fpu_arith.h revision 176491
1/* $NetBSD: fpu_arith.h,v 1.4 2005/12/24 20:07:28 perry Exp $ */ 2/* $FreeBSD: head/sys/powerpc/fpu/fpu_arith.h 176491 2008-02-23 20:05:26Z marcel $ */ 3 4/* 5 * Copyright (c) 1992, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * This software was developed by the Computer Systems Engineering group 9 * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and 10 * contributed to Berkeley. 11 * 12 * All advertising materials mentioning features or use of this software 13 * must display the following acknowledgement: 14 * This product includes software developed by the University of 15 * California, Lawrence Berkeley Laboratory. 16 * 17 * Redistribution and use in source and binary forms, with or without 18 * modification, are permitted provided that the following conditions 19 * are met: 20 * 1. Redistributions of source code must retain the above copyright 21 * notice, this list of conditions and the following disclaimer. 22 * 2. Redistributions in binary form must reproduce the above copyright 23 * notice, this list of conditions and the following disclaimer in the 24 * documentation and/or other materials provided with the distribution. 25 * 3. Neither the name of the University nor the names of its contributors 26 * may be used to endorse or promote products derived from this software 27 * without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 39 * SUCH DAMAGE. 40 * 41 * @(#)fpu_arith.h 8.1 (Berkeley) 6/11/93 42 */ 43 44/* 45 * Extended-precision arithmetic. 46 * 47 * We hold the notion of a `carry register', which may or may not be a 48 * machine carry bit or register. On the SPARC, it is just the machine's 49 * carry bit. 50 * 51 * In the worst case, you can compute the carry from x+y as 52 * (unsigned)(x + y) < (unsigned)x 53 * and from x+y+c as 54 * ((unsigned)(x + y + c) <= (unsigned)x && (y|c) != 0) 55 * for example. 56 */ 57 58 59#ifndef FPE_USE_ASM 60 61/* set up for extended-precision arithemtic */ 62#define FPU_DECL_CARRY quad_t fpu_carry, fpu_tmp; 63 64/* 65 * We have three kinds of add: 66 * add with carry: r = x + y + c 67 * add (ignoring current carry) and set carry: c'r = x + y + 0 68 * add with carry and set carry: c'r = x + y + c 69 * The macros use `C' for `use carry' and `S' for `set carry'. 70 * Note that the state of the carry is undefined after ADDC and SUBC, 71 * so if all you have for these is `add with carry and set carry', 72 * that is OK. 73 * 74 * The same goes for subtract, except that we compute x - y - c. 75 * 76 * Finally, we have a way to get the carry into a `regular' variable, 77 * or set it from a value. SET_CARRY turns 0 into no-carry, nonzero 78 * into carry; GET_CARRY sets its argument to 0 or 1. 79 */ 80#define FPU_ADDC(r, x, y) \ 81 (r) = (x) + (y) + (!!fpu_carry) 82#define FPU_ADDS(r, x, y) \ 83 { \ 84 fpu_tmp = (quad_t)(x) + (quad_t)(y); \ 85 (r) = (u_int)fpu_tmp; \ 86 fpu_carry = ((fpu_tmp & 0xffffffff00000000LL) != 0); \ 87 } 88#define FPU_ADDCS(r, x, y) \ 89 { \ 90 fpu_tmp = (quad_t)(x) + (quad_t)(y) + (!!fpu_carry); \ 91 (r) = (u_int)fpu_tmp; \ 92 fpu_carry = ((fpu_tmp & 0xffffffff00000000LL) != 0); \ 93 } 94#define FPU_SUBC(r, x, y) \ 95 (r) = (x) - (y) - (!!fpu_carry) 96#define FPU_SUBS(r, x, y) \ 97 { \ 98 fpu_tmp = (quad_t)(x) - (quad_t)(y); \ 99 (r) = (u_int)fpu_tmp; \ 100 fpu_carry = ((fpu_tmp & 0xffffffff00000000LL) != 0); \ 101 } 102#define FPU_SUBCS(r, x, y) \ 103 { \ 104 fpu_tmp = (quad_t)(x) - (quad_t)(y) - (!!fpu_carry); \ 105 (r) = (u_int)fpu_tmp; \ 106 fpu_carry = ((fpu_tmp & 0xffffffff00000000LL) != 0); \ 107 } 108 109#define FPU_GET_CARRY(r) (r) = (!!fpu_carry) 110#define FPU_SET_CARRY(v) fpu_carry = ((v) != 0) 111 112#else 113/* set up for extended-precision arithemtic */ 114#define FPU_DECL_CARRY 115 116/* 117 * We have three kinds of add: 118 * add with carry: r = x + y + c 119 * add (ignoring current carry) and set carry: c'r = x + y + 0 120 * add with carry and set carry: c'r = x + y + c 121 * The macros use `C' for `use carry' and `S' for `set carry'. 122 * Note that the state of the carry is undefined after ADDC and SUBC, 123 * so if all you have for these is `add with carry and set carry', 124 * that is OK. 125 * 126 * The same goes for subtract, except that we compute x - y - c. 127 * 128 * Finally, we have a way to get the carry into a `regular' variable, 129 * or set it from a value. SET_CARRY turns 0 into no-carry, nonzero 130 * into carry; GET_CARRY sets its argument to 0 or 1. 131 */ 132#define FPU_ADDC(r, x, y) \ 133 __asm volatile("adde %0,%1,%2" : "=r"(r) : "r"(x), "r"(y)) 134#define FPU_ADDS(r, x, y) \ 135 __asm volatile("addc %0,%1,%2" : "=r"(r) : "r"(x), "r"(y)) 136#define FPU_ADDCS(r, x, y) \ 137 __asm volatile("adde %0,%1,%2" : "=r"(r) : "r"(x), "r"(y)) 138#define FPU_SUBC(r, x, y) \ 139 __asm volatile("subfe %0,%2,%1" : "=r"(r) : "r"(x), "r"(y)) 140#define FPU_SUBS(r, x, y) \ 141 __asm volatile("subfc %0,%2,%1" : "=r"(r) : "r"(x), "r"(y)) 142#define FPU_SUBCS(r, x, y) \ 143 __asm volatile("subfe %0,%2,%1" : "=r"(r) : "r"(x), "r"(y)) 144 145#define FPU_GET_CARRY(r) __asm volatile("li %0,0; addie %0,%0,0" : "=r"(r)) 146/* This one needs to destroy a temp register. */ 147#define FPU_SET_CARRY(v) do { int __tmp; \ 148 __asm volatile("addic %0,%0,-1" : "r"(__tmp) : "r"(v)); \ 149 } while (0) 150 151#define FPU_SHL1_BY_ADD /* shift left 1 faster by ADDC than (a<<1)|(b>>31) */ 152#endif 153