| 1 |
// SPDX-License-Identifier: GPL-2.0-only |
| 2 |
/* IEEE754 floating point arithmetic |
| 3 |
* double precision: common utilities |
| 4 |
*/ |
| 5 |
/* |
| 6 |
* MIPS floating point support |
| 7 |
* Copyright (C) 1994-2000 Algorithmics Ltd. |
| 8 |
*/ |
| 9 |
|
| 10 |
#include "ieee754dp.h" |
| 11 |
|
| 12 |
union ieee754dp ieee754dp_mul(union ieee754dp x, union ieee754dp y) |
| 13 |
{ |
| 14 |
int re; |
| 15 |
int rs; |
| 16 |
u64 rm; |
| 17 |
unsigned int lxm; |
| 18 |
unsigned int hxm; |
| 19 |
unsigned int lym; |
| 20 |
unsigned int hym; |
| 21 |
u64 lrm; |
| 22 |
u64 hrm; |
| 23 |
u64 t; |
| 24 |
u64 at; |
| 25 |
|
| 26 |
COMPXDP; |
| 27 |
COMPYDP; |
| 28 |
|
| 29 |
EXPLODEXDP; |
| 30 |
EXPLODEYDP; |
| 31 |
|
| 32 |
ieee754_clearcx(); |
| 33 |
|
| 34 |
FLUSHXDP; |
| 35 |
FLUSHYDP; |
| 36 |
|
| 37 |
switch (CLPAIR(xc, yc)) { |
| 38 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_SNAN): |
| 39 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_SNAN): |
| 40 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_SNAN): |
| 41 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_SNAN): |
| 42 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_SNAN): |
| 43 |
return ieee754dp_nanxcpt(y); |
| 44 |
|
| 45 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_SNAN): |
| 46 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_QNAN): |
| 47 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_ZERO): |
| 48 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_NORM): |
| 49 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_DNORM): |
| 50 |
case CLPAIR(IEEE754_CLASS_SNAN, IEEE754_CLASS_INF): |
| 51 |
return ieee754dp_nanxcpt(x); |
| 52 |
|
| 53 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_QNAN): |
| 54 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_QNAN): |
| 55 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_QNAN): |
| 56 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_QNAN): |
| 57 |
return y; |
| 58 |
|
| 59 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_QNAN): |
| 60 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_ZERO): |
| 61 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_NORM): |
| 62 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_DNORM): |
| 63 |
case CLPAIR(IEEE754_CLASS_QNAN, IEEE754_CLASS_INF): |
| 64 |
return x; |
| 65 |
|
| 66 |
|
| 67 |
/* |
| 68 |
* Infinity handling |
| 69 |
*/ |
| 70 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_ZERO): |
| 71 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_INF): |
| 72 |
ieee754_setcx(IEEE754_INVALID_OPERATION); |
| 73 |
return ieee754dp_indef(); |
| 74 |
|
| 75 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_INF): |
| 76 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_INF): |
| 77 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_NORM): |
| 78 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_DNORM): |
| 79 |
case CLPAIR(IEEE754_CLASS_INF, IEEE754_CLASS_INF): |
| 80 |
return ieee754dp_inf(xs ^ ys); |
| 81 |
|
| 82 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_ZERO): |
| 83 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_NORM): |
| 84 |
case CLPAIR(IEEE754_CLASS_ZERO, IEEE754_CLASS_DNORM): |
| 85 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_ZERO): |
| 86 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_ZERO): |
| 87 |
return ieee754dp_zero(xs ^ ys); |
| 88 |
|
| 89 |
|
| 90 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_DNORM): |
| 91 |
DPDNORMX; |
| 92 |
fallthrough; |
| 93 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_DNORM): |
| 94 |
DPDNORMY; |
| 95 |
break; |
| 96 |
|
| 97 |
case CLPAIR(IEEE754_CLASS_DNORM, IEEE754_CLASS_NORM): |
| 98 |
DPDNORMX; |
| 99 |
break; |
| 100 |
|
| 101 |
case CLPAIR(IEEE754_CLASS_NORM, IEEE754_CLASS_NORM): |
| 102 |
break; |
| 103 |
} |
| 104 |
/* rm = xm * ym, re = xe+ye basically */ |
| 105 |
assert(xm & DP_HIDDEN_BIT); |
| 106 |
assert(ym & DP_HIDDEN_BIT); |
| 107 |
|
| 108 |
re = xe + ye; |
| 109 |
rs = xs ^ ys; |
| 110 |
|
| 111 |
/* shunt to top of word */ |
| 112 |
xm <<= 64 - (DP_FBITS + 1); |
| 113 |
ym <<= 64 - (DP_FBITS + 1); |
| 114 |
|
| 115 |
/* |
| 116 |
* Multiply 64 bits xm, ym to give high 64 bits rm with stickness. |
| 117 |
*/ |
| 118 |
|
| 119 |
lxm = xm; |
| 120 |
hxm = xm >> 32; |
| 121 |
lym = ym; |
| 122 |
hym = ym >> 32; |
| 123 |
|
| 124 |
lrm = DPXMULT(lxm, lym); |
| 125 |
hrm = DPXMULT(hxm, hym); |
| 126 |
|
| 127 |
t = DPXMULT(lxm, hym); |
| 128 |
|
| 129 |
at = lrm + (t << 32); |
| 130 |
hrm += at < lrm; |
| 131 |
lrm = at; |
| 132 |
|
| 133 |
hrm = hrm + (t >> 32); |
| 134 |
|
| 135 |
t = DPXMULT(hxm, lym); |
| 136 |
|
| 137 |
at = lrm + (t << 32); |
| 138 |
hrm += at < lrm; |
| 139 |
lrm = at; |
| 140 |
|
| 141 |
hrm = hrm + (t >> 32); |
| 142 |
|
| 143 |
rm = hrm | (lrm != 0); |
| 144 |
|
| 145 |
/* |
| 146 |
* Sticky shift down to normal rounding precision. |
| 147 |
*/ |
| 148 |
if ((s64) rm < 0) { |
| 149 |
rm = (rm >> (64 - (DP_FBITS + 1 + 3))) | |
| 150 |
((rm << (DP_FBITS + 1 + 3)) != 0); |
| 151 |
re++; |
| 152 |
} else { |
| 153 |
rm = (rm >> (64 - (DP_FBITS + 1 + 3 + 1))) | |
| 154 |
((rm << (DP_FBITS + 1 + 3 + 1)) != 0); |
| 155 |
} |
| 156 |
assert(rm & (DP_HIDDEN_BIT << 3)); |
| 157 |
|
| 158 |
return ieee754dp_format(rs, re, rm); |
| 159 |
} |