Add subclass data to the FoldingSetNode for MemIntrinsicSDNodes.

Not having the subclass data on an MemIntrinsicSDNodes means it was possible
to try to fold 2 nodes with the same operands but differing MMO flags. This
would trip an assertion when trying to refine the alignment between the 2
MachineMemOperands.

Differential Revision: https://reviews.llvm.org/D38898

llvm-svn: 316737

llvm/include/llvm/CodeGen/SelectionDAG.h

@@ -336,6 +336,14 @@ class SelectionDAG {
         .getRawSubclassData();
   }
 
+  template <typename SDNodeTy>
+  static uint16_t getSyntheticNodeSubclassData(unsigned Opc, unsigned Order,
+                                                SDVTList VTs, EVT MemoryVT,
+                                                MachineMemOperand *MMO) {
+    return SDNodeTy(Opc, Order, DebugLoc(), VTs, MemoryVT, MMO)
+         .getRawSubclassData();
+  }
+
   void createOperands(SDNode *Node, ArrayRef<SDValue> Vals) {
     assert(!Node->OperandList && "Node already has operands");
     SDUse *Ops = OperandRecycler.allocate(

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

@@ -5746,6 +5746,8 @@ SDValue SelectionDAG::getMemIntrinsicNode(unsigned Opcode, const SDLoc &dl,
   if (VTList.VTs[VTList.NumVTs-1] != MVT::Glue) {
     FoldingSetNodeID ID;
     AddNodeIDNode(ID, Opcode, VTList, Ops);
+    ID.AddInteger(getSyntheticNodeSubclassData<MemIntrinsicSDNode>(
+        Opcode, dl.getIROrder(), VTList, MemVT, MMO));
     ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
     void *IP = nullptr;
     if (SDNode *E = FindNodeOrInsertPos(ID, dl, IP)) {

llvm/test/CodeGen/PowerPC/MMO-flags-assertion.ll

@@ -0,0 +1,37 @@
+; RUN: llc < %s -mtriple powerpc64le-unknown-linux-gnu
+
+; void llvm::MachineMemOperand::refineAlignment(const llvm::MachineMemOperand*):
+; Assertion `MMO->getFlags() == getFlags() && "Flags mismatch !"' failed.
+
+declare void @_Z3fn11F(%class.F* byval align 8) local_unnamed_addr
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i32, i1)
+declare signext i32 @_ZN1F11isGlobalRegEv(%class.F*) local_unnamed_addr
+declare void @llvm.lifetime.start.p0i8(i64, i8* nocapture)
+declare void @_Z10EmitLValuev(%class.F* sret) local_unnamed_addr
+declare void @llvm.lifetime.end.p0i8(i64, i8* nocapture)
+
+%class.F = type { i32, i64, i8, [64 x i8], i8, i32* }
+
+define signext i32 @_Z29EmitOMPAtomicSimpleUpdateExpr1F(%class.F* byval align 8 %p1) local_unnamed_addr {
+entry:
+  call void @_Z3fn11F(%class.F* byval nonnull align 8 %p1)
+  %call = call signext i32 @_ZN1F11isGlobalRegEv(%class.F* nonnull %p1)
+  ret i32 %call
+}
+
+define void @_Z3fn2v() local_unnamed_addr {
+entry:
+  %agg.tmp1 = alloca %class.F, align 8
+  %XLValue = alloca %class.F, align 8
+  %0 = bitcast %class.F* %XLValue to i8*
+  call void @llvm.lifetime.start.p0i8(i64 96, i8* nonnull %0)
+  call void @_Z10EmitLValuev(%class.F* nonnull sret %XLValue)
+  %1 = bitcast %class.F* %agg.tmp1 to i8*
+  call void @llvm.lifetime.start.p0i8(i64 96, i8* nonnull %1)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* nonnull %1, i8* nonnull %0, i64 96, i32 8, i1 false)
+  call void @_Z3fn11F(%class.F* byval nonnull align 8 %XLValue)
+  %call.i = call signext i32 @_ZN1F11isGlobalRegEv(%class.F* nonnull %agg.tmp1)
+  call void @llvm.lifetime.end.p0i8(i64 96, i8* nonnull %1)
+  call void @llvm.lifetime.end.p0i8(i64 96, i8* nonnull %0)
+  ret void
+}