aarch64/fenv: fix fenv_t (#319)

* aarch64: copy `aarch64/fenv.h` from FreeBSD

https://github.com/freebsd/freebsd-src/blob/de1aa3dab23c06fec962a14da3e7b4755c5880cf/lib/msun/aarch64/fenv.h

* aarch64: clean code

* arm: remove AARCH64 ifdef

include/openlibm_fenv.h

@@ -2,7 +2,9 @@
 #include <fenv.h>
 #else /* !OPENLIBM_USE_HOST_FENV_H */
 
-#if defined(__aarch64__) || defined(__arm__)
+#if defined(__aarch64__)
+#include <openlibm_fenv_aarch64.h>
+#elif defined(__arm__)
 #include <openlibm_fenv_arm.h>
 #elif defined(__x86_64__)
 #include <openlibm_fenv_amd64.h>

include/openlibm_fenv_aarch64.h

@@ -0,0 +1,247 @@
+/*-
+ * Copyright (c) 2004-2005 David Schultz <[email protected]>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. 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.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 THE AUTHOR OR 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.
+ */
+
+#ifndef	_FENV_H_
+#define	_FENV_H_
+
+#include <stdint.h>
+
+#include "cdefs-compat.h"
+
+#ifndef	__fenv_static
+#define	__fenv_static	static
+#endif
+
+/* The high 32 bits contain fpcr, low 32 contain fpsr. */
+typedef	__uint64_t	fenv_t;
+typedef	__uint64_t	fexcept_t;
+
+/* Exception flags */
+#define	FE_INVALID	0x00000001
+#define	FE_DIVBYZERO	0x00000002
+#define	FE_OVERFLOW	0x00000004
+#define	FE_UNDERFLOW	0x00000008
+#define	FE_INEXACT	0x00000010
+#define	FE_ALL_EXCEPT	(FE_DIVBYZERO | FE_INEXACT | \
+			 FE_INVALID | FE_OVERFLOW | FE_UNDERFLOW)
+
+/*
+ * Rounding modes
+ *
+ * We can't just use the hardware bit values here, because that would
+ * make FE_UPWARD and FE_DOWNWARD negative, which is not allowed.
+ */
+#define	FE_TONEAREST	0x0
+#define	FE_UPWARD	0x1
+#define	FE_DOWNWARD	0x2
+#define	FE_TOWARDZERO	0x3
+#define	_ROUND_MASK	(FE_TONEAREST | FE_DOWNWARD | \
+			 FE_UPWARD | FE_TOWARDZERO)
+#define	_ROUND_SHIFT	22
+
+__BEGIN_DECLS
+
+/* Default floating-point environment */
+extern const fenv_t	__fe_dfl_env;
+#define	FE_DFL_ENV	(&__fe_dfl_env)
+
+/* We need to be able to map status flag positions to mask flag positions */
+#define _FPUSW_SHIFT	8
+#define	_ENABLE_MASK	(FE_ALL_EXCEPT << _FPUSW_SHIFT)
+
+#define	__mrs_fpcr(__r)	__asm __volatile("mrs %0, fpcr" : "=r" (__r))
+#define	__msr_fpcr(__r)	__asm __volatile("msr fpcr, %0" : : "r" (__r))
+
+#define	__mrs_fpsr(__r)	__asm __volatile("mrs %0, fpsr" : "=r" (__r))
+#define	__msr_fpsr(__r)	__asm __volatile("msr fpsr, %0" : : "r" (__r))
+
+
+__fenv_static inline int
+feclearexcept(int __excepts)
+{
+	fexcept_t __r;
+
+	__mrs_fpsr(__r);
+	__r &= ~__excepts;
+	__msr_fpsr(__r);
+	return (0);
+}
+
+__fenv_static inline int
+fegetexceptflag(fexcept_t *__flagp, int __excepts)
+{
+	fexcept_t __r;
+
+	__mrs_fpsr(__r);
+	*__flagp = __r & __excepts;
+	return (0);
+}
+
+__fenv_static inline int
+fesetexceptflag(const fexcept_t *__flagp, int __excepts)
+{
+	fexcept_t __r;
+
+	__mrs_fpsr(__r);
+	__r &= ~__excepts;
+	__r |= *__flagp & __excepts;
+	__msr_fpsr(__r);
+	return (0);
+}
+
+__fenv_static inline int
+feraiseexcept(int __excepts)
+{
+	fexcept_t __r;
+
+	__mrs_fpsr(__r);
+	__r |= __excepts;
+	__msr_fpsr(__r);
+	return (0);
+}
+
+__fenv_static inline int
+fetestexcept(int __excepts)
+{
+	fexcept_t __r;
+
+	__mrs_fpsr(__r);
+	return (__r & __excepts);
+}
+
+__fenv_static inline int
+fegetround(void)
+{
+	fenv_t __r;
+
+	__mrs_fpcr(__r);
+	return ((__r >> _ROUND_SHIFT) & _ROUND_MASK);
+}
+
+__fenv_static inline int
+fesetround(int __round)
+{
+	fenv_t __r;
+
+	if (__round & ~_ROUND_MASK)
+		return (-1);
+	__mrs_fpcr(__r);
+	__r &= ~(_ROUND_MASK << _ROUND_SHIFT);
+	__r |= __round << _ROUND_SHIFT;
+	__msr_fpcr(__r);
+	return (0);
+}
+
+__fenv_static inline int
+fegetenv(fenv_t *__envp)
+{
+	__uint64_t fpcr;
+	__uint64_t fpsr;
+
+	__mrs_fpcr(fpcr);
+	__mrs_fpsr(fpsr);
+	*__envp = fpsr | (fpcr << 32);
+
+	return (0);
+}
+
+__fenv_static inline int
+feholdexcept(fenv_t *__envp)
+{
+	fenv_t __r;
+
+	__mrs_fpcr(__r);
+	*__envp = __r << 32;
+	__r &= ~(_ENABLE_MASK);
+	__msr_fpcr(__r);
+
+	__mrs_fpsr(__r);
+	*__envp |= (__uint32_t)__r;
+	__r &= ~(_ENABLE_MASK);
+	__msr_fpsr(__r);
+	return (0);
+}
+
+__fenv_static inline int
+fesetenv(const fenv_t *__envp)
+{
+
+	__msr_fpcr((*__envp) >> 32);
+	__msr_fpsr((fenv_t)(__uint32_t)*__envp);
+	return (0);
+}
+
+__fenv_static inline int
+feupdateenv(const fenv_t *__envp)
+{
+	fexcept_t __r;
+
+	__mrs_fpsr(__r);
+	fesetenv(__envp);
+	feraiseexcept(__r & FE_ALL_EXCEPT);
+	return (0);
+}
+
+#if __BSD_VISIBLE
+
+/* We currently provide no external definitions of the functions below. */
+
+static inline int
+feenableexcept(int __mask)
+{
+	fenv_t __old_r, __new_r;
+
+	__mrs_fpcr(__old_r);
+	__new_r = __old_r | ((__mask & FE_ALL_EXCEPT) << _FPUSW_SHIFT);
+	__msr_fpcr(__new_r);
+	return ((__old_r >> _FPUSW_SHIFT) & FE_ALL_EXCEPT);
+}
+
+static inline int
+fedisableexcept(int __mask)
+{
+	fenv_t __old_r, __new_r;
+
+	__mrs_fpcr(__old_r);
+	__new_r = __old_r & ~((__mask & FE_ALL_EXCEPT) << _FPUSW_SHIFT);
+	__msr_fpcr(__new_r);
+	return ((__old_r >> _FPUSW_SHIFT) & FE_ALL_EXCEPT);
+}
+
+static inline int
+fegetexcept(void)
+{
+	fenv_t __r;
+
+	__mrs_fpcr(__r);
+	return ((__r & _ENABLE_MASK) >> _FPUSW_SHIFT);
+}
+
+#endif /* __BSD_VISIBLE */
+
+__END_DECLS
+
+#endif	/* !_FENV_H_ */

include/openlibm_fenv_arm.h

@@ -66,12 +66,9 @@ extern const fenv_t	__fe_dfl_env;
 #define _FPUSW_SHIFT	16
 #define	_ENABLE_MASK	(FE_ALL_EXCEPT << _FPUSW_SHIFT)
 
-#if defined(__aarch64__)
-#define __rfs(__fpsr)   __asm __volatile("mrs %0,fpsr" : "=r" (*(__fpsr)))
-#define __wfs(__fpsr)   __asm __volatile("msr fpsr,%0" : : "r" (__fpsr))
 /* Test for hardware support for ARM floating point operations, explicitly
 checking for float and double support, see "ARM C Language Extensions", 6.5.1 */
-#elif defined(__ARM_FP) && (__ARM_FP & 0x0C) != 0
+#if defined(__ARM_FP) && (__ARM_FP & 0x0C) != 0
 #define __rfs(__fpsr)   __asm __volatile("vmrs %0,fpscr" : "=&r" (*(__fpsr)))
 #define __wfs(__fpsr)   __asm __volatile("vmsr fpscr,%0" : : "r" (__fpsr))
 #else