[Triple][MachO] Define "arm64e", an AArch64 subarch for Pointer Auth.

This also teaches MachO writers/readers about the MachO cpu subtype,
beyond the minimal subtype reader support present at the moment.

This also defines a preprocessor macro to allow users to distinguish
__arm64__ from __arm64e__.

arm64e defaults to an "apple-a12" CPU, which supports v8.3a, allowing
pointer-authentication codegen.
It also currently defaults to ios14 and macos11.

Differential Revision: https://reviews.llvm.org/D87095

include/llvm/ADT/Triple.h

@@ -129,6 +129,8 @@ public:
     ARMSubArch_v5te,
     ARMSubArch_v4t,
 
+    AArch64SubArch_arm64e,
+
     KalimbaSubArch_v3,
     KalimbaSubArch_v4,
     KalimbaSubArch_v5,
@@ -777,6 +779,12 @@ public:
     return getArch() == Triple::csky;
   }
 
+  /// Tests whether the target is the Apple "arm64e" AArch64 subarch.
+  bool isArm64e() const {
+    return getArch() == Triple::aarch64 &&
+           getSubArch() == Triple::AArch64SubArch_arm64e;
+  }
+
   /// Tests whether the target supports comdat
   bool supportsCOMDAT() const {
     return !(isOSBinFormatMachO() || isOSBinFormatXCOFF());

lib/BinaryFormat/MachO.cpp

@@ -58,7 +58,7 @@ static MachO::CPUSubTypeARM64 getARM64SubType(const Triple &T) {
   assert(T.isAArch64());
   if (T.isArch32Bit())
     return (MachO::CPUSubTypeARM64)MachO::CPU_SUBTYPE_ARM64_32_V8;
-  if (T.getArchName() == "arm64e")
+  if (T.isArm64e())
     return MachO::CPU_SUBTYPE_ARM64E;
 
   return MachO::CPU_SUBTYPE_ARM64_ALL;

lib/LTO/LTOCodeGenerator.cpp

@@ -377,6 +377,8 @@ bool LTOCodeGenerator::determineTarget() {
       MCpu = "core2";
     else if (Triple.getArch() == llvm::Triple::x86)
       MCpu = "yonah";
+    else if (Triple.isArm64e())
+      MCpu = "apple-a12";
     else if (Triple.getArch() == llvm::Triple::aarch64 ||
              Triple.getArch() == llvm::Triple::aarch64_32)
       MCpu = "cyclone";

lib/LTO/LTOModule.cpp

@@ -222,6 +222,8 @@ LTOModule::makeLTOModule(MemoryBufferRef Buffer, const TargetOptions &options,
       CPU = "core2";
     else if (Triple.getArch() == llvm::Triple::x86)
       CPU = "yonah";
+    else if (Triple.isArm64e())
+      CPU = "apple-a12";
     else if (Triple.getArch() == llvm::Triple::aarch64 ||
              Triple.getArch() == llvm::Triple::aarch64_32)
       CPU = "cyclone";

lib/Object/MachOObjectFile.cpp

@@ -2694,6 +2694,12 @@ Triple MachOObjectFile::getArchTriple(uint32_t CPUType, uint32_t CPUSubType,
       if (ArchFlag)
         *ArchFlag = "arm64";
       return Triple("arm64-apple-darwin");
+    case MachO::CPU_SUBTYPE_ARM64E:
+      if (McpuDefault)
+        *McpuDefault = "apple-a12";
+      if (ArchFlag)
+        *ArchFlag = "arm64e";
+      return Triple("arm64e-apple-darwin");
     default:
       return Triple();
     }
@@ -2741,13 +2747,28 @@ bool MachOObjectFile::isValidArch(StringRef ArchFlag) {
 }
 
 ArrayRef<StringRef> MachOObjectFile::getValidArchs() {
-  static const std::array<StringRef, 17> validArchs = {{
+  static const std::array<StringRef, 18> ValidArchs = {{
-      "i386",   "x86_64", "x86_64h",  "armv4t",  "arm",    "armv5e",
+      "i386",
-      "armv6",  "armv6m", "armv7",    "armv7em", "armv7k", "armv7m",
+      "x86_64",
-      "armv7s", "arm64",  "arm64_32", "ppc",     "ppc64",
+      "x86_64h",
+      "armv4t",
+      "arm",
+      "armv5e",
+      "armv6",
+      "armv6m",
+      "armv7",
+      "armv7em",
+      "armv7k",
+      "armv7m",
+      "armv7s",
+      "arm64",
+      "arm64e",
+      "arm64_32",
+      "ppc",
+      "ppc64",
   }};
 
-  return validArchs;
+  return ValidArchs;
 }
 
 Triple::ArchType MachOObjectFile::getArch() const {

lib/Support/ARMTargetParser.cpp

@@ -280,6 +280,8 @@ StringRef ARM::getCanonicalArchName(StringRef Arch) {
   // Begins with "arm" / "thumb", move past it.
   if (A.startswith("arm64_32"))
     offset = 8;
+  else if (A.startswith("arm64e"))
+    offset = 6;
   else if (A.startswith("arm64"))
     offset = 5;
   else if (A.startswith("aarch64_32"))

lib/Support/Triple.cpp

@@ -407,6 +407,7 @@ static Triple::ArchType parseArch(StringRef ArchName) {
     .Case("arc", Triple::arc)
     .Case("arm64", Triple::aarch64)
     .Case("arm64_32", Triple::aarch64_32)
+    .Case("arm64e", Triple::aarch64)
     .Case("arm", Triple::arm)
     .Case("armeb", Triple::armeb)
     .Case("thumb", Triple::thumb)
@@ -572,6 +573,9 @@ static Triple::SubArchType parseSubArch(StringRef SubArchName) {
   if (SubArchName == "powerpcspe")
     return Triple::PPCSubArch_spe;
 
+  if (SubArchName == "arm64e")
+    return Triple::AArch64SubArch_arm64e;
+
   StringRef ARMSubArch = ARM::getCanonicalArchName(SubArchName);
 
   // For now, this is the small part. Early return.
@@ -1648,6 +1652,9 @@ VersionTuple Triple::getMinimumSupportedOSVersion() const {
     // ARM64 simulators are supported for iOS 14+.
     if (isMacCatalystEnvironment() || isSimulatorEnvironment())
       return VersionTuple(14, 0, 0);
+    // ARM64e slice is supported starting from iOS 14.
+    if (isArm64e())
+      return VersionTuple(14, 0, 0);
     break;
   case Triple::TvOS:
     // ARM64 simulators are supported for tvOS 14+.

lib/Target/AArch64/AArch64TargetMachine.cpp

@@ -229,6 +229,12 @@ static std::string computeDataLayout(const Triple &TT,
   return "E-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128";
 }
 
+static StringRef computeDefaultCPU(const Triple &TT, StringRef CPU) {
+  if (CPU.empty() && TT.isArm64e())
+    return "apple-a12";
+  return CPU;
+}
+
 static Reloc::Model getEffectiveRelocModel(const Triple &TT,
                                            Optional<Reloc::Model> RM) {
   // AArch64 Darwin and Windows are always PIC.
@@ -276,7 +282,8 @@ AArch64TargetMachine::AArch64TargetMachine(const Target &T, const Triple &TT,
                                            bool LittleEndian)
     : LLVMTargetMachine(T,
                         computeDataLayout(TT, Options.MCOptions, LittleEndian),
-                        TT, CPU, FS, Options, getEffectiveRelocModel(TT, RM),
+                        TT, computeDefaultCPU(TT, CPU), FS, Options,
+                        getEffectiveRelocModel(TT, RM),
                         getEffectiveAArch64CodeModel(TT, CM, JIT), OL),
       TLOF(createTLOF(getTargetTriple())), isLittle(LittleEndian) {
   initAsmInfo();

lib/Target/AArch64/MCTargetDesc/AArch64MCTargetDesc.cpp

@@ -50,9 +50,13 @@ static MCInstrInfo *createAArch64MCInstrInfo() {
 
 static MCSubtargetInfo *
 createAArch64MCSubtargetInfo(const Triple &TT, StringRef CPU, StringRef FS) {
-  if (CPU.empty())
+  if (CPU.empty()) {
     CPU = "generic";
 
+    if (TT.isArm64e())
+      CPU = "apple-a12";
+  }
+
   return createAArch64MCSubtargetInfoImpl(TT, CPU, /*TuneCPU*/ CPU, FS);
 }
 

test/MC/AArch64/arm64e-subtype.s

@@ -0,0 +1,12 @@
+; RUN: llvm-mc -triple=arm64e-apple-ios -filetype=obj %s -o - | llvm-objdump --macho -d -p - | FileCheck %s
+
+; CHECK: _foo:
+; CHECK: 0: c0 03 5f d6   ret
+
+; CHECK: Mach header
+; CHECK: magic cputype cpusubtype  caps    filetype ncmds sizeofcmds      flags
+; CHECK: MH_MAGIC_64   ARM64          E  0x00      OBJECT     3        256 0x00000000
+
+.globl _foo
+_foo:
+  ret

test/MC/AArch64/arm64e.s

@@ -0,0 +1,9 @@
+// RUN: not llvm-mc -triple arm64-- -show-encoding < %s 2> %t
+// RUN: FileCheck --check-prefix=CHECK-GENERIC < %t %s
+
+// RUN: llvm-mc -triple arm64e-- -show-encoding < %s |\
+// RUN: FileCheck %s --check-prefix=CHECK-ARM64E
+
+// CHECK-GENERIC:  error: instruction requires: pa
+// CHECK-ARM64E:  pacia x0, x1 // encoding: [0x20,0x00,0xc1,0xda]
+  pacia x0, x1

test/MC/MachO/AArch64/arm-darwin-version-min-load-command.s

@@ -1,16 +1,39 @@
 // RUN: llvm-mc -triple arm64-apple-macos10.10.2 %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-MACOS-ARM64
+// RUN: llvm-mc -triple arm64e-apple-macos10.10 %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-MACOS-ARM64
 // RUN: llvm-mc -triple arm64-apple-macos11 %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-MACOS-ARM64
 // RUN: llvm-mc -triple arm64-apple-macos11.1 %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-MACOS-ARM64_1
 // RUN: llvm-mc -triple arm64-apple-ios13.0-macabi %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-MACCATALYST-ARM64
+// RUN: llvm-mc -triple arm64e-apple-ios13.0-macabi %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-MACCATALYST-ARM64
 // RUN: llvm-mc -triple arm64-apple-ios14.1-macabi %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-MACCATALYST-ARM64_1
 
+// RUN: llvm-mc -triple arm64e-apple-ios10.3 %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-IOS-ARM64E
+// RUN: llvm-mc -triple arm64e-apple-ios13 %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-IOS-ARM64E
+// RUN: llvm-mc -triple arm64e-apple-ios14.1 %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-IOS-ARM64E3
+
 // RUN: llvm-mc -triple arm64-apple-ios10-simulator %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-IOSSIM2
+// RUN: llvm-mc -triple arm64e-apple-ios10-simulator %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-IOSSIM2
 // RUN: llvm-mc -triple arm64-apple-ios13-simulator %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-IOSSIM2
 // RUN: llvm-mc -triple arm64-apple-ios14-simulator %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-IOSSIM2
 // RUN: llvm-mc -triple arm64-apple-ios14.1-simulator %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-IOSSIM3
 // RUN: llvm-mc -triple arm64-apple-tvos10-simulator %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-TVOSSIM2
 // RUN: llvm-mc -triple arm64-apple-watchos3-simulator %s -filetype=obj -o - | llvm-objdump --macho --private-headers - | FileCheck %s --check-prefix=CHECK-BUILD-WATCHOSSIM2
 
+// CHECK-BUILD-IOS-ARM64E:           cmd LC_BUILD_VERSION
+// CHECK-BUILD-IOS-ARM64E-NEXT:  cmdsize 24
+// CHECK-BUILD-IOS-ARM64E-NEXT: platform ios
+// CHECK-BUILD-IOS-ARM64E-NEXT:      sdk n/a
+// CHECK-BUILD-IOS-ARM64E-NEXT:    minos 14.0
+// CHECK-BUILD-IOS-ARM64E-NEXT:   ntools 0
+// CHECK-BUILD-IOS-ARM64E-NOT: LC_VERSION_MIN
+
+// CHECK-BUILD-IOS-ARM64E3:           cmd LC_BUILD_VERSION
+// CHECK-BUILD-IOS-ARM64E3-NEXT:  cmdsize 24
+// CHECK-BUILD-IOS-ARM64E3-NEXT: platform ios
+// CHECK-BUILD-IOS-ARM64E3-NEXT:      sdk n/a
+// CHECK-BUILD-IOS-ARM64E3-NEXT:    minos 14.1
+// CHECK-BUILD-IOS-ARM64E3-NEXT:   ntools 0
+// CHECK-BUILD-IOS-ARM64E3-NOT: LC_VERSION_MIN
+
 // CHECK-BUILD-IOSSIM2:           cmd LC_BUILD_VERSION
 // CHECK-BUILD-IOSSIM2-NEXT:  cmdsize 24
 // CHECK-BUILD-IOSSIM2-NEXT: platform iossim

test/tools/llvm-dwarfdump/AArch64/arm64e.ll

@@ -0,0 +1,17 @@
+; RUN: llc -O0 %s -filetype=obj -o - \
+; RUN:   | llvm-dwarfdump -arch arm64e - | FileCheck %s
+; CHECK: file format Mach-O arm64
+
+source_filename = "/tmp/empty.c"
+target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
+target triple = "arm64e-apple-ios"
+
+!llvm.module.flags = !{!1, !2, !3, !4}
+!llvm.dbg.cu = !{!5}
+
+!1 = !{i32 2, !"Dwarf Version", i32 4}
+!2 = !{i32 2, !"Debug Info Version", i32 3}
+!3 = !{i32 1, !"wchar_size", i32 4}
+!4 = !{i32 7, !"PIC Level", i32 2}
+!5 = distinct !DICompileUnit(language: DW_LANG_C99, file: !6, producer: "Apple clang", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug)
+!6 = !DIFile(filename: "/tmp/empty.c", directory: "/Volumes/Data/llvm-project")

test/tools/llvm-objdump/MachO/universal-arm64.test

@@ -22,7 +22,7 @@
 # CHECK-NEXT:     offset 16384
 # CHECK-NEXT:     size 384
 # CHECK-NEXT:     align 2^14 (16384)
-# CHECK-NEXT: architecture
+# CHECK-NEXT: architecture arm64e
 # CHECK-NEXT:     cputype CPU_TYPE_ARM64
 # CHECK-NEXT:     cpusubtype CPU_SUBTYPE_ARM64E
 # CHECK-NEXT:     capabilities 0x0

test/tools/llvm-readobj/macho-arm64e.test

@@ -0,0 +1,17 @@
+# RUN: yaml2obj %s -o %t.o
+# RUN: llvm-readobj -h %t.o | FileCheck %s
+
+# CHECK: Magic: Magic64 (0xFEEDFACF)
+# CHECK: CpuType: Arm64 (0x100000C)
+# CHECK: CpuSubType: CPU_SUBTYPE_ARM64E (0x2)
+
+--- !mach-o
+FileHeader:
+  magic:           0xFEEDFACF
+  cputype:         0x0100000C
+  cpusubtype:      0x00000002
+  filetype:        0x00000001
+  ncmds:           0
+  sizeofcmds:      0
+  flags:           0x00000000
+  reserved:        0x00000000

unittests/ADT/TripleTest.cpp

@@ -1590,5 +1590,10 @@ TEST(TripleTest, ParseARMArch) {
     Triple T = Triple("aarch64_be");
     EXPECT_EQ(Triple::aarch64_be, T.getArch());
   }
+  {
+    Triple T = Triple("arm64e");
+    EXPECT_EQ(Triple::aarch64, T.getArch());
+    EXPECT_EQ(Triple::AArch64SubArch_arm64e, T.getSubArch());
+  }
 }
 } // end anonymous namespace

utils/UpdateTestChecks/asm.py

@@ -336,6 +336,7 @@ def build_function_body_dictionary_for_triple(args, raw_tool_output, triple,
       'amdgcn': (scrub_asm_amdgpu, ASM_FUNCTION_AMDGPU_RE),
       'arm': (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_RE),
       'arm64': (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_RE),
+      'arm64e': (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE),
       'arm64-apple-ios': (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE),
       'armv7-apple-ios' : (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_IOS_RE),
       'armv7-apple-darwin': (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_DARWIN_RE),