diff --git a/llvm/include/llvm/Transforms/Utils/MemoryTaggingSupport.h b/llvm/include/llvm/Transforms/Utils/MemoryTaggingSupport.h
--- a/llvm/include/llvm/Transforms/Utils/MemoryTaggingSupport.h
+++ b/llvm/include/llvm/Transforms/Utils/MemoryTaggingSupport.h
@@ -102,7 +102,7 @@
 };
 
 uint64_t getAllocaSizeInBytes(const AllocaInst &AI);
-void alignAndPadAlloca(memtag::AllocaInfo &Info, llvm::Align Align);
+bool alignAndPadAlloca(memtag::AllocaInfo &Info, llvm::Align Align);
 
 } // namespace memtag
 } // namespace llvm
diff --git a/llvm/lib/Target/AArch64/AArch64StackTagging.cpp b/llvm/lib/Target/AArch64/AArch64StackTagging.cpp
--- a/llvm/lib/Target/AArch64/AArch64StackTagging.cpp
+++ b/llvm/lib/Target/AArch64/AArch64StackTagging.cpp
@@ -505,7 +505,9 @@
   for (auto &I : SInfo.AllocasToInstrument) {
     memtag::AllocaInfo &Info = I.second;
     assert(Info.AI && isInterestingAlloca(*Info.AI));
-    memtag::alignAndPadAlloca(Info, kTagGranuleSize);
+    auto *PrevAI = Info.AI;
+    if (memtag::alignAndPadAlloca(Info, kTagGranuleSize))
+      PrevAI->eraseFromParent();
   }
 
   std::unique_ptr<DominatorTree> DeleteDT;
diff --git a/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp b/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
--- a/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
@@ -261,8 +261,6 @@
 
   void setSSI(const StackSafetyGlobalInfo *S) { SSI = S; }
 
-  DenseMap<AllocaInst *, AllocaInst *> padInterestingAllocas(
-      const MapVector<AllocaInst *, memtag::AllocaInfo> &AllocasToInstrument);
   bool sanitizeFunction(Function &F,
                         llvm::function_ref<const DominatorTree &()> GetDT,
                         llvm::function_ref<const PostDominatorTree &()> GetPDT);
@@ -1380,6 +1378,19 @@
           II->eraseFromParent();
       }
     }
+    if (memtag::alignAndPadAlloca(Info, Align(Mapping.getObjectAlignment()))) {
+      for (auto DVI : Info.DbgVariableIntrinsics) {
+        SmallDenseSet<Value *> LocationOps(DVI->location_ops().begin(),
+                                           DVI->location_ops().end());
+        for (Value *V : LocationOps) {
+          if (auto *AI = dyn_cast_or_null<AllocaInst>(V)) {
+            if (V == AI)
+              DVI->replaceVariableLocationOp(V, Info.AI);
+          }
+        }
+      }
+      AI->eraseFromParent();
+    }
   }
   for (auto &I : SInfo.UnrecognizedLifetimes)
     I->eraseFromParent();
@@ -1404,39 +1415,6 @@
          !(SSI && SSI->isSafe(AI));
 }
 
-DenseMap<AllocaInst *, AllocaInst *> HWAddressSanitizer::padInterestingAllocas(
-    const MapVector<AllocaInst *, memtag::AllocaInfo> &AllocasToInstrument) {
-  DenseMap<AllocaInst *, AllocaInst *> AllocaToPaddedAllocaMap;
-  for (auto &KV : AllocasToInstrument) {
-    AllocaInst *AI = KV.first;
-    uint64_t Size = memtag::getAllocaSizeInBytes(*AI);
-    uint64_t AlignedSize = alignTo(Size, Mapping.getObjectAlignment());
-    AI->setAlignment(
-        Align(std::max(AI->getAlignment(), Mapping.getObjectAlignment())));
-    if (Size != AlignedSize) {
-      Type *AllocatedType = AI->getAllocatedType();
-      if (AI->isArrayAllocation()) {
-        uint64_t ArraySize =
-            cast<ConstantInt>(AI->getArraySize())->getZExtValue();
-        AllocatedType = ArrayType::get(AllocatedType, ArraySize);
-      }
-      Type *TypeWithPadding = StructType::get(
-          AllocatedType, ArrayType::get(Int8Ty, AlignedSize - Size));
-      auto *NewAI = new AllocaInst(
-          TypeWithPadding, AI->getType()->getAddressSpace(), nullptr, "", AI);
-      NewAI->takeName(AI);
-      NewAI->setAlignment(AI->getAlign());
-      NewAI->setUsedWithInAlloca(AI->isUsedWithInAlloca());
-      NewAI->setSwiftError(AI->isSwiftError());
-      NewAI->copyMetadata(*AI);
-      auto *Bitcast = new BitCastInst(NewAI, AI->getType(), "", AI);
-      AI->replaceAllUsesWith(Bitcast);
-      AllocaToPaddedAllocaMap[AI] = NewAI;
-    }
-  }
-  return AllocaToPaddedAllocaMap;
-}
-
 bool HWAddressSanitizer::sanitizeFunction(
     Function &F, llvm::function_ref<const DominatorTree &()> GetDT,
     llvm::function_ref<const PostDominatorTree &()> GetPDT) {
@@ -1509,28 +1487,6 @@
     instrumentStack(DetectUseAfterScope && !SInfo.CallsReturnTwice, SIB.get(),
                     StackTag, GetDT, GetPDT);
   }
-  // Pad and align each of the allocas that we instrumented to stop small
-  // uninteresting allocas from hiding in instrumented alloca's padding and so
-  // that we have enough space to store real tags for short granules.
-  DenseMap<AllocaInst *, AllocaInst *> AllocaToPaddedAllocaMap =
-      padInterestingAllocas(SInfo.AllocasToInstrument);
-
-  if (!AllocaToPaddedAllocaMap.empty()) {
-    for (auto &Inst : instructions(F)) {
-      if (auto *DVI = dyn_cast<DbgVariableIntrinsic>(&Inst)) {
-        SmallDenseSet<Value *> LocationOps(DVI->location_ops().begin(),
-                                           DVI->location_ops().end());
-        for (Value *V : LocationOps) {
-          if (auto *AI = dyn_cast_or_null<AllocaInst>(V)) {
-            if (auto *NewAI = AllocaToPaddedAllocaMap.lookup(AI))
-              DVI->replaceVariableLocationOp(V, NewAI);
-          }
-        }
-      }
-    }
-    for (auto &P : AllocaToPaddedAllocaMap)
-      P.first->eraseFromParent();
-  }
 
   // If we split the entry block, move any allocas that were originally in the
   // entry block back into the entry block so that they aren't treated as
diff --git a/llvm/lib/Transforms/Utils/MemoryTaggingSupport.cpp b/llvm/lib/Transforms/Utils/MemoryTaggingSupport.cpp
--- a/llvm/lib/Transforms/Utils/MemoryTaggingSupport.cpp
+++ b/llvm/lib/Transforms/Utils/MemoryTaggingSupport.cpp
@@ -109,7 +109,7 @@
   return AI.getAllocationSizeInBits(DL).getValue() / 8;
 }
 
-void alignAndPadAlloca(memtag::AllocaInfo &Info, llvm::Align Alignment) {
+bool alignAndPadAlloca(memtag::AllocaInfo &Info, llvm::Align Alignment) {
   const Align NewAlignment = max(MaybeAlign(Info.AI->getAlign()), Alignment);
   Info.AI->setAlignment(NewAlignment);
   auto &Ctx = Info.AI->getFunction()->getContext();
@@ -117,7 +117,7 @@
   uint64_t Size = getAllocaSizeInBytes(*Info.AI);
   uint64_t AlignedSize = alignTo(Size, Alignment);
   if (Size == AlignedSize)
-    return;
+    return false;
 
   // Add padding to the alloca.
   Type *AllocatedType =
@@ -139,8 +139,8 @@
 
   auto *NewPtr = new BitCastInst(NewAI, Info.AI->getType(), "", Info.AI);
   Info.AI->replaceAllUsesWith(NewPtr);
-  Info.AI->eraseFromParent();
   Info.AI = NewAI;
+  return true;
 }
 
 } // namespace memtag