[SelectionDAG] set dereferenceable flag in MergeConsecutiveStores to fix assetion failure

[inouehrs]

Summary

When SelectionDAG merges consecutive stores and loads in MergeConsecutiveStores, it does not set dereferenceable flag for a created load instruction. This results in an assertion failure if SelectionDAG commonizes this load instruction with other load instructions, as well as it may miss optimization opportunities.

This patch sat dereferenceable flag for the newly created load instruction if all the load instruction to be merged are dereferenceable.

[hfinkel]

LGTM

llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

while (StoreNodes.size() > 1) {
 
     // Scan the memory operations on the chain and find the first
     // non-consecutive load memory address. These variables hold the index in
     // the store node array.
     unsigned LastConsecutiveLoad = 1;
     // This variable refers to the size and not index in the array.
     unsigned LastLegalVectorType = 1;
     unsigned LastLegalIntegerType = 1;
+    bool isDereferenceable = true;
     bool DoIntegerTruncate = false;
     StartAddress = LoadNodes[0].OffsetFromBase;
     SDValue FirstChain = FirstLoad->getChain();
     for (unsigned i = 1; i < LoadNodes.size(); ++i) {
       // All loads must share the same chain.
       if (LoadNodes[i].MemNode->getChain() != FirstChain)
         break;
 
       int64_t CurrAddress = LoadNodes[i].OffsetFromBase;
       if (CurrAddress - StartAddress != (ElementSizeBytes * i))
         break;
       LastConsecutiveLoad = i;
+
+      if (isDereferenceable && !LoadNodes[i].MemNode->isDereferenceable())
+        isDereferenceable = false;
+
       // Find a legal type for the vector store.
       EVT StoreTy = EVT::getVectorVT(Context, MemVT, i + 1);
       bool IsFastSt, IsFastLd;
       if (TLI.isTypeLegal(StoreTy) &&
           TLI.canMergeStoresTo(FirstStoreAS, StoreTy) &&
           TLI.allowsMemoryAccess(Context, DL, StoreTy, FirstStoreAS,
                                  FirstStoreAlign, &IsFastSt) &&
           IsFastSt &&
while (StoreNodes.size() > 1) {
     SDLoc StoreDL(StoreNodes[0].MemNode);
 
     // The merged loads are required to have the same incoming chain, so
     // using the first's chain is acceptable.
 
     SDValue NewStoreChain = getMergeStoreChains(StoreNodes, NumElem);
     AddToWorklist(NewStoreChain.getNode());
 
+    MachineMemOperand::Flags MMOFlags = isDereferenceable ? 
+                                          MachineMemOperand::MODereferenceable:
+                                          MachineMemOperand::MONone;
+
     SDValue NewLoad, NewStore;
     if (UseVectorTy || !DoIntegerTruncate) {
       NewLoad = DAG.getLoad(JointMemOpVT, LoadDL, FirstLoad->getChain(),
                             FirstLoad->getBasePtr(),
-                            FirstLoad->getPointerInfo(), FirstLoadAlign);
+                            FirstLoad->getPointerInfo(), FirstLoadAlign,
+                            MMOFlags);
       NewStore = DAG.getStore(NewStoreChain, StoreDL, NewLoad,
                               FirstInChain->getBasePtr(),
                               FirstInChain->getPointerInfo(), FirstStoreAlign);
     } else { // This must be the truncstore/extload case
       EVT ExtendedTy =
           TLI.getTypeToTransformTo(*DAG.getContext(), JointMemOpVT);
       NewLoad =
           DAG.getExtLoad(ISD::EXTLOAD, LoadDL, ExtendedTy, FirstLoad->getChain(),
                          FirstLoad->getBasePtr(), FirstLoad->getPointerInfo(),
-                         JointMemOpVT, FirstLoadAlign);
+                         JointMemOpVT, FirstLoadAlign, MMOFlags);
       NewStore = DAG.getTruncStore(NewStoreChain, StoreDL, NewLoad,
                                    FirstInChain->getBasePtr(),
                                    FirstInChain->getPointerInfo(), JointMemOpVT,
                                    FirstInChain->getAlignment(),
                                    FirstInChain->getMemOperand()->getFlags());
     }
 
     // Transfer chain users from old loads to the new load.

llvm/trunk/test/CodeGen/PowerPC/merge_stores_dereferenceable.ll

+; RUN: llc -verify-machineinstrs -mcpu=pwr8 -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck %s
+
+; This code causes an assertion failure if dereferenceable flag is not properly set when in merging consecutive stores
+; CHECK-LABEL: func:
+; CHECK: lxvd2x [[REG1:[0-9]+]], {{[0-9]+}}, {{[0-9]+}}
+; CHECK-NOT: lxvd2x
+; CHECK: stxvd2x [[REG1:[0-9]+]], {{[0-9]+}}, {{[0-9]+}}
+
+define <2 x i64> @func(i64* %pdst) {
+entry:
+  %a = alloca [4 x i64], align 8
+  %psrc0 = bitcast [4 x i64]* %a to i64*
+  %psrc1 = getelementptr inbounds i64, i64* %psrc0, i64 1
+  %d0 = load i64, i64* %psrc0
+  %d1 = load i64, i64* %psrc1
+  %pdst0 = getelementptr inbounds i64, i64* %pdst, i64 0
+  %pdst1 = getelementptr inbounds i64, i64* %pdst, i64 1
+  store i64 %d0, i64* %pdst0, align 8
+  store i64 %d1, i64* %pdst1, align 8
+  %psrcd = bitcast [4 x i64]* %a to <2 x i64>*
+  %vec = load <2 x i64>, <2 x i64>* %psrcd
+  ret <2 x i64> %vec
+}
+