Updated x86_64 fenv.h and added fenv.c.

This commit is contained in:
Alex Smith
2012-07-27 15:21:41 +01:00
parent 5f119ed78f
commit 3fde53501d
5 changed files with 386 additions and 7 deletions
+8 -5
View File
@@ -111,7 +111,8 @@ feclearexcept(int __excepts)
static __inline int
fegetexceptflag(fexcept_t *__flagp, int __excepts)
{
int __mxcsr, __status;
uint32_t __mxcsr;
uint16_t __status;
__stmxcsr(&__mxcsr);
__fnstsw(&__status);
@@ -125,7 +126,8 @@ int feraiseexcept(int __excepts);
static __inline int
fetestexcept(int __excepts)
{
int __mxcsr, __status;
uint32_t __mxcsr;
uint16_t __status;
__stmxcsr(&__mxcsr);
__fnstsw(&__status);
@@ -135,7 +137,7 @@ fetestexcept(int __excepts)
static __inline int
fegetround(void)
{
int __control;
uint16_t __control;
/*
* We assume that the x87 and the SSE unit agree on the
@@ -150,7 +152,8 @@ fegetround(void)
static __inline int
fesetround(int __round)
{
int __mxcsr, __control;
uint32_t __mxcsr;
uint16_t __control;
if (__round & ~_ROUND_MASK)
return (-1);
@@ -198,7 +201,7 @@ int fedisableexcept(int __mask);
static __inline int
fegetexcept(void)
{
int __control;
uint16_t __control;
/*
* We assume that the masks for the x87 and the SSE unit are
+218
View File
@@ -0,0 +1,218 @@
/*-
* Copyright (c) 1990 The Regents of the University of California.
* All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* William Jolitz.
*
* 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.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
*
* from: @(#)npx.h 5.3 (Berkeley) 1/18/91
* $FreeBSD$
*/
/*
* Floating Point Data Structures and Constants
* W. Jolitz 1/90
*/
#ifndef _X86_FPU_H_
#define _X86_FPU_H_
#define __aligned(x) __attribute__((__aligned__(x)))
/* Environment information of floating point unit. */
struct env87 {
int32_t en_cw; /* control word (16bits) */
int32_t en_sw; /* status word (16bits) */
int32_t en_tw; /* tag word (16bits) */
int32_t en_fip; /* fp instruction pointer */
uint16_t en_fcs; /* fp code segment selector */
uint16_t en_opcode; /* opcode last executed (11 bits) */
int32_t en_foo; /* fp operand offset */
int32_t en_fos; /* fp operand segment selector */
};
/* Contents of each x87 floating point accumulator. */
struct fpacc87 {
uint8_t fp_bytes[10];
};
/* Floating point context. (i386 fnsave/frstor) */
struct save87 {
struct env87 sv_env; /* floating point control/status */
struct fpacc87 sv_ac[8]; /* accumulator contents, 0-7 */
uint8_t sv_pad0[4]; /* saved status word (now unused) */
/*
* Bogus padding for emulators. Emulators should use their own
* struct and arrange to store into this struct (ending here)
* before it is inspected for ptracing or for core dumps. Some
* emulators overwrite the whole struct. We have no good way of
* knowing how much padding to leave. Leave just enough for the
* GPL emulator's i387_union (176 bytes total).
*/
uint8_t sv_pad[64]; /* padding; used by emulators */
};
/* Contents of each SSE extended accumulator. */
struct xmmacc {
uint8_t xmm_bytes[16];
};
/* Contents of the upper 16 bytes of each AVX extended accumulator. */
struct ymmacc {
uint8_t ymm_bytes[16];
};
/* Rename structs below depending on machine architecture. */
#ifdef __i386__
#define __envxmm32 envxmm
#else
#define __envxmm32 envxmm32
#define __envxmm64 envxmm
#endif
struct __envxmm32 {
uint16_t en_cw; /* control word (16bits) */
uint16_t en_sw; /* status word (16bits) */
uint16_t en_tw; /* tag word (16bits) */
uint16_t en_opcode; /* opcode last executed (11 bits) */
uint32_t en_fip; /* fp instruction pointer */
uint16_t en_fcs; /* fp code segment selector */
uint16_t en_pad0; /* padding */
uint32_t en_foo; /* fp operand offset */
uint16_t en_fos; /* fp operand segment selector */
uint16_t en_pad1; /* padding */
uint32_t en_mxcsr; /* SSE control/status register */
uint32_t en_mxcsr_mask; /* valid bits in mxcsr */
};
struct __envxmm64 {
uint16_t en_cw; /* control word (16bits) */
uint16_t en_sw; /* status word (16bits) */
uint8_t en_tw; /* tag word (8bits) */
uint8_t en_zero;
uint16_t en_opcode; /* opcode last executed (11 bits ) */
uint64_t en_rip; /* fp instruction pointer */
uint64_t en_rdp; /* fp operand pointer */
uint32_t en_mxcsr; /* SSE control/status register */
uint32_t en_mxcsr_mask; /* valid bits in mxcsr */
};
/* Floating point context. (i386 fxsave/fxrstor) */
struct savexmm {
struct __envxmm32 sv_env;
struct {
struct fpacc87 fp_acc;
uint8_t fp_pad[6]; /* padding */
} sv_fp[8];
struct xmmacc sv_xmm[8];
uint8_t sv_pad[224];
} __aligned(16);
#ifdef __i386__
union savefpu {
struct save87 sv_87;
struct savexmm sv_xmm;
};
#else
/* Floating point context. (amd64 fxsave/fxrstor) */
struct savefpu {
struct __envxmm64 sv_env;
struct {
struct fpacc87 fp_acc;
uint8_t fp_pad[6]; /* padding */
} sv_fp[8];
struct xmmacc sv_xmm[16];
uint8_t sv_pad[96];
} __aligned(16);
#endif
struct xstate_hdr {
uint64_t xstate_bv;
uint8_t xstate_rsrv0[16];
uint8_t xstate_rsrv[40];
};
struct savexmm_xstate {
struct xstate_hdr sx_hd;
struct ymmacc sx_ymm[16];
};
struct savexmm_ymm {
struct __envxmm32 sv_env;
struct {
struct fpacc87 fp_acc;
int8_t fp_pad[6]; /* padding */
} sv_fp[8];
struct xmmacc sv_xmm[16];
uint8_t sv_pad[96];
struct savexmm_xstate sv_xstate;
} __aligned(64);
struct savefpu_xstate {
struct xstate_hdr sx_hd;
struct ymmacc sx_ymm[16];
};
struct savefpu_ymm {
struct __envxmm64 sv_env;
struct {
struct fpacc87 fp_acc;
int8_t fp_pad[6]; /* padding */
} sv_fp[8];
struct xmmacc sv_xmm[16];
uint8_t sv_pad[96];
struct savefpu_xstate sv_xstate;
} __aligned(64);
#undef __envxmm32
#undef __envxmm64
/*
* The hardware default control word for i387's and later coprocessors is
* 0x37F, giving:
*
* round to nearest
* 64-bit precision
* all exceptions masked.
*
* FreeBSD/i386 uses 53 bit precision for things like fadd/fsub/fsqrt etc
* because of the difference between memory and fpu register stack arguments.
* If its using an intermediate fpu register, it has 80/64 bits to work
* with. If it uses memory, it has 64/53 bits to work with. However,
* gcc is aware of this and goes to a fair bit of trouble to make the
* best use of it.
*
* This is mostly academic for AMD64, because the ABI prefers the use
* SSE2 based math. For FreeBSD/amd64, we go with the default settings.
*/
#define __INITIAL_FPUCW__ 0x037F
#define __INITIAL_FPUCW_I386__ 0x127F
#define __INITIAL_NPXCW__ __INITIAL_FPUCW_I386__
#define __INITIAL_MXCSR__ 0x1F80
#define __INITIAL_MXCSR_MASK__ 0xFFBF
#endif /* !_X86_FPU_H_ */
+1 -1
View File
@@ -3,7 +3,7 @@
#if defined(__INTEL__)
# include <arch/x86/fenv.h>
#elif defined(_x86_64_)
#elif defined(__x86_64__)
# include <arch/x86_64/fenv.h>
#elif defined(__ARM__)
# include <arch/arm/fenv.h>
+1 -1
View File
@@ -8,7 +8,7 @@ local genericSources =
;
MergeObject posix_arch_$(TARGET_ARCH).o :
#fenv.c
fenv.c
sigsetjmp.S
siglongjmp.S
+158
View File
@@ -0,0 +1,158 @@
/*-
* 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.
*
* $FreeBSD$
*/
#include <sys/cdefs.h>
#include <sys/types.h>
#include <posix/fenv.h>
#include <arch/x86_64/fpu.h>
const fenv_t __fe_dfl_env = {
{ 0xffff0000 | __INITIAL_FPUCW__,
0xffff0000,
0xffffffff,
{ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xff, 0xff }
},
__INITIAL_MXCSR__
};
extern inline int feclearexcept(int __excepts);
extern inline int fegetexceptflag(fexcept_t *__flagp, int __excepts);
int
fesetexceptflag(const fexcept_t *flagp, int excepts)
{
fenv_t env;
__fnstenv(&env.__x87);
env.__x87.__status &= ~excepts;
env.__x87.__status |= *flagp & excepts;
__fldenv(env.__x87);
__stmxcsr(&env.__mxcsr);
env.__mxcsr &= ~excepts;
env.__mxcsr |= *flagp & excepts;
__ldmxcsr(env.__mxcsr);
return (0);
}
int
feraiseexcept(int excepts)
{
fexcept_t ex = excepts;
fesetexceptflag(&ex, excepts);
__fwait();
return (0);
}
extern inline int fetestexcept(int __excepts);
extern inline int fegetround(void);
extern inline int fesetround(int __round);
int
fegetenv(fenv_t *envp)
{
__fnstenv(&envp->__x87);
__stmxcsr(&envp->__mxcsr);
/*
* fnstenv masks all exceptions, so we need to restore the
* control word to avoid this side effect.
*/
__fldcw(envp->__x87.__control);
return (0);
}
int
feholdexcept(fenv_t *envp)
{
uint32_t mxcsr;
__stmxcsr(&mxcsr);
__fnstenv(&envp->__x87);
__fnclex();
envp->__mxcsr = mxcsr;
mxcsr &= ~FE_ALL_EXCEPT;
mxcsr |= FE_ALL_EXCEPT << _SSE_EMASK_SHIFT;
__ldmxcsr(mxcsr);
return (0);
}
extern inline int fesetenv(const fenv_t *__envp);
int
feupdateenv(const fenv_t *envp)
{
uint32_t mxcsr;
uint16_t status;
__fnstsw(&status);
__stmxcsr(&mxcsr);
fesetenv(envp);
feraiseexcept((mxcsr | status) & FE_ALL_EXCEPT);
return (0);
}
int
__feenableexcept(int mask)
{
uint32_t mxcsr, omask;
uint16_t control;
mask &= FE_ALL_EXCEPT;
__fnstcw(&control);
__stmxcsr(&mxcsr);
omask = ~(control | mxcsr >> _SSE_EMASK_SHIFT) & FE_ALL_EXCEPT;
control &= ~mask;
__fldcw(control);
mxcsr &= ~(mask << _SSE_EMASK_SHIFT);
__ldmxcsr(mxcsr);
return (omask);
}
int
__fedisableexcept(int mask)
{
uint32_t mxcsr, omask;
uint16_t control;
mask &= FE_ALL_EXCEPT;
__fnstcw(&control);
__stmxcsr(&mxcsr);
omask = ~(control | mxcsr >> _SSE_EMASK_SHIFT) & FE_ALL_EXCEPT;
control |= mask;
__fldcw(control);
mxcsr |= mask << _SSE_EMASK_SHIFT;
__ldmxcsr(mxcsr);
return (omask);
}
__weak_reference(__feenableexcept, feenableexcept);
__weak_reference(__fedisableexcept, fedisableexcept);