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
141
142
|
//===- DisassemblerEmitter.cpp - Generate a disassembler ------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#include "DisassemblerEmitter.h"
#include "CodeGenTarget.h"
#include "X86DisassemblerTables.h"
#include "X86RecognizableInstr.h"
#include "FixedLenDecoderEmitter.h"
#include "llvm/TableGen/Error.h"
#include "llvm/TableGen/Record.h"
using namespace llvm;
using namespace llvm::X86Disassembler;
/// DisassemblerEmitter - Contains disassembler table emitters for various
/// architectures.
/// X86 Disassembler Emitter
///
/// *** IF YOU'RE HERE TO RESOLVE A "Primary decode conflict", LOOK DOWN NEAR
/// THE END OF THIS COMMENT!
///
/// The X86 disassembler emitter is part of the X86 Disassembler, which is
/// documented in lib/Target/X86/X86Disassembler.h.
///
/// The emitter produces the tables that the disassembler uses to translate
/// instructions. The emitter generates the following tables:
///
/// - One table (CONTEXTS_SYM) that contains a mapping of attribute masks to
/// instruction contexts. Although for each attribute there are cases where
/// that attribute determines decoding, in the majority of cases decoding is
/// the same whether or not an attribute is present. For example, a 64-bit
/// instruction with an OPSIZE prefix and an XS prefix decodes the same way in
/// all cases as a 64-bit instruction with only OPSIZE set. (The XS prefix
/// may have effects on its execution, but does not change the instruction
/// returned.) This allows considerable space savings in other tables.
/// - Six tables (ONEBYTE_SYM, TWOBYTE_SYM, THREEBYTE38_SYM, THREEBYTE3A_SYM,
/// THREEBYTEA6_SYM, and THREEBYTEA7_SYM contain the hierarchy that the
/// decoder traverses while decoding an instruction. At the lowest level of
/// this hierarchy are instruction UIDs, 16-bit integers that can be used to
/// uniquely identify the instruction and correspond exactly to its position
/// in the list of CodeGenInstructions for the target.
/// - One table (INSTRUCTIONS_SYM) contains information about the operands of
/// each instruction and how to decode them.
///
/// During table generation, there may be conflicts between instructions that
/// occupy the same space in the decode tables. These conflicts are resolved as
/// follows in setTableFields() (X86DisassemblerTables.cpp)
///
/// - If the current context is the native context for one of the instructions
/// (that is, the attributes specified for it in the LLVM tables specify
/// precisely the current context), then it has priority.
/// - If the current context isn't native for either of the instructions, then
/// the higher-priority context wins (that is, the one that is more specific).
/// That hierarchy is determined by outranks() (X86DisassemblerTables.cpp)
/// - If the current context is native for both instructions, then the table
/// emitter reports a conflict and dies.
///
/// *** RESOLUTION FOR "Primary decode conflict"S
///
/// If two instructions collide, typically the solution is (in order of
/// likelihood):
///
/// (1) to filter out one of the instructions by editing filter()
/// (X86RecognizableInstr.cpp). This is the most common resolution, but
/// check the Intel manuals first to make sure that (2) and (3) are not the
/// problem.
/// (2) to fix the tables (X86.td and its subsidiaries) so the opcodes are
/// accurate. Sometimes they are not.
/// (3) to fix the tables to reflect the actual context (for example, required
/// prefixes), and possibly to add a new context by editing
/// lib/Target/X86/X86DisassemblerDecoderCommon.h. This is unlikely to be
/// the cause.
///
/// DisassemblerEmitter.cpp contains the implementation for the emitter,
/// which simply pulls out instructions from the CodeGenTarget and pushes them
/// into X86DisassemblerTables.
/// X86DisassemblerTables.h contains the interface for the instruction tables,
/// which manage and emit the structures discussed above.
/// X86DisassemblerTables.cpp contains the implementation for the instruction
/// tables.
/// X86ModRMFilters.h contains filters that can be used to determine which
/// ModR/M values are valid for a particular instruction. These are used to
/// populate ModRMDecisions.
/// X86RecognizableInstr.h contains the interface for a single instruction,
/// which knows how to translate itself from a CodeGenInstruction and provide
/// the information necessary for integration into the tables.
/// X86RecognizableInstr.cpp contains the implementation for a single
/// instruction.
void DisassemblerEmitter::run(raw_ostream &OS) {
CodeGenTarget Target(Records);
OS << "/*===- TableGen'erated file "
<< "---------------------------------------*- C -*-===*\n"
<< " *\n"
<< " * " << Target.getName() << " Disassembler\n"
<< " *\n"
<< " * Automatically generated file, do not edit!\n"
<< " *\n"
<< " *===---------------------------------------------------------------"
<< "-------===*/\n";
// X86 uses a custom disassembler.
if (Target.getName() == "X86") {
DisassemblerTables Tables;
const std::vector<const CodeGenInstruction*> &numberedInstructions =
Target.getInstructionsByEnumValue();
for (unsigned i = 0, e = numberedInstructions.size(); i != e; ++i)
RecognizableInstr::processInstr(Tables, *numberedInstructions[i], i);
// FIXME: As long as we are using exceptions, might as well drop this to the
// actual conflict site.
if (Tables.hasConflicts())
throw TGError(Target.getTargetRecord()->getLoc(),
"Primary decode conflict");
Tables.emit(OS);
return;
}
// ARM and Thumb have a CHECK() macro to deal with DecodeStatuses.
if (Target.getName() == "ARM" ||
Target.getName() == "Thumb") {
FixedLenDecoderEmitter(Records,
"ARM",
"if (!Check(S, ", ")) return MCDisassembler::Fail;",
"S", "MCDisassembler::Fail",
" MCDisassembler::DecodeStatus S = MCDisassembler::Success;\n(void)S;").run(OS);
return;
}
FixedLenDecoderEmitter(Records, Target.getName()).run(OS);
}
|
