits) */ 61 uint16_t en_sw; /* status word (16bits) */ 62 uint8_t en_tw; /* tag word (8bits) */ 63 uint8_t en_zero; 64 uint16_t en_opcode; /* opcode last executed (11 bits ) */ 65 uint64_t en_rip; /* floating point instruction pointer */ 66 uint64_t en_rdp; /* floating operand pointer */ 67 uint32_t en_mxcsr; /* SSE sontorol/status register */ 68 uint32_t en_mxcsr_mask; /* valid bits in mxcsr */ 69 }; 70 71 struct savefpu { 72 struct envxmm sv_env; 73 struct { 74 struct fpacc87 fp_acc; 75 uint8_t fp_pad[6]; /* padding */ 76 } sv_fp[8]; 77 struct xmmacc sv_xmm[16]; 78 uint8_t sv_pad[96]; 79 } __aligned(16); 80 81 struct xstate_hdr { 82 uint64_t xstate_bv; 83 uint8_t xstate_rsrv0[16]; 84 uint8_t xstate_rsrv[40]; 85 }; 86 87 struct savefpu_xstate { 88 struct xstate_hdr sx_hd; 89 struct ymmacc sx_ymm[16]; 90 }; 91 92 struct savefpu_ymm { 93 struct envxmm sv_env; 94 struct { 95 struct fpacc87 fp_acc; 96 int8_t fp_pad[6]; /* padding */ 97 } sv_fp[8]; 98 struct xmmacc sv_xmm[16]; 99 uint8_t sv_pad[96]; 100 struct savefpu_xstate sv_xstate; 101 } __aligned(64); 102 103 #ifdef _KERNEL 104 105 struct fpu_kern_ctx; 106 107 #define PCB_USER_FPU(pcb) (((pcb)->pcb_flags & PCB_KERNFPU) == 0) 108 109 #define XSAVE_AREA_ALIGN 64 110 111 #endif 112 113 /* 114 * The hardware default control word for i387's and later coprocessors is 115 * 0x37F, giving: 116 * 117 * round to nearest 118 * 64-bit precision 119 * all exceptions masked. 120 * 121 * FreeBSD/i386 uses 53 bit precision for things like fadd/fsub/fsqrt etc 122 * because of the difference between memory and fpu register stack arguments. 123 * If its using an intermediate fpu register, it has 80/64 bits to work 124 * with. If it uses memory, it has 64/53 bits to work with. However, 125 * gcc is aware of this and goes to a fair bit of trouble to make the 126 * best use of it. 127 * 128 * This is mostly academic for AMD64, because the ABI prefers the use 129 * SSE2 based math. For FreeBSD/amd64, we go with the default settings. 130 */ 131 #define __INITIAL_FPUCW__ 0x037F 132 #define __INITIAL_FPUCW_I386__ 0x127F 133 #define __INITIAL_MXCSR__ 0x1F80 134 #define __INITIAL_MXCSR_MASK__ 0xFFBF 135 136 #ifdef _KERNEL 137 void fpudna(void); 138 void fpudrop(void); 139 void fpuexit(struct thread *td); 140 int fpuformat(void); 141 int fpugetregs(struct thread *td); 142 void fpuinit(void); 143 void fpusave(void *addr); 144 int fpusetregs(struct thread *td, struct savefpu *addr, 145 char *xfpustate, size_t xfpustate_size); 146 int fpusetxstate(struct thread *td, char *xfpustate, 147 size_t xfpustate_size); 148 int fputrap(void); 149 void fpuuserinited(struct thread *td); 150 struct fpu_kern_ctx *fpu_kern_alloc_ctx(u_int flags); 151 void fpu_kern_free_ctx(struct fpu_kern_ctx *ctx); 152 int fpu_kern_enter(struct thread *td, struct fpu_kern_ctx *ctx, 153 u_int flags); 154 int fpu_kern_leave(struct thread *td, struct fpu_kern_ctx *ctx); 155 int fpu_kern_thread(u_int flags); 156 int is_fpu_kern_thread(u_int flags); 157 158 /* 159 * Flags for fpu_kern_alloc_ctx(), fpu_kern_enter() and fpu_kern_thread(). 160 */ 161 #define FPU_KERN_NORMAL 0x0000 162 #define FPU_KERN_NOWAIT 0x0001 163 164 #endif 165 166 #endif /* !_MACHINE_FPU_H_ */ 167