Add a command-line flag to control the Swift extended async frame info.

[aschwaighofer]

Summary

Introduce a new command-line flag -swift-async-fp={auto|always|never}
that controls how code generation sets the Swift extended async frame
info bit. There are three possibilities:

• auto: the default, which determines how to set the bit based on

deployment target, either statically or dynamically via
swift_async_extendedFramePointerFlags.

• always: always set the bit statically, regardless of deployment

target.

• never: never set the bit, regardless of deployment target.

Patch by Doug Gregor <[e-mail]>

[aschwaighofer]

Trying to set the dependence on https://reviews.llvm.org/D109218

llvm/include/llvm/CodeGen/CommandFlags.h

 DenormalMode::DenormalModeKind getDenormalFP32Math();
 
 bool getEnableHonorSignDependentRoundingFPMath();
 
 llvm::FloatABI::ABIType getFloatABIForCalls();
 
 llvm::FPOpFusion::FPOpFusionMode getFuseFPOps();
 
+SwiftAsyncFramePointerMode getSwiftAsyncFramePointer();
+
 bool getDontPlaceZerosInBSS();
 
 bool getEnableGuaranteedTailCallOpt();
 
 bool getEnableAIXExtendedAltivecABI();
 
 bool getDisableTailCalls();
 

llvm/include/llvm/Target/TargetOptions.h

namespace llvm {
   /// Enable abort calls when global instruction selection fails to lower/select
   /// an instruction.
   enum class GlobalISelAbortMode {
     Disable,        // Disable the abort.
     Enable,         // Enable the abort.
     DisableWithDiag // Disable the abort but emit a diagnostic on failure.
   };
 
+  /// Indicates when and how the Swift async frame pointer bit should be set.
+  enum class SwiftAsyncFramePointerMode {
+    /// Determine whether to set the bit statically or dynamically based
+    /// on the deployment target.
+    DeploymentBased,
+    /// Always set the bit.
+    Always,
+    /// Never set the bit.
+    Never,
+  };
+
   class TargetOptions {
   public:
     TargetOptions()
         : UnsafeFPMath(false), NoInfsFPMath(false), NoNaNsFPMath(false),
           NoTrappingFPMath(true), NoSignedZerosFPMath(false),
           EnableAIXExtendedAltivecABI(false),
           HonorSignDependentRoundingFPMathOption(false), NoZerosInBSS(false),
           GuaranteedTailCallOpt(false), StackSymbolOrdering(true),
public:
 
     /// EnableGlobalISel - This flag enables global instruction selection.
     unsigned EnableGlobalISel : 1;
 
     /// EnableGlobalISelAbort - Control abort behaviour when global instruction
     /// selection fails to lower/select an instruction.
     GlobalISelAbortMode GlobalISelAbort = GlobalISelAbortMode::Enable;
 
+    /// Control when and how the Swift async frame pointer bit should
+    /// be set.
+    SwiftAsyncFramePointerMode SwiftAsyncFramePointer =
+        SwiftAsyncFramePointerMode::Always;
+
     /// UseInitArray - Use .init_array instead of .ctors for static
     /// constructors.
     unsigned UseInitArray : 1;
 
     /// Disable the integrated assembler.
     unsigned DisableIntegratedAS : 1;
 
     /// Compress DWARF debug sections.

llvm/lib/CodeGen/CommandFlags.cpp

 CGOPT(bool, EnableNoSignedZerosFPMath)
 CGOPT(bool, EnableNoTrappingFPMath)
 CGOPT(bool, EnableAIXExtendedAltivecABI)
 CGOPT(DenormalMode::DenormalModeKind, DenormalFPMath)
 CGOPT(DenormalMode::DenormalModeKind, DenormalFP32Math)
 CGOPT(bool, EnableHonorSignDependentRoundingFPMath)
 CGOPT(FloatABI::ABIType, FloatABIForCalls)
 CGOPT(FPOpFusion::FPOpFusionMode, FuseFPOps)
+CGOPT(SwiftAsyncFramePointerMode, SwiftAsyncFramePointer)
 CGOPT(bool, DontPlaceZerosInBSS)
 CGOPT(bool, EnableGuaranteedTailCallOpt)
 CGOPT(bool, DisableTailCalls)
 CGOPT(bool, StackSymbolOrdering)
 CGOPT(bool, StackRealign)
 CGOPT(std::string, TrapFuncName)
 CGOPT(bool, UseCtors)
 CGOPT(bool, RelaxELFRelocations)
static cl::opt<FPOpFusion::FPOpFusionMode> FuseFPOps(
       cl::values(
           clEnumValN(FPOpFusion::Fast, "fast",
                      "Fuse FP ops whenever profitable"),
           clEnumValN(FPOpFusion::Standard, "on", "Only fuse 'blessed' FP ops."),
           clEnumValN(FPOpFusion::Strict, "off",
                      "Only fuse FP ops when the result won't be affected.")));
   CGBINDOPT(FuseFPOps);
 
+  static cl::opt<SwiftAsyncFramePointerMode> SwiftAsyncFramePointer(
+      "swift-async-fp",
+      cl::desc("Determine when the Swift async frame pointer should be set"),
+      cl::init(SwiftAsyncFramePointerMode::Always),
+      cl::values(clEnumValN(SwiftAsyncFramePointerMode::DeploymentBased, "auto",
+                            "Determine based on deployment target"),
+                 clEnumValN(SwiftAsyncFramePointerMode::Always, "always",
+                            "Always set the bit"),
+                 clEnumValN(SwiftAsyncFramePointerMode::Never, "never",
+                            "Never set the bit")));
+  CGBINDOPT(SwiftAsyncFramePointer);
+
   static cl::opt<bool> DontPlaceZerosInBSS(
       "nozero-initialized-in-bss",
       cl::desc("Don't place zero-initialized symbols into bss section"),
       cl::init(false));
   CGBINDOPT(DontPlaceZerosInBSS);
 
   static cl::opt<bool> EnableAIXExtendedAltivecABI(
       "vec-extabi", cl::desc("Enable the AIX Extended Altivec ABI."),
codegen::InitTargetOptionsFromCodeGenFlags(const Triple &TheTriple) {
   Options.DebugStrictDwarf = getDebugStrictDwarf();
   Options.LoopAlignment = getAlignLoops();
 
   Options.MCOptions = mc::InitMCTargetOptionsFromFlags();
 
   Options.ThreadModel = getThreadModel();
   Options.EABIVersion = getEABIVersion();
   Options.DebuggerTuning = getDebuggerTuningOpt();
-
+  Options.SwiftAsyncFramePointer = getSwiftAsyncFramePointer();
   return Options;
 }
 
 std::string codegen::getCPUStr() {
   // If user asked for the 'native' CPU, autodetect here. If autodection fails,
   // this will set the CPU to an empty string which tells the target to
   // pick a basic default.
   if (getMCPU() == "native")

llvm/lib/Target/AArch64/AArch64FrameLowering.cpp

BuildMI(MBB, MBBI, DL, TII->get(TargetOpcode::CFI_INSTRUCTION))
         .setMIFlags(MachineInstr::FrameSetup);
   }
 
   // We signal the presence of a Swift extended frame to external tools by
   // storing FP with 0b0001 in bits 63:60. In normal userland operation a simple
   // ORR is sufficient, it is assumed a Swift kernel would initialize the TBI
   // bits so that is still true.
   if (HasFP && AFI->hasSwiftAsyncContext()) {
-    if  (Subtarget.swiftAsyncContextIsDynamicallySet()) {
+    switch (MF.getTarget().Options.SwiftAsyncFramePointer) {
+    case SwiftAsyncFramePointerMode::DeploymentBased:
+      if (Subtarget.swiftAsyncContextIsDynamicallySet()) {
         // The special symbol below is absolute and has a *value* that can be
         // combined with the frame pointer to signal an extended frame.
         BuildMI(MBB, MBBI, DL, TII->get(AArch64::LOADgot), AArch64::X16)
             .addExternalSymbol("swift_async_extendedFramePointerFlags",
                                AArch64II::MO_GOT);
         BuildMI(MBB, MBBI, DL, TII->get(AArch64::ORRXrs), AArch64::FP)
             .addUse(AArch64::FP)
             .addUse(AArch64::X16)
             .addImm(Subtarget.isTargetILP32() ? 32 : 0);
-    } else {
+        break;
+      }
+      LLVM_FALLTHROUGH;
+
+    case SwiftAsyncFramePointerMode::Always:
       // ORR x29, x29, #0x1000_0000_0000_0000
       BuildMI(MBB, MBBI, DL, TII->get(AArch64::ORRXri), AArch64::FP)
           .addUse(AArch64::FP)
           .addImm(0x1100)
           .setMIFlag(MachineInstr::FrameSetup);
+      break;
+
+    case SwiftAsyncFramePointerMode::Never:
+      break;
     }
   }
 
   // All calls are tail calls in GHC calling conv, and functions have no
   // prologue/epilogue.
   if (MF.getFunction().getCallingConv() == CallingConv::GHC)
     return;
 

llvm/lib/Target/X86/X86FrameLowering.cpp

void X86FrameLowering::emitPrologue(MachineFunction &MF,
   if (TailCallArgReserveSize  && IsWin64Prologue)
     report_fatal_error("Can't handle guaranteed tail call under win64 yet");
 
   const bool EmitStackProbeCall =
       STI.getTargetLowering()->hasStackProbeSymbol(MF);
   unsigned StackProbeSize = STI.getTargetLowering()->getStackProbeSize(MF);
 
   if (HasFP && X86FI->hasSwiftAsyncContext()) {
+    switch (MF.getTarget().Options.SwiftAsyncFramePointer) {
+    case SwiftAsyncFramePointerMode::DeploymentBased:
       if (STI.swiftAsyncContextIsDynamicallySet()) {
         // The special symbol below is absolute and has a *value* suitable to be
         // combined with the frame pointer directly.
         BuildMI(MBB, MBBI, DL, TII.get(X86::OR64rm), MachineFramePtr)
             .addUse(MachineFramePtr)
             .addUse(X86::RIP)
             .addImm(1)
             .addUse(X86::NoRegister)
             .addExternalSymbol("swift_async_extendedFramePointerFlags",
                                X86II::MO_GOTPCREL)
             .addUse(X86::NoRegister);
-    } else {
+        break;
+      }
+      LLVM_FALLTHROUGH;
+
+    case SwiftAsyncFramePointerMode::Always:
       BuildMI(MBB, MBBI, DL, TII.get(X86::BTS64ri8), MachineFramePtr)
           .addUse(MachineFramePtr)
           .addImm(60)
           .setMIFlag(MachineInstr::FrameSetup);
+      break;
+
+    case SwiftAsyncFramePointerMode::Never:
+      break;
     }
   }
 
   // Re-align the stack on 64-bit if the x86-interrupt calling convention is
   // used and an error code was pushed, since the x86-64 ABI requires a 16-byte
   // stack alignment.
   if (Fn.getCallingConv() == CallingConv::X86_INTR && Is64Bit &&
       Fn.arg_size() == 2) {

llvm/test/CodeGen/AArch64/swift-dynamic-async-frame.ll

 ; RUN: llc -mtriple arm64-apple-ios15.0.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
-; RUN: llc -mtriple arm64-apple-ios14.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-DYNAMIC
+; RUN: llc -mtriple arm64-apple-ios15.0.0 -swift-async-fp=auto %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple arm64-apple-ios15.0.0 -swift-async-fp=never %s -o - | FileCheck %s --check-prefix=CHECK-NEVER
+; RUN: llc -mtriple arm64-apple-ios14.9.0 -swift-async-fp=always %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple arm64-apple-ios14.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple arm64-apple-ios14.9.0 -swift-async-fp=auto %s -o - | FileCheck %s --check-prefix=CHECK-DYNAMIC
 ; RUN: llc -mtriple arm64-apple-tvos15.0.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
-; RUN: llc -mtriple arm64-apple-tvos14.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-DYNAMIC
+; RUN: llc -mtriple arm64-apple-tvos14.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
 ; RUN: llc -mtriple arm64-apple-macosx12.0.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
-; RUN: llc -mtriple arm64-apple-macosx11.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-DYNAMIC
+; RUN: llc -mtriple arm64-apple-macosx11.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
 ; RUN: llc -mtriple arm64_32-apple-watchos8.0.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
-; RUN: llc -mtriple arm64_32-apple-watchos7.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-DYNAMIC-32
+; RUN: llc -mtriple arm64_32-apple-watchos7.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple arm64_32-apple-watchos7.9.0 -swift-async-fp=auto %s -o - | FileCheck %s --check-prefix=CHECK-DYNAMIC-32
 
 ; CHECK-STATIC-LABEL: foo:
 ; CHECK-STATIC: orr x29, x29, #0x1000000000000000
 
+; CHECK-NEVER-LABEL: foo:
+; CHECK-NEVER-NOT: orr x29, x29, #0x1000000000000000
+; CHECK-NEVER-NOT: _swift_async_extendedFramePointerFlags
+
 ; CHECK-DYNAMIC-LABEL: foo:
 ; CHECK-DYNAMIC: adrp x16, _swift_async_extendedFramePointerFlags@GOTPAGE
 ; CHECK-DYNAMIC: ldr x16, [x16, _swift_async_extendedFramePointerFlags@GOTPAGEOFF]
 ; CHECK-DYNAMIC: orr x29, x29, x16
 
 ; CHECK-DYNAMIC-32-LABEL: foo:
 ; CHECK-DYNAMIC-32: adrp x16, _swift_async_extendedFramePointerFlags@GOTPAGE
 ; CHECK-DYNAMIC-32: ldr w16, [x16, _swift_async_extendedFramePointerFlags@GOTPAGEOFF]
 ; CHECK-DYNAMIC-32: orr x29, x29, x16, lsl #32
 
 define void @foo(i8* swiftasync) "frame-pointer"="all" {
   ret void
 }

llvm/test/CodeGen/X86/swift-dynamic-async-frame.ll

 ; RUN: llc -mtriple x86_64-apple-macosx12.0.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
-; RUN: llc -mtriple x86_64-apple-macosx11.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-DYNAMIC
+; RUN: llc -mtriple x86_64-apple-macosx12.0.0 -swift-async-fp=always %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple x86_64-apple-macosx12.0.0 -swift-async-fp=auto %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple x86_64-apple-macosx11.9.0 -swift-async-fp=always %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple x86_64-apple-macosx11.9.0 %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple x86_64-apple-macosx11.9.0 -swift-async-fp=auto %s -o - | FileCheck %s --check-prefix=CHECK-DYNAMIC
+; RUN: llc -mtriple x86_64-apple-macosx11.9.0 -swift-async-fp=always %s -o - | FileCheck %s --check-prefix=CHECK-STATIC
+; RUN: llc -mtriple x86_64-apple-macosx11.9.0 -swift-async-fp=never %s -o - | FileCheck %s --check-prefix=CHECK-NEVER
 
 ; CHECK-STATIC-LABEL: foo:
 ; CHECK-STATIC: btsq $60, %rbp
 
 ; CHECK-DYNAMIC-LABEL: foo:
 ; CHECK-DYNAMIC: orq _swift_async_extendedFramePointerFlags@GOTPCREL(%rip), %rbp
 
+; CHECK-NEVER-LABEL: foo:
+; CHECK-NEVER-NOT: btsq $60, %rbp
+; CHECK-NEVER-NOT: _swift_async_extendedFramePointerFlags
+
 define void @foo(i8* swiftasync) "frame-pointer"="all" {
   ret void
 }