1/*
2 * Copyright (C) 2005 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 2, or (at your option) any
7 * later version.
8 *
9 * In addition to the permissions in the GNU General Public License, the
10 * Free Software Foundation gives you unlimited permission to link the
11 * compiled version of this file with other programs, and to distribute
12 * those programs without any restriction coming from the use of this
13 * file.  (The General Public License restrictions do apply in other
14 * respects; for example, they cover modification of the file, and
15 * distribution when not linked into another program.)
16 *
17 * This file is distributed in the hope that it will be useful, but
18 * WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
20 * General Public License for more details.
21 *
22 * You should have received a copy of the GNU General Public License
23 * along with this program; see the file COPYING.  If not, write to
24 * the Free Software Foundation, 51 Franklin Street, Fifth Floor,
25 * Boston, MA 02110-1301, USA.
26 *
27 *    As a special exception, if you link this library with files
28 *    compiled with GCC to produce an executable, this does not cause
29 *    the resulting executable to be covered by the GNU General Public License.
30 *    This exception does not however invalidate any other reasons why
31 *    the executable file might be covered by the GNU General Public License.
32 */
33
34#define MXCSR_DAZ (1 << 6)	/* Enable denormals are zero mode */
35#define MXCSR_FTZ (1 << 15)	/* Enable flush to zero mode */
36
37#define FXSAVE	(1 << 24)
38#define SSE	(1 << 25)
39
40static void __attribute__((constructor))
41#ifndef __x86_64__
42/* The i386 ABI only requires 4-byte stack alignment, so this is necessary
43   to make sure the fxsave struct gets correct alignment.
44   See PR27537 and PR28621.  */
45__attribute__ ((force_align_arg_pointer))
46#endif
47set_fast_math (void)
48{
49#ifndef __x86_64__
50  /* All 64-bit targets have SSE and DAZ; only check them explicitly
51     for 32-bit ones. */
52  unsigned int eax, ebx, ecx, edx;
53
54  /* See if we can use cpuid.  */
55  asm volatile ("pushfl; pushfl; popl %0; movl %0,%1; xorl %2,%0;"
56		"pushl %0; popfl; pushfl; popl %0; popfl"
57		: "=&r" (eax), "=&r" (ebx)
58		: "i" (0x00200000));
59
60  if (((eax ^ ebx) & 0x00200000) == 0)
61    return;
62
63  /* Check the highest input value for eax.  */
64  asm volatile ("xchgl %%ebx, %1; cpuid; xchgl %%ebx, %1"
65		: "=a" (eax), "=r" (ebx), "=c" (ecx), "=d" (edx)
66		: "0" (0));
67
68  if (eax == 0)
69    return;
70
71  asm volatile ("xchgl %%ebx, %1; cpuid; xchgl %%ebx, %1"
72		: "=a" (eax), "=r" (ebx), "=c" (ecx), "=d" (edx)
73		: "0" (1));
74
75  if (edx & SSE)
76    {
77      unsigned int mxcsr = __builtin_ia32_stmxcsr ();
78
79      mxcsr |= MXCSR_FTZ;
80
81      if (edx & FXSAVE)
82	{
83	  /* Check if DAZ is available.  */
84	  struct
85	    {
86	      unsigned short int cwd;
87	      unsigned short int swd;
88	      unsigned short int twd;
89	      unsigned short int fop;
90	      long int fip;
91	      long int fcs;
92	      long int foo;
93	      long int fos;
94	      long int mxcsr;
95	      long int mxcsr_mask;
96	      long int st_space[32];
97	      long int xmm_space[32];
98	      long int padding[56];
99	    } __attribute__ ((aligned (16))) fxsave;
100
101	  __builtin_memset (&fxsave, 0, sizeof (fxsave));
102
103	  asm volatile ("fxsave %0" : "=m" (fxsave) : "m" (fxsave));
104
105	  if (fxsave.mxcsr_mask & MXCSR_DAZ)
106	    mxcsr |= MXCSR_DAZ;
107	}
108
109      __builtin_ia32_ldmxcsr (mxcsr);
110    }
111#else
112  unsigned int mxcsr = __builtin_ia32_stmxcsr ();
113  mxcsr |= MXCSR_DAZ | MXCSR_FTZ;
114  __builtin_ia32_ldmxcsr (mxcsr);
115#endif
116}
117