1/* 2 * Copyright (c) 1992 The Regents of the University of California. 3 * All rights reserved. 4 * 5 * This software was developed by the Computer Systems Engineering group 6 * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and 7 * contributed to Berkeley. 8 * 9 * %sccs.include.redist.c% 10 * 11 * from: $Header: mul.s,v 1.5 92/06/25 13:24:03 torek Exp $ 12 */ 13 14#if defined(LIBC_SCCS) && !defined(lint) 15 .asciz "@(#)mul.s 5.1 (Berkeley) 06/25/92" 16#endif /* LIBC_SCCS and not lint */ 17 18/* 19 * Signed multiply, from Appendix E of the Sparc Version 8 20 * Architecture Manual. 21 * 22 * Returns %o0 * %o1 in %o1%o0 (i.e., %o1 holds the upper 32 bits of 23 * the 64-bit product). 24 * 25 * This code optimizes short (less than 13-bit) multiplies. 26 */ 27 28#include "DEFS.h" 29FUNC(.mul) 30 mov %o0, %y ! multiplier -> Y 31 andncc %o0, 0xfff, %g0 ! test bits 12..31 32 be Lmul_shortway ! if zero, can do it the short way 33 andcc %g0, %g0, %o4 ! zero the partial product and clear N and V 34 35 /* 36 * Long multiply. 32 steps, followed by a final shift step. 37 */ 38 mulscc %o4, %o1, %o4 ! 1 39 mulscc %o4, %o1, %o4 ! 2 40 mulscc %o4, %o1, %o4 ! 3 41 mulscc %o4, %o1, %o4 ! 4 42 mulscc %o4, %o1, %o4 ! 5 43 mulscc %o4, %o1, %o4 ! 6 44 mulscc %o4, %o1, %o4 ! 7 45 mulscc %o4, %o1, %o4 ! 8 46 mulscc %o4, %o1, %o4 ! 9 47 mulscc %o4, %o1, %o4 ! 10 48 mulscc %o4, %o1, %o4 ! 11 49 mulscc %o4, %o1, %o4 ! 12 50 mulscc %o4, %o1, %o4 ! 13 51 mulscc %o4, %o1, %o4 ! 14 52 mulscc %o4, %o1, %o4 ! 15 53 mulscc %o4, %o1, %o4 ! 16 54 mulscc %o4, %o1, %o4 ! 17 55 mulscc %o4, %o1, %o4 ! 18 56 mulscc %o4, %o1, %o4 ! 19 57 mulscc %o4, %o1, %o4 ! 20 58 mulscc %o4, %o1, %o4 ! 21 59 mulscc %o4, %o1, %o4 ! 22 60 mulscc %o4, %o1, %o4 ! 23 61 mulscc %o4, %o1, %o4 ! 24 62 mulscc %o4, %o1, %o4 ! 25 63 mulscc %o4, %o1, %o4 ! 26 64 mulscc %o4, %o1, %o4 ! 27 65 mulscc %o4, %o1, %o4 ! 28 66 mulscc %o4, %o1, %o4 ! 29 67 mulscc %o4, %o1, %o4 ! 30 68 mulscc %o4, %o1, %o4 ! 31 69 mulscc %o4, %o1, %o4 ! 32 70 mulscc %o4, %g0, %o4 ! final shift 71 72 ! If %o0 was negative, the result is 73 ! (%o0 * %o1) + (%o1 << 32)) 74 ! We fix that here. 75 76 tst %o0 77 bge 1f 78 rd %y, %o0 79 80 ! %o0 was indeed negative; fix upper 32 bits of result by subtracting 81 ! %o1 (i.e., return %o4 - %o1 in %o1). 82 retl 83 sub %o4, %o1, %o1 84 851: 86 retl 87 mov %o4, %o1 88 89Lmul_shortway: 90 /* 91 * Short multiply. 12 steps, followed by a final shift step. 92 * The resulting bits are off by 12 and (32-12) = 20 bit positions, 93 * but there is no problem with %o0 being negative (unlike above). 94 */ 95 mulscc %o4, %o1, %o4 ! 1 96 mulscc %o4, %o1, %o4 ! 2 97 mulscc %o4, %o1, %o4 ! 3 98 mulscc %o4, %o1, %o4 ! 4 99 mulscc %o4, %o1, %o4 ! 5 100 mulscc %o4, %o1, %o4 ! 6 101 mulscc %o4, %o1, %o4 ! 7 102 mulscc %o4, %o1, %o4 ! 8 103 mulscc %o4, %o1, %o4 ! 9 104 mulscc %o4, %o1, %o4 ! 10 105 mulscc %o4, %o1, %o4 ! 11 106 mulscc %o4, %o1, %o4 ! 12 107 mulscc %o4, %g0, %o4 ! final shift 108 109 /* 110 * %o4 has 20 of the bits that should be in the low part of the 111 * result; %y has the bottom 12 (as %y's top 12). That is: 112 * 113 * %o4 %y 114 * +----------------+----------------+ 115 * | -12- | -20- | -12- | -20- | 116 * +------(---------+------)---------+ 117 * --hi-- ----low-part---- 118 * 119 * The upper 12 bits of %o4 should be sign-extended to form the 120 * high part of the product (i.e., highpart = %o4 >> 20). 121 */ 122 123 rd %y, %o5 124 sll %o4, 12, %o0 ! shift middle bits left 12 125 srl %o5, 20, %o5 ! shift low bits right 20, zero fill at left 126 or %o5, %o0, %o0 ! construct low part of result 127 retl 128 sra %o4, 20, %o1 ! ... and extract high part of result 129