[VPlan] Clean up dead recipes after UF & VF specific simplification.

Recursively remove dead recipes after simplifying vector loop exit
branch.

Author: fhahn

llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp

@@ -641,6 +641,25 @@ static void removeRedundantExpandSCEVRecipes(VPlan &Plan) {
   }
 }
 
+static void recursivelyDeleteDeadRecipes(VPValue *V) {
+  SmallVector<VPValue *> WorkList;
+  SmallPtrSet<VPValue *, 8> Seen;
+  WorkList.push_back(V);
+
+  while (!WorkList.empty()) {
+    VPValue *Cur = WorkList.pop_back_val();
+    if (!Seen.insert(Cur).second)
+      continue;
+    VPRecipeBase *R = Cur->getDefiningRecipe();
+    if (!R)
+      continue;
+    if (!isDeadRecipe(*R))
+      continue;
+    WorkList.append(R->op_begin(), R->op_end());
+    R->eraseFromParent();
+  }
+}
+
 void VPlanTransforms::optimizeForVFAndUF(VPlan &Plan, ElementCount BestVF,
                                          unsigned BestUF,
                                          PredicatedScalarEvolution &PSE) {
@@ -674,7 +693,11 @@ void VPlanTransforms::optimizeForVFAndUF(VPlan &Plan, ElementCount BestVF,
   auto *BOC =
       new VPInstruction(VPInstruction::BranchOnCond,
                         {Plan.getOrAddLiveIn(ConstantInt::getTrue(Ctx))});
+
+  SmallVector<VPValue *> PossiblyDead(Term->operands());
   Term->eraseFromParent();
+  for (VPValue *Op : PossiblyDead)
+    recursivelyDeleteDeadRecipes(Op);
   ExitingVPBB->appendRecipe(BOC);
   Plan.setVF(BestVF);
   Plan.setUF(BestUF);

llvm/test/Transforms/LoopVectorize/AArch64/clamped-trip-count.ll

@@ -43,9 +43,8 @@ define void @clamped_tc_8(ptr nocapture %dst, i32 %n, i64 %val) vscale_range(1,1
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP16]], ptr [[TMP17]], i32 1, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_NEXT]], i64 8)
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i64> [[VEC_IND]], [[DOTSPLAT]]
-; CHECK-NEXT:    [[TMP18:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_NEXT]], i64 8)
 ; CHECK-NEXT:    br i1 true, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
@@ -135,9 +134,8 @@ define void @clamped_tc_max_8(ptr nocapture %dst, i32 %n, i64 %val) vscale_range
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP16]], ptr [[TMP17]], i32 1, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_NEXT]], i64 [[WIDE_TRIP_COUNT]])
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i64> [[VEC_IND]], [[DOTSPLAT]]
-; CHECK-NEXT:    [[TMP18:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_NEXT]], i64 [[WIDE_TRIP_COUNT]])
 ; CHECK-NEXT:    br i1 true, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]