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
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
| //===-- AVRRegisterInfo.cpp - AVR Register Information --------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file contains the AVR implementation of the TargetRegisterInfo class.
//
//===----------------------------------------------------------------------===//
#include "AVRRegisterInfo.h"
#include "llvm/ADT/BitVector.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/IR/Function.h"
#include "llvm/CodeGen/TargetFrameLowering.h"
#include "AVR.h"
#include "AVRInstrInfo.h"
#include "AVRTargetMachine.h"
#include "MCTargetDesc/AVRMCTargetDesc.h"
#define GET_REGINFO_TARGET_DESC
#include "AVRGenRegisterInfo.inc"
namespace llvm {
AVRRegisterInfo::AVRRegisterInfo() : AVRGenRegisterInfo(0) {}
const uint16_t *
AVRRegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
CallingConv::ID CC = MF->getFunction().getCallingConv();
return ((CC == CallingConv::AVR_INTR || CC == CallingConv::AVR_SIGNAL)
? CSR_Interrupts_SaveList
: CSR_Normal_SaveList);
}
const uint32_t *
AVRRegisterInfo::getCallPreservedMask(const MachineFunction &MF,
CallingConv::ID CC) const {
return ((CC == CallingConv::AVR_INTR || CC == CallingConv::AVR_SIGNAL)
? CSR_Interrupts_RegMask
: CSR_Normal_RegMask);
}
BitVector AVRRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
BitVector Reserved(getNumRegs());
// Reserve the intermediate result registers r1 and r2
// The result of instructions like 'mul' is always stored here.
Reserved.set(AVR::R0);
Reserved.set(AVR::R1);
Reserved.set(AVR::R1R0);
// Reserve the stack pointer.
Reserved.set(AVR::SPL);
Reserved.set(AVR::SPH);
Reserved.set(AVR::SP);
// We tenatively reserve the frame pointer register r29:r28 because the
// function may require one, but we cannot tell until register allocation
// is complete, which can be too late.
//
// Instead we just unconditionally reserve the Y register.
//
// TODO: Write a pass to enumerate functions which reserved the Y register
// but didn't end up needing a frame pointer. In these, we can
// convert one or two of the spills inside to use the Y register.
Reserved.set(AVR::R28);
Reserved.set(AVR::R29);
Reserved.set(AVR::R29R28);
return Reserved;
}
const TargetRegisterClass *
AVRRegisterInfo::getLargestLegalSuperClass(const TargetRegisterClass *RC,
const MachineFunction &MF) const {
const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo();
if (TRI->isTypeLegalForClass(*RC, MVT::i16)) {
return &AVR::DREGSRegClass;
}
if (TRI->isTypeLegalForClass(*RC, MVT::i8)) {
return &AVR::GPR8RegClass;
}
llvm_unreachable("Invalid register size");
}
/// Fold a frame offset shared between two add instructions into a single one.
static void foldFrameOffset(MachineBasicBlock::iterator &II, int &Offset, unsigned DstReg) {
MachineInstr &MI = *II;
int Opcode = MI.getOpcode();
// Don't bother trying if the next instruction is not an add or a sub.
if ((Opcode != AVR::SUBIWRdK) && (Opcode != AVR::ADIWRdK)) {
return;
}
// Check that DstReg matches with next instruction, otherwise the instruction
// is not related to stack address manipulation.
if (DstReg != MI.getOperand(0).getReg()) {
return;
}
// Add the offset in the next instruction to our offset.
switch (Opcode) {
case AVR::SUBIWRdK:
Offset += -MI.getOperand(2).getImm();
break;
case AVR::ADIWRdK:
Offset += MI.getOperand(2).getImm();
break;
}
// Finally remove the instruction.
II++;
MI.eraseFromParent();
}
void AVRRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
int SPAdj, unsigned FIOperandNum,
RegScavenger *RS) const {
assert(SPAdj == 0 && "Unexpected SPAdj value");
MachineInstr &MI = *II;
DebugLoc dl = MI.getDebugLoc();
MachineBasicBlock &MBB = *MI.getParent();
const MachineFunction &MF = *MBB.getParent();
const AVRTargetMachine &TM = (const AVRTargetMachine &)MF.getTarget();
const TargetInstrInfo &TII = *TM.getSubtargetImpl()->getInstrInfo();
const MachineFrameInfo &MFI = MF.getFrameInfo();
const TargetFrameLowering *TFI = TM.getSubtargetImpl()->getFrameLowering();
int FrameIndex = MI.getOperand(FIOperandNum).getIndex();
int Offset = MFI.getObjectOffset(FrameIndex);
// Add one to the offset because SP points to an empty slot.
Offset += MFI.getStackSize() - TFI->getOffsetOfLocalArea() + 1;
// Fold incoming offset.
Offset += MI.getOperand(FIOperandNum + 1).getImm();
// This is actually "load effective address" of the stack slot
// instruction. We have only two-address instructions, thus we need to
// expand it into move + add.
if (MI.getOpcode() == AVR::FRMIDX) {
MI.setDesc(TII.get(AVR::MOVWRdRr));
MI.getOperand(FIOperandNum).ChangeToRegister(AVR::R29R28, false);
MI.RemoveOperand(2);
assert(Offset > 0 && "Invalid offset");
// We need to materialize the offset via an add instruction.
unsigned Opcode;
Register DstReg = MI.getOperand(0).getReg();
assert(DstReg != AVR::R29R28 && "Dest reg cannot be the frame pointer");
II++; // Skip over the FRMIDX (and now MOVW) instruction.
// Generally, to load a frame address two add instructions are emitted that
// could get folded into a single one:
// movw r31:r30, r29:r28
// adiw r31:r30, 29
// adiw r31:r30, 16
// to:
// movw r31:r30, r29:r28
// adiw r31:r30, 45
if (II != MBB.end())
foldFrameOffset(II, Offset, DstReg);
// Select the best opcode based on DstReg and the offset size.
switch (DstReg) {
case AVR::R25R24:
case AVR::R27R26:
case AVR::R31R30: {
if (isUInt<6>(Offset)) {
Opcode = AVR::ADIWRdK;
break;
}
LLVM_FALLTHROUGH;
}
default: {
// This opcode will get expanded into a pair of subi/sbci.
Opcode = AVR::SUBIWRdK;
Offset = -Offset;
break;
}
}
MachineInstr *New = BuildMI(MBB, II, dl, TII.get(Opcode), DstReg)
.addReg(DstReg, RegState::Kill)
.addImm(Offset);
New->getOperand(3).setIsDead();
return;
}
// If the offset is too big we have to adjust and restore the frame pointer
// to materialize a valid load/store with displacement.
//:TODO: consider using only one adiw/sbiw chain for more than one frame index
if (Offset > 62) {
unsigned AddOpc = AVR::ADIWRdK, SubOpc = AVR::SBIWRdK;
int AddOffset = Offset - 63 + 1;
// For huge offsets where adiw/sbiw cannot be used use a pair of subi/sbci.
if ((Offset - 63 + 1) > 63) {
AddOpc = AVR::SUBIWRdK;
SubOpc = AVR::SUBIWRdK;
AddOffset = -AddOffset;
}
// It is possible that the spiller places this frame instruction in between
// a compare and branch, invalidating the contents of SREG set by the
// compare instruction because of the add/sub pairs. Conservatively save and
// restore SREG before and after each add/sub pair.
BuildMI(MBB, II, dl, TII.get(AVR::INRdA), AVR::R0).addImm(0x3f);
MachineInstr *New = BuildMI(MBB, II, dl, TII.get(AddOpc), AVR::R29R28)
.addReg(AVR::R29R28, RegState::Kill)
.addImm(AddOffset);
New->getOperand(3).setIsDead();
// Restore SREG.
BuildMI(MBB, std::next(II), dl, TII.get(AVR::OUTARr))
.addImm(0x3f)
.addReg(AVR::R0, RegState::Kill);
// No need to set SREG as dead here otherwise if the next instruction is a
// cond branch it will be using a dead register.
BuildMI(MBB, std::next(II), dl, TII.get(SubOpc), AVR::R29R28)
.addReg(AVR::R29R28, RegState::Kill)
.addImm(Offset - 63 + 1);
Offset = 62;
}
MI.getOperand(FIOperandNum).ChangeToRegister(AVR::R29R28, false);
assert(isUInt<6>(Offset) && "Offset is out of range");
MI.getOperand(FIOperandNum + 1).ChangeToImmediate(Offset);
}
Register AVRRegisterInfo::getFrameRegister(const MachineFunction &MF) const {
const TargetFrameLowering *TFI = MF.getSubtarget().getFrameLowering();
if (TFI->hasFP(MF)) {
// The Y pointer register
return AVR::R28;
}
return AVR::SP;
}
const TargetRegisterClass *
AVRRegisterInfo::getPointerRegClass(const MachineFunction &MF,
unsigned Kind) const {
// FIXME: Currently we're using avr-gcc as reference, so we restrict
// ptrs to Y and Z regs. Though avr-gcc has buggy implementation
// of memory constraint, so we can fix it and bit avr-gcc here ;-)
return &AVR::PTRDISPREGSRegClass;
}
void AVRRegisterInfo::splitReg(unsigned Reg,
unsigned &LoReg,
unsigned &HiReg) const {
assert(AVR::DREGSRegClass.contains(Reg) && "can only split 16-bit registers");
LoReg = getSubReg(Reg, AVR::sub_lo);
HiReg = getSubReg(Reg, AVR::sub_hi);
}
bool AVRRegisterInfo::shouldCoalesce(MachineInstr *MI,
const TargetRegisterClass *SrcRC,
unsigned SubReg,
const TargetRegisterClass *DstRC,
unsigned DstSubReg,
const TargetRegisterClass *NewRC,
LiveIntervals &LIS) const {
if(this->getRegClass(AVR::PTRDISPREGSRegClassID)->hasSubClassEq(NewRC)) {
return false;
}
return TargetRegisterInfo::shouldCoalesce(MI, SrcRC, SubReg, DstRC, DstSubReg, NewRC, LIS);
}
} // end of namespace llvm
|