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
|
.file "nearbyint.s"
// Copyright (C) 2000, 2001, Intel Corporation
// All rights reserved.
//
// Contributed 10/19/2000 by John Harrison, Cristina Iordache, Ted Kubaska,
// Bob Norin, Tom Rowan, Shane Story, and Ping Tak Peter Tang of the
// Computational Software Lab, 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://developer.intel.com/opensource.
//
// History
//==============================================================
// 10/19/2000: Created
// 2/08/01 Corrected behavior for all rounding modes.
//==============================================================
//
// API
//==============================================================
// double nearbyint(double x)
#include "libm_support.h"
//
// general registers used:
//
nearbyint_GR_signexp = r14
nearbyint_GR_exponent = r15
nearbyint_GR_17ones = r16
nearbyint_GR_10033 = r17
nearbyint_GR_fpsr = r18
nearbyint_GR_rcs0 = r19
nearbyint_GR_rcs0_mask = r20
// predicate registers used:
// p6-11
// floating-point registers used:
NEARBYINT_NORM_f8 = f9
NEARBYINT_FLOAT_INT_f8 = f10
NEARBYINT_INT_f8 = f11
// Overview of operation
//==============================================================
// double nearbyint(double x)
// Return an integer value (represented as a double) that is x rounded to integer in current
// rounding mode
// *******************************************************************************
// Set denormal flag for denormal input and
// and take denormal fault if necessary.
// Is the input an integer value already?
// double_extended
// if the exponent is >= 1003e => 3F(true) = 63(decimal)
// we have a significand of 64 bits 1.63-bits.
// If we multiply by 2^63, we no longer have a fractional part
// So input is an integer value already.
// double
// if the exponent is >= 10033 => 34(true) = 52(decimal)
// 34 + 3ff = 433
// we have a significand of 53 bits 1.52-bits. (implicit 1)
// If we multiply by 2^52, we no longer have a fractional part
// So input is an integer value already.
// single
// if the exponent is >= 10016 => 17(true) = 23(decimal)
// we have a significand of 53 bits 1.52-bits. (implicit 1)
// If we multiply by 2^52, we no longer have a fractional part
// So input is an integer value already.
// If x is NAN, ZERO, or INFINITY, then return
// qnan snan inf norm unorm 0 -+
// 1 1 1 0 0 1 11 0xe7
.align 32
.global nearbyint#
.section .text
.proc nearbyint#
.align 32
nearbyint:
{ .mfi
mov nearbyint_GR_fpsr = ar40 // Read the fpsr--need to check rc.s0
fcvt.fx.s1 NEARBYINT_INT_f8 = f8
addl nearbyint_GR_10033 = 0x10033, r0
}
{ .mfi
nop.m 999
fnorm.s1 NEARBYINT_NORM_f8 = f8
mov nearbyint_GR_17ones = 0x1FFFF
;;
}
{ .mfi
nop.m 999
fclass.m.unc p6,p0 = f8, 0xe7
mov nearbyint_GR_rcs0_mask = 0x0c00
;;
}
{ .mfb
nop.m 999
(p6) fnorm.d f8 = f8
(p6) br.ret.spnt b0 // Exit if x nan, inf, zero
;;
}
{ .mfi
nop.m 999
fcvt.xf NEARBYINT_FLOAT_INT_f8 = NEARBYINT_INT_f8
nop.i 999
;;
}
{ .mfi
getf.exp nearbyint_GR_signexp = NEARBYINT_NORM_f8
fcmp.eq.s0 p8,p0 = f8,f0 // Dummy op to set denormal
nop.i 999
;;
}
{ .mii
nop.m 999
nop.i 999
and nearbyint_GR_exponent = nearbyint_GR_signexp, nearbyint_GR_17ones
;;
}
{ .mmi
cmp.ge.unc p7,p6 = nearbyint_GR_exponent, nearbyint_GR_10033
and nearbyint_GR_rcs0 = nearbyint_GR_rcs0_mask, nearbyint_GR_fpsr
nop.i 999
;;
}
// Check to see if s0 rounding mode is round to nearest. If not then set s2
// rounding mode to that of s0 and repeat conversions.
L(NEARBYINT_COMMON):
{ .mfb
cmp.ne p11,p0 = nearbyint_GR_rcs0, r0
(p6) fclass.m.unc p9,p10 = NEARBYINT_FLOAT_INT_f8, 0x07 // Test for result=0
(p11) br.cond.spnt L(NEARBYINT_NOT_ROUND_NEAREST) // Branch if not round to nearest
;;
}
{ .mfi
nop.m 999
(p7) fnorm.d.s0 f8 = f8
nop.i 999
;;
}
// If result is zero, merge sign of input
{ .mfi
nop.m 999
(p9) fmerge.s f8 = f8, NEARBYINT_FLOAT_INT_f8
nop.i 999
}
{ .mfb
nop.m 999
(p10) fnorm.d f8 = NEARBYINT_FLOAT_INT_f8
br.ret.sptk b0
;;
}
L(NEARBYINT_NOT_ROUND_NEAREST):
// Set rounding mode of s2 to that of s0
{ .mfi
mov nearbyint_GR_rcs0 = r0 // Clear so we don't come back here
fsetc.s2 0x7f, 0x40
nop.i 999
;;
}
{ .mfi
nop.m 999
fcvt.fx.s2 NEARBYINT_INT_f8 = f8
nop.i 999
;;
}
{ .mfb
nop.m 999
fcvt.xf NEARBYINT_FLOAT_INT_f8 = NEARBYINT_INT_f8
br.cond.sptk L(NEARBYINT_COMMON)
;;
}
.endp nearbyint
ASM_SIZE_DIRECTIVE(nearbyint)
|
