1 | //===--- SPIRVUtils.h ---- SPIR-V Utility Functions -------------*- C++ -*-===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | // |
9 | // This file contains miscellaneous utility functions. |
10 | // |
11 | //===----------------------------------------------------------------------===// |
12 | |
13 | #ifndef LLVM_LIB_TARGET_SPIRV_SPIRVUTILS_H |
14 | #define LLVM_LIB_TARGET_SPIRV_SPIRVUTILS_H |
15 | |
16 | #include "MCTargetDesc/SPIRVBaseInfo.h" |
17 | #include "llvm/Analysis/LoopInfo.h" |
18 | #include "llvm/CodeGen/MachineBasicBlock.h" |
19 | #include "llvm/IR/Dominators.h" |
20 | #include "llvm/IR/GlobalVariable.h" |
21 | #include "llvm/IR/IRBuilder.h" |
22 | #include "llvm/IR/TypedPointerType.h" |
23 | #include <queue> |
24 | #include <string> |
25 | #include <unordered_map> |
26 | #include <unordered_set> |
27 | |
28 | namespace llvm { |
29 | class MCInst; |
30 | class MachineFunction; |
31 | class MachineInstr; |
32 | class MachineInstrBuilder; |
33 | class MachineIRBuilder; |
34 | class MachineRegisterInfo; |
35 | class Register; |
36 | class StringRef; |
37 | class SPIRVInstrInfo; |
38 | class SPIRVSubtarget; |
39 | class SPIRVGlobalRegistry; |
40 | |
41 | // This class implements a partial ordering visitor, which visits a cyclic graph |
42 | // in natural topological-like ordering. Topological ordering is not defined for |
43 | // directed graphs with cycles, so this assumes cycles are a single node, and |
44 | // ignores back-edges. The cycle is visited from the entry in the same |
45 | // topological-like ordering. |
46 | // |
47 | // Note: this visitor REQUIRES a reducible graph. |
48 | // |
49 | // This means once we visit a node, we know all the possible ancestors have been |
50 | // visited. |
51 | // |
52 | // clang-format off |
53 | // |
54 | // Given this graph: |
55 | // |
56 | // ,-> B -\ |
57 | // A -+ +---> D ----> E -> F -> G -> H |
58 | // `-> C -/ ^ | |
59 | // +-----------------+ |
60 | // |
61 | // Visit order is: |
62 | // A, [B, C in any order], D, E, F, G, H |
63 | // |
64 | // clang-format on |
65 | // |
66 | // Changing the function CFG between the construction of the visitor and |
67 | // visiting is undefined. The visitor can be reused, but if the CFG is updated, |
68 | // the visitor must be rebuilt. |
69 | class PartialOrderingVisitor { |
70 | DomTreeBuilder::BBDomTree DT; |
71 | LoopInfo LI; |
72 | |
73 | std::unordered_set<BasicBlock *> Queued = {}; |
74 | std::queue<BasicBlock *> ToVisit = {}; |
75 | |
76 | struct OrderInfo { |
77 | size_t Rank; |
78 | size_t TraversalIndex; |
79 | }; |
80 | |
81 | using BlockToOrderInfoMap = std::unordered_map<BasicBlock *, OrderInfo>; |
82 | BlockToOrderInfoMap BlockToOrder; |
83 | std::vector<BasicBlock *> Order = {}; |
84 | |
85 | // Get all basic-blocks reachable from Start. |
86 | std::unordered_set<BasicBlock *> getReachableFrom(BasicBlock *Start); |
87 | |
88 | // Internal function used to determine the partial ordering. |
89 | // Visits |BB| with the current rank being |Rank|. |
90 | size_t visit(BasicBlock *BB, size_t Rank); |
91 | |
92 | bool CanBeVisited(BasicBlock *BB) const; |
93 | |
94 | public: |
95 | size_t GetNodeRank(BasicBlock *BB) const; |
96 | |
97 | // Build the visitor to operate on the function F. |
98 | PartialOrderingVisitor(Function &F); |
99 | |
100 | // Returns true is |LHS| comes before |RHS| in the partial ordering. |
101 | // If |LHS| and |RHS| have the same rank, the traversal order determines the |
102 | // order (order is stable). |
103 | bool compare(const BasicBlock *LHS, const BasicBlock *RHS) const; |
104 | |
105 | // Visit the function starting from the basic block |Start|, and calling |Op| |
106 | // on each visited BB. This traversal ignores back-edges, meaning this won't |
107 | // visit a node to which |Start| is not an ancestor. |
108 | // If Op returns |true|, the visitor continues. If |Op| returns false, the |
109 | // visitor will stop at that rank. This means if 2 nodes share the same rank, |
110 | // and Op returns false when visiting the first, the second will be visited |
111 | // afterwards. But none of their successors will. |
112 | void partialOrderVisit(BasicBlock &Start, |
113 | std::function<bool(BasicBlock *)> Op); |
114 | }; |
115 | |
116 | // Add the given string as a series of integer operand, inserting null |
117 | // terminators and padding to make sure the operands all have 32-bit |
118 | // little-endian words. |
119 | void addStringImm(const StringRef &Str, MCInst &Inst); |
120 | void addStringImm(const StringRef &Str, MachineInstrBuilder &MIB); |
121 | void addStringImm(const StringRef &Str, IRBuilder<> &B, |
122 | std::vector<Value *> &Args); |
123 | |
124 | // Read the series of integer operands back as a null-terminated string using |
125 | // the reverse of the logic in addStringImm. |
126 | std::string getStringImm(const MachineInstr &MI, unsigned StartIndex); |
127 | |
128 | // Returns the string constant that the register refers to. It is assumed that |
129 | // Reg is a global value that contains a string. |
130 | std::string getStringValueFromReg(Register Reg, MachineRegisterInfo &MRI); |
131 | |
132 | // Add the given numerical immediate to MIB. |
133 | void addNumImm(const APInt &Imm, MachineInstrBuilder &MIB); |
134 | |
135 | // Add an OpName instruction for the given target register. |
136 | void buildOpName(Register Target, const StringRef &Name, |
137 | MachineIRBuilder &MIRBuilder); |
138 | void buildOpName(Register Target, const StringRef &Name, MachineInstr &I, |
139 | const SPIRVInstrInfo &TII); |
140 | |
141 | // Add an OpDecorate instruction for the given Reg. |
142 | void buildOpDecorate(Register Reg, MachineIRBuilder &MIRBuilder, |
143 | SPIRV::Decoration::Decoration Dec, |
144 | const std::vector<uint32_t> &DecArgs, |
145 | StringRef StrImm = "" ); |
146 | void buildOpDecorate(Register Reg, MachineInstr &I, const SPIRVInstrInfo &TII, |
147 | SPIRV::Decoration::Decoration Dec, |
148 | const std::vector<uint32_t> &DecArgs, |
149 | StringRef StrImm = "" ); |
150 | |
151 | // Add an OpDecorate instruction for the given Reg. |
152 | void buildOpMemberDecorate(Register Reg, MachineIRBuilder &MIRBuilder, |
153 | SPIRV::Decoration::Decoration Dec, uint32_t Member, |
154 | const std::vector<uint32_t> &DecArgs, |
155 | StringRef StrImm = "" ); |
156 | void buildOpMemberDecorate(Register Reg, MachineInstr &I, |
157 | const SPIRVInstrInfo &TII, |
158 | SPIRV::Decoration::Decoration Dec, uint32_t Member, |
159 | const std::vector<uint32_t> &DecArgs, |
160 | StringRef StrImm = "" ); |
161 | |
162 | // Add an OpDecorate instruction by "spirv.Decorations" metadata node. |
163 | void buildOpSpirvDecorations(Register Reg, MachineIRBuilder &MIRBuilder, |
164 | const MDNode *GVarMD); |
165 | |
166 | // Return a valid position for the OpVariable instruction inside a function, |
167 | // i.e., at the beginning of the first block of the function. |
168 | MachineBasicBlock::iterator getOpVariableMBBIt(MachineInstr &I); |
169 | |
170 | // Return a valid position for the instruction at the end of the block before |
171 | // terminators and debug instructions. |
172 | MachineBasicBlock::iterator getInsertPtValidEnd(MachineBasicBlock *MBB); |
173 | |
174 | // Returns true if a pointer to the storage class can be casted to/from a |
175 | // pointer to the Generic storage class. |
176 | constexpr bool isGenericCastablePtr(SPIRV::StorageClass::StorageClass SC) { |
177 | switch (SC) { |
178 | case SPIRV::StorageClass::Workgroup: |
179 | case SPIRV::StorageClass::CrossWorkgroup: |
180 | case SPIRV::StorageClass::Function: |
181 | return true; |
182 | default: |
183 | return false; |
184 | } |
185 | } |
186 | |
187 | // Convert a SPIR-V storage class to the corresponding LLVM IR address space. |
188 | // TODO: maybe the following two functions should be handled in the subtarget |
189 | // to allow for different OpenCL vs Vulkan handling. |
190 | constexpr unsigned |
191 | storageClassToAddressSpace(SPIRV::StorageClass::StorageClass SC) { |
192 | switch (SC) { |
193 | case SPIRV::StorageClass::Function: |
194 | return 0; |
195 | case SPIRV::StorageClass::CrossWorkgroup: |
196 | return 1; |
197 | case SPIRV::StorageClass::UniformConstant: |
198 | return 2; |
199 | case SPIRV::StorageClass::Workgroup: |
200 | return 3; |
201 | case SPIRV::StorageClass::Generic: |
202 | return 4; |
203 | case SPIRV::StorageClass::DeviceOnlyINTEL: |
204 | return 5; |
205 | case SPIRV::StorageClass::HostOnlyINTEL: |
206 | return 6; |
207 | case SPIRV::StorageClass::Input: |
208 | return 7; |
209 | case SPIRV::StorageClass::Output: |
210 | return 8; |
211 | case SPIRV::StorageClass::CodeSectionINTEL: |
212 | return 9; |
213 | case SPIRV::StorageClass::Private: |
214 | return 10; |
215 | case SPIRV::StorageClass::StorageBuffer: |
216 | return 11; |
217 | case SPIRV::StorageClass::Uniform: |
218 | return 12; |
219 | default: |
220 | report_fatal_error(reason: "Unable to get address space id" ); |
221 | } |
222 | } |
223 | |
224 | // Convert an LLVM IR address space to a SPIR-V storage class. |
225 | SPIRV::StorageClass::StorageClass |
226 | addressSpaceToStorageClass(unsigned AddrSpace, const SPIRVSubtarget &STI); |
227 | |
228 | SPIRV::MemorySemantics::MemorySemantics |
229 | getMemSemanticsForStorageClass(SPIRV::StorageClass::StorageClass SC); |
230 | |
231 | SPIRV::MemorySemantics::MemorySemantics getMemSemantics(AtomicOrdering Ord); |
232 | |
233 | SPIRV::Scope::Scope getMemScope(LLVMContext &Ctx, SyncScope::ID Id); |
234 | |
235 | // Find def instruction for the given ConstReg, walking through |
236 | // spv_track_constant and ASSIGN_TYPE instructions. Updates ConstReg by def |
237 | // of OpConstant instruction. |
238 | MachineInstr *getDefInstrMaybeConstant(Register &ConstReg, |
239 | const MachineRegisterInfo *MRI); |
240 | |
241 | // Get constant integer value of the given ConstReg. |
242 | uint64_t getIConstVal(Register ConstReg, const MachineRegisterInfo *MRI); |
243 | |
244 | // Check if MI is a SPIR-V specific intrinsic call. |
245 | bool isSpvIntrinsic(const MachineInstr &MI, Intrinsic::ID IntrinsicID); |
246 | // Check if it's a SPIR-V specific intrinsic call. |
247 | bool isSpvIntrinsic(const Value *Arg); |
248 | |
249 | // Get type of i-th operand of the metadata node. |
250 | Type *getMDOperandAsType(const MDNode *N, unsigned I); |
251 | |
252 | // If OpenCL or SPIR-V builtin function name is recognized, return a demangled |
253 | // name, otherwise return an empty string. |
254 | std::string getOclOrSpirvBuiltinDemangledName(StringRef Name); |
255 | |
256 | // Check if a string contains a builtin prefix. |
257 | bool hasBuiltinTypePrefix(StringRef Name); |
258 | |
259 | // Check if given LLVM type is a special opaque builtin type. |
260 | bool isSpecialOpaqueType(const Type *Ty); |
261 | |
262 | // Check if the function is an SPIR-V entry point |
263 | bool isEntryPoint(const Function &F); |
264 | |
265 | // Parse basic scalar type name, substring TypeName, and return LLVM type. |
266 | Type *parseBasicTypeName(StringRef &TypeName, LLVMContext &Ctx); |
267 | |
268 | // Sort blocks in a partial ordering, so each block is after all its |
269 | // dominators. This should match both the SPIR-V and the MIR requirements. |
270 | // Returns true if the function was changed. |
271 | bool sortBlocks(Function &F); |
272 | |
273 | inline bool hasInitializer(const GlobalVariable *GV) { |
274 | return GV->hasInitializer() && !isa<UndefValue>(Val: GV->getInitializer()); |
275 | } |
276 | |
277 | // True if this is an instance of TypedPointerType. |
278 | inline bool isTypedPointerTy(const Type *T) { |
279 | return T && T->getTypeID() == Type::TypedPointerTyID; |
280 | } |
281 | |
282 | // True if this is an instance of PointerType. |
283 | inline bool isUntypedPointerTy(const Type *T) { |
284 | return T && T->getTypeID() == Type::PointerTyID; |
285 | } |
286 | |
287 | // True if this is an instance of PointerType or TypedPointerType. |
288 | inline bool isPointerTy(const Type *T) { |
289 | return isUntypedPointerTy(T) || isTypedPointerTy(T); |
290 | } |
291 | |
292 | // Get the address space of this pointer or pointer vector type for instances of |
293 | // PointerType or TypedPointerType. |
294 | inline unsigned getPointerAddressSpace(const Type *T) { |
295 | Type *SubT = T->getScalarType(); |
296 | return SubT->getTypeID() == Type::PointerTyID |
297 | ? cast<PointerType>(Val: SubT)->getAddressSpace() |
298 | : cast<TypedPointerType>(Val: SubT)->getAddressSpace(); |
299 | } |
300 | |
301 | // Return true if the Argument is decorated with a pointee type |
302 | inline bool hasPointeeTypeAttr(Argument *Arg) { |
303 | return Arg->hasByValAttr() || Arg->hasByRefAttr() || Arg->hasStructRetAttr(); |
304 | } |
305 | |
306 | // Return the pointee type of the argument or nullptr otherwise |
307 | inline Type *getPointeeTypeByAttr(Argument *Arg) { |
308 | if (Arg->hasByValAttr()) |
309 | return Arg->getParamByValType(); |
310 | if (Arg->hasStructRetAttr()) |
311 | return Arg->getParamStructRetType(); |
312 | if (Arg->hasByRefAttr()) |
313 | return Arg->getParamByRefType(); |
314 | return nullptr; |
315 | } |
316 | |
317 | inline Type *reconstructFunctionType(Function *F) { |
318 | SmallVector<Type *> ArgTys; |
319 | for (unsigned i = 0; i < F->arg_size(); ++i) |
320 | ArgTys.push_back(Elt: F->getArg(i)->getType()); |
321 | return FunctionType::get(Result: F->getReturnType(), Params: ArgTys, isVarArg: F->isVarArg()); |
322 | } |
323 | |
324 | #define TYPED_PTR_TARGET_EXT_NAME "spirv.$TypedPointerType" |
325 | inline Type *getTypedPointerWrapper(Type *ElemTy, unsigned AS) { |
326 | return TargetExtType::get(Context&: ElemTy->getContext(), TYPED_PTR_TARGET_EXT_NAME, |
327 | Types: {ElemTy}, Ints: {AS}); |
328 | } |
329 | |
330 | inline bool isTypedPointerWrapper(const TargetExtType *ExtTy) { |
331 | return ExtTy->getName() == TYPED_PTR_TARGET_EXT_NAME && |
332 | ExtTy->getNumIntParameters() == 1 && |
333 | ExtTy->getNumTypeParameters() == 1; |
334 | } |
335 | |
336 | // True if this is an instance of PointerType or TypedPointerType. |
337 | inline bool isPointerTyOrWrapper(const Type *Ty) { |
338 | if (auto *ExtTy = dyn_cast<TargetExtType>(Val: Ty)) |
339 | return isTypedPointerWrapper(ExtTy); |
340 | return isPointerTy(T: Ty); |
341 | } |
342 | |
343 | inline Type *applyWrappers(Type *Ty) { |
344 | if (auto *ExtTy = dyn_cast<TargetExtType>(Val: Ty)) { |
345 | if (isTypedPointerWrapper(ExtTy)) |
346 | return TypedPointerType::get(ElementType: applyWrappers(Ty: ExtTy->getTypeParameter(i: 0)), |
347 | AddressSpace: ExtTy->getIntParameter(i: 0)); |
348 | } else if (auto *VecTy = dyn_cast<VectorType>(Val: Ty)) { |
349 | Type *ElemTy = VecTy->getElementType(); |
350 | Type *NewElemTy = ElemTy->isTargetExtTy() ? applyWrappers(Ty: ElemTy) : ElemTy; |
351 | if (NewElemTy != ElemTy) |
352 | return VectorType::get(ElementType: NewElemTy, EC: VecTy->getElementCount()); |
353 | } |
354 | return Ty; |
355 | } |
356 | |
357 | inline Type *getPointeeType(const Type *Ty) { |
358 | if (Ty) { |
359 | if (auto PType = dyn_cast<TypedPointerType>(Val: Ty)) |
360 | return PType->getElementType(); |
361 | else if (auto *ExtTy = dyn_cast<TargetExtType>(Val: Ty)) |
362 | if (isTypedPointerWrapper(ExtTy)) |
363 | return ExtTy->getTypeParameter(i: 0); |
364 | } |
365 | return nullptr; |
366 | } |
367 | |
368 | inline bool isUntypedEquivalentToTyExt(Type *Ty1, Type *Ty2) { |
369 | if (!isUntypedPointerTy(T: Ty1) || !Ty2) |
370 | return false; |
371 | if (auto *ExtTy = dyn_cast<TargetExtType>(Val: Ty2)) |
372 | if (isTypedPointerWrapper(ExtTy) && |
373 | ExtTy->getTypeParameter(i: 0) == |
374 | IntegerType::getInt8Ty(C&: Ty1->getContext()) && |
375 | ExtTy->getIntParameter(i: 0) == cast<PointerType>(Val: Ty1)->getAddressSpace()) |
376 | return true; |
377 | return false; |
378 | } |
379 | |
380 | inline bool isEquivalentTypes(Type *Ty1, Type *Ty2) { |
381 | return isUntypedEquivalentToTyExt(Ty1, Ty2) || |
382 | isUntypedEquivalentToTyExt(Ty1: Ty2, Ty2: Ty1); |
383 | } |
384 | |
385 | inline Type *toTypedPointer(Type *Ty) { |
386 | if (Type *NewTy = applyWrappers(Ty); NewTy != Ty) |
387 | return NewTy; |
388 | return isUntypedPointerTy(T: Ty) |
389 | ? TypedPointerType::get(ElementType: IntegerType::getInt8Ty(C&: Ty->getContext()), |
390 | AddressSpace: getPointerAddressSpace(T: Ty)) |
391 | : Ty; |
392 | } |
393 | |
394 | inline Type *toTypedFunPointer(FunctionType *FTy) { |
395 | Type *OrigRetTy = FTy->getReturnType(); |
396 | Type *RetTy = toTypedPointer(Ty: OrigRetTy); |
397 | bool IsUntypedPtr = false; |
398 | for (Type *PTy : FTy->params()) { |
399 | if (isUntypedPointerTy(T: PTy)) { |
400 | IsUntypedPtr = true; |
401 | break; |
402 | } |
403 | } |
404 | if (!IsUntypedPtr && RetTy == OrigRetTy) |
405 | return FTy; |
406 | SmallVector<Type *> ParamTys; |
407 | for (Type *PTy : FTy->params()) |
408 | ParamTys.push_back(Elt: toTypedPointer(Ty: PTy)); |
409 | return FunctionType::get(Result: RetTy, Params: ParamTys, isVarArg: FTy->isVarArg()); |
410 | } |
411 | |
412 | inline const Type *unifyPtrType(const Type *Ty) { |
413 | if (auto FTy = dyn_cast<FunctionType>(Val: Ty)) |
414 | return toTypedFunPointer(FTy: const_cast<FunctionType *>(FTy)); |
415 | return toTypedPointer(Ty: const_cast<Type *>(Ty)); |
416 | } |
417 | |
418 | inline bool isVector1(Type *Ty) { |
419 | auto *FVTy = dyn_cast<FixedVectorType>(Val: Ty); |
420 | return FVTy && FVTy->getNumElements() == 1; |
421 | } |
422 | |
423 | // Modify an LLVM type to conform with future transformations in IRTranslator. |
424 | // At the moment use cases comprise only a <1 x Type> vector. To extend when/if |
425 | // needed. |
426 | inline Type *normalizeType(Type *Ty) { |
427 | auto *FVTy = dyn_cast<FixedVectorType>(Val: Ty); |
428 | if (!FVTy || FVTy->getNumElements() != 1) |
429 | return Ty; |
430 | // If it's a <1 x Type> vector type, replace it by the element type, because |
431 | // it's not a legal vector type in LLT and IRTranslator will represent it as |
432 | // the scalar eventually. |
433 | return normalizeType(Ty: FVTy->getElementType()); |
434 | } |
435 | |
436 | inline PoisonValue *getNormalizedPoisonValue(Type *Ty) { |
437 | return PoisonValue::get(T: normalizeType(Ty)); |
438 | } |
439 | |
440 | inline MetadataAsValue *buildMD(Value *Arg) { |
441 | LLVMContext &Ctx = Arg->getContext(); |
442 | return MetadataAsValue::get( |
443 | Context&: Ctx, MD: MDNode::get(Context&: Ctx, MDs: ValueAsMetadata::getConstant(C: Arg))); |
444 | } |
445 | |
446 | CallInst *buildIntrWithMD(Intrinsic::ID IntrID, ArrayRef<Type *> Types, |
447 | Value *Arg, Value *Arg2, ArrayRef<Constant *> Imms, |
448 | IRBuilder<> &B); |
449 | |
450 | MachineInstr *getVRegDef(MachineRegisterInfo &MRI, Register Reg); |
451 | |
452 | #define SPIRV_BACKEND_SERVICE_FUN_NAME "__spirv_backend_service_fun" |
453 | bool getVacantFunctionName(Module &M, std::string &Name); |
454 | |
455 | void setRegClassType(Register Reg, const Type *Ty, SPIRVGlobalRegistry *GR, |
456 | MachineIRBuilder &MIRBuilder, |
457 | SPIRV::AccessQualifier::AccessQualifier AccessQual, |
458 | bool EmitIR, bool Force = false); |
459 | void setRegClassType(Register Reg, const MachineInstr *SpvType, |
460 | SPIRVGlobalRegistry *GR, MachineRegisterInfo *MRI, |
461 | const MachineFunction &MF, bool Force = false); |
462 | Register createVirtualRegister(const MachineInstr *SpvType, |
463 | SPIRVGlobalRegistry *GR, |
464 | MachineRegisterInfo *MRI, |
465 | const MachineFunction &MF); |
466 | Register createVirtualRegister(const MachineInstr *SpvType, |
467 | SPIRVGlobalRegistry *GR, |
468 | MachineIRBuilder &MIRBuilder); |
469 | Register createVirtualRegister( |
470 | const Type *Ty, SPIRVGlobalRegistry *GR, MachineIRBuilder &MIRBuilder, |
471 | SPIRV::AccessQualifier::AccessQualifier AccessQual, bool EmitIR); |
472 | |
473 | // Return true if there is an opaque pointer type nested in the argument. |
474 | bool isNestedPointer(const Type *Ty); |
475 | |
476 | enum FPDecorationId { NONE, RTE, RTZ, RTP, RTN, SAT }; |
477 | |
478 | inline FPDecorationId demangledPostfixToDecorationId(const std::string &S) { |
479 | static std::unordered_map<std::string, FPDecorationId> Mapping = { |
480 | {"rte" , FPDecorationId::RTE}, |
481 | {"rtz" , FPDecorationId::RTZ}, |
482 | {"rtp" , FPDecorationId::RTP}, |
483 | {"rtn" , FPDecorationId::RTN}, |
484 | {"sat" , FPDecorationId::SAT}}; |
485 | auto It = Mapping.find(x: S); |
486 | return It == Mapping.end() ? FPDecorationId::NONE : It->second; |
487 | } |
488 | |
489 | SmallVector<MachineInstr *, 4> |
490 | createContinuedInstructions(MachineIRBuilder &MIRBuilder, unsigned Opcode, |
491 | unsigned MinWC, unsigned ContinuedOpcode, |
492 | ArrayRef<Register> Args, Register ReturnRegister, |
493 | Register TypeID); |
494 | |
495 | // Instruction selection directed by type folding. |
496 | const std::set<unsigned> &getTypeFoldingSupportedOpcodes(); |
497 | bool isTypeFoldingSupported(unsigned Opcode); |
498 | |
499 | // Get loop controls from llvm.loop. metadata. |
500 | SmallVector<unsigned, 1> getSpirvLoopControlOperandsFromLoopMetadata(Loop *L); |
501 | |
502 | // Traversing [g]MIR accounting for pseudo-instructions. |
503 | MachineInstr *passCopy(MachineInstr *Def, const MachineRegisterInfo *MRI); |
504 | MachineInstr *getDef(const MachineOperand &MO, const MachineRegisterInfo *MRI); |
505 | MachineInstr *getImm(const MachineOperand &MO, const MachineRegisterInfo *MRI); |
506 | int64_t foldImm(const MachineOperand &MO, const MachineRegisterInfo *MRI); |
507 | unsigned getArrayComponentCount(const MachineRegisterInfo *MRI, |
508 | const MachineInstr *ResType); |
509 | |
510 | } // namespace llvm |
511 | #endif // LLVM_LIB_TARGET_SPIRV_SPIRVUTILS_H |
512 | |