[llvm] [InterleavedAccess] Construct interleaved access store with shuffles (PR #164000)

[via llvm-commits]

================
@@ -4801,19 +4801,46 @@ InstructionCost AArch64TTIImpl::getInterleavedMemoryOpCost(
   if (!VecTy->isScalableTy() && (UseMaskForCond || UseMaskForGaps))
     return InstructionCost::getInvalid();
 
-  if (!UseMaskForGaps && Factor <= TLI->getMaxSupportedInterleaveFactor()) {
+  unsigned NumLoadStores = 1;
+  InstructionCost ShuffleCost = 0;
+  bool isInterleaveWithShuffle = false;
+  unsigned MaxSupportedFactor = TLI->getMaxSupportedInterleaveFactor();
+
+  auto *SubVecTy =
+      VectorType::get(VecVTy->getElementType(),
+                      VecVTy->getElementCount().divideCoefficientBy(Factor));
+
+  if (TLI->isProfitableToInterleaveWithGatherScatter() &&
+      Opcode == Instruction::Store && (0 == Factor % MaxSupportedFactor) &&
+      Factor > MaxSupportedFactor) {
+    isInterleaveWithShuffle = true;
+    SmallVector<int, 16> Mask;
+    // preparing interleave Mask.
+    for (unsigned i = 0; i < VecVTy->getElementCount().getKnownMinValue() / 2;
+         i++)
+      for (unsigned j = 0; j < 2; j++)
+        Mask.push_back(j * Factor + i);
----------------
ram-NK wrote:

Fixed the issue.

https://github.com/llvm/llvm-project/pull/164000