[IA]: Construct (de)interleave4 out of (de)interleave2

llvm/include/llvm/CodeGen/TargetLowering.h

@@ -3155,8 +3155,11 @@ class TargetLoweringBase {
   ///
   /// \p DI is the deinterleave intrinsic.
   /// \p LI is the accompanying load instruction
-  virtual bool lowerDeinterleaveIntrinsicToLoad(IntrinsicInst *DI,
-                                                LoadInst *LI) const {
+  /// \p DeadInsts is a reference to a vector that keeps track of dead
+  /// instruction during transformations.
+  virtual bool lowerDeinterleaveIntrinsicToLoad(
+      IntrinsicInst *DI, LoadInst *LI,
+      SmallVectorImpl<Instruction *> &DeadInsts) const {
     return false;
   }
 
@@ -3166,8 +3169,11 @@ class TargetLoweringBase {
   ///
   /// \p II is the interleave intrinsic.
   /// \p SI is the accompanying store instruction
-  virtual bool lowerInterleaveIntrinsicToStore(IntrinsicInst *II,
-                                               StoreInst *SI) const {
+  /// \p DeadInsts is a reference to a vector that keeps track of dead
+  /// instruction during transformations.
+  virtual bool lowerInterleaveIntrinsicToStore(
+      IntrinsicInst *II, StoreInst *SI,
+      SmallVectorImpl<Instruction *> &DeadInsts) const {
     return false;
   }
 

llvm/include/llvm/IR/PatternMatch.h

@@ -2924,6 +2924,17 @@ inline VScaleVal_match m_VScale() {
   return VScaleVal_match();
 }
 
+template <typename Opnd0, typename Opnd1>
+inline typename m_Intrinsic_Ty<Opnd0, Opnd1>::Ty
+m_Interleave2(const Opnd0 &Op0, const Opnd1 &Op1) {
+  return m_Intrinsic<Intrinsic::vector_interleave2>(Op0, Op1);
+}
+
+template <typename Opnd>
+inline typename m_Intrinsic_Ty<Opnd>::Ty m_Deinterleave2(const Opnd &Op) {
+  return m_Intrinsic<Intrinsic::vector_deinterleave2>(Op);
+}
+
 template <typename LHS, typename RHS, unsigned Opcode, bool Commutable = false>
 struct LogicalOp_match {
   LHS L;

llvm/lib/CodeGen/InterleavedAccessPass.cpp

@@ -489,7 +489,7 @@ bool InterleavedAccessImpl::lowerDeinterleaveIntrinsic(
   LLVM_DEBUG(dbgs() << "IA: Found a deinterleave intrinsic: " << *DI << "\n");
 
   // Try and match this with target specific intrinsics.
-  if (!TLI->lowerDeinterleaveIntrinsicToLoad(DI, LI))
+  if (!TLI->lowerDeinterleaveIntrinsicToLoad(DI, LI, DeadInsts))
     return false;
 
   // We now have a target-specific load, so delete the old one.
@@ -510,13 +510,16 @@ bool InterleavedAccessImpl::lowerInterleaveIntrinsic(
 
   LLVM_DEBUG(dbgs() << "IA: Found an interleave intrinsic: " << *II << "\n");
 
+  SmallVector<Instruction *, 4> InterleaveDeadInsts;
   // Try and match this with target specific intrinsics.
-  if (!TLI->lowerInterleaveIntrinsicToStore(II, SI))
+  if (!TLI->lowerInterleaveIntrinsicToStore(II, SI, InterleaveDeadInsts))
     return false;
 
   // We now have a target-specific store, so delete the old one.
   DeadInsts.push_back(SI);
   DeadInsts.push_back(II);
+  DeadInsts.insert(DeadInsts.end(), InterleaveDeadInsts.begin(),
+                   InterleaveDeadInsts.end());
   return true;
 }
 
@@ -537,7 +540,7 @@ bool InterleavedAccessImpl::runOnFunction(Function &F) {
       // with a factor of 2.
       if (II->getIntrinsicID() == Intrinsic::vector_deinterleave2)
         Changed |= lowerDeinterleaveIntrinsic(II, DeadInsts);
-      if (II->getIntrinsicID() == Intrinsic::vector_interleave2)
+      else if (II->getIntrinsicID() == Intrinsic::vector_interleave2)
         Changed |= lowerInterleaveIntrinsic(II, DeadInsts);
     }
   }