about summary refs log tree commit diff
path: root/sysdeps/i386/fpu/s_cexpf.S
blob: 680f8e5ee39deebac29e57878ca2c3df5d863a57 (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
/* ix87 specific implementation of complex exponential function for double.
   Copyright (C) 1997, 2005, 2012 Free Software Foundation, Inc.
   This file is part of the GNU C Library.
   Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.

   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 <sysdep.h>

	.section .rodata

	.align ALIGNARG(4)
	ASM_TYPE_DIRECTIVE(huge_nan_null_null,@object)
huge_nan_null_null:
	.byte 0, 0, 0x80, 0x7f
	.byte 0, 0, 0xc0, 0x7f
	.float	0.0
zero:	.float	0.0
infinity:
	.byte 0, 0, 0x80, 0x7f
	.byte 0, 0, 0xc0, 0x7f
	.float 0.0
	.byte 0, 0, 0, 0x80
	ASM_SIZE_DIRECTIVE(huge_nan_null_null)

	ASM_TYPE_DIRECTIVE(twopi,@object)
twopi:
	.byte 0x35, 0xc2, 0x68, 0x21, 0xa2, 0xda, 0xf, 0xc9, 0x1, 0x40
	.byte 0, 0, 0, 0, 0, 0
	ASM_SIZE_DIRECTIVE(twopi)

	ASM_TYPE_DIRECTIVE(l2e,@object)
l2e:
	.byte 0xbc, 0xf0, 0x17, 0x5c, 0x29, 0x3b, 0xaa, 0xb8, 0xff, 0x3f
	.byte 0, 0, 0, 0, 0, 0
	ASM_SIZE_DIRECTIVE(l2e)

	ASM_TYPE_DIRECTIVE(one,@object)
one:	.double 1.0
	ASM_SIZE_DIRECTIVE(one)


#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(__cexpf)
	flds	4(%esp)			/* x */
	fxam
	fnstsw
	flds	8(%esp)			/* y : x */
#ifdef  PIC
	LOAD_PIC_REG (cx)
#endif
	movb	%ah, %dh
	andb	$0x45, %ah
	cmpb	$0x05, %ah
	je	1f			/* Jump if real part is +-Inf */
	cmpb	$0x01, %ah
	je	2f			/* Jump if real part is NaN */

	fxam				/* y : x */
	fnstsw
	/* If the imaginary part is not finite we return NaN+i NaN, as
	   for the case when the real part is NaN.  A test for +-Inf and
	   NaN would be necessary.  But since we know the stack register
	   we applied `fxam' to is not empty we can simply use one test.
	   Check your FPU manual for more information.  */
	andb	$0x01, %ah
	cmpb	$0x01, %ah
	je	20f

	/* We have finite numbers in the real and imaginary part.  Do
	   the real work now.  */
	fxch			/* x : y */
	fldt	MO(l2e)		/* log2(e) : x : y */
	fmulp			/* x * log2(e) : y */
	fld	%st		/* x * log2(e) : x * log2(e) : y */
	frndint			/* int(x * log2(e)) : x * log2(e) : y */
	fsubr	%st, %st(1)	/* int(x * log2(e)) : frac(x * log2(e)) : y */
	fxch			/* frac(x * log2(e)) : int(x * log2(e)) : y */
	f2xm1			/* 2^frac(x * log2(e))-1 : int(x * log2(e)) : y */
	faddl	MO(one)		/* 2^frac(x * log2(e)) : int(x * log2(e)) : y */
	fscale			/* e^x : int(x * log2(e)) : y */
	fst	%st(1)		/* e^x : e^x : y */
	fxch	%st(2)		/* y : e^x : e^x */
	fsincos			/* cos(y) : sin(y) : e^x : e^x */
	fnstsw
	testl	$0x400, %eax
	jnz	7f
	fmulp	%st, %st(3)	/* sin(y) : e^x : e^x * cos(y) */
	fmulp	%st, %st(1)	/* e^x * sin(y) : e^x * cos(y) */
	subl	$8, %esp
	cfi_adjust_cfa_offset (8)
	fstps	4(%esp)
	fstps	(%esp)
	popl	%eax
	cfi_adjust_cfa_offset (-4)
	popl	%edx
	cfi_adjust_cfa_offset (-4)
	ret

	/* We have to reduce the argument to fsincos.  */
	.align ALIGNARG(4)
7:	fldt	MO(twopi)	/* 2*pi : y : e^x : e^x */
	fxch			/* y : 2*pi : e^x : e^x */
8:	fprem1			/* y%(2*pi) : 2*pi : e^x : e^x */
	fnstsw
	testl	$0x400, %eax
	jnz	8b
	fstp	%st(1)		/* y%(2*pi) : e^x : e^x */
	fsincos			/* cos(y) : sin(y) : e^x : e^x */
	fmulp	%st, %st(3)
	fmulp	%st, %st(1)
	subl	$8, %esp
	cfi_adjust_cfa_offset (8)
	fstps	4(%esp)
	fstps	(%esp)
	popl	%eax
	cfi_adjust_cfa_offset (-4)
	popl	%edx
	cfi_adjust_cfa_offset (-4)
	ret

	/* The real part is +-inf.  We must make further differences.  */
	.align ALIGNARG(4)
1:	fxam			/* y : x */
	fnstsw
	movb	%ah, %dl
	testb	$0x01, %ah	/* See above why 0x01 is usable here.  */
	jne	3f


	/* The real part is +-Inf and the imaginary part is finite.  */
	andl	$0x245, %edx
	cmpb	$0x40, %dl	/* Imaginary part == 0?  */
	je	4f		/* Yes ->  */

	fxch			/* x : y */
	shrl	$6, %edx
	fstp	%st(0)		/* y */ /* Drop the real part.  */
	andl	$8, %edx	/* This puts the sign bit of the real part
				   in bit 3.  So we can use it to index a
				   small array to select 0 or Inf.  */
	fsincos			/* cos(y) : sin(y) */
	fnstsw
	testl	$0x0400, %eax
	jnz	5f
	fxch
	ftst
	fnstsw
	fstp	%st(0)
	shll	$23, %eax
	andl	$0x80000000, %eax
	orl	MOX(huge_nan_null_null,%edx,1), %eax
	movl	MOX(huge_nan_null_null,%edx,1), %ecx
	movl	%eax, %edx
	ftst
	fnstsw
	fstp	%st(0)
	shll	$23, %eax
	andl	$0x80000000, %eax
	orl	%ecx, %eax
	ret
	/* We must reduce the argument to fsincos.  */
	.align ALIGNARG(4)
5:	fldt	MO(twopi)
	fxch
6:	fprem1
	fnstsw
	testl	$0x400, %eax
	jnz	6b
	fstp	%st(1)
	fsincos
	fxch
	ftst
	fnstsw
	fstp	%st(0)
	shll	$23, %eax
	andl	$0x80000000, %eax
	orl	MOX(huge_nan_null_null,%edx,1), %eax
	movl	MOX(huge_nan_null_null,%edx,1), %ecx
	movl	%eax, %edx
	ftst
	fnstsw
	fstp	%st(0)
	shll	$23, %eax
	andl	$0x80000000, %eax
	orl	%ecx, %eax
	ret

	/* The real part is +-Inf and the imaginary part is +-0.  So return
	   +-Inf+-0i.  */
	.align ALIGNARG(4)
4:	subl	$4, %esp
	cfi_adjust_cfa_offset (4)
	fstps	(%esp)
	shrl	$6, %edx
	fstp	%st(0)
	andl	$8, %edx
	movl	MOX(huge_nan_null_null,%edx,1), %eax
	popl	%edx
	cfi_adjust_cfa_offset (-4)
	ret

	/* The real part is +-Inf, the imaginary is also is not finite.  */
	.align ALIGNARG(4)
3:	fstp	%st(0)
	fstp	%st(0)		/* <empty> */
	andb	$0x45, %ah
	andb	$0x47, %dh
	xorb	%dh, %ah
	jnz	30f
	flds	MO(infinity)	/* Raise invalid exception.  */
	fmuls	MO(zero)
	fstp	%st(0)
30:	movl	%edx, %eax
	shrl	$6, %edx
	shll	$3, %eax
	andl	$8, %edx
	andl	$16, %eax
	orl	%eax, %edx

	movl	MOX(huge_nan_null_null,%edx,1), %eax
	movl	MOX(huge_nan_null_null+4,%edx,1), %edx
	ret

	/* The real part is NaN.  */
	.align ALIGNARG(4)
20:	flds	MO(infinity)		/* Raise invalid exception.  */
	fmuls	MO(zero)
	fstp	%st(0)
2:	fstp	%st(0)
	fstp	%st(0)
	movl	MO(huge_nan_null_null+4), %eax
	movl	%eax, %edx
	ret

END(__cexpf)
weak_alias (__cexpf, cexpf)