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
|
//===-- AMDGPUISelDAGToDAG.h - A dag to dag inst selector for AMDGPU ----===//
//
// 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
//
//==-----------------------------------------------------------------------===//
//
/// \file
/// Defines an instruction selector for the AMDGPU target.
//
//===----------------------------------------------------------------------===//
#ifndef LLVM_LIB_TARGET_AMDGPU_AMDGPUISELDAGTODAG_H
#define LLVM_LIB_TARGET_AMDGPU_AMDGPUISELDAGTODAG_H
#include "GCNSubtarget.h"
#include "SIMachineFunctionInfo.h"
#include "SIModeRegisterDefaults.h"
#include "llvm/CodeGen/SelectionDAGISel.h"
#include "llvm/Target/TargetMachine.h"
using namespace llvm;
namespace {
static inline bool isNullConstantOrUndef(SDValue V) {
return V.isUndef() || isNullConstant(V);
}
static inline bool getConstantValue(SDValue N, uint32_t &Out) {
// This is only used for packed vectors, where using 0 for undef should
// always be good.
if (N.isUndef()) {
Out = 0;
return true;
}
if (const ConstantSDNode *C = dyn_cast<ConstantSDNode>(N)) {
Out = C->getAPIntValue().getSExtValue();
return true;
}
if (const ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(N)) {
Out = C->getValueAPF().bitcastToAPInt().getSExtValue();
return true;
}
return false;
}
// TODO: Handle undef as zero
static inline SDNode *packConstantV2I16(const SDNode *N, SelectionDAG &DAG,
bool Negate = false) {
assert(N->getOpcode() == ISD::BUILD_VECTOR && N->getNumOperands() == 2);
uint32_t LHSVal, RHSVal;
if (getConstantValue(N->getOperand(0), LHSVal) &&
getConstantValue(N->getOperand(1), RHSVal)) {
SDLoc SL(N);
uint32_t K = Negate ? (-LHSVal & 0xffff) | (-RHSVal << 16)
: (LHSVal & 0xffff) | (RHSVal << 16);
return DAG.getMachineNode(AMDGPU::S_MOV_B32, SL, N->getValueType(0),
DAG.getTargetConstant(K, SL, MVT::i32));
}
return nullptr;
}
static inline SDNode *packNegConstantV2I16(const SDNode *N, SelectionDAG &DAG) {
return packConstantV2I16(N, DAG, true);
}
} // namespace
/// AMDGPU specific code to select AMDGPU machine instructions for
/// SelectionDAG operations.
class AMDGPUDAGToDAGISel : public SelectionDAGISel {
// Subtarget - Keep a pointer to the AMDGPU Subtarget around so that we can
// make the right decision when generating code for different targets.
const GCNSubtarget *Subtarget;
// Default FP mode for the current function.
SIModeRegisterDefaults Mode;
bool EnableLateStructurizeCFG;
// Instructions that will be lowered with a final instruction that zeros the
// high result bits.
bool fp16SrcZerosHighBits(unsigned Opc) const;
public:
static char ID;
AMDGPUDAGToDAGISel() = delete;
explicit AMDGPUDAGToDAGISel(TargetMachine &TM, CodeGenOpt::Level OptLevel);
~AMDGPUDAGToDAGISel() override = default;
void getAnalysisUsage(AnalysisUsage &AU) const override;
bool matchLoadD16FromBuildVector(SDNode *N) const;
bool runOnMachineFunction(MachineFunction &MF) override;
void PreprocessISelDAG() override;
void Select(SDNode *N) override;
StringRef getPassName() const override;
void PostprocessISelDAG() override;
protected:
void SelectBuildVector(SDNode *N, unsigned RegClassID);
private:
std::pair<SDValue, SDValue> foldFrameIndex(SDValue N) const;
bool isInlineImmediate(const SDNode *N, bool Negated = false) const;
bool isNegInlineImmediate(const SDNode *N) const {
return isInlineImmediate(N, true);
}
bool isInlineImmediate16(int64_t Imm) const {
return AMDGPU::isInlinableLiteral16(Imm, Subtarget->hasInv2PiInlineImm());
}
bool isInlineImmediate32(int64_t Imm) const {
return AMDGPU::isInlinableLiteral32(Imm, Subtarget->hasInv2PiInlineImm());
}
bool isInlineImmediate64(int64_t Imm) const {
return AMDGPU::isInlinableLiteral64(Imm, Subtarget->hasInv2PiInlineImm());
}
bool isInlineImmediate(const APFloat &Imm) const {
return Subtarget->getInstrInfo()->isInlineConstant(Imm);
}
bool isVGPRImm(const SDNode *N) const;
bool isUniformLoad(const SDNode *N) const;
bool isUniformBr(const SDNode *N) const;
// Returns true if ISD::AND SDNode `N`'s masking of the shift amount operand's
// `ShAmtBits` bits is unneeded.
bool isUnneededShiftMask(const SDNode *N, unsigned ShAmtBits) const;
bool isBaseWithConstantOffset64(SDValue Addr, SDValue &LHS,
SDValue &RHS) const;
MachineSDNode *buildSMovImm64(SDLoc &DL, uint64_t Val, EVT VT) const;
SDNode *glueCopyToOp(SDNode *N, SDValue NewChain, SDValue Glue) const;
SDNode *glueCopyToM0(SDNode *N, SDValue Val) const;
SDNode *glueCopyToM0LDSInit(SDNode *N) const;
const TargetRegisterClass *getOperandRegClass(SDNode *N, unsigned OpNo) const;
virtual bool SelectADDRVTX_READ(SDValue Addr, SDValue &Base, SDValue &Offset);
virtual bool SelectADDRIndirect(SDValue Addr, SDValue &Base, SDValue &Offset);
bool isDSOffsetLegal(SDValue Base, unsigned Offset) const;
bool isDSOffset2Legal(SDValue Base, unsigned Offset0, unsigned Offset1,
unsigned Size) const;
bool isFlatScratchBaseLegal(
SDValue Base, uint64_t FlatVariant = SIInstrFlags::FlatScratch) const;
bool SelectDS1Addr1Offset(SDValue Ptr, SDValue &Base, SDValue &Offset) const;
bool SelectDS64Bit4ByteAligned(SDValue Ptr, SDValue &Base, SDValue &Offset0,
SDValue &Offset1) const;
bool SelectDS128Bit8ByteAligned(SDValue Ptr, SDValue &Base, SDValue &Offset0,
SDValue &Offset1) const;
bool SelectDSReadWrite2(SDValue Ptr, SDValue &Base, SDValue &Offset0,
SDValue &Offset1, unsigned Size) const;
bool SelectMUBUF(SDValue Addr, SDValue &SRsrc, SDValue &VAddr,
SDValue &SOffset, SDValue &Offset, SDValue &Offen,
SDValue &Idxen, SDValue &Addr64) const;
bool SelectMUBUFAddr64(SDValue Addr, SDValue &SRsrc, SDValue &VAddr,
SDValue &SOffset, SDValue &Offset) const;
bool SelectMUBUFScratchOffen(SDNode *Parent, SDValue Addr, SDValue &RSrc,
SDValue &VAddr, SDValue &SOffset,
SDValue &ImmOffset) const;
bool SelectMUBUFScratchOffset(SDNode *Parent, SDValue Addr, SDValue &SRsrc,
SDValue &Soffset, SDValue &Offset) const;
bool SelectMUBUFOffset(SDValue Addr, SDValue &SRsrc, SDValue &Soffset,
SDValue &Offset) const;
bool SelectFlatOffsetImpl(SDNode *N, SDValue Addr, SDValue &VAddr,
SDValue &Offset, uint64_t FlatVariant) const;
bool SelectFlatOffset(SDNode *N, SDValue Addr, SDValue &VAddr,
SDValue &Offset) const;
bool SelectGlobalOffset(SDNode *N, SDValue Addr, SDValue &VAddr,
SDValue &Offset) const;
bool SelectScratchOffset(SDNode *N, SDValue Addr, SDValue &VAddr,
SDValue &Offset) const;
bool SelectGlobalSAddr(SDNode *N, SDValue Addr, SDValue &SAddr,
SDValue &VOffset, SDValue &Offset) const;
bool SelectScratchSAddr(SDNode *N, SDValue Addr, SDValue &SAddr,
SDValue &Offset) const;
bool checkFlatScratchSVSSwizzleBug(SDValue VAddr, SDValue SAddr,
uint64_t ImmOffset) const;
bool SelectScratchSVAddr(SDNode *N, SDValue Addr, SDValue &VAddr,
SDValue &SAddr, SDValue &Offset) const;
bool SelectSMRDOffset(SDValue ByteOffsetNode, SDValue *SOffset,
SDValue *Offset, bool Imm32Only = false,
bool IsBuffer = false) const;
SDValue Expand32BitAddress(SDValue Addr) const;
bool SelectSMRDBaseOffset(SDValue Addr, SDValue &SBase, SDValue *SOffset,
SDValue *Offset, bool Imm32Only = false,
bool IsBuffer = false) const;
bool SelectSMRD(SDValue Addr, SDValue &SBase, SDValue *SOffset,
SDValue *Offset, bool Imm32Only = false) const;
bool SelectSMRDImm(SDValue Addr, SDValue &SBase, SDValue &Offset) const;
bool SelectSMRDImm32(SDValue Addr, SDValue &SBase, SDValue &Offset) const;
bool SelectSMRDSgpr(SDValue Addr, SDValue &SBase, SDValue &SOffset) const;
bool SelectSMRDSgprImm(SDValue Addr, SDValue &SBase, SDValue &SOffset,
SDValue &Offset) const;
bool SelectSMRDBufferImm(SDValue N, SDValue &Offset) const;
bool SelectSMRDBufferImm32(SDValue N, SDValue &Offset) const;
bool SelectSMRDBufferSgprImm(SDValue N, SDValue &SOffset,
SDValue &Offset) const;
bool SelectMOVRELOffset(SDValue Index, SDValue &Base, SDValue &Offset) const;
bool SelectVOP3ModsImpl(SDValue In, SDValue &Src, unsigned &SrcMods,
bool IsCanonicalizing = true,
bool AllowAbs = true) const;
bool SelectVOP3Mods(SDValue In, SDValue &Src, SDValue &SrcMods) const;
bool SelectVOP3ModsNonCanonicalizing(SDValue In, SDValue &Src,
SDValue &SrcMods) const;
bool SelectVOP3BMods(SDValue In, SDValue &Src, SDValue &SrcMods) const;
bool SelectVOP3NoMods(SDValue In, SDValue &Src) const;
bool SelectVOP3Mods0(SDValue In, SDValue &Src, SDValue &SrcMods,
SDValue &Clamp, SDValue &Omod) const;
bool SelectVOP3BMods0(SDValue In, SDValue &Src, SDValue &SrcMods,
SDValue &Clamp, SDValue &Omod) const;
bool SelectVOP3NoMods0(SDValue In, SDValue &Src, SDValue &SrcMods,
SDValue &Clamp, SDValue &Omod) const;
bool SelectVINTERPModsImpl(SDValue In, SDValue &Src, SDValue &SrcMods,
bool OpSel) const;
bool SelectVINTERPMods(SDValue In, SDValue &Src, SDValue &SrcMods) const;
bool SelectVINTERPModsHi(SDValue In, SDValue &Src, SDValue &SrcMods) const;
bool SelectVOP3OMods(SDValue In, SDValue &Src, SDValue &Clamp,
SDValue &Omod) const;
bool SelectVOP3PMods(SDValue In, SDValue &Src, SDValue &SrcMods,
bool IsDOT = false) const;
bool SelectVOP3PModsDOT(SDValue In, SDValue &Src, SDValue &SrcMods) const;
bool SelectDotIUVOP3PMods(SDValue In, SDValue &Src) const;
bool SelectWMMAOpSelVOP3PMods(SDValue In, SDValue &Src) const;
bool SelectVOP3OpSel(SDValue In, SDValue &Src, SDValue &SrcMods) const;
bool SelectVOP3OpSelMods(SDValue In, SDValue &Src, SDValue &SrcMods) const;
bool SelectVOP3PMadMixModsImpl(SDValue In, SDValue &Src,
unsigned &Mods) const;
bool SelectVOP3PMadMixModsExt(SDValue In, SDValue &Src,
SDValue &SrcMods) const;
bool SelectVOP3PMadMixMods(SDValue In, SDValue &Src, SDValue &SrcMods) const;
SDValue getHi16Elt(SDValue In) const;
SDValue getMaterializedScalarImm32(int64_t Val, const SDLoc &DL) const;
void SelectADD_SUB_I64(SDNode *N);
void SelectAddcSubb(SDNode *N);
void SelectUADDO_USUBO(SDNode *N);
void SelectDIV_SCALE(SDNode *N);
void SelectMAD_64_32(SDNode *N);
void SelectMUL_LOHI(SDNode *N);
void SelectFMA_W_CHAIN(SDNode *N);
void SelectFMUL_W_CHAIN(SDNode *N);
SDNode *getBFE32(bool IsSigned, const SDLoc &DL, SDValue Val, uint32_t Offset,
uint32_t Width);
void SelectS_BFEFromShifts(SDNode *N);
void SelectS_BFE(SDNode *N);
bool isCBranchSCC(const SDNode *N) const;
void SelectBRCOND(SDNode *N);
void SelectFMAD_FMA(SDNode *N);
void SelectDSAppendConsume(SDNode *N, unsigned IntrID);
void SelectDSBvhStackIntrinsic(SDNode *N);
void SelectDS_GWS(SDNode *N, unsigned IntrID);
void SelectInterpP1F16(SDNode *N);
void SelectINTRINSIC_W_CHAIN(SDNode *N);
void SelectINTRINSIC_WO_CHAIN(SDNode *N);
void SelectINTRINSIC_VOID(SDNode *N);
protected:
// Include the pieces autogenerated from the target description.
#include "AMDGPUGenDAGISel.inc"
};
#endif // LLVM_LIB_TARGET_AMDGPU_AMDGPUISELDAGTODAG_H
|
