[VPlan] Fix header masks in EVL tail folding

With EVL tail folding, the EVL may not always be VF on the second-to-last iteration.

Recipes that have been converted to VP intrinsics via optimizeMaskToEVL account for this, but recipes that are left behind will still use the old header mask which may end up having a different vector length.

This is effectively the same as llvm#95368, and fixes this by converting header masks from icmp ule wide-canonical-iv, backedge-trip-count -> icmp ult step-vector, evl. Without it, recipes that fall through optimizeMaskToEVL may use the wrong vector length, e.g. in llvm#150074 and llvm#149981.

We really need to split off optimizeMaskToEVL into VPlanTransforms::optimize and move transformRecipestoEVLRecipes into tryToBuildVPlanWithVPRecipes, so we don't mix up what is needed for correctness and what is needed to optimize away the mask computations. We should be able to still generate a correct albeit suboptimal VPlan without running optimizeMaskToEVL. I've added a TODO for this.

Fixes llvm#150197

Author: lukel97

llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp

@@ -2235,6 +2235,10 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
   }
 
   // Try to optimize header mask recipes away to their EVL variants.
+  //
+  // TODO: Split this out and move into VPlanTransforms::optimize.
+  // transformRecipestoEVLRecipes should be run in tryToBuildVPlanWithVPRecipes
+  // beforehand.
   for (VPValue *HeaderMask : collectAllHeaderMasks(Plan)) {
     for (VPUser *U : collectUsersRecursively(HeaderMask)) {
       auto *CurRecipe = cast<VPRecipeBase>(U);
@@ -2265,6 +2269,27 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
     for (VPValue *Op : PossiblyDead)
       recursivelyDeleteDeadRecipes(Op);
   }
+
+  // Replace header masks with a mask equivalent to predicating by EVL:
+  //
+  // icmp ule widen-canonical-iv backedge-taken-count
+  // ->
+  // icmp ult step-vector, EVL
+  Type *EVLType = TypeInfo.inferScalarType(&EVL);
+  for (VPValue *HeaderMask : collectAllHeaderMasks(Plan)) {
+    if (HeaderMask->users().empty())
+      continue;
+    VPRecipeBase *EVLR = EVL.getDefiningRecipe();
+    VPBuilder Builder(Plan.getVectorPreheader());
+    VPValue *StepVector =
+        Builder.createNaryOp(VPInstruction::StepVector, {}, EVLType);
+    Builder.setInsertPoint(EVLR->getParent(), std::next(EVLR->getIterator()));
+    VPValue *EVLMask = Builder.createICmp(
+        CmpInst::ICMP_ULT, StepVector,
+        Builder.createNaryOp(VPInstruction::Broadcast, {&EVL}));
+    HeaderMask->replaceAllUsesWith(EVLMask);
+    HeaderMask->getDefiningRecipe()->eraseFromParent();
+  }
 }
 
 /// Add a VPEVLBasedIVPHIRecipe and related recipes to \p Plan and

llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp

@@ -173,6 +173,8 @@ bool VPlanVerifier::verifyEVLRecipe(const VPInstruction &EVL) const {
         .Case<VPInstruction>([&](const VPInstruction *I) {
           if (I->getOpcode() == Instruction::PHI)
             return VerifyEVLUse(*I, 1);
+          if (I->getOpcode() == VPInstruction::Broadcast)
+            return VerifyEVLUse(*I, 0);
           switch (I->getOpcode()) {
           case Instruction::Add:
             break;

llvm/test/Transforms/LoopVectorize/RISCV/pr88802.ll

@@ -23,6 +23,7 @@ define void @test(ptr %p, i64 %a, i8 %b) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = ashr <vscale x 2 x i64> [[TMP5]], splat (i64 52)
 ; CHECK-NEXT:    [[TMP7:%.*]] = trunc <vscale x 2 x i64> [[TMP6]] to <vscale x 2 x i32>
 ; CHECK-NEXT:    [[TMP8:%.*]] = zext <vscale x 2 x i8> [[BROADCAST_SPLAT]] to <vscale x 2 x i32>
+; CHECK-NEXT:    [[TMP19:%.*]] = call <vscale x 2 x i32> @llvm.stepvector.nxv2i32()
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <vscale x 2 x ptr> poison, ptr [[P]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <vscale x 2 x ptr> [[BROADCAST_SPLATINSERT3]], <vscale x 2 x ptr> poison, <vscale x 2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <vscale x 2 x i32> @llvm.stepvector.nxv2i32()
@@ -38,7 +39,9 @@ define void @test(ptr %p, i64 %a, i8 %b) {
 ; CHECK-NEXT:    [[TMP12:%.*]] = mul i32 1, [[TMP11]]
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <vscale x 2 x i32> poison, i32 [[TMP12]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <vscale x 2 x i32> [[BROADCAST_SPLATINSERT5]], <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP13:%.*]] = icmp ule <vscale x 2 x i32> [[VEC_IND]], splat (i32 8)
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT7:%.*]] = insertelement <vscale x 2 x i32> poison, i32 [[TMP11]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT8:%.*]] = shufflevector <vscale x 2 x i32> [[BROADCAST_SPLATINSERT7]], <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp ult <vscale x 2 x i32> [[TMP19]], [[BROADCAST_SPLAT8]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp sge <vscale x 2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP15:%.*]] = select <vscale x 2 x i1> [[TMP13]], <vscale x 2 x i1> [[TMP14]], <vscale x 2 x i1> zeroinitializer
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP15]], <vscale x 2 x i32> [[TMP7]], <vscale x 2 x i32> [[TMP8]]

llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll

@@ -383,6 +383,7 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
 ; TF-SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
 ; TF-SCALABLE-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
 ; TF-SCALABLE-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; TF-SCALABLE-NEXT:    [[TMP16:%.*]] = call <vscale x 4 x i32> @llvm.stepvector.nxv4i32()
 ; TF-SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x ptr> poison, ptr [[B]], i64 0
 ; TF-SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x ptr> [[BROADCAST_SPLATINSERT]], <vscale x 4 x ptr> poison, <vscale x 4 x i32> zeroinitializer
 ; TF-SCALABLE-NEXT:    [[TMP5:%.*]] = call <vscale x 4 x i64> @llvm.stepvector.nxv4i64()
@@ -399,7 +400,9 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
 ; TF-SCALABLE-NEXT:    [[TMP8:%.*]] = mul i64 1, [[TMP11]]
 ; TF-SCALABLE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TMP8]], i64 0
 ; TF-SCALABLE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 4 x i64> [[DOTSPLATINSERT]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
-; TF-SCALABLE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IND]], splat (i64 1024)
+; TF-SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP7]], i64 0
+; TF-SCALABLE-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT3]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; TF-SCALABLE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = icmp ult <vscale x 4 x i32> [[TMP16]], [[BROADCAST_SPLAT4]]
 ; TF-SCALABLE-NEXT:    [[TMP10:%.*]] = icmp ugt <vscale x 4 x i64> [[VEC_IND]], splat (i64 10)
 ; TF-SCALABLE-NEXT:    [[TMP9:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP10]], <vscale x 4 x i1> zeroinitializer
 ; TF-SCALABLE-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i64> @llvm.vp.gather.nxv4i64.nxv4p0(<vscale x 4 x ptr> align 8 [[BROADCAST_SPLAT]], <vscale x 4 x i1> [[TMP10]], i32 [[TMP7]])

llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cond-reduction.ll

@@ -262,25 +262,20 @@ define i32 @cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-OUTLOOP-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP6]]
 ; IF-EVL-OUTLOOP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
 ; IF-EVL-OUTLOOP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; IF-EVL-OUTLOOP-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
-; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+; IF-EVL-OUTLOOP-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.stepvector.nxv4i32()
 ; IF-EVL-OUTLOOP-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; IF-EVL-OUTLOOP:       vector.body:
 ; IF-EVL-OUTLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-OUTLOOP-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP9]], [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP10:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 4, i1 true)
-; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[EVL_BASED_IV]], i64 0
-; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
-; IF-EVL-OUTLOOP-NEXT:    [[TMP13:%.*]] = call <vscale x 4 x i64> @llvm.stepvector.nxv4i64()
-; IF-EVL-OUTLOOP-NEXT:    [[TMP14:%.*]] = add <vscale x 4 x i64> zeroinitializer, [[TMP13]]
-; IF-EVL-OUTLOOP-NEXT:    [[VEC_IV:%.*]] = add <vscale x 4 x i64> [[BROADCAST_SPLAT]], [[TMP14]]
-; IF-EVL-OUTLOOP-NEXT:    [[TMP15:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]]
+; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP11]], i64 0
+; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; IF-EVL-OUTLOOP-NEXT:    [[TMP15:%.*]] = icmp ult <vscale x 4 x i32> [[TMP12]], [[BROADCAST_SPLAT]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
 ; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
@@ -790,35 +785,27 @@ define i32 @step_cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-OUTLOOP-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP6]]
 ; IF-EVL-OUTLOOP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
 ; IF-EVL-OUTLOOP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; IF-EVL-OUTLOOP-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x i32> zeroinitializer, i32 [[START]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
-; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
-; IF-EVL-OUTLOOP-NEXT:    [[TMP10:%.*]] = call <vscale x 4 x i64> @llvm.stepvector.nxv4i64()
-; IF-EVL-OUTLOOP-NEXT:    [[TMP11:%.*]] = mul <vscale x 4 x i64> [[TMP10]], splat (i64 1)
-; IF-EVL-OUTLOOP-NEXT:    [[INDUCTION:%.*]] = add <vscale x 4 x i64> zeroinitializer, [[TMP11]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.stepvector.nxv4i32()
-; IF-EVL-OUTLOOP-NEXT:    [[TMP13:%.*]] = mul <vscale x 4 x i32> [[TMP12]], splat (i32 1)
+; IF-EVL-OUTLOOP-NEXT:    [[TMP10:%.*]] = call <vscale x 4 x i32> @llvm.stepvector.nxv4i32()
+; IF-EVL-OUTLOOP-NEXT:    [[TMP13:%.*]] = mul <vscale x 4 x i32> [[TMP10]], splat (i32 1)
 ; IF-EVL-OUTLOOP-NEXT:    [[INDUCTION1:%.*]] = add <vscale x 4 x i32> zeroinitializer, [[TMP13]]
 ; IF-EVL-OUTLOOP-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; IF-EVL-OUTLOOP:       vector.body:
 ; IF-EVL-OUTLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-OUTLOOP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
-; IF-EVL-OUTLOOP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 4 x i64> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-OUTLOOP-NEXT:    [[VEC_IND2:%.*]] = phi <vscale x 4 x i32> [ [[INDUCTION1]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-OUTLOOP-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP15:%.*]] = mul i32 1, [[TMP14]]
 ; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP15]], i64 0
 ; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT3]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
-; IF-EVL-OUTLOOP-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP14]] to i64
-; IF-EVL-OUTLOOP-NEXT:    [[TMP17:%.*]] = mul i64 1, [[TMP16]]
-; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TMP17]], i64 0
-; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT5]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
-; IF-EVL-OUTLOOP-NEXT:    [[TMP18:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP14]], i64 0
+; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; IF-EVL-OUTLOOP-NEXT:    [[TMP18:%.*]] = icmp ult <vscale x 4 x i32> [[TMP12]], [[BROADCAST_SPLAT2]]
 ; IF-EVL-OUTLOOP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 0
 ; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
@@ -830,7 +817,6 @@ define i32 @step_cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP14]] to i64
 ; IF-EVL-OUTLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP25]], [[IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
-; IF-EVL-OUTLOOP-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT6]]
 ; IF-EVL-OUTLOOP-NEXT:    [[VEC_IND_NEXT7]] = add <vscale x 4 x i32> [[VEC_IND2]], [[BROADCAST_SPLAT4]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; IF-EVL-OUTLOOP-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK1:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]