[LoongArch] Implement mayBeEmittedAsTailCall for tail call optimization

[wangleiat]

Summary

Implements TargetLowering callback mayBeEmittedAsTailCall that enables
CodeGenPrepare to duplicate returns when they might enable a tail-call.

llvm/lib/Target/LoongArch/LoongArchISelLowering.h

public:
   getExceptionSelectorRegister(const Constant *PersonalityFn) const override;
 
   ISD::NodeType getExtendForAtomicOps() const override {
     return ISD::SIGN_EXTEND;
   }
 
   Register getRegisterByName(const char *RegName, LLT VT,
                              const MachineFunction &MF) const override;
+  bool mayBeEmittedAsTailCall(const CallInst *CI) const override;
 
   bool decomposeMulByConstant(LLVMContext &Context, EVT VT,
                               SDValue C) const override;
 
 private:
   /// Target-specific function used to lower LoongArch calling conventions.
   typedef bool LoongArchCCAssignFn(const DataLayout &DL, LoongArchABI::ABI ABI,
                                    unsigned ValNo, MVT ValVT,

llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp

SDValue LoongArchTargetLowering::LowerFormalArguments(
   if (!OutChains.empty()) {
     OutChains.push_back(Chain);
     Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, OutChains);
   }
 
   return Chain;
 }
 
+bool LoongArchTargetLowering::mayBeEmittedAsTailCall(const CallInst *CI) const {
+  return CI->isTailCall();
+}
+
 // Check whether the call is eligible for tail call optimization.
 bool LoongArchTargetLowering::isEligibleForTailCallOptimization(
     CCState &CCInfo, CallLoweringInfo &CLI, MachineFunction &MF,
     const SmallVectorImpl<CCValAssign> &ArgLocs) const {
 
   auto CalleeCC = CLI.CallConv;
   auto &Outs = CLI.Outs;
   auto &Caller = MF.getFunction();

llvm/test/CodeGen/LoongArch/bnez-beqz.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc --mtriple=loongarch32 < %s | FileCheck %s --check-prefix=LA32
 ; RUN: llc --mtriple=loongarch64 < %s | FileCheck %s --check-prefix=LA64
 
 declare void @bar()
 
 define void @bnez_i32(i32 signext %0) nounwind {
 ; LA32-LABEL: bnez_i32:
 ; LA32:       # %bb.0: # %start
-; LA32-NEXT:    addi.w $sp, $sp, -16
-; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
-; LA32-NEXT:    bnez $a0, .LBB0_2
-; LA32-NEXT:  # %bb.1: # %t
-; LA32-NEXT:    bl %plt(bar)
-; LA32-NEXT:  .LBB0_2: # %f
-; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
-; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    beqz $a0, .LBB0_2
+; LA32-NEXT:  # %bb.1: # %f
 ; LA32-NEXT:    ret
+; LA32-NEXT:  .LBB0_2: # %t
+; LA32-NEXT:    b %plt(bar)
 ;
 ; LA64-LABEL: bnez_i32:
 ; LA64:       # %bb.0: # %start
-; LA64-NEXT:    addi.d $sp, $sp, -16
-; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
-; LA64-NEXT:    bnez $a0, .LBB0_2
-; LA64-NEXT:  # %bb.1: # %t
-; LA64-NEXT:    bl %plt(bar)
-; LA64-NEXT:  .LBB0_2: # %f
-; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
-; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    beqz $a0, .LBB0_2
+; LA64-NEXT:  # %bb.1: # %f
 ; LA64-NEXT:    ret
+; LA64-NEXT:  .LBB0_2: # %t
+; LA64-NEXT:    b %plt(bar)
 start:
   %1 = icmp eq i32 %0, 0
   br i1 %1, label %t, label %f
 
 t:
   tail call void @bar()
   br label %f
 
 f:
   ret void
 }
 
 define void @beqz_i32(i32 signext %0) nounwind {
 ; LA32-LABEL: beqz_i32:
 ; LA32:       # %bb.0: # %start
-; LA32-NEXT:    addi.w $sp, $sp, -16
-; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
 ; LA32-NEXT:    beqz $a0, .LBB1_2
 ; LA32-NEXT:  # %bb.1: # %t
-; LA32-NEXT:    bl %plt(bar)
+; LA32-NEXT:    b %plt(bar)
 ; LA32-NEXT:  .LBB1_2: # %f
-; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
-; LA32-NEXT:    addi.w $sp, $sp, 16
 ; LA32-NEXT:    ret
 ;
 ; LA64-LABEL: beqz_i32:
 ; LA64:       # %bb.0: # %start
-; LA64-NEXT:    addi.d $sp, $sp, -16
-; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
 ; LA64-NEXT:    beqz $a0, .LBB1_2
 ; LA64-NEXT:  # %bb.1: # %t
-; LA64-NEXT:    bl %plt(bar)
+; LA64-NEXT:    b %plt(bar)
 ; LA64-NEXT:  .LBB1_2: # %f
-; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
-; LA64-NEXT:    addi.d $sp, $sp, 16
 ; LA64-NEXT:    ret
 start:
   %1 = icmp ne i32 %0, 0
   br i1 %1, label %t, label %f
 
 t:
   tail call void @bar()
   br label %f
 
 f:
   ret void
 }
 
 define void @bnez_i64(i64 %0) nounwind {
 ; LA32-LABEL: bnez_i64:
 ; LA32:       # %bb.0: # %start
-; LA32-NEXT:    addi.w $sp, $sp, -16
-; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
 ; LA32-NEXT:    or $a0, $a0, $a1
-; LA32-NEXT:    bnez $a0, .LBB2_2
-; LA32-NEXT:  # %bb.1: # %t
-; LA32-NEXT:    bl %plt(bar)
-; LA32-NEXT:  .LBB2_2: # %f
-; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
-; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    beqz $a0, .LBB2_2
+; LA32-NEXT:  # %bb.1: # %f
 ; LA32-NEXT:    ret
+; LA32-NEXT:  .LBB2_2: # %t
+; LA32-NEXT:    b %plt(bar)
 ;
 ; LA64-LABEL: bnez_i64:
 ; LA64:       # %bb.0: # %start
-; LA64-NEXT:    addi.d $sp, $sp, -16
-; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
-; LA64-NEXT:    bnez $a0, .LBB2_2
-; LA64-NEXT:  # %bb.1: # %t
-; LA64-NEXT:    bl %plt(bar)
-; LA64-NEXT:  .LBB2_2: # %f
-; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
-; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    beqz $a0, .LBB2_2
+; LA64-NEXT:  # %bb.1: # %f
 ; LA64-NEXT:    ret
+; LA64-NEXT:  .LBB2_2: # %t
+; LA64-NEXT:    b %plt(bar)
 start:
   %1 = icmp eq i64 %0, 0
   br i1 %1, label %t, label %f
 
 t:
   tail call void @bar()
   br label %f
 
 f:
   ret void
 }
 
 define void @beqz_i64(i64 %0) nounwind {
 ; LA32-LABEL: beqz_i64:
 ; LA32:       # %bb.0: # %start
-; LA32-NEXT:    addi.w $sp, $sp, -16
-; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
 ; LA32-NEXT:    or $a0, $a0, $a1
 ; LA32-NEXT:    beqz $a0, .LBB3_2
 ; LA32-NEXT:  # %bb.1: # %t
-; LA32-NEXT:    bl %plt(bar)
+; LA32-NEXT:    b %plt(bar)
 ; LA32-NEXT:  .LBB3_2: # %f
-; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
-; LA32-NEXT:    addi.w $sp, $sp, 16
 ; LA32-NEXT:    ret
 ;
 ; LA64-LABEL: beqz_i64:
 ; LA64:       # %bb.0: # %start
-; LA64-NEXT:    addi.d $sp, $sp, -16
-; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
 ; LA64-NEXT:    beqz $a0, .LBB3_2
 ; LA64-NEXT:  # %bb.1: # %t
-; LA64-NEXT:    bl %plt(bar)
+; LA64-NEXT:    b %plt(bar)
 ; LA64-NEXT:  .LBB3_2: # %f
-; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
-; LA64-NEXT:    addi.d $sp, $sp, 16
 ; LA64-NEXT:    ret
 start:
   %1 = icmp ne i64 %0, 0
   br i1 %1, label %t, label %f
 
 t:
   tail call void @bar()
   br label %f
 
 f:
   ret void
 }

llvm/test/CodeGen/LoongArch/duplicate-returns-for-tailcall.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc --mtriple=loongarch64 < %s | FileCheck %s
+
+;; Perform tail call optimization for duplicate returns.
+declare i32 @test()
+declare i32 @test1()
+declare i32 @test2()
+declare i32 @test3()
+define i32 @duplicate_returns(i32 %a, i32 %b) nounwind {
+; CHECK-LABEL: duplicate_returns:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    bstrpick.d $a2, $a0, 31, 0
+; CHECK-NEXT:    beqz $a2, .LBB0_4
+; CHECK-NEXT:  # %bb.1: # %if.else
+; CHECK-NEXT:    bstrpick.d $a2, $a1, 31, 0
+; CHECK-NEXT:    beqz $a2, .LBB0_5
+; CHECK-NEXT:  # %bb.2: # %if.else2
+; CHECK-NEXT:    addi.w $a0, $a0, 0
+; CHECK-NEXT:    addi.w $a1, $a1, 0
+; CHECK-NEXT:    bge $a1, $a0, .LBB0_6
+; CHECK-NEXT:  # %bb.3: # %if.then3
+; CHECK-NEXT:    b %plt(test2)
+; CHECK-NEXT:  .LBB0_4: # %if.then
+; CHECK-NEXT:    b %plt(test)
+; CHECK-NEXT:  .LBB0_5: # %if.then2
+; CHECK-NEXT:    b %plt(test1)
+; CHECK-NEXT:  .LBB0_6: # %if.else3
+; CHECK-NEXT:    b %plt(test3)
+entry:
+  %cmp = icmp eq i32 %a, 0
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %call = tail call i32 @test()
+  br label %return
+
+if.else:
+  %cmp1 = icmp eq i32 %b, 0
+  br i1 %cmp1, label %if.then2, label %if.else2
+
+if.then2:
+  %call1 = tail call i32 @test1()
+  br label %return
+
+if.else2:
+  %cmp5 = icmp sgt i32 %a, %b
+  br i1 %cmp5, label %if.then3, label %if.else3
+
+if.then3:
+  %call2 = tail call i32 @test2()
+  br label %return
+
+if.else3:
+  %call3 = tail call i32 @test3()
+  br label %return
+
+return:
+  %retval = phi i32 [ %call, %if.then ], [ %call1, %if.then2 ], [ %call2, %if.then3 ], [ %call3, %if.else3 ]
+  ret i32 %retval
+}