| 1 |
282 |
jeremybenn |
/*
|
| 2 |
|
|
* Copyright (C) 2005, 2007, 2009 Free Software Foundation, Inc.
|
| 3 |
|
|
*
|
| 4 |
|
|
* This file is free software; you can redistribute it and/or modify it
|
| 5 |
|
|
* under the terms of the GNU General Public License as published by the
|
| 6 |
|
|
* Free Software Foundation; either version 3, or (at your option) any
|
| 7 |
|
|
* later version.
|
| 8 |
|
|
*
|
| 9 |
|
|
* This file is distributed in the hope that it will be useful, but
|
| 10 |
|
|
* WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 11 |
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
| 12 |
|
|
* General Public License for more details.
|
| 13 |
|
|
*
|
| 14 |
|
|
* Under Section 7 of GPL version 3, you are granted additional
|
| 15 |
|
|
* permissions described in the GCC Runtime Library Exception, version
|
| 16 |
|
|
* 3.1, as published by the Free Software Foundation.
|
| 17 |
|
|
*
|
| 18 |
|
|
* You should have received a copy of the GNU General Public License and
|
| 19 |
|
|
* a copy of the GCC Runtime Library Exception along with this program;
|
| 20 |
|
|
* see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
| 21 |
|
|
* <http://www.gnu.org/licenses/>.
|
| 22 |
|
|
*/
|
| 23 |
|
|
|
| 24 |
|
|
#define MXCSR_DAZ (1 << 6) /* Enable denormals are zero mode */
|
| 25 |
|
|
#define MXCSR_FTZ (1 << 15) /* Enable flush to zero mode */
|
| 26 |
|
|
|
| 27 |
|
|
#ifndef __x86_64__
|
| 28 |
|
|
/* All 64-bit targets have SSE and DAZ;
|
| 29 |
|
|
only check them explicitly for 32-bit ones. */
|
| 30 |
|
|
#include "cpuid.h"
|
| 31 |
|
|
#endif
|
| 32 |
|
|
|
| 33 |
|
|
static void __attribute__((constructor))
|
| 34 |
|
|
#ifndef __x86_64__
|
| 35 |
|
|
/* The i386 ABI only requires 4-byte stack alignment, so this is necessary
|
| 36 |
|
|
to make sure the fxsave struct gets correct alignment.
|
| 37 |
|
|
See PR27537 and PR28621. */
|
| 38 |
|
|
__attribute__ ((force_align_arg_pointer))
|
| 39 |
|
|
#endif
|
| 40 |
|
|
set_fast_math (void)
|
| 41 |
|
|
{
|
| 42 |
|
|
#ifndef __x86_64__
|
| 43 |
|
|
unsigned int eax, ebx, ecx, edx;
|
| 44 |
|
|
|
| 45 |
|
|
if (!__get_cpuid (1, &eax, &ebx, &ecx, &edx))
|
| 46 |
|
|
return;
|
| 47 |
|
|
|
| 48 |
|
|
if (edx & bit_SSE)
|
| 49 |
|
|
{
|
| 50 |
|
|
unsigned int mxcsr = __builtin_ia32_stmxcsr ();
|
| 51 |
|
|
|
| 52 |
|
|
mxcsr |= MXCSR_FTZ;
|
| 53 |
|
|
|
| 54 |
|
|
if (edx & bit_FXSAVE)
|
| 55 |
|
|
{
|
| 56 |
|
|
/* Check if DAZ is available. */
|
| 57 |
|
|
struct
|
| 58 |
|
|
{
|
| 59 |
|
|
unsigned short int cwd;
|
| 60 |
|
|
unsigned short int swd;
|
| 61 |
|
|
unsigned short int twd;
|
| 62 |
|
|
unsigned short int fop;
|
| 63 |
|
|
long int fip;
|
| 64 |
|
|
long int fcs;
|
| 65 |
|
|
long int foo;
|
| 66 |
|
|
long int fos;
|
| 67 |
|
|
long int mxcsr;
|
| 68 |
|
|
long int mxcsr_mask;
|
| 69 |
|
|
long int st_space[32];
|
| 70 |
|
|
long int xmm_space[32];
|
| 71 |
|
|
long int padding[56];
|
| 72 |
|
|
} __attribute__ ((aligned (16))) fxsave;
|
| 73 |
|
|
|
| 74 |
|
|
__builtin_memset (&fxsave, 0, sizeof (fxsave));
|
| 75 |
|
|
|
| 76 |
|
|
asm volatile ("fxsave %0" : "=m" (fxsave) : "m" (fxsave));
|
| 77 |
|
|
|
| 78 |
|
|
if (fxsave.mxcsr_mask & MXCSR_DAZ)
|
| 79 |
|
|
mxcsr |= MXCSR_DAZ;
|
| 80 |
|
|
}
|
| 81 |
|
|
|
| 82 |
|
|
__builtin_ia32_ldmxcsr (mxcsr);
|
| 83 |
|
|
}
|
| 84 |
|
|
#else
|
| 85 |
|
|
unsigned int mxcsr = __builtin_ia32_stmxcsr ();
|
| 86 |
|
|
mxcsr |= MXCSR_DAZ | MXCSR_FTZ;
|
| 87 |
|
|
__builtin_ia32_ldmxcsr (mxcsr);
|
| 88 |
|
|
#endif
|
| 89 |
|
|
}
|
