[ARM64] [Windows] Follow MS X86_64 C++ ABI when passing structs

[ssijaric]

Summary

Microsoft's C++ object model for ARM64 is the same as that for X86_64. For example, small structs with non-trivial copy constructors or virtual function tables are passed indirectly. Currently, they are passed in registers when compiled with clang.

[mstorsjo]

This seems to be another case of all other arches being the same while x86_32 is different. The thumb case (with same actual code as x86_64) has a comment about a corner case where this doesn't match MSVC exactly, does that also apply here?

[ssijaric]

Yes, it looks like the comment for the thumb case is applicable here as well. VC++ on ARM64 passes small structs with a destructor and a default copy constructor in a register, while clang passes them on the stack. I will probably address this in a separate patch.

[ssijaric]

Updated to address Martin's observation. Verified that small structs with default copy constructors and non-trivial destructors are passed in registers on ARM64 Windows.

[mstorsjo]

LGTM. Should the same tweak also be done for arm/thumb?

[rnk]

Please avoid the extra enum case, and commit.

include/clang/Basic/TargetInfo.h

1223

This is poorly named. I'll try to send a patch to rename it after this lands.

1226–1227

IMO it'd be better to collapse these into something like CCK_MicrosoftWin64, since we don't treat them differently.

[ssijaric]

Last change before committing to address Reid's point of using CCK_MicrosoftWin64 for Windows on both X86_64 and ARM64.

include/clang/Basic/TargetInfo.h

virtual CallingConvCheckResult checkCallingConvention(CallingConv CC) const {
     switch (CC) {
       default:
         return CCCR_Warning;
       case CC_C:
         return CCCR_OK;
     }
   }
 
   enum CallingConvKind {
     CCK_Default,
     CCK_ClangABI4OrPS4,
-    CCK_MicrosoftX86_64
+    CCK_MicrosoftWin64
   };
 
   virtual CallingConvKind getCallingConvKind(bool ClangABICompat4) const;
 
   /// Controls if __builtin_longjmp / __builtin_setjmp can be lowered to
   /// llvm.eh.sjlj.longjmp / llvm.eh.sjlj.setjmp.
   virtual bool hasSjLjLowering() const {
     return false;
   }

lib/Basic/Targets/AArch64.h

 public:
   MicrosoftARM64TargetInfo(const llvm::Triple &Triple,
                            const TargetOptions &Opts);
 
   void getVisualStudioDefines(const LangOptions &Opts,
                               MacroBuilder &Builder) const;
   void getTargetDefines(const LangOptions &Opts,
                         MacroBuilder &Builder) const override;
+  TargetInfo::CallingConvKind
+  getCallingConvKind(bool ClangABICompat4) const override;
 };
 
 // ARM64 MinGW target
 class LLVM_LIBRARY_VISIBILITY MinGWARM64TargetInfo
     : public WindowsARM64TargetInfo {
 public:
   MinGWARM64TargetInfo(const llvm::Triple &Triple, const TargetOptions &Opts);
 };

lib/Basic/Targets/AArch64.cpp

 }
 
 void MicrosoftARM64TargetInfo::getTargetDefines(const LangOptions &Opts,
                                                 MacroBuilder &Builder) const {
   WindowsTargetInfo::getTargetDefines(Opts, Builder);
   getVisualStudioDefines(Opts, Builder);
 }
 
+TargetInfo::CallingConvKind
+MicrosoftARM64TargetInfo::getCallingConvKind(bool ClangABICompat4) const {
+  return CCK_MicrosoftWin64;
+}
+
 MinGWARM64TargetInfo::MinGWARM64TargetInfo(const llvm::Triple &Triple,
                                            const TargetOptions &Opts)
     : WindowsARM64TargetInfo(Triple, Opts) {
   TheCXXABI.set(TargetCXXABI::GenericAArch64);
 }
 
 DarwinAArch64TargetInfo::DarwinAArch64TargetInfo(const llvm::Triple &Triple,
                                                  const TargetOptions &Opts)

lib/Basic/Targets/X86.h

void getTargetDefines(const LangOptions &Opts,
     WindowsX86_64TargetInfo::getTargetDefines(Opts, Builder);
     WindowsX86_64TargetInfo::getVisualStudioDefines(Opts, Builder);
     Builder.defineMacro("_M_X64", "100");
     Builder.defineMacro("_M_AMD64", "100");
   }
 
   TargetInfo::CallingConvKind
   getCallingConvKind(bool ClangABICompat4) const override {
-    return CCK_MicrosoftX86_64;
+    return CCK_MicrosoftWin64;
   }
 };
 
 // x86-64 MinGW target
 class LLVM_LIBRARY_VISIBILITY MinGWX86_64TargetInfo
     : public WindowsX86_64TargetInfo {
 public:
   MinGWX86_64TargetInfo(const llvm::Triple &Triple, const TargetOptions &Opts)

lib/CodeGen/MicrosoftCXXABI.cpp

case llvm::Triple::x86:
     if (!canCopyArgument(RD))
       return RAA_DirectInMemory;
 
     // Otherwise, construct the argument into a temporary and copy the bytes
     // into the outgoing argument memory.
     return RAA_Default;
 
   case llvm::Triple::x86_64:
+  case llvm::Triple::aarch64:
     return !canCopyArgument(RD) ? RAA_Indirect : RAA_Default;
   }
 
   llvm_unreachable("invalid enum");
 }
 
 void MicrosoftCXXABI::emitVirtualObjectDelete(CodeGenFunction &CGF,
                                               const CXXDeleteExpr *DE,

lib/Sema/SemaDeclCXX.cpp

if (D->isDependentType() || D->isInvalidDecl())
     return false;
 
   // Clang <= 4 used the pre-C++11 rule, which ignores move operations.
   // The PS4 platform ABI follows the behavior of Clang 3.2.
   if (CCK == TargetInfo::CCK_ClangABI4OrPS4)
     return !D->hasNonTrivialDestructorForCall() &&
            !D->hasNonTrivialCopyConstructorForCall();
 
-  if (CCK == TargetInfo::CCK_MicrosoftX86_64) {
+  if (CCK == TargetInfo::CCK_MicrosoftWin64) {
     bool CopyCtorIsTrivial = false, CopyCtorIsTrivialForCall = false;
     bool DtorIsTrivialForCall = false;
 
     // If a class has at least one non-deleted, trivial copy constructor, it
     // is passed according to the C ABI. Otherwise, it is passed indirectly.
     //
     // Note: This permits classes with non-trivial copy or move ctors to be
     // passed in registers, so long as they *also* have a trivial copy ctor,

test/CodeGenCXX/microsoft-abi-sret-and-byval.cpp

 // RUN: %clang_cc1 -std=c++11 -emit-llvm %s -o - -triple=i386-pc-linux | FileCheck -check-prefix LINUX %s
 // RUN: %clang_cc1 -std=c++11 -emit-llvm %s -o - -triple=i386-pc-win32 -mconstructor-aliases -fno-rtti | FileCheck -check-prefix WIN32 %s
 // RUN: %clang_cc1 -std=c++11 -emit-llvm %s -o - -triple=thumb-pc-win32 -mconstructor-aliases -fno-rtti | FileCheck -check-prefix WOA %s
 // RUN: %clang_cc1 -std=c++11 -emit-llvm %s -o - -triple=x86_64-pc-win32 -mconstructor-aliases -fno-rtti | FileCheck -check-prefix WIN64 %s
+// RUN: %clang_cc1 -std=c++11 -emit-llvm %s -o - -triple=aarch64-windows-msvc -mconstructor-aliases -fno-rtti | FileCheck -check-prefix WOA64 %s
 
 struct Empty {};
 
 struct EmptyWithCtor {
   EmptyWithCtor() {}
 };
 
 struct Small {

 // Test that dtors are invoked in the callee.
 void small_arg_with_dtor(SmallWithDtor s) {}
 // WIN32: define dso_local void @"?small_arg_with_dtor@@YAXUSmallWithDtor@@@Z"(<{ %struct.SmallWithDtor }>* inalloca) {{.*}} {
 // WIN32:   call x86_thiscallcc void @"??1SmallWithDtor@@QAE@XZ"
 // WIN32: }
 // WIN64: define dso_local void @"?small_arg_with_dtor@@YAXUSmallWithDtor@@@Z"(i32 %s.coerce) {{.*}} {
 // WIN64:   call void @"??1SmallWithDtor@@QEAA@XZ"
 // WIN64: }
+// WOA64: define dso_local void @"?small_arg_with_dtor@@YAXUSmallWithDtor@@@Z"(i64 %s.coerce) {{.*}} {
+// WOA64:   call void @"??1SmallWithDtor@@QEAA@XZ"
+// WOA64: }
 
 // FIXME: MSVC incompatible!
 // WOA: define dso_local arm_aapcs_vfpcc void @"?small_arg_with_dtor@@YAXUSmallWithDtor@@@Z"(%struct.SmallWithDtor* %s) {{.*}} {
 // WOA:   call arm_aapcs_vfpcc void @"??1SmallWithDtor@@QAA@XZ"(%struct.SmallWithDtor* %s)
 // WOA: }
 
 void call_small_arg_with_dtor() {
   small_arg_with_dtor(SmallWithDtor());

 // WIN32:   call void @"?takes_two_by_val_with_dtor@@YAXUSmallWithDtor@@0@Z"
 // WIN32-NOT: call x86_thiscallcc void @"??1SmallWithDtor@@QAE@XZ"
 // WIN32: }
 
 void small_arg_with_vftable(SmallWithVftable s) {}
 // LINUX-LABEL: define void @_Z22small_arg_with_vftable16SmallWithVftable(%struct.SmallWithVftable* %s)
 // WIN32: define dso_local void @"?small_arg_with_vftable@@YAXUSmallWithVftable@@@Z"(<{ %struct.SmallWithVftable }>* inalloca)
 // WIN64: define dso_local void @"?small_arg_with_vftable@@YAXUSmallWithVftable@@@Z"(%struct.SmallWithVftable* %s)
+// WOA64: define dso_local void @"?small_arg_with_vftable@@YAXUSmallWithVftable@@@Z"(%struct.SmallWithVftable* %s)
 
 void medium_arg_with_copy_ctor(MediumWithCopyCtor s) {}
 // LINUX-LABEL: define void @_Z25medium_arg_with_copy_ctor18MediumWithCopyCtor(%struct.MediumWithCopyCtor* %s)
 // WIN32: define dso_local void @"?medium_arg_with_copy_ctor@@YAXUMediumWithCopyCtor@@@Z"(<{ %struct.MediumWithCopyCtor }>* inalloca)
 // WIN64: define dso_local void @"?medium_arg_with_copy_ctor@@YAXUMediumWithCopyCtor@@@Z"(%struct.MediumWithCopyCtor* %s)
 // WOA: define dso_local arm_aapcs_vfpcc void @"?medium_arg_with_copy_ctor@@YAXUMediumWithCopyCtor@@@Z"(%struct.MediumWithCopyCtor* %s)
+// WOA64: define dso_local void @"?medium_arg_with_copy_ctor@@YAXUMediumWithCopyCtor@@@Z"(%struct.MediumWithCopyCtor* %s)
 
 void big_arg(Big s) {}
 // LINUX-LABEL: define void @_Z7big_arg3Big(%struct.Big* byval align 4 %s)
 // WIN32: define dso_local void @"?big_arg@@YAXUBig@@@Z"(%struct.Big* byval align 4 %s)
 // WIN64: define dso_local void @"?big_arg@@YAXUBig@@@Z"(%struct.Big* %s)
 
 // PR27607: We would attempt to load i32 value out of the reference instead of
 // just loading the pointer from the struct during argument expansion.