[llvm] [VPlan] Dissolve replicate regions with vector live-outs. (PR #189022)

[via llvm-commits]

================
@@ -808,6 +819,73 @@ static void dissolveReplicateRegion(VPRegionBlock *Region, ElementCount VF,
   // FirstLaneExiting.
   VPBlockUtils::connectBlocks(Predecessor, FirstLaneEntry);
   VPBlockUtils::connectBlocks(FirstLaneExiting, NextLaneEntry);
+
+  // Collect per-lane predicated scalar phis and create InsertElement chains or
+  // BuildStructVector to pack them into a vector result. Phis are in the same
+  // order across all cloned exit blocks, so we can match them by position.
+  SmallVector<SmallVector<VPValue *, 4>> PhisByLane;
+  for (VPBasicBlock *Exit : ExitingBlocks) {
+    auto &Phis = PhisByLane.emplace_back();
+    for (auto &Phi : Exit->phis())
+      Phis.push_back(Phi.getVPSingleValue());
+  }
+
+  VPTypeAnalysis TypeInfo(Plan);
+  SmallVector<VPValue *> LaneValues;
+  for (unsigned I = 0; I < PhisByLane[0].size(); ++I) {
+    LaneValues.clear();
+    for (auto &LanePhis : PhisByLane)
+      LaneValues.push_back(LanePhis[I]);
+
+    // If only the first lane is used, no need to pack into a vector.
+    // The lane-0 scalar phi can be used directly. Only lane 0 has external
+    // users; the other lane phis are clones without downstream users.
+    if (vputils::onlyFirstLaneUsed(LaneValues[0])) {
+      assert(all_of(drop_begin(LaneValues),
+                    [](VPValue *V) { return V->getNumUsers() == 0; }) &&
+             "cloned lane phis should have no users");
+      continue;
+    }
+
+    Type *ScalarTy = TypeInfo.inferScalarType(LaneValues[0]);
+
+    // Struct types need BuildStructVector as InsertElement doesn't apply.
+    if (isa<StructType>(ScalarTy)) {
----------------
ayalz wrote:

Can the BuildStructVector be built as a user of struct-typed PredInstPhi's of original lane 0, prior to cloning, and have each of its replicas add itself as an operand when visited during processLaneForReplicateRegion()?

https://github.com/llvm/llvm-project/pull/189022