1 |
733 |
jeremybenn |
/* FPU-related code for x86 and x86_64 processors.
|
2 |
|
|
Copyright 2005, 2007, 2009, 2010, 2011 Free Software Foundation, Inc.
|
3 |
|
|
Contributed by Francois-Xavier Coudert <coudert@clipper.ens.fr>
|
4 |
|
|
|
5 |
|
|
This file is part of the GNU Fortran 95 runtime library (libgfortran).
|
6 |
|
|
|
7 |
|
|
Libgfortran is free software; you can redistribute it and/or
|
8 |
|
|
modify it under the terms of the GNU General Public
|
9 |
|
|
License as published by the Free Software Foundation; either
|
10 |
|
|
version 3 of the License, or (at your option) any later version.
|
11 |
|
|
|
12 |
|
|
Libgfortran is distributed in the hope that it will be useful,
|
13 |
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
14 |
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
15 |
|
|
GNU General Public License for more details.
|
16 |
|
|
|
17 |
|
|
Under Section 7 of GPL version 3, you are granted additional
|
18 |
|
|
permissions described in the GCC Runtime Library Exception, version
|
19 |
|
|
3.1, as published by the Free Software Foundation.
|
20 |
|
|
|
21 |
|
|
You should have received a copy of the GNU General Public License and
|
22 |
|
|
a copy of the GCC Runtime Library Exception along with this program;
|
23 |
|
|
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
24 |
|
|
<http://www.gnu.org/licenses/>. */
|
25 |
|
|
|
26 |
|
|
#ifndef __x86_64__
|
27 |
|
|
#include "cpuid.h"
|
28 |
|
|
#endif
|
29 |
|
|
|
30 |
|
|
#if defined(__sun__) && defined(__svr4__)
|
31 |
|
|
#include <signal.h>
|
32 |
|
|
#include <ucontext.h>
|
33 |
|
|
|
34 |
|
|
static volatile sig_atomic_t sigill_caught;
|
35 |
|
|
|
36 |
|
|
static void
|
37 |
|
|
sigill_hdlr (int sig __attribute((unused)),
|
38 |
|
|
siginfo_t *sip __attribute__((unused)),
|
39 |
|
|
ucontext_t *ucp)
|
40 |
|
|
{
|
41 |
|
|
sigill_caught = 1;
|
42 |
|
|
/* Set PC to the instruction after the faulting one to skip over it,
|
43 |
|
|
otherwise we enter an infinite loop. 3 is the size of the movaps
|
44 |
|
|
instruction. */
|
45 |
|
|
ucp->uc_mcontext.gregs[EIP] += 3;
|
46 |
|
|
setcontext (ucp);
|
47 |
|
|
}
|
48 |
|
|
#endif
|
49 |
|
|
|
50 |
|
|
static int
|
51 |
|
|
has_sse (void)
|
52 |
|
|
{
|
53 |
|
|
#ifndef __x86_64__
|
54 |
|
|
unsigned int eax, ebx, ecx, edx;
|
55 |
|
|
|
56 |
|
|
if (!__get_cpuid (1, &eax, &ebx, &ecx, &edx))
|
57 |
|
|
return 0;
|
58 |
|
|
|
59 |
|
|
#if defined(__sun__) && defined(__svr4__)
|
60 |
|
|
/* Solaris 2 before Solaris 9 4/04 cannot execute SSE instructions even
|
61 |
|
|
if the CPU supports them. Programs receive SIGILL instead, so check
|
62 |
|
|
for that at runtime. */
|
63 |
|
|
|
64 |
|
|
if (edx & bit_SSE)
|
65 |
|
|
{
|
66 |
|
|
struct sigaction act, oact;
|
67 |
|
|
|
68 |
|
|
act.sa_handler = sigill_hdlr;
|
69 |
|
|
sigemptyset (&act.sa_mask);
|
70 |
|
|
/* Need to set SA_SIGINFO so a ucontext_t * is passed to the handler. */
|
71 |
|
|
act.sa_flags = SA_SIGINFO;
|
72 |
|
|
sigaction (SIGILL, &act, &oact);
|
73 |
|
|
|
74 |
|
|
/* We need a single SSE instruction here so the handler can safely skip
|
75 |
|
|
over it. */
|
76 |
|
|
__asm__ volatile ("movaps %xmm0,%xmm0");
|
77 |
|
|
|
78 |
|
|
sigaction (SIGILL, &oact, NULL);
|
79 |
|
|
|
80 |
|
|
if (sigill_caught)
|
81 |
|
|
return 0;
|
82 |
|
|
}
|
83 |
|
|
#endif /* __sun__ && __svr4__ */
|
84 |
|
|
|
85 |
|
|
return edx & bit_SSE;
|
86 |
|
|
#else
|
87 |
|
|
return 1;
|
88 |
|
|
#endif
|
89 |
|
|
}
|
90 |
|
|
|
91 |
|
|
/* i387 -- see linux <fpu_control.h> header file for details. */
|
92 |
|
|
#define _FPU_MASK_IM 0x01
|
93 |
|
|
#define _FPU_MASK_DM 0x02
|
94 |
|
|
#define _FPU_MASK_ZM 0x04
|
95 |
|
|
#define _FPU_MASK_OM 0x08
|
96 |
|
|
#define _FPU_MASK_UM 0x10
|
97 |
|
|
#define _FPU_MASK_PM 0x20
|
98 |
|
|
|
99 |
|
|
void set_fpu (void)
|
100 |
|
|
{
|
101 |
|
|
unsigned short cw;
|
102 |
|
|
|
103 |
|
|
asm volatile ("fnstcw %0" : "=m" (cw));
|
104 |
|
|
|
105 |
|
|
cw |= (_FPU_MASK_IM | _FPU_MASK_DM | _FPU_MASK_ZM | _FPU_MASK_OM
|
106 |
|
|
| _FPU_MASK_UM | _FPU_MASK_PM);
|
107 |
|
|
|
108 |
|
|
if (options.fpe & GFC_FPE_INVALID) cw &= ~_FPU_MASK_IM;
|
109 |
|
|
if (options.fpe & GFC_FPE_DENORMAL) cw &= ~_FPU_MASK_DM;
|
110 |
|
|
if (options.fpe & GFC_FPE_ZERO) cw &= ~_FPU_MASK_ZM;
|
111 |
|
|
if (options.fpe & GFC_FPE_OVERFLOW) cw &= ~_FPU_MASK_OM;
|
112 |
|
|
if (options.fpe & GFC_FPE_UNDERFLOW) cw &= ~_FPU_MASK_UM;
|
113 |
|
|
if (options.fpe & GFC_FPE_INEXACT) cw &= ~_FPU_MASK_PM;
|
114 |
|
|
|
115 |
|
|
asm volatile ("fldcw %0" : : "m" (cw));
|
116 |
|
|
|
117 |
|
|
if (has_sse())
|
118 |
|
|
{
|
119 |
|
|
unsigned int cw_sse;
|
120 |
|
|
|
121 |
|
|
asm volatile ("stmxcsr %0" : "=m" (cw_sse));
|
122 |
|
|
|
123 |
|
|
cw_sse &= 0xffff0000;
|
124 |
|
|
cw_sse |= (_FPU_MASK_IM | _FPU_MASK_DM | _FPU_MASK_ZM | _FPU_MASK_OM
|
125 |
|
|
| _FPU_MASK_UM | _FPU_MASK_PM ) << 7;
|
126 |
|
|
|
127 |
|
|
if (options.fpe & GFC_FPE_INVALID) cw_sse &= ~(_FPU_MASK_IM << 7);
|
128 |
|
|
if (options.fpe & GFC_FPE_DENORMAL) cw_sse &= ~(_FPU_MASK_DM << 7);
|
129 |
|
|
if (options.fpe & GFC_FPE_ZERO) cw_sse &= ~(_FPU_MASK_ZM << 7);
|
130 |
|
|
if (options.fpe & GFC_FPE_OVERFLOW) cw_sse &= ~(_FPU_MASK_OM << 7);
|
131 |
|
|
if (options.fpe & GFC_FPE_UNDERFLOW) cw_sse &= ~(_FPU_MASK_UM << 7);
|
132 |
|
|
if (options.fpe & GFC_FPE_INEXACT) cw_sse &= ~(_FPU_MASK_PM << 7);
|
133 |
|
|
|
134 |
|
|
asm volatile ("ldmxcsr %0" : : "m" (cw_sse));
|
135 |
|
|
}
|
136 |
|
|
}
|