[WebAssembly] Make llvm.wasm.throw invokable

`llvm.wasm.throw` intrinsic can throw but it was not invokable. Not sure
what the rationale was when it was first written that way, but I think
at least in Emscripten's C++ exception support with the Wasm port of
libunwind, `__builtin_wasm_throw`, which is lowered down to
`llvm.wasm.rethrow`, is used only within `_Unwind_RaiseException`, which
is a one-liner and thus does not need an `invoke`:
https://github.com/emscripten-core/emscripten/blob/720e97f76d6f19e0c6a2d6988988cfe23f0517fb/system/lib/libunwind/src/Unwind-wasm.c#L69
(`_Unwind_RaiseException` is called by `__cxa_throw`, which is generated
by the `throw` C++ keyword)

But this does not address other direct uses of the builtin in C++, whose
use I'm not sure about but is not prohibited. Also other language
frontends may need to use the builtin in different functions, which has
`try`-`catch`es or destructors.

This makes `llvm.wasm.throw` invokable in the backend. To do that, this
adds a custom lowering routine to `SelectionDAGBuilder::visitInvoke`,
like we did for `llvm.wasm.rethrow`.

This does not generate `invoke`s for `__builtin_wasm_throw` yet, which
will be done by a follow-up PR.

Addresses llvm#124710.

Author: aheejin

llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

@@ -3360,10 +3360,23 @@ void SelectionDAGBuilder::visitInvoke(const InvokeInst &I) {
     case Intrinsic::experimental_gc_statepoint:
       LowerStatepoint(cast<GCStatepointInst>(I), EHPadBB);
       break;
+    // wasm_throw, wasm_rethrow: This is usually done in visitTargetIntrinsic,
+    // but this intrinsic is special because it can be invoked, so we manually
+    // lower it to a DAG node here.
+    case Intrinsic::wasm_throw: {
+      SmallVector<SDValue, 8> Ops;
+      Ops.push_back(getControlRoot()); // inchain for the terminator node
+      const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+      Ops.push_back(
+          DAG.getTargetConstant(Intrinsic::wasm_throw, getCurSDLoc(),
+                                TLI.getPointerTy(DAG.getDataLayout())));
+      Ops.push_back(getValue(I.getArgOperand(0))); // tag
+      Ops.push_back(getValue(I.getArgOperand(1))); // thrown value
+      SDVTList VTs = DAG.getVTList(ArrayRef<EVT>({MVT::Other})); // outchain
+      DAG.setRoot(DAG.getNode(ISD::INTRINSIC_VOID, getCurSDLoc(), VTs, Ops));
+      break;
+    }
     case Intrinsic::wasm_rethrow: {
-      // This is usually done in visitTargetIntrinsic, but this intrinsic is
-      // special because it can be invoked, so we manually lower it to a DAG
-      // node here.
       SmallVector<SDValue, 8> Ops;
       Ops.push_back(getControlRoot()); // inchain for the terminator node
       const TargetLowering &TLI = DAG.getTargetLoweringInfo();

llvm/lib/CodeGen/WasmEHPrepare.cpp

@@ -201,10 +201,8 @@ bool WasmEHPrepareImpl::prepareThrows(Function &F) {
   // delete all following instructions within the BB, and delete all the dead
   // children of the BB as well.
   for (User *U : ThrowF->users()) {
-    // A call to @llvm.wasm.throw() is only generated from __cxa_throw()
-    // builtin call within libcxxabi, and cannot be an InvokeInst.
-    auto *ThrowI = cast<CallInst>(U);
-    if (ThrowI->getFunction() != &F)
+    auto *ThrowI = dyn_cast<CallInst>(U);
+    if (!ThrowI || ThrowI->getFunction() != &F)
       continue;
     Changed = true;
     auto *BB = ThrowI->getParent();

llvm/lib/IR/Verifier.cpp

@@ -5203,10 +5203,12 @@ void Verifier::visitInstruction(Instruction &I) {
                 F->getIntrinsicID() == Intrinsic::experimental_patchpoint ||
                 F->getIntrinsicID() == Intrinsic::fake_use ||
                 F->getIntrinsicID() == Intrinsic::experimental_gc_statepoint ||
+                F->getIntrinsicID() == Intrinsic::wasm_throw ||
                 F->getIntrinsicID() == Intrinsic::wasm_rethrow ||
                 IsAttachedCallOperand(F, CBI, i),
             "Cannot invoke an intrinsic other than donothing, patchpoint, "
-            "statepoint, coro_resume, coro_destroy or clang.arc.attachedcall",
+            "statepoint, coro_resume, coro_destroy, clang.arc.attachedcall or "
+            "wasm.(re)throw",
             &I);
       Check(F->getParent() == &M, "Referencing function in another module!", &I,
             &M, F, F->getParent());

llvm/test/CodeGen/WebAssembly/exception.ll

@@ -566,6 +566,32 @@ unreachable:                                      ; preds = %entry
   unreachable
 }
 
+; This tests whether llvm.wasm.throw intrinsic can invoked and iseled correctly.
+
+; CHECK-LABEL: invoke_throw:
+; CHECK: try_table    (catch __cpp_exception 0)
+; CHECK:   local.get  0
+; CHECK:   throw     __cpp_exception
+; CHECK: end_try_table
+define void @invoke_throw(ptr %p) personality ptr @__gxx_wasm_personality_v0 {
+entry:
+  invoke void @llvm.wasm.throw(i32 0, ptr %p)
+          to label %try.cont unwind label %catch.dispatch
+
+catch.dispatch:                                   ; preds = %entry
+  %0 = catchswitch within none [label %catch.start] unwind to caller
+
+catch.start:                                      ; preds = %catch.dispatch
+  %1 = catchpad within %0 [ptr null]
+  %2 = call ptr @llvm.wasm.get.exception(token %1)
+  %3 = call i32 @llvm.wasm.get.ehselector(token %1)
+  %4 = call ptr @__cxa_begin_catch(ptr %2) #4 [ "funclet"(token %1) ]
+  call void @__cxa_end_catch() [ "funclet"(token %1) ]
+  catchret from %1 to label %try.cont
+
+try.cont:                                         ; preds = %catch, %entry
+  ret void
+}
 
 declare void @foo()
 declare void @bar(ptr)