1/*-
2 * Copyright (c) 2004-2005 David Schultz <das@FreeBSD.ORG>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD$
27 */
28
29#ifndef	_FENV_H_
30#define	_FENV_H_
31
32#include <stdint.h>
33#include <sys/cdefs.h>
34#include <sys/types.h>
35
36typedef struct {
37	struct {
38		uint32_t	__control;
39		uint32_t	__status;
40		uint32_t	__tag;
41		char		__other[16];
42	} __x87;
43	uint32_t		__mxcsr;
44} fenv_t;
45
46typedef	uint16_t	fexcept_t;
47
48/* Exception flags */
49#define	FE_INVALID	0x01
50#define	FE_DENORMAL	0x02
51#define	FE_DIVBYZERO	0x04
52#define	FE_OVERFLOW	0x08
53#define	FE_UNDERFLOW	0x10
54#define	FE_INEXACT	0x20
55#define	FE_ALL_EXCEPT	(FE_DIVBYZERO | FE_DENORMAL | FE_INEXACT | \
56			 FE_INVALID | FE_OVERFLOW | FE_UNDERFLOW)
57
58/* Rounding modes */
59#define	FE_TONEAREST	0x0000
60#define	FE_DOWNWARD	0x0400
61#define	FE_UPWARD	0x0800
62#define	FE_TOWARDZERO	0x0c00
63#define	_ROUND_MASK	(FE_TONEAREST | FE_DOWNWARD | \
64			 FE_UPWARD | FE_TOWARDZERO)
65
66/*
67 * As compared to the x87 control word, the SSE unit's control word
68 * has the rounding control bits offset by 3 and the exception mask
69 * bits offset by 7.
70 */
71#define	_SSE_ROUND_SHIFT	3
72#define	_SSE_EMASK_SHIFT	7
73
74__BEGIN_DECLS
75
76/* Default floating-point environment */
77extern const fenv_t	__fe_dfl_env;
78#define	FE_DFL_ENV	(&__fe_dfl_env)
79
80#define	__fldcw(__cw)		__asm __volatile("fldcw %0" : : "m" (__cw))
81#define	__fldenv(__env)		__asm __volatile("fldenv %0" : : "m" (__env))
82#define	__fldenvx(__env)	__asm __volatile("fldenv %0" : : "m" (__env)  \
83				: "st", "st(1)", "st(2)", "st(3)", "st(4)",   \
84				"st(5)", "st(6)", "st(7)")
85#define	__fnclex()		__asm __volatile("fnclex")
86#define	__fnstenv(__env)	__asm __volatile("fnstenv %0" : "=m" (*(__env)))
87#define	__fnstcw(__cw)		__asm __volatile("fnstcw %0" : "=m" (*(__cw)))
88#define	__fnstsw(__sw)		__asm __volatile("fnstsw %0" : "=am" (*(__sw)))
89#define	__fwait()		__asm __volatile("fwait")
90#define	__ldmxcsr(__csr)	__asm __volatile("ldmxcsr %0" : : "m" (__csr))
91#define	__stmxcsr(__csr)	__asm __volatile("stmxcsr %0" : "=m" (*(__csr)))
92
93static __inline int
94feclearexcept(int __excepts)
95{
96	fenv_t __env;
97
98	if (__excepts == FE_ALL_EXCEPT) {
99		__fnclex();
100	} else {
101		__fnstenv(&__env.__x87);
102		__env.__x87.__status &= ~__excepts;
103		__fldenv(__env.__x87);
104	}
105	__stmxcsr(&__env.__mxcsr);
106	__env.__mxcsr &= ~__excepts;
107	__ldmxcsr(__env.__mxcsr);
108	return (0);
109}
110
111static __inline int
112fegetexceptflag(fexcept_t *__flagp, int __excepts)
113{
114	uint32_t __mxcsr;
115	uint16_t __status;
116
117	__stmxcsr(&__mxcsr);
118	__fnstsw(&__status);
119	*__flagp = (__mxcsr | __status) & __excepts;
120	return (0);
121}
122
123int fesetexceptflag(const fexcept_t *__flagp, int __excepts);
124int feraiseexcept(int __excepts);
125
126static __inline int
127fetestexcept(int __excepts)
128{
129	uint32_t __mxcsr;
130	uint16_t __status;
131
132	__stmxcsr(&__mxcsr);
133	__fnstsw(&__status);
134	return ((__status | __mxcsr) & __excepts);
135}
136
137static __inline int
138fegetround(void)
139{
140	uint16_t __control;
141
142	/*
143	 * We assume that the x87 and the SSE unit agree on the
144	 * rounding mode.  Reading the control word on the x87 turns
145	 * out to be about 5 times faster than reading it on the SSE
146	 * unit on an Opteron 244.
147	 */
148	__fnstcw(&__control);
149	return (__control & _ROUND_MASK);
150}
151
152static __inline int
153fesetround(int __round)
154{
155	uint32_t __mxcsr;
156	uint16_t __control;
157
158	if (__round & ~_ROUND_MASK)
159		return (-1);
160
161	__fnstcw(&__control);
162	__control &= ~_ROUND_MASK;
163	__control |= __round;
164	__fldcw(__control);
165
166	__stmxcsr(&__mxcsr);
167	__mxcsr &= ~(_ROUND_MASK << _SSE_ROUND_SHIFT);
168	__mxcsr |= __round << _SSE_ROUND_SHIFT;
169	__ldmxcsr(__mxcsr);
170
171	return (0);
172}
173
174int fegetenv(fenv_t *__envp);
175int feholdexcept(fenv_t *__envp);
176
177static __inline int
178fesetenv(const fenv_t *__envp)
179{
180
181	/*
182	 * XXX Using fldenvx() instead of fldenv() tells the compiler that this
183	 * instruction clobbers the i387 register stack.  This happens because
184	 * we restore the tag word from the saved environment.  Normally, this
185	 * would happen anyway and we wouldn't care, because the ABI allows
186	 * function calls to clobber the i387 regs.  However, fesetenv() is
187	 * inlined, so we need to be more careful.
188	 */
189	__fldenvx(__envp->__x87);
190	__ldmxcsr(__envp->__mxcsr);
191	return (0);
192}
193
194int feupdateenv(const fenv_t *__envp);
195
196#if __BSD_VISIBLE
197
198int feenableexcept(int __mask);
199int fedisableexcept(int __mask);
200
201static __inline int
202fegetexcept(void)
203{
204	uint16_t __control;
205
206	/*
207	 * We assume that the masks for the x87 and the SSE unit are
208	 * the same.
209	 */
210	__fnstcw(&__control);
211	return (~__control & FE_ALL_EXCEPT);
212}
213
214#endif /* __BSD_VISIBLE */
215
216__END_DECLS
217
218#endif	/* !_FENV_H_ */
219