summaryrefslogtreecommitdiff
path: root/sysdeps/i386/fpu/e_pow.S
blob: 997dd30e9dba4799350d6a14ca124eef9a35c264 (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
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
315
316
317
318
319
320
321
322
323
324
325
326
327
328
/* ix87 specific implementation of pow function.
   Copyright (C) 1996, 1997, 1998, 1999, 2001 Free Software Foundation, Inc.
   This file is part of the GNU C Library.
   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1996.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, write to the Free
   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
   02111-1307 USA.  */

#include <machine/asm.h>

#ifdef __ELF__
	.section .rodata
#else
	.text
#endif

	.align ALIGNARG(4)
	ASM_TYPE_DIRECTIVE(infinity,@object)
inf_zero:
infinity:
	.byte 0, 0, 0, 0, 0, 0, 0xf0, 0x7f
	ASM_SIZE_DIRECTIVE(infinity)
	ASM_TYPE_DIRECTIVE(zero,@object)
zero:	.double 0.0
	ASM_SIZE_DIRECTIVE(zero)
	ASM_TYPE_DIRECTIVE(minf_mzero,@object)
minf_mzero:
minfinity:
	.byte 0, 0, 0, 0, 0, 0, 0xf0, 0xff
mzero:
	.byte 0, 0, 0, 0, 0, 0, 0, 0x80
	ASM_SIZE_DIRECTIVE(minf_mzero)
	ASM_TYPE_DIRECTIVE(one,@object)
one:	.double 1.0
	ASM_SIZE_DIRECTIVE(one)
	ASM_TYPE_DIRECTIVE(limit,@object)
limit:	.double 0.29
	ASM_SIZE_DIRECTIVE(limit)

#ifdef PIC
#define MO(op) op##@GOTOFF(%ecx)
#define MOX(op,x,f) op##@GOTOFF(%ecx,x,f)
#else
#define MO(op) op
#define MOX(op,x,f) op(,x,f)
#endif

	.text
ENTRY(__ieee754_pow)
	fldl	12(%esp)	// y
	fxam

#ifdef	PIC
	call	1f
1:	popl	%ecx
	addl	$_GLOBAL_OFFSET_TABLE_+[.-1b], %ecx
#endif

	fnstsw
	movb	%ah, %dl
	andb	$0x45, %ah
	cmpb	$0x40, %ah	// is y == 0 ?
	je	11f

	cmpb	$0x05, %ah	// is y == ±inf ?
	je	12f

	cmpb	$0x01, %ah	// is y == NaN ?
	je	30f

	fldl	4(%esp)		// x : y

	subl	$8,%esp

	fxam
	fnstsw
	movb	%ah, %dh
	andb	$0x45, %ah
	cmpb	$0x40, %ah
	je	20f		// x is ±0

	cmpb	$0x05, %ah
	je	15f		// x is ±inf

	fxch			// y : x

	/* First see whether `y' is a natural number.  In this case we
	   can use a more precise algorithm.  */
	fld	%st		// y : y : x
	fistpll	(%esp)		// y : x
	fildll	(%esp)		// int(y) : y : x
	fucomp	%st(1)		// y : x
	fnstsw
	sahf
	jne	2f

	/* OK, we have an integer value for y.  */
	popl	%eax
	popl	%edx
	orl	$0, %edx
	fstp	%st(0)		// x
	jns	4f		// y >= 0, jump
	fdivrl	MO(one)		// 1/x		(now referred to as x)
	negl	%eax
	adcl	$0, %edx
	negl	%edx
4:	fldl	MO(one)		// 1 : x
	fxch

6:	shrdl	$1, %edx, %eax
	jnc	5f
	fxch
	fmul	%st(1)		// x : ST*x
	fxch
5:	fmul	%st(0), %st	// x*x : ST*x
	shrl	$1, %edx
	movl	%eax, %ecx
	orl	%edx, %ecx
	jnz	6b
	fstp	%st(0)		// ST*x
	ret

	/* y is ±NAN */
30:	fldl	4(%esp)		// x : y
	fldl	MO(one)		// 1.0 : x : y
	fucomp	%st(1)		// x : y
	fnstsw
	sahf
	je	31f
	fxch			// y : x
31:	fstp	%st(1)
	ret

	.align ALIGNARG(4)
2:	/* y is a real number.  */
	fxch			// x : y
	fldl	MO(one)		// 1.0 : x : y
	fld	%st(1)		// x : 1.0 : x : y
	fsub	%st(1)		// x-1 : 1.0 : x : y
	fabs			// |x-1| : 1.0 : x : y
	fcompl	MO(limit)	// 1.0 : x : y
	fnstsw
	fxch			// x : 1.0 : y
	sahf
	ja	7f
	fsub	%st(1)		// x-1 : 1.0 : y
	fyl2xp1			// log2(x) : y
	jmp	8f

7:	fyl2x			// log2(x) : y
8:	fmul	%st(1)		// y*log2(x) : y
	fst	%st(1)		// y*log2(x) : y*log2(x)
	frndint			// int(y*log2(x)) : y*log2(x)
	fsubr	%st, %st(1)	// int(y*log2(x)) : fract(y*log2(x))
	fxch			// fract(y*log2(x)) : int(y*log2(x))
	f2xm1			// 2^fract(y*log2(x))-1 : int(y*log2(x))
	faddl	MO(one)		// 2^fract(y*log2(x)) : int(y*log2(x))
	fscale			// 2^fract(y*log2(x))*2^int(y*log2(x)) : int(y*log2(x))
	addl	$8, %esp
	fstp	%st(1)		// 2^fract(y*log2(x))*2^int(y*log2(x))
	ret


	// pow(x,±0) = 1
	.align ALIGNARG(4)
11:	fstp	%st(0)		// pop y
	fldl	MO(one)
	ret

	// y == ±inf
	.align ALIGNARG(4)
12:	fstp	%st(0)		// pop y
	fldl	4(%esp)		// x
	fabs
	fcompl	MO(one)		// < 1, == 1, or > 1
	fnstsw
	andb	$0x45, %ah
	cmpb	$0x45, %ah
	je	13f		// jump if x is NaN

	cmpb	$0x40, %ah
	je	14f		// jump if |x| == 1

	shlb	$1, %ah
	xorb	%ah, %dl
	andl	$2, %edx
	fldl	MOX(inf_zero, %edx, 4)
	ret

	.align ALIGNARG(4)
14:	fldl	MO(one)
	ret

	.align ALIGNARG(4)
13:	fldl	4(%esp)		// load x == NaN
	ret

	.align ALIGNARG(4)
	// x is ±inf
15:	fstp	%st(0)		// y
	testb	$2, %dh
	jz	16f		// jump if x == +inf

	// We must find out whether y is an odd integer.
	fld	%st		// y : y
	fistpll	(%esp)		// y
	fildll	(%esp)		// int(y) : y
	fucompp			// <empty>
	fnstsw
	sahf
	jne	17f

	// OK, the value is an integer, but is the number of bits small
	// enough so that all are coming from the mantissa?
	popl	%eax
	popl	%edx
	andb	$1, %al
	jz	18f		// jump if not odd
	movl	%edx, %eax
	orl	%edx, %edx
	jns	155f
	negl	%eax
155:	cmpl	$0x00200000, %eax
	ja	18f		// does not fit in mantissa bits
	// It's an odd integer.
	shrl	$31, %edx
	fldl	MOX(minf_mzero, %edx, 8)
	ret

	.align ALIGNARG(4)
16:	fcompl	MO(zero)
	addl	$8, %esp
	fnstsw
	shrl	$5, %eax
	andl	$8, %eax
	fldl	MOX(inf_zero, %eax, 1)
	ret

	.align ALIGNARG(4)
17:	shll	$30, %edx	// sign bit for y in right position
	addl	$8, %esp
18:	shrl	$31, %edx
	fldl	MOX(inf_zero, %edx, 8)
	ret

	.align ALIGNARG(4)
	// x is ±0
20:	fstp	%st(0)		// y
	testb	$2, %dl
	jz	21f		// y > 0

	// x is ±0 and y is < 0.  We must find out whether y is an odd integer.
	testb	$2, %dh
	jz	25f

	fld	%st		// y : y
	fistpll	(%esp)		// y
	fildll	(%esp)		// int(y) : y
	fucompp			// <empty>
	fnstsw
	sahf
	jne	26f

	// OK, the value is an integer, but is the number of bits small
	// enough so that all are coming from the mantissa?
	popl	%eax
	popl	%edx
	andb	$1, %al
	jz	27f		// jump if not odd
	cmpl	$0xffe00000, %edx
	jbe	27f		// does not fit in mantissa bits
	// It's an odd integer.
	// Raise divide-by-zero exception and get minus infinity value.
	fldl	MO(one)
	fdivl	MO(zero)
	fchs
	ret

25:	fstp	%st(0)
26:	addl	$8, %esp
27:	// Raise divide-by-zero exception and get infinity value.
	fldl	MO(one)
	fdivl	MO(zero)
	ret

	.align ALIGNARG(4)
	// x is ±0 and y is > 0.  We must find out whether y is an odd integer.
21:	testb	$2, %dh
	jz	22f

	fld	%st		// y : y
	fistpll	(%esp)		// y
	fildll	(%esp)		// int(y) : y
	fucompp			// <empty>
	fnstsw
	sahf
	jne	23f

	// OK, the value is an integer, but is the number of bits small
	// enough so that all are coming from the mantissa?
	popl	%eax
	popl	%edx
	andb	$1, %al
	jz	24f		// jump if not odd
	cmpl	$0xffe00000, %edx
	jae	24f		// does not fit in mantissa bits
	// It's an odd integer.
	fldl	MO(mzero)
	ret

22:	fstp	%st(0)
23:	addl	$8, %esp	// Don't use 2 x pop
24:	fldl	MO(zero)
	ret

END(__ieee754_pow)