[WebAssembly] Exception handling: Switch to the new proposal

Summary:
This switches the EH implementation to the new proposal:
https://github.com/WebAssembly/exception-handling/blob/master/proposals/Exceptions.md
(The previous proposal was
 https://github.com/WebAssembly/exception-handling/blob/master/proposals/old/Exceptions.md)

- Instruction changes
  - Now we have one single `catch` instruction that returns a except_ref
    value
  - `throw` now can take variable number of operations
  - `rethrow` does not have 'depth' argument anymore
  - `br_on_exn` queries an except_ref to see if it matches the tag and
    branches to the given label if true.
  - `extract_exception` is a pseudo instruction that simulates popping
    values from wasm stack. This is to make `br_on_exn`, a very special
    instruction, work: `br_on_exn` puts values onto the stack only if it
    is taken, and the # of values can vay depending on the tag.

- Now there's only one `catch` per `try`, this patch removes all special
  handling for terminate pad with a call to `__clang_call_terminate`.
  Before it was the only case there are two catch clauses (a normal
  `catch` and `catch_all` per `try`).

- Make `rethrow` act as a terminator like `throw`. This splits BB after
  `rethrow` in WasmEHPrepare, and deletes an unnecessary `unreachable`
  after `rethrow` in LateEHPrepare.

- Now we stop at all catchpads (because we add wasm `catch` instruction
  that catches all exceptions), this creates new
  `findWasmUnwindDestinations` function in SelectionDAGBuilder.

- Now we use `br_on_exn` instrution to figure out if an except_ref
  matches the current tag or not, LateEHPrepare generates this sequence
  for catch pads:
```
  catch
  block i32
  br_on_exn $__cpp_exception
  end_block
  extract_exception
```

- Branch analysis for `br_on_exn` in WebAssemblyInstrInfo

- Other various misc. changes to switch to the new proposal.

Reviewers: dschuff

Subscribers: sbc100, jgravelle-google, sunfish, llvm-commits

Differential Revision: https://reviews.llvm.org/D57134

llvm-svn: 352598

16 files changed, 351 insertions, 513 deletions

diff --git a/llvm/lib/Target/WebAssembly/InstPrinter/WebAssemblyInstPrinter.cpp b/llvm/lib/Target/WebAssembly/InstPrinter/WebAssemblyInstPrinter.cpp
index c70890052f2..225f4f2f9f4 100644
--- a/llvm/lib/Target/WebAssembly/InstPrinter/WebAssemblyInstPrinter.cpp
+++ b/llvm/lib/Target/WebAssembly/InstPrinter/WebAssemblyInstPrinter.cpp
@@ -122,61 +122,48 @@ void WebAssemblyInstPrinter::printInst(const MCInst *MI, raw_ostream &OS,
       }
       break;
 
-    case WebAssembly::CATCH_I32:
-    case WebAssembly::CATCH_I32_S:
-    case WebAssembly::CATCH_I64:
-    case WebAssembly::CATCH_I64_S:
-    case WebAssembly::CATCH_ALL:
-    case WebAssembly::CATCH_ALL_S:
-      // There can be multiple catch instructions for one try instruction, so we
-      // print a label only for the first 'catch' label.
-      if (LastSeenEHInst != CATCH) {
-        if (EHPadStack.empty()) {
-          printAnnotation(OS, "try-catch mismatch!");
-        } else {
-          printAnnotation(OS,
-                          "catch" + utostr(EHPadStack.pop_back_val()) + ':');
-        }
+    case WebAssembly::CATCH:
+    case WebAssembly::CATCH_S:
+      if (EHPadStack.empty()) {
+        printAnnotation(OS, "try-catch mismatch!");
+      } else {
+        printAnnotation(OS, "catch" + utostr(EHPadStack.pop_back_val()) + ':');
       }
-      LastSeenEHInst = CATCH;
       break;
     }
 
     // Annotate any control flow label references.
-    unsigned NumFixedOperands = Desc.NumOperands;
-    SmallSet<uint64_t, 8> Printed;
-    for (unsigned i = 0, e = MI->getNumOperands(); i < e; ++i) {
-      // See if this operand denotes a basic block target.
-      if (i < NumFixedOperands) {
-        // A non-variable_ops operand, check its type.
-        if (Desc.OpInfo[i].OperandType != WebAssembly::OPERAND_BASIC_BLOCK)
-          continue;
+
+    // rethrow instruction does not take any depth argument and rethrows to the
+    // nearest enclosing catch scope, if any. If there's no enclosing catch
+    // scope, it throws up to the caller.
+    if (Opc == WebAssembly::RETHROW || Opc == WebAssembly::RETHROW_S) {
+      if (EHPadStack.empty()) {
+        printAnnotation(OS, "to caller");
       } else {
-        // A variable_ops operand, which currently can be immediates (used in
-        // br_table) which are basic block targets, or for call instructions
-        // when using -wasm-keep-registers (in which case they are registers,
-        // and should not be processed).
-        if (!MI->getOperand(i).isImm())
-          continue;
+        printAnnotation(OS, "down to catch" + utostr(EHPadStack.back()));
       }
-      uint64_t Depth = MI->getOperand(i).getImm();
-      if (!Printed.insert(Depth).second)
-        continue;
 
-      if (Opc == WebAssembly::RETHROW || Opc == WebAssembly::RETHROW_S) {
-        if (Depth > EHPadStack.size()) {
-          printAnnotation(OS, "Invalid depth argument!");
-        } else if (Depth == EHPadStack.size()) {
-          // This can happen when rethrow instruction breaks out of all nests
-          // and throws up to the current function's caller.
-          printAnnotation(OS, utostr(Depth) + ": " + "to caller");
+    } else {
+      unsigned NumFixedOperands = Desc.NumOperands;
+      SmallSet<uint64_t, 8> Printed;
+      for (unsigned I = 0, E = MI->getNumOperands(); I < E; ++I) {
+        // See if this operand denotes a basic block target.
+        if (I < NumFixedOperands) {
+          // A non-variable_ops operand, check its type.
+          if (Desc.OpInfo[I].OperandType != WebAssembly::OPERAND_BASIC_BLOCK)
+            continue;
         } else {
-          uint64_t CatchNo = EHPadStack.rbegin()[Depth];
-          printAnnotation(OS, utostr(Depth) + ": " + "down to catch" +
-                                  utostr(CatchNo));
+          // A variable_ops operand, which currently can be immediates (used in
+          // br_table) which are basic block targets, or for call instructions
+          // when using -wasm-keep-registers (in which case they are registers,
+          // and should not be processed).
+          if (!MI->getOperand(I).isImm())
+            continue;
         }
-
-      } else {
+        uint64_t Depth = MI->getOperand(I).getImm();
+        if (!Printed.insert(Depth).second)
+          continue;
         if (Depth >= ControlFlowStack.size()) {
           printAnnotation(OS, "Invalid depth argument!");
         } else {
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyAsmPrinter.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyAsmPrinter.cpp
index 573e4c2af84..f1bb524ad58 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyAsmPrinter.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyAsmPrinter.cpp
@@ -43,6 +43,8 @@ using namespace llvm;
 
 #define DEBUG_TYPE "asm-printer"
 
+extern cl::opt<bool> WasmKeepRegisters;
+
 //===----------------------------------------------------------------------===//
 // Helpers.
 //===----------------------------------------------------------------------===//
@@ -304,6 +306,14 @@ void WebAssemblyAsmPrinter::EmitInstruction(const MachineInstr *MI) {
       OutStreamer->AddBlankLine();
     }
     break;
+  case WebAssembly::EXTRACT_EXCEPTION_I32:
+  case WebAssembly::EXTRACT_EXCEPTION_I32_S:
+    // These are pseudo instructions that simulates popping values from stack.
+    // We print these only when we have -wasm-keep-registers on for assembly
+    // readability.
+    if (!WasmKeepRegisters)
+      break;
+    LLVM_FALLTHROUGH;
   default: {
     WebAssemblyMCInstLower MCInstLowering(OutContext, *this);
     MCInst TmpInst;
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyCFGStackify.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyCFGStackify.cpp
index 6b30f31ab1d..566b9a5e460 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyCFGStackify.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyCFGStackify.cpp
@@ -37,6 +37,7 @@
 #include "llvm/MC/MCAsmInfo.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
+#include <cstring>
 using namespace llvm;
 
 #define DEBUG_TYPE "wasm-cfg-stackify"
@@ -110,11 +111,9 @@ FunctionPass *llvm::createWebAssemblyCFGStackify() {
 static bool ExplicitlyBranchesTo(MachineBasicBlock *Pred,
                                  MachineBasicBlock *MBB) {
   for (MachineInstr &MI : Pred->terminators())
-    // Even if a rethrow takes a BB argument, it is not a branch
-    if (!WebAssembly::isRethrow(MI))
-      for (MachineOperand &MO : MI.explicit_operands())
-        if (MO.isMBB() && MO.getMBB() == MBB)
-          return true;
+    for (MachineOperand &MO : MI.explicit_operands())
+      if (MO.isMBB() && MO.getMBB() == MBB)
+        return true;
   return false;
 }
 
@@ -217,12 +216,20 @@ void WebAssemblyCFGStackify::placeBlockMarker(MachineBasicBlock &MBB) {
   // which reduces overall stack height.
   MachineBasicBlock *Header = nullptr;
   bool IsBranchedTo = false;
+  bool IsBrOnExn = false;
+  MachineInstr *BrOnExn = nullptr;
   int MBBNumber = MBB.getNumber();
   for (MachineBasicBlock *Pred : MBB.predecessors()) {
     if (Pred->getNumber() < MBBNumber) {
       Header = Header ? MDT.findNearestCommonDominator(Header, Pred) : Pred;
-      if (ExplicitlyBranchesTo(Pred, &MBB))
+      if (ExplicitlyBranchesTo(Pred, &MBB)) {
         IsBranchedTo = true;
+        if (Pred->getFirstTerminator()->getOpcode() == WebAssembly::BR_ON_EXN) {
+          IsBrOnExn = true;
+          assert(!BrOnExn && "There should be only one br_on_exn per block");
+          BrOnExn = &*Pred->getFirstTerminator();
+        }
+      }
     }
   }
   if (!Header)
@@ -299,11 +306,27 @@ void WebAssemblyCFGStackify::placeBlockMarker(MachineBasicBlock &MBB) {
   }
 
   // Add the BLOCK.
+
+  // 'br_on_exn' extracts except_ref object and pushes variable number of values
+  // depending on its tag. For C++ exception, its a single i32 value, and the
+  // generated code will be in the form of:
+  // block i32
+  //   br_on_exn 0, $__cpp_exception
+  //   rethrow
+  // end_block
+  WebAssembly::ExprType ReturnType = WebAssembly::ExprType::Void;
+  if (IsBrOnExn) {
+    const char *TagName = BrOnExn->getOperand(1).getSymbolName();
+    if (std::strcmp(TagName, "__cpp_exception") != 0)
+      llvm_unreachable("Only C++ exception is supported");
+    ReturnType = WebAssembly::ExprType::I32;
+  }
+
   auto InsertPos = GetLatestInsertPos(Header, BeforeSet, AfterSet);
   MachineInstr *Begin =
       BuildMI(*Header, InsertPos, Header->findDebugLoc(InsertPos),
               TII.get(WebAssembly::BLOCK))
-          .addImm(int64_t(WebAssembly::ExprType::Void));
+          .addImm(int64_t(ReturnType));
 
   // Decide where in Header to put the END_BLOCK.
   BeforeSet.clear();
@@ -416,11 +439,6 @@ void WebAssemblyCFGStackify::placeTryMarker(MachineBasicBlock &MBB) {
   if (!MBB.isEHPad())
     return;
 
-  // catch_all terminate pad is grouped together with catch terminate pad and
-  // does not need a separate TRY and END_TRY marker.
-  if (WebAssembly::isCatchAllTerminatePad(MBB))
-    return;
-
   MachineFunction &MF = *MBB.getParent();
   auto &MDT = getAnalysis<MachineDominatorTree>();
   const auto &TII = *MF.getSubtarget<WebAssemblySubtarget>().getInstrInfo();
@@ -529,7 +547,8 @@ void WebAssemblyCFGStackify::placeTryMarker(MachineBasicBlock &MBB) {
   // throw.
   if (MBB.isPredecessor(Header)) {
     auto TermPos = Header->getFirstTerminator();
-    if (TermPos == Header->end() || !WebAssembly::isRethrow(*TermPos)) {
+    if (TermPos == Header->end() ||
+        TermPos->getOpcode() != WebAssembly::RETHROW) {
       for (const auto &MI : reverse(*Header)) {
         if (MI.isCall()) {
           AfterSet.insert(&MI);
@@ -674,7 +693,6 @@ static void AppendEndToFunction(MachineFunction &MF,
 
 /// Insert LOOP/TRY/BLOCK markers at appropriate places.
 void WebAssemblyCFGStackify::placeMarkers(MachineFunction &MF) {
-  const MCAsmInfo *MCAI = MF.getTarget().getMCAsmInfo();
   // We allocate one more than the number of blocks in the function to
   // accommodate for the possible fake block we may insert at the end.
   ScopeTops.resize(MF.getNumBlockIDs() + 1);
@@ -682,6 +700,7 @@ void WebAssemblyCFGStackify::placeMarkers(MachineFunction &MF) {
   for (auto &MBB : MF)
     placeLoopMarker(MBB);
   // Place the TRY for MBB if MBB is the EH pad of an exception.
+  const MCAsmInfo *MCAI = MF.getTarget().getMCAsmInfo();
   if (MCAI->getExceptionHandlingType() == ExceptionHandling::Wasm &&
       MF.getFunction().hasPersonalityFn())
     for (auto &MBB : MF)
@@ -692,12 +711,8 @@ void WebAssemblyCFGStackify::placeMarkers(MachineFunction &MF) {
 }
 
 void WebAssemblyCFGStackify::rewriteDepthImmediates(MachineFunction &MF) {
-  const auto &TII = *MF.getSubtarget<WebAssemblySubtarget>().getInstrInfo();
   // Now rewrite references to basic blocks to be depth immediates.
-  // We need two stacks: one for normal scopes and the other for EH pad scopes.
-  // EH pad stack is used to rewrite depths in rethrow instructions.
   SmallVector<const MachineBasicBlock *, 8> Stack;
-  SmallVector<const MachineBasicBlock *, 8> EHPadStack;
   for (auto &MBB : reverse(MF)) {
     for (auto I = MBB.rbegin(), E = MBB.rend(); I != E; ++I) {
       MachineInstr &MI = *I;
@@ -714,26 +729,6 @@ void WebAssemblyCFGStackify::rewriteDepthImmediates(MachineFunction &MF) {
                    MBB.getNumber() &&
                "Block/try marker should be balanced");
         Stack.pop_back();
-        EHPadStack.pop_back();
-        break;
-
-      case WebAssembly::CATCH_I32:
-      case WebAssembly::CATCH_I64:
-      case WebAssembly::CATCH_ALL:
-        // Currently the only case there are more than one catch for a try is
-        // for catch terminate pad, in the form of
-        //   try
-        //   catch
-        //     call @__clang_call_terminate
-        //     unreachable
-        //   catch_all
-        //     call @std::terminate
-        //     unreachable
-        //   end
-        // So we shouldn't push the current BB for the second catch_all block
-        // here.
-        if (!WebAssembly::isCatchAllTerminatePad(MBB))
-          EHPadStack.push_back(&MBB);
         break;
 
       case WebAssembly::LOOP:
@@ -750,23 +745,6 @@ void WebAssemblyCFGStackify::rewriteDepthImmediates(MachineFunction &MF) {
         Stack.push_back(EndToBegin[&MI]->getParent());
         break;
 
-      case WebAssembly::RETHROW: {
-        // Rewrite MBB operands to be depth immediates.
-        unsigned EHPadDepth = GetDepth(EHPadStack, MI.getOperand(0).getMBB());
-        MI.RemoveOperand(0);
-        MI.addOperand(MF, MachineOperand::CreateImm(EHPadDepth));
-        break;
-      }
-
-      case WebAssembly::RETHROW_TO_CALLER: {
-        MachineInstr *Rethrow =
-            BuildMI(MBB, MI, MI.getDebugLoc(), TII.get(WebAssembly::RETHROW))
-                .addImm(EHPadStack.size());
-        MI.eraseFromParent();
-        I = MachineBasicBlock::reverse_iterator(Rethrow);
-        break;
-      }
-
       default:
         if (MI.isTerminator()) {
           // Rewrite MBB operands to be depth immediates.
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyEHRestoreStackPointer.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyEHRestoreStackPointer.cpp
index dc0f05c4536..bed4d1ede99 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyEHRestoreStackPointer.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyEHRestoreStackPointer.cpp
@@ -77,7 +77,7 @@ bool WebAssemblyEHRestoreStackPointer::runOnMachineFunction(
     // function uses the red zone, but that only happens with leaf functions,
     // and we don't restore __stack_pointer in leaf functions anyway.
     auto InsertPos = MBB.begin();
-    if (WebAssembly::isCatch(*MBB.begin()))
+    if (MBB.begin()->getOpcode() == WebAssembly::CATCH)
       InsertPos++;
     FrameLowering->writeSPToGlobal(WebAssembly::SP32, MF, MBB, InsertPos,
                                    MBB.begin()->getDebugLoc());
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyExceptionInfo.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyExceptionInfo.cpp
index 4a316e5c11c..0387957b14c 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyExceptionInfo.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyExceptionInfo.cpp
@@ -50,10 +50,6 @@ void WebAssemblyExceptionInfo::recalculate(
     MachineBasicBlock *EHPad = DomNode->getBlock();
     if (!EHPad->isEHPad())
       continue;
-    // We group catch & catch-all terminate pads together, so skip the second
-    // one
-    if (WebAssembly::isCatchAllTerminatePad(*EHPad))
-      continue;
     auto *WE = new WebAssemblyException(EHPad);
     discoverAndMapException(WE, MDT, MDF);
     Exceptions.push_back(WE);
@@ -104,16 +100,6 @@ void WebAssemblyExceptionInfo::discoverAndMapException(
 
   // Map blocks that belong to a catchpad / cleanuppad
   MachineBasicBlock *EHPad = WE->getEHPad();
-
-  // We group catch & catch-all terminate pads together within an exception
-  if (WebAssembly::isCatchTerminatePad(*EHPad)) {
-    assert(EHPad->succ_size() == 1 &&
-           "Catch terminate pad has more than one successors");
-    changeExceptionFor(EHPad, WE);
-    changeExceptionFor(*(EHPad->succ_begin()), WE);
-    return;
-  }
-
   SmallVector<MachineBasicBlock *, 8> WL;
   WL.push_back(EHPad);
   while (!WL.empty()) {
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
index 7cb7f2750ff..52d9d282159 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
@@ -884,13 +884,13 @@ SDValue WebAssemblyTargetLowering::LowerOperation(SDValue Op,
     return LowerFRAMEADDR(Op, DAG);
   case ISD::CopyToReg:
     return LowerCopyToReg(Op, DAG);
-  case ISD::INTRINSIC_WO_CHAIN:
-    return LowerINTRINSIC_WO_CHAIN(Op, DAG);
   case ISD::EXTRACT_VECTOR_ELT:
   case ISD::INSERT_VECTOR_ELT:
     return LowerAccessVectorElement(Op, DAG);
   case ISD::INTRINSIC_VOID:
-    return LowerINTRINSIC_VOID(Op, DAG);
+  case ISD::INTRINSIC_WO_CHAIN:
+  case ISD::INTRINSIC_W_CHAIN:
+    return LowerIntrinsic(Op, DAG);
   case ISD::SIGN_EXTEND_INREG:
     return LowerSIGN_EXTEND_INREG(Op, DAG);
   case ISD::BUILD_VECTOR:
@@ -1035,17 +1035,28 @@ SDValue WebAssemblyTargetLowering::LowerVASTART(SDValue Op,
                       MachinePointerInfo(SV), 0);
 }
 
-SDValue
-WebAssemblyTargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
-                                                   SelectionDAG &DAG) const {
-  unsigned IntNo = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
+SDValue WebAssemblyTargetLowering::LowerIntrinsic(SDValue Op,
+                                                  SelectionDAG &DAG) const {
+  MachineFunction &MF = DAG.getMachineFunction();
+  unsigned IntNo;
+  switch (Op.getOpcode()) {
+  case ISD::INTRINSIC_VOID:
+  case ISD::INTRINSIC_W_CHAIN:
+    IntNo = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();
+    break;
+  case ISD::INTRINSIC_WO_CHAIN:
+    IntNo = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
+    break;
+  default:
+    llvm_unreachable("Invalid intrinsic");
+  }
   SDLoc DL(Op);
+
   switch (IntNo) {
   default:
     return {}; // Don't custom lower most intrinsics.
 
   case Intrinsic::wasm_lsda: {
-    MachineFunction &MF = DAG.getMachineFunction();
     EVT VT = Op.getValueType();
     const TargetLowering &TLI = DAG.getTargetLoweringInfo();
     MVT PtrVT = TLI.getPointerTy(DAG.getDataLayout());
@@ -1055,43 +1066,26 @@ WebAssemblyTargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
     return DAG.getNode(WebAssemblyISD::Wrapper, DL, VT,
                        DAG.getMCSymbol(S, PtrVT));
   }
-  }
-}
-
-SDValue
-WebAssemblyTargetLowering::LowerINTRINSIC_VOID(SDValue Op,
-                                               SelectionDAG &DAG) const {
-  MachineFunction &MF = DAG.getMachineFunction();
-  unsigned IntNo = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();
-  SDLoc DL(Op);
-
-  switch (IntNo) {
-  default:
-    return {}; // Don't custom lower most intrinsics.
 
   case Intrinsic::wasm_throw: {
+    // We only support C++ exceptions for now
     int Tag = cast<ConstantSDNode>(Op.getOperand(2).getNode())->getZExtValue();
-    switch (Tag) {
-    case CPP_EXCEPTION: {
-      const TargetLowering &TLI = DAG.getTargetLoweringInfo();
-      MVT PtrVT = TLI.getPointerTy(DAG.getDataLayout());
-      const char *SymName = MF.createExternalSymbolName("__cpp_exception");
-      SDValue SymNode =
-          DAG.getNode(WebAssemblyISD::Wrapper, DL, PtrVT,
-                      DAG.getTargetExternalSymbol(
-                          SymName, PtrVT, WebAssemblyII::MO_SYMBOL_EVENT));
-      return DAG.getNode(WebAssemblyISD::THROW, DL,
-                         MVT::Other, // outchain type
-                         {
-                             Op.getOperand(0), // inchain
-                             SymNode,          // exception symbol
-                             Op.getOperand(3)  // thrown value
-                         });
-    }
-    default:
+    if (Tag != CPP_EXCEPTION)
       llvm_unreachable("Invalid tag!");
-    }
-    break;
+    const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+    MVT PtrVT = TLI.getPointerTy(DAG.getDataLayout());
+    const char *SymName = MF.createExternalSymbolName("__cpp_exception");
+    SDValue SymNode =
+        DAG.getNode(WebAssemblyISD::Wrapper, DL, PtrVT,
+                    DAG.getTargetExternalSymbol(
+                        SymName, PtrVT, WebAssemblyII::MO_SYMBOL_EVENT));
+    return DAG.getNode(WebAssemblyISD::THROW, DL,
+                       MVT::Other, // outchain type
+                       {
+                           Op.getOperand(0), // inchain
+                           SymNode,          // exception symbol
+                           Op.getOperand(3)  // thrown value
+                       });
   }
   }
 }
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.h b/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.h
index d4b6dcacafc..ee2e863685a 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.h
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.h
@@ -96,8 +96,7 @@ private:
   SDValue LowerJumpTable(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerVASTART(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerCopyToReg(SDValue Op, SelectionDAG &DAG) const;
-  SDValue LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) const;
-  SDValue LowerINTRINSIC_VOID(SDValue Op, SelectionDAG &DAG) const;
+  SDValue LowerIntrinsic(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerSIGN_EXTEND_INREG(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG) const;
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyInstrControl.td b/llvm/lib/Target/WebAssembly/WebAssemblyInstrControl.td
index 1832c74ef59..8160cab7520 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyInstrControl.td
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyInstrControl.td
@@ -141,23 +141,11 @@ let Predicates = [HasExceptionHandling] in {
 
 // Throwing an exception: throw / rethrow
 let isTerminator = 1, hasCtrlDep = 1, isBarrier = 1 in {
-defm THROW_I32 : I<(outs), (ins event_op:$tag, I32:$val),
-                   (outs), (ins event_op:$tag),
-                   [(WebAssemblythrow (WebAssemblywrapper texternalsym:$tag),
-                                      I32:$val)],
-                   "throw   \t$tag, $val", "throw   \t$tag",
-                   0x08>;
-defm THROW_I64 : I<(outs), (ins event_op:$tag, I64:$val),
-                   (outs), (ins event_op:$tag),
-                   [(WebAssemblythrow (WebAssemblywrapper texternalsym:$tag),
-                                      I64:$val)],
-                   "throw   \t$tag, $val", "throw   \t$tag",
-                   0x08>;
-defm RETHROW : NRI<(outs), (ins bb_op:$dst), [], "rethrow \t$dst", 0x09>;
-let isCodeGenOnly = 1 in
-// This is used when the destination for rethrow is the caller function. This
-// will be converted to a rethrow in CFGStackify.
-defm RETHROW_TO_CALLER : NRI<(outs), (ins), [], "rethrow">;
+defm THROW : I<(outs), (ins event_op:$tag, variable_ops),
+               (outs), (ins event_op:$tag),
+               [(WebAssemblythrow (WebAssemblywrapper texternalsym:$tag))],
+               "throw   \t$tag", "throw   \t$tag", 0x08>;
+defm RETHROW : NRI<(outs), (ins), [(int_wasm_rethrow)], "rethrow", 0x09>;
 } // isTerminator = 1, hasCtrlDep = 1, isBarrier = 1
 
 // Region within which an exception is caught: try / end_try
@@ -166,24 +154,33 @@ defm TRY     : NRI<(outs), (ins Signature:$sig), [], "try     \t$sig", 0x06>;
 defm END_TRY : NRI<(outs), (ins), [], "end_try", 0x0b>;
 } // Uses = [VALUE_STACK], Defs = [VALUE_STACK]
 
-// Catching an exception: catch / catch_all
-let hasCtrlDep = 1, hasSideEffects = 1 in {
-defm CATCH_I32 : I<(outs I32:$dst), (ins i32imm:$tag),
-                   (outs), (ins i32imm:$tag),
-                   [(set I32:$dst, (int_wasm_catch imm:$tag))],
-                   "i32.catch   \t$dst, $tag", "i32.catch   \t$tag", 0x07>;
-defm CATCH_I64 : I<(outs I64:$dst), (ins i32imm:$tag),
-                   (outs), (ins i32imm:$tag),
-                   [(set I64:$dst, (int_wasm_catch imm:$tag))],
-                   "i64.catch   \t$dst, $tag", "i64.catch   \t$tag", 0x07>;
-defm CATCH_ALL : NRI<(outs), (ins), [], "catch_all", 0x05>;
-}
+// Catching an exception: catch / extract_exception
+let hasCtrlDep = 1, hasSideEffects = 1 in
+defm CATCH : I<(outs EXCEPT_REF:$dst), (ins), (outs), (ins), [],
+               "catch   \t$dst", "catch", 0x07>;
+
+// Querying / extracing exception: br_on_exn
+// br_on_exn queries an except_ref to see if it matches the corresponding
+// exception tag index. If true it branches to the given label and pushes the
+// corresponding argument values of the exception onto the stack.
+let isBranch = 1, isTerminator = 1, hasCtrlDep = 1 in
+defm BR_ON_EXN : I<(outs), (ins bb_op:$dst, event_op:$tag, EXCEPT_REF:$exn),
+                   (outs), (ins bb_op:$dst, event_op:$tag), [],
+                   "br_on_exn \t$dst, $tag, $exn", "br_on_exn \t$dst, $tag",
+                   0x0a>;
+// This is a pseudo instruction that simulates popping a value from stack, which
+// has been pushed by br_on_exn
+let isCodeGenOnly = 1, hasSideEffects = 1 in
+defm EXTRACT_EXCEPTION_I32 : NRI<(outs I32:$dst), (ins),
+                                 [(set I32:$dst, (int_wasm_extract_exception))],
+                                 "extract_exception\t$dst">;
 
 // Pseudo instructions: cleanupret / catchret
 let isTerminator = 1, hasSideEffects = 1, isBarrier = 1, hasCtrlDep = 1,
-    isCodeGenOnly = 1, isEHScopeReturn = 1 in {
-  defm CLEANUPRET : NRI<(outs), (ins), [(cleanupret)], "", 0>;
+    isPseudo = 1, isEHScopeReturn = 1 in {
+  defm CLEANUPRET : NRI<(outs), (ins), [(cleanupret)], "cleanupret", 0>;
   defm CATCHRET : NRI<(outs), (ins bb_op:$dst, bb_op:$from),
-                   [(catchret bb:$dst, bb:$from)], "", 0>;
-}
+                      [(catchret bb:$dst, bb:$from)], "catchret", 0>;
+} // isTerminator = 1, hasSideEffects = 1, isBarrier = 1, hasCtrlDep = 1,
+  // isPseudo = 1, isEHScopeReturn = 1
 }
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.cpp
index 9592130a536..995dc88bb33 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.cpp
@@ -134,6 +134,17 @@ bool WebAssemblyInstrInfo::analyzeBranch(MachineBasicBlock &MBB,
       else
         FBB = MI.getOperand(0).getMBB();
       break;
+    case WebAssembly::BR_ON_EXN:
+      if (HaveCond)
+        return true;
+      // If we're running after CFGStackify, we can't optimize further.
+      if (!MI.getOperand(0).isMBB())
+        return true;
+      Cond.push_back(MachineOperand::CreateImm(true));
+      Cond.push_back(MI.getOperand(2));
+      TBB = MI.getOperand(0).getMBB();
+      HaveCond = true;
+      break;
     }
     if (MI.isBarrier())
       break;
@@ -179,9 +190,22 @@ unsigned WebAssemblyInstrInfo::insertBranch(
 
   assert(Cond.size() == 2 && "Expected a flag and a successor block");
 
+  MachineFunction &MF = *MBB.getParent();
+  auto &MRI = MF.getRegInfo();
+  bool IsBrOnExn = Cond[1].isReg() && MRI.getRegClass(Cond[1].getReg()) ==
+                                          &WebAssembly::EXCEPT_REFRegClass;
+
   if (Cond[0].getImm()) {
-    BuildMI(&MBB, DL, get(WebAssembly::BR_IF)).addMBB(TBB).add(Cond[1]);
+    if (IsBrOnExn) {
+      const char *CPPExnSymbol = MF.createExternalSymbolName("__cpp_exception");
+      BuildMI(&MBB, DL, get(WebAssembly::BR_ON_EXN))
+          .addMBB(TBB)
+          .addExternalSymbol(CPPExnSymbol, WebAssemblyII::MO_SYMBOL_EVENT)
+          .add(Cond[1]);
+    } else
+      BuildMI(&MBB, DL, get(WebAssembly::BR_IF)).addMBB(TBB).add(Cond[1]);
   } else {
+    assert(!IsBrOnExn && "br_on_exn does not have a reversed condition");
     BuildMI(&MBB, DL, get(WebAssembly::BR_UNLESS)).addMBB(TBB).add(Cond[1]);
   }
   if (!FBB)
@@ -193,7 +217,15 @@ unsigned WebAssemblyInstrInfo::insertBranch(
 
 bool WebAssemblyInstrInfo::reverseBranchCondition(
     SmallVectorImpl<MachineOperand> &Cond) const {
-  assert(Cond.size() == 2 && "Expected a flag and a successor block");
+  assert(Cond.size() == 2 && "Expected a flag and a condition expression");
+
+  // br_on_exn's condition cannot be reversed
+  MachineFunction &MF = *Cond[1].getParent()->getParent()->getParent();
+  auto &MRI = MF.getRegInfo();
+  if (Cond[1].isReg() &&
+      MRI.getRegClass(Cond[1].getReg()) == &WebAssembly::EXCEPT_REFRegClass)
+    return true;
+
   Cond.front() = MachineOperand::CreateImm(!Cond.front().getImm());
   return false;
 }
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.td b/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.td
index a609267aa5e..35d663903e4 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.td
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.td
@@ -66,7 +66,7 @@ def SDT_WebAssemblyArgument : SDTypeProfile<1, 1, [SDTCisVT<1, i32>]>;
 def SDT_WebAssemblyReturn   : SDTypeProfile<0, -1, []>;
 def SDT_WebAssemblyWrapper  : SDTypeProfile<1, 1, [SDTCisSameAs<0, 1>,
                                                    SDTCisPtrTy<0>]>;
-def SDT_WebAssemblyThrow    : SDTypeProfile<0, 2, [SDTCisPtrTy<0>]>;
+def SDT_WebAssemblyThrow    : SDTypeProfile<0, -1, [SDTCisPtrTy<0>]>;
 
 //===----------------------------------------------------------------------===//
 // WebAssembly-specific DAG Nodes.
@@ -94,7 +94,7 @@ def WebAssemblyreturn   : SDNode<"WebAssemblyISD::RETURN",
 def WebAssemblywrapper  : SDNode<"WebAssemblyISD::Wrapper",
                                  SDT_WebAssemblyWrapper>;
 def WebAssemblythrow : SDNode<"WebAssemblyISD::THROW", SDT_WebAssemblyThrow,
-                              [SDNPHasChain]>;
+                              [SDNPHasChain, SDNPVariadic]>;
 
 //===----------------------------------------------------------------------===//
 // WebAssembly-specific Operands.
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyLateEHPrepare.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyLateEHPrepare.cpp
index caedd140a1c..030f09b8bf4 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyLateEHPrepare.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyLateEHPrepare.cpp
@@ -15,6 +15,7 @@
 #include "WebAssembly.h"
 #include "WebAssemblySubtarget.h"
 #include "WebAssemblyUtilities.h"
+#include "llvm/ADT/SmallSet.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/WasmEHFuncInfo.h"
 #include "llvm/MC/MCAsmInfo.h"
@@ -25,19 +26,14 @@ using namespace llvm;
 namespace {
 class WebAssemblyLateEHPrepare final : public MachineFunctionPass {
   StringRef getPassName() const override {
-    return "WebAssembly Prepare Exception";
+    return "WebAssembly Late Prepare Exception";
   }
 
   bool runOnMachineFunction(MachineFunction &MF) override;
-
   bool removeUnnecessaryUnreachables(MachineFunction &MF);
   bool replaceFuncletReturns(MachineFunction &MF);
-  bool hoistCatches(MachineFunction &MF);
-  bool addCatchAlls(MachineFunction &MF);
-  bool addRethrows(MachineFunction &MF);
-  bool ensureSingleBBTermPads(MachineFunction &MF);
-  bool mergeTerminatePads(MachineFunction &MF);
-  bool addCatchAllTerminatePads(MachineFunction &MF);
+  bool addCatches(MachineFunction &MF);
+  bool addExceptionExtraction(MachineFunction &MF);
 
 public:
   static char ID; // Pass identification, replacement for typeid
@@ -112,15 +108,11 @@ bool WebAssemblyLateEHPrepare::runOnMachineFunction(MachineFunction &MF) {
 
   bool Changed = false;
   Changed |= removeUnnecessaryUnreachables(MF);
-  Changed |= addRethrows(MF);
   if (!MF.getFunction().hasPersonalityFn())
     return Changed;
   Changed |= replaceFuncletReturns(MF);
-  Changed |= hoistCatches(MF);
-  Changed |= addCatchAlls(MF);
-  Changed |= ensureSingleBBTermPads(MF);
-  Changed |= mergeTerminatePads(MF);
-  Changed |= addCatchAllTerminatePads(MF);
+  Changed |= addCatches(MF);
+  Changed |= addExceptionExtraction(MF);
   return Changed;
 }
 
@@ -129,7 +121,8 @@ bool WebAssemblyLateEHPrepare::removeUnnecessaryUnreachables(
   bool Changed = false;
   for (auto &MBB : MF) {
     for (auto &MI : MBB) {
-      if (!WebAssembly::isThrow(MI))
+      if (MI.getOpcode() != WebAssembly::THROW &&
+          MI.getOpcode() != WebAssembly::RETHROW)
         continue;
       Changed = true;
 
@@ -152,7 +145,6 @@ bool WebAssemblyLateEHPrepare::removeUnnecessaryUnreachables(
 bool WebAssemblyLateEHPrepare::replaceFuncletReturns(MachineFunction &MF) {
   bool Changed = false;
   const auto &TII = *MF.getSubtarget<WebAssemblySubtarget>().getInstrInfo();
-  auto *EHInfo = MF.getWasmEHFuncInfo();
 
   for (auto &MBB : MF) {
     auto Pos = MBB.getFirstTerminator();
@@ -173,13 +165,7 @@ bool WebAssemblyLateEHPrepare::replaceFuncletReturns(MachineFunction &MF) {
     }
     case WebAssembly::CLEANUPRET: {
       // Replace a cleanupret with a rethrow
-      if (EHInfo->hasThrowUnwindDest(&MBB))
-        BuildMI(MBB, TI, TI->getDebugLoc(), TII.get(WebAssembly::RETHROW))
-            .addMBB(EHInfo->getThrowUnwindDest(&MBB));
-      else
-        BuildMI(MBB, TI, TI->getDebugLoc(),
-                TII.get(WebAssembly::RETHROW_TO_CALLER));
-
+      BuildMI(MBB, TI, TI->getDebugLoc(), TII.get(WebAssembly::RETHROW));
       TI->eraseFromParent();
       Changed = true;
       break;
@@ -189,233 +175,158 @@ bool WebAssemblyLateEHPrepare::replaceFuncletReturns(MachineFunction &MF) {
   return Changed;
 }
 
-// Hoist catch instructions to the beginning of their matching EH pad BBs in
-// case,
-// (1) catch instruction is not the first instruction in EH pad.
-// ehpad:
-//   some_other_instruction
-//   ...
-//   %exn = catch 0
-// (2) catch instruction is in a non-EH pad BB. For example,
-// ehpad:
-//   br bb0
-// bb0:
-//   %exn = catch 0
-bool WebAssemblyLateEHPrepare::hoistCatches(MachineFunction &MF) {
-  bool Changed = false;
-  SmallVector<MachineInstr *, 16> Catches;
-  for (auto &MBB : MF)
-    for (auto &MI : MBB)
-      if (WebAssembly::isCatch(MI))
-        Catches.push_back(&MI);
-
-  for (auto *Catch : Catches) {
-    MachineBasicBlock *EHPad = getMatchingEHPad(Catch);
-    assert(EHPad && "No matching EH pad for catch");
-    if (EHPad->begin() == Catch)
-      continue;
-    Changed = true;
-    EHPad->insert(EHPad->begin(), Catch->removeFromParent());
-  }
-  return Changed;
-}
-
-// Add catch_all to beginning of cleanup pads.
-bool WebAssemblyLateEHPrepare::addCatchAlls(MachineFunction &MF) {
+// Add catch instruction to beginning of catchpads and cleanuppads.
+bool WebAssemblyLateEHPrepare::addCatches(MachineFunction &MF) {
   bool Changed = false;
   const auto &TII = *MF.getSubtarget<WebAssemblySubtarget>().getInstrInfo();
-
+  MachineRegisterInfo &MRI = MF.getRegInfo();
   for (auto &MBB : MF) {
-    if (!MBB.isEHPad())
-      continue;
-    // This runs after hoistCatches(), so we assume that if there is a catch,
-    // that should be the first instruction in an EH pad.
-    if (!WebAssembly::isCatch(*MBB.begin())) {
+    if (MBB.isEHPad()) {
       Changed = true;
+      unsigned DstReg =
+          MRI.createVirtualRegister(&WebAssembly::EXCEPT_REFRegClass);
       BuildMI(MBB, MBB.begin(), MBB.begin()->getDebugLoc(),
-              TII.get(WebAssembly::CATCH_ALL));
+              TII.get(WebAssembly::CATCH), DstReg);
     }
   }
   return Changed;
 }
 
-// Add a 'rethrow' instruction after __cxa_rethrow() call
-bool WebAssemblyLateEHPrepare::addRethrows(MachineFunction &MF) {
-  bool Changed = false;
+// Wasm uses 'br_on_exn' instruction to check the tag of an exception. It takes
+// except_ref type object returned by 'catch', and branches to the destination
+// if it matches a given tag. We currently use __cpp_exception symbol to
+// represent the tag for all C++ exceptions.
+//
+// block $l (result i32)
+//   ...
+//   ;; except_ref $e is on the stack at this point
+//   br_on_exn $l $e ;; branch to $l with $e's arguments
+//   ...
+// end
+// ;; Here we expect the extracted values are on top of the wasm value stack
+// ... Handle exception using values ...
+//
+// br_on_exn takes an except_ref object and branches if it matches the given
+// tag. There can be multiple br_on_exn instructions if we want to match for
+// another tag, but for now we only test for __cpp_exception tag, and if it does
+// not match, i.e., it is a foreign exception, we rethrow it.
+//
+// In the destination BB that's the target of br_on_exn, extracted exception
+// values (in C++'s case a single i32, which represents an exception pointer)
+// are placed on top of the wasm stack. Because we can't model wasm stack in
+// LLVM instruction, we use 'extract_exception' pseudo instruction to retrieve
+// it. The pseudo instruction will be deleted later.
+bool WebAssemblyLateEHPrepare::addExceptionExtraction(MachineFunction &MF) {
   const auto &TII = *MF.getSubtarget<WebAssemblySubtarget>().getInstrInfo();
   auto *EHInfo = MF.getWasmEHFuncInfo();
-
-  for (auto &MBB : MF)
+  SmallVector<MachineInstr *, 16> ExtractInstrs;
+  for (auto &MBB : MF) {
     for (auto &MI : MBB) {
-      // Check if it is a call to __cxa_rethrow()
-      if (!MI.isCall())
-        continue;
-      MachineOperand &CalleeOp = MI.getOperand(0);
-      if (!CalleeOp.isGlobal() ||
-          CalleeOp.getGlobal()->getName() != WebAssembly::CxaRethrowFn)
-        continue;
-
-      // Now we have __cxa_rethrow() call
-      Changed = true;
-      auto InsertPt = std::next(MachineBasicBlock::iterator(MI));
-      while (InsertPt != MBB.end() && InsertPt->isLabel()) // Skip EH_LABELs
-        ++InsertPt;
-      MachineInstr *Rethrow = nullptr;
-      if (EHInfo->hasThrowUnwindDest(&MBB))
-        Rethrow = BuildMI(MBB, InsertPt, MI.getDebugLoc(),
-                          TII.get(WebAssembly::RETHROW))
-                      .addMBB(EHInfo->getThrowUnwindDest(&MBB));
-      else
-        Rethrow = BuildMI(MBB, InsertPt, MI.getDebugLoc(),
-                          TII.get(WebAssembly::RETHROW_TO_CALLER));
-
-      // Because __cxa_rethrow does not return, the instruction after the
-      // rethrow should be an unreachable or a branch to another BB that should
-      // eventually lead to an unreachable. Delete it because rethrow itself is
-      // a terminator, and also delete non-EH pad successors if any.
-      MBB.erase(std::next(MachineBasicBlock::iterator(Rethrow)), MBB.end());
-      SmallVector<MachineBasicBlock *, 8> NonPadSuccessors;
-      for (auto *Succ : MBB.successors())
-        if (!Succ->isEHPad())
-          NonPadSuccessors.push_back(Succ);
-      for (auto *Succ : NonPadSuccessors)
-        MBB.removeSuccessor(Succ);
-      eraseDeadBBsAndChildren(NonPadSuccessors);
+      if (MI.getOpcode() == WebAssembly::EXTRACT_EXCEPTION_I32) {
+        if (MI.getOperand(0).isDead())
+          MI.eraseFromParent();
+        else
+          ExtractInstrs.push_back(&MI);
+      }
     }
-  return Changed;
-}
-
-// Terminate pads are an single-BB EH pad in the form of
-// termpad:
-//   %exn = catch 0
-//   call @__clang_call_terminate(%exn)
-//   unreachable
-// (There can be local.set and local.gets before the call if we didn't run
-// RegStackify)
-// But code transformations can change or add more control flow, so the call to
-// __clang_call_terminate() function may not be in the original EH pad anymore.
-// This ensures every terminate pad is a single BB in the form illustrated
-// above.
-bool WebAssemblyLateEHPrepare::ensureSingleBBTermPads(MachineFunction &MF) {
-  const auto &TII = *MF.getSubtarget<WebAssemblySubtarget>().getInstrInfo();
+  }
+  if (ExtractInstrs.empty())
+    return false;
 
-  // Find calls to __clang_call_terminate()
-  SmallVector<MachineInstr *, 8> ClangCallTerminateCalls;
-  for (auto &MBB : MF)
-    for (auto &MI : MBB)
+  // Find terminate pads.
+  SmallSet<MachineBasicBlock *, 8> TerminatePads;
+  for (auto &MBB : MF) {
+    for (auto &MI : MBB) {
       if (MI.isCall()) {
         const MachineOperand &CalleeOp = MI.getOperand(0);
         if (CalleeOp.isGlobal() && CalleeOp.getGlobal()->getName() ==
                                        WebAssembly::ClangCallTerminateFn)
-          ClangCallTerminateCalls.push_back(&MI);
+          TerminatePads.insert(getMatchingEHPad(&MI));
       }
-
-  bool Changed = false;
-  for (auto *Call : ClangCallTerminateCalls) {
-    MachineBasicBlock *EHPad = getMatchingEHPad(Call);
-    assert(EHPad && "No matching EH pad for catch");
-
-    // If it is already the form we want, skip it
-    if (Call->getParent() == EHPad &&
-        Call->getNextNode()->getOpcode() == WebAssembly::UNREACHABLE)
-      continue;
-
-    // In case the __clang_call_terminate() call is not in its matching EH pad,
-    // move the call to the end of EH pad and add an unreachable instruction
-    // after that. Delete all successors and their children if any, because here
-    // the program terminates.
-    Changed = true;
-    MachineInstr *Catch = &*EHPad->begin();
-    // This runs after hoistCatches(), so catch instruction should be at the top
-    assert(WebAssembly::isCatch(*Catch));
-    // Takes the result register of the catch instruction as argument. There may
-    // have been some other local.set/local.gets in between, but at this point
-    // we don't care.
-    Call->getOperand(1).setReg(Catch->getOperand(0).getReg());
-    auto InsertPos = std::next(MachineBasicBlock::iterator(Catch));
-    EHPad->insert(InsertPos, Call->removeFromParent());
-    BuildMI(*EHPad, InsertPos, Call->getDebugLoc(),
-            TII.get(WebAssembly::UNREACHABLE));
-    EHPad->erase(InsertPos, EHPad->end());
-    SmallVector<MachineBasicBlock *, 8> Succs(EHPad->succ_begin(),
-                                              EHPad->succ_end());
-    for (auto *Succ : Succs)
-      EHPad->removeSuccessor(Succ);
-    eraseDeadBBsAndChildren(Succs);
+    }
   }
-  return Changed;
-}
 
-// In case there are multiple terminate pads, merge them into one for code size.
-// This runs after ensureSingleBBTermPads() and assumes every terminate pad is a
-// single BB.
-// In principle this violates EH scope relationship because it can merge
-// multiple inner EH scopes, each of which is in different outer EH scope. But
-// getEHScopeMembership() function will not be called after this, so it is fine.
-bool WebAssemblyLateEHPrepare::mergeTerminatePads(MachineFunction &MF) {
-  SmallVector<MachineBasicBlock *, 8> TermPads;
-  for (auto &MBB : MF)
-    if (WebAssembly::isCatchTerminatePad(MBB))
-      TermPads.push_back(&MBB);
-  if (TermPads.empty())
-    return false;
-
-  MachineBasicBlock *UniqueTermPad = TermPads.front();
-  for (auto *TermPad :
-       llvm::make_range(std::next(TermPads.begin()), TermPads.end())) {
-    SmallVector<MachineBasicBlock *, 2> Preds(TermPad->pred_begin(),
-                                              TermPad->pred_end());
-    for (auto *Pred : Preds)
-      Pred->replaceSuccessor(TermPad, UniqueTermPad);
-    TermPad->eraseFromParent();
+  for (auto *Extract : ExtractInstrs) {
+    MachineBasicBlock *EHPad = getMatchingEHPad(Extract);
+    assert(EHPad && "No matching EH pad for extract_exception");
+    MachineInstr *Catch = &*EHPad->begin();
+    if (Catch->getNextNode() != Extract)
+      EHPad->insert(Catch->getNextNode(), Extract->removeFromParent());
+
+    // - Before:
+    // ehpad:
+    //   %exnref:except_ref = catch
+    //   %exn:i32 = extract_exception
+    //   ... use exn ...
+    //
+    // - After:
+    // ehpad:
+    //   %exnref:except_ref = catch
+    //   br_on_exn %thenbb, $__cpp_exception, %exnref
+    //   br %elsebb
+    // elsebb:
+    //   rethrow
+    // thenbb:
+    //   %exn:i32 = extract_exception
+    //   ... use exn ...
+    unsigned ExnRefReg = Catch->getOperand(0).getReg();
+    auto *ThenMBB = MF.CreateMachineBasicBlock();
+    auto *ElseMBB = MF.CreateMachineBasicBlock();
+    MF.insert(std::next(MachineFunction::iterator(EHPad)), ElseMBB);
+    MF.insert(std::next(MachineFunction::iterator(ElseMBB)), ThenMBB);
+    ThenMBB->splice(ThenMBB->end(), EHPad, Extract, EHPad->end());
+    ThenMBB->transferSuccessors(EHPad);
+    EHPad->addSuccessor(ThenMBB);
+    EHPad->addSuccessor(ElseMBB);
+
+    DebugLoc DL = Extract->getDebugLoc();
+    const char *CPPExnSymbol = MF.createExternalSymbolName("__cpp_exception");
+    BuildMI(EHPad, DL, TII.get(WebAssembly::BR_ON_EXN))
+        .addMBB(ThenMBB)
+        .addExternalSymbol(CPPExnSymbol, WebAssemblyII::MO_SYMBOL_EVENT)
+        .addReg(ExnRefReg);
+    BuildMI(EHPad, DL, TII.get(WebAssembly::BR)).addMBB(ElseMBB);
+
+    // When this is a terminate pad with __clang_call_terminate() call, we don't
+    // rethrow it anymore and call __clang_call_terminate() with a nullptr
+    // argument, which will call std::terminate().
+    //
+    // - Before:
+    // ehpad:
+    //   %exnref:except_ref = catch
+    //   %exn:i32 = extract_exception
+    //   call @__clang_call_terminate(%exn)
+    //   unreachable
+    //
+    // - After:
+    // ehpad:
+    //   %exnref:except_ref = catch
+    //   br_on_exn %thenbb, $__cpp_exception, %exnref
+    //   br %elsebb
+    // elsebb:
+    //   call @__clang_call_terminate(0)
+    //   unreachable
+    // thenbb:
+    //   %exn:i32 = extract_exception
+    //   call @__clang_call_terminate(%exn)
+    //   unreachable
+    if (TerminatePads.count(EHPad)) {
+      Function *ClangCallTerminateFn =
+          MF.getFunction().getParent()->getFunction(
+              WebAssembly::ClangCallTerminateFn);
+      assert(ClangCallTerminateFn &&
+             "There is no __clang_call_terminate() function");
+      BuildMI(ElseMBB, DL, TII.get(WebAssembly::CALL_VOID))
+          .addGlobalAddress(ClangCallTerminateFn)
+          .addImm(0);
+      BuildMI(ElseMBB, DL, TII.get(WebAssembly::UNREACHABLE));
+
+    } else {
+      BuildMI(ElseMBB, DL, TII.get(WebAssembly::RETHROW));
+      if (EHInfo->hasEHPadUnwindDest(EHPad))
+        EHInfo->setThrowUnwindDest(ElseMBB, EHInfo->getEHPadUnwindDest(EHPad));
+    }
   }
-  return true;
-}
 
-// Terminate pads are cleanup pads, so they should start with a 'catch_all'
-// instruction. But in the Itanium model, when we have a C++ exception object,
-// we pass them to __clang_call_terminate function, which calls __cxa_end_catch
-// with the passed exception pointer and then std::terminate. This is the reason
-// that terminate pads are generated with not a catch_all but a catch
-// instruction in clang and earlier llvm passes. Here we append a terminate pad
-// with a catch_all after each existing terminate pad so we can also catch
-// foreign exceptions. For every terminate pad:
-//   %exn = catch 0
-//   call @__clang_call_terminate(%exn)
-//   unreachable
-// We append this BB right after that:
-//   catch_all
-//   call @std::terminate()
-//   unreachable
-bool WebAssemblyLateEHPrepare::addCatchAllTerminatePads(MachineFunction &MF) {
-  const auto &TII = *MF.getSubtarget<WebAssemblySubtarget>().getInstrInfo();
-  SmallVector<MachineBasicBlock *, 8> TermPads;
-  for (auto &MBB : MF)
-    if (WebAssembly::isCatchTerminatePad(MBB))
-      TermPads.push_back(&MBB);
-  if (TermPads.empty())
-    return false;
-
-  Function *StdTerminateFn =
-      MF.getFunction().getParent()->getFunction(WebAssembly::StdTerminateFn);
-  assert(StdTerminateFn && "There is no std::terminate() function");
-  for (auto *CatchTermPad : TermPads) {
-    DebugLoc DL = CatchTermPad->findDebugLoc(CatchTermPad->begin());
-    auto *CatchAllTermPad = MF.CreateMachineBasicBlock();
-    MF.insert(std::next(MachineFunction::iterator(CatchTermPad)),
-              CatchAllTermPad);
-    CatchAllTermPad->setIsEHPad();
-    BuildMI(CatchAllTermPad, DL, TII.get(WebAssembly::CATCH_ALL));
-    BuildMI(CatchAllTermPad, DL, TII.get(WebAssembly::CALL_VOID))
-        .addGlobalAddress(StdTerminateFn);
-    BuildMI(CatchAllTermPad, DL, TII.get(WebAssembly::UNREACHABLE));
-
-    // Actually this CatchAllTermPad (new terminate pad with a catch_all) is not
-    // a successor of an existing terminate pad. CatchAllTermPad should have all
-    // predecessors CatchTermPad has instead. This is a hack to force
-    // CatchAllTermPad be always sorted right after CatchTermPad; the correct
-    // predecessor-successor relationships will be restored in CFGStackify pass.
-    CatchTermPad->addSuccessor(CatchAllTermPad);
-  }
   return true;
 }
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyMCInstLower.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyMCInstLower.cpp
index 482131fa549..e34ebdbb14d 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyMCInstLower.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyMCInstLower.cpp
@@ -36,7 +36,7 @@ using namespace llvm;
 
 // This disables the removal of registers when lowering into MC, as required
 // by some current tests.
-static cl::opt<bool>
+cl::opt<bool>
     WasmKeepRegisters("wasm-keep-registers", cl::Hidden,
                       cl::desc("WebAssembly: output stack registers in"
                                " instruction output for test purposes only."),
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyRegStackify.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyRegStackify.cpp
index 7b9b494527b..957a713734b 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyRegStackify.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyRegStackify.cpp
@@ -317,6 +317,18 @@ static bool IsSafeToMove(const MachineInstr *Def, const MachineInstr *Insert,
                          AliasAnalysis &AA, const MachineRegisterInfo &MRI) {
   assert(Def->getParent() == Insert->getParent());
 
+  // 'catch' and 'extract_exception' should be the first instruction of a BB and
+  // cannot move.
+  if (Def->getOpcode() == WebAssembly::CATCH ||
+      Def->getOpcode() == WebAssembly::EXTRACT_EXCEPTION_I32) {
+    const MachineBasicBlock *MBB = Def->getParent();
+    auto NextI = std::next(MachineBasicBlock::const_iterator(Def));
+    for (auto E = MBB->end(); NextI != E && NextI->isDebugInstr(); ++NextI)
+      ;
+    if (NextI != Insert)
+      return false;
+  }
+
   // Check for register dependencies.
   SmallVector<unsigned, 4> MutableRegisters;
   for (const MachineOperand &MO : Def->operands()) {
@@ -819,6 +831,24 @@ bool WebAssemblyRegStackify::runOnMachineFunction(MachineFunction &MF) {
         if (WebAssembly::isArgument(*Def))
           continue;
 
+        // Currently catch's return value register cannot be stackified, because
+        // the wasm LLVM backend currently does not support live-in values
+        // entering blocks, which is a part of multi-value proposal.
+        //
+        // Once we support live-in values of wasm blocks, this can be:
+        // catch                           ; push except_ref value onto stack
+        // block except_ref -> i32
+        // br_on_exn $__cpp_exception      ; pop the except_ref value
+        // end_block
+        //
+        // But because we don't support it yet, the catch instruction's dst
+        // register should be assigned to a local to be propagated across
+        // 'block' boundary now.
+        //
+        // TODO Fix this once we support the multi-value proposal.
+        if (Def->getOpcode() == WebAssembly::CATCH)
+          continue;
+
         // Decide which strategy to take. Prefer to move a single-use value
         // over cloning it, and prefer cloning over introducing a tee.
         // For moving, we require the def to be in the same block as the use;
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
index b2713daac2a..c235fd21d06 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
@@ -301,8 +301,10 @@ void WebAssemblyPassConfig::addPreEmitPass() {
   addPass(createWebAssemblyFixIrreducibleControlFlow());
 
   // Do various transformations for exception handling.
+  // Every CFG-changing optimizations should come before this.
   addPass(createWebAssemblyLateEHPrepare());
 
+  // Preparations and optimizations related to register stackification.
   if (getOptLevel() != CodeGenOpt::None) {
     // LiveIntervals isn't commonly run this late. Re-establish preconditions.
     addPass(createWebAssemblyPrepareForLiveIntervals());
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyUtilities.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyUtilities.cpp
index f1cb5799da2..4edd54602fb 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyUtilities.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyUtilities.cpp
@@ -242,50 +242,10 @@ bool WebAssembly::isMarker(const MachineInstr &MI) {
   }
 }
 
-bool WebAssembly::isThrow(const MachineInstr &MI) {
-  switch (MI.getOpcode()) {
-  case WebAssembly::THROW_I32:
-  case WebAssembly::THROW_I32_S:
-  case WebAssembly::THROW_I64:
-  case WebAssembly::THROW_I64_S:
-    return true;
-  default:
-    return false;
-  }
-}
-
-bool WebAssembly::isRethrow(const MachineInstr &MI) {
-  switch (MI.getOpcode()) {
-  case WebAssembly::RETHROW:
-  case WebAssembly::RETHROW_S:
-  case WebAssembly::RETHROW_TO_CALLER:
-  case WebAssembly::RETHROW_TO_CALLER_S:
-    return true;
-  default:
-    return false;
-  }
-}
-
-bool WebAssembly::isCatch(const MachineInstr &MI) {
-  switch (MI.getOpcode()) {
-  case WebAssembly::CATCH_I32:
-  case WebAssembly::CATCH_I32_S:
-  case WebAssembly::CATCH_I64:
-  case WebAssembly::CATCH_I64_S:
-  case WebAssembly::CATCH_ALL:
-  case WebAssembly::CATCH_ALL_S:
-    return true;
-  default:
-    return false;
-  }
-}
-
 bool WebAssembly::mayThrow(const MachineInstr &MI) {
   switch (MI.getOpcode()) {
-  case WebAssembly::THROW_I32:
-  case WebAssembly::THROW_I32_S:
-  case WebAssembly::THROW_I64:
-  case WebAssembly::THROW_I64_S:
+  case WebAssembly::THROW:
+  case WebAssembly::THROW_S:
   case WebAssembly::RETHROW:
   case WebAssembly::RETHROW_S:
     return true;
@@ -308,41 +268,3 @@ bool WebAssembly::mayThrow(const MachineInstr &MI) {
     return false;
   return true;
 }
-
-bool WebAssembly::isCatchTerminatePad(const MachineBasicBlock &MBB) {
-  if (!MBB.isEHPad())
-    return false;
-  bool SeenCatch = false;
-  for (auto &MI : MBB) {
-    if (MI.getOpcode() == WebAssembly::CATCH_I32 ||
-        MI.getOpcode() == WebAssembly::CATCH_I64 ||
-        MI.getOpcode() == WebAssembly::CATCH_I32_S ||
-        MI.getOpcode() == WebAssembly::CATCH_I64_S)
-      SeenCatch = true;
-    if (SeenCatch && MI.isCall()) {
-      const MachineOperand &CalleeOp = MI.getOperand(getCalleeOpNo(MI));
-      if (CalleeOp.isGlobal() &&
-          CalleeOp.getGlobal()->getName() == ClangCallTerminateFn)
-        return true;
-    }
-  }
-  return false;
-}
-
-bool WebAssembly::isCatchAllTerminatePad(const MachineBasicBlock &MBB) {
-  if (!MBB.isEHPad())
-    return false;
-  bool SeenCatchAll = false;
-  for (auto &MI : MBB) {
-    if (MI.getOpcode() == WebAssembly::CATCH_ALL ||
-        MI.getOpcode() == WebAssembly::CATCH_ALL_S)
-      SeenCatchAll = true;
-    if (SeenCatchAll && MI.isCall()) {
-      const MachineOperand &CalleeOp = MI.getOperand(getCalleeOpNo(MI));
-      if (CalleeOp.isGlobal() &&
-          CalleeOp.getGlobal()->getName() == StdTerminateFn)
-        return true;
-    }
-  }
-  return false;
-}
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyUtilities.h b/llvm/lib/Target/WebAssembly/WebAssemblyUtilities.h
index 72dc945d53f..f80b49662ea 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyUtilities.h
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyUtilities.h
@@ -30,22 +30,12 @@ bool isChild(const MachineInstr &MI, const WebAssemblyFunctionInfo &MFI);
 bool isCallDirect(const MachineInstr &MI);
 bool isCallIndirect(const MachineInstr &MI);
 bool isMarker(const MachineInstr &MI);
-bool isThrow(const MachineInstr &MI);
-bool isRethrow(const MachineInstr &MI);
-bool isCatch(const MachineInstr &MI);
 bool mayThrow(const MachineInstr &MI);
 
 /// Returns the operand number of a callee, assuming the argument is a call
 /// instruction.
 unsigned getCalleeOpNo(const MachineInstr &MI);
 
-/// Returns if the given BB is a single BB terminate pad which starts with a
-/// 'catch' instruction.
-bool isCatchTerminatePad(const MachineBasicBlock &MBB);
-/// Returns if the given BB is a single BB terminate pad which starts with a
-/// 'catch_all' insrtruction.
-bool isCatchAllTerminatePad(const MachineBasicBlock &MBB);
-
 // Exception-related function names
 extern const char *const ClangCallTerminateFn;
 extern const char *const CxaBeginCatchFn;