| 1 | //===- Sparc.cpp ----------------------------------------------------------===// |
| 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 | #include "ABIInfoImpl.h" |
| 10 | #include "TargetInfo.h" |
| 11 | #include <algorithm> |
| 12 | |
| 13 | using namespace clang; |
| 14 | using namespace clang::CodeGen; |
| 15 | |
| 16 | //===----------------------------------------------------------------------===// |
| 17 | // SPARC v8 ABI Implementation. |
| 18 | // Based on the SPARC Compliance Definition version 2.4.1. |
| 19 | // |
| 20 | // Ensures that complex values are passed in registers. |
| 21 | // |
| 22 | namespace { |
| 23 | class SparcV8ABIInfo : public DefaultABIInfo { |
| 24 | public: |
| 25 | SparcV8ABIInfo(CodeGenTypes &CGT) : DefaultABIInfo(CGT) {} |
| 26 | |
| 27 | private: |
| 28 | ABIArgInfo classifyReturnType(QualType RetTy) const; |
| 29 | ABIArgInfo classifyArgumentType(QualType Ty) const; |
| 30 | void computeInfo(CGFunctionInfo &FI) const override; |
| 31 | }; |
| 32 | } // end anonymous namespace |
| 33 | |
| 34 | ABIArgInfo SparcV8ABIInfo::classifyReturnType(QualType Ty) const { |
| 35 | const auto *CT = Ty->getAs<ComplexType>(); |
| 36 | const auto *BT = Ty->getAs<BuiltinType>(); |
| 37 | if (CT) |
| 38 | BT = CT->getElementType()->getAs<BuiltinType>(); |
| 39 | bool IsLongDouble = BT && BT->getKind() == BuiltinType::LongDouble; |
| 40 | |
| 41 | // long double _Complex is special in that it should be marked as inreg. |
| 42 | if (CT) |
| 43 | return IsLongDouble ? ABIArgInfo::getDirectInReg() |
| 44 | : ABIArgInfo::getDirect(); |
| 45 | |
| 46 | if (IsLongDouble) |
| 47 | return getNaturalAlignIndirect(Ty, AddrSpace: getDataLayout().getAllocaAddrSpace(), |
| 48 | /*ByVal=*/false); |
| 49 | |
| 50 | return DefaultABIInfo::classifyReturnType(RetTy: Ty); |
| 51 | } |
| 52 | |
| 53 | ABIArgInfo SparcV8ABIInfo::classifyArgumentType(QualType Ty) const { |
| 54 | if (const auto *BT = Ty->getAs<BuiltinType>(); |
| 55 | BT && BT->getKind() == BuiltinType::LongDouble) |
| 56 | return getNaturalAlignIndirect(Ty, AddrSpace: getDataLayout().getAllocaAddrSpace()); |
| 57 | |
| 58 | return DefaultABIInfo::classifyArgumentType(RetTy: Ty); |
| 59 | } |
| 60 | |
| 61 | void SparcV8ABIInfo::computeInfo(CGFunctionInfo &FI) const { |
| 62 | FI.getReturnInfo() = classifyReturnType(Ty: FI.getReturnType()); |
| 63 | for (auto &Arg : FI.arguments()) |
| 64 | Arg.info = classifyArgumentType(Ty: Arg.type); |
| 65 | } |
| 66 | |
| 67 | namespace { |
| 68 | class SparcV8TargetCodeGenInfo : public TargetCodeGenInfo { |
| 69 | public: |
| 70 | SparcV8TargetCodeGenInfo(CodeGenTypes &CGT) |
| 71 | : TargetCodeGenInfo(std::make_unique<SparcV8ABIInfo>(args&: CGT)) {} |
| 72 | |
| 73 | llvm::Value *decodeReturnAddress(CodeGen::CodeGenFunction &CGF, |
| 74 | llvm::Value *Address) const override { |
| 75 | int Offset; |
| 76 | if (isAggregateTypeForABI(T: CGF.CurFnInfo->getReturnType())) |
| 77 | Offset = 12; |
| 78 | else |
| 79 | Offset = 8; |
| 80 | return CGF.Builder.CreateGEP(Ty: CGF.Int8Ty, Ptr: Address, |
| 81 | IdxList: llvm::ConstantInt::get(Ty: CGF.Int32Ty, V: Offset)); |
| 82 | } |
| 83 | |
| 84 | llvm::Value *encodeReturnAddress(CodeGen::CodeGenFunction &CGF, |
| 85 | llvm::Value *Address) const override { |
| 86 | int Offset; |
| 87 | if (isAggregateTypeForABI(T: CGF.CurFnInfo->getReturnType())) |
| 88 | Offset = -12; |
| 89 | else |
| 90 | Offset = -8; |
| 91 | return CGF.Builder.CreateGEP(Ty: CGF.Int8Ty, Ptr: Address, |
| 92 | IdxList: llvm::ConstantInt::get(Ty: CGF.Int32Ty, V: Offset)); |
| 93 | } |
| 94 | }; |
| 95 | } // end anonymous namespace |
| 96 | |
| 97 | //===----------------------------------------------------------------------===// |
| 98 | // SPARC v9 ABI Implementation. |
| 99 | // Based on the SPARC Compliance Definition version 2.4.1. |
| 100 | // |
| 101 | // Function arguments a mapped to a nominal "parameter array" and promoted to |
| 102 | // registers depending on their type. Each argument occupies 8 or 16 bytes in |
| 103 | // the array, structs larger than 16 bytes are passed indirectly. |
| 104 | // |
| 105 | // One case requires special care: |
| 106 | // |
| 107 | // struct mixed { |
| 108 | // int i; |
| 109 | // float f; |
| 110 | // }; |
| 111 | // |
| 112 | // When a struct mixed is passed by value, it only occupies 8 bytes in the |
| 113 | // parameter array, but the int is passed in an integer register, and the float |
| 114 | // is passed in a floating point register. This is represented as two arguments |
| 115 | // with the LLVM IR inreg attribute: |
| 116 | // |
| 117 | // declare void f(i32 inreg %i, float inreg %f) |
| 118 | // |
| 119 | // The code generator will only allocate 4 bytes from the parameter array for |
| 120 | // the inreg arguments. All other arguments are allocated a multiple of 8 |
| 121 | // bytes. |
| 122 | // |
| 123 | namespace { |
| 124 | class SparcV9ABIInfo : public ABIInfo { |
| 125 | public: |
| 126 | SparcV9ABIInfo(CodeGenTypes &CGT) : ABIInfo(CGT) {} |
| 127 | |
| 128 | private: |
| 129 | ABIArgInfo classifyType(QualType RetTy, unsigned SizeLimit, |
| 130 | unsigned &RegOffset) const; |
| 131 | void computeInfo(CGFunctionInfo &FI) const override; |
| 132 | RValue EmitVAArg(CodeGenFunction &CGF, Address VAListAddr, QualType Ty, |
| 133 | AggValueSlot Slot) const override; |
| 134 | |
| 135 | // Coercion type builder for structs passed in registers. The coercion type |
| 136 | // serves two purposes: |
| 137 | // |
| 138 | // 1. Pad structs to a multiple of 64 bits, so they are passed 'left-aligned' |
| 139 | // in registers. |
| 140 | // 2. Expose aligned floating point elements as first-level elements, so the |
| 141 | // code generator knows to pass them in floating point registers. |
| 142 | // |
| 143 | // We also compute the InReg flag which indicates that the struct contains |
| 144 | // aligned 32-bit floats. |
| 145 | // |
| 146 | struct CoerceBuilder { |
| 147 | llvm::LLVMContext &Context; |
| 148 | const llvm::DataLayout &DL; |
| 149 | SmallVector<llvm::Type*, 8> Elems; |
| 150 | uint64_t Size; |
| 151 | bool InReg; |
| 152 | |
| 153 | CoerceBuilder(llvm::LLVMContext &c, const llvm::DataLayout &dl) |
| 154 | : Context(c), DL(dl), Size(0), InReg(false) {} |
| 155 | |
| 156 | // Pad Elems with integers until Size is ToSize. |
| 157 | void pad(uint64_t ToSize) { |
| 158 | assert(ToSize >= Size && "Cannot remove elements" ); |
| 159 | if (ToSize == Size) |
| 160 | return; |
| 161 | |
| 162 | // Finish the current 64-bit word. |
| 163 | uint64_t Aligned = llvm::alignTo(Value: Size, Align: 64); |
| 164 | if (Aligned > Size && Aligned <= ToSize) { |
| 165 | Elems.push_back(Elt: llvm::IntegerType::get(C&: Context, NumBits: Aligned - Size)); |
| 166 | Size = Aligned; |
| 167 | } |
| 168 | |
| 169 | // Add whole 64-bit words. |
| 170 | while (Size + 64 <= ToSize) { |
| 171 | Elems.push_back(Elt: llvm::Type::getInt64Ty(C&: Context)); |
| 172 | Size += 64; |
| 173 | } |
| 174 | |
| 175 | // Final in-word padding. |
| 176 | if (Size < ToSize) { |
| 177 | Elems.push_back(Elt: llvm::IntegerType::get(C&: Context, NumBits: ToSize - Size)); |
| 178 | Size = ToSize; |
| 179 | } |
| 180 | } |
| 181 | |
| 182 | // Add a floating point element at Offset. |
| 183 | void addFloat(uint64_t Offset, llvm::Type *Ty, unsigned Bits) { |
| 184 | // Unaligned floats are treated as integers. |
| 185 | if (Offset % Bits) |
| 186 | return; |
| 187 | // The InReg flag is only required if there are any floats < 64 bits. |
| 188 | if (Bits < 64) |
| 189 | InReg = true; |
| 190 | pad(ToSize: Offset); |
| 191 | Elems.push_back(Elt: Ty); |
| 192 | Size = Offset + Bits; |
| 193 | } |
| 194 | |
| 195 | // Add a struct type to the coercion type, starting at Offset (in bits). |
| 196 | void addStruct(uint64_t Offset, llvm::StructType *StrTy) { |
| 197 | const llvm::StructLayout *Layout = DL.getStructLayout(Ty: StrTy); |
| 198 | for (unsigned i = 0, e = StrTy->getNumElements(); i != e; ++i) { |
| 199 | llvm::Type *ElemTy = StrTy->getElementType(N: i); |
| 200 | uint64_t ElemOffset = Offset + Layout->getElementOffsetInBits(Idx: i); |
| 201 | switch (ElemTy->getTypeID()) { |
| 202 | case llvm::Type::StructTyID: |
| 203 | addStruct(Offset: ElemOffset, StrTy: cast<llvm::StructType>(Val: ElemTy)); |
| 204 | break; |
| 205 | case llvm::Type::FloatTyID: |
| 206 | addFloat(Offset: ElemOffset, Ty: ElemTy, Bits: 32); |
| 207 | break; |
| 208 | case llvm::Type::DoubleTyID: |
| 209 | addFloat(Offset: ElemOffset, Ty: ElemTy, Bits: 64); |
| 210 | break; |
| 211 | case llvm::Type::FP128TyID: |
| 212 | addFloat(Offset: ElemOffset, Ty: ElemTy, Bits: 128); |
| 213 | break; |
| 214 | case llvm::Type::PointerTyID: |
| 215 | if (ElemOffset % 64 == 0) { |
| 216 | pad(ToSize: ElemOffset); |
| 217 | Elems.push_back(Elt: ElemTy); |
| 218 | Size += 64; |
| 219 | } |
| 220 | break; |
| 221 | default: |
| 222 | break; |
| 223 | } |
| 224 | } |
| 225 | } |
| 226 | |
| 227 | // Check if Ty is a usable substitute for the coercion type. |
| 228 | bool isUsableType(llvm::StructType *Ty) const { |
| 229 | return llvm::ArrayRef(Elems) == Ty->elements(); |
| 230 | } |
| 231 | |
| 232 | // Get the coercion type as a literal struct type. |
| 233 | llvm::Type *getType() const { |
| 234 | if (Elems.size() == 1) |
| 235 | return Elems.front(); |
| 236 | else |
| 237 | return llvm::StructType::get(Context, Elements: Elems); |
| 238 | } |
| 239 | }; |
| 240 | }; |
| 241 | } // end anonymous namespace |
| 242 | |
| 243 | ABIArgInfo SparcV9ABIInfo::classifyType(QualType Ty, unsigned SizeLimit, |
| 244 | unsigned &RegOffset) const { |
| 245 | if (Ty->isVoidType()) |
| 246 | return ABIArgInfo::getIgnore(); |
| 247 | |
| 248 | auto &Context = getContext(); |
| 249 | auto &VMContext = getVMContext(); |
| 250 | |
| 251 | uint64_t Size = Context.getTypeSize(T: Ty); |
| 252 | unsigned Alignment = Context.getTypeAlign(T: Ty); |
| 253 | bool NeedPadding = (Alignment > 64) && (RegOffset % 2 != 0); |
| 254 | |
| 255 | // Anything too big to fit in registers is passed with an explicit indirect |
| 256 | // pointer / sret pointer. |
| 257 | if (Size > SizeLimit) { |
| 258 | RegOffset += 1; |
| 259 | return getNaturalAlignIndirect( |
| 260 | Ty, /*AddrSpace=*/getDataLayout().getAllocaAddrSpace(), |
| 261 | /*ByVal=*/false); |
| 262 | } |
| 263 | |
| 264 | // Treat an enum type as its underlying type. |
| 265 | if (const auto *ED = Ty->getAsEnumDecl()) |
| 266 | Ty = ED->getIntegerType(); |
| 267 | |
| 268 | // Integer types smaller than a register are extended. |
| 269 | if (Size < 64 && Ty->isIntegerType()) { |
| 270 | RegOffset += 1; |
| 271 | return ABIArgInfo::getExtend(Ty); |
| 272 | } |
| 273 | |
| 274 | if (const auto *EIT = Ty->getAs<BitIntType>()) |
| 275 | if (EIT->getNumBits() < 64) { |
| 276 | RegOffset += 1; |
| 277 | return ABIArgInfo::getExtend(Ty); |
| 278 | } |
| 279 | |
| 280 | // Other non-aggregates go in registers. |
| 281 | if (!isAggregateTypeForABI(T: Ty)) { |
| 282 | RegOffset += Size / 64; |
| 283 | return ABIArgInfo::getDirect(); |
| 284 | } |
| 285 | |
| 286 | // If a C++ object has either a non-trivial copy constructor or a non-trivial |
| 287 | // destructor, it is passed with an explicit indirect pointer / sret pointer. |
| 288 | if (CGCXXABI::RecordArgABI RAA = getRecordArgABI(T: Ty, CXXABI&: getCXXABI())) { |
| 289 | RegOffset += 1; |
| 290 | return getNaturalAlignIndirect(Ty, AddrSpace: getDataLayout().getAllocaAddrSpace(), |
| 291 | ByVal: RAA == CGCXXABI::RAA_DirectInMemory); |
| 292 | } |
| 293 | |
| 294 | // This is a small aggregate type that should be passed in registers. |
| 295 | // Build a coercion type from the LLVM struct type. |
| 296 | llvm::StructType *StrTy = dyn_cast<llvm::StructType>(Val: CGT.ConvertType(T: Ty)); |
| 297 | if (!StrTy) { |
| 298 | RegOffset += Size / 64; |
| 299 | return ABIArgInfo::getDirect(); |
| 300 | } |
| 301 | |
| 302 | CoerceBuilder CB(VMContext, getDataLayout()); |
| 303 | CB.addStruct(Offset: 0, StrTy); |
| 304 | // All structs, even empty ones, should take up a register argument slot, |
| 305 | // so pin the minimum struct size to one bit. |
| 306 | CB.pad(ToSize: llvm::alignTo( |
| 307 | Value: std::max(a: CB.DL.getTypeSizeInBits(Ty: StrTy).getKnownMinValue(), b: uint64_t(1)), |
| 308 | Align: 64)); |
| 309 | RegOffset += CB.Size / 64; |
| 310 | |
| 311 | // If we're dealing with overaligned structs we may need to add a padding in |
| 312 | // the front, to preserve the correct register-memory mapping. |
| 313 | // |
| 314 | // See SCD 2.4.1, pages 3P-11 and 3P-12. |
| 315 | llvm::Type *Padding = |
| 316 | NeedPadding ? llvm::Type::getInt64Ty(C&: VMContext) : nullptr; |
| 317 | RegOffset += NeedPadding ? 1 : 0; |
| 318 | |
| 319 | // Try to use the original type for coercion. |
| 320 | llvm::Type *CoerceTy = CB.isUsableType(Ty: StrTy) ? StrTy : CB.getType(); |
| 321 | |
| 322 | ABIArgInfo AAI = ABIArgInfo::getDirect(T: CoerceTy, Offset: 0, Padding); |
| 323 | AAI.setInReg(CB.InReg); |
| 324 | return AAI; |
| 325 | } |
| 326 | |
| 327 | RValue SparcV9ABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAListAddr, |
| 328 | QualType Ty, AggValueSlot Slot) const { |
| 329 | CharUnits SlotSize = CharUnits::fromQuantity(Quantity: 8); |
| 330 | auto TInfo = getContext().getTypeInfoInChars(T: Ty); |
| 331 | |
| 332 | // Zero-sized types have a width of one byte for parameter passing purposes. |
| 333 | TInfo.Width = std::max(a: TInfo.Width, b: CharUnits::fromQuantity(Quantity: 1)); |
| 334 | |
| 335 | // Arguments bigger than 2*SlotSize bytes are passed indirectly. |
| 336 | return emitVoidPtrVAArg(CGF, VAListAddr, ValueTy: Ty, |
| 337 | /*IsIndirect=*/TInfo.Width > 2 * SlotSize, ValueInfo: TInfo, |
| 338 | SlotSizeAndAlign: SlotSize, |
| 339 | /*AllowHigherAlign=*/true, Slot); |
| 340 | } |
| 341 | |
| 342 | void SparcV9ABIInfo::computeInfo(CGFunctionInfo &FI) const { |
| 343 | unsigned RetOffset = 0; |
| 344 | ABIArgInfo RetType = classifyType(Ty: FI.getReturnType(), SizeLimit: 32 * 8, RegOffset&: RetOffset); |
| 345 | FI.getReturnInfo() = RetType; |
| 346 | |
| 347 | // Indirect returns will have its pointer passed as an argument. |
| 348 | unsigned ArgOffset = RetType.isIndirect() ? RetOffset : 0; |
| 349 | for (auto &I : FI.arguments()) |
| 350 | I.info = classifyType(Ty: I.type, SizeLimit: 16 * 8, RegOffset&: ArgOffset); |
| 351 | } |
| 352 | |
| 353 | namespace { |
| 354 | class SparcV9TargetCodeGenInfo : public TargetCodeGenInfo { |
| 355 | public: |
| 356 | SparcV9TargetCodeGenInfo(CodeGenTypes &CGT) |
| 357 | : TargetCodeGenInfo(std::make_unique<SparcV9ABIInfo>(args&: CGT)) {} |
| 358 | |
| 359 | int getDwarfEHStackPointer(CodeGen::CodeGenModule &M) const override { |
| 360 | return 14; |
| 361 | } |
| 362 | |
| 363 | bool initDwarfEHRegSizeTable(CodeGen::CodeGenFunction &CGF, |
| 364 | llvm::Value *Address) const override; |
| 365 | |
| 366 | llvm::Value *decodeReturnAddress(CodeGen::CodeGenFunction &CGF, |
| 367 | llvm::Value *Address) const override { |
| 368 | return CGF.Builder.CreateGEP(Ty: CGF.Int8Ty, Ptr: Address, |
| 369 | IdxList: llvm::ConstantInt::get(Ty: CGF.Int32Ty, V: 8)); |
| 370 | } |
| 371 | |
| 372 | llvm::Value *encodeReturnAddress(CodeGen::CodeGenFunction &CGF, |
| 373 | llvm::Value *Address) const override { |
| 374 | return CGF.Builder.CreateGEP(Ty: CGF.Int8Ty, Ptr: Address, |
| 375 | IdxList: llvm::ConstantInt::get(Ty: CGF.Int32Ty, V: -8)); |
| 376 | } |
| 377 | }; |
| 378 | } // end anonymous namespace |
| 379 | |
| 380 | bool |
| 381 | SparcV9TargetCodeGenInfo::initDwarfEHRegSizeTable(CodeGen::CodeGenFunction &CGF, |
| 382 | llvm::Value *Address) const { |
| 383 | // This is calculated from the LLVM and GCC tables and verified |
| 384 | // against gcc output. AFAIK all ABIs use the same encoding. |
| 385 | |
| 386 | CodeGen::CGBuilderTy &Builder = CGF.Builder; |
| 387 | |
| 388 | llvm::IntegerType *i8 = CGF.Int8Ty; |
| 389 | llvm::Value *Four8 = llvm::ConstantInt::get(Ty: i8, V: 4); |
| 390 | llvm::Value *Eight8 = llvm::ConstantInt::get(Ty: i8, V: 8); |
| 391 | |
| 392 | // 0-31: the 8-byte general-purpose registers |
| 393 | AssignToArrayRange(Builder, Array: Address, Value: Eight8, FirstIndex: 0, LastIndex: 31); |
| 394 | |
| 395 | // 32-63: f0-31, the 4-byte floating-point registers |
| 396 | AssignToArrayRange(Builder, Array: Address, Value: Four8, FirstIndex: 32, LastIndex: 63); |
| 397 | |
| 398 | // Y = 64 |
| 399 | // PSR = 65 |
| 400 | // WIM = 66 |
| 401 | // TBR = 67 |
| 402 | // PC = 68 |
| 403 | // NPC = 69 |
| 404 | // FSR = 70 |
| 405 | // CSR = 71 |
| 406 | AssignToArrayRange(Builder, Array: Address, Value: Eight8, FirstIndex: 64, LastIndex: 71); |
| 407 | |
| 408 | // 72-87: d0-15, the 8-byte floating-point registers |
| 409 | AssignToArrayRange(Builder, Array: Address, Value: Eight8, FirstIndex: 72, LastIndex: 87); |
| 410 | |
| 411 | return false; |
| 412 | } |
| 413 | |
| 414 | std::unique_ptr<TargetCodeGenInfo> |
| 415 | CodeGen::createSparcV8TargetCodeGenInfo(CodeGenModule &CGM) { |
| 416 | return std::make_unique<SparcV8TargetCodeGenInfo>(args&: CGM.getTypes()); |
| 417 | } |
| 418 | |
| 419 | std::unique_ptr<TargetCodeGenInfo> |
| 420 | CodeGen::createSparcV9TargetCodeGenInfo(CodeGenModule &CGM) { |
| 421 | return std::make_unique<SparcV9TargetCodeGenInfo>(args&: CGM.getTypes()); |
| 422 | } |
| 423 | |