about summary refs log tree commit diff
path: root/sysdeps/ia64/fpu/s_logbf.S
blob: f2f671f8920c1d85cccc2e867f8b82f59658f573 (plain) (blame)
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
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
.file "logbf.s"

// Copyright (C) 2000, 2001, Intel Corporation
// All rights reserved.
// 
// Contributed 2/2/2000 by John Harrison, Ted Kubaska, Bob Norin, 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
//==============================================================
// 2/02/00  Initial version
// 2/16/00  Modified to conform to C9X 
// 3/16/00  Improved speed
// 4/04/00  Unwind support added
// 5/30/00  Fixed bug when x double-extended denormal
// 8/15/00  Bundle added after call to __libm_error_support to properly
//          set [the previously overwritten] GR_Parameter_RESULT.
//
// API
//==============================================================
// float logbf( float x);
//
// Overview of operation
//==============================================================
// The logbf function extracts the exponent of x as an integer in 
// floating-point format. 
// logbf computes log2 of x as a float

// logbf is similar to ilogbf but differs in the  following ways:
//         +-inf
//            ilogbf: returns INT_MAX
//             logbf: returns +inf
//         Nan  returns FP_LOGBNAN (which is either INT_MAX or INT_MIN)
//            ilogbf: returns INT_MAX (7fffffff)
//             logbf: returns QNAN (quietized SNAN)
//         0    returns FP_ILOGB0 (which is either INT_MIN or -INT_MAX)
//            ilogbf: returns -INT_MAX (80000001)
//             logbf: returns -inf, raises the divide-by-zero exception,
//                   and calls libm_error_support to set domain error
//
// Registers used
//==============================================================
// general registers used: 
// ar.pfs r32 
// r33 -> r37
// r38 -> r41 used as parameters to error path
//
// predicate registers used: 
// p6, p7, p8
//
// floating-point registers used: 
// f9, f10, f11
// f8, input

#include "libm_support.h"

GR_SAVE_B0                    = r34
// r40 is address of table of coefficients
GR_SAVE_PFS                   = r32
GR_SAVE_GP                    = r35 

GR_Parameter_X                = r38
GR_Parameter_Y                = r39
GR_Parameter_RESULT           = r40
GR_Parameter_TAG              = r41

FR_X             = f8
FR_Y             = f0
FR_RESULT        = f10


.align 32
.global logbf#

.section .text
.proc  logbf#
.align 32


logbf: 

// qnan snan inf norm     unorm 0 -+
// 0    0    0   0        1     0 11
// 0                      b
{ .mfi
      alloc          r32=ar.pfs,1,5,4,0
(p0)  fclass.m.unc  p8,p0 = f8, 0x0b
      nop.i 999
}
// X NORMAL
// r37 = exp(f8) - - 0xffff
// sig(f8) = r37
// f8 = convert_to_fp (sig))
{ .mfi
(p0)  getf.exp      r35 = f8
(p0)  fnorm         f10=f8
      nop.i 999 ;;
}

// qnan snan inf norm     unorm 0 -+
// 1    1    1   0        0     0 11
// e                      3
{ .mmf
(p0)  mov      r33 = 0xffff
(p0)  mov      r34 = 0x1ffff
(p0)  fclass.m.unc  p6,p0 = f8, 0xe3 ;;
}

{ .mfb
(p0)  and           r36 = r35, r34
(p0)  fclass.m.unc  p7,p0 = f8, 0x07
(p8)  br.cond.spnt  L(LOGB_DENORM) ;;
}

{ .mib
(p0)  sub           r37 = r36, r33
      nop.i 999
(p6)  br.cond.spnt  L(LOGB_NAN_INF) ;;
}

{ .mib
(p0)  setf.sig      f9  = r37
      nop.i 999
(p7)  br.cond.spnt  L(LOGB_ZERO) ;;
}

{ .mfi
      nop.m 999
(p0)  fcvt.xf       f10 = f9
      nop.i 999 ;;
}

{ .mfb
      nop.m 999
(p0)  fnorm.s       f8 = f10
(p0)  br.ret.sptk    b0 ;;
}

L(LOGB_DENORM):
// Form signexp of 2^64 in case need to scale denormal
// Check to see if double-extended denormal
{ .mfi
(p0)  mov r38 = 0x1003f
(p0)  fclass.m.unc  p8,p0 = f10, 0x0b
      nop.i 999 ;;
}

// Form 2^64 in case need to scale denormal
{ .mfi
(p0)  setf.exp f11 = r38
      nop.f 999
      nop.i 999 ;;
}

// If double-extended denormal add 64 to exponent bias for scaling
// If double-extended denormal form x * 2^64 which is normal
{ .mfi
(p8)  add r33 = 64, r33
(p8)  fmpy f10 = f10, f11    
      nop.i 999 ;;
}

// Logic is the same as normal path but use normalized input
{ .mmi
(p0)  getf.exp      r35 = f10 ;;
      nop.m 999
      nop.i 999 ;;
}

{ .mmi
(p0)  and           r36 = r35, r34 ;;
(p0)  sub           r37 = r36, r33
      nop.i 999 ;;
}

{ .mmi
(p0)  setf.sig      f9  = r37
      nop.m 999
      nop.i 999 ;;
}

{ .mfi
      nop.m 999
(p0)  fcvt.xf       f10 = f9
      nop.i 999 ;;
}

{ .mfb
      nop.m 999
(p0)  fnorm.s       f8 = f10
(p0)  br.ret.sptk    b0 ;;
}

L(LOGB_NAN_INF): 

// X NAN or INFINITY, return f8 * f8
{ .mfb
      nop.m 999
(p0)  fma.s      f8= f8,f8,f0
(p0)  br.ret.sptk   b0 ;;
}

L(LOGB_ZERO): 

// X ZERO
// return -1.0/fabs(f8)=-inf, set divide-by-zero flag, call error support
{ .mfi
      nop.m 999
(p0)  fmerge.s      f9 = f0,f8
      nop.i 999 ;;
}

{ .mfi
      nop.m 999
(p0)  fmerge.ns     f10 = f0,f9
      nop.i 999 ;;
}

{ .mfi
      nop.m 999
(p0)  frcpa         f10,p6 = f1,f10
      nop.i 999 ;;
}

.endp logbf
ASM_SIZE_DIRECTIVE(logbf)


.proc __libm_error_region
__libm_error_region:
.prologue
{ .mii
        add   GR_Parameter_Y=-32,sp             // Parameter 2 value
(p0)    mov   GR_Parameter_TAG = 152                   
.save   ar.pfs,GR_SAVE_PFS
        mov  GR_SAVE_PFS=ar.pfs                 // Save ar.pfs
}
{ .mfi
.fframe 64
        add sp=-64,sp                           // Create new stack
        nop.f 0
        mov GR_SAVE_GP=gp                       // Save gp
};;
{ .mmi
        stfs [GR_Parameter_Y] = FR_Y,16         // Store Parameter 2 on stack
        add GR_Parameter_X = 16,sp              // Parameter 1 address
.save   b0, GR_SAVE_B0
        mov GR_SAVE_B0=b0                       // Save b0
};;
.body
{ .mib
        stfs [GR_Parameter_X] = FR_X            // Store Parameter 1 on stack
        add   GR_Parameter_RESULT = 0,GR_Parameter_Y
        nop.b 0                                 // Parameter 3 address
}
{ .mib
        stfs [GR_Parameter_Y] = FR_RESULT       // Store Parameter 3 on stack
        add   GR_Parameter_Y = -16,GR_Parameter_Y
        br.call.sptk b0=__libm_error_support#   // Call error handling function
};;
{ .mmi
        nop.m 0
        nop.m 0
        add   GR_Parameter_RESULT = 48,sp
};;
{ .mmi
        ldfs  f8 = [GR_Parameter_RESULT]       // Get return result off stack
.restore sp
        add   sp = 64,sp                       // Restore stack pointer
        mov   b0 = GR_SAVE_B0                  // Restore return address
};;
{ .mib
        mov   gp = GR_SAVE_GP                  // Restore gp 
        mov   ar.pfs = GR_SAVE_PFS             // Restore ar.pfs
        br.ret.sptk     b0                     // Return
};; 

.endp __libm_error_region
ASM_SIZE_DIRECTIVE(__libm_error_region)


.type   __libm_error_support#,@function
.global __libm_error_support#