[AArch64][GlobalISel] Tail call memory intrinsics

[paquette]

Summary

Because memory intrinsics are handled differently than other calls, we need to check them for tail call eligiblity in the legalizer. This allows us to still inline them when it's beneficial to do so, but also tail call when possible.

This adds simple tail calling support for when the intrinsic is followed by a return.

It ports the attribute checks from TargetLowering::isInTailCallPosition into a similarly-named function in LegalizerHelper.cpp. The target-specific isUsedByReturnOnly hook is not ported here.

Update tailcall-mem-intrinsics.ll to show that GlobalISel can now tail call memory intrinsics.

Update legalize-memcpy-et-al.mir to have a case where we don't tail call.

[paquette]

Add missing check for when the return is a tail call.

[aemerson]

llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h

98

Can you make it clearer in this comment that it's done by the legalizer.

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

335

A comment here would be useful.

llvm/test/CodeGen/AArch64/tailcall-mem-intrinsics.ll

2

-global-isel-abort=1?

[paquette]

Update comments, add -global-isel-abort=1 to test

llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h

struct CallLoweringInfo {
 
     /// True if the call must be tail call optimized.
     bool IsMustTailCall = false;
 
     /// True if the call passes all target-independent checks for tail call
     /// optimization.
     bool IsTailCall = false;
 
+    /// True if the call was lowered as a tail call.
+    bool LoweredTailCall = false;
+
     /// True if the call is to a vararg function.
     bool IsVarArg = false;
   };
 
   /// Argument handling is mostly uniform between the four places that
   /// make these decisions: function formal arguments, call
   /// instruction args, call instruction returns and function
   /// returns. However, once a decision has been made on where an

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

case TargetOpcode::G_FCEIL:
     return Size == 64 ? RTLIB::CEIL_F64 : RTLIB::CEIL_F32;
   case TargetOpcode::G_FFLOOR:
     assert((Size == 32 || Size == 64) && "Unsupported size");
     return Size == 64 ? RTLIB::FLOOR_F64 : RTLIB::FLOOR_F32;
   }
   llvm_unreachable("Unknown libcall function");
 }
 
+/// True if an instruction is in tail position in its caller. Intended for
+/// legalizing libcalls as tail calls when possible.
+static bool isLibCallInTailPosition(MachineInstr &MI) {
+  const Function &F = MI.getParent()->getParent()->getFunction();
+  AttributeList CallerAttrs = F.getAttributes();
+  if (AttrBuilder(CallerAttrs, AttributeList::ReturnIndex)
+          .removeAttribute(Attribute::NoAlias)
+          .removeAttribute(Attribute::NonNull)
+          .hasAttributes())
+    return false;
+
+  // It's not safe to eliminate the sign / zero extension of the return value.
+  if (CallerAttrs.hasAttribute(AttributeList::ReturnIndex, Attribute::ZExt) ||
+      CallerAttrs.hasAttribute(AttributeList::ReturnIndex, Attribute::SExt))
+    return false;
+
+  // Only tail call if the following instruction is a standard return.
+  auto &TII = *MI.getMF()->getSubtarget().getInstrInfo();
+  MachineInstr *Next = MI.getNextNode();
+  if (!Next || TII.isTailCall(*Next) || !Next->isReturn())
+    return false;
+
+  return true;
+}
+
 LegalizerHelper::LegalizeResult
 llvm::createLibcall(MachineIRBuilder &MIRBuilder, RTLIB::Libcall Libcall,
                     const CallLowering::ArgInfo &Result,
                     ArrayRef<CallLowering::ArgInfo> Args) {
   auto &CLI = *MIRBuilder.getMF().getSubtarget().getCallLowering();
   auto &TLI = *MIRBuilder.getMF().getSubtarget().getTargetLowering();
   const char *Name = TLI.getLibcallName(Libcall);
 
llvm::createMemLibcall(MachineIRBuilder &MIRBuilder, MachineRegisterInfo &MRI,
 
   MIRBuilder.setInstr(MI);
   MIRBuilder.getMF().getFrameInfo().setHasCalls(true);
 
   CallLowering::CallLoweringInfo Info;
   Info.CallConv = TLI.getLibcallCallingConv(RTLibcall);
   Info.Callee = MachineOperand::CreateES(Name);
   Info.OrigRet = CallLowering::ArgInfo({0}, Type::getVoidTy(Ctx));
+  Info.IsTailCall = isLibCallInTailPosition(MI);
+
   std::copy(Args.begin(), Args.end(), std::back_inserter(Info.OrigArgs));
   if (!CLI.lowerCall(MIRBuilder, Info))
     return LegalizerHelper::UnableToLegalize;
 
+  if (Info.LoweredTailCall) {
+    assert(Info.IsTailCall && "Lowered tail call when it wasn't a tail call?");
+    // We must have a return following the call to get past
+    // isLibCallInTailPosition.
+    assert(MI.getNextNode() && MI.getNextNode()->isReturn() &&
+           "Expected instr following MI to be a return?");
+
+    // We lowered a tail call, so the call is now the return from the block.
+    // Delete the old return.
+    MI.getNextNode()->eraseFromParent();
+  }
+
   return LegalizerHelper::Legalized;
 }
 
 static RTLIB::Libcall getConvRTLibDesc(unsigned Opcode, Type *ToType,
                                        Type *FromType) {
   auto ToMVT = MVT::getVT(ToType);
   auto FromMVT = MVT::getVT(FromType);
 

llvm/lib/Target/AArch64/AArch64CallLowering.cpp

bool AArch64CallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
   if (Info.Callee.isReg())
     MIB->getOperand(0).setReg(constrainOperandRegClass(
         MF, *TRI, MRI, *MF.getSubtarget().getInstrInfo(),
         *MF.getSubtarget().getRegBankInfo(), *MIB, MIB->getDesc(), Info.Callee,
         0));
 
   // If we're tail calling, then we're the return from the block. So, we don't
   // want to copy anything.
-  if (IsSibCall)
+  if (IsSibCall) {
+    Info.LoweredTailCall = true;
     return true;
+  }
 
   // Finally we can copy the returned value back into its virtual-register. In
   // symmetry with the arugments, the physical register must be an
   // implicit-define of the call instruction.
   if (!Info.OrigRet.Ty->isVoidTy()) {
     CCAssignFn *RetAssignFn = TLI.CCAssignFnForReturn(F.getCallingConv());
     CallReturnHandler Handler(MIRBuilder, MRI, MIB, RetAssignFn);
     if (!handleAssignments(MIRBuilder, InArgs, Handler))

llvm/test/CodeGen/AArch64/GlobalISel/legalize-memcpy-et-al.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
 # RUN: llc -march=aarch64 -run-pass=legalizer -verify-machineinstrs %s -o - | FileCheck %s
 ---
 name:            test_memcpy
 tracksRegLiveness: true
 body:             |
   bb.1:
     liveins: $w2, $x0, $x1
 
     ; CHECK-LABEL: name: test_memcpy
     ; CHECK: liveins: $w2, $x0, $x1
     ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
     ; CHECK: [[COPY1:%[0-9]+]]:_(p0) = COPY $x1
     ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $w2
     ; CHECK: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
-    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
     ; CHECK: $x0 = COPY [[COPY]](p0)
     ; CHECK: $x1 = COPY [[COPY1]](p0)
     ; CHECK: $x2 = COPY [[ZEXT]](s64)
-    ; CHECK: BL &memcpy, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $x0, implicit $x1, implicit $x2
-    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
-    ; CHECK: RET_ReallyLR
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+    ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[C1]], [[C]]
+    ; CHECK: $w3 = COPY [[AND]](s32)
+    ; CHECK: TCRETURNdi &memcpy, 0, csr_aarch64_aapcs, implicit $sp, implicit $x0, implicit $x1, implicit $x2, implicit $w3
     %0:_(p0) = COPY $x0
     %1:_(p0) = COPY $x1
     %2:_(s32) = COPY $w2
     %4:_(s1) = G_CONSTANT i1 false
     %3:_(s64) = G_ZEXT %2(s32)
     G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.memcpy), %0(p0), %1(p0), %3(s64), %4(s1)
     RET_ReallyLR
 
 ...
 ---
 name:            test_memmove
 tracksRegLiveness: true
 body:             |
   bb.1:
     liveins: $w2, $x0, $x1
 
     ; CHECK-LABEL: name: test_memmove
     ; CHECK: liveins: $w2, $x0, $x1
     ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
     ; CHECK: [[COPY1:%[0-9]+]]:_(p0) = COPY $x1
     ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $w2
     ; CHECK: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
-    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
     ; CHECK: $x0 = COPY [[COPY]](p0)
     ; CHECK: $x1 = COPY [[COPY1]](p0)
     ; CHECK: $x2 = COPY [[ZEXT]](s64)
-    ; CHECK: BL &memmove, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $x0, implicit $x1, implicit $x2
-    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
-    ; CHECK: RET_ReallyLR
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+    ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[C1]], [[C]]
+    ; CHECK: $w3 = COPY [[AND]](s32)
+    ; CHECK: TCRETURNdi &memmove, 0, csr_aarch64_aapcs, implicit $sp, implicit $x0, implicit $x1, implicit $x2, implicit $w3
     %0:_(p0) = COPY $x0
     %1:_(p0) = COPY $x1
     %2:_(s32) = COPY $w2
     %4:_(s1) = G_CONSTANT i1 false
     %3:_(s64) = G_ZEXT %2(s32)
     G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.memmove), %0(p0), %1(p0), %3(s64), %4(s1)
     RET_ReallyLR
 
 ...
 ---
 name:            test_memset
 tracksRegLiveness: true
 body:             |
   bb.1:
     liveins: $w1, $w2, $x0
 
     ; CHECK-LABEL: name: test_memset
     ; CHECK: liveins: $w1, $w2, $x0
     ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
     ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $w1
     ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $w2
     ; CHECK: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
-    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
     ; CHECK: $x0 = COPY [[COPY]](p0)
     ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
     ; CHECK: $w1 = COPY [[COPY3]](s32)
     ; CHECK: $x2 = COPY [[ZEXT]](s64)
-    ; CHECK: BL &memset, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $x0, implicit $w1, implicit $x2
-    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
-    ; CHECK: RET_ReallyLR
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+    ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[C1]], [[C]]
+    ; CHECK: $w3 = COPY [[AND]](s32)
+    ; CHECK: TCRETURNdi &memset, 0, csr_aarch64_aapcs, implicit $sp, implicit $x0, implicit $w1, implicit $x2, implicit $w3
     %0:_(p0) = COPY $x0
     %1:_(s32) = COPY $w1
     %2:_(s32) = COPY $w2
     %5:_(s1) = G_CONSTANT i1 false
     %3:_(s8) = G_TRUNC %1(s32)
     %4:_(s64) = G_ZEXT %2(s32)
     G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.memset), %0(p0), %3(s8), %4(s64), %5(s1)
     RET_ReallyLR
 
 ...
+---
+name:            no_tail_call
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w2, $x0, $x1
+
+    ; CHECK-LABEL: name: no_tail_call
+    ; CHECK: liveins: $w2, $x0, $x1
+    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(p0) = COPY $x1
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $w2
+    ; CHECK: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $x0 = COPY [[COPY]](p0)
+    ; CHECK: $x1 = COPY [[COPY1]](p0)
+    ; CHECK: $x2 = COPY [[ZEXT]](s64)
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+    ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[C1]], [[C]]
+    ; CHECK: $w3 = COPY [[AND]](s32)
+    ; CHECK: BL &memcpy, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $x0, implicit $x1, implicit $x2, implicit $w3
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $x0 = COPY [[ZEXT]](s64)
+    ; CHECK: RET_ReallyLR implicit $x0
+    %0:_(p0) = COPY $x0
+    %1:_(p0) = COPY $x1
+    %2:_(s32) = COPY $w2
+    %4:_(s1) = G_CONSTANT i1 false
+    %3:_(s64) = G_ZEXT %2(s32)
+    G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.memcpy), %0(p0), %1(p0), %3(s64), %4(s1)
+    $x0 = COPY %3
+    RET_ReallyLR implicit $x0
+
+...
+---
+name:            dont_tc_twice
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w2, $x0, $x1
+    ; CHECK-LABEL: name: dont_tc_twice
+    ; CHECK: liveins: $w2, $x0, $x1
+    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(p0) = COPY $x1
+    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $w2
+    ; CHECK: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[COPY2]](s32)
+    ; CHECK: ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: $x0 = COPY [[COPY]](p0)
+    ; CHECK: $x1 = COPY [[COPY1]](p0)
+    ; CHECK: $x2 = COPY [[ZEXT]](s64)
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+    ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[C1]], [[C]]
+    ; CHECK: $w3 = COPY [[AND]](s32)
+    ; CHECK: BL &memcpy, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $x0, implicit $x1, implicit $x2, implicit $w3
+    ; CHECK: ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
+    ; CHECK: TCRETURNdi &memset, 0, csr_aarch64_aapcs, implicit $sp
+    %0:_(p0) = COPY $x0
+    %1:_(p0) = COPY $x1
+    %2:_(s32) = COPY $w2
+    %4:_(s1) = G_CONSTANT i1 false
+    %3:_(s64) = G_ZEXT %2(s32)
+    G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.memcpy), %0(p0), %1(p0), %3(s64), %4(s1)
+    TCRETURNdi &memset, 0, csr_aarch64_aapcs, implicit $sp

llvm/test/CodeGen/AArch64/tailcall-mem-intrinsics.ll

 ; RUN: llc -mtriple=aarch64-unknown-unknown < %s | FileCheck %s
+; RUN: llc -verify-machineinstrs -mtriple=aarch64-unknown-unknown -global-isel < %s | FileCheck %s
 
 ; CHECK-LABEL: tail_memcpy:
 ; CHECK: b memcpy
 define void @tail_memcpy(i8* nocapture %p, i8* nocapture readonly %q, i32 %n) #0 {
 entry:
   tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %p, i8* %q, i32 %n, i1 false)
   ret void
 }