WholeProgramDevirt: Change internal vcall data structures to match summary.

[pcc]

Summary

Group calls into constant and non-constant arguments up front, and use uint64_t
instead of ConstantInt to represent constant arguments. The goal is to allow
the information from the summary to fit naturally into this data structure in
a future change (specifically, it will be added to CallSiteInfo).

This has two side effects:

• We disallow VCP for constant integer arguments of width >64 bits.
• We remove the restriction that the bitwidth of a vcall's argument and return types must match those of the vfunc definitions.

I don't expect either of these to matter in practice. The first case is
uncommon, and the second one will lead to UB (so we can do anything we like).

[pcc]

• Small tweak to make my next change NFC

[tejohnson]

llvm/lib/Transforms/IPO/WholeProgramDevirt.cpp

285

Document the new structures

291

Please document what this map is holding (vs the single CSInfo above). I believe I deduced it from code below, but better to document so one doesn't need to see how these are accessed to understand.

667

Why are the ZExt here and in below function now required?

llvm/test/Transforms/WholeProgramDevirt/vcp-too-wide-ints.ll

12

Why are these changes useful/necessary? Is there less coverage with the return type being dropped to i64?

llvm/test/Transforms/WholeProgramDevirt/vcp-type-mismatch.ll

34

Comment about why we can do optimization (function name still just says "bad_arg_type", which implies not being able to do it).

64

ditto

[pcc]

• Address review comments

llvm/lib/Transforms/IPO/WholeProgramDevirt.cpp

667

This one is needed because there is now the possibility of a mismatch between the return type of the function and the type of Cmp (always i1 before). I've added a test showing that we handle that case.

The one in tryVirtualConstProp is unnecessary, I've removed it.

llvm/test/Transforms/WholeProgramDevirt/vcp-too-wide-ints.ll

12

This change was to cover specifically the rejection due to the wide argument type. Good catch for the return type, I've re-added coverage for that.

llvm/test/Transforms/WholeProgramDevirt/vcp-type-mismatch.ll

34

Added a comment to the top of the file.

[pcc]

• Revert an irrelevant change

[pcc]

• Revert "Revert an irrelevant change", this was needed after all

[tejohnson]

LGTM

llvm/lib/Transforms/IPO/WholeProgramDevirt.cpp

void replaceAndErase(const Twine &OptName, const Twine &TargetName,
     }
     CS->eraseFromParent();
     // This use is no longer unsafe.
     if (NumUnsafeUses)
       --*NumUnsafeUses;
   }
 };
 
+// Call site information collected for a specific VTableSlot and possibly a list
+// of constant integer arguments. The grouping by arguments is handled by the
+// VTableSlotInfo class.
+struct CallSiteInfo {
+  std::vector<VirtualCallSite> CallSites;
+};
+
+// Call site information collected for a specific VTableSlot.
+struct VTableSlotInfo {
+  // The set of call sites which do not have all constant integer arguments
+  // (excluding "this").
+  CallSiteInfo CSInfo;
+
+  // The set of call sites with all constant integer arguments (excluding
+  // "this"), grouped by argument list.
+  std::map<std::vector<uint64_t>, CallSiteInfo> ConstCSInfo;
+
+  void addCallSite(Value *VTable, CallSite CS, unsigned *NumUnsafeUses);
+
+private:
+  CallSiteInfo &findCallSiteInfo(CallSite CS);
+};
+
+CallSiteInfo &VTableSlotInfo::findCallSiteInfo(CallSite CS) {
+  std::vector<uint64_t> Args;
+  auto *CI = dyn_cast<IntegerType>(CS.getType());
+  if (!CI || CI->getBitWidth() > 64)
+    return CSInfo;
+  for (auto &&Arg : make_range(CS.arg_begin() + 1, CS.arg_end())) {
+    auto *CI = dyn_cast<ConstantInt>(Arg);
+    if (!CI || CI->getBitWidth() > 64)
+      return CSInfo;
+    Args.push_back(CI->getZExtValue());
+  }
+  return ConstCSInfo[Args];
+}
+
+void VTableSlotInfo::addCallSite(Value *VTable, CallSite CS,
+                                 unsigned *NumUnsafeUses) {
+  findCallSiteInfo(CS).CallSites.push_back({VTable, CS, NumUnsafeUses});
+}
+
 struct DevirtModule {
   Module &M;
 
   PassSummaryAction Action;
   ModuleSummaryIndex *Summary;
 
   IntegerType *Int8Ty;
   PointerType *Int8PtrTy;
   IntegerType *Int32Ty;
 
   bool RemarksEnabled;
 
-  MapVector<VTableSlot, std::vector<VirtualCallSite>> CallSlots;
+  MapVector<VTableSlot, VTableSlotInfo> CallSlots;
 
   // This map keeps track of the number of "unsafe" uses of a loaded function
   // pointer. The key is the associated llvm.type.test intrinsic call generated
   // by this pass. An unsafe use is one that calls the loaded function pointer
   // directly. Every time we eliminate an unsafe use (for example, by
   // devirtualizing it or by applying virtual constant propagation), we
   // decrement the value stored in this map. If a value reaches zero, we can
   // eliminate the type check by RAUWing the associated llvm.type.test call with
void buildTypeIdentifierMap(
       std::vector<VTableBits> &Bits,
       DenseMap<Metadata *, std::set<TypeMemberInfo>> &TypeIdMap);
   Constant *getPointerAtOffset(Constant *I, uint64_t Offset);
   bool
   tryFindVirtualCallTargets(std::vector<VirtualCallTarget> &TargetsForSlot,
                             const std::set<TypeMemberInfo> &TypeMemberInfos,
                             uint64_t ByteOffset);
   bool trySingleImplDevirt(MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-                           MutableArrayRef<VirtualCallSite> CallSites);
+                           VTableSlotInfo &SlotInfo);
   bool tryEvaluateFunctionsWithArgs(
       MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-      ArrayRef<ConstantInt *> Args);
-  bool tryUniformRetValOpt(IntegerType *RetType,
-                           MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-                           MutableArrayRef<VirtualCallSite> CallSites);
+      ArrayRef<uint64_t> Args);
+  bool tryUniformRetValOpt(MutableArrayRef<VirtualCallTarget> TargetsForSlot,
+                           CallSiteInfo &CSInfo);
   bool tryUniqueRetValOpt(unsigned BitWidth,
                           MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-                          MutableArrayRef<VirtualCallSite> CallSites);
+                          CallSiteInfo &CSInfo);
   bool tryVirtualConstProp(MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-                           ArrayRef<VirtualCallSite> CallSites);
+                           VTableSlotInfo &SlotInfo);
 
   void rebuildGlobal(VTableBits &B);
 
   bool run();
 
   // Lower the module using the action and summary passed as command line
   // arguments. For testing purposes only.
   static bool runForTesting(Module &M);
bool DevirtModule::tryFindVirtualCallTargets(
   }
 
   // Give up if we couldn't find any targets.
   return !TargetsForSlot.empty();
 }
 
 bool DevirtModule::trySingleImplDevirt(
     MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-    MutableArrayRef<VirtualCallSite> CallSites) {
+    VTableSlotInfo &SlotInfo) {
   // See if the program contains a single implementation of this virtual
   // function.
   Function *TheFn = TargetsForSlot[0].Fn;
   for (auto &&Target : TargetsForSlot)
     if (TheFn != Target.Fn)
       return false;
 
   if (RemarksEnabled)
     TargetsForSlot[0].WasDevirt = true;
   // If so, update each call site to call that implementation directly.
-  for (auto &&VCallSite : CallSites) {
+  auto Apply = [&](CallSiteInfo &CSInfo) {
+    for (auto &&VCallSite : CSInfo.CallSites) {
       if (RemarksEnabled)
         VCallSite.emitRemark("single-impl", TheFn->getName());
       VCallSite.CS.setCalledFunction(ConstantExpr::getBitCast(
           TheFn, VCallSite.CS.getCalledValue()->getType()));
       // This use is no longer unsafe.
       if (VCallSite.NumUnsafeUses)
         --*VCallSite.NumUnsafeUses;
     }
+  };
+  Apply(SlotInfo.CSInfo);
+  for (auto &P : SlotInfo.ConstCSInfo)
+    Apply(P.second);
   return true;
 }
 
 bool DevirtModule::tryEvaluateFunctionsWithArgs(
     MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-    ArrayRef<ConstantInt *> Args) {
+    ArrayRef<uint64_t> Args) {
   // Evaluate each function and store the result in each target's RetVal
   // field.
   for (VirtualCallTarget &Target : TargetsForSlot) {
     if (Target.Fn->arg_size() != Args.size() + 1)
       return false;
-    for (unsigned I = 0; I != Args.size(); ++I)
-      if (Target.Fn->getFunctionType()->getParamType(I + 1) !=
-          Args[I]->getType())
-        return false;
 
     Evaluator Eval(M.getDataLayout(), nullptr);
     SmallVector<Constant *, 2> EvalArgs;
     EvalArgs.push_back(
         Constant::getNullValue(Target.Fn->getFunctionType()->getParamType(0)));
-    EvalArgs.insert(EvalArgs.end(), Args.begin(), Args.end());
+    for (unsigned I = 0; I != Args.size(); ++I) {
+      auto *ArgTy = dyn_cast<IntegerType>(
+          Target.Fn->getFunctionType()->getParamType(I + 1));
+      if (!ArgTy)
+        return false;
+      EvalArgs.push_back(ConstantInt::get(ArgTy, Args[I]));
+    }
+
     Constant *RetVal;
     if (!Eval.EvaluateFunction(Target.Fn, RetVal, EvalArgs) ||
         !isa<ConstantInt>(RetVal))
       return false;
     Target.RetVal = cast<ConstantInt>(RetVal)->getZExtValue();
   }
   return true;
 }
 
 bool DevirtModule::tryUniformRetValOpt(
-    IntegerType *RetType, MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-    MutableArrayRef<VirtualCallSite> CallSites) {
+    MutableArrayRef<VirtualCallTarget> TargetsForSlot, CallSiteInfo &CSInfo) {
   // Uniform return value optimization. If all functions return the same
   // constant, replace all calls with that constant.
   uint64_t TheRetVal = TargetsForSlot[0].RetVal;
   for (const VirtualCallTarget &Target : TargetsForSlot)
     if (Target.RetVal != TheRetVal)
       return false;
 
-  auto TheRetValConst = ConstantInt::get(RetType, TheRetVal);
-  for (auto Call : CallSites)
+  for (auto Call : CSInfo.CallSites)
     Call.replaceAndErase("uniform-ret-val", TargetsForSlot[0].Fn->getName(),
-                         RemarksEnabled, TheRetValConst);
+                         RemarksEnabled,
+                         ConstantInt::get(Call.CS->getType(), TheRetVal));
   if (RemarksEnabled)
     for (auto &&Target : TargetsForSlot)
       Target.WasDevirt = true;
   return true;
 }
 
 bool DevirtModule::tryUniqueRetValOpt(
     unsigned BitWidth, MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-    MutableArrayRef<VirtualCallSite> CallSites) {
+    CallSiteInfo &CSInfo) {
   // IsOne controls whether we look for a 0 or a 1.
   auto tryUniqueRetValOptFor = [&](bool IsOne) {
     const TypeMemberInfo *UniqueMember = nullptr;
     for (const VirtualCallTarget &Target : TargetsForSlot) {
       if (Target.RetVal == (IsOne ? 1 : 0)) {
         if (UniqueMember)
           return false;
         UniqueMember = Target.TM;
       }
     }
 
     // We should have found a unique member or bailed out by now. We already
     // checked for a uniform return value in tryUniformRetValOpt.
     assert(UniqueMember);
 
     // Replace each call with the comparison.
-    for (auto &&Call : CallSites) {
+    for (auto &&Call : CSInfo.CallSites) {
       IRBuilder<> B(Call.CS.getInstruction());
       Value *OneAddr = B.CreateBitCast(UniqueMember->Bits->GV, Int8PtrTy);
       OneAddr = B.CreateConstGEP1_64(OneAddr, UniqueMember->Offset);
       Value *Cmp = B.CreateICmp(IsOne ? ICmpInst::ICMP_EQ : ICmpInst::ICMP_NE,
                                 Call.VTable, OneAddr);
+      Cmp = B.CreateZExt(Cmp, Call.CS->getType());
       Call.replaceAndErase("unique-ret-val", TargetsForSlot[0].Fn->getName(),
                            RemarksEnabled, Cmp);
     }
     // Update devirtualization statistics for targets.
     if (RemarksEnabled)
       for (auto &&Target : TargetsForSlot)
         Target.WasDevirt = true;
 
     return true;
   };
 
   if (BitWidth == 1) {
     if (tryUniqueRetValOptFor(true))
       return true;
     if (tryUniqueRetValOptFor(false))
       return true;
   }
   return false;
 }
 
 bool DevirtModule::tryVirtualConstProp(
     MutableArrayRef<VirtualCallTarget> TargetsForSlot,
-    ArrayRef<VirtualCallSite> CallSites) {
+    VTableSlotInfo &SlotInfo) {
   // This only works if the function returns an integer.
   auto RetType = dyn_cast<IntegerType>(TargetsForSlot[0].Fn->getReturnType());
   if (!RetType)
     return false;
   unsigned BitWidth = RetType->getBitWidth();
   if (BitWidth > 64)
     return false;
 
   // Make sure that each function is defined, does not access memory, takes at
   // least one argument, does not use its first argument (which we assume is
   // 'this'), and has the same return type.
   for (VirtualCallTarget &Target : TargetsForSlot) {
     if (Target.Fn->isDeclaration() || !Target.Fn->doesNotAccessMemory() ||
         Target.Fn->arg_empty() || !Target.Fn->arg_begin()->use_empty() ||
         Target.Fn->getReturnType() != RetType)
       return false;
   }
 
-  // Group call sites by the list of constant arguments they pass.
-  // The comparator ensures deterministic ordering.
-  struct ByAPIntValue {
-    bool operator()(const std::vector<ConstantInt *> &A,
-                    const std::vector<ConstantInt *> &B) const {
-      return std::lexicographical_compare(
-          A.begin(), A.end(), B.begin(), B.end(),
-          [](ConstantInt *AI, ConstantInt *BI) {
-            return AI->getValue().ult(BI->getValue());
-          });
-    }
-  };
-  std::map<std::vector<ConstantInt *>, std::vector<VirtualCallSite>,
-           ByAPIntValue>
-      VCallSitesByConstantArg;
-  for (auto &&VCallSite : CallSites) {
-    std::vector<ConstantInt *> Args;
-    if (VCallSite.CS.getType() != RetType)
-      continue;
-    for (auto &&Arg :
-         make_range(VCallSite.CS.arg_begin() + 1, VCallSite.CS.arg_end())) {
-      if (!isa<ConstantInt>(Arg))
-        break;
-      Args.push_back(cast<ConstantInt>(&Arg));
-    }
-    if (Args.size() + 1 != VCallSite.CS.arg_size())
-      continue;
-
-    VCallSitesByConstantArg[Args].push_back(VCallSite);
-  }
-
-  for (auto &&CSByConstantArg : VCallSitesByConstantArg) {
+  for (auto &&CSByConstantArg : SlotInfo.ConstCSInfo) {
     if (!tryEvaluateFunctionsWithArgs(TargetsForSlot, CSByConstantArg.first))
       continue;
 
-    if (tryUniformRetValOpt(RetType, TargetsForSlot, CSByConstantArg.second))
+    if (tryUniformRetValOpt(TargetsForSlot, CSByConstantArg.second))
       continue;
 
     if (tryUniqueRetValOpt(BitWidth, TargetsForSlot, CSByConstantArg.second))
       continue;
 
     // Find an allocation offset in bits in all vtables associated with the
     // type.
     uint64_t AllocBefore =
else
       setAfterReturnValues(TargetsForSlot, AllocAfter, BitWidth, OffsetByte,
                            OffsetBit);
 
     if (RemarksEnabled)
       for (auto &&Target : TargetsForSlot)
         Target.WasDevirt = true;
 
     // Rewrite each call to a load from OffsetByte/OffsetBit.
-    for (auto Call : CSByConstantArg.second) {
+    for (auto Call : CSByConstantArg.second.CallSites) {
+      auto *CSRetType = cast<IntegerType>(Call.CS.getType());
       IRBuilder<> B(Call.CS.getInstruction());
       Value *Addr = B.CreateConstGEP1_64(Call.VTable, OffsetByte);
-      if (BitWidth == 1) {
+      if (CSRetType->getBitWidth() == 1) {
         Value *Bits = B.CreateLoad(Addr);
         Value *Bit = ConstantInt::get(Int8Ty, 1ULL << OffsetBit);
         Value *BitsAndBit = B.CreateAnd(Bits, Bit);
-        auto IsBitSet = B.CreateICmpNE(BitsAndBit, ConstantInt::get(Int8Ty, 0));
+        Value *IsBitSet =
+            B.CreateICmpNE(BitsAndBit, ConstantInt::get(Int8Ty, 0));
         Call.replaceAndErase("virtual-const-prop-1-bit",
                              TargetsForSlot[0].Fn->getName(),
                              RemarksEnabled, IsBitSet);
       } else {
-        Value *ValAddr = B.CreateBitCast(Addr, RetType->getPointerTo());
-        Value *Val = B.CreateLoad(RetType, ValAddr);
+        Value *ValAddr = B.CreateBitCast(Addr, CSRetType->getPointerTo());
+        Value *Val = B.CreateLoad(CSRetType, ValAddr);
         Call.replaceAndErase("virtual-const-prop",
                              TargetsForSlot[0].Fn->getName(),
                              RemarksEnabled, Val);
       }
     }
   }
   return true;
 }
for (auto I = TypeTestFunc->use_begin(), E = TypeTestFunc->use_end();
     // the vtable pointer before, as it may have been CSE'd with pointers from
     // other call sites, and we don't want to process call sites multiple times.
     if (!Assumes.empty()) {
       Metadata *TypeId =
           cast<MetadataAsValue>(CI->getArgOperand(1))->getMetadata();
       Value *Ptr = CI->getArgOperand(0)->stripPointerCasts();
       if (SeenPtrs.insert(Ptr).second) {
         for (DevirtCallSite Call : DevirtCalls) {
-          CallSlots[{TypeId, Call.Offset}].push_back(
-              {CI->getArgOperand(0), Call.CS, nullptr});
+          CallSlots[{TypeId, Call.Offset}].addCallSite(CI->getArgOperand(0),
+                                                       Call.CS, nullptr);
         }
       }
     }
 
     // We no longer need the assumes or the type test.
     for (auto Assume : Assumes)
       Assume->eraseFromParent();
     // We can't use RecursivelyDeleteTriviallyDeadInstructions here because we
for (auto I = TypeCheckedLoadFunc->use_begin(),
     NumUnsafeUses = DevirtCalls.size();
 
     // If the function pointer has a non-call user, we cannot eliminate the type
     // check, as one of those users may eventually call the pointer. Increment
     // the unsafe use count to make sure it cannot reach zero.
     if (HasNonCallUses)
       ++NumUnsafeUses;
     for (DevirtCallSite Call : DevirtCalls) {
-      CallSlots[{TypeId, Call.Offset}].push_back(
-          {Ptr, Call.CS, &NumUnsafeUses});
+      CallSlots[{TypeId, Call.Offset}].addCallSite(Ptr, Call.CS,
+                                                   &NumUnsafeUses);
     }
 
     CI->eraseFromParent();
   }
 }
 
 bool DevirtModule::run() {
   Function *TypeTestFunc =

llvm/test/Transforms/WholeProgramDevirt/unique-retval.ll

define i1 @call1(i8* %obj) {
   %fptr = load i8*, i8** %fptrptr
   %fptr_casted = bitcast i8* %fptr to i1 (i8*)*
   ; CHECK: [[RES1:%[^ ]*]] = icmp eq i8* [[VT1]], bitcast ([1 x i8*]* @vt3 to i8*)
   %result = call i1 %fptr_casted(i8* %obj)
   ; CHECK: ret i1 [[RES1]]
   ret i1 %result
 }
 
-; CHECK: define i1 @call2
-define i1 @call2(i8* %obj) {
+; CHECK: define i32 @call2
+define i32 @call2(i8* %obj) {
   %vtableptr = bitcast i8* %obj to [1 x i8*]**
   %vtable = load [1 x i8*]*, [1 x i8*]** %vtableptr
   ; CHECK: [[VT2:%[^ ]*]] = bitcast [1 x i8*]* {{.*}} to i8*
   %vtablei8 = bitcast [1 x i8*]* %vtable to i8*
   %p = call i1 @llvm.type.test(i8* %vtablei8, metadata !"typeid2")
   call void @llvm.assume(i1 %p)
   %fptrptr = getelementptr [1 x i8*], [1 x i8*]* %vtable, i32 0, i32 0
   %fptr = load i8*, i8** %fptrptr
-  %fptr_casted = bitcast i8* %fptr to i1 (i8*)*
-  ; CHECK: [[RES1:%[^ ]*]] = icmp ne i8* [[VT1]], bitcast ([1 x i8*]* @vt2 to i8*)
-  %result = call i1 %fptr_casted(i8* %obj)
-  ret i1 %result
+  ; Intentional type mismatch to test zero extend.
+  %fptr_casted = bitcast i8* %fptr to i32 (i8*)*
+  ; CHECK: [[RES2:%[^ ]*]] = icmp ne i8* [[VT1]], bitcast ([1 x i8*]* @vt2 to i8*)
+  %result = call i32 %fptr_casted(i8* %obj)
+  ; CHECK: [[ZEXT2:%[^ ]*]] = zext i1 [[RES2]] to i32
+  ; CHECK: ret i32 [[ZEXT2:%[^ ]*]]
+  ret i32 %result
 }
 
 declare i1 @llvm.type.test(i8*, metadata)
 declare void @llvm.assume(i1)
 
 !0 = !{i32 0, !"typeid1"}
 !1 = !{i32 0, !"typeid2"}

llvm/test/Transforms/WholeProgramDevirt/vcp-too-wide-ints.ll

 ; RUN: opt -S -wholeprogramdevirt %s | FileCheck %s
 
 target datalayout = "e-p:64:64"
 target triple = "x86_64-unknown-linux-gnu"
 
-@vt1 = constant [1 x i8*] [i8* bitcast (i128 (i8*, i128)* @vf1 to i8*)], !type !0
-@vt2 = constant [1 x i8*] [i8* bitcast (i128 (i8*, i128)* @vf2 to i8*)], !type !0
+@vt1 = constant [1 x i8*] [i8* bitcast (i64 (i8*, i128)* @vf1 to i8*)], !type !0
+@vt2 = constant [1 x i8*] [i8* bitcast (i64 (i8*, i128)* @vf2 to i8*)], !type !0
+@vt3 = constant [1 x i8*] [i8* bitcast (i128 (i8*, i64)* @vf3 to i8*)], !type !1
+@vt4 = constant [1 x i8*] [i8* bitcast (i128 (i8*, i64)* @vf4 to i8*)], !type !1
+
+define i64 @vf1(i8* %this, i128 %arg) readnone {
+  %argtrunc = trunc i128 %arg to i64
+  ret i64 %argtrunc
+}
+
+define i64 @vf2(i8* %this, i128 %arg) readnone {
+  %argtrunc = trunc i128 %arg to i64
+  ret i64 %argtrunc
+}
 
-define i128 @vf1(i8* %this, i128 %arg) readnone {
-  ret i128 %arg
+define i128 @vf3(i8* %this, i64 %arg) readnone {
+  %argzext = zext i64 %arg to i128
+  ret i128 %argzext
 }
 
-define i128 @vf2(i8* %this, i128 %arg) readnone {
-  ret i128 %arg
+define i128 @vf4(i8* %this, i64 %arg) readnone {
+  %argzext = zext i64 %arg to i128
+  ret i128 %argzext
+}
+
+; CHECK: define i64 @call1
+define i64 @call1(i8* %obj) {
+  %vtableptr = bitcast i8* %obj to [1 x i8*]**
+  %vtable = load [1 x i8*]*, [1 x i8*]** %vtableptr
+  %vtablei8 = bitcast [1 x i8*]* %vtable to i8*
+  %p = call i1 @llvm.type.test(i8* %vtablei8, metadata !"typeid1")
+  call void @llvm.assume(i1 %p)
+  %fptrptr = getelementptr [1 x i8*], [1 x i8*]* %vtable, i32 0, i32 0
+  %fptr = load i8*, i8** %fptrptr
+  %fptr_casted = bitcast i8* %fptr to i64 (i8*, i128)*
+  ; CHECK: call i64 %
+  %result = call i64 %fptr_casted(i8* %obj, i128 1)
+  ret i64 %result
 }
 
-; CHECK: define i128 @call
-define i128 @call(i8* %obj) {
+; CHECK: define i128 @call2
+define i128 @call2(i8* %obj) {
   %vtableptr = bitcast i8* %obj to [1 x i8*]**
   %vtable = load [1 x i8*]*, [1 x i8*]** %vtableptr
   %vtablei8 = bitcast [1 x i8*]* %vtable to i8*
-  %p = call i1 @llvm.type.test(i8* %vtablei8, metadata !"typeid")
+  %p = call i1 @llvm.type.test(i8* %vtablei8, metadata !"typeid2")
   call void @llvm.assume(i1 %p)
   %fptrptr = getelementptr [1 x i8*], [1 x i8*]* %vtable, i32 0, i32 0
   %fptr = load i8*, i8** %fptrptr
-  %fptr_casted = bitcast i8* %fptr to i128 (i8*, i128)*
+  %fptr_casted = bitcast i8* %fptr to i128 (i8*, i64)*
   ; CHECK: call i128 %
-  %result = call i128 %fptr_casted(i8* %obj, i128 1)
+  %result = call i128 %fptr_casted(i8* %obj, i64 1)
   ret i128 %result
 }
 
 declare i1 @llvm.type.test(i8*, metadata)
 declare void @llvm.assume(i1)
 
-!0 = !{i32 0, !"typeid"}
+!0 = !{i32 0, !"typeid1"}
+!1 = !{i32 0, !"typeid2"}

llvm/test/Transforms/WholeProgramDevirt/vcp-type-mismatch.ll

 ; RUN: opt -S -wholeprogramdevirt %s | FileCheck %s
 
+; Test that we correctly handle function type mismatches in argument counts
+; and bitwidths. We handle an argument count mismatch by refusing
+; to optimize. For bitwidth mismatches, we allow the optimization in order
+; to simplify the implementation. This is legal because the bitwidth mismatch
+; gives the call undefined behavior.
+
 target datalayout = "e-p:64:64"
 target triple = "x86_64-unknown-linux-gnu"
 
 @vt1 = constant [1 x i8*] [i8* bitcast (i32 (i8*, i32)* @vf1 to i8*)], !type !0
 @vt2 = constant [1 x i8*] [i8* bitcast (i32 (i8*, i32)* @vf2 to i8*)], !type !0
 
 define i32 @vf1(i8* %this, i32 %arg) readnone {
   ret i32 %arg
 }
 
 define i32 @vf2(i8* %this, i32 %arg) readnone {
   ret i32 %arg
 }
 
 ; CHECK: define i32 @bad_arg_type
 define i32 @bad_arg_type(i8* %obj) {
   %vtableptr = bitcast i8* %obj to [1 x i8*]**
   %vtable = load [1 x i8*]*, [1 x i8*]** %vtableptr
   %vtablei8 = bitcast [1 x i8*]* %vtable to i8*
   %p = call i1 @llvm.type.test(i8* %vtablei8, metadata !"typeid")
   call void @llvm.assume(i1 %p)
   %fptrptr = getelementptr [1 x i8*], [1 x i8*]* %vtable, i32 0, i32 0
   %fptr = load i8*, i8** %fptrptr
   %fptr_casted = bitcast i8* %fptr to i32 (i8*, i64)*
-  ; CHECK: call i32 %
   %result = call i32 %fptr_casted(i8* %obj, i64 1)
+  ; CHECK: ret i32 1
   ret i32 %result
 }
 
 ; CHECK: define i32 @bad_arg_count
 define i32 @bad_arg_count(i8* %obj) {
   %vtableptr = bitcast i8* %obj to [1 x i8*]**
   %vtable = load [1 x i8*]*, [1 x i8*]** %vtableptr
   %vtablei8 = bitcast [1 x i8*]* %vtable to i8*
define i64 @bad_return_type(i8* %obj) {
   %vtableptr = bitcast i8* %obj to [1 x i8*]**
   %vtable = load [1 x i8*]*, [1 x i8*]** %vtableptr
   %vtablei8 = bitcast [1 x i8*]* %vtable to i8*
   %p = call i1 @llvm.type.test(i8* %vtablei8, metadata !"typeid")
   call void @llvm.assume(i1 %p)
   %fptrptr = getelementptr [1 x i8*], [1 x i8*]* %vtable, i32 0, i32 0
   %fptr = load i8*, i8** %fptrptr
   %fptr_casted = bitcast i8* %fptr to i64 (i8*, i32)*
-  ; CHECK: call i64 %
   %result = call i64 %fptr_casted(i8* %obj, i32 1)
+  ; CHECK: ret i64 1
   ret i64 %result
 }
 
 declare i1 @llvm.type.test(i8*, metadata)
 declare void @llvm.assume(i1)
 
 !0 = !{i32 0, !"typeid"}