[Transform][LoadStoreVectorizer] Allow partial redundant

Author: cmc-rep

llvm/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp

@@ -157,7 +157,7 @@ using EqClassKey =
 struct ChainElem {
   Instruction *Inst;
   APInt OffsetFromLeader;
-  bool Redundant = false; // Set to true when load is redundant.
+  unsigned IncrementalBytes;
   ChainElem(Instruction *Inst, APInt OffsetFromLeader)
       : Inst(std::move(Inst)), OffsetFromLeader(std::move(OffsetFromLeader)) {}
 };
@@ -638,21 +638,23 @@ std::vector<Chain> Vectorizer::splitChainByContiguity(Chain &C) {
 
     // Add this instruction to the end of the current chain, or start a new one.
     APInt ReadEnd = It->OffsetFromLeader + SzBits / 8;
-    bool IsRedundant = ReadEnd.sle(PrevReadEnd);
-    bool AreContiguous = It->OffsetFromLeader == PrevReadEnd;
+    unsigned ExtraBytes =
+        PrevReadEnd.sle(ReadEnd) ? (ReadEnd - PrevReadEnd).getSExtValue() : 0u;
+    bool AreContiguous = It->OffsetFromLeader.sle(PrevReadEnd);
 
     LLVM_DEBUG(dbgs() << "LSV: Instruction is "
-                      << (AreContiguous
-                              ? "contiguous"
-                              : ((IsRedundant ? "redundant" : "chain-breaker")))
+                      << (AreContiguous ? "contiguous" : "chain-breaker")
                       << *It->Inst << " (starts at offset "
                       << It->OffsetFromLeader << ")\n");
 
-    It->Redundant = IsRedundant;
-    if (AreContiguous || IsRedundant)
+    if (AreContiguous) {
+      It->IncrementalBytes = ExtraBytes;
       CurChain.push_back(*It);
-    else
+    }
+    else {
+      It->IncrementalBytes = SzBits / 8;
       Ret.push_back({*It});
+    }
     PrevReadEnd = APIntOps::smax(PrevReadEnd, ReadEnd);
   }
 
@@ -883,11 +885,9 @@ bool Vectorizer::vectorizeChain(Chain &C) {
   bool IsLoadChain = isa<LoadInst>(C[0].Inst);
   unsigned AS = getLoadStoreAddressSpace(C[0].Inst);
   unsigned ChainBytes = 0;
-  for (auto &E : C) {
-    if (E.Redundant)
-      continue;
-    ChainBytes += DL.getTypeStoreSize(getLoadStoreType(E.Inst));
-  }
+  for (auto &E : C)
+    ChainBytes += E.IncrementalBytes;
+
   assert(ChainBytes % DL.getTypeStoreSize(VecElemTy) == 0);
   // VecTy is a power of 2 and 1 byte at smallest, but VecElemTy may be smaller
   // than 1 byte (e.g. VecTy == <32 x i1>).