1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
|
/* GRAPHITE2 LICENSING
Copyright 2010, SIL International
All rights reserved.
This library is free software; you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published
by the Free Software Foundation; either version 2.1 of License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should also have received a copy of the GNU Lesser General Public
License along with this library in the file named "LICENSE".
If not, write to the Free Software Foundation, 51 Franklin Street,
Suite 500, Boston, MA 02110-1335, USA or visit their web page on the
internet at http://www.fsf.org/licenses/lgpl.html.
Alternatively, the contents of this file may be used under the terms of the
Mozilla Public License (http://mozilla.org/MPL) or the GNU General Public
License, as published by the Free Software Foundation, either version 2
of the License or (at your option) any later version.
*/
// This direct threaded interpreter implmentation for machine.h
// Author: Tim Eves
// Build either this interpreter or the call_machine implementation.
// The direct threaded interpreter is relies upon a gcc feature called
// labels-as-values so is only portable to compilers that support the
// extension (gcc only as far as I know) however it should build on any
// architecture gcc supports.
// This is twice as fast as the call threaded model and is likely faster on
// inorder processors with short pipelines and little branch prediction such
// as the ARM and possibly Atom chips.
#include <cassert>
#include <cstring>
#include "inc/Machine.h"
#include "inc/Segment.h"
#include "inc/Slot.h"
#include "inc/Rule.h"
#define STARTOP(name) name: {
#define ENDOP }; goto *((sp - sb)/Machine::STACK_MAX ? &&end : *++ip);
#define EXIT(status) { push(status); goto end; }
#define do_(name) &&name
using namespace graphite2;
using namespace vm;
namespace {
// The GCC manual has this to say about labels as values:
// The &&foo expressions for the same label might have different values
// if the containing function is inlined or cloned. If a program relies
// on them being always the same, __attribute__((__noinline__,__noclone__))
// should be used to prevent inlining and cloning.
//
// is_return in Code.cpp relies on being able to do comparisons, so it needs
// them to be always the same.
//
// The GCC manual further adds:
// If &&foo is used in a static variable initializer, inlining and
// cloning is forbidden.
//
// In this file, &&foo *is* used in a static variable initializer, and it's not
// entirely clear whether this should prevent inlining of the function or not.
// In practice, though, clang 7 can end up inlining the function with ThinLTO,
// which breaks at least is_return. https://bugs.llvm.org/show_bug.cgi?id=39241
// So all in all, we need at least the __noinline__ attribute. __noclone__
// is not supported by clang.
__attribute__((__noinline__))
const void * direct_run(const bool get_table_mode,
const instr * program,
const byte * data,
Machine::stack_t * stack,
slotref * & __map,
uint8 _dir,
Machine::status_t & status,
SlotMap * __smap=0)
{
// We need to define and return to opcode table from within this function
// other inorder to take the addresses of the instruction bodies.
#include "inc/opcode_table.h"
if (get_table_mode)
return opcode_table;
// Declare virtual machine registers
const instr * ip = program;
const byte * dp = data;
Machine::stack_t * sp = stack + Machine::STACK_GUARD,
* const sb = sp;
SlotMap & smap = *__smap;
Segment & seg = smap.segment;
slotref is = *__map,
* map = __map,
* const mapb = smap.begin()+smap.context();
uint8 dir = _dir;
int8 flags = 0;
// start the program
goto **ip;
// Pull in the opcode definitions
#include "inc/opcodes.h"
end:
__map = map;
*__map = is;
return sp;
}
}
const opcode_t * Machine::getOpcodeTable() throw()
{
slotref * dummy;
Machine::status_t dumstat = Machine::finished;
return static_cast<const opcode_t *>(direct_run(true, 0, 0, 0, dummy, 0, dumstat));
}
Machine::stack_t Machine::run(const instr * program,
const byte * data,
slotref * & is)
{
assert(program != 0);
const stack_t *sp = static_cast<const stack_t *>(
direct_run(false, program, data, _stack, is, _map.dir(), _status, &_map));
const stack_t ret = sp == _stack+STACK_GUARD+1 ? *sp-- : 0;
check_final_stack(sp);
return ret;
}
|