1 //===--- SwiftCallingConv.cpp - Lowering for the Swift calling convention -===// 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 // Implementation of the abstract lowering for the Swift calling convention. 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "clang/CodeGen/SwiftCallingConv.h" 14 #include "ABIInfo.h" 15 #include "CodeGenModule.h" 16 #include "TargetInfo.h" 17 #include "clang/Basic/TargetInfo.h" 18 19 using namespace clang; 20 using namespace CodeGen; 21 using namespace swiftcall; 22 23 static const SwiftABIInfo &getSwiftABIInfo(CodeGenModule &CGM) { 24 return CGM.getTargetCodeGenInfo().getSwiftABIInfo(); 25 } 26 27 static bool isPowerOf2(unsigned n) { 28 return n == (n & -n); 29 } 30 31 /// Given two types with the same size, try to find a common type. 32 static llvm::Type *getCommonType(llvm::Type *first, llvm::Type *second) { 33 assert(first != second); 34 35 // Allow pointers to merge with integers, but prefer the integer type. 36 if (first->isIntegerTy()) { 37 if (second->isPointerTy()) return first; 38 } else if (first->isPointerTy()) { 39 if (second->isIntegerTy()) return second; 40 if (second->isPointerTy()) return first; 41 42 // Allow two vectors to be merged (given that they have the same size). 43 // This assumes that we never have two different vector register sets. 44 } else if (auto firstVecTy = dyn_cast<llvm::VectorType>(first)) { 45 if (auto secondVecTy = dyn_cast<llvm::VectorType>(second)) { 46 if (auto commonTy = getCommonType(firstVecTy->getElementType(), 47 secondVecTy->getElementType())) { 48 return (commonTy == firstVecTy->getElementType() ? first : second); 49 } 50 } 51 } 52 53 return nullptr; 54 } 55 56 static CharUnits getTypeStoreSize(CodeGenModule &CGM, llvm::Type *type) { 57 return CharUnits::fromQuantity(CGM.getDataLayout().getTypeStoreSize(type)); 58 } 59 60 static CharUnits getTypeAllocSize(CodeGenModule &CGM, llvm::Type *type) { 61 return CharUnits::fromQuantity(CGM.getDataLayout().getTypeAllocSize(type)); 62 } 63 64 void SwiftAggLowering::addTypedData(QualType type, CharUnits begin) { 65 // Deal with various aggregate types as special cases: 66 67 // Record types. 68 if (auto recType = type->getAs<RecordType>()) { 69 addTypedData(recType->getDecl(), begin); 70 71 // Array types. 72 } else if (type->isArrayType()) { 73 // Incomplete array types (flexible array members?) don't provide 74 // data to lay out, and the other cases shouldn't be possible. 75 auto arrayType = CGM.getContext().getAsConstantArrayType(type); 76 if (!arrayType) return; 77 78 QualType eltType = arrayType->getElementType(); 79 auto eltSize = CGM.getContext().getTypeSizeInChars(eltType); 80 for (uint64_t i = 0, e = arrayType->getZExtSize(); i != e; ++i) { 81 addTypedData(eltType, begin + i * eltSize); 82 } 83 84 // Complex types. 85 } else if (auto complexType = type->getAs<ComplexType>()) { 86 auto eltType = complexType->getElementType(); 87 auto eltSize = CGM.getContext().getTypeSizeInChars(eltType); 88 auto eltLLVMType = CGM.getTypes().ConvertType(eltType); 89 addTypedData(eltLLVMType, begin, begin + eltSize); 90 addTypedData(eltLLVMType, begin + eltSize, begin + 2 * eltSize); 91 92 // Member pointer types. 93 } else if (type->getAs<MemberPointerType>()) { 94 // Just add it all as opaque. 95 addOpaqueData(begin, begin + CGM.getContext().getTypeSizeInChars(type)); 96 97 // Atomic types. 98 } else if (const auto *atomicType = type->getAs<AtomicType>()) { 99 auto valueType = atomicType->getValueType(); 100 auto atomicSize = CGM.getContext().getTypeSizeInChars(atomicType); 101 auto valueSize = CGM.getContext().getTypeSizeInChars(valueType); 102 103 addTypedData(atomicType->getValueType(), begin); 104 105 // Add atomic padding. 106 auto atomicPadding = atomicSize - valueSize; 107 if (atomicPadding > CharUnits::Zero()) 108 addOpaqueData(begin + valueSize, begin + atomicSize); 109 110 // Everything else is scalar and should not convert as an LLVM aggregate. 111 } else { 112 // We intentionally convert as !ForMem because we want to preserve 113 // that a type was an i1. 114 auto *llvmType = CGM.getTypes().ConvertType(type); 115 addTypedData(llvmType, begin); 116 } 117 } 118 119 void SwiftAggLowering::addTypedData(const RecordDecl *record, CharUnits begin) { 120 addTypedData(record, begin, CGM.getContext().getASTRecordLayout(record)); 121 } 122 123 void SwiftAggLowering::addTypedData(const RecordDecl *record, CharUnits begin, 124 const ASTRecordLayout &layout) { 125 // Unions are a special case. 126 if (record->isUnion()) { 127 for (auto *field : record->fields()) { 128 if (field->isBitField()) { 129 addBitFieldData(field, begin, 0); 130 } else { 131 addTypedData(field->getType(), begin); 132 } 133 } 134 return; 135 } 136 137 // Note that correctness does not rely on us adding things in 138 // their actual order of layout; it's just somewhat more efficient 139 // for the builder. 140 141 // With that in mind, add "early" C++ data. 142 auto cxxRecord = dyn_cast<CXXRecordDecl>(record); 143 if (cxxRecord) { 144 // - a v-table pointer, if the class adds its own 145 if (layout.hasOwnVFPtr()) { 146 addTypedData(CGM.Int8PtrTy, begin); 147 } 148 149 // - non-virtual bases 150 for (auto &baseSpecifier : cxxRecord->bases()) { 151 if (baseSpecifier.isVirtual()) continue; 152 153 auto baseRecord = baseSpecifier.getType()->getAsCXXRecordDecl(); 154 addTypedData(baseRecord, begin + layout.getBaseClassOffset(baseRecord)); 155 } 156 157 // - a vbptr if the class adds its own 158 if (layout.hasOwnVBPtr()) { 159 addTypedData(CGM.Int8PtrTy, begin + layout.getVBPtrOffset()); 160 } 161 } 162 163 // Add fields. 164 for (auto *field : record->fields()) { 165 auto fieldOffsetInBits = layout.getFieldOffset(field->getFieldIndex()); 166 if (field->isBitField()) { 167 addBitFieldData(field, begin, fieldOffsetInBits); 168 } else { 169 addTypedData(field->getType(), 170 begin + CGM.getContext().toCharUnitsFromBits(fieldOffsetInBits)); 171 } 172 } 173 174 // Add "late" C++ data: 175 if (cxxRecord) { 176 // - virtual bases 177 for (auto &vbaseSpecifier : cxxRecord->vbases()) { 178 auto baseRecord = vbaseSpecifier.getType()->getAsCXXRecordDecl(); 179 addTypedData(baseRecord, begin + layout.getVBaseClassOffset(baseRecord)); 180 } 181 } 182 } 183 184 void SwiftAggLowering::addBitFieldData(const FieldDecl *bitfield, 185 CharUnits recordBegin, 186 uint64_t bitfieldBitBegin) { 187 assert(bitfield->isBitField()); 188 auto &ctx = CGM.getContext(); 189 auto width = bitfield->getBitWidthValue(); 190 191 // We can ignore zero-width bit-fields. 192 if (width == 0) return; 193 194 // toCharUnitsFromBits rounds down. 195 CharUnits bitfieldByteBegin = ctx.toCharUnitsFromBits(bitfieldBitBegin); 196 197 // Find the offset of the last byte that is partially occupied by the 198 // bit-field; since we otherwise expect exclusive ends, the end is the 199 // next byte. 200 uint64_t bitfieldBitLast = bitfieldBitBegin + width - 1; 201 CharUnits bitfieldByteEnd = 202 ctx.toCharUnitsFromBits(bitfieldBitLast) + CharUnits::One(); 203 addOpaqueData(recordBegin + bitfieldByteBegin, 204 recordBegin + bitfieldByteEnd); 205 } 206 207 void SwiftAggLowering::addTypedData(llvm::Type *type, CharUnits begin) { 208 assert(type && "didn't provide type for typed data"); 209 addTypedData(type, begin, begin + getTypeStoreSize(CGM, type)); 210 } 211 212 void SwiftAggLowering::addTypedData(llvm::Type *type, 213 CharUnits begin, CharUnits end) { 214 assert(type && "didn't provide type for typed data"); 215 assert(getTypeStoreSize(CGM, type) == end - begin); 216 217 // Legalize vector types. 218 if (auto vecTy = dyn_cast<llvm::VectorType>(type)) { 219 SmallVector<llvm::Type*, 4> componentTys; 220 legalizeVectorType(CGM, end - begin, vecTy, componentTys); 221 assert(componentTys.size() >= 1); 222 223 // Walk the initial components. 224 for (size_t i = 0, e = componentTys.size(); i != e - 1; ++i) { 225 llvm::Type *componentTy = componentTys[i]; 226 auto componentSize = getTypeStoreSize(CGM, componentTy); 227 assert(componentSize < end - begin); 228 addLegalTypedData(componentTy, begin, begin + componentSize); 229 begin += componentSize; 230 } 231 232 return addLegalTypedData(componentTys.back(), begin, end); 233 } 234 235 // Legalize integer types. 236 if (auto intTy = dyn_cast<llvm::IntegerType>(type)) { 237 if (!isLegalIntegerType(CGM, intTy)) 238 return addOpaqueData(begin, end); 239 } 240 241 // All other types should be legal. 242 return addLegalTypedData(type, begin, end); 243 } 244 245 void SwiftAggLowering::addLegalTypedData(llvm::Type *type, 246 CharUnits begin, CharUnits end) { 247 // Require the type to be naturally aligned. 248 if (!begin.isZero() && !begin.isMultipleOf(getNaturalAlignment(CGM, type))) { 249 250 // Try splitting vector types. 251 if (auto vecTy = dyn_cast<llvm::VectorType>(type)) { 252 auto split = splitLegalVectorType(CGM, end - begin, vecTy); 253 auto eltTy = split.first; 254 auto numElts = split.second; 255 256 auto eltSize = (end - begin) / numElts; 257 assert(eltSize == getTypeStoreSize(CGM, eltTy)); 258 for (size_t i = 0, e = numElts; i != e; ++i) { 259 addLegalTypedData(eltTy, begin, begin + eltSize); 260 begin += eltSize; 261 } 262 assert(begin == end); 263 return; 264 } 265 266 return addOpaqueData(begin, end); 267 } 268 269 addEntry(type, begin, end); 270 } 271 272 void SwiftAggLowering::addEntry(llvm::Type *type, 273 CharUnits begin, CharUnits end) { 274 assert((!type || 275 (!isa<llvm::StructType>(type) && !isa<llvm::ArrayType>(type))) && 276 "cannot add aggregate-typed data"); 277 assert(!type || begin.isMultipleOf(getNaturalAlignment(CGM, type))); 278 279 // Fast path: we can just add entries to the end. 280 if (Entries.empty() || Entries.back().End <= begin) { 281 Entries.push_back({begin, end, type}); 282 return; 283 } 284 285 // Find the first existing entry that ends after the start of the new data. 286 // TODO: do a binary search if Entries is big enough for it to matter. 287 size_t index = Entries.size() - 1; 288 while (index != 0) { 289 if (Entries[index - 1].End <= begin) break; 290 --index; 291 } 292 293 // The entry ends after the start of the new data. 294 // If the entry starts after the end of the new data, there's no conflict. 295 if (Entries[index].Begin >= end) { 296 // This insertion is potentially O(n), but the way we generally build 297 // these layouts makes that unlikely to matter: we'd need a union of 298 // several very large types. 299 Entries.insert(Entries.begin() + index, {begin, end, type}); 300 return; 301 } 302 303 // Otherwise, the ranges overlap. The new range might also overlap 304 // with later ranges. 305 restartAfterSplit: 306 307 // Simplest case: an exact overlap. 308 if (Entries[index].Begin == begin && Entries[index].End == end) { 309 // If the types match exactly, great. 310 if (Entries[index].Type == type) return; 311 312 // If either type is opaque, make the entry opaque and return. 313 if (Entries[index].Type == nullptr) { 314 return; 315 } else if (type == nullptr) { 316 Entries[index].Type = nullptr; 317 return; 318 } 319 320 // If they disagree in an ABI-agnostic way, just resolve the conflict 321 // arbitrarily. 322 if (auto entryType = getCommonType(Entries[index].Type, type)) { 323 Entries[index].Type = entryType; 324 return; 325 } 326 327 // Otherwise, make the entry opaque. 328 Entries[index].Type = nullptr; 329 return; 330 } 331 332 // Okay, we have an overlapping conflict of some sort. 333 334 // If we have a vector type, split it. 335 if (auto vecTy = dyn_cast_or_null<llvm::VectorType>(type)) { 336 auto eltTy = vecTy->getElementType(); 337 CharUnits eltSize = 338 (end - begin) / cast<llvm::FixedVectorType>(vecTy)->getNumElements(); 339 assert(eltSize == getTypeStoreSize(CGM, eltTy)); 340 for (unsigned i = 0, 341 e = cast<llvm::FixedVectorType>(vecTy)->getNumElements(); 342 i != e; ++i) { 343 addEntry(eltTy, begin, begin + eltSize); 344 begin += eltSize; 345 } 346 assert(begin == end); 347 return; 348 } 349 350 // If the entry is a vector type, split it and try again. 351 if (Entries[index].Type && Entries[index].Type->isVectorTy()) { 352 splitVectorEntry(index); 353 goto restartAfterSplit; 354 } 355 356 // Okay, we have no choice but to make the existing entry opaque. 357 358 Entries[index].Type = nullptr; 359 360 // Stretch the start of the entry to the beginning of the range. 361 if (begin < Entries[index].Begin) { 362 Entries[index].Begin = begin; 363 assert(index == 0 || begin >= Entries[index - 1].End); 364 } 365 366 // Stretch the end of the entry to the end of the range; but if we run 367 // into the start of the next entry, just leave the range there and repeat. 368 while (end > Entries[index].End) { 369 assert(Entries[index].Type == nullptr); 370 371 // If the range doesn't overlap the next entry, we're done. 372 if (index == Entries.size() - 1 || end <= Entries[index + 1].Begin) { 373 Entries[index].End = end; 374 break; 375 } 376 377 // Otherwise, stretch to the start of the next entry. 378 Entries[index].End = Entries[index + 1].Begin; 379 380 // Continue with the next entry. 381 index++; 382 383 // This entry needs to be made opaque if it is not already. 384 if (Entries[index].Type == nullptr) 385 continue; 386 387 // Split vector entries unless we completely subsume them. 388 if (Entries[index].Type->isVectorTy() && 389 end < Entries[index].End) { 390 splitVectorEntry(index); 391 } 392 393 // Make the entry opaque. 394 Entries[index].Type = nullptr; 395 } 396 } 397 398 /// Replace the entry of vector type at offset 'index' with a sequence 399 /// of its component vectors. 400 void SwiftAggLowering::splitVectorEntry(unsigned index) { 401 auto vecTy = cast<llvm::VectorType>(Entries[index].Type); 402 auto split = splitLegalVectorType(CGM, Entries[index].getWidth(), vecTy); 403 404 auto eltTy = split.first; 405 CharUnits eltSize = getTypeStoreSize(CGM, eltTy); 406 auto numElts = split.second; 407 Entries.insert(Entries.begin() + index + 1, numElts - 1, StorageEntry()); 408 409 CharUnits begin = Entries[index].Begin; 410 for (unsigned i = 0; i != numElts; ++i) { 411 unsigned idx = index + i; 412 Entries[idx].Type = eltTy; 413 Entries[idx].Begin = begin; 414 Entries[idx].End = begin + eltSize; 415 begin += eltSize; 416 } 417 } 418 419 /// Given a power-of-two unit size, return the offset of the aligned unit 420 /// of that size which contains the given offset. 421 /// 422 /// In other words, round down to the nearest multiple of the unit size. 423 static CharUnits getOffsetAtStartOfUnit(CharUnits offset, CharUnits unitSize) { 424 assert(isPowerOf2(unitSize.getQuantity())); 425 auto unitMask = ~(unitSize.getQuantity() - 1); 426 return CharUnits::fromQuantity(offset.getQuantity() & unitMask); 427 } 428 429 static bool areBytesInSameUnit(CharUnits first, CharUnits second, 430 CharUnits chunkSize) { 431 return getOffsetAtStartOfUnit(first, chunkSize) 432 == getOffsetAtStartOfUnit(second, chunkSize); 433 } 434 435 static bool isMergeableEntryType(llvm::Type *type) { 436 // Opaquely-typed memory is always mergeable. 437 if (type == nullptr) return true; 438 439 // Pointers and integers are always mergeable. In theory we should not 440 // merge pointers, but (1) it doesn't currently matter in practice because 441 // the chunk size is never greater than the size of a pointer and (2) 442 // Swift IRGen uses integer types for a lot of things that are "really" 443 // just storing pointers (like std::optional<SomePointer>). If we ever have a 444 // target that would otherwise combine pointers, we should put some effort 445 // into fixing those cases in Swift IRGen and then call out pointer types 446 // here. 447 448 // Floating-point and vector types should never be merged. 449 // Most such types are too large and highly-aligned to ever trigger merging 450 // in practice, but it's important for the rule to cover at least 'half' 451 // and 'float', as well as things like small vectors of 'i1' or 'i8'. 452 return (!type->isFloatingPointTy() && !type->isVectorTy()); 453 } 454 455 bool SwiftAggLowering::shouldMergeEntries(const StorageEntry &first, 456 const StorageEntry &second, 457 CharUnits chunkSize) { 458 // Only merge entries that overlap the same chunk. We test this first 459 // despite being a bit more expensive because this is the condition that 460 // tends to prevent merging. 461 if (!areBytesInSameUnit(first.End - CharUnits::One(), second.Begin, 462 chunkSize)) 463 return false; 464 465 return (isMergeableEntryType(first.Type) && 466 isMergeableEntryType(second.Type)); 467 } 468 469 void SwiftAggLowering::finish() { 470 if (Entries.empty()) { 471 Finished = true; 472 return; 473 } 474 475 // We logically split the layout down into a series of chunks of this size, 476 // which is generally the size of a pointer. 477 const CharUnits chunkSize = getMaximumVoluntaryIntegerSize(CGM); 478 479 // First pass: if two entries should be merged, make them both opaque 480 // and stretch one to meet the next. 481 // Also, remember if there are any opaque entries. 482 bool hasOpaqueEntries = (Entries[0].Type == nullptr); 483 for (size_t i = 1, e = Entries.size(); i != e; ++i) { 484 if (shouldMergeEntries(Entries[i - 1], Entries[i], chunkSize)) { 485 Entries[i - 1].Type = nullptr; 486 Entries[i].Type = nullptr; 487 Entries[i - 1].End = Entries[i].Begin; 488 hasOpaqueEntries = true; 489 490 } else if (Entries[i].Type == nullptr) { 491 hasOpaqueEntries = true; 492 } 493 } 494 495 // The rest of the algorithm leaves non-opaque entries alone, so if we 496 // have no opaque entries, we're done. 497 if (!hasOpaqueEntries) { 498 Finished = true; 499 return; 500 } 501 502 // Okay, move the entries to a temporary and rebuild Entries. 503 auto orig = std::move(Entries); 504 assert(Entries.empty()); 505 506 for (size_t i = 0, e = orig.size(); i != e; ++i) { 507 // Just copy over non-opaque entries. 508 if (orig[i].Type != nullptr) { 509 Entries.push_back(orig[i]); 510 continue; 511 } 512 513 // Scan forward to determine the full extent of the next opaque range. 514 // We know from the first pass that only contiguous ranges will overlap 515 // the same aligned chunk. 516 auto begin = orig[i].Begin; 517 auto end = orig[i].End; 518 while (i + 1 != e && 519 orig[i + 1].Type == nullptr && 520 end == orig[i + 1].Begin) { 521 end = orig[i + 1].End; 522 i++; 523 } 524 525 // Add an entry per intersected chunk. 526 do { 527 // Find the smallest aligned storage unit in the maximal aligned 528 // storage unit containing 'begin' that contains all the bytes in 529 // the intersection between the range and this chunk. 530 CharUnits localBegin = begin; 531 CharUnits chunkBegin = getOffsetAtStartOfUnit(localBegin, chunkSize); 532 CharUnits chunkEnd = chunkBegin + chunkSize; 533 CharUnits localEnd = std::min(end, chunkEnd); 534 535 // Just do a simple loop over ever-increasing unit sizes. 536 CharUnits unitSize = CharUnits::One(); 537 CharUnits unitBegin, unitEnd; 538 for (; ; unitSize *= 2) { 539 assert(unitSize <= chunkSize); 540 unitBegin = getOffsetAtStartOfUnit(localBegin, unitSize); 541 unitEnd = unitBegin + unitSize; 542 if (unitEnd >= localEnd) break; 543 } 544 545 // Add an entry for this unit. 546 auto entryTy = 547 llvm::IntegerType::get(CGM.getLLVMContext(), 548 CGM.getContext().toBits(unitSize)); 549 Entries.push_back({unitBegin, unitEnd, entryTy}); 550 551 // The next chunk starts where this chunk left off. 552 begin = localEnd; 553 } while (begin != end); 554 } 555 556 // Okay, finally finished. 557 Finished = true; 558 } 559 560 void SwiftAggLowering::enumerateComponents(EnumerationCallback callback) const { 561 assert(Finished && "haven't yet finished lowering"); 562 563 for (auto &entry : Entries) { 564 callback(entry.Begin, entry.End, entry.Type); 565 } 566 } 567 568 std::pair<llvm::StructType*, llvm::Type*> 569 SwiftAggLowering::getCoerceAndExpandTypes() const { 570 assert(Finished && "haven't yet finished lowering"); 571 572 auto &ctx = CGM.getLLVMContext(); 573 574 if (Entries.empty()) { 575 auto type = llvm::StructType::get(ctx); 576 return { type, type }; 577 } 578 579 SmallVector<llvm::Type*, 8> elts; 580 CharUnits lastEnd = CharUnits::Zero(); 581 bool hasPadding = false; 582 bool packed = false; 583 for (auto &entry : Entries) { 584 if (entry.Begin != lastEnd) { 585 auto paddingSize = entry.Begin - lastEnd; 586 assert(!paddingSize.isNegative()); 587 588 auto padding = llvm::ArrayType::get(llvm::Type::getInt8Ty(ctx), 589 paddingSize.getQuantity()); 590 elts.push_back(padding); 591 hasPadding = true; 592 } 593 594 if (!packed && !entry.Begin.isMultipleOf(CharUnits::fromQuantity( 595 CGM.getDataLayout().getABITypeAlign(entry.Type)))) 596 packed = true; 597 598 elts.push_back(entry.Type); 599 600 lastEnd = entry.Begin + getTypeAllocSize(CGM, entry.Type); 601 assert(entry.End <= lastEnd); 602 } 603 604 // We don't need to adjust 'packed' to deal with possible tail padding 605 // because we never do that kind of access through the coercion type. 606 auto coercionType = llvm::StructType::get(ctx, elts, packed); 607 608 llvm::Type *unpaddedType = coercionType; 609 if (hasPadding) { 610 elts.clear(); 611 for (auto &entry : Entries) { 612 elts.push_back(entry.Type); 613 } 614 if (elts.size() == 1) { 615 unpaddedType = elts[0]; 616 } else { 617 unpaddedType = llvm::StructType::get(ctx, elts, /*packed*/ false); 618 } 619 } else if (Entries.size() == 1) { 620 unpaddedType = Entries[0].Type; 621 } 622 623 return { coercionType, unpaddedType }; 624 } 625 626 bool SwiftAggLowering::shouldPassIndirectly(bool asReturnValue) const { 627 assert(Finished && "haven't yet finished lowering"); 628 629 // Empty types don't need to be passed indirectly. 630 if (Entries.empty()) return false; 631 632 // Avoid copying the array of types when there's just a single element. 633 if (Entries.size() == 1) { 634 return getSwiftABIInfo(CGM).shouldPassIndirectly(Entries.back().Type, 635 asReturnValue); 636 } 637 638 SmallVector<llvm::Type*, 8> componentTys; 639 componentTys.reserve(Entries.size()); 640 for (auto &entry : Entries) { 641 componentTys.push_back(entry.Type); 642 } 643 return getSwiftABIInfo(CGM).shouldPassIndirectly(componentTys, asReturnValue); 644 } 645 646 bool swiftcall::shouldPassIndirectly(CodeGenModule &CGM, 647 ArrayRef<llvm::Type*> componentTys, 648 bool asReturnValue) { 649 return getSwiftABIInfo(CGM).shouldPassIndirectly(componentTys, asReturnValue); 650 } 651 652 CharUnits swiftcall::getMaximumVoluntaryIntegerSize(CodeGenModule &CGM) { 653 // Currently always the size of an ordinary pointer. 654 return CGM.getContext().toCharUnitsFromBits( 655 CGM.getContext().getTargetInfo().getPointerWidth(LangAS::Default)); 656 } 657 658 CharUnits swiftcall::getNaturalAlignment(CodeGenModule &CGM, llvm::Type *type) { 659 // For Swift's purposes, this is always just the store size of the type 660 // rounded up to a power of 2. 661 auto size = (unsigned long long) getTypeStoreSize(CGM, type).getQuantity(); 662 size = llvm::bit_ceil(size); 663 assert(CGM.getDataLayout().getABITypeAlign(type) <= size); 664 return CharUnits::fromQuantity(size); 665 } 666 667 bool swiftcall::isLegalIntegerType(CodeGenModule &CGM, 668 llvm::IntegerType *intTy) { 669 auto size = intTy->getBitWidth(); 670 switch (size) { 671 case 1: 672 case 8: 673 case 16: 674 case 32: 675 case 64: 676 // Just assume that the above are always legal. 677 return true; 678 679 case 128: 680 return CGM.getContext().getTargetInfo().hasInt128Type(); 681 682 default: 683 return false; 684 } 685 } 686 687 bool swiftcall::isLegalVectorType(CodeGenModule &CGM, CharUnits vectorSize, 688 llvm::VectorType *vectorTy) { 689 return isLegalVectorType( 690 CGM, vectorSize, vectorTy->getElementType(), 691 cast<llvm::FixedVectorType>(vectorTy)->getNumElements()); 692 } 693 694 bool swiftcall::isLegalVectorType(CodeGenModule &CGM, CharUnits vectorSize, 695 llvm::Type *eltTy, unsigned numElts) { 696 assert(numElts > 1 && "illegal vector length"); 697 return getSwiftABIInfo(CGM).isLegalVectorType(vectorSize, eltTy, numElts); 698 } 699 700 std::pair<llvm::Type*, unsigned> 701 swiftcall::splitLegalVectorType(CodeGenModule &CGM, CharUnits vectorSize, 702 llvm::VectorType *vectorTy) { 703 auto numElts = cast<llvm::FixedVectorType>(vectorTy)->getNumElements(); 704 auto eltTy = vectorTy->getElementType(); 705 706 // Try to split the vector type in half. 707 if (numElts >= 4 && isPowerOf2(numElts)) { 708 if (isLegalVectorType(CGM, vectorSize / 2, eltTy, numElts / 2)) 709 return {llvm::FixedVectorType::get(eltTy, numElts / 2), 2}; 710 } 711 712 return {eltTy, numElts}; 713 } 714 715 void swiftcall::legalizeVectorType(CodeGenModule &CGM, CharUnits origVectorSize, 716 llvm::VectorType *origVectorTy, 717 llvm::SmallVectorImpl<llvm::Type*> &components) { 718 // If it's already a legal vector type, use it. 719 if (isLegalVectorType(CGM, origVectorSize, origVectorTy)) { 720 components.push_back(origVectorTy); 721 return; 722 } 723 724 // Try to split the vector into legal subvectors. 725 auto numElts = cast<llvm::FixedVectorType>(origVectorTy)->getNumElements(); 726 auto eltTy = origVectorTy->getElementType(); 727 assert(numElts != 1); 728 729 // The largest size that we're still considering making subvectors of. 730 // Always a power of 2. 731 unsigned logCandidateNumElts = llvm::Log2_32(numElts); 732 unsigned candidateNumElts = 1U << logCandidateNumElts; 733 assert(candidateNumElts <= numElts && candidateNumElts * 2 > numElts); 734 735 // Minor optimization: don't check the legality of this exact size twice. 736 if (candidateNumElts == numElts) { 737 logCandidateNumElts--; 738 candidateNumElts >>= 1; 739 } 740 741 CharUnits eltSize = (origVectorSize / numElts); 742 CharUnits candidateSize = eltSize * candidateNumElts; 743 744 // The sensibility of this algorithm relies on the fact that we never 745 // have a legal non-power-of-2 vector size without having the power of 2 746 // also be legal. 747 while (logCandidateNumElts > 0) { 748 assert(candidateNumElts == 1U << logCandidateNumElts); 749 assert(candidateNumElts <= numElts); 750 assert(candidateSize == eltSize * candidateNumElts); 751 752 // Skip illegal vector sizes. 753 if (!isLegalVectorType(CGM, candidateSize, eltTy, candidateNumElts)) { 754 logCandidateNumElts--; 755 candidateNumElts /= 2; 756 candidateSize /= 2; 757 continue; 758 } 759 760 // Add the right number of vectors of this size. 761 auto numVecs = numElts >> logCandidateNumElts; 762 components.append(numVecs, 763 llvm::FixedVectorType::get(eltTy, candidateNumElts)); 764 numElts -= (numVecs << logCandidateNumElts); 765 766 if (numElts == 0) return; 767 768 // It's possible that the number of elements remaining will be legal. 769 // This can happen with e.g. <7 x float> when <3 x float> is legal. 770 // This only needs to be separately checked if it's not a power of 2. 771 if (numElts > 2 && !isPowerOf2(numElts) && 772 isLegalVectorType(CGM, eltSize * numElts, eltTy, numElts)) { 773 components.push_back(llvm::FixedVectorType::get(eltTy, numElts)); 774 return; 775 } 776 777 // Bring vecSize down to something no larger than numElts. 778 do { 779 logCandidateNumElts--; 780 candidateNumElts /= 2; 781 candidateSize /= 2; 782 } while (candidateNumElts > numElts); 783 } 784 785 // Otherwise, just append a bunch of individual elements. 786 components.append(numElts, eltTy); 787 } 788 789 bool swiftcall::mustPassRecordIndirectly(CodeGenModule &CGM, 790 const RecordDecl *record) { 791 // FIXME: should we not rely on the standard computation in Sema, just in 792 // case we want to diverge from the platform ABI (e.g. on targets where 793 // that uses the MSVC rule)? 794 return !record->canPassInRegisters(); 795 } 796 797 static ABIArgInfo classifyExpandedType(SwiftAggLowering &lowering, 798 bool forReturn, 799 CharUnits alignmentForIndirect) { 800 if (lowering.empty()) { 801 return ABIArgInfo::getIgnore(); 802 } else if (lowering.shouldPassIndirectly(forReturn)) { 803 return ABIArgInfo::getIndirect(alignmentForIndirect, /*byval*/ false); 804 } else { 805 auto types = lowering.getCoerceAndExpandTypes(); 806 return ABIArgInfo::getCoerceAndExpand(types.first, types.second); 807 } 808 } 809 810 static ABIArgInfo classifyType(CodeGenModule &CGM, CanQualType type, 811 bool forReturn) { 812 if (auto recordType = dyn_cast<RecordType>(type)) { 813 auto record = recordType->getDecl(); 814 auto &layout = CGM.getContext().getASTRecordLayout(record); 815 816 if (mustPassRecordIndirectly(CGM, record)) 817 return ABIArgInfo::getIndirect(layout.getAlignment(), /*byval*/ false); 818 819 SwiftAggLowering lowering(CGM); 820 lowering.addTypedData(recordType->getDecl(), CharUnits::Zero(), layout); 821 lowering.finish(); 822 823 return classifyExpandedType(lowering, forReturn, layout.getAlignment()); 824 } 825 826 // Just assume that all of our target ABIs can support returning at least 827 // two integer or floating-point values. 828 if (isa<ComplexType>(type)) { 829 return (forReturn ? ABIArgInfo::getDirect() : ABIArgInfo::getExpand()); 830 } 831 832 // Vector types may need to be legalized. 833 if (isa<VectorType>(type)) { 834 SwiftAggLowering lowering(CGM); 835 lowering.addTypedData(type, CharUnits::Zero()); 836 lowering.finish(); 837 838 CharUnits alignment = CGM.getContext().getTypeAlignInChars(type); 839 return classifyExpandedType(lowering, forReturn, alignment); 840 } 841 842 // Member pointer types need to be expanded, but it's a simple form of 843 // expansion that 'Direct' can handle. Note that CanBeFlattened should be 844 // true for this to work. 845 846 // 'void' needs to be ignored. 847 if (type->isVoidType()) { 848 return ABIArgInfo::getIgnore(); 849 } 850 851 // Everything else can be passed directly. 852 return ABIArgInfo::getDirect(); 853 } 854 855 ABIArgInfo swiftcall::classifyReturnType(CodeGenModule &CGM, CanQualType type) { 856 return classifyType(CGM, type, /*forReturn*/ true); 857 } 858 859 ABIArgInfo swiftcall::classifyArgumentType(CodeGenModule &CGM, 860 CanQualType type) { 861 return classifyType(CGM, type, /*forReturn*/ false); 862 } 863 864 void swiftcall::computeABIInfo(CodeGenModule &CGM, CGFunctionInfo &FI) { 865 auto &retInfo = FI.getReturnInfo(); 866 retInfo = classifyReturnType(CGM, FI.getReturnType()); 867 868 for (unsigned i = 0, e = FI.arg_size(); i != e; ++i) { 869 auto &argInfo = FI.arg_begin()[i]; 870 argInfo.info = classifyArgumentType(CGM, argInfo.type); 871 } 872 } 873 874 // Is swifterror lowered to a register by the target ABI. 875 bool swiftcall::isSwiftErrorLoweredInRegister(CodeGenModule &CGM) { 876 return getSwiftABIInfo(CGM).isSwiftErrorInRegister(); 877 } 878