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
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
|
//===-- SIInstrInfo.cpp - SI Instruction Information ---------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
/// \file
/// \brief SI Implementation of TargetInstrInfo.
//
//===----------------------------------------------------------------------===//
#include "SIInstrInfo.h"
#include "AMDGPUTargetMachine.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/MC/MCInstrDesc.h"
#include <stdio.h>
using namespace llvm;
SIInstrInfo::SIInstrInfo(AMDGPUTargetMachine &tm)
: AMDGPUInstrInfo(tm),
RI(tm, *this)
{ }
const SIRegisterInfo &SIInstrInfo::getRegisterInfo() const {
return RI;
}
void
SIInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI, DebugLoc DL,
unsigned DestReg, unsigned SrcReg,
bool KillSrc) const {
// If we are trying to copy to or from SCC, there is a bug somewhere else in
// the backend. While it may be theoretically possible to do this, it should
// never be necessary.
assert(DestReg != AMDGPU::SCC && SrcReg != AMDGPU::SCC);
const int16_t Sub0_15[] = {
AMDGPU::sub0, AMDGPU::sub1, AMDGPU::sub2, AMDGPU::sub3,
AMDGPU::sub4, AMDGPU::sub5, AMDGPU::sub6, AMDGPU::sub7,
AMDGPU::sub8, AMDGPU::sub9, AMDGPU::sub10, AMDGPU::sub11,
AMDGPU::sub12, AMDGPU::sub13, AMDGPU::sub14, AMDGPU::sub15, 0
};
const int16_t Sub0_7[] = {
AMDGPU::sub0, AMDGPU::sub1, AMDGPU::sub2, AMDGPU::sub3,
AMDGPU::sub4, AMDGPU::sub5, AMDGPU::sub6, AMDGPU::sub7, 0
};
const int16_t Sub0_3[] = {
AMDGPU::sub0, AMDGPU::sub1, AMDGPU::sub2, AMDGPU::sub3, 0
};
const int16_t Sub0_1[] = {
AMDGPU::sub0, AMDGPU::sub1, 0
};
unsigned Opcode;
const int16_t *SubIndices;
if (AMDGPU::SReg_32RegClass.contains(DestReg)) {
assert(AMDGPU::SReg_32RegClass.contains(SrcReg));
BuildMI(MBB, MI, DL, get(AMDGPU::S_MOV_B32), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc));
return;
} else if (AMDGPU::SReg_64RegClass.contains(DestReg)) {
assert(AMDGPU::SReg_64RegClass.contains(SrcReg));
BuildMI(MBB, MI, DL, get(AMDGPU::S_MOV_B64), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc));
return;
} else if (AMDGPU::SReg_128RegClass.contains(DestReg)) {
assert(AMDGPU::SReg_128RegClass.contains(SrcReg));
Opcode = AMDGPU::S_MOV_B32;
SubIndices = Sub0_3;
} else if (AMDGPU::SReg_256RegClass.contains(DestReg)) {
assert(AMDGPU::SReg_256RegClass.contains(SrcReg));
Opcode = AMDGPU::S_MOV_B32;
SubIndices = Sub0_7;
} else if (AMDGPU::SReg_512RegClass.contains(DestReg)) {
assert(AMDGPU::SReg_512RegClass.contains(SrcReg));
Opcode = AMDGPU::S_MOV_B32;
SubIndices = Sub0_15;
} else if (AMDGPU::VReg_32RegClass.contains(DestReg)) {
assert(AMDGPU::VReg_32RegClass.contains(SrcReg) ||
AMDGPU::SReg_32RegClass.contains(SrcReg));
BuildMI(MBB, MI, DL, get(AMDGPU::V_MOV_B32_e32), DestReg)
.addReg(SrcReg, getKillRegState(KillSrc));
return;
} else if (AMDGPU::VReg_64RegClass.contains(DestReg)) {
assert(AMDGPU::VReg_64RegClass.contains(SrcReg) ||
AMDGPU::SReg_64RegClass.contains(SrcReg));
Opcode = AMDGPU::V_MOV_B32_e32;
SubIndices = Sub0_1;
} else if (AMDGPU::VReg_128RegClass.contains(DestReg)) {
assert(AMDGPU::VReg_128RegClass.contains(SrcReg) ||
AMDGPU::SReg_128RegClass.contains(SrcReg));
Opcode = AMDGPU::V_MOV_B32_e32;
SubIndices = Sub0_3;
} else if (AMDGPU::VReg_256RegClass.contains(DestReg)) {
assert(AMDGPU::VReg_256RegClass.contains(SrcReg) ||
AMDGPU::SReg_256RegClass.contains(SrcReg));
Opcode = AMDGPU::V_MOV_B32_e32;
SubIndices = Sub0_7;
} else if (AMDGPU::VReg_512RegClass.contains(DestReg)) {
assert(AMDGPU::VReg_512RegClass.contains(SrcReg) ||
AMDGPU::SReg_512RegClass.contains(SrcReg));
Opcode = AMDGPU::V_MOV_B32_e32;
SubIndices = Sub0_15;
} else {
llvm_unreachable("Can't copy register!");
}
while (unsigned SubIdx = *SubIndices++) {
MachineInstrBuilder Builder = BuildMI(MBB, MI, DL,
get(Opcode), RI.getSubReg(DestReg, SubIdx));
Builder.addReg(RI.getSubReg(SrcReg, SubIdx), getKillRegState(KillSrc));
if (*SubIndices)
Builder.addReg(DestReg, RegState::Define | RegState::Implicit);
}
}
MachineInstr *SIInstrInfo::commuteInstruction(MachineInstr *MI,
bool NewMI) const {
if (MI->getNumOperands() < 3 || !MI->getOperand(1).isReg() ||
!MI->getOperand(2).isReg())
return 0;
return TargetInstrInfo::commuteInstruction(MI, NewMI);
}
MachineInstr * SIInstrInfo::getMovImmInstr(MachineFunction *MF, unsigned DstReg,
int64_t Imm) const {
MachineInstr * MI = MF->CreateMachineInstr(get(AMDGPU::V_MOV_B32_e32), DebugLoc());
MachineInstrBuilder MIB(*MF, MI);
MIB.addReg(DstReg, RegState::Define);
MIB.addImm(Imm);
return MI;
}
bool SIInstrInfo::isMov(unsigned Opcode) const {
switch(Opcode) {
default: return false;
case AMDGPU::S_MOV_B32:
case AMDGPU::S_MOV_B64:
case AMDGPU::V_MOV_B32_e32:
case AMDGPU::V_MOV_B32_e64:
return true;
}
}
bool
SIInstrInfo::isSafeToMoveRegClassDefs(const TargetRegisterClass *RC) const {
return RC != &AMDGPU::EXECRegRegClass;
}
//===----------------------------------------------------------------------===//
// Indirect addressing callbacks
//===----------------------------------------------------------------------===//
unsigned SIInstrInfo::calculateIndirectAddress(unsigned RegIndex,
unsigned Channel) const {
assert(Channel == 0);
return RegIndex;
}
int SIInstrInfo::getIndirectIndexBegin(const MachineFunction &MF) const {
llvm_unreachable("Unimplemented");
}
int SIInstrInfo::getIndirectIndexEnd(const MachineFunction &MF) const {
llvm_unreachable("Unimplemented");
}
const TargetRegisterClass *SIInstrInfo::getIndirectAddrStoreRegClass(
unsigned SourceReg) const {
llvm_unreachable("Unimplemented");
}
const TargetRegisterClass *SIInstrInfo::getIndirectAddrLoadRegClass() const {
llvm_unreachable("Unimplemented");
}
MachineInstrBuilder SIInstrInfo::buildIndirectWrite(
MachineBasicBlock *MBB,
MachineBasicBlock::iterator I,
unsigned ValueReg,
unsigned Address, unsigned OffsetReg) const {
llvm_unreachable("Unimplemented");
}
MachineInstrBuilder SIInstrInfo::buildIndirectRead(
MachineBasicBlock *MBB,
MachineBasicBlock::iterator I,
unsigned ValueReg,
unsigned Address, unsigned OffsetReg) const {
llvm_unreachable("Unimplemented");
}
const TargetRegisterClass *SIInstrInfo::getSuperIndirectRegClass() const {
llvm_unreachable("Unimplemented");
}
|
