AArch64: add arm64_32 support to Clang.

11 files changed, 90 insertions, 29 deletions

diff --git a/clang/lib/Basic/Targets.cpp b/clang/lib/Basic/Targets.cpp
index 63a64ed2931..664260d184f 100644
--- a/clang/lib/Basic/Targets.cpp
+++ b/clang/lib/Basic/Targets.cpp
@@ -122,6 +122,11 @@ TargetInfo *AllocateTarget(const llvm::Triple &Triple,
   case llvm::Triple::lanai:
     return new LanaiTargetInfo(Triple, Opts);
 
+  case llvm::Triple::aarch64_32:
+    if (Triple.isOSDarwin())
+      return new DarwinAArch64TargetInfo(Triple, Opts);
+
+    return nullptr;
   case llvm::Triple::aarch64:
     if (Triple.isOSDarwin())
       return new DarwinAArch64TargetInfo(Triple, Opts);
diff --git a/clang/lib/Basic/Targets/AArch64.cpp b/clang/lib/Basic/Targets/AArch64.cpp
index c86cc63e3d8..bdfb5700b46 100644
--- a/clang/lib/Basic/Targets/AArch64.cpp
+++ b/clang/lib/Basic/Targets/AArch64.cpp
@@ -51,7 +51,11 @@ AArch64TargetInfo::AArch64TargetInfo(const llvm::Triple &Triple,
   HasLegalHalfType = true;
   HasFloat16 = true;
 
-  LongWidth = LongAlign = PointerWidth = PointerAlign = 64;
+  if (Triple.isArch64Bit())
+    LongWidth = LongAlign = PointerWidth = PointerAlign = 64;
+  else
+    LongWidth = LongAlign = PointerWidth = PointerAlign = 32;
+
   MaxVectorAlign = 128;
   MaxAtomicInlineWidth = 128;
   MaxAtomicPromoteWidth = 128;
@@ -160,7 +164,7 @@ void AArch64TargetInfo::getTargetDefines(const LangOptions &Opts,
     Builder.defineMacro("__ELF__");
 
   // Target properties.
-  if (!getTriple().isOSWindows()) {
+  if (!getTriple().isOSWindows() && getTriple().isArch64Bit()) {
     Builder.defineMacro("_LP64");
     Builder.defineMacro("__LP64__");
   }
@@ -506,14 +510,19 @@ int AArch64TargetInfo::getEHDataRegisterNumber(unsigned RegNo) const {
   return -1;
 }
 
+bool AArch64TargetInfo::hasInt128Type() const { return true; }
+
 AArch64leTargetInfo::AArch64leTargetInfo(const llvm::Triple &Triple,
                                          const TargetOptions &Opts)
     : AArch64TargetInfo(Triple, Opts) {}
 
 void AArch64leTargetInfo::setDataLayout() {
-  if (getTriple().isOSBinFormatMachO())
-    resetDataLayout("e-m:o-i64:64-i128:128-n32:64-S128");
-  else
+  if (getTriple().isOSBinFormatMachO()) {
+    if(getTriple().isArch32Bit())
+      resetDataLayout("e-m:o-p:32:32-i64:64-i128:128-n32:64-S128");
+    else
+      resetDataLayout("e-m:o-i64:64-i128:128-n32:64-S128");
+  } else
     resetDataLayout("e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128");
 }
 
@@ -631,19 +640,34 @@ DarwinAArch64TargetInfo::DarwinAArch64TargetInfo(const llvm::Triple &Triple,
                                                  const TargetOptions &Opts)
     : DarwinTargetInfo<AArch64leTargetInfo>(Triple, Opts) {
   Int64Type = SignedLongLong;
+  if (getTriple().isArch32Bit())
+    IntMaxType = SignedLongLong;
+
+  WCharType = SignedInt;
   UseSignedCharForObjCBool = false;
 
   LongDoubleWidth = LongDoubleAlign = SuitableAlign = 64;
   LongDoubleFormat = &llvm::APFloat::IEEEdouble();
 
-  TheCXXABI.set(TargetCXXABI::iOS64);
+  UseZeroLengthBitfieldAlignment = false;
+
+  if (getTriple().isArch32Bit()) {
+    UseBitFieldTypeAlignment = false;
+    ZeroLengthBitfieldBoundary = 32;
+    UseZeroLengthBitfieldAlignment = true;
+    TheCXXABI.set(TargetCXXABI::WatchOS);
+  } else
+    TheCXXABI.set(TargetCXXABI::iOS64);
 }
 
 void DarwinAArch64TargetInfo::getOSDefines(const LangOptions &Opts,
                                            const llvm::Triple &Triple,
                                            MacroBuilder &Builder) const {
   Builder.defineMacro("__AARCH64_SIMD__");
-  Builder.defineMacro("__ARM64_ARCH_8__");
+  if (Triple.isArch32Bit())
+    Builder.defineMacro("__ARM64_ARCH_8_32__");
+  else
+    Builder.defineMacro("__ARM64_ARCH_8__");
   Builder.defineMacro("__ARM_NEON__");
   Builder.defineMacro("__LITTLE_ENDIAN__");
   Builder.defineMacro("__REGISTER_PREFIX__", "");
diff --git a/clang/lib/Basic/Targets/AArch64.h b/clang/lib/Basic/Targets/AArch64.h
index b6aa07780ed..7062ea1ae50 100644
--- a/clang/lib/Basic/Targets/AArch64.h
+++ b/clang/lib/Basic/Targets/AArch64.h
@@ -97,6 +97,8 @@ public:
   }
 
   int getEHDataRegisterNumber(unsigned RegNo) const override;
+
+  bool hasInt128Type() const override;
 };
 
 class LLVM_LIBRARY_VISIBILITY AArch64leTargetInfo : public AArch64TargetInfo {
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index cc2cbb90707..0a16636c98f 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -4238,6 +4238,7 @@ static Value *EmitTargetArchBuiltinExpr(CodeGenFunction *CGF,
   case llvm::Triple::thumbeb:
     return CGF->EmitARMBuiltinExpr(BuiltinID, E, ReturnValue, Arch);
   case llvm::Triple::aarch64:
+  case llvm::Triple::aarch64_32:
   case llvm::Triple::aarch64_be:
     return CGF->EmitAArch64BuiltinExpr(BuiltinID, E, Arch);
   case llvm::Triple::bpfeb:
@@ -5670,7 +5671,8 @@ Value *CodeGenFunction::EmitCommonNeonBuiltinExpr(
     llvm::Type *PTy = llvm::PointerType::getUnqual(VTy->getVectorElementType());
     // TODO: Currently in AArch32 mode the pointer operand comes first, whereas
     // in AArch64 it comes last. We may want to stick to one or another.
-    if (Arch == llvm::Triple::aarch64 || Arch == llvm::Triple::aarch64_be) {
+    if (Arch == llvm::Triple::aarch64 || Arch == llvm::Triple::aarch64_be ||
+        Arch == llvm::Triple::aarch64_32) {
       llvm::Type *Tys[2] = { VTy, PTy };
       std::rotate(Ops.begin(), Ops.begin() + 1, Ops.end());
       return EmitNeonCall(CGM.getIntrinsic(LLVMIntrinsic, Tys), Ops, "");
diff --git a/clang/lib/CodeGen/TargetInfo.cpp b/clang/lib/CodeGen/TargetInfo.cpp
index 26c527d7c98..b139f8a7d43 100644
--- a/clang/lib/CodeGen/TargetInfo.cpp
+++ b/clang/lib/CodeGen/TargetInfo.cpp
@@ -4991,7 +4991,7 @@ private:
   ABIKind getABIKind() const { return Kind; }
   bool isDarwinPCS() const { return Kind == DarwinPCS; }
 
-  ABIArgInfo classifyReturnType(QualType RetTy) const;
+  ABIArgInfo classifyReturnType(QualType RetTy, bool IsVariadic) const;
   ABIArgInfo classifyArgumentType(QualType RetTy) const;
   bool isHomogeneousAggregateBaseType(QualType Ty) const override;
   bool isHomogeneousAggregateSmallEnough(const Type *Ty,
@@ -5001,7 +5001,8 @@ private:
 
   void computeInfo(CGFunctionInfo &FI) const override {
     if (!::classifyReturnType(getCXXABI(), FI, *this))
-      FI.getReturnInfo() = classifyReturnType(FI.getReturnType());
+      FI.getReturnInfo() =
+          classifyReturnType(FI.getReturnType(), FI.isVariadic());
 
     for (auto &it : FI.arguments())
       it.info = classifyArgumentType(it.type);
@@ -5184,23 +5185,24 @@ ABIArgInfo AArch64ABIInfo::classifyArgumentType(QualType Ty) const {
       Alignment = getContext().getTypeUnadjustedAlign(Ty);
       Alignment = Alignment < 128 ? 64 : 128;
     } else {
-      Alignment = getContext().getTypeAlign(Ty);
+      Alignment = std::max(getContext().getTypeAlign(Ty),
+                           (unsigned)getTarget().getPointerWidth(0));
     }
-    Size = llvm::alignTo(Size, 64); // round up to multiple of 8 bytes
+    Size = llvm::alignTo(Size, Alignment);
 
     // We use a pair of i64 for 16-byte aggregate with 8-byte alignment.
     // For aggregates with 16-byte alignment, we use i128.
-    if (Alignment < 128 && Size == 128) {
-      llvm::Type *BaseTy = llvm::Type::getInt64Ty(getVMContext());
-      return ABIArgInfo::getDirect(llvm::ArrayType::get(BaseTy, Size / 64));
-    }
-    return ABIArgInfo::getDirect(llvm::IntegerType::get(getVMContext(), Size));
+    llvm::Type *BaseTy = llvm::Type::getIntNTy(getVMContext(), Alignment);
+    return ABIArgInfo::getDirect(
+        Size == Alignment ? BaseTy
+                          : llvm::ArrayType::get(BaseTy, Size / Alignment));
   }
 
   return getNaturalAlignIndirect(Ty, /*ByVal=*/false);
 }
 
-ABIArgInfo AArch64ABIInfo::classifyReturnType(QualType RetTy) const {
+ABIArgInfo AArch64ABIInfo::classifyReturnType(QualType RetTy,
+                                              bool IsVariadic) const {
   if (RetTy->isVoidType())
     return ABIArgInfo::getIgnore();
 
@@ -5224,7 +5226,9 @@ ABIArgInfo AArch64ABIInfo::classifyReturnType(QualType RetTy) const {
 
   const Type *Base = nullptr;
   uint64_t Members = 0;
-  if (isHomogeneousAggregate(RetTy, Base, Members))
+  if (isHomogeneousAggregate(RetTy, Base, Members) &&
+      !(getTarget().getTriple().getArch() == llvm::Triple::aarch64_32 &&
+        IsVariadic))
     // Homogeneous Floating-point Aggregates (HFAs) are returned directly.
     return ABIArgInfo::getDirect();
 
@@ -5259,6 +5263,14 @@ bool AArch64ABIInfo::isIllegalVectorType(QualType Ty) const {
     // NumElements should be power of 2.
     if (!llvm::isPowerOf2_32(NumElements))
       return true;
+
+    // arm64_32 has to be compatible with the ARM logic here, which allows huge
+    // vectors for some reason.
+    llvm::Triple Triple = getTarget().getTriple();
+    if (Triple.getArch() == llvm::Triple::aarch64_32 &&
+        Triple.isOSBinFormatMachO())
+      return Size <= 32;
+
     return Size != 64 && (Size != 128 || NumElements == 1);
   }
   return false;
@@ -5550,7 +5562,8 @@ Address AArch64ABIInfo::EmitDarwinVAArg(Address VAListAddr, QualType Ty,
   if (!isAggregateTypeForABI(Ty) && !isIllegalVectorType(Ty))
     return EmitVAArgInstr(CGF, VAListAddr, Ty, ABIArgInfo::getDirect());
 
-  CharUnits SlotSize = CharUnits::fromQuantity(8);
+  uint64_t PointerSize = getTarget().getPointerWidth(0) / 8;
+  CharUnits SlotSize = CharUnits::fromQuantity(PointerSize);
 
   // Empty records are ignored for parameter passing purposes.
   if (isEmptyRecord(getContext(), Ty, true)) {
@@ -9773,6 +9786,7 @@ const TargetCodeGenInfo &CodeGenModule::getTargetCodeGenInfo() {
     return SetCGInfo(new AVRTargetCodeGenInfo(Types));
 
   case llvm::Triple::aarch64:
+  case llvm::Triple::aarch64_32:
   case llvm::Triple::aarch64_be: {
     AArch64ABIInfo::ABIKind Kind = AArch64ABIInfo::AAPCS;
     if (getTarget().getABI() == "darwinpcs")
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index a014c611ee2..ba4128e0622 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -637,6 +637,8 @@ std::string ToolChain::ComputeLLVMTriple(const ArgList &Args,
     Triple.setArchName("arm64");
     return Triple.getTriple();
   }
+  case llvm::Triple::aarch64_32:
+    return getTripleString();
   case llvm::Triple::arm:
   case llvm::Triple::armeb:
   case llvm::Triple::thumb:
diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
index 04589d65e6d..2ee649f200a 100644
--- a/clang/lib/Driver/ToolChains/Clang.cpp
+++ b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -342,6 +342,7 @@ static void getTargetFeatures(const ToolChain &TC, const llvm::Triple &Triple,
     systemz::getSystemZTargetFeatures(Args, Features);
     break;
   case llvm::Triple::aarch64:
+  case llvm::Triple::aarch64_32:
   case llvm::Triple::aarch64_be:
     aarch64::getAArch64TargetFeatures(D, Triple, Args, Features);
     break;
@@ -1351,6 +1352,7 @@ static bool isSignedCharDefault(const llvm::Triple &Triple) {
     return true;
 
   case llvm::Triple::aarch64:
+  case llvm::Triple::aarch64_32:
   case llvm::Triple::aarch64_be:
   case llvm::Triple::arm:
   case llvm::Triple::armeb:
@@ -1473,6 +1475,7 @@ void Clang::RenderTargetOptions(const llvm::Triple &EffectiveTriple,
     break;
 
   case llvm::Triple::aarch64:
+  case llvm::Triple::aarch64_32:
   case llvm::Triple::aarch64_be:
     AddAArch64TargetArgs(Args, CmdArgs);
     CmdArgs.push_back("-fallow-half-arguments-and-returns");
@@ -4032,6 +4035,7 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
       RenderARMABI(Triple, Args, CmdArgs);
       break;
     case llvm::Triple::aarch64:
+    case llvm::Triple::aarch64_32:
     case llvm::Triple::aarch64_be:
       RenderAArch64ABI(Triple, Args, CmdArgs);
       break;
@@ -5789,11 +5793,12 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
       // We only support -moutline in AArch64 right now. If we're not compiling
       // for AArch64, emit a warning and ignore the flag. Otherwise, add the
       // proper mllvm flags.
-      if (Triple.getArch() != llvm::Triple::aarch64) {
+      if (Triple.getArch() != llvm::Triple::aarch64 &&
+          Triple.getArch() != llvm::Triple::aarch64_32) {
         D.Diag(diag::warn_drv_moutline_unsupported_opt) << Triple.getArchName();
       } else {
-          CmdArgs.push_back("-mllvm");
-          CmdArgs.push_back("-enable-machine-outliner");
+        CmdArgs.push_back("-mllvm");
+        CmdArgs.push_back("-enable-machine-outliner");
       }
     } else {
       // Disable all outlining behaviour.
diff --git a/clang/lib/Driver/ToolChains/CommonArgs.cpp b/clang/lib/Driver/ToolChains/CommonArgs.cpp
index 615e96d7ec4..58705a2aa93 100644
--- a/clang/lib/Driver/ToolChains/CommonArgs.cpp
+++ b/clang/lib/Driver/ToolChains/CommonArgs.cpp
@@ -253,6 +253,7 @@ std::string tools::getCPUName(const ArgList &Args, const llvm::Triple &T,
     return "";
 
   case llvm::Triple::aarch64:
+  case llvm::Triple::aarch64_32:
   case llvm::Triple::aarch64_be:
     return aarch64::getAArch64TargetCPU(Args, T, A);
 
diff --git a/clang/lib/Driver/ToolChains/Darwin.cpp b/clang/lib/Driver/ToolChains/Darwin.cpp
index b47150d1cee..75a47d2c3a2 100644
--- a/clang/lib/Driver/ToolChains/Darwin.cpp
+++ b/clang/lib/Driver/ToolChains/Darwin.cpp
@@ -58,6 +58,7 @@ llvm::Triple::ArchType darwin::getArchTypeForMachOArchName(StringRef Str) {
       .Cases("armv7", "armv7em", "armv7k", "armv7m", llvm::Triple::arm)
       .Cases("armv7s", "xscale", llvm::Triple::arm)
       .Case("arm64", llvm::Triple::aarch64)
+      .Case("arm64_32", llvm::Triple::aarch64_32)
       .Case("r600", llvm::Triple::r600)
       .Case("amdgcn", llvm::Triple::amdgcn)
       .Case("nvptx", llvm::Triple::nvptx)
@@ -832,6 +833,9 @@ StringRef MachO::getMachOArchName(const ArgList &Args) const {
   default:
     return getDefaultUniversalArchName();
 
+  case llvm::Triple::aarch64_32:
+    return "arm64_32";
+
   case llvm::Triple::aarch64:
     return "arm64";
 
@@ -1640,7 +1644,7 @@ inferDeploymentTargetFromArch(DerivedArgList &Args, const Darwin &Toolchain,
   if (MachOArchName == "armv7" || MachOArchName == "armv7s" ||
       MachOArchName == "arm64")
     OSTy = llvm::Triple::IOS;
-  else if (MachOArchName == "armv7k")
+  else if (MachOArchName == "armv7k" || MachOArchName == "arm64_32")
     OSTy = llvm::Triple::WatchOS;
   else if (MachOArchName != "armv6m" && MachOArchName != "armv7m" &&
            MachOArchName != "armv7em")
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
index 8322a9bf147..806df77cb27 100644
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -1536,6 +1536,7 @@ Sema::CheckBuiltinFunctionCall(FunctionDecl *FDecl, unsigned BuiltinID,
           return ExprError();
         break;
       case llvm::Triple::aarch64:
+      case llvm::Triple::aarch64_32:
       case llvm::Triple::aarch64_be:
         if (CheckAArch64BuiltinFunctionCall(BuiltinID, TheCall))
           return ExprError();
@@ -1685,6 +1686,7 @@ bool Sema::CheckNeonBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
 
     llvm::Triple::ArchType Arch = Context.getTargetInfo().getTriple().getArch();
     bool IsPolyUnsigned = Arch == llvm::Triple::aarch64 ||
+                          Arch == llvm::Triple::aarch64_32 ||
                           Arch == llvm::Triple::aarch64_be;
     bool IsInt64Long =
         Context.getTargetInfo().getInt64Type() == TargetInfo::SignedLong;
@@ -5516,7 +5518,8 @@ ExprResult Sema::CheckOSLogFormatStringArg(Expr *Arg) {
 static bool checkVAStartABI(Sema &S, unsigned BuiltinID, Expr *Fn) {
   const llvm::Triple &TT = S.Context.getTargetInfo().getTriple();
   bool IsX64 = TT.getArch() == llvm::Triple::x86_64;
-  bool IsAArch64 = TT.getArch() == llvm::Triple::aarch64;
+  bool IsAArch64 = (TT.getArch() == llvm::Triple::aarch64 ||
+                    TT.getArch() == llvm::Triple::aarch64_32);
   bool IsWindows = TT.isOSWindows();
   bool IsMSVAStart = BuiltinID == Builtin::BI__builtin_ms_va_start;
   if (IsX64 || IsAArch64) {
diff --git a/clang/lib/Sema/SemaType.cpp b/clang/lib/Sema/SemaType.cpp
index 91945e23515..abee6e68c0e 100644
--- a/clang/lib/Sema/SemaType.cpp
+++ b/clang/lib/Sema/SemaType.cpp
@@ -7215,6 +7215,7 @@ static bool isPermittedNeonBaseType(QualType &Ty,
   // Signed poly is mathematically wrong, but has been baked into some ABIs by
   // now.
   bool IsPolyUnsigned = Triple.getArch() == llvm::Triple::aarch64 ||
+                        Triple.getArch() == llvm::Triple::aarch64_32 ||
                         Triple.getArch() == llvm::Triple::aarch64_be;
   if (VecKind == VectorType::NeonPolyVector) {
     if (IsPolyUnsigned) {
@@ -7232,10 +7233,8 @@ static bool isPermittedNeonBaseType(QualType &Ty,
 
   // Non-polynomial vector types: the usual suspects are allowed, as well as
   // float64_t on AArch64.
-  bool Is64Bit = Triple.getArch() == llvm::Triple::aarch64 ||
-                 Triple.getArch() == llvm::Triple::aarch64_be;
-
-  if (Is64Bit && BTy->getKind() == BuiltinType::Double)
+  if ((Triple.isArch64Bit() || Triple.getArch() == llvm::Triple::aarch64_32) &&
+      BTy->getKind() == BuiltinType::Double)
     return true;
 
   return BTy->getKind() == BuiltinType::SChar ||