[FPEnv][EarlyCSE] Add support for CSE of constrained FP intrinsics

[kpn]

Summary

Currently we only CSE constrained FP intrinsics in the default floating point environment.

This patch adds support for CSE when exception behavior is "maytrap". No CSE of constrained intrinsics will happen across a function call or other unsafe event. Since any function call can change the floating point status register, the restrictions on exception behavior are required.

[kpn]

This patch may be overkill. If the reading of the floating point status register is restricted to "fpexcept.strict" then perhaps we don't need all of this code. But if we want to allow checking the status register after "fpexcept.maytrap" instructions then we do need all this. The LangRef restricts those reads to "fpexcept.strict". Are we certain we want to stick with that language?

[andrew.w.kaylor]

In D105895#2874128, @kpn wrote:

This patch may be overkill. If the reading of the floating point status register is restricted to "fpexcept.strict" then perhaps we don't need all of this code. But if we want to allow checking the status register after "fpexcept.maytrap" instructions then we do need all this. The LangRef restricts those reads to "fpexcept.strict". Are we certain we want to stick with that language?

Where does the LangRef say this?

I think reading the status register should be allowed with fpexcept.maytrap, but the results of such a test must be understood to apply only to the generated code, not the strict semantics of the source. That is, if I check the status register and see, for example, that the DIVZERO flag is set, I can be sure that something in my code caused that, but if I check the register and no exception flags are set, that might have been because code that would have caused an exception was optimized away. The "maytrap" state is intended to avoid the introduction of spurious exceptions, but it isn't guaranteed to preserve all existing exceptions.

Strictly speaking, this is an off-the-map option. The relevant standards don't let you test the FP status flag or unmask exceptions unless fenv access is enabled, which leads to the fpexcept.strict behavior. The "maytrap" mode is meant to enable some useful optimizations, like DCE of code that might contain FP exceptions, while preventing optimizations that could cause rogue FP exceptions through speculative execution.

[kpn]

OK, this is overkill per Andy's comment. I'll update.

[kpn]

Replaced with D112256. This patch here has restrictions that are not needed. To avoid confusing reading I'm replacing the review.

llvm/lib/Transforms/Scalar/EarlyCSE.cpp

 // SimpleValue
 //===----------------------------------------------------------------------===//
 
 namespace {
 
 /// Struct representing the available values in the scoped hash table.
 struct SimpleValue {
   Instruction *Inst;
+  unsigned StrictFPGeneration;
 
-  SimpleValue(Instruction *I) : Inst(I) {
+  SimpleValue(Instruction *I) : Inst(I), StrictFPGeneration(0) {
+    assert((isSentinel() || canHandle(I)) && "Inst can't be handled!");
+  }
+
+  SimpleValue(Instruction *I, unsigned StrictFPGeneration)
+      : Inst(I), StrictFPGeneration(StrictFPGeneration) {
     assert((isSentinel() || canHandle(I)) && "Inst can't be handled!");
   }
 
   bool isSentinel() const {
     return Inst == DenseMapInfo<Instruction *>::getEmptyKey() ||
            Inst == DenseMapInfo<Instruction *>::getTombstoneKey();
   }
 
if (CallInst *CI = dyn_cast<CallInst>(Inst)) {
         case Intrinsic::experimental_constrained_frem:
         case Intrinsic::experimental_constrained_fptosi:
         case Intrinsic::experimental_constrained_sitofp:
         case Intrinsic::experimental_constrained_fptoui:
         case Intrinsic::experimental_constrained_uitofp:
         case Intrinsic::experimental_constrained_fcmp:
         case Intrinsic::experimental_constrained_fcmps: {
           auto *CFP = cast<ConstrainedFPIntrinsic>(CI);
-          return CFP->isDefaultFPEnvironment();
+          if (CFP->getExceptionBehavior().getValue() == fp::ebStrict)
+             return false;
+          return true;
         }
         }
       }
       return CI->doesNotAccessMemory() && !CI->getType()->isVoidTy();
     }
     return isa<CastInst>(Inst) || isa<UnaryOperator>(Inst) ||
            isa<BinaryOperator>(Inst) || isa<GetElementPtrInst>(Inst) ||
            isa<CmpInst>(Inst) || isa<SelectInst>(Inst) ||
            isa<ExtractElementInst>(Inst) || isa<InsertElementInst>(Inst) ||
            isa<ShuffleVectorInst>(Inst) || isa<ExtractValueInst>(Inst) ||
            isa<InsertValueInst>(Inst) || isa<FreezeInst>(Inst);
   }
+  static bool canHandle(Instruction *Inst,
+                        enum fp::ExceptionBehavior MaxExceptBehavior) {
+    bool isOK = canHandle(Inst);
+    if (isOK) {
+      if (auto *FPI = dyn_cast<ConstrainedFPIntrinsic>(Inst)) {
+        Optional<fp::ExceptionBehavior> InstExceptBehavior =
+            FPI->getExceptionBehavior();
+        if (InstExceptBehavior) {
+          switch (InstExceptBehavior.getValue()) {
+          case fp::ebIgnore:
+            // The weakest value is always fine. It is equivalent to a
+            // normal FP instruction.
+            break;
+          case fp::ebMayTrap:
+            isOK = (MaxExceptBehavior == fp::ebMayTrap ||
+                    MaxExceptBehavior == fp::ebStrict);
+            break;
+          case fp::ebStrict:
+            isOK = (MaxExceptBehavior == fp::ebStrict);
+            break;
+          }
+        }
+      }
+    }
+    return isOK;
+  }
 };
 
 } // end anonymous namespace
 
 namespace llvm {
 
 template <> struct DenseMapInfo<SimpleValue> {
   static inline SimpleValue getEmptyKey() {

 }
 
 static bool isEqualImpl(SimpleValue LHS, SimpleValue RHS) {
   Instruction *LHSI = LHS.Inst, *RHSI = RHS.Inst;
 
   if (LHS.isSentinel() || RHS.isSentinel())
     return LHSI == RHSI;
 
+  // Non-constrained anything will have a StrictFPGeneration of zero.
+  if (LHS.StrictFPGeneration != RHS.StrictFPGeneration)
+    return false;
   if (LHSI->getOpcode() != RHSI->getOpcode())
     return false;
   if (LHSI->isIdenticalToWhenDefined(RHSI))
     return true;
 
   // If we're not strictly identical, we still might be a commutable instruction
   if (BinaryOperator *LHSBinOp = dyn_cast<BinaryOperator>(LHSI)) {
     if (!LHSBinOp->isCommutative())
if (CmpInst *LHSCmp = dyn_cast<CmpInst>(LHSI)) {
     CmpInst *RHSCmp = cast<CmpInst>(RHSI);
     // Commuted equality
     return LHSCmp->getOperand(0) == RHSCmp->getOperand(1) &&
            LHSCmp->getOperand(1) == RHSCmp->getOperand(0) &&
            LHSCmp->getSwappedPredicate() == RHSCmp->getPredicate();
   }
 
   // TODO: Extend this for >2 args by matching the trailing N-2 args.
+  // TODO: Constrained FP intrinsics require matching the first two args.
   auto *LII = dyn_cast<IntrinsicInst>(LHSI);
   auto *RII = dyn_cast<IntrinsicInst>(RHSI);
   if (LII && RII && LII->getIntrinsicID() == RII->getIntrinsicID() &&
       LII->isCommutative() && LII->getNumArgOperands() == 2) {
     return LII->getArgOperand(0) == RII->getArgOperand(1) &&
            LII->getArgOperand(1) == RII->getArgOperand(0);
   }
 

 public:
   const TargetLibraryInfo &TLI;
   const TargetTransformInfo &TTI;
   DominatorTree &DT;
   AssumptionCache &AC;
   const SimplifyQuery SQ;
   MemorySSA *MSSA;
   std::unique_ptr<MemorySSAUpdater> MSSAUpdater;
+  bool isFunctionStrictFP;
 
   using AllocatorTy =
       RecyclingAllocator<BumpPtrAllocator,
                          ScopedHashTableVal<SimpleValue, Value *>>;
   using ScopedHTType =
       ScopedHashTable<SimpleValue, Value *, DenseMapInfo<SimpleValue>,
                       AllocatorTy>;
 
public:
   /// It uses the same generation count as loads.
   using CallHTType =
       ScopedHashTable<CallValue, std::pair<Instruction *, unsigned>>;
   CallHTType AvailableCalls;
 
   /// This is the current generation of the memory value.
   unsigned CurrentGeneration = 0;
 
+  /// This is the current generation of the memory value, taking into account
+  /// the crossing of function calls.
+  unsigned StrictFPGeneration = 0;
+
   /// Set up the EarlyCSE runner for a particular function.
   EarlyCSE(const DataLayout &DL, const TargetLibraryInfo &TLI,
            const TargetTransformInfo &TTI, DominatorTree &DT,
-           AssumptionCache &AC, MemorySSA *MSSA)
+           AssumptionCache &AC, MemorySSA *MSSA, bool isFunctionStrictFP)
       : TLI(TLI), TTI(TTI), DT(DT), AC(AC), SQ(DL, &TLI, &DT, &AC), MSSA(MSSA),
-        MSSAUpdater(std::make_unique<MemorySSAUpdater>(MSSA)) {}
+        MSSAUpdater(std::make_unique<MemorySSAUpdater>(MSSA)), isFunctionStrictFP(isFunctionStrictFP) {}
 
   bool run();
 
 private:
   unsigned ClobberCounter = 0;
   // Almost a POD, but needs to call the constructors for the scoped hash
   // tables so that a new scope gets pushed on. These are RAII so that the
   // scope gets popped when the NodeScope is destroyed.
private:
   // Contains all the needed information to create a stack for doing a depth
   // first traversal of the tree. This includes scopes for values, loads, and
   // calls as well as the generation. There is a child iterator so that the
   // children do not need to be store separately.
   class StackNode {
   public:
     StackNode(ScopedHTType &AvailableValues, LoadHTType &AvailableLoads,
               InvariantHTType &AvailableInvariants, CallHTType &AvailableCalls,
-              unsigned cg, DomTreeNode *n, DomTreeNode::const_iterator child,
+              unsigned cg, unsigned sfpcg, DomTreeNode *n,
+              DomTreeNode::const_iterator child,
               DomTreeNode::const_iterator end)
-        : CurrentGeneration(cg), ChildGeneration(cg), Node(n), ChildIter(child),
-          EndIter(end),
-          Scopes(AvailableValues, AvailableLoads, AvailableInvariants,
-                 AvailableCalls)
-          {}
+        : CurrentGeneration(cg), ChildGeneration(cg), StrictFPGeneration(sfpcg),
+          ChildStrictFPGeneration(sfpcg), Node(n), ChildIter(child),
+          EndIter(end), Scopes(AvailableValues, AvailableLoads,
+                               AvailableInvariants, AvailableCalls) {}
     StackNode(const StackNode &) = delete;
     StackNode &operator=(const StackNode &) = delete;
 
     // Accessors.
     unsigned currentGeneration() const { return CurrentGeneration; }
     unsigned childGeneration() const { return ChildGeneration; }
     void childGeneration(unsigned generation) { ChildGeneration = generation; }
+    unsigned strictFPGeneration() const { return StrictFPGeneration; }
+    unsigned childStrictFPGeneration() const { return ChildStrictFPGeneration; }
+    void childStrictFPGeneration(unsigned generation) {
+      ChildStrictFPGeneration = generation;
+    }
     DomTreeNode *node() { return Node; }
     DomTreeNode::const_iterator childIter() const { return ChildIter; }
 
     DomTreeNode *nextChild() {
       DomTreeNode *child = *ChildIter;
       ++ChildIter;
       return child;
     }
 
     DomTreeNode::const_iterator end() const { return EndIter; }
     bool isProcessed() const { return Processed; }
     void process() { Processed = true; }
 
   private:
     unsigned CurrentGeneration;
     unsigned ChildGeneration;
+    unsigned StrictFPGeneration;
+    unsigned ChildStrictFPGeneration;
     DomTreeNode *Node;
     DomTreeNode::const_iterator ChildIter;
     DomTreeNode::const_iterator EndIter;
     NodeScope Scopes;
     bool Processed = false;
   };
 
   /// Wrapper class to handle memory instructions, including loads,
private:
   }
 
   bool processNode(DomTreeNode *Node);
 
   bool handleBranchCondition(Instruction *CondInst, const BranchInst *BI,
                              const BasicBlock *BB, const BasicBlock *Pred);
 
   Value *getMatchingValue(LoadValue &InVal, ParseMemoryInst &MemInst,
-                          unsigned CurrentGeneration);
+                          unsigned CurrentGeneration,
+                          unsigned StrictFPGeneration);
 
   bool overridingStores(const ParseMemoryInst &Earlier,
                         const ParseMemoryInst &Later);
 
   Value *getOrCreateResult(Value *Inst, Type *ExpectedType) const {
     if (auto *LI = dyn_cast<LoadInst>(Inst))
       return LI;
     if (auto *SI = dyn_cast<StoreInst>(Inst))
private:
 
   /// Return true if the instruction is known to only operate on memory
   /// provably invariant in the given "generation".
   bool isOperatingOnInvariantMemAt(Instruction *I, unsigned GenAt);
 
   bool isSameMemGeneration(unsigned EarlierGeneration, unsigned LaterGeneration,
                            Instruction *EarlierInst, Instruction *LaterInst);
 
+  bool isSameStrictFPGeneration(unsigned EarlierGeneration,
+                                unsigned LaterGeneration,
+                                Instruction *EarlierInst,
+                                Instruction *LaterInst);
+
+  unsigned getStrictFPGeneration(Instruction *I);
+
   bool isNonTargetIntrinsicMatch(const IntrinsicInst *Earlier,
                                  const IntrinsicInst *Later) {
     auto IsSubmask = [](const Value *Mask0, const Value *Mask1) {
       // Is Mask0 a submask of Mask1?
       if (Mask0 == Mask1)
         return true;
       if (isa<UndefValue>(Mask0) || isa<UndefValue>(Mask1))
         return false;
if (ClobberCounter < EarlyCSEMssaOptCap) {
     LaterDef = MSSA->getWalker()->getClobberingMemoryAccess(LaterInst);
     ClobberCounter++;
   } else
     LaterDef = LaterMA->getDefiningAccess();
 
   return MSSA->dominates(LaterDef, EarlierMA);
 }
 
+/// Determine if the two instructions are together where CSE is safe.
+/// This is before any function calls, just before a return, or between
+/// two adjacent function calls. The two adjacent function calls may be
+/// on different basic blocks only if there is exactly one path from the
+/// first function call to the second function call.
+bool EarlyCSE::isSameStrictFPGeneration(unsigned EarlierGeneration,
+                                        unsigned LaterGeneration,
+                                        Instruction *EarlierInst,
+                                        Instruction *LaterInst) {
+  return (EarlierGeneration == LaterGeneration);
+}
+
+/// Return the value used for isSameStrictFPGeneration(), and for
+/// instructions where strictfp is not used return zero.
+unsigned EarlyCSE::getStrictFPGeneration(Instruction *I) {
+  if (ConstrainedFPIntrinsic *CI = dyn_cast<ConstrainedFPIntrinsic>(I)) {
+    if (CI->isDefaultFPEnvironment())
+      return 0;
+    return StrictFPGeneration;
+  }
+  return 0;
+}
+
 bool EarlyCSE::isOperatingOnInvariantMemAt(Instruction *I, unsigned GenAt) {
   // A location loaded from with an invariant_load is assumed to *never* change
   // within the visible scope of the compilation.
   if (auto *LI = dyn_cast<LoadInst>(I))
     if (LI->hasMetadata(LLVMContext::MD_invariant_load))
       return true;
 
   auto MemLocOpt = MemoryLocation::getOrNone(I);
if (MatchBinOp(Curr, PropagateOpcode, LHS, RHS))
           if (SimpleValue::canHandle(OPI) && Visited.insert(OPI).second)
             WorkList.push_back(OPI);
   }
 
   return MadeChanges;
 }
 
 Value *EarlyCSE::getMatchingValue(LoadValue &InVal, ParseMemoryInst &MemInst,
-                                  unsigned CurrentGeneration) {
+                                  unsigned CurrentGeneration,
+                                  unsigned StrictFPGeneration) {
   if (InVal.DefInst == nullptr)
     return nullptr;
   if (InVal.MatchingId != MemInst.getMatchingId())
     return nullptr;
   // We don't yet handle removing loads with ordering of any kind.
   if (MemInst.isVolatile() || !MemInst.isUnordered())
     return nullptr;
   // We can't replace an atomic load with one which isn't also atomic.
bool EarlyCSE::processNode(DomTreeNode *Node) {
   BasicBlock *BB = Node->getBlock();
 
   // If this block has a single predecessor, then the predecessor is the parent
   // of the domtree node and all of the live out memory values are still current
   // in this block.  If this block has multiple predecessors, then they could
   // have invalidated the live-out memory values of our parent value.  For now,
   // just be conservative and invalidate memory if this block has multiple
   // predecessors.
-  if (!BB->getSinglePredecessor())
+  if (!BB->getSinglePredecessor()) {
     ++CurrentGeneration;
+    if (StrictFPGeneration > 0) {
+      ++StrictFPGeneration;
+    }
+  }
 
   // If this node has a single predecessor which ends in a conditional branch,
   // we can infer the value of the branch condition given that we took this
   // path.  We need the single predecessor to ensure there's not another path
   // which reaches this block where the condition might hold a different
   // value.  Since we're adding this to the scoped hash table (like any other
   // def), it will have been popped if we encounter a future merge block.
   if (BasicBlock *Pred = BB->getSinglePredecessor()) {
if (Value *V = SimplifyInstruction(&Inst, SQ)) {
         if (Changed)
           ++NumSimplify;
         if (Killed)
           continue;
       }
     }
 
     // If this is a simple instruction that we can value number, process it.
+    Value *DeferredElideWith = nullptr;
     if (SimpleValue::canHandle(&Inst)) {
+      bool ConstrainedFPMayWriteMemory = false;
+      if (auto *CI = dyn_cast<ConstrainedFPIntrinsic>(&Inst)) {
+        assert(CI->getExceptionBehavior().getValue() != fp::ebStrict &&
+               "Unexpected ebStrict from SimpleValue::canHandle()");
+        if (CI->getExceptionBehavior().getValue() == fp::ebMayTrap) {
+          // Can write to memory in, for example, an exception handler.
+          ConstrainedFPMayWriteMemory = true;
+        }
+      }
       // See if the instruction has an available value.  If so, use it.
-      if (Value *V = AvailableValues.lookup(&Inst)) {
-        LLVM_DEBUG(dbgs() << "EarlyCSE CSE: " << Inst << "  to: " << *V
-                          << '\n');
+      if (Value *V = AvailableValues.lookup(
+              SimpleValue(&Inst, getStrictFPGeneration(&Inst)))) {
         if (!DebugCounter::shouldExecute(CSECounter)) {
           LLVM_DEBUG(dbgs() << "Skipping due to debug counter\n");
           continue;
         }
+        if (ConstrainedFPMayWriteMemory) {
+          DeferredElideWith = V;
+          LLVM_DEBUG(dbgs() << "EarlyCSE CSE Deferred: " << V << " for " << Inst
+                            << "  to: " << '\n');
+        } else {
+          LLVM_DEBUG(dbgs()
+                     << "EarlyCSE CSE: " << Inst << "  to: " << *V << '\n');
           if (auto *I = dyn_cast<Instruction>(V))
             I->andIRFlags(&Inst);
           Inst.replaceAllUsesWith(V);
           salvageKnowledge(&Inst, &AC);
           removeMSSA(Inst);
           Inst.eraseFromParent();
           Changed = true;
           ++NumCSE;
           continue;
         }
+      }
 
       // Otherwise, just remember that this value is available.
-      AvailableValues.insert(&Inst, &Inst);
+      if (!DeferredElideWith) {
+        AvailableValues.insert(SimpleValue(&Inst, getStrictFPGeneration(&Inst)),
+                               &Inst);
         continue;
       }
+    }
+    else {
+      if (Inst.getOpcode() == Instruction::Call ||
+          Inst.getOpcode() == Instruction::Invoke) {
+        if (StrictFPGeneration > 0)
+          ++StrictFPGeneration;
+      }
+    }
 
     ParseMemoryInst MemInst(&Inst, TTI);
     // If this is a non-volatile load, process it.
-    if (MemInst.isValid() && MemInst.isLoad()) {
+    if (MemInst.isValid() && MemInst.isLoad() && !DeferredElideWith) {
       // (conservatively) we can't peak past the ordering implied by this
       // operation, but we can add this load to our set of available values
       if (MemInst.isVolatile() || !MemInst.isUnordered()) {
         LastStore = nullptr;
         ++CurrentGeneration;
       }
 
       if (MemInst.isInvariantLoad()) {
if (MemInst.isValid() && MemInst.isLoad() && !DeferredElideWith) {
       // If we have an available version of this load, and if it is the right
       // generation or the load is known to be from an invariant location,
       // replace this instruction.
       //
       // If either the dominating load or the current load are invariant, then
       // we can assume the current load loads the same value as the dominating
       // load.
       LoadValue InVal = AvailableLoads.lookup(MemInst.getPointerOperand());
-      if (Value *Op = getMatchingValue(InVal, MemInst, CurrentGeneration)) {
-        LLVM_DEBUG(dbgs() << "EarlyCSE CSE LOAD: " << Inst
+      if (Value *Op = getMatchingValue(InVal, MemInst, CurrentGeneration,
+                                       StrictFPGeneration)) {
+        LLVM_DEBUG(dbgs() << "EarlyCSE CSE LOAD: (curgen " << CurrentGeneration
+                          << " sfpgen " << StrictFPGeneration << ") " << Inst
                           << "  to: " << *InVal.DefInst << '\n');
         if (!DebugCounter::shouldExecute(CSECounter)) {
           LLVM_DEBUG(dbgs() << "Skipping due to debug counter\n");
           continue;
         }
         if (!Inst.use_empty())
           Inst.replaceAllUsesWith(Op);
         salvageKnowledge(&Inst, &AC);
for (Instruction &Inst : make_early_inc_range(BB->getInstList())) {
     // may override this (e.g. so that a store intrinsic does not read from
     // memory, and thus will be treated the same as a regular store for
     // commoning purposes).
     if ((Inst.mayReadFromMemory() || Inst.mayThrow()) &&
         !(MemInst.isValid() && !MemInst.mayReadFromMemory()))
       LastStore = nullptr;
 
     // If this is a read-only call, process it.
-    if (CallValue::canHandle(&Inst)) {
+    if (CallValue::canHandle(&Inst) && !DeferredElideWith) {
       // If we have an available version of this call, and if it is the right
       // generation, replace this instruction.
       std::pair<Instruction *, unsigned> InVal = AvailableCalls.lookup(&Inst);
       if (InVal.first != nullptr &&
           isSameMemGeneration(InVal.second, CurrentGeneration, InVal.first,
                               &Inst)) {
         LLVM_DEBUG(dbgs() << "EarlyCSE CSE CALL: " << Inst
                           << "  to: " << *InVal.first << '\n');
if (auto *FI = dyn_cast<FenceInst>(&Inst))
         continue;
       }
 
     // write back DSE - If we write back the same value we just loaded from
     // the same location and haven't passed any intervening writes or ordering
     // operations, we can remove the write.  The primary benefit is in allowing
     // the available load table to remain valid and value forward past where
     // the store originally was.
-    if (MemInst.isValid() && MemInst.isStore()) {
+    if (MemInst.isValid() && MemInst.isStore() && !DeferredElideWith) {
       LoadValue InVal = AvailableLoads.lookup(MemInst.getPointerOperand());
       if (InVal.DefInst &&
-          InVal.DefInst == getMatchingValue(InVal, MemInst, CurrentGeneration)) {
+          InVal.DefInst == getMatchingValue(InVal, MemInst, CurrentGeneration,
+                                            StrictFPGeneration)) {
         // It is okay to have a LastStore to a different pointer here if MemorySSA
         // tells us that the load and store are from the same memory generation.
         // In that case, LastStore should keep its present value since we're
         // removing the current store.
         assert((!LastStore ||
                 ParseMemoryInst(LastStore, TTI).getPointerOperand() ==
                     MemInst.getPointerOperand() ||
                 MSSA) &&
if (MemInst.isValid() && MemInst.isStore() && !DeferredElideWith) {
         // to eliminate this store.
         continue;
       }
     }
 
     // Okay, this isn't something we can CSE at all.  Check to see if it is
     // something that could modify memory.  If so, our available memory values
     // cannot be used so bump the generation count.
+    // Well, constrained FP may end up here, and we do handle that, but we
+    // still need to treat it like a memory write.
     if (Inst.mayWriteToMemory()) {
       ++CurrentGeneration;
 
       if (MemInst.isValid() && MemInst.isStore()) {
         // We do a trivial form of DSE if there are two stores to the same
         // location with no intervening loads.  Delete the earlier store.
         if (LastStore) {
           if (overridingStores(ParseMemoryInst(LastStore, TTI), MemInst)) {
if (Inst.mayWriteToMemory()) {
         // it's not clear this is a profitable transform. Another option would
         // be to merge the ordering with that of the post dominating store.
         if (MemInst.isUnordered() && !MemInst.isVolatile())
           LastStore = &Inst;
         else
           LastStore = nullptr;
       }
     }
+    if (DeferredElideWith) {
+      LLVM_DEBUG(dbgs() << "EarlyCSE CSE Finally: " << Inst
+                        << "  to: " << *DeferredElideWith << '\n');
+      if (auto *I = dyn_cast<Instruction>(DeferredElideWith))
+        I->andIRFlags(&Inst);
+      Inst.replaceAllUsesWith(DeferredElideWith);
+      salvageKnowledge(&Inst, &AC);
+      removeMSSA(Inst);
+      Inst.eraseFromParent();
+      Changed = true;
+      ++NumCSE;
+      DeferredElideWith = nullptr;
+    }
   }
 
   return Changed;
 }
 
 bool EarlyCSE::run() {
   // Note, deque is being used here because there is significant performance
   // gains over vector when the container becomes very large due to the
   // specific access patterns. For more information see the mailing list
   // discussion on this:
   // http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20120116/135228.html
   std::deque<StackNode *> nodesToProcess;
 
   bool Changed = false;
 
+  StrictFPGeneration = (isFunctionStrictFP ? 1 : 0);
   // Process the root node.
   nodesToProcess.push_back(new StackNode(
       AvailableValues, AvailableLoads, AvailableInvariants, AvailableCalls,
-      CurrentGeneration, DT.getRootNode(),
+      CurrentGeneration, StrictFPGeneration, DT.getRootNode(),
       DT.getRootNode()->begin(), DT.getRootNode()->end()));
 
   assert(!CurrentGeneration && "Create a new EarlyCSE instance to rerun it.");
 
   // Process the stack.
   while (!nodesToProcess.empty()) {
     // Grab the first item off the stack. Set the current generation, remove
     // the node from the stack, and process it.
     StackNode *NodeToProcess = nodesToProcess.back();
 
     // Initialize class members.
     CurrentGeneration = NodeToProcess->currentGeneration();
+    StrictFPGeneration = NodeToProcess->strictFPGeneration();
 
     // Check if the node needs to be processed.
     if (!NodeToProcess->isProcessed()) {
       // Process the node.
       Changed |= processNode(NodeToProcess->node());
       NodeToProcess->childGeneration(CurrentGeneration);
       NodeToProcess->process();
     } else if (NodeToProcess->childIter() != NodeToProcess->end()) {
       // Push the next child onto the stack.
       DomTreeNode *child = NodeToProcess->nextChild();
       nodesToProcess.push_back(
           new StackNode(AvailableValues, AvailableLoads, AvailableInvariants,
                         AvailableCalls, NodeToProcess->childGeneration(),
-                        child, child->begin(), child->end()));
+                        NodeToProcess->childStrictFPGeneration(), child,
+                        child->begin(), child->end()));
     } else {
       // It has been processed, and there are no more children to process,
       // so delete it and pop it off the stack.
       delete NodeToProcess;
       nodesToProcess.pop_back();
     }
   } // while (!nodes...)
 
   return Changed;
 }
 
 PreservedAnalyses EarlyCSEPass::run(Function &F,
                                     FunctionAnalysisManager &AM) {
   auto &TLI = AM.getResult<TargetLibraryAnalysis>(F);
   auto &TTI = AM.getResult<TargetIRAnalysis>(F);
   auto &DT = AM.getResult<DominatorTreeAnalysis>(F);
   auto &AC = AM.getResult<AssumptionAnalysis>(F);
   auto *MSSA =
       UseMemorySSA ? &AM.getResult<MemorySSAAnalysis>(F).getMSSA() : nullptr;
 
-  EarlyCSE CSE(F.getParent()->getDataLayout(), TLI, TTI, DT, AC, MSSA);
+  EarlyCSE CSE(F.getParent()->getDataLayout(), TLI, TTI, DT, AC, MSSA, F.hasFnAttribute(Attribute::StrictFP));
 
   if (!CSE.run())
     return PreservedAnalyses::all();
 
   PreservedAnalyses PA;
   PA.preserveSet<CFGAnalyses>();
   if (UseMemorySSA)
     PA.preserve<MemorySSAAnalysis>();
bool runOnFunction(Function &F) override {
 
     auto &TLI = getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(F);
     auto &TTI = getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
     auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
     auto &AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
     auto *MSSA =
         UseMemorySSA ? &getAnalysis<MemorySSAWrapperPass>().getMSSA() : nullptr;
 
-    EarlyCSE CSE(F.getParent()->getDataLayout(), TLI, TTI, DT, AC, MSSA);
+    EarlyCSE CSE(F.getParent()->getDataLayout(), TLI, TTI, DT, AC, MSSA, F.hasFnAttribute(Attribute::StrictFP));
 
     return CSE.run();
   }
 
   void getAnalysisUsage(AnalysisUsage &AU) const override {
     AU.addRequired<AssumptionCacheTracker>();
     AU.addRequired<DominatorTreeWrapperPass>();
     AU.addRequired<TargetLibraryInfoWrapperPass>();

llvm/test/Transforms/EarlyCSE/maytrap-strictfp.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -S -early-cse -earlycse-debug-hash | FileCheck %s
+; RUN: opt < %s -S -basic-aa -early-cse-memssa | FileCheck %s
+
+; Test use of constrained floating point intrinsics in the default
+; floating point environment.
+
+define double @multiple_fadd(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fadd(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fadd.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0:[0-9]+]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %2 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fadd_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fadd_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fadd.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fadd.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fsub(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fsub(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fsub.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %2 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fsub_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fsub_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fsub.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fsub.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fmul(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fmul(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fmul.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %2 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fmul_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fmul_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fmul.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fmul.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fdiv(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fdiv(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fdiv.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %2 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fdiv_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fdiv_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fdiv.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fdiv.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_frem(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_frem(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.frem.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %2 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_frem_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_frem_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.frem.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.frem.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define i32 @multiple_fptoui(double %a) #0 {
+; CHECK-LABEL: @multiple_fptoui(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double [[A:%.*]], metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP1]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.maytrap") #0
+  %2 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.maytrap") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define i32 @multiple_fptoui_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fptoui_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double [[A:%.*]], metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double [[A]], metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.maytrap") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define double @multiple_uitofp(i32 %a) #0 {
+; CHECK-LABEL: @multiple_uitofp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %2 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define double @multiple_uitofp_split(i32 %a) #0 {
+; CHECK-LABEL: @multiple_uitofp_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 [[A]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define i32 @multiple_fptosi(double %a) #0 {
+; CHECK-LABEL: @multiple_fptosi(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double [[A:%.*]], metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP1]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.maytrap") #0
+  %2 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.maytrap") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define i32 @multiple_fptosi_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fptosi_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double [[A:%.*]], metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double [[A]], metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.maytrap") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define double @multiple_sitofp(i32 %a) #0 {
+; CHECK-LABEL: @multiple_sitofp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %2 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define double @multiple_sitofp_split(i32 %a) #0 {
+; CHECK-LABEL: @multiple_sitofp_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 [[A]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define i1 @multiple_fcmp(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmp.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP2]], i32 [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.maytrap") #0
+  %2 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.maytrap") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+define i1 @multiple_fcmp_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmp_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmp.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i1 @llvm.experimental.constrained.fcmp.f64(double [[A]], double [[B]], metadata !"oeq", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i1 [[TMP2]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @bar.i32(i32 [[TMP3]], i32 [[TMP4]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP2]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.maytrap") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+define i1 @multiple_fcmps(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmps(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP2]], i32 [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.maytrap") #0
+  %2 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.maytrap") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+define i1 @multiple_fcmps_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmps_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double [[A]], double [[B]], metadata !"oeq", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i1 [[TMP2]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @bar.i32(i32 [[TMP3]], i32 [[TMP4]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP2]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.maytrap") #0
+  call void @arbitraryfunc() #0
+  %2 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.maytrap") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+attributes #0 = { strictfp }
+
+declare void @arbitraryfunc() #0
+declare double @foo.f64(double, double) #0
+declare i32 @bar.i32(i32, i32) #0
+
+declare double @llvm.experimental.constrained.fadd.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fsub.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fmul.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fdiv.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.frem.f64(double, double, metadata, metadata)
+declare i32 @llvm.experimental.constrained.fptoui.i32.f64(double, metadata)
+declare double @llvm.experimental.constrained.uitofp.f64.i32(i32, metadata, metadata)
+declare i32 @llvm.experimental.constrained.fptosi.i32.f64(double, metadata)
+declare double @llvm.experimental.constrained.sitofp.f64.i32(i32, metadata, metadata)
+declare i1 @llvm.experimental.constrained.fcmp.i1.f64(double, double, metadata, metadata)
+declare i1 @llvm.experimental.constrained.fcmps.i1.f64(double, double, metadata, metadata)

llvm/test/Transforms/EarlyCSE/mixed-strictfp.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -S -early-cse -earlycse-debug-hash | FileCheck %s
+; RUN: opt < %s -S -basic-aa -early-cse-memssa | FileCheck %s
+
+; Test use of constrained floating point intrinsics in the default
+; floating point environment.
+
+define double @multiple_fadd(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fadd(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fadd.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore") #[[ATTR0:[0-9]+]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fadd.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fsub(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fsub(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fsub.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fsub.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fmul(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fmul(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fmul.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fmul.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fdiv(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fdiv(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fdiv.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fdiv.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_frem(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_frem(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.frem.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.frem.f64(double [[A]], double [[B]], metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define i32 @multiple_fptoui(double %a) #0 {
+; CHECK-LABEL: @multiple_fptoui(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double [[A:%.*]], metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double [[A]], metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  %2 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.strict") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define double @multiple_uitofp(i32 %a) #0 {
+; CHECK-LABEL: @multiple_uitofp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 [[A]], metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define i32 @multiple_fptosi(double %a) #0 {
+; CHECK-LABEL: @multiple_fptosi(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double [[A:%.*]], metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double [[A]], metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  %2 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.strict") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define double @multiple_sitofp(i32 %a) #0 {
+; CHECK-LABEL: @multiple_sitofp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 [[A]], metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define i1 @multiple_fcmp(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmp.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i1 @llvm.experimental.constrained.fcmp.f64(double [[A]], double [[B]], metadata !"oeq", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i1 [[TMP2]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @bar.i32(i32 [[TMP3]], i32 [[TMP4]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP2]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  %2 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.strict") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+define i1 @multiple_fcmps(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmps(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double [[A]], double [[B]], metadata !"oeq", metadata !"fpexcept.strict") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i1 [[TMP2]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @bar.i32(i32 [[TMP3]], i32 [[TMP4]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP2]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  %2 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.strict") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+attributes #0 = { strictfp }
+
+declare void @arbitraryfunc() #0
+declare double @foo.f64(double, double) #0
+declare i32 @bar.i32(i32, i32) #0
+
+declare double @llvm.experimental.constrained.fadd.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fsub.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fmul.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fdiv.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.frem.f64(double, double, metadata, metadata)
+declare i32 @llvm.experimental.constrained.fptoui.i32.f64(double, metadata)
+declare double @llvm.experimental.constrained.uitofp.f64.i32(i32, metadata, metadata)
+declare i32 @llvm.experimental.constrained.fptosi.i32.f64(double, metadata)
+declare double @llvm.experimental.constrained.sitofp.f64.i32(i32, metadata, metadata)
+declare i1 @llvm.experimental.constrained.fcmp.i1.f64(double, double, metadata, metadata)
+declare i1 @llvm.experimental.constrained.fcmps.i1.f64(double, double, metadata, metadata)

llvm/test/Transforms/EarlyCSE/round-dyn-strictfp.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -S -early-cse -earlycse-debug-hash | FileCheck %s
+; RUN: opt < %s -S -basic-aa -early-cse-memssa | FileCheck %s
+
+; Test use of constrained floating point intrinsics in the default
+; floating point environment.
+
+define double @multiple_fadd(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fadd(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fadd.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0:[0-9]+]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fadd_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fadd_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fadd.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fadd.f64(double [[A]], double [[B]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.fadd.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fsub(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fsub(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fsub.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fsub_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fsub_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fsub.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fsub.f64(double [[A]], double [[B]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.fsub.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fmul(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fmul(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fmul.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fmul_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fmul_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fmul.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fmul.f64(double [[A]], double [[B]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.fmul.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fdiv(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fdiv(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fdiv.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_fdiv_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fdiv_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.fdiv.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.fdiv.f64(double [[A]], double [[B]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.fdiv.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_frem(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_frem(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.frem.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define double @multiple_frem_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_frem_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.frem.f64(double [[A:%.*]], double [[B:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.frem.f64(double [[A]], double [[B]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.frem.f64(double %a, double %b, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %2) #0
+  ret double %2
+}
+
+define i32 @multiple_fptoui(double %a) #0 {
+; CHECK-LABEL: @multiple_fptoui(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double [[A:%.*]], metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP1]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  %2 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define i32 @multiple_fptoui_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fptoui_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double [[A:%.*]], metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP1]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call i32 @llvm.experimental.constrained.fptoui.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define double @multiple_uitofp(i32 %a) #0 {
+; CHECK-LABEL: @multiple_uitofp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 [[A:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define double @multiple_uitofp_split(i32 %a) #0 {
+; CHECK-LABEL: @multiple_uitofp_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 [[A:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 [[A]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %a, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define i32 @multiple_fptosi(double %a) #0 {
+; CHECK-LABEL: @multiple_fptosi(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double [[A:%.*]], metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP1]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  %2 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define i32 @multiple_fptosi_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fptosi_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double [[A:%.*]], metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @bar.i32(i32 [[TMP1]], i32 [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i32 [[TMP1]]
+;
+  %1 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call i32 @llvm.experimental.constrained.fptosi.i32.f64(double %a, metadata !"fpexcept.ignore") #0
+  %3 = call i32 @bar.i32(i32 %1, i32 %1) #0
+  ret i32 %2
+}
+
+define double @multiple_sitofp(i32 %a) #0 {
+; CHECK-LABEL: @multiple_sitofp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 [[A:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP1]]
+;
+  %1 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %2 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define double @multiple_sitofp_split(i32 %a) #0 {
+; CHECK-LABEL: @multiple_sitofp_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 [[A:%.*]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 [[A]], metadata !"round.dynamic", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = call double @foo.f64(double [[TMP1]], double [[TMP1]]) #[[ATTR0]]
+; CHECK-NEXT:    ret double [[TMP2]]
+;
+  %1 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call double @llvm.experimental.constrained.sitofp.f64.i32(i32 %a, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0
+  %3 = call double @foo.f64(double %1, double %1) #0
+  ret double %2
+}
+
+define i1 @multiple_fcmp(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmp(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmp.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP2]], i32 [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  %2 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+define i1 @multiple_fcmp_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmp_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmp.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP2]], i32 [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call i1 @llvm.experimental.constrained.fcmp.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+define i1 @multiple_fcmps(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmps(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP2]], i32 [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  %2 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+define i1 @multiple_fcmps_split(double %a, double %b) #0 {
+; CHECK-LABEL: @multiple_fcmps_split(
+; CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double [[A:%.*]], double [[B:%.*]], metadata !"oeq", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    call void @arbitraryfunc() #[[ATTR0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i1 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @bar.i32(i32 [[TMP2]], i32 [[TMP2]]) #[[ATTR0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %1 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  call void @arbitraryfunc() #0
+  %2 = call i1 @llvm.experimental.constrained.fcmps.i1.f64(double %a, double %b, metadata !"oeq", metadata !"fpexcept.ignore") #0
+  %3 = zext i1 %1 to i32
+  %4 = zext i1 %2 to i32
+  %5 = call i32 @bar.i32(i32 %3, i32 %4) #0
+  ret i1 %2
+}
+
+attributes #0 = { strictfp }
+
+declare void @arbitraryfunc() #0
+declare double @foo.f64(double, double) #0
+declare i32 @bar.i32(i32, i32) #0
+
+declare double @llvm.experimental.constrained.fadd.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fsub.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fmul.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.fdiv.f64(double, double, metadata, metadata)
+declare double @llvm.experimental.constrained.frem.f64(double, double, metadata, metadata)
+declare i32 @llvm.experimental.constrained.fptoui.i32.f64(double, metadata)
+declare double @llvm.experimental.constrained.uitofp.f64.i32(i32, metadata, metadata)
+declare i32 @llvm.experimental.constrained.fptosi.i32.f64(double, metadata)
+declare double @llvm.experimental.constrained.sitofp.f64.i32(i32, metadata, metadata)
+declare i1 @llvm.experimental.constrained.fcmp.i1.f64(double, double, metadata, metadata)
+declare i1 @llvm.experimental.constrained.fcmps.i1.f64(double, double, metadata, metadata)

llvm/test/Transforms/EarlyCSE/tfpropagation.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -S -early-cse -earlycse-debug-hash | FileCheck %s
+; RUN: opt < %s -S -basic-aa -early-cse-memssa | FileCheck %s
+
+define i64 @branching_int(i32 %a) {
+; CHECK-LABEL: @branching_int(
+; CHECK-NEXT:    [[CONV1:%.*]] = zext i32 [[A:%.*]] to i64
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp ugt i64 1, [[CONV1]]
+; CHECK-NEXT:    br i1 [[CMP2]], label [[IF_THEN3:%.*]], label [[IF_END3:%.*]]
+; CHECK:       if.then3:
+; CHECK-NEXT:    [[C:%.*]] = call double @truefunc.f64.i1(i1 true)
+; CHECK-NEXT:    br label [[OUT:%.*]]
+; CHECK:       if.end3:
+; CHECK-NEXT:    [[D:%.*]] = call double @falsefunc.f64.i1(i1 false)
+; CHECK-NEXT:    br label [[OUT]]
+; CHECK:       out:
+; CHECK-NEXT:    ret i64 [[CONV1]]
+;
+  %conv1 = zext i32 %a to i64
+  %cmp2 = icmp ugt i64 1, %conv1
+  br i1 %cmp2, label %if.then3, label %if.end3
+
+if.then3:
+  %c = call double @truefunc.f64.i1(i1 %cmp2)
+  br label %out
+
+if.end3:
+  %d = call double @falsefunc.f64.i1(i1 %cmp2)
+  br label %out
+
+out:
+  ret i64 %conv1
+}
+
+define double @branching_fp(i64 %a) {
+; CHECK-LABEL: @branching_fp(
+; CHECK-NEXT:    [[CONV1:%.*]] = uitofp i64 [[A:%.*]] to double
+; CHECK-NEXT:    [[CMP2:%.*]] = fcmp ogt double 1.000000e+00, [[CONV1]]
+; CHECK-NEXT:    br i1 [[CMP2]], label [[IF_THEN3:%.*]], label [[IF_END3:%.*]]
+; CHECK:       if.then3:
+; CHECK-NEXT:    [[C:%.*]] = call double @truefunc.f64.i1(i1 true)
+; CHECK-NEXT:    br label [[OUT:%.*]]
+; CHECK:       if.end3:
+; CHECK-NEXT:    [[D:%.*]] = call double @falsefunc.f64.i1(i1 false)
+; CHECK-NEXT:    br label [[OUT]]
+; CHECK:       out:
+; CHECK-NEXT:    ret double [[CONV1]]
+;
+  %conv1 = uitofp i64 %a to double
+  %cmp2 = fcmp ogt double 1.000000e+00, %conv1
+  br i1 %cmp2, label %if.then3, label %if.end3
+
+if.then3:
+  %c = call double @truefunc.f64.i1(i1 %cmp2)
+  br label %out
+
+if.end3:
+  %d = call double @falsefunc.f64.i1(i1 %cmp2)
+  br label %out
+
+out:
+  ret double %conv1
+}
+
+define double @branching_exceptignore(i64 %a) #0 {
+; CHECK-LABEL: @branching_exceptignore(
+; CHECK-NEXT:    [[CONV1:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i64(i64 [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore") #[[ATTR0:[0-9]+]]
+; CHECK-NEXT:    [[CMP2:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double 1.000000e+00, double [[CONV1]], metadata !"ogt", metadata !"fpexcept.ignore") #[[ATTR0]]
+; CHECK-NEXT:    br i1 [[CMP2]], label [[IF_THEN3:%.*]], label [[IF_END3:%.*]]
+; CHECK:       if.then3:
+; CHECK-NEXT:    [[C:%.*]] = call double @truefunc.f64.i1(i1 true) #[[ATTR0]]
+; CHECK-NEXT:    br label [[OUT:%.*]]
+; CHECK:       if.end3:
+; CHECK-NEXT:    [[D:%.*]] = call double @falsefunc.f64.i1(i1 false) #[[ATTR0]]
+; CHECK-NEXT:    br label [[OUT]]
+; CHECK:       out:
+; CHECK-NEXT:    ret double [[CONV1]]
+;
+  %conv1 = call double @llvm.experimental.constrained.uitofp.f64.i64(i64 %a, metadata !"round.tonearest", metadata !"fpexcept.ignore") #0
+  %cmp2 = call i1 @llvm.experimental.constrained.fcmps.f64(double 1.000000e+00, double %conv1, metadata !"ogt", metadata !"fpexcept.ignore") #0
+  br i1 %cmp2, label %if.then3, label %if.end3
+
+if.then3:
+  %c = call double @truefunc.f64.i1(i1 %cmp2) #0
+  br label %out
+
+if.end3:
+  %d = call double @falsefunc.f64.i1(i1 %cmp2) #0
+  br label %out
+
+out:
+  ret double %conv1
+}
+
+define double @branching_maytrap(i64 %a) #0 {
+; CHECK-LABEL: @branching_maytrap(
+; CHECK-NEXT:    [[CONV1:%.*]] = call double @llvm.experimental.constrained.uitofp.f64.i64(i64 [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    [[CMP2:%.*]] = call i1 @llvm.experimental.constrained.fcmps.f64(double 1.000000e+00, double [[CONV1]], metadata !"ogt", metadata !"fpexcept.maytrap") #[[ATTR0]]
+; CHECK-NEXT:    br i1 [[CMP2]], label [[IF_THEN3:%.*]], label [[IF_END3:%.*]]
+; CHECK:       if.then3:
+; CHECK-NEXT:    [[C:%.*]] = call double @truefunc.f64.i1(i1 true) #[[ATTR0]]
+; CHECK-NEXT:    br label [[OUT:%.*]]
+; CHECK:       if.end3:
+; CHECK-NEXT:    [[D:%.*]] = call double @falsefunc.f64.i1(i1 false) #[[ATTR0]]
+; CHECK-NEXT:    br label [[OUT]]
+; CHECK:       out:
+; CHECK-NEXT:    ret double [[CONV1]]
+;
+  %conv1 = call double @llvm.experimental.constrained.uitofp.f64.i64(i64 %a, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
+  %cmp2 = call i1 @llvm.experimental.constrained.fcmps.f64(double 1.000000e+00, double %conv1, metadata !"ogt", metadata !"fpexcept.maytrap") #0
+  br i1 %cmp2, label %if.then3, label %if.end3
+
+if.then3:
+  %c = call double @truefunc.f64.i1(i1 %cmp2) #0
+  br label %out
+
+if.end3:
+  %d = call double @falsefunc.f64.i1(i1 %cmp2) #0
+  br label %out
+
+out:
+  ret double %conv1
+}
+
+declare double @truefunc.f64.i1(i1)
+declare double @falsefunc.f64.i1(i1)
+declare double @llvm.experimental.constrained.uitofp.f64.i64(i64, metadata, metadata) #0
+declare i1 @llvm.experimental.constrained.fcmps.f64(double, double, metadata, metadata) #0
+
+attributes #0 = { strictfp }
+
+declare <4 x float> @llvm.experimental.constrained.fadd.v4f32(<4 x float>, <4 x float>, metadata, metadata) strictfp