[llvm] [IA][RISCV] Add support for vp.load/vp.store with shufflevector (PR #135445)

[Luke Lau via llvm-commits]

================
@@ -339,25 +371,48 @@ bool InterleavedAccessImpl::lowerInterleavedLoad(
     return false;
 
   bool BinOpShuffleChanged =
-      replaceBinOpShuffles(BinOpShuffles.getArrayRef(), Shuffles, LI);
-
-  LLVM_DEBUG(dbgs() << "IA: Found an interleaved load: " << *LI << "\n");
+      replaceBinOpShuffles(BinOpShuffles.getArrayRef(), Shuffles, LoadOp);
+
+  // Check if the de-interleaved vp.load masks are the same.
+  unsigned ShuffleMaskLen = Shuffles[0]->getShuffleMask().size();
+  SmallVector<Constant *, 8> LaneMask(ShuffleMaskLen, nullptr);
+  if (auto *VPLoad = dyn_cast<VPIntrinsic>(LoadOp)) {
+    if (!isInterleavedConstantMask(
+            Factor, cast<ConstantVector>(VPLoad->getArgOperand(1)), LaneMask))
+      return false;
+  }
 
-  // Try to create target specific intrinsics to replace the load and shuffles.
-  if (!TLI->lowerInterleavedLoad(LI, Shuffles, Indices, Factor)) {
-    // If Extracts is not empty, tryReplaceExtracts made changes earlier.
-    return !Extracts.empty() || BinOpShuffleChanged;
+  LLVM_DEBUG(dbgs() << "IA: Found an interleaved load: " << *LoadOp << "\n");
+
+  if (auto *VPLoad = dyn_cast<VPIntrinsic>(LoadOp)) {
+    auto *MaskVec = ConstantVector::get(LaneMask);
+    // Sometimes the number of Shuffles might be less than Factor, we have to
+    // fill the gaps with null. Also, lowerDeinterleavedVPLoad
+    // expects them to be sorted.
+    SmallVector<Value *, 4> ShuffleValues(Factor, nullptr);
+    for (auto [Idx, ShuffleMaskIdx] : enumerate(Indices))
+      ShuffleValues[ShuffleMaskIdx] = Shuffles[Idx];
----------------
lukel97 wrote:

Ah I see, thanks for explaining it. Would be simpler to add a Factor argument to `lowerDeinterleavedVPLoad` to bring it inline with `lowerInterleavedLoad`? Non-blocking, just a suggestion. 

https://github.com/llvm/llvm-project/pull/135445