1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright (c) 1988 by Sun Microsystems, Inc. 24 */ 25 26 #ident "%Z%%M% %I% %E% SMI" /* SunOS-4.1 1.8 88/12/06 */ 27 28 #include <sys/fpu/fpu_simulator.h> 29 #include <sys/fpu/globals.h> 30 31 void 32 _fp_mul(pfpsd, px, py, pz) 33 fp_simd_type *pfpsd; 34 unpacked *px, *py, *pz; 35 36 { 37 unpacked *pt; 38 unsigned acc[4]; /* Product accumulator. */ 39 unsigned j, y, *x, s, r, c; 40 41 if ((int) px->fpclass <= (int) py->fpclass) { 42 pt = px; 43 px = py; 44 py = pt; 45 } 46 /* Now class(x) >= class(y). */ 47 48 *pz = *px; 49 if (pz->fpclass < fp_quiet) 50 pz->sign = px->sign ^ py->sign; 51 52 switch (px->fpclass) { 53 case fp_quiet: 54 case fp_signaling: 55 case fp_zero: 56 return; 57 case fp_infinity: 58 if (py->fpclass == fp_zero) { 59 fpu_error_nan(pfpsd, pz); 60 pz->fpclass = fp_quiet; 61 } 62 return; 63 case fp_normal: 64 if (py->fpclass == fp_zero) { 65 pz->fpclass = fp_zero; 66 return; 67 } 68 } 69 70 /* Now x and y are both normal or subnormal. */ 71 72 x = px->significand; /* save typing */ 73 74 /* intialize acc to zero */ 75 s = r = acc[0] = acc[1] = acc[2] = acc[3] = 0; 76 77 y = py->significand[3]; /* py->significand[3] * x */ 78 if (y != 0) { 79 j = 1; 80 do { 81 s |= r; /* shift acc right one bit */ 82 r = acc[3]&1; 83 acc[3] = ((acc[2]&1)<<31)|(acc[3]>>1); 84 acc[2] = ((acc[1]&1)<<31)|(acc[2]>>1); 85 acc[1] = ((acc[0]&1)<<31)|(acc[1]>>1); 86 acc[0] = (acc[0]>>1); 87 if (j&y) { /* bit i of y != 0, add x to acc */ 88 c = 0; 89 c = fpu_add3wc(&acc[3], acc[3], x[3], c); 90 c = fpu_add3wc(&acc[2], acc[2], x[2], c); 91 c = fpu_add3wc(&acc[1], acc[1], x[1], c); 92 c = fpu_add3wc(&acc[0], acc[0], x[0], c); 93 } 94 j += j; 95 } while (j != 0); 96 } 97 98 y = py->significand[2]; /* py->significand[2] * x */ 99 if (y != 0) { 100 j = 1; 101 do { 102 s |= r; /* shift acc right one bit */ 103 r = acc[3]&1; 104 acc[3] = ((acc[2]&1)<<31)|(acc[3]>>1); 105 acc[2] = ((acc[1]&1)<<31)|(acc[2]>>1); 106 acc[1] = ((acc[0]&1)<<31)|(acc[1]>>1); 107 acc[0] = (acc[0]>>1); 108 if (j&y) { /* bit i of y != 0, add x to acc */ 109 c = 0; 110 c = fpu_add3wc(&acc[3], acc[3], x[3], c); 111 c = fpu_add3wc(&acc[2], acc[2], x[2], c); 112 c = fpu_add3wc(&acc[1], acc[1], x[1], c); 113 c = fpu_add3wc(&acc[0], acc[0], x[0], c); 114 } 115 j += j; 116 } while (j != 0); 117 } else { 118 s |= r|(acc[3]&0x7fffffff); 119 r = (acc[3]&0x80000000)>>31; 120 acc[3] = acc[2]; acc[2] = acc[1]; acc[1] = acc[0]; acc[0] = 0; 121 } 122 123 y = py->significand[1]; /* py->significand[1] * x */ 124 if (y != 0) { 125 j = 1; 126 do { 127 s |= r; /* shift acc right one bit */ 128 r = acc[3]&1; 129 acc[3] = ((acc[2]&1)<<31)|(acc[3]>>1); 130 acc[2] = ((acc[1]&1)<<31)|(acc[2]>>1); 131 acc[1] = ((acc[0]&1)<<31)|(acc[1]>>1); 132 acc[0] = (acc[0]>>1); 133 if (j&y) { /* bit i of y != 0, add x to acc */ 134 c = 0; 135 c = fpu_add3wc(&acc[3], acc[3], x[3], c); 136 c = fpu_add3wc(&acc[2], acc[2], x[2], c); 137 c = fpu_add3wc(&acc[1], acc[1], x[1], c); 138 c = fpu_add3wc(&acc[0], acc[0], x[0], c); 139 } 140 j += j; 141 } while (j != 0); 142 } else { 143 s |= r|(acc[3]&0x7fffffff); 144 r = (acc[3]&0x80000000)>>31; 145 acc[3] = acc[2]; acc[2] = acc[1]; acc[1] = acc[0]; acc[0] = 0; 146 } 147 148 /* py->significand[0] * x */ 149 y = py->significand[0]; /* y is of form 0x0001???? */ 150 j = 1; 151 do { 152 s |= r; /* shift acc right one bit */ 153 r = acc[3]&1; 154 acc[3] = ((acc[2]&1)<<31)|(acc[3]>>1); 155 acc[2] = ((acc[1]&1)<<31)|(acc[2]>>1); 156 acc[1] = ((acc[0]&1)<<31)|(acc[1]>>1); 157 acc[0] = (acc[0]>>1); 158 if (j&y) { /* bit i of y != 0, add x to acc */ 159 c = 0; 160 c = fpu_add3wc(&acc[3], acc[3], x[3], c); 161 c = fpu_add3wc(&acc[2], acc[2], x[2], c); 162 c = fpu_add3wc(&acc[1], acc[1], x[1], c); 163 c = fpu_add3wc(&acc[0], acc[0], x[0], c); 164 } 165 j += j; 166 } while (j <= y); 167 168 if (acc[0] >= 0x20000) { /* right shift one bit to normalize */ 169 pz->exponent = px->exponent + py->exponent + 1; 170 pz->sticky = s|r; 171 pz->rounded = acc[3]&1; 172 pz->significand[3] = ((acc[2]&1)<<31)|(acc[3]>>1); 173 pz->significand[2] = ((acc[1]&1)<<31)|(acc[2]>>1); 174 pz->significand[1] = ((acc[0]&1)<<31)|(acc[1]>>1); 175 pz->significand[0] = (acc[0]>>1); 176 } else { 177 pz->exponent = px->exponent + py->exponent; 178 pz->sticky = s; 179 pz->rounded = r; 180 pz->significand[3] = acc[3]; 181 pz->significand[2] = acc[2]; 182 pz->significand[1] = acc[1]; 183 pz->significand[0] = acc[0]; 184 } 185 } 186