summaryrefslogtreecommitdiffstats
path: root/arch/parisc/math-emu/dfsqrt.c
blob: e3a3a19b966b013d5bf52b9d0d47c8c3abe1d4e6 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * Linux/PA-RISC Project (http://www.parisc-linux.org/)
 *
 * Floating-point emulation code
 *  Copyright (C) 2001 Hewlett-Packard (Paul Bame) <bame@debian.org>
 */
/*
 * BEGIN_DESC
 *
 *  File:
 *	@(#)	pa/spmath/dfsqrt.c		$Revision: 1.1 $
 *
 *  Purpose:
 *	Double Floating-point Square Root
 *
 *  External Interfaces:
 *	dbl_fsqrt(srcptr,_nullptr,dstptr,status)
 *
 *  Internal Interfaces:
 *
 *  Theory:
 *	<<please update with a overview of the operation of this file>>
 *
 * END_DESC
*/


#include "float.h"
#include "dbl_float.h"

/*
 *  Double Floating-point Square Root
 */

/*ARGSUSED*/
unsigned int
dbl_fsqrt(
	    dbl_floating_point *srcptr,
	    unsigned int *_nullptr,
	    dbl_floating_point *dstptr,
	    unsigned int *status)
{
	register unsigned int srcp1, srcp2, resultp1, resultp2;
	register unsigned int newbitp1, newbitp2, sump1, sump2;
	register int src_exponent;
	register boolean guardbit = FALSE, even_exponent;

	Dbl_copyfromptr(srcptr,srcp1,srcp2);
        /*
         * check source operand for NaN or infinity
         */
        if ((src_exponent = Dbl_exponent(srcp1)) == DBL_INFINITY_EXPONENT) {
                /*
                 * is signaling NaN?
                 */
                if (Dbl_isone_signaling(srcp1)) {
                        /* trap if INVALIDTRAP enabled */
                        if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION);
                        /* make NaN quiet */
                        Set_invalidflag();
                        Dbl_set_quiet(srcp1);
                }
                /*
                 * Return quiet NaN or positive infinity.
		 *  Fall through to negative test if negative infinity.
                 */
		if (Dbl_iszero_sign(srcp1) || 
		    Dbl_isnotzero_mantissa(srcp1,srcp2)) {
                	Dbl_copytoptr(srcp1,srcp2,dstptr);
                	return(NOEXCEPTION);
		}
        }

        /*
         * check for zero source operand
         */
	if (Dbl_iszero_exponentmantissa(srcp1,srcp2)) {
		Dbl_copytoptr(srcp1,srcp2,dstptr);
		return(NOEXCEPTION);
	}

        /*
         * check for negative source operand 
         */
	if (Dbl_isone_sign(srcp1)) {
		/* trap if INVALIDTRAP enabled */
		if (Is_invalidtrap_enabled()) return(INVALIDEXCEPTION);
		/* make NaN quiet */
		Set_invalidflag();
		Dbl_makequietnan(srcp1,srcp2);
		Dbl_copytoptr(srcp1,srcp2,dstptr);
		return(NOEXCEPTION);
	}

	/*
	 * Generate result
	 */
	if (src_exponent > 0) {
		even_exponent = Dbl_hidden(srcp1);
		Dbl_clear_signexponent_set_hidden(srcp1);
	}
	else {
		/* normalize operand */
		Dbl_clear_signexponent(srcp1);
		src_exponent++;
		Dbl_normalize(srcp1,srcp2,src_exponent);
		even_exponent = src_exponent & 1;
	}
	if (even_exponent) {
		/* exponent is even */
		/* Add comment here.  Explain why odd exponent needs correction */
		Dbl_leftshiftby1(srcp1,srcp2);
	}
	/*
	 * Add comment here.  Explain following algorithm.
	 * 
	 * Trust me, it works.
	 *
	 */
	Dbl_setzero(resultp1,resultp2);
	Dbl_allp1(newbitp1) = 1 << (DBL_P - 32);
	Dbl_setzero_mantissap2(newbitp2);
	while (Dbl_isnotzero(newbitp1,newbitp2) && Dbl_isnotzero(srcp1,srcp2)) {
		Dbl_addition(resultp1,resultp2,newbitp1,newbitp2,sump1,sump2);
		if(Dbl_isnotgreaterthan(sump1,sump2,srcp1,srcp2)) {
			Dbl_leftshiftby1(newbitp1,newbitp2);
			/* update result */
			Dbl_addition(resultp1,resultp2,newbitp1,newbitp2,
			 resultp1,resultp2);  
			Dbl_subtract(srcp1,srcp2,sump1,sump2,srcp1,srcp2);
			Dbl_rightshiftby2(newbitp1,newbitp2);
		}
		else {
			Dbl_rightshiftby1(newbitp1,newbitp2);
		}
		Dbl_leftshiftby1(srcp1,srcp2);
	}
	/* correct exponent for pre-shift */
	if (even_exponent) {
		Dbl_rightshiftby1(resultp1,resultp2);
	}

	/* check for inexact */
	if (Dbl_isnotzero(srcp1,srcp2)) {
		if (!even_exponent && Dbl_islessthan(resultp1,resultp2,srcp1,srcp2)) {
			Dbl_increment(resultp1,resultp2);
		}
		guardbit = Dbl_lowmantissap2(resultp2);
		Dbl_rightshiftby1(resultp1,resultp2);

		/*  now round result  */
		switch (Rounding_mode()) {
		case ROUNDPLUS:
		     Dbl_increment(resultp1,resultp2);
		     break;
		case ROUNDNEAREST:
		     /* stickybit is always true, so guardbit 
		      * is enough to determine rounding */
		     if (guardbit) {
			    Dbl_increment(resultp1,resultp2);
		     }
		     break;
		}
		/* increment result exponent by 1 if mantissa overflowed */
		if (Dbl_isone_hiddenoverflow(resultp1)) src_exponent+=2;

		if (Is_inexacttrap_enabled()) {
			Dbl_set_exponent(resultp1,
			 ((src_exponent-DBL_BIAS)>>1)+DBL_BIAS);
			Dbl_copytoptr(resultp1,resultp2,dstptr);
			return(INEXACTEXCEPTION);
		}
		else Set_inexactflag();
	}
	else {
		Dbl_rightshiftby1(resultp1,resultp2);
	}
	Dbl_set_exponent(resultp1,((src_exponent-DBL_BIAS)>>1)+DBL_BIAS);
	Dbl_copytoptr(resultp1,resultp2,dstptr);
	return(NOEXCEPTION);
}