about summary refs log tree commit diff
path: root/sysdeps/i386/fpu/e_atanhl.S
blob: df3f1b8f842addff524bf0115ccaed4d0fd46260 (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
/* ix87 specific implementation of arctanh function.
   Copyright (C) 1996-2022 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   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, see
   <https://www.gnu.org/licenses/>.  */

#include <machine/asm.h>
#include <libm-alias-finite.h>

	.section .rodata

	.align ALIGNARG(4)
	/* Please note that we use double values for 0.5 and 1.0.  These
	   numbers have exact representations and so we don't get accuracy
	   problems.  The advantage is that the code is simpler.  */
	.type half,@object
half:	.double 0.5
	ASM_SIZE_DIRECTIVE(half)
	.type one,@object
one:	.double 1.0
	ASM_SIZE_DIRECTIVE(one)
	/* It is not important that this constant is precise.  It is only
	   a value which is known to be on the safe side for using the
	   fyl2xp1 instruction.  */
	.type limit,@object
limit:	.double 0.29
	ASM_SIZE_DIRECTIVE(limit)
	.align ALIGNARG(4)
	.type ln2_2,@object
ln2_2:	.quad   0xb17217f7d1cf79ac  /* 0.3465735902799726547086160  */
	.short  0x3ffd
	ASM_SIZE_DIRECTIVE(ln2_2)

#ifdef PIC
#define MO(op) op##@GOTOFF(%edx)
#else
#define MO(op) op
#endif

	.text
ENTRY(__ieee754_atanhl)
	movl	12(%esp), %ecx

	movl	%ecx, %eax
	andl	$0x7fff, %eax
	cmpl	$0x7fff, %eax
	je	5f
	cmpl	$0x3fdf, %eax
	jge	7f
	// Exponent below -32; return x, with underflow if subnormal.
	fldt	4(%esp)
	cmpl	$0, %eax
	jne	8f
	fld	%st(0)
	fmul	%st(0)
	fstp	%st(0)
8:	ret
7:

#ifdef PIC
	LOAD_PIC_REG (dx)
#endif

	andl	$0x8000, %ecx	// ECX == 0 iff X >= 0

	fldt	MO(ln2_2)	// 0.5*ln2
	xorl	%ecx, 12(%esp)
	fldt	4(%esp)		// |x| : 0.5*ln2
	fcoml	MO(half)	// |x| : 0.5*ln2
	fld	%st(0)		// |x| : |x| : 0.5*ln2
	fnstsw			// |x| : |x| : 0.5*ln2
	sahf
	jae	2f
	fadd	%st, %st(1)	// |x| : 2*|x| : 0.5*ln2
	fld	%st		// |x| : |x| : 2*|x| : 0.5*ln2
	fsubrl	MO(one)		// 1-|x| : |x| : 2*|x| : 0.5*ln2
	fxch			// |x| : 1-|x| : 2*|x| : 0.5*ln2
	fmul	%st(2)		// 2*|x|^2 : 1-|x| : 2*|x| : 0.5*ln2
	fdivp			// (2*|x|^2)/(1-|x|) : 2*|x| : 0.5*ln2
	faddp			// 2*|x|+(2*|x|^2)/(1-|x|) : 0.5*ln2
	fcoml	MO(limit)	// 2*|x|+(2*|x|^2)/(1-|x|) : 0.5*ln2
	fnstsw			// 2*|x|+(2*|x|^2)/(1-|x|) : 0.5*ln2
	sahf
	jae	4f
	fyl2xp1			// 0.5*ln2*ld(1+2*|x|+(2*|x|^2)/(1-|x|))
	jecxz	3f
	fchs			// 0.5*ln2*ld(1+2*x+(2*x^2)/(1-x))
3:	ret

	.align ALIGNARG(4)
4:	faddl	MO(one)		// 1+2*|x|+(2*|x|^2)/(1-|x|) : 0.5*ln2
	fyl2x			// 0.5*ln2*ld(1+2*|x|+(2*|x|^2)/(1-|x|))
	jecxz	3f
	fchs			// 0.5*ln2*ld(1+2*x+(2*x^2)/(1-x))
3:	ret

	.align ALIGNARG(4)
2:	faddl	MO(one)		// 1+|x| : |x| : 0.5*ln2
	fxch			// |x| : 1+|x| : 0.5*ln2
	fsubrl	MO(one)		// 1-|x| : 1+|x| : 0.5*ln2
	fdivrp			// (1+|x|)/(1-|x|) : 0.5*ln2
	fyl2x			// 0.5*ln2*ld((1+|x|)/(1-|x|))
	jecxz	3f
	fchs			// 0.5*ln2*ld((1+x)/(1-x))
3:	ret

	// x == NaN or ±Inf
5:	cmpl	$0x80000000, 8(%esp)
	ja	6f
	cmpl	$0, 4(%esp)
	je	7b
6:	fldt	4(%esp)
	fadd	%st(0)
	ret
END(__ieee754_atanhl)
libm_alias_finite (__ieee754_atanhl, __atanhl)