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
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
|
.file "modf.s"
// Copyright (c) 2000 - 2003, Intel Corporation
// All rights reserved.
//
// Contributed 2000 by the Intel Numerics Group, Intel Corporation
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//
// * Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
//
// * The name of Intel Corporation may not be used to endorse or promote
// products derived from this software without specific prior written
// permission.
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Intel Corporation is the author of this code, and requests that all
// problem reports or change requests be submitted to it directly at
// http://www.intel.com/software/products/opensource/libraries/num.htm.
//
// History
//==============================================================
// 02/02/00 Initial version
// 04/04/00 Improved speed, corrected result for NaN input
// 12/22/00 Fixed so inexact flag is never set, and invalid is not set for
// qnans nor for inputs larger than 2^63.
// 05/20/02 Cleaned up namespace and sf0 syntax
// 02/10/03 Reordered header: .section, .global, .proc, .align
//
// API
//==============================================================
// double modf(double x, double *iptr)
// break a floating point x number into fraction and an exponent
//
// input floating point f8, address in r33
// output floating point f8 (x fraction), and *iptr (x integral part)
//
// OVERVIEW
//==============================================================
//
// NO FRACTIONAL PART: HUGE
// If
// for double-extended
// If the true exponent is greater than or equal 63
// 1003e ==> 1003e -ffff = 3f = 63(dec)
// for double
// If the true exponent is greater than or equal 52
// 10033 -ffff = 34 = 52(dec)
// for single
// If the true exponent is greater than or equal 23
// 10016 -ffff = 17 = 23(dec)
// then
// we are already an integer (p9 true)
// NO INTEGER PART: SMALL
// Is f8 exponent less than register bias (that is, is it
// less than 1). If it is, get the right sign of
// zero and store this in iptr.
// CALCULATION: NOT HUGE, NOT SMALL
// To get the integer part
// Take the floating-point input and truncate
// then convert this integer to fp Call it MODF_INTEGER_PART
// Subtract MODF_INTEGER_PART from MODF_NORM_F8 to get fraction part
// Then put fraction part in f8
// put integer part MODF_INTEGER_PART into *iptr
// Registers used
//==============================================================
// predicate registers used:
// p6 - p13
// 0xFFFF 0x10033
// -----------------------+-----------------+-------------
// SMALL | NORMAL | HUGE
// p11 --------------->|<----- p12 ----->| <-------------- p9
// p10 --------------------------------->|
// p13 --------------------------------------------------->|
//
// floating-point registers used:
MODF_NORM_F8 = f9
MODF_FRACTION_PART = f10
MODF_INTEGER_PART = f11
MODF_INT_INTEGER_PART = f12
// general registers used
modf_signexp = r14
modf_GR_no_frac = r15
modf_GR_FFFF = r16
modf_17_ones = r17
modf_exp = r18
// r33 = iptr
.section .text
GLOBAL_LIBM_ENTRY(modf)
// Main path is p9, p11, p8 FALSE and p12 TRUE
// Assume input is normalized and get signexp
// Normalize input just in case
// Form exponent bias
{ .mfi
getf.exp modf_signexp = f8
fnorm.s0 MODF_NORM_F8 = f8
addl modf_GR_FFFF = 0xffff, r0
}
// Get integer part of input
// Form exponent mask
{ .mfi
nop.m 999
fcvt.fx.trunc.s1 MODF_INT_INTEGER_PART = f8
mov modf_17_ones = 0x1ffff ;;
}
// Is x nan or inf?
// qnan snan inf norm unorm 0 -+
// 1 1 1 0 0 0 11 = 0xe3 NAN_INF
// Form biased exponent where input only has an integer part
{ .mfi
nop.m 999
fclass.m.unc p6,p13 = f8, 0xe3
addl modf_GR_no_frac = 0x10033, r0 ;;
}
// Mask to get exponent
// Is x unnorm?
// qnan snan inf norm unorm 0 -+
// 0 0 0 0 1 0 11 = 0x0b UNORM
// Set p13 to indicate calculation path, else p6 if nan or inf
{ .mfi
and modf_exp = modf_17_ones, modf_signexp
fclass.m.unc p8,p0 = f8, 0x0b
nop.i 999 ;;
}
// p11 <== SMALL, no integer part, fraction is everyting
// p9 <== HUGE, no fraction part, integer is everything
// p12 <== NORMAL, fraction part and integer part
{ .mii
(p13) cmp.lt.unc p11,p10 = modf_exp, modf_GR_FFFF
nop.i 999
nop.i 999 ;;
}
// Is x inf? p6 if inf, p7 if nan
{ .mfb
(p10) cmp.ge.unc p9,p12 = modf_exp, modf_GR_no_frac
(p6) fclass.m.unc p6,p7 = f8, 0x23
(p8) br.cond.spnt MODF_DENORM ;;
}
MODF_COMMON:
// For HUGE set fraction to signed 0
{ .mfi
nop.m 999
(p9) fmerge.s f8 = f8,f0
nop.i 999
}
// For HUGE set integer part to normalized input
{ .mfi
nop.m 999
(p9) fnorm.d.s0 MODF_INTEGER_PART = MODF_NORM_F8
nop.i 999 ;;
}
// For SMALL set fraction to normalized input, integer part to signed 0
{ .mfi
nop.m 999
(p11) fmerge.s MODF_INTEGER_PART = f8,f0
nop.i 999
}
{ .mfi
nop.m 999
(p11) fnorm.d.s0 f8 = MODF_NORM_F8
nop.i 999 ;;
}
// For NORMAL float the integer part
{ .mfi
nop.m 999
(p12) fcvt.xf MODF_INTEGER_PART = MODF_INT_INTEGER_PART
nop.i 999 ;;
}
// If x inf set integer part to INF, fraction to signed 0
{ .mfi
(p6) stfd [r33] = MODF_NORM_F8
(p6) fmerge.s f8 = f8,f0
nop.i 999 ;;
}
// If x nan set integer and fraction parts to NaN (quietized)
{ .mfi
(p7) stfd [r33] = MODF_NORM_F8
(p7) fmerge.s f8 = MODF_NORM_F8, MODF_NORM_F8
nop.i 999 ;;
}
{ .mmi
(p9) stfd [r33] = MODF_INTEGER_PART
nop.m 999
nop.i 999 ;;
}
// For NORMAL compute fraction part
{ .mfi
(p11) stfd [r33] = MODF_INTEGER_PART
(p12) fms.d.s0 f8 = MODF_NORM_F8,f1, MODF_INTEGER_PART
nop.i 999 ;;
}
// For NORMAL test if fraction part is zero; if so append correct sign
{ .mfi
nop.m 999
(p12) fcmp.eq.unc.s0 p7,p0 = MODF_NORM_F8, MODF_INTEGER_PART
nop.i 999 ;;
}
{ .mfi
(p12) stfd [r33] = MODF_INTEGER_PART
nop.f 999
nop.i 999 ;;
}
// For NORMAL if fraction part is zero append sign of input
{ .mfb
nop.m 999
(p7) fmerge.s f8 = MODF_NORM_F8, f0
br.ret.sptk b0 ;;
}
MODF_DENORM:
// If x unorm get signexp from normalized input
// If x unorm get integer part from normalized input
{ .mfi
getf.exp modf_signexp = MODF_NORM_F8
fcvt.fx.trunc.s1 MODF_INT_INTEGER_PART = MODF_NORM_F8
nop.i 999 ;;
}
// If x unorm mask to get exponent
{ .mmi
and modf_exp = modf_17_ones, modf_signexp ;;
cmp.lt.unc p11,p10 = modf_exp, modf_GR_FFFF
nop.i 999 ;;
}
{ .mfb
(p10) cmp.ge.unc p9,p12 = modf_exp, modf_GR_no_frac
nop.f 999
br.cond.spnt MODF_COMMON ;;
}
GLOBAL_LIBM_END(modf)
|