1 //===--- SPIRVCallLowering.cpp - Call lowering ------------------*- 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 implements the lowering of LLVM calls to machine code calls for
10 // GlobalISel.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "SPIRVCallLowering.h"
15 #include "MCTargetDesc/SPIRVBaseInfo.h"
16 #include "SPIRV.h"
17 #include "SPIRVBuiltins.h"
18 #include "SPIRVGlobalRegistry.h"
19 #include "SPIRVISelLowering.h"
20 #include "SPIRVRegisterInfo.h"
21 #include "SPIRVSubtarget.h"
22 #include "SPIRVUtils.h"
23 #include "llvm/CodeGen/FunctionLoweringInfo.h"
24 #include "llvm/Support/ModRef.h"
25 
26 using namespace llvm;
27 
28 SPIRVCallLowering::SPIRVCallLowering(const SPIRVTargetLowering &TLI,
29                                      SPIRVGlobalRegistry *GR)
30     : CallLowering(&TLI), GR(GR) {}
31 
32 bool SPIRVCallLowering::lowerReturn(MachineIRBuilder &MIRBuilder,
33                                     const Value *Val, ArrayRef<Register> VRegs,
34                                     FunctionLoweringInfo &FLI,
35                                     Register SwiftErrorVReg) const {
36   // Currently all return types should use a single register.
37   // TODO: handle the case of multiple registers.
38   if (VRegs.size() > 1)
39     return false;
40   if (Val) {
41     const auto &STI = MIRBuilder.getMF().getSubtarget();
42     return MIRBuilder.buildInstr(SPIRV::OpReturnValue)
43         .addUse(VRegs[0])
44         .constrainAllUses(MIRBuilder.getTII(), *STI.getRegisterInfo(),
45                           *STI.getRegBankInfo());
46   }
47   MIRBuilder.buildInstr(SPIRV::OpReturn);
48   return true;
49 }
50 
51 // Based on the LLVM function attributes, get a SPIR-V FunctionControl.
52 static uint32_t getFunctionControl(const Function &F) {
53   MemoryEffects MemEffects = F.getMemoryEffects();
54 
55   uint32_t FuncControl = static_cast<uint32_t>(SPIRV::FunctionControl::None);
56 
57   if (F.hasFnAttribute(Attribute::AttrKind::NoInline))
58     FuncControl |= static_cast<uint32_t>(SPIRV::FunctionControl::DontInline);
59   else if (F.hasFnAttribute(Attribute::AttrKind::AlwaysInline))
60     FuncControl |= static_cast<uint32_t>(SPIRV::FunctionControl::Inline);
61 
62   if (MemEffects.doesNotAccessMemory())
63     FuncControl |= static_cast<uint32_t>(SPIRV::FunctionControl::Pure);
64   else if (MemEffects.onlyReadsMemory())
65     FuncControl |= static_cast<uint32_t>(SPIRV::FunctionControl::Const);
66 
67   return FuncControl;
68 }
69 
70 static ConstantInt *getConstInt(MDNode *MD, unsigned NumOp) {
71   if (MD->getNumOperands() > NumOp) {
72     auto *CMeta = dyn_cast<ConstantAsMetadata>(MD->getOperand(NumOp));
73     if (CMeta)
74       return dyn_cast<ConstantInt>(CMeta->getValue());
75   }
76   return nullptr;
77 }
78 
79 // This code restores function args/retvalue types for composite cases
80 // because the final types should still be aggregate whereas they're i32
81 // during the translation to cope with aggregate flattening etc.
82 static FunctionType *getOriginalFunctionType(const Function &F) {
83   auto *NamedMD = F.getParent()->getNamedMetadata("spv.cloned_funcs");
84   if (NamedMD == nullptr)
85     return F.getFunctionType();
86 
87   Type *RetTy = F.getFunctionType()->getReturnType();
88   SmallVector<Type *, 4> ArgTypes;
89   for (auto &Arg : F.args())
90     ArgTypes.push_back(Arg.getType());
91 
92   auto ThisFuncMDIt =
93       std::find_if(NamedMD->op_begin(), NamedMD->op_end(), [&F](MDNode *N) {
94         return isa<MDString>(N->getOperand(0)) &&
95                cast<MDString>(N->getOperand(0))->getString() == F.getName();
96       });
97   // TODO: probably one function can have numerous type mutations,
98   // so we should support this.
99   if (ThisFuncMDIt != NamedMD->op_end()) {
100     auto *ThisFuncMD = *ThisFuncMDIt;
101     MDNode *MD = dyn_cast<MDNode>(ThisFuncMD->getOperand(1));
102     assert(MD && "MDNode operand is expected");
103     ConstantInt *Const = getConstInt(MD, 0);
104     if (Const) {
105       auto *CMeta = dyn_cast<ConstantAsMetadata>(MD->getOperand(1));
106       assert(CMeta && "ConstantAsMetadata operand is expected");
107       assert(Const->getSExtValue() >= -1);
108       // Currently -1 indicates return value, greater values mean
109       // argument numbers.
110       if (Const->getSExtValue() == -1)
111         RetTy = CMeta->getType();
112       else
113         ArgTypes[Const->getSExtValue()] = CMeta->getType();
114     }
115   }
116 
117   return FunctionType::get(RetTy, ArgTypes, F.isVarArg());
118 }
119 
120 static MDString *getKernelArgAttribute(const Function &KernelFunction,
121                                        unsigned ArgIdx,
122                                        const StringRef AttributeName) {
123   assert(KernelFunction.getCallingConv() == CallingConv::SPIR_KERNEL &&
124          "Kernel attributes are attached/belong only to kernel functions");
125 
126   // Lookup the argument attribute in metadata attached to the kernel function.
127   MDNode *Node = KernelFunction.getMetadata(AttributeName);
128   if (Node && ArgIdx < Node->getNumOperands())
129     return cast<MDString>(Node->getOperand(ArgIdx));
130 
131   // Sometimes metadata containing kernel attributes is not attached to the
132   // function, but can be found in the named module-level metadata instead.
133   // For example:
134   //   !opencl.kernels = !{!0}
135   //   !0 = !{void ()* @someKernelFunction, !1, ...}
136   //   !1 = !{!"kernel_arg_addr_space", ...}
137   // In this case the actual index of searched argument attribute is ArgIdx + 1,
138   // since the first metadata node operand is occupied by attribute name
139   // ("kernel_arg_addr_space" in the example above).
140   unsigned MDArgIdx = ArgIdx + 1;
141   NamedMDNode *OpenCLKernelsMD =
142       KernelFunction.getParent()->getNamedMetadata("opencl.kernels");
143   if (!OpenCLKernelsMD || OpenCLKernelsMD->getNumOperands() == 0)
144     return nullptr;
145 
146   // KernelToMDNodeList contains kernel function declarations followed by
147   // corresponding MDNodes for each attribute. Search only MDNodes "belonging"
148   // to the currently lowered kernel function.
149   MDNode *KernelToMDNodeList = OpenCLKernelsMD->getOperand(0);
150   bool FoundLoweredKernelFunction = false;
151   for (const MDOperand &Operand : KernelToMDNodeList->operands()) {
152     ValueAsMetadata *MaybeValue = dyn_cast<ValueAsMetadata>(Operand);
153     if (MaybeValue && dyn_cast<Function>(MaybeValue->getValue())->getName() ==
154                           KernelFunction.getName()) {
155       FoundLoweredKernelFunction = true;
156       continue;
157     }
158     if (MaybeValue && FoundLoweredKernelFunction)
159       return nullptr;
160 
161     MDNode *MaybeNode = dyn_cast<MDNode>(Operand);
162     if (FoundLoweredKernelFunction && MaybeNode &&
163         cast<MDString>(MaybeNode->getOperand(0))->getString() ==
164             AttributeName &&
165         MDArgIdx < MaybeNode->getNumOperands())
166       return cast<MDString>(MaybeNode->getOperand(MDArgIdx));
167   }
168   return nullptr;
169 }
170 
171 static SPIRV::AccessQualifier::AccessQualifier
172 getArgAccessQual(const Function &F, unsigned ArgIdx) {
173   if (F.getCallingConv() != CallingConv::SPIR_KERNEL)
174     return SPIRV::AccessQualifier::ReadWrite;
175 
176   MDString *ArgAttribute =
177       getKernelArgAttribute(F, ArgIdx, "kernel_arg_access_qual");
178   if (!ArgAttribute)
179     return SPIRV::AccessQualifier::ReadWrite;
180 
181   if (ArgAttribute->getString().compare("read_only") == 0)
182     return SPIRV::AccessQualifier::ReadOnly;
183   if (ArgAttribute->getString().compare("write_only") == 0)
184     return SPIRV::AccessQualifier::WriteOnly;
185   return SPIRV::AccessQualifier::ReadWrite;
186 }
187 
188 static std::vector<SPIRV::Decoration::Decoration>
189 getKernelArgTypeQual(const Function &KernelFunction, unsigned ArgIdx) {
190   MDString *ArgAttribute =
191       getKernelArgAttribute(KernelFunction, ArgIdx, "kernel_arg_type_qual");
192   if (ArgAttribute && ArgAttribute->getString().compare("volatile") == 0)
193     return {SPIRV::Decoration::Volatile};
194   return {};
195 }
196 
197 static Type *getArgType(const Function &F, unsigned ArgIdx) {
198   Type *OriginalArgType = getOriginalFunctionType(F)->getParamType(ArgIdx);
199   if (F.getCallingConv() != CallingConv::SPIR_KERNEL ||
200       isSpecialOpaqueType(OriginalArgType))
201     return OriginalArgType;
202 
203   MDString *MDKernelArgType =
204       getKernelArgAttribute(F, ArgIdx, "kernel_arg_type");
205   if (!MDKernelArgType || !MDKernelArgType->getString().endswith("_t"))
206     return OriginalArgType;
207 
208   std::string KernelArgTypeStr = "opencl." + MDKernelArgType->getString().str();
209   Type *ExistingOpaqueType =
210       StructType::getTypeByName(F.getContext(), KernelArgTypeStr);
211   return ExistingOpaqueType
212              ? ExistingOpaqueType
213              : StructType::create(F.getContext(), KernelArgTypeStr);
214 }
215 
216 bool SPIRVCallLowering::lowerFormalArguments(MachineIRBuilder &MIRBuilder,
217                                              const Function &F,
218                                              ArrayRef<ArrayRef<Register>> VRegs,
219                                              FunctionLoweringInfo &FLI) const {
220   assert(GR && "Must initialize the SPIRV type registry before lowering args.");
221   GR->setCurrentFunc(MIRBuilder.getMF());
222 
223   // Assign types and names to all args, and store their types for later.
224   FunctionType *FTy = getOriginalFunctionType(F);
225   SmallVector<SPIRVType *, 4> ArgTypeVRegs;
226   if (VRegs.size() > 0) {
227     unsigned i = 0;
228     for (const auto &Arg : F.args()) {
229       // Currently formal args should use single registers.
230       // TODO: handle the case of multiple registers.
231       if (VRegs[i].size() > 1)
232         return false;
233       SPIRV::AccessQualifier::AccessQualifier ArgAccessQual =
234           getArgAccessQual(F, i);
235       auto *SpirvTy = GR->assignTypeToVReg(getArgType(F, i), VRegs[i][0],
236                                            MIRBuilder, ArgAccessQual);
237       ArgTypeVRegs.push_back(SpirvTy);
238 
239       if (Arg.hasName())
240         buildOpName(VRegs[i][0], Arg.getName(), MIRBuilder);
241       if (Arg.getType()->isPointerTy()) {
242         auto DerefBytes = static_cast<unsigned>(Arg.getDereferenceableBytes());
243         if (DerefBytes != 0)
244           buildOpDecorate(VRegs[i][0], MIRBuilder,
245                           SPIRV::Decoration::MaxByteOffset, {DerefBytes});
246       }
247       if (Arg.hasAttribute(Attribute::Alignment)) {
248         auto Alignment = static_cast<unsigned>(
249             Arg.getAttribute(Attribute::Alignment).getValueAsInt());
250         buildOpDecorate(VRegs[i][0], MIRBuilder, SPIRV::Decoration::Alignment,
251                         {Alignment});
252       }
253       if (Arg.hasAttribute(Attribute::ReadOnly)) {
254         auto Attr =
255             static_cast<unsigned>(SPIRV::FunctionParameterAttribute::NoWrite);
256         buildOpDecorate(VRegs[i][0], MIRBuilder,
257                         SPIRV::Decoration::FuncParamAttr, {Attr});
258       }
259       if (Arg.hasAttribute(Attribute::ZExt)) {
260         auto Attr =
261             static_cast<unsigned>(SPIRV::FunctionParameterAttribute::Zext);
262         buildOpDecorate(VRegs[i][0], MIRBuilder,
263                         SPIRV::Decoration::FuncParamAttr, {Attr});
264       }
265       if (Arg.hasAttribute(Attribute::NoAlias)) {
266         auto Attr =
267             static_cast<unsigned>(SPIRV::FunctionParameterAttribute::NoAlias);
268         buildOpDecorate(VRegs[i][0], MIRBuilder,
269                         SPIRV::Decoration::FuncParamAttr, {Attr});
270       }
271 
272       if (F.getCallingConv() == CallingConv::SPIR_KERNEL) {
273         std::vector<SPIRV::Decoration::Decoration> ArgTypeQualDecs =
274             getKernelArgTypeQual(F, i);
275         for (SPIRV::Decoration::Decoration Decoration : ArgTypeQualDecs)
276           buildOpDecorate(VRegs[i][0], MIRBuilder, Decoration, {});
277       }
278 
279       MDNode *Node = F.getMetadata("spirv.ParameterDecorations");
280       if (Node && i < Node->getNumOperands() &&
281           isa<MDNode>(Node->getOperand(i))) {
282         MDNode *MD = cast<MDNode>(Node->getOperand(i));
283         for (const MDOperand &MDOp : MD->operands()) {
284           MDNode *MD2 = dyn_cast<MDNode>(MDOp);
285           assert(MD2 && "Metadata operand is expected");
286           ConstantInt *Const = getConstInt(MD2, 0);
287           assert(Const && "MDOperand should be ConstantInt");
288           auto Dec =
289               static_cast<SPIRV::Decoration::Decoration>(Const->getZExtValue());
290           std::vector<uint32_t> DecVec;
291           for (unsigned j = 1; j < MD2->getNumOperands(); j++) {
292             ConstantInt *Const = getConstInt(MD2, j);
293             assert(Const && "MDOperand should be ConstantInt");
294             DecVec.push_back(static_cast<uint32_t>(Const->getZExtValue()));
295           }
296           buildOpDecorate(VRegs[i][0], MIRBuilder, Dec, DecVec);
297         }
298       }
299       ++i;
300     }
301   }
302 
303   // Generate a SPIR-V type for the function.
304   auto MRI = MIRBuilder.getMRI();
305   Register FuncVReg = MRI->createGenericVirtualRegister(LLT::scalar(32));
306   MRI->setRegClass(FuncVReg, &SPIRV::IDRegClass);
307   if (F.isDeclaration())
308     GR->add(&F, &MIRBuilder.getMF(), FuncVReg);
309   SPIRVType *RetTy = GR->getOrCreateSPIRVType(FTy->getReturnType(), MIRBuilder);
310   SPIRVType *FuncTy = GR->getOrCreateOpTypeFunctionWithArgs(
311       FTy, RetTy, ArgTypeVRegs, MIRBuilder);
312 
313   // Build the OpTypeFunction declaring it.
314   uint32_t FuncControl = getFunctionControl(F);
315 
316   MIRBuilder.buildInstr(SPIRV::OpFunction)
317       .addDef(FuncVReg)
318       .addUse(GR->getSPIRVTypeID(RetTy))
319       .addImm(FuncControl)
320       .addUse(GR->getSPIRVTypeID(FuncTy));
321 
322   // Add OpFunctionParameters.
323   int i = 0;
324   for (const auto &Arg : F.args()) {
325     assert(VRegs[i].size() == 1 && "Formal arg has multiple vregs");
326     MRI->setRegClass(VRegs[i][0], &SPIRV::IDRegClass);
327     MIRBuilder.buildInstr(SPIRV::OpFunctionParameter)
328         .addDef(VRegs[i][0])
329         .addUse(GR->getSPIRVTypeID(ArgTypeVRegs[i]));
330     if (F.isDeclaration())
331       GR->add(&Arg, &MIRBuilder.getMF(), VRegs[i][0]);
332     i++;
333   }
334   // Name the function.
335   if (F.hasName())
336     buildOpName(FuncVReg, F.getName(), MIRBuilder);
337 
338   // Handle entry points and function linkage.
339   if (F.getCallingConv() == CallingConv::SPIR_KERNEL) {
340     auto MIB = MIRBuilder.buildInstr(SPIRV::OpEntryPoint)
341                    .addImm(static_cast<uint32_t>(SPIRV::ExecutionModel::Kernel))
342                    .addUse(FuncVReg);
343     addStringImm(F.getName(), MIB);
344   } else if (F.getLinkage() == GlobalValue::LinkageTypes::ExternalLinkage ||
345              F.getLinkage() == GlobalValue::LinkOnceODRLinkage) {
346     auto LnkTy = F.isDeclaration() ? SPIRV::LinkageType::Import
347                                    : SPIRV::LinkageType::Export;
348     buildOpDecorate(FuncVReg, MIRBuilder, SPIRV::Decoration::LinkageAttributes,
349                     {static_cast<uint32_t>(LnkTy)}, F.getGlobalIdentifier());
350   }
351 
352   return true;
353 }
354 
355 bool SPIRVCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
356                                   CallLoweringInfo &Info) const {
357   // Currently call returns should have single vregs.
358   // TODO: handle the case of multiple registers.
359   if (Info.OrigRet.Regs.size() > 1)
360     return false;
361   MachineFunction &MF = MIRBuilder.getMF();
362   GR->setCurrentFunc(MF);
363   FunctionType *FTy = nullptr;
364   const Function *CF = nullptr;
365 
366   // Emit a regular OpFunctionCall. If it's an externally declared function,
367   // be sure to emit its type and function declaration here. It will be hoisted
368   // globally later.
369   if (Info.Callee.isGlobal()) {
370     CF = dyn_cast_or_null<const Function>(Info.Callee.getGlobal());
371     // TODO: support constexpr casts and indirect calls.
372     if (CF == nullptr)
373       return false;
374     FTy = getOriginalFunctionType(*CF);
375   }
376 
377   MachineRegisterInfo *MRI = MIRBuilder.getMRI();
378   Register ResVReg =
379       Info.OrigRet.Regs.empty() ? Register(0) : Info.OrigRet.Regs[0];
380   std::string FuncName = Info.Callee.getGlobal()->getName().str();
381   std::string DemangledName = getOclOrSpirvBuiltinDemangledName(FuncName);
382   const auto *ST = static_cast<const SPIRVSubtarget *>(&MF.getSubtarget());
383   // TODO: check that it's OCL builtin, then apply OpenCL_std.
384   if (!DemangledName.empty() && CF && CF->isDeclaration() &&
385       ST->canUseExtInstSet(SPIRV::InstructionSet::OpenCL_std)) {
386     const Type *OrigRetTy = Info.OrigRet.Ty;
387     if (FTy)
388       OrigRetTy = FTy->getReturnType();
389     SmallVector<Register, 8> ArgVRegs;
390     for (auto Arg : Info.OrigArgs) {
391       assert(Arg.Regs.size() == 1 && "Call arg has multiple VRegs");
392       ArgVRegs.push_back(Arg.Regs[0]);
393       SPIRVType *SPIRVTy = GR->getOrCreateSPIRVType(Arg.Ty, MIRBuilder);
394       GR->assignSPIRVTypeToVReg(SPIRVTy, Arg.Regs[0], MIRBuilder.getMF());
395     }
396     if (auto Res = SPIRV::lowerBuiltin(
397             DemangledName, SPIRV::InstructionSet::OpenCL_std, MIRBuilder,
398             ResVReg, OrigRetTy, ArgVRegs, GR))
399       return *Res;
400   }
401   if (CF && CF->isDeclaration() &&
402       !GR->find(CF, &MIRBuilder.getMF()).isValid()) {
403     // Emit the type info and forward function declaration to the first MBB
404     // to ensure VReg definition dependencies are valid across all MBBs.
405     MachineIRBuilder FirstBlockBuilder;
406     FirstBlockBuilder.setMF(MF);
407     FirstBlockBuilder.setMBB(*MF.getBlockNumbered(0));
408 
409     SmallVector<ArrayRef<Register>, 8> VRegArgs;
410     SmallVector<SmallVector<Register, 1>, 8> ToInsert;
411     for (const Argument &Arg : CF->args()) {
412       if (MIRBuilder.getDataLayout().getTypeStoreSize(Arg.getType()).isZero())
413         continue; // Don't handle zero sized types.
414       Register Reg = MRI->createGenericVirtualRegister(LLT::scalar(32));
415       MRI->setRegClass(Reg, &SPIRV::IDRegClass);
416       ToInsert.push_back({Reg});
417       VRegArgs.push_back(ToInsert.back());
418     }
419     // TODO: Reuse FunctionLoweringInfo
420     FunctionLoweringInfo FuncInfo;
421     lowerFormalArguments(FirstBlockBuilder, *CF, VRegArgs, FuncInfo);
422   }
423 
424   // Make sure there's a valid return reg, even for functions returning void.
425   if (!ResVReg.isValid())
426     ResVReg = MIRBuilder.getMRI()->createVirtualRegister(&SPIRV::IDRegClass);
427   SPIRVType *RetType =
428       GR->assignTypeToVReg(FTy->getReturnType(), ResVReg, MIRBuilder);
429 
430   // Emit the OpFunctionCall and its args.
431   auto MIB = MIRBuilder.buildInstr(SPIRV::OpFunctionCall)
432                  .addDef(ResVReg)
433                  .addUse(GR->getSPIRVTypeID(RetType))
434                  .add(Info.Callee);
435 
436   for (const auto &Arg : Info.OrigArgs) {
437     // Currently call args should have single vregs.
438     if (Arg.Regs.size() > 1)
439       return false;
440     MIB.addUse(Arg.Regs[0]);
441   }
442   return MIB.constrainAllUses(MIRBuilder.getTII(), *ST->getRegisterInfo(),
443                               *ST->getRegBankInfo());
444 }
445