[llvm] bbfaf0b - Autogen more vectorizer tests in advance of D115387.

[Philip Reames via llvm-commits]

Author: Philip Reames
Date: 2021-12-13T11:04:20-08:00
New Revision: bbfaf0b170b6070e08f1dc22419dfedc75b9a0fe

URL: https://github.com/llvm/llvm-project/commit/bbfaf0b170b6070e08f1dc22419dfedc75b9a0fe
DIFF: https://github.com/llvm/llvm-project/commit/bbfaf0b170b6070e08f1dc22419dfedc75b9a0fe.diff

LOG: Autogen more vectorizer tests in advance of D115387.

These are the ones my first round of scripting couldn't handle that required a bit of manual messaging.  This should be the last batch in llvm-check.

Added: 
    

Modified: 
    llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-uniforms.ll
    llvm/test/Transforms/LoopVectorize/bsd_regex.ll
    llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
    llvm/test/Transforms/LoopVectorize/float-induction.ll
    llvm/test/Transforms/LoopVectorize/induction.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-uniforms.ll b/llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-uniforms.ll
index 1a030ab6d72e..85dc043d33c7 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-uniforms.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/consecutive-ptr-uniforms.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; REQUIRES: asserts
 ; RUN: opt < %s -aa-pipeline=basic-aa -passes=loop-vectorize,instcombine -S -debug-only=loop-vectorize -disable-output -print-after=instcombine 2>&1 | FileCheck %s
 ; RUN: opt < %s -loop-vectorize -instcombine -S -debug-only=loop-vectorize -disable-output -print-after=instcombine -enable-new-pm=0 2>&1 | FileCheck %s
@@ -6,33 +7,35 @@
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-linux-gnu"
 
-; CHECK-LABEL: PR31671
 ;
 ; Check a pointer in which one of its uses is consecutive-like and another of
-; its uses is non-consecutive-like. In the test case below, %tmp3 is the
+; its uses is non-consecutive-like. In the test case below, %var3 is the
 ; pointer operand of an interleaved load, making it consecutive-like. However,
 ; it is also the pointer operand of a non-interleaved store that will become a
-; scatter operation. %tmp3 (and the induction variable) should not be marked
+; scatter operation. %var3 (and the induction variable) should not be marked
 ; uniform-after-vectorization.
 ;
-; CHECK:       LV: Found uniform instruction: %tmp0 = getelementptr inbounds %data, %data* %d, i64 0, i32 3, i64 %i
-; CHECK-NOT:   LV: Found uniform instruction: %tmp3 = getelementptr inbounds %data, %data* %d, i64 0, i32 0, i64 %i
-; CHECK-NOT:   LV: Found uniform instruction: %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]
-; CHECK-NOT:   LV: Found uniform instruction: %i.next = add nuw nsw i64 %i, 5
+
+%data = type { [32000 x float], [3 x i32], [4 x i8], [32000 x float] }
+
+define void @PR31671(float %x, %data* %d) #0 {
+; CHECK-LABEL: @PR31671(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <16 x float> poison, float %x, i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <16 x float> poison, float [[X:%.*]], i32 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <16 x float> [[BROADCAST_SPLATINSERT]], <16 x float> poison, <16 x i32> zeroinitializer
-; CHECK-NEXT:    br label %vector.body
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
-; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <16 x i64> [ <i64 0, i64 5, i64 10, i64 15, i64 20, i64 25, i64 30, i64 35, i64 40, i64 45, i64 50, i64 55, i64 60, i64 65, i64 70, i64 75>, %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <16 x i64> [ <i64 0, i64 5, i64 10, i64 15, i64 20, i64 25, i64 30, i64 35, i64 40, i64 45, i64 50, i64 55, i64 60, i64 65, i64 70, i64 75>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 5
-; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds %data, %data* %d, i64 0, i32 3, i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[DATA:%.*]], %data* [[D:%.*]], i64 0, i32 3, i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[TMP0]] to <80 x float>*
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <80 x float>, <80 x float>* [[TMP1]], align 4
 ; CHECK-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <80 x float> [[WIDE_VEC]], <80 x float> poison, <16 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75>
 ; CHECK-NEXT:    [[TMP2:%.*]] = fmul <16 x float> [[BROADCAST_SPLAT]], [[STRIDED_VEC]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds %data, %data* %d, i64 0, i32 0, <16 x i64> [[VEC_IND]]
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, <16 x i64> [[VEC_IND]]
 ; CHECK-NEXT:    [[BC:%.*]] = bitcast <16 x float*> [[TMP3]] to <16 x <80 x float>*>
 ; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <16 x <80 x float>*> [[BC]], i32 0
 ; CHECK-NEXT:    [[WIDE_VEC1:%.*]] = load <80 x float>, <80 x float>* [[TMP4]], align 4
@@ -41,23 +44,155 @@ target triple = "x86_64-unknown-linux-gnu"
 ; CHECK-NEXT:    call void @llvm.masked.scatter.v16f32.v16p0f32(<16 x float> [[TMP5]], <16 x float*> [[TMP3]], i32 4, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <16 x i64> [[VEC_IND]], <i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80, i64 80>
-; CHECK:         br i1 {{.*}}, label %middle.block, label %vector.body
-
-%data = type { [32000 x float], [3 x i32], [4 x i8], [32000 x float] }
-
-define void @PR31671(float %x, %data* %d) #0 {
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 6384
+; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 31920, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[VAR0:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 3, i64 [[I]]
+; CHECK-NEXT:    [[VAR1:%.*]] = load float, float* [[VAR0]], align 4
+; CHECK-NEXT:    [[VAR2:%.*]] = fmul float [[VAR1]], [[X]]
+; CHECK-NEXT:    [[VAR3:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[I]]
+; CHECK-NEXT:    [[VAR4:%.*]] = load float, float* [[VAR3]], align 4
+; CHECK-NEXT:    [[VAR5:%.*]] = fadd float [[VAR4]], [[VAR2]]
+; CHECK-NEXT:    store float [[VAR5]], float* [[VAR3]], align 4
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 5
+; CHECK-NEXT:    [[COND:%.*]] = icmp ult i64 [[I]], 31995
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP2:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; FORCE-LABEL: @PR31671(
+; FORCE-NEXT:  entry:
+; FORCE-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; FORCE:       vector.ph:
+; FORCE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[X:%.*]], i32 0
+; FORCE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; FORCE-NEXT:    [[BROADCAST_SPLATINSERT7:%.*]] = insertelement <2 x float> poison, float [[X]], i32 0
+; FORCE-NEXT:    [[BROADCAST_SPLAT8:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT7]], <2 x float> poison, <2 x i32> zeroinitializer
+; FORCE-NEXT:    [[BROADCAST_SPLATINSERT9:%.*]] = insertelement <2 x float> poison, float [[X]], i32 0
+; FORCE-NEXT:    [[BROADCAST_SPLAT10:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT9]], <2 x float> poison, <2 x i32> zeroinitializer
+; FORCE-NEXT:    [[BROADCAST_SPLATINSERT11:%.*]] = insertelement <2 x float> poison, float [[X]], i32 0
+; FORCE-NEXT:    [[BROADCAST_SPLAT12:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT11]], <2 x float> poison, <2 x i32> zeroinitializer
+; FORCE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; FORCE:       vector.body:
+; FORCE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; FORCE-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 5
+; FORCE-NEXT:    [[TMP0:%.*]] = add i64 [[OFFSET_IDX]], 0
+; FORCE-NEXT:    [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 5
+; FORCE-NEXT:    [[TMP2:%.*]] = add i64 [[OFFSET_IDX]], 10
+; FORCE-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 15
+; FORCE-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 20
+; FORCE-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 25
+; FORCE-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 30
+; FORCE-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], 35
+; FORCE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[DATA:%.*]], %data* [[D:%.*]], i64 0, i32 3, i64 [[TMP0]]
+; FORCE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 3, i64 [[TMP2]]
+; FORCE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 3, i64 [[TMP4]]
+; FORCE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 3, i64 [[TMP6]]
+; FORCE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 0
+; FORCE-NEXT:    [[TMP13:%.*]] = bitcast float* [[TMP12]] to <10 x float>*
+; FORCE-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[TMP9]], i32 0
+; FORCE-NEXT:    [[TMP15:%.*]] = bitcast float* [[TMP14]] to <10 x float>*
+; FORCE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP10]], i32 0
+; FORCE-NEXT:    [[TMP17:%.*]] = bitcast float* [[TMP16]] to <10 x float>*
+; FORCE-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[TMP11]], i32 0
+; FORCE-NEXT:    [[TMP19:%.*]] = bitcast float* [[TMP18]] to <10 x float>*
+; FORCE-NEXT:    [[WIDE_VEC:%.*]] = load <10 x float>, <10 x float>* [[TMP13]], align 4
+; FORCE-NEXT:    [[WIDE_VEC1:%.*]] = load <10 x float>, <10 x float>* [[TMP15]], align 4
+; FORCE-NEXT:    [[WIDE_VEC2:%.*]] = load <10 x float>, <10 x float>* [[TMP17]], align 4
+; FORCE-NEXT:    [[WIDE_VEC3:%.*]] = load <10 x float>, <10 x float>* [[TMP19]], align 4
+; FORCE-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <10 x float> [[WIDE_VEC]], <10 x float> poison, <2 x i32> <i32 0, i32 5>
+; FORCE-NEXT:    [[STRIDED_VEC4:%.*]] = shufflevector <10 x float> [[WIDE_VEC1]], <10 x float> poison, <2 x i32> <i32 0, i32 5>
+; FORCE-NEXT:    [[STRIDED_VEC5:%.*]] = shufflevector <10 x float> [[WIDE_VEC2]], <10 x float> poison, <2 x i32> <i32 0, i32 5>
+; FORCE-NEXT:    [[STRIDED_VEC6:%.*]] = shufflevector <10 x float> [[WIDE_VEC3]], <10 x float> poison, <2 x i32> <i32 0, i32 5>
+; FORCE-NEXT:    [[TMP20:%.*]] = fmul <2 x float> [[BROADCAST_SPLAT]], [[STRIDED_VEC]]
+; FORCE-NEXT:    [[TMP21:%.*]] = fmul <2 x float> [[BROADCAST_SPLAT8]], [[STRIDED_VEC4]]
+; FORCE-NEXT:    [[TMP22:%.*]] = fmul <2 x float> [[BROADCAST_SPLAT10]], [[STRIDED_VEC5]]
+; FORCE-NEXT:    [[TMP23:%.*]] = fmul <2 x float> [[BROADCAST_SPLAT12]], [[STRIDED_VEC6]]
+; FORCE-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[TMP0]]
+; FORCE-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[TMP1]]
+; FORCE-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[TMP2]]
+; FORCE-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[TMP3]]
+; FORCE-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[TMP4]]
+; FORCE-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[TMP5]]
+; FORCE-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[TMP6]]
+; FORCE-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[TMP7]]
+; FORCE-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP24]], i32 0
+; FORCE-NEXT:    [[TMP33:%.*]] = bitcast float* [[TMP32]] to <10 x float>*
+; FORCE-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, float* [[TMP26]], i32 0
+; FORCE-NEXT:    [[TMP35:%.*]] = bitcast float* [[TMP34]] to <10 x float>*
+; FORCE-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, float* [[TMP28]], i32 0
+; FORCE-NEXT:    [[TMP37:%.*]] = bitcast float* [[TMP36]] to <10 x float>*
+; FORCE-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, float* [[TMP30]], i32 0
+; FORCE-NEXT:    [[TMP39:%.*]] = bitcast float* [[TMP38]] to <10 x float>*
+; FORCE-NEXT:    [[WIDE_VEC13:%.*]] = load <10 x float>, <10 x float>* [[TMP33]], align 4
+; FORCE-NEXT:    [[WIDE_VEC14:%.*]] = load <10 x float>, <10 x float>* [[TMP35]], align 4
+; FORCE-NEXT:    [[WIDE_VEC15:%.*]] = load <10 x float>, <10 x float>* [[TMP37]], align 4
+; FORCE-NEXT:    [[WIDE_VEC16:%.*]] = load <10 x float>, <10 x float>* [[TMP39]], align 4
+; FORCE-NEXT:    [[STRIDED_VEC17:%.*]] = shufflevector <10 x float> [[WIDE_VEC13]], <10 x float> poison, <2 x i32> <i32 0, i32 5>
+; FORCE-NEXT:    [[STRIDED_VEC18:%.*]] = shufflevector <10 x float> [[WIDE_VEC14]], <10 x float> poison, <2 x i32> <i32 0, i32 5>
+; FORCE-NEXT:    [[STRIDED_VEC19:%.*]] = shufflevector <10 x float> [[WIDE_VEC15]], <10 x float> poison, <2 x i32> <i32 0, i32 5>
+; FORCE-NEXT:    [[STRIDED_VEC20:%.*]] = shufflevector <10 x float> [[WIDE_VEC16]], <10 x float> poison, <2 x i32> <i32 0, i32 5>
+; FORCE-NEXT:    [[TMP40:%.*]] = fadd <2 x float> [[STRIDED_VEC17]], [[TMP20]]
+; FORCE-NEXT:    [[TMP41:%.*]] = fadd <2 x float> [[STRIDED_VEC18]], [[TMP21]]
+; FORCE-NEXT:    [[TMP42:%.*]] = fadd <2 x float> [[STRIDED_VEC19]], [[TMP22]]
+; FORCE-NEXT:    [[TMP43:%.*]] = fadd <2 x float> [[STRIDED_VEC20]], [[TMP23]]
+; FORCE-NEXT:    [[TMP44:%.*]] = extractelement <2 x float> [[TMP40]], i32 0
+; FORCE-NEXT:    store float [[TMP44]], float* [[TMP24]], align 4
+; FORCE-NEXT:    [[TMP45:%.*]] = extractelement <2 x float> [[TMP40]], i32 1
+; FORCE-NEXT:    store float [[TMP45]], float* [[TMP25]], align 4
+; FORCE-NEXT:    [[TMP46:%.*]] = extractelement <2 x float> [[TMP41]], i32 0
+; FORCE-NEXT:    store float [[TMP46]], float* [[TMP26]], align 4
+; FORCE-NEXT:    [[TMP47:%.*]] = extractelement <2 x float> [[TMP41]], i32 1
+; FORCE-NEXT:    store float [[TMP47]], float* [[TMP27]], align 4
+; FORCE-NEXT:    [[TMP48:%.*]] = extractelement <2 x float> [[TMP42]], i32 0
+; FORCE-NEXT:    store float [[TMP48]], float* [[TMP28]], align 4
+; FORCE-NEXT:    [[TMP49:%.*]] = extractelement <2 x float> [[TMP42]], i32 1
+; FORCE-NEXT:    store float [[TMP49]], float* [[TMP29]], align 4
+; FORCE-NEXT:    [[TMP50:%.*]] = extractelement <2 x float> [[TMP43]], i32 0
+; FORCE-NEXT:    store float [[TMP50]], float* [[TMP30]], align 4
+; FORCE-NEXT:    [[TMP51:%.*]] = extractelement <2 x float> [[TMP43]], i32 1
+; FORCE-NEXT:    store float [[TMP51]], float* [[TMP31]], align 4
+; FORCE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; FORCE-NEXT:    [[TMP52:%.*]] = icmp eq i64 [[INDEX_NEXT]], 6392
+; FORCE-NEXT:    br i1 [[TMP52]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; FORCE:       middle.block:
+; FORCE-NEXT:    br label [[SCALAR_PH]]
+; FORCE:       scalar.ph:
+; FORCE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 31960, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FORCE-NEXT:    br label [[FOR_BODY:%.*]]
+; FORCE:       for.body:
+; FORCE-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; FORCE-NEXT:    [[VAR0:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 3, i64 [[I]]
+; FORCE-NEXT:    [[VAR1:%.*]] = load float, float* [[VAR0]], align 4
+; FORCE-NEXT:    [[VAR2:%.*]] = fmul float [[X]], [[VAR1]]
+; FORCE-NEXT:    [[VAR3:%.*]] = getelementptr inbounds [[DATA]], %data* [[D]], i64 0, i32 0, i64 [[I]]
+; FORCE-NEXT:    [[VAR4:%.*]] = load float, float* [[VAR3]], align 4
+; FORCE-NEXT:    [[VAR5:%.*]] = fadd float [[VAR4]], [[VAR2]]
+; FORCE-NEXT:    store float [[VAR5]], float* [[VAR3]], align 4
+; FORCE-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 5
+; FORCE-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], 32000
+; FORCE-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP2:![0-9]+]]
+; FORCE:       for.end:
+; FORCE-NEXT:    ret void
+;
 entry:
   br label %for.body
 
 for.body:
   %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]
-  %tmp0 = getelementptr inbounds %data, %data* %d, i64 0, i32 3, i64 %i
-  %tmp1 = load float, float* %tmp0, align 4
-  %tmp2 = fmul float %x, %tmp1
-  %tmp3 = getelementptr inbounds %data, %data* %d, i64 0, i32 0, i64 %i
-  %tmp4 = load float, float* %tmp3, align 4
-  %tmp5 = fadd float %tmp4, %tmp2
-  store float %tmp5, float* %tmp3, align 4
+  %var0 = getelementptr inbounds %data, %data* %d, i64 0, i32 3, i64 %i
+  %var1 = load float, float* %var0, align 4
+  %var2 = fmul float %x, %var1
+  %var3 = getelementptr inbounds %data, %data* %d, i64 0, i32 0, i64 %i
+  %var4 = load float, float* %var3, align 4
+  %var5 = fadd float %var4, %var2
+  store float %var5, float* %var3, align 4
   %i.next = add nuw nsw i64 %i, 5
   %cond = icmp slt i64 %i.next, 32000
   br i1 %cond, label %for.body, label %for.end
@@ -68,60 +203,81 @@ for.end:
 
 attributes #0 = { "target-cpu"="knl" }
 
-; CHECK-LABEL: PR40816
 ;
 ; Check that scalar with predication instructions are not considered uniform
 ; after vectorization, because that results in replicating a region instead of
 ; having a single instance (out of VF). The predication stems from a tiny count
 ; of 3 leading to folding the tail by masking using icmp ule <i, i+1> <= <2, 2>.
 ;
-; CHECK:     LV: Found trip count: 3
-; CHECK:     LV: Found uniform instruction:   {{%.*}} = icmp eq i32 {{%.*}}, 0
-; CHECK-NOT: LV: Found uniform instruction:   {{%.*}} = load i32, i32* {{%.*}}, align 1
-; CHECK:     LV: Found not uniform being ScalarWithPredication:  {{%.*}} = load i32, i32* {{%.*}}, align 1
-; CHECK:     LV: Found scalar instruction:   {{%.*}} = getelementptr inbounds [3 x i32], [3 x i32]* @a, i32 0, i32 {{%.*}}
+;
+;
+ at a = internal constant [3 x i32] [i32 7, i32 7, i32 0], align 1
+ at b = external global i32, align 1
+
+define void @PR40816() #1 {
+; CHECK-LABEL: @PR40816(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    store i32 [[TMP0]], i32* @b, align 1
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp eq i32 [[TMP0]], 2
+; CHECK-NEXT:    [[INC]] = add nuw nsw i32 [[TMP0]], 1
+; CHECK-NEXT:    br i1 [[CMP2]], label [[RETURN:%.*]], label [[FOR_BODY]]
+; CHECK:       return:
+; CHECK-NEXT:    ret void
 ;
 ; FORCE-LABEL: @PR40816(
 ; FORCE-NEXT:  entry:
-; FORCE-NEXT:    br i1 false, label {{%.*}}, label [[VECTOR_PH:%.*]]
+; FORCE-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; FORCE:       vector.ph:
 ; FORCE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; FORCE:       vector.body:
-; FORCE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE4:%.*]] ]
-; FORCE-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_LOAD_CONTINUE4]] ]
-; FORCE-NEXT:    [[TMP2:%.*]] = icmp ule <2 x i32> [[VEC_IND]], <i32 2, i32 2>
-; FORCE-NEXT:    [[TMP3:%.*]] = extractelement <2 x i1> [[TMP2]], i32 0
-; FORCE-NEXT:    br i1 [[TMP3]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
+; FORCE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE2:%.*]] ]
+; FORCE-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_LOAD_CONTINUE2]] ]
+; FORCE-NEXT:    [[TMP0:%.*]] = icmp ule <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; FORCE-NEXT:    [[TMP1:%.*]] = extractelement <2 x i1> [[TMP0]], i32 0
+; FORCE-NEXT:    br i1 [[TMP1]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
 ; FORCE:       pred.load.if:
-; FORCE-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
-; FORCE-NEXT:    store i32 [[TMP0]], i32* @b, align 1
-; FORCE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i32], [3 x i32]* @a, i32 0, i32 [[TMP0]]
-; FORCE-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 1
-; FORCE-NEXT:    [[TMP8:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i32 0
+; FORCE-NEXT:    [[TMP2:%.*]] = add i32 [[INDEX]], 0
+; FORCE-NEXT:    store i32 [[TMP2]], i32* @b, align 1
+; FORCE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [3 x i32], [3 x i32]* @a, i32 0, i32 [[TMP2]]
+; FORCE-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 1
+; FORCE-NEXT:    [[TMP5:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i32 0
 ; FORCE-NEXT:    br label [[PRED_LOAD_CONTINUE]]
 ; FORCE:       pred.load.continue:
-; FORCE-NEXT:    [[TMP9:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_LOAD_IF]] ]
-; FORCE-NEXT:    [[TMP10:%.*]] = extractelement <2 x i1> [[TMP2]], i32 1
-; FORCE-NEXT:    br i1 [[TMP10]], label [[PRED_LOAD_IF3:%.*]], label [[PRED_LOAD_CONTINUE4]]
+; FORCE-NEXT:    [[TMP6:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP5]], [[PRED_LOAD_IF]] ]
+; FORCE-NEXT:    [[TMP7:%.*]] = extractelement <2 x i1> [[TMP0]], i32 1
+; FORCE-NEXT:    br i1 [[TMP7]], label [[PRED_LOAD_IF1:%.*]], label [[PRED_LOAD_CONTINUE2]]
 ; FORCE:       pred.load.if1:
-; FORCE-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 1
-; FORCE-NEXT:    store i32 [[TMP1]], i32* @b, align 1
-; FORCE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i32], [3 x i32]* @a, i32 0, i32 [[TMP1]]
-; FORCE-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 1
-; FORCE-NEXT:    [[TMP13:%.*]] = insertelement <2 x i32> [[TMP9]], i32 [[TMP12]], i32 1
-; FORCE-NEXT:    br label [[PRED_LOAD_CONTINUE4]]
+; FORCE-NEXT:    [[TMP8:%.*]] = add i32 [[INDEX]], 1
+; FORCE-NEXT:    store i32 [[TMP8]], i32* @b, align 1
+; FORCE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i32], [3 x i32]* @a, i32 0, i32 [[TMP8]]
+; FORCE-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 1
+; FORCE-NEXT:    [[TMP11:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP10]], i32 1
+; FORCE-NEXT:    br label [[PRED_LOAD_CONTINUE2]]
 ; FORCE:       pred.load.continue2:
-; FORCE-NEXT:    [[TMP14:%.*]] = phi <2 x i32> [ [[TMP9]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP13]], [[PRED_LOAD_IF3]] ]
+; FORCE-NEXT:    [[TMP12:%.*]] = phi <2 x i32> [ [[TMP6]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP11]], [[PRED_LOAD_IF1]] ]
 ; FORCE-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 2
 ; FORCE-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
-; FORCE-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], 4
-; FORCE-NEXT:    br i1 [[TMP15]], label {{%.*}}, label [[VECTOR_BODY]]
+; FORCE-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], 4
+; FORCE-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; FORCE:       middle.block:
+; FORCE-NEXT:    br i1 true, label [[RETURN:%.*]], label [[SCALAR_PH]]
+; FORCE:       scalar.ph:
+; FORCE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 4, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FORCE-NEXT:    br label [[FOR_BODY:%.*]]
+; FORCE:       for.body:
+; FORCE-NEXT:    [[TMP14:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; FORCE-NEXT:    store i32 [[TMP14]], i32* @b, align 1
+; FORCE-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [3 x i32], [3 x i32]* @a, i32 0, i32 [[TMP14]]
+; FORCE-NEXT:    [[TMP15:%.*]] = load i32, i32* [[ARRAYIDX1]], align 1
+; FORCE-NEXT:    [[CMP2:%.*]] = icmp eq i32 [[TMP15]], 0
+; FORCE-NEXT:    [[INC]] = add nuw nsw i32 [[TMP14]], 1
+; FORCE-NEXT:    br i1 [[CMP2]], label [[RETURN]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; FORCE:       return:
+; FORCE-NEXT:    ret void
 ;
- at a = internal constant [3 x i32] [i32 7, i32 7, i32 0], align 1
- at b = external global i32, align 1
-
-define void @PR40816() #1 {
-
 entry:
   br label %for.body
 

diff  --git a/llvm/test/Transforms/LoopVectorize/bsd_regex.ll b/llvm/test/Transforms/LoopVectorize/bsd_regex.ll
index ddb00ba31e14..55e67d7f81f3 100644
--- a/llvm/test/Transforms/LoopVectorize/bsd_regex.ll
+++ b/llvm/test/Transforms/LoopVectorize/bsd_regex.ll
@@ -1,23 +1,50 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -loop-vectorize -dce -instcombine -force-vector-width=2 -force-vector-interleave=2 < %s | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 
 ;PR 15830.
 
-;CHECK-LABEL: @foo(
 ; When scalarizing stores we need to preserve the original order.
 ; Make sure that we are extracting in the correct order (0101, and not 0011).
-;CHECK: extractelement <2 x i64> {{.*}}, i32 0
-;CHECK: extractelement <2 x i64> {{.*}}, i32 1
-;CHECK: extractelement <2 x i64> {{.*}}, i32 0
-;CHECK: extractelement <2 x i64> {{.*}}, i32 1
-;CHECK: store
-;CHECK: store
-;CHECK: store
-;CHECK: store
-;CHECK: ret
 
 define i32 @foo(i32* nocapture %A) {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[STEP_ADD:%.*]] = shl <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i64> [[STEP_ADD]], <i64 8, i64 8>
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i64> [[TMP0]], i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x i64> [[TMP0]], i32 1
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x i64> [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP6]]
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP1]], i32 1
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
+; CHECK-NEXT:    store i32 4, i32* [[TMP3]], align 4
+; CHECK-NEXT:    store i32 4, i32* [[TMP5]], align 4
+; CHECK-NEXT:    store i32 4, i32* [[TMP7]], align 4
+; CHECK-NEXT:    store i32 4, i32* [[TMP9]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 4, i64 4>
+; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
+; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    br i1 undef, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret i32 undef
+;
 entry:
   br label %for.body
 

diff  --git a/llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
index c65f62cef65e..68c886fe3d33 100644
--- a/llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
+++ b/llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -dce -instcombine -S | FileCheck %s
 ; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-interleave=2 -dce -instcombine -S | FileCheck %s --check-prefix=UNROLL
 ; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-interleave=2 -S | FileCheck %s --check-prefix=UNROLL-NO-IC
@@ -12,31 +13,452 @@ target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
 ;     b[i] =  a[i] + a[i - 1]
 ; }
 ;
+;
+;
+define void @recurrence_1(i32* nocapture readonly %a, i32* nocapture %b, i32 %n) {
 ; CHECK-LABEL: @recurrence_1(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[FOR_PREHEADER:%.*]]
+; CHECK:       for.preheader:
+; CHECK-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[A:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK:       vector.memcheck:
+; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[TMP5]]
+; CHECK-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i32, i32* [[A]], i64 1
+; CHECK-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; CHECK-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP6]]
+; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[SCEVGEP5]], [[B]]
+; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult i32* [[SCEVGEP3]], [[SCEVGEP]]
+; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK:         %vector.recur.init = insertelement <4 x i32> poison, i32 %pre_load, i32 3
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
+; CHECK-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD]], i32 3
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK:         %vector.recur = phi <4 x i32> [ %vector.recur.init, %vector.ph ], [ [[L1:%[a-zA-Z0-9.]+]], %vector.body ]
-; CHECK:         [[L1]] = load <4 x i32>
-; CHECK:         {{.*}} = shufflevector <4 x i32> %vector.recur, <4 x i32> [[L1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP7:%.*]] = or i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = bitcast i32* [[TMP8]] to <4 x i32>*
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, <4 x i32>* [[TMP9]], align 4, !alias.scope !0
+; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP12:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP10]]
+; CHECK-NEXT:    [[TMP13:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>*
+; CHECK-NEXT:    store <4 x i32> [[TMP12]], <4 x i32>* [[TMP13]], align 4, !alias.scope !3, !noalias !0
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK:         %vector.recur.extract = extractelement <4 x i32> [[L1]], i32 3
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
-; CHECK:         %scalar.recur.init = phi i32 [ %pre_load, %vector.memcheck ], [ %pre_load, %for.preheader ], [ %vector.recur.extract, %middle.block ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[VECTOR_MEMCHECK]] ], [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[FOR_PREHEADER]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[SCALAR_BODY:%.*]]
 ; CHECK:       scalar.body:
-; CHECK:         %scalar.recur = phi i32 [ %scalar.recur.init, %scalar.ph ], [ {{.*}}, %scalar.body ]
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[SCALAR_BODY]] ]
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
+; CHECK-NEXT:    [[TMP15]] = load i32, i32* [[ARRAYIDX32]], align 4
+; CHECK-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    [[ADD35:%.*]] = add i32 [[TMP15]], [[SCALAR_RECUR]]
+; CHECK-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
+; CHECK-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK:       for.exit:
+; CHECK-NEXT:    ret void
 ;
 ; UNROLL-LABEL: @recurrence_1(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br label [[FOR_PREHEADER:%.*]]
+; UNROLL:       for.preheader:
+; UNROLL-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[A:%.*]], align 4
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL:       vector.memcheck:
+; UNROLL-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; UNROLL-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; UNROLL-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; UNROLL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[TMP5]]
+; UNROLL-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i32, i32* [[A]], i64 1
+; UNROLL-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; UNROLL-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP6]]
+; UNROLL-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[SCEVGEP5]], [[B]]
+; UNROLL-NEXT:    [[BOUND1:%.*]] = icmp ult i32* [[SCEVGEP3]], [[SCEVGEP]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
+; UNROLL-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD]], i32 3
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL:       vector.body:
-; UNROLL:         %vector.recur = phi <4 x i32> [ %vector.recur.init, %vector.ph ], [ [[L2:%[a-zA-Z0-9.]+]], %vector.body ]
-; UNROLL:         [[L1:%[a-zA-Z0-9.]+]] = load <4 x i32>
-; UNROLL:         [[L2]] = load <4 x i32>
-; UNROLL:         {{.*}} = shufflevector <4 x i32> %vector.recur, <4 x i32> [[L1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; UNROLL:         {{.*}} = shufflevector <4 x i32> [[L1]], <4 x i32> [[L2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP7:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = bitcast i32* [[TMP8]] to <4 x i32>*
+; UNROLL-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP9]], align 4, !alias.scope !0
+; UNROLL-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP8]], i64 4
+; UNROLL-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <4 x i32>*
+; UNROLL-NEXT:    [[WIDE_LOAD7]] = load <4 x i32>, <4 x i32>* [[TMP11]], align 4, !alias.scope !0
+; UNROLL-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> [[WIDE_LOAD7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP15:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP12]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = add <4 x i32> [[WIDE_LOAD7]], [[TMP13]]
+; UNROLL-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP14]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> [[TMP15]], <4 x i32>* [[TMP17]], align 4, !alias.scope !3, !noalias !0
+; UNROLL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[TMP14]], i64 4
+; UNROLL-NEXT:    [[TMP19:%.*]] = bitcast i32* [[TMP18]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> [[TMP16]], <4 x i32>* [[TMP19]], align 4, !alias.scope !3, !noalias !0
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; UNROLL:       middle.block:
-; UNROLL:         %vector.recur.extract = extractelement <4 x i32> [[L2]], i32 3
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD7]], i32 3
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[VECTOR_MEMCHECK]] ], [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[FOR_PREHEADER]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL:       scalar.body:
+; UNROLL-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP21:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NEXT:    [[TMP21]] = load i32, i32* [[ARRAYIDX32]], align 4
+; UNROLL-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NEXT:    [[ADD35:%.*]] = add i32 [[TMP21]], [[SCALAR_RECUR]]
+; UNROLL-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
+; UNROLL-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; UNROLL:       for.exit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @recurrence_1(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; UNROLL-NO-IC-NEXT:    br label [[FOR_PREHEADER:%.*]]
+; UNROLL-NO-IC:       for.preheader:
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 0
+; UNROLL-NO-IC-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[ARRAYIDX_PHI_TRANS_INSERT]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-IC:       vector.memcheck:
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i32, i32* [[A]], i64 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP34:%.*]] = bitcast i32* [[SCEVGEP3]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP6]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP56:%.*]] = bitcast i32* [[SCEVGEP5]] to i8*
+; UNROLL-NO-IC-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; UNROLL-NO-IC-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-IC-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add nuw nsw i64 [[TMP7]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add nuw nsw i64 [[TMP8]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP9]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP10]]
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[TMP11]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = bitcast i32* [[TMP13]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP14]], align 4, !alias.scope !0
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP11]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD7]] = load <4 x i32>, <4 x i32>* [[TMP16]], align 4, !alias.scope !0
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> [[WIDE_LOAD7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP17]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = add <4 x i32> [[WIDE_LOAD7]], [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = bitcast i32* [[TMP23]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP21]], <4 x i32>* [[TMP24]], align 4, !alias.scope !3, !noalias !0
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = bitcast i32* [[TMP25]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP22]], <4 x i32>* [[TMP26]], align 4, !alias.scope !3, !noalias !0
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD7]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD7]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[VECTOR_MEMCHECK]] ], [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-IC:       scalar.body:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NO-IC-NEXT:    [[TMP28]] = load i32, i32* [[ARRAYIDX32]], align 4
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    [[ADD35:%.*]] = add i32 [[TMP28]], [[SCALAR_RECUR]]
+; UNROLL-NO-IC-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
+; UNROLL-NO-IC-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; UNROLL-NO-IC:       for.exit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @recurrence_1(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; UNROLL-NO-VF-NEXT:    br label [[FOR_PREHEADER:%.*]]
+; UNROLL-NO-VF:       for.preheader:
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 0
+; UNROLL-NO-VF-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[ARRAYIDX_PHI_TRANS_INSERT]], align 4
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-VF:       vector.memcheck:
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[TMP5]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i32, i32* [[A]], i64 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP34:%.*]] = bitcast i32* [[SCEVGEP3]] to i8*
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP6]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP56:%.*]] = bitcast i32* [[SCEVGEP5]] to i8*
+; UNROLL-NO-VF-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; UNROLL-NO-VF-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; UNROLL-NO-VF-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-VF-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ [[PRE_LOAD]], [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION7:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[INDUCTION]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[INDUCTION7]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP9]], align 4, !alias.scope !0
+; UNROLL-NO-VF-NEXT:    [[TMP12]] = load i32, i32* [[TMP10]], align 4, !alias.scope !0
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION7]]
+; UNROLL-NO-VF-NEXT:    [[TMP15:%.*]] = add i32 [[TMP11]], [[VECTOR_RECUR]]
+; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = add i32 [[TMP12]], [[TMP11]]
+; UNROLL-NO-VF-NEXT:    store i32 [[TMP15]], i32* [[TMP13]], align 4, !alias.scope !3, !noalias !0
+; UNROLL-NO-VF-NEXT:    store i32 [[TMP16]], i32* [[TMP14]], align 4, !alias.scope !3, !noalias !0
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[VECTOR_MEMCHECK]] ], [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-VF:       scalar.body:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NO-VF-NEXT:    [[TMP18]] = load i32, i32* [[ARRAYIDX32]], align 4
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NO-VF-NEXT:    [[ADD35:%.*]] = add i32 [[TMP18]], [[SCALAR_RECUR]]
+; UNROLL-NO-VF-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
+; UNROLL-NO-VF-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; UNROLL-NO-VF:       for.exit:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @recurrence_1(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; SINK-AFTER-NEXT:    br label [[FOR_PREHEADER:%.*]]
+; SINK-AFTER:       for.preheader:
+; SINK-AFTER-NEXT:    [[ARRAYIDX_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 0
+; SINK-AFTER-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[ARRAYIDX_PHI_TRANS_INSERT]], align 4
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; SINK-AFTER:       vector.memcheck:
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[TMP5]]
+; SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; SINK-AFTER-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i32, i32* [[A]], i64 1
+; SINK-AFTER-NEXT:    [[SCEVGEP34:%.*]] = bitcast i32* [[SCEVGEP3]] to i8*
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; SINK-AFTER-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP6]]
+; SINK-AFTER-NEXT:    [[SCEVGEP56:%.*]] = bitcast i32* [[SCEVGEP5]] to i8*
+; SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; SINK-AFTER-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD]], i32 3
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[TMP7]], 1
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i32 0
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <4 x i32>*
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, <4 x i32>* [[TMP11]], align 4, !alias.scope !0
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP7]]
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP12]]
+; SINK-AFTER-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP13]], i32 0
+; SINK-AFTER-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <4 x i32>*
+; SINK-AFTER-NEXT:    store <4 x i32> [[TMP14]], <4 x i32>* [[TMP16]], align 4, !alias.scope !3, !noalias !0
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[VECTOR_MEMCHECK]] ], [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; SINK-AFTER:       scalar.body:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SINK-AFTER-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
+; SINK-AFTER-NEXT:    [[TMP18]] = load i32, i32* [[ARRAYIDX32]], align 4
+; SINK-AFTER-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; SINK-AFTER-NEXT:    [[ADD35:%.*]] = add i32 [[TMP18]], [[SCALAR_RECUR]]
+; SINK-AFTER-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
+; SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; SINK-AFTER:       for.exit:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @recurrence_1(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; NO-SINK-AFTER-NEXT:    br label [[FOR_PREHEADER:%.*]]
+; NO-SINK-AFTER:       for.preheader:
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 0
+; NO-SINK-AFTER-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[ARRAYIDX_PHI_TRANS_INSERT]], align 4
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; NO-SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; NO-SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; NO-SINK-AFTER:       vector.memcheck:
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[TMP5]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i32, i32* [[A]], i64 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP34:%.*]] = bitcast i32* [[SCEVGEP3]] to i8*
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP6]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP56:%.*]] = bitcast i32* [[SCEVGEP5]] to i8*
+; NO-SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; NO-SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; NO-SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-SINK-AFTER-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[TMP7]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <4 x i32>*
+; NO-SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, <4 x i32>* [[TMP11]], align 4, !alias.scope !0
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP7]]
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP12]]
+; NO-SINK-AFTER-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP13]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <4 x i32>*
+; NO-SINK-AFTER-NEXT:    store <4 x i32> [[TMP14]], <4 x i32>* [[TMP16]], align 4, !alias.scope !3, !noalias !0
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[VECTOR_MEMCHECK]] ], [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; NO-SINK-AFTER:       scalar.body:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
+; NO-SINK-AFTER-NEXT:    [[TMP18]] = load i32, i32* [[ARRAYIDX32]], align 4
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; NO-SINK-AFTER-NEXT:    [[ADD35:%.*]] = add i32 [[TMP18]], [[SCALAR_RECUR]]
+; NO-SINK-AFTER-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
+; NO-SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; NO-SINK-AFTER:       for.exit:
+; NO-SINK-AFTER-NEXT:    ret void
 ;
-define void @recurrence_1(i32* nocapture readonly %a, i32* nocapture %b, i32 %n) {
 entry:
   br label %for.preheader
 
@@ -69,31 +491,439 @@ for.exit:
 ;   return minmax;
 ; }
 ;
+;
+;
+define i32 @recurrence_2(i32* nocapture readonly %a, i32 %n) {
 ; CHECK-LABEL: @recurrence_2(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK:       for.preheader:
+; CHECK-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
+; CHECK-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK:         %vector.recur.init = insertelement <4 x i32> poison, i32 %.pre, i32 3
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
+; CHECK-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[DOTPRE]], i32 3
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK:         %vector.recur = phi <4 x i32> [ %vector.recur.init, %vector.ph ], [ [[L1:%[a-zA-Z0-9.]+]], %vector.body ]
-; CHECK:         [[L1]] = load <4 x i32>
-; CHECK:         {{.*}} = shufflevector <4 x i32> %vector.recur, <4 x i32> [[L1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <4 x i32>*
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, <4 x i32>* [[TMP4]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[TMP6:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD]], [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp sgt <4 x i32> [[TMP6]], zeroinitializer
+; CHECK-NEXT:    [[TMP8:%.*]] = select <4 x i1> [[TMP7]], <4 x i32> [[TMP6]], <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[TMP8]]
+; CHECK-NEXT:    [[TMP10]] = select <4 x i1> [[TMP9]], <4 x i32> [[VEC_PHI]], <4 x i32> [[TMP8]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK:         %vector.recur.extract = extractelement <4 x i32> [[L1]], i32 3
+; CHECK-NEXT:    [[TMP12:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[TMP10]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
-; CHECK:         %scalar.recur.init = phi i32 [ %.pre, %for.preheader ], [ %vector.recur.extract, %middle.block ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[FOR_PREHEADER]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[SCALAR_BODY:%.*]]
+; CHECK:       for.cond.cleanup.loopexit:
+; CHECK-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
 ; CHECK:       scalar.body:
-; CHECK:         %scalar.recur = phi i32 [ %scalar.recur.init, %scalar.ph ], [ {{.*}}, %scalar.body ]
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP13:%.*]], [[SCALAR_BODY]] ]
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; CHECK-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    [[TMP13]] = load i32, i32* [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP13]], [[SCALAR_RECUR]]
+; CHECK-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
+; CHECK-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
+; CHECK-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
+; CHECK-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ;
 ; UNROLL-LABEL: @recurrence_2(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; UNROLL-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; UNROLL:       for.preheader:
+; UNROLL-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
+; UNROLL-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; UNROLL-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
+; UNROLL-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[DOTPRE]], i32 3
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL:       vector.body:
-; UNROLL:         %vector.recur = phi <4 x i32> [ %vector.recur.init, %vector.ph ], [ [[L2:%[a-zA-Z0-9.]+]], %vector.body ]
-; UNROLL:         [[L1:%[a-zA-Z0-9.]+]] = load <4 x i32>
-; UNROLL:         [[L2]] = load <4 x i32>
-; UNROLL:         {{.*}} = shufflevector <4 x i32> %vector.recur, <4 x i32> [[L1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; UNROLL:         {{.*}} = shufflevector <4 x i32> [[L1]], <4 x i32> [[L2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD2:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <4 x i32>*
+; UNROLL-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP4]], align 4
+; UNROLL-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i64 4
+; UNROLL-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
+; UNROLL-NEXT:    [[WIDE_LOAD2]] = load <4 x i32>, <4 x i32>* [[TMP6]], align 4
+; UNROLL-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> [[WIDE_LOAD2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP9:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD]], [[TMP7]]
+; UNROLL-NEXT:    [[TMP10:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD2]], [[TMP8]]
+; UNROLL-NEXT:    [[TMP11:%.*]] = icmp sgt <4 x i32> [[TMP9]], zeroinitializer
+; UNROLL-NEXT:    [[TMP12:%.*]] = icmp sgt <4 x i32> [[TMP10]], zeroinitializer
+; UNROLL-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP11]], <4 x i32> [[TMP9]], <4 x i32> zeroinitializer
+; UNROLL-NEXT:    [[TMP14:%.*]] = select <4 x i1> [[TMP12]], <4 x i32> [[TMP10]], <4 x i32> zeroinitializer
+; UNROLL-NEXT:    [[TMP15:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[TMP13]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = icmp slt <4 x i32> [[VEC_PHI1]], [[TMP14]]
+; UNROLL-NEXT:    [[TMP17]] = select <4 x i1> [[TMP15]], <4 x i32> [[VEC_PHI]], <4 x i32> [[TMP13]]
+; UNROLL-NEXT:    [[TMP18]] = select <4 x i1> [[TMP16]], <4 x i32> [[VEC_PHI1]], <4 x i32> [[TMP14]]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; UNROLL:       middle.block:
-; UNROLL:         %vector.recur.extract = extractelement <4 x i32> [[L2]], i32 3
+; UNROLL-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp slt <4 x i32> [[TMP17]], [[TMP18]]
+; UNROLL-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP17]], <4 x i32> [[TMP18]]
+; UNROLL-NEXT:    [[TMP20:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[RDX_MINMAX_SELECT]])
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i32 3
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[FOR_PREHEADER]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL:       for.cond.cleanup.loopexit:
+; UNROLL-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[FOR_COND_CLEANUP]]
+; UNROLL:       for.cond.cleanup:
+; UNROLL-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
+; UNROLL-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
+; UNROLL:       scalar.body:
+; UNROLL-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP21:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
+; UNROLL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; UNROLL-NEXT:    [[TMP21]] = load i32, i32* [[ARRAYIDX]], align 4
+; UNROLL-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP21]], [[SCALAR_RECUR]]
+; UNROLL-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
+; UNROLL-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
+; UNROLL-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
+; UNROLL-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+;
+; UNROLL-NO-IC-LABEL: @recurrence_2(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; UNROLL-NO-IC:       for.preheader:
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
+; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[DOTPRE]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD2:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP8]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD2]] = load <4 x i32>, <4 x i32>* [[TMP10]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> [[WIDE_LOAD2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD]], [[TMP11]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD2]], [[TMP12]]
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = icmp sgt <4 x i32> [[TMP13]], zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = icmp sgt <4 x i32> [[TMP14]], zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = select <4 x i1> [[TMP15]], <4 x i32> [[TMP13]], <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = select <4 x i1> [[TMP16]], <4 x i32> [[TMP14]], <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[TMP17]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = icmp slt <4 x i32> [[VEC_PHI1]], [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[TMP21]] = select <4 x i1> [[TMP19]], <4 x i32> [[VEC_PHI]], <4 x i32> [[TMP17]]
+; UNROLL-NO-IC-NEXT:    [[TMP22]] = select <4 x i1> [[TMP20]], <4 x i32> [[VEC_PHI1]], <4 x i32> [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp slt <4 x i32> [[TMP21]], [[TMP22]]
+; UNROLL-NO-IC-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP21]], <4 x i32> [[TMP22]]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[RDX_MINMAX_SELECT]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[TMP24]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-IC:       for.cond.cleanup.loopexit:
+; UNROLL-NO-IC-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[TMP24]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_COND_CLEANUP]]
+; UNROLL-NO-IC:       for.cond.cleanup:
+; UNROLL-NO-IC-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
+; UNROLL-NO-IC:       scalar.body:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP25:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    [[TMP25]] = load i32, i32* [[ARRAYIDX]], align 4
+; UNROLL-NO-IC-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP25]], [[SCALAR_RECUR]]
+; UNROLL-NO-IC-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
+; UNROLL-NO-IC-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
+; UNROLL-NO-IC-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+;
+; UNROLL-NO-VF-LABEL: @recurrence_2(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; UNROLL-NO-VF:       for.preheader:
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
+; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ [[DOTPRE]], [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ poison, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VEC_PHI2:%.*]] = phi i32 [ poison, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDUCTION1]]
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 4
+; UNROLL-NO-VF-NEXT:    [[TMP6]] = load i32, i32* [[TMP4]], align 4
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sub nsw i32 [[TMP5]], [[VECTOR_RECUR]]
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sub nsw i32 [[TMP6]], [[TMP5]]
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP7]], 0
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], 0
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP9]], i32 [[TMP7]], i32 0
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 0
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = icmp slt i32 [[VEC_PHI]], [[TMP11]]
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = icmp slt i32 [[VEC_PHI2]], [[TMP12]]
+; UNROLL-NO-VF-NEXT:    [[TMP15]] = select i1 [[TMP13]], i32 [[VEC_PHI]], i32 [[TMP11]]
+; UNROLL-NO-VF-NEXT:    [[TMP16]] = select i1 [[TMP14]], i32 [[VEC_PHI2]], i32 [[TMP12]]
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp slt i32 [[TMP15]], [[TMP16]]
+; UNROLL-NO-VF-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select i1 [[RDX_MINMAX_CMP]], i32 [[TMP15]], i32 [[TMP16]]
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[RDX_MINMAX_SELECT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-VF:       for.cond.cleanup.loopexit:
+; UNROLL-NO-VF-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[RDX_MINMAX_SELECT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_COND_CLEANUP]]
+; UNROLL-NO-VF:       for.cond.cleanup:
+; UNROLL-NO-VF-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
+; UNROLL-NO-VF-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
+; UNROLL-NO-VF:       scalar.body:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; UNROLL-NO-VF-NEXT:    [[TMP18]] = load i32, i32* [[ARRAYIDX]], align 4
+; UNROLL-NO-VF-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP18]], [[SCALAR_RECUR]]
+; UNROLL-NO-VF-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
+; UNROLL-NO-VF-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
+; UNROLL-NO-VF-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
+; UNROLL-NO-VF-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-VF-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+;
+; SINK-AFTER-LABEL: @recurrence_2(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; SINK-AFTER-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; SINK-AFTER:       for.preheader:
+; SINK-AFTER-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
+; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[DOTPRE]], i32 3
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP3]]
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, <4 x i32>* [[TMP6]], align 4
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD]], [[TMP7]]
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = icmp sgt <4 x i32> [[TMP8]], zeroinitializer
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = select <4 x i1> [[TMP9]], <4 x i32> [[TMP8]], <4 x i32> zeroinitializer
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[TMP10]]
+; SINK-AFTER-NEXT:    [[TMP12]] = select <4 x i1> [[TMP11]], <4 x i32> [[VEC_PHI]], <4 x i32> [[TMP10]]
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[TMP12]])
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
+; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; SINK-AFTER:       for.cond.cleanup.loopexit:
+; SINK-AFTER-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    br label [[FOR_COND_CLEANUP]]
+; SINK-AFTER:       for.cond.cleanup:
+; SINK-AFTER-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
+; SINK-AFTER-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
+; SINK-AFTER:       scalar.body:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; SINK-AFTER-NEXT:    [[TMP15]] = load i32, i32* [[ARRAYIDX]], align 4
+; SINK-AFTER-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP15]], [[SCALAR_RECUR]]
+; SINK-AFTER-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
+; SINK-AFTER-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
+; SINK-AFTER-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
+; SINK-AFTER-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
+; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+;
+; NO-SINK-AFTER-LABEL: @recurrence_2(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; NO-SINK-AFTER:       for.preheader:
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
+; NO-SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; NO-SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; NO-SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[DOTPRE]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP3]]
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
+; NO-SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, <4 x i32>* [[TMP6]], align 4
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD]], [[TMP7]]
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = icmp sgt <4 x i32> [[TMP8]], zeroinitializer
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = select <4 x i1> [[TMP9]], <4 x i32> [[TMP8]], <4 x i32> zeroinitializer
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[TMP10]]
+; NO-SINK-AFTER-NEXT:    [[TMP12]] = select <4 x i1> [[TMP11]], <4 x i32> [[VEC_PHI]], <4 x i32> [[TMP10]]
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[TMP12]])
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; NO-SINK-AFTER:       for.cond.cleanup.loopexit:
+; NO-SINK-AFTER-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_COND_CLEANUP]]
+; NO-SINK-AFTER:       for.cond.cleanup:
+; NO-SINK-AFTER-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
+; NO-SINK-AFTER-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
+; NO-SINK-AFTER:       scalar.body:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; NO-SINK-AFTER-NEXT:    [[TMP15]] = load i32, i32* [[ARRAYIDX]], align 4
+; NO-SINK-AFTER-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP15]], [[SCALAR_RECUR]]
+; NO-SINK-AFTER-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
+; NO-SINK-AFTER-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
+; NO-SINK-AFTER-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
+; NO-SINK-AFTER-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; NO-SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ;
-define i32 @recurrence_2(i32* nocapture readonly %a, i32 %n) {
 entry:
   %cmp27 = icmp sgt i32 %n, 0
   br i1 %cmp27, label %for.preheader, label %for.cond.cleanup
@@ -134,34 +964,570 @@ scalar.body:
 ;     b[i] = (double)a[i] - f * (double)a[i - 1];
 ; }
 ;
+; Check also that the casts were not moved needlessly.
+;
+;
+define void @recurrence_3(i16* nocapture readonly %a, double* nocapture %b, i32 %n, float %f, i16 %p) {
 ; CHECK-LABEL: @recurrence_3(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; CHECK-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
+; CHECK-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
+; CHECK-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
+; CHECK-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
+; CHECK-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
+; CHECK-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
+; CHECK-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
+; CHECK-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; CHECK:       for.preheader:
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; CHECK-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP1]], 3
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK:       vector.memcheck:
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr double, double* [[B]], i64 1
+; CHECK-NEXT:    [[TMP4:%.*]] = add i32 [[N]], -2
+; CHECK-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
+; CHECK-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; CHECK-NEXT:    [[SCEVGEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP6]]
+; CHECK-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; CHECK-NEXT:    [[SCEVGEP6:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP6]]
+; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i16* [[SCEVGEP6]] to double*
+; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ult double* [[SCEVGEP]], [[TMP7]]
+; CHECK-NEXT:    [[TMP8:%.*]] = bitcast double* [[SCEVGEP2]] to i16*
+; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult i16* [[SCEVGEP4]], [[TMP8]]
+; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK:         %vector.recur.init = insertelement <4 x i16> poison, i16 %0, i32 3
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP3]], 8589934588
+; CHECK-NEXT:    [[IND_END:%.*]] = or i64 [[N_VEC]], 1
+; CHECK-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i32 3
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK:         %vector.recur = phi <4 x i16> [ %vector.recur.init, %vector.ph ], [ [[L1:%[a-zA-Z0-9.]+]], %vector.body ]
-; CHECK:         [[L1]] = load <4 x i16>
-; CHECK:         [[SHUF:%[a-zA-Z0-9.]+]] = shufflevector <4 x i16> %vector.recur, <4 x i16> [[L1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; Check also that the casts were not moved needlessly.
-; CHECK:         sitofp <4 x i16> [[L1]] to <4 x double>
-; CHECK:         sitofp <4 x i16> [[SHUF]] to <4 x double>
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = or i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP10:%.*]] = bitcast i16* [[TMP9]] to <4 x i16>*
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP10]], align 2, !alias.scope !11
+; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[TMP12:%.*]] = sitofp <4 x i16> [[WIDE_LOAD]] to <4 x double>
+; CHECK-NEXT:    [[TMP13:%.*]] = sitofp <4 x i16> [[TMP11]] to <4 x double>
+; CHECK-NEXT:    [[TMP14:%.*]] = fmul fast <4 x double> [[BROADCAST_SPLAT]], [[TMP13]]
+; CHECK-NEXT:    [[TMP15:%.*]] = fsub fast <4 x double> [[TMP12]], [[TMP14]]
+; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP17:%.*]] = bitcast double* [[TMP16]] to <4 x double>*
+; CHECK-NEXT:    store <4 x double> [[TMP15]], <4 x double>* [[TMP17]], align 8, !alias.scope !14, !noalias !11
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK:         %vector.recur.extract = extractelement <4 x i16> [[L1]], i32 3
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
-; CHECK:         %scalar.recur.init = phi i16 [ %0, %vector.memcheck ], [ %0, %for.preheader ], [ %vector.recur.extract, %middle.block ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[VECTOR_MEMCHECK]] ], [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[VECTOR_MEMCHECK]] ], [ 1, [[FOR_PREHEADER]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[SCALAR_BODY:%.*]]
 ; CHECK:       scalar.body:
-; CHECK:         %scalar.recur = phi i16 [ %scalar.recur.init, %scalar.ph ], [ {{.*}}, %scalar.body ]
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP19:%.*]], [[SCALAR_BODY]] ]
+; CHECK-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
+; CHECK-NEXT:    [[TMP19]] = load i16, i16* [[ARRAYIDX5]], align 2
+; CHECK-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP19]] to double
+; CHECK-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
+; CHECK-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
+; CHECK-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
+; CHECK-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
+; CHECK-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
+; CHECK-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
+; CHECK-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; CHECK:       for.end.loopexit:
+; CHECK-NEXT:    br label [[FOR_END]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
 ;
 ; UNROLL-LABEL: @recurrence_3(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
+; UNROLL-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
+; UNROLL-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
+; UNROLL-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
+; UNROLL-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
+; UNROLL-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
+; UNROLL-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
+; UNROLL-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; UNROLL:       for.preheader:
+; UNROLL-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
+; UNROLL-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; UNROLL-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP1]], 7
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL:       vector.memcheck:
+; UNROLL-NEXT:    [[SCEVGEP:%.*]] = getelementptr double, double* [[B]], i64 1
+; UNROLL-NEXT:    [[TMP4:%.*]] = add i32 [[N]], -2
+; UNROLL-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
+; UNROLL-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; UNROLL-NEXT:    [[SCEVGEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP6]]
+; UNROLL-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NEXT:    [[SCEVGEP6:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP6]]
+; UNROLL-NEXT:    [[TMP7:%.*]] = bitcast i16* [[SCEVGEP6]] to double*
+; UNROLL-NEXT:    [[BOUND0:%.*]] = icmp ult double* [[SCEVGEP]], [[TMP7]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = bitcast double* [[SCEVGEP2]] to i16*
+; UNROLL-NEXT:    [[BOUND1:%.*]] = icmp ult i16* [[SCEVGEP4]], [[TMP8]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP3]], 8589934584
+; UNROLL-NEXT:    [[IND_END:%.*]] = or i64 [[N_VEC]], 1
+; UNROLL-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i32 3
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT9:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT10:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT9]], <4 x double> poison, <4 x i32> zeroinitializer
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL:       vector.body:
-; UNROLL:         %vector.recur = phi <4 x i16> [ %vector.recur.init, %vector.ph ], [ [[L2:%[a-zA-Z0-9.]+]], %vector.body ]
-; UNROLL:         [[L1:%[a-zA-Z0-9.]+]] = load <4 x i16>
-; UNROLL:         [[L2]] = load <4 x i16>
-; UNROLL:         {{.*}} = shufflevector <4 x i16> %vector.recur, <4 x i16> [[L1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; UNROLL:         {{.*}} = shufflevector <4 x i16> [[L1]], <4 x i16> [[L2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD8:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[OFFSET_IDX]]
+; UNROLL-NEXT:    [[TMP10:%.*]] = bitcast i16* [[TMP9]] to <4 x i16>*
+; UNROLL-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP10]], align 2, !alias.scope !11
+; UNROLL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i16, i16* [[TMP9]], i64 4
+; UNROLL-NEXT:    [[TMP12:%.*]] = bitcast i16* [[TMP11]] to <4 x i16>*
+; UNROLL-NEXT:    [[WIDE_LOAD8]] = load <4 x i16>, <4 x i16>* [[TMP12]], align 2, !alias.scope !11
+; UNROLL-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD8]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP15:%.*]] = sitofp <4 x i16> [[WIDE_LOAD]] to <4 x double>
+; UNROLL-NEXT:    [[TMP16:%.*]] = sitofp <4 x i16> [[WIDE_LOAD8]] to <4 x double>
+; UNROLL-NEXT:    [[TMP17:%.*]] = sitofp <4 x i16> [[TMP13]] to <4 x double>
+; UNROLL-NEXT:    [[TMP18:%.*]] = sitofp <4 x i16> [[TMP14]] to <4 x double>
+; UNROLL-NEXT:    [[TMP19:%.*]] = fmul fast <4 x double> [[BROADCAST_SPLAT]], [[TMP17]]
+; UNROLL-NEXT:    [[TMP20:%.*]] = fmul fast <4 x double> [[BROADCAST_SPLAT10]], [[TMP18]]
+; UNROLL-NEXT:    [[TMP21:%.*]] = fsub fast <4 x double> [[TMP15]], [[TMP19]]
+; UNROLL-NEXT:    [[TMP22:%.*]] = fsub fast <4 x double> [[TMP16]], [[TMP20]]
+; UNROLL-NEXT:    [[TMP23:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[OFFSET_IDX]]
+; UNROLL-NEXT:    [[TMP24:%.*]] = bitcast double* [[TMP23]] to <4 x double>*
+; UNROLL-NEXT:    store <4 x double> [[TMP21]], <4 x double>* [[TMP24]], align 8, !alias.scope !14, !noalias !11
+; UNROLL-NEXT:    [[TMP25:%.*]] = getelementptr inbounds double, double* [[TMP23]], i64 4
+; UNROLL-NEXT:    [[TMP26:%.*]] = bitcast double* [[TMP25]] to <4 x double>*
+; UNROLL-NEXT:    store <4 x double> [[TMP22]], <4 x double>* [[TMP26]], align 8, !alias.scope !14, !noalias !11
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; UNROLL:       middle.block:
-; UNROLL:         %vector.recur.extract = extractelement <4 x i16> [[L2]], i32 3
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD8]], i32 3
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[VECTOR_MEMCHECK]] ], [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[VECTOR_MEMCHECK]] ], [ 1, [[FOR_PREHEADER]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL:       scalar.body:
+; UNROLL-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
+; UNROLL-NEXT:    [[TMP28]] = load i16, i16* [[ARRAYIDX5]], align 2
+; UNROLL-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP28]] to double
+; UNROLL-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
+; UNROLL-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
+; UNROLL-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
+; UNROLL-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
+; UNROLL-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
+; UNROLL-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
+; UNROLL-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; UNROLL:       for.end.loopexit:
+; UNROLL-NEXT:    br label [[FOR_END]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @recurrence_3(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NO-IC-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
+; UNROLL-NO-IC-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
+; UNROLL-NO-IC-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
+; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
+; UNROLL-NO-IC-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
+; UNROLL-NO-IC-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
+; UNROLL-NO-IC-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; UNROLL-NO-IC:       for.preheader:
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-IC:       vector.memcheck:
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP:%.*]] = getelementptr double, double* [[B]], i64 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP1:%.*]] = bitcast double* [[SCEVGEP]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i32 [[N]], -2
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP6]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP23:%.*]] = bitcast double* [[SCEVGEP2]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP45:%.*]] = bitcast i16* [[SCEVGEP4]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP6:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP6]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP67:%.*]] = bitcast i16* [[SCEVGEP6]] to i8*
+; UNROLL-NO-IC-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP67]]
+; UNROLL-NO-IC-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP45]], [[SCEVGEP23]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-IC-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = add i64 1, [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i32 3
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT9:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT10:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT9]], <4 x double> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD8:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i64 [[OFFSET_IDX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i16, i16* [[TMP9]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = bitcast i16* [[TMP11]] to <4 x i16>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP12]], align 2, !alias.scope !11
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, i16* [[TMP9]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = bitcast i16* [[TMP13]] to <4 x i16>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD8]] = load <4 x i16>, <4 x i16>* [[TMP14]], align 2, !alias.scope !11
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD8]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = sitofp <4 x i16> [[WIDE_LOAD]] to <4 x double>
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = sitofp <4 x i16> [[WIDE_LOAD8]] to <4 x double>
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = sitofp <4 x i16> [[TMP15]] to <4 x double>
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = sitofp <4 x i16> [[TMP16]] to <4 x double>
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = fmul fast <4 x double> [[TMP19]], [[BROADCAST_SPLAT]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = fmul fast <4 x double> [[TMP20]], [[BROADCAST_SPLAT10]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = fsub fast <4 x double> [[TMP17]], [[TMP21]]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = fsub fast <4 x double> [[TMP18]], [[TMP22]]
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = getelementptr inbounds double, double* [[TMP25]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = bitcast double* [[TMP27]] to <4 x double>*
+; UNROLL-NO-IC-NEXT:    store <4 x double> [[TMP23]], <4 x double>* [[TMP28]], align 8, !alias.scope !14, !noalias !11
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = getelementptr inbounds double, double* [[TMP25]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = bitcast double* [[TMP29]] to <4 x double>*
+; UNROLL-NO-IC-NEXT:    store <4 x double> [[TMP24]], <4 x double>* [[TMP30]], align 8, !alias.scope !14, !noalias !11
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD8]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD8]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[VECTOR_MEMCHECK]] ], [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[FOR_PREHEADER]] ], [ 1, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-IC:       scalar.body:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP32:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
+; UNROLL-NO-IC-NEXT:    [[TMP32]] = load i16, i16* [[ARRAYIDX5]], align 2
+; UNROLL-NO-IC-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP32]] to double
+; UNROLL-NO-IC-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
+; UNROLL-NO-IC-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
+; UNROLL-NO-IC-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
+; UNROLL-NO-IC-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
+; UNROLL-NO-IC-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; UNROLL-NO-IC:       for.end.loopexit:
+; UNROLL-NO-IC-NEXT:    br label [[FOR_END]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @recurrence_3(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NO-VF-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
+; UNROLL-NO-VF-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
+; UNROLL-NO-VF-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
+; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
+; UNROLL-NO-VF-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
+; UNROLL-NO-VF-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
+; UNROLL-NO-VF-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; UNROLL-NO-VF:       for.preheader:
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-VF:       vector.memcheck:
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP:%.*]] = getelementptr double, double* [[B]], i64 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP1:%.*]] = bitcast double* [[SCEVGEP]] to i8*
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = add i32 [[N]], -2
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP6]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP23:%.*]] = bitcast double* [[SCEVGEP2]] to i8*
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP45:%.*]] = bitcast i16* [[SCEVGEP4]] to i8*
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP6:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP6]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP67:%.*]] = bitcast i16* [[SCEVGEP6]] to i8*
+; UNROLL-NO-VF-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP67]]
+; UNROLL-NO-VF-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP45]], [[SCEVGEP23]]
+; UNROLL-NO-VF-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-VF-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = add i64 1, [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[OFFSET_IDX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION8:%.*]] = add i64 [[OFFSET_IDX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDUCTION8]]
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = load i16, i16* [[TMP7]], align 2, !alias.scope !10
+; UNROLL-NO-VF-NEXT:    [[TMP10]] = load i16, i16* [[TMP8]], align 2, !alias.scope !10
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = sitofp i16 [[TMP9]] to double
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = sitofp i16 [[TMP10]] to double
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = sitofp i16 [[VECTOR_RECUR]] to double
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = sitofp i16 [[TMP9]] to double
+; UNROLL-NO-VF-NEXT:    [[TMP15:%.*]] = fmul fast double [[TMP13]], [[CONV1]]
+; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = fmul fast double [[TMP14]], [[CONV1]]
+; UNROLL-NO-VF-NEXT:    [[TMP17:%.*]] = fsub fast double [[TMP11]], [[TMP15]]
+; UNROLL-NO-VF-NEXT:    [[TMP18:%.*]] = fsub fast double [[TMP12]], [[TMP16]]
+; UNROLL-NO-VF-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP20:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[INDUCTION8]]
+; UNROLL-NO-VF-NEXT:    store double [[TMP17]], double* [[TMP19]], align 8, !alias.scope !13, !noalias !10
+; UNROLL-NO-VF-NEXT:    store double [[TMP18]], double* [[TMP20]], align 8, !alias.scope !13, !noalias !10
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[VECTOR_MEMCHECK]] ], [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[TMP10]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[FOR_PREHEADER]] ], [ 1, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-VF:       scalar.body:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP22:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
+; UNROLL-NO-VF-NEXT:    [[TMP22]] = load i16, i16* [[ARRAYIDX5]], align 2
+; UNROLL-NO-VF-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP22]] to double
+; UNROLL-NO-VF-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
+; UNROLL-NO-VF-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
+; UNROLL-NO-VF-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
+; UNROLL-NO-VF-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
+; UNROLL-NO-VF-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
+; UNROLL-NO-VF-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; UNROLL-NO-VF:       for.end.loopexit:
+; UNROLL-NO-VF-NEXT:    br label [[FOR_END]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @recurrence_3(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; SINK-AFTER-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
+; SINK-AFTER-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
+; SINK-AFTER-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
+; SINK-AFTER-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
+; SINK-AFTER-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
+; SINK-AFTER-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
+; SINK-AFTER-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
+; SINK-AFTER-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; SINK-AFTER:       for.preheader:
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; SINK-AFTER:       vector.memcheck:
+; SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr double, double* [[B]], i64 1
+; SINK-AFTER-NEXT:    [[SCEVGEP1:%.*]] = bitcast double* [[SCEVGEP]] to i8*
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i32 [[N]], -2
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP6]]
+; SINK-AFTER-NEXT:    [[SCEVGEP23:%.*]] = bitcast double* [[SCEVGEP2]] to i8*
+; SINK-AFTER-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; SINK-AFTER-NEXT:    [[SCEVGEP45:%.*]] = bitcast i16* [[SCEVGEP4]] to i8*
+; SINK-AFTER-NEXT:    [[SCEVGEP6:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP6]]
+; SINK-AFTER-NEXT:    [[SCEVGEP67:%.*]] = bitcast i16* [[SCEVGEP6]] to i8*
+; SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP67]]
+; SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP45]], [[SCEVGEP23]]
+; SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; SINK-AFTER-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[IND_END:%.*]] = add i64 1, [[N_VEC]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i32 3
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], 0
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP7]]
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, i16* [[TMP8]], i32 0
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = bitcast i16* [[TMP9]] to <4 x i16>*
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP10]], align 2, !alias.scope !11
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = sitofp <4 x i16> [[WIDE_LOAD]] to <4 x double>
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = sitofp <4 x i16> [[TMP11]] to <4 x double>
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = fmul fast <4 x double> [[TMP13]], [[BROADCAST_SPLAT]]
+; SINK-AFTER-NEXT:    [[TMP15:%.*]] = fsub fast <4 x double> [[TMP12]], [[TMP14]]
+; SINK-AFTER-NEXT:    [[TMP16:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[TMP7]]
+; SINK-AFTER-NEXT:    [[TMP17:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 0
+; SINK-AFTER-NEXT:    [[TMP18:%.*]] = bitcast double* [[TMP17]] to <4 x double>*
+; SINK-AFTER-NEXT:    store <4 x double> [[TMP15]], <4 x double>* [[TMP18]], align 8, !alias.scope !14, !noalias !11
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[VECTOR_MEMCHECK]] ], [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[FOR_PREHEADER]] ], [ 1, [[VECTOR_MEMCHECK]] ]
+; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; SINK-AFTER:       scalar.body:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP20:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
+; SINK-AFTER-NEXT:    [[TMP20]] = load i16, i16* [[ARRAYIDX5]], align 2
+; SINK-AFTER-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP20]] to double
+; SINK-AFTER-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
+; SINK-AFTER-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
+; SINK-AFTER-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
+; SINK-AFTER-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
+; SINK-AFTER-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
+; SINK-AFTER-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
+; SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; SINK-AFTER:       for.end.loopexit:
+; SINK-AFTER-NEXT:    br label [[FOR_END]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @recurrence_3(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; NO-SINK-AFTER-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
+; NO-SINK-AFTER-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
+; NO-SINK-AFTER-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
+; NO-SINK-AFTER-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
+; NO-SINK-AFTER-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
+; NO-SINK-AFTER-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
+; NO-SINK-AFTER-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; NO-SINK-AFTER:       for.preheader:
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; NO-SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 4
+; NO-SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; NO-SINK-AFTER:       vector.memcheck:
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr double, double* [[B]], i64 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP1:%.*]] = bitcast double* [[SCEVGEP]] to i8*
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i32 [[N]], -2
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP6]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP23:%.*]] = bitcast double* [[SCEVGEP2]] to i8*
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP45:%.*]] = bitcast i16* [[SCEVGEP4]] to i8*
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP6:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP6]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP67:%.*]] = bitcast i16* [[SCEVGEP6]] to i8*
+; NO-SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP67]]
+; NO-SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP45]], [[SCEVGEP23]]
+; NO-SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-SINK-AFTER-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[IND_END:%.*]] = add i64 1, [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i32 3
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP7]]
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, i16* [[TMP8]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = bitcast i16* [[TMP9]] to <4 x i16>*
+; NO-SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP10]], align 2, !alias.scope !11
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = sitofp <4 x i16> [[WIDE_LOAD]] to <4 x double>
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = sitofp <4 x i16> [[TMP11]] to <4 x double>
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = fmul fast <4 x double> [[TMP13]], [[BROADCAST_SPLAT]]
+; NO-SINK-AFTER-NEXT:    [[TMP15:%.*]] = fsub fast <4 x double> [[TMP12]], [[TMP14]]
+; NO-SINK-AFTER-NEXT:    [[TMP16:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[TMP7]]
+; NO-SINK-AFTER-NEXT:    [[TMP17:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP18:%.*]] = bitcast double* [[TMP17]] to <4 x double>*
+; NO-SINK-AFTER-NEXT:    store <4 x double> [[TMP15]], <4 x double>* [[TMP18]], align 8, !alias.scope !14, !noalias !11
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[VECTOR_MEMCHECK]] ], [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[FOR_PREHEADER]] ], [ 1, [[VECTOR_MEMCHECK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; NO-SINK-AFTER:       scalar.body:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP20:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
+; NO-SINK-AFTER-NEXT:    [[TMP20]] = load i16, i16* [[ARRAYIDX5]], align 2
+; NO-SINK-AFTER-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP20]] to double
+; NO-SINK-AFTER-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
+; NO-SINK-AFTER-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
+; NO-SINK-AFTER-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
+; NO-SINK-AFTER-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
+; NO-SINK-AFTER-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
+; NO-SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; NO-SINK-AFTER:       for.end.loopexit:
+; NO-SINK-AFTER-NEXT:    br label [[FOR_END]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
 ;
-define void @recurrence_3(i16* nocapture readonly %a, double* nocapture %b, i32 %n, float %f, i16 %p) {
 entry:
   %0 = load i16, i16* %a, align 2
   %conv = sitofp i16 %0 to double
@@ -207,11 +1573,190 @@ for.end:
 ;   }
 ; }
 ;
-; CHECK-LABEL: @PR26734(
-; CHECK-NOT:   vector.ph:
-; CHECK:       }
 ;
 define void @PR26734(i16* %a, i32* %b, i32* %c, i32 %d, i16* %e) {
+; CHECK-LABEL: @PR26734(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
+; CHECK-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
+; CHECK:       entry.for.end_crit_edge:
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body.lr.ph:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; CHECK-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
+; CHECK-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
+; CHECK-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
+; CHECK-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B:%.*]], align 4
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
+; CHECK-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
+; CHECK:       for.cond.for.end_crit_edge:
+; CHECK-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
+; CHECK-NEXT:    store i32 [[AND]], i32* [[B]], align 4
+; CHECK-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
+; CHECK-NEXT:    br label [[FOR_END]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @PR26734(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
+; UNROLL-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
+; UNROLL:       entry.for.end_crit_edge:
+; UNROLL-NEXT:    br label [[FOR_END:%.*]]
+; UNROLL:       for.body.lr.ph:
+; UNROLL-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
+; UNROLL-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
+; UNROLL-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
+; UNROLL-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B:%.*]], align 4
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
+; UNROLL-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
+; UNROLL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
+; UNROLL-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
+; UNROLL:       for.cond.for.end_crit_edge:
+; UNROLL-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
+; UNROLL-NEXT:    store i32 [[AND]], i32* [[B]], align 4
+; UNROLL-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
+; UNROLL-NEXT:    br label [[FOR_END]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @PR26734(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
+; UNROLL-NO-IC:       entry.for.end_crit_edge:
+; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[B:%.*]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[FOR_END:%.*]]
+; UNROLL-NO-IC:       for.body.lr.ph:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NO-IC-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
+; UNROLL-NO-IC-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
+; UNROLL-NO-IC-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
+; UNROLL-NO-IC-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
+; UNROLL-NO-IC-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
+; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
+; UNROLL-NO-IC:       for.cond.for.end_crit_edge:
+; UNROLL-NO-IC-NEXT:    [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
+; UNROLL-NO-IC-NEXT:    store i32 [[AND_LCSSA]], i32* [[B]], align 4
+; UNROLL-NO-IC-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
+; UNROLL-NO-IC-NEXT:    br label [[FOR_END]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @PR26734(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
+; UNROLL-NO-VF:       entry.for.end_crit_edge:
+; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[B:%.*]], align 4
+; UNROLL-NO-VF-NEXT:    br label [[FOR_END:%.*]]
+; UNROLL-NO-VF:       for.body.lr.ph:
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NO-VF-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
+; UNROLL-NO-VF-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
+; UNROLL-NO-VF-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
+; UNROLL-NO-VF-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B]], align 4
+; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-VF:       for.body:
+; UNROLL-NO-VF-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
+; UNROLL-NO-VF-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
+; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
+; UNROLL-NO-VF:       for.cond.for.end_crit_edge:
+; UNROLL-NO-VF-NEXT:    [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
+; UNROLL-NO-VF-NEXT:    store i32 [[AND_LCSSA]], i32* [[B]], align 4
+; UNROLL-NO-VF-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
+; UNROLL-NO-VF-NEXT:    br label [[FOR_END]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @PR26734(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
+; SINK-AFTER-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
+; SINK-AFTER:       entry.for.end_crit_edge:
+; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[B:%.*]], align 4
+; SINK-AFTER-NEXT:    br label [[FOR_END:%.*]]
+; SINK-AFTER:       for.body.lr.ph:
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; SINK-AFTER-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
+; SINK-AFTER-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
+; SINK-AFTER-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
+; SINK-AFTER-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B]], align 4
+; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; SINK-AFTER:       for.body:
+; SINK-AFTER-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
+; SINK-AFTER-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
+; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
+; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
+; SINK-AFTER:       for.cond.for.end_crit_edge:
+; SINK-AFTER-NEXT:    [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
+; SINK-AFTER-NEXT:    store i32 [[AND_LCSSA]], i32* [[B]], align 4
+; SINK-AFTER-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
+; SINK-AFTER-NEXT:    br label [[FOR_END]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @PR26734(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
+; NO-SINK-AFTER:       entry.for.end_crit_edge:
+; NO-SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[B:%.*]], align 4
+; NO-SINK-AFTER-NEXT:    br label [[FOR_END:%.*]]
+; NO-SINK-AFTER:       for.body.lr.ph:
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
+; NO-SINK-AFTER-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
+; NO-SINK-AFTER-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
+; NO-SINK-AFTER-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
+; NO-SINK-AFTER-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B]], align 4
+; NO-SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-SINK-AFTER:       for.body:
+; NO-SINK-AFTER-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
+; NO-SINK-AFTER-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
+; NO-SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
+; NO-SINK-AFTER:       for.cond.for.end_crit_edge:
+; NO-SINK-AFTER-NEXT:    [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
+; NO-SINK-AFTER-NEXT:    store i32 [[AND_LCSSA]], i32* [[B]], align 4
+; NO-SINK-AFTER-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
+; NO-SINK-AFTER-NEXT:    br label [[FOR_END]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
+;
 entry:
   %cmp4 = icmp eq i32 %d, 21
   br i1 %cmp4, label %entry.for.end_crit_edge, label %for.body.lr.ph
@@ -258,11 +1803,136 @@ for.end:
 ;   return n;
 ; }
 ;
-; CHECK-LABEL: @PR27246(
-; CHECK-NOT:   vector.ph:
-; CHECK:       }
 ;
 define i32 @PR27246() {
+; CHECK-LABEL: @PR27246(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
+; CHECK:       for.cond1.preheader:
+; CHECK-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
+; CHECK-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1:%.*]], [[FOR_COND_CLEANUP3]] ]
+; CHECK-NEXT:    br label [[FOR_COND1:%.*]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    ret i32 [[E_1]]
+; CHECK:       for.cond1:
+; CHECK-NEXT:    [[E_1]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
+; CHECK-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
+; CHECK-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
+; CHECK-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
+; CHECK:       for.cond.cleanup3:
+; CHECK-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
+;
+; UNROLL-LABEL: @PR27246(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
+; UNROLL:       for.cond1.preheader:
+; UNROLL-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
+; UNROLL-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1:%.*]], [[FOR_COND_CLEANUP3]] ]
+; UNROLL-NEXT:    br label [[FOR_COND1:%.*]]
+; UNROLL:       for.cond.cleanup:
+; UNROLL-NEXT:    ret i32 [[E_1]]
+; UNROLL:       for.cond1:
+; UNROLL-NEXT:    [[E_1]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
+; UNROLL-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
+; UNROLL-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
+; UNROLL-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
+; UNROLL-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
+; UNROLL:       for.cond.cleanup3:
+; UNROLL-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
+;
+; UNROLL-NO-IC-LABEL: @PR27246(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
+; UNROLL-NO-IC:       for.cond1.preheader:
+; UNROLL-NO-IC-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1_LCSSA:%.*]], [[FOR_COND_CLEANUP3]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_COND1:%.*]]
+; UNROLL-NO-IC:       for.cond.cleanup:
+; UNROLL-NO-IC-NEXT:    [[E_1_LCSSA_LCSSA:%.*]] = phi i32 [ [[E_1_LCSSA]], [[FOR_COND_CLEANUP3]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[E_1_LCSSA_LCSSA]]
+; UNROLL-NO-IC:       for.cond1:
+; UNROLL-NO-IC-NEXT:    [[E_1:%.*]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
+; UNROLL-NO-IC-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
+; UNROLL-NO-IC-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
+; UNROLL-NO-IC-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
+; UNROLL-NO-IC:       for.cond.cleanup3:
+; UNROLL-NO-IC-NEXT:    [[E_1_LCSSA]] = phi i32 [ [[E_1]], [[FOR_COND1]] ]
+; UNROLL-NO-IC-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
+;
+; UNROLL-NO-VF-LABEL: @PR27246(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
+; UNROLL-NO-VF:       for.cond1.preheader:
+; UNROLL-NO-VF-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1_LCSSA:%.*]], [[FOR_COND_CLEANUP3]] ]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_COND1:%.*]]
+; UNROLL-NO-VF:       for.cond.cleanup:
+; UNROLL-NO-VF-NEXT:    [[E_1_LCSSA_LCSSA:%.*]] = phi i32 [ [[E_1_LCSSA]], [[FOR_COND_CLEANUP3]] ]
+; UNROLL-NO-VF-NEXT:    ret i32 [[E_1_LCSSA_LCSSA]]
+; UNROLL-NO-VF:       for.cond1:
+; UNROLL-NO-VF-NEXT:    [[E_1:%.*]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
+; UNROLL-NO-VF-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
+; UNROLL-NO-VF-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
+; UNROLL-NO-VF-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
+; UNROLL-NO-VF:       for.cond.cleanup3:
+; UNROLL-NO-VF-NEXT:    [[E_1_LCSSA]] = phi i32 [ [[E_1]], [[FOR_COND1]] ]
+; UNROLL-NO-VF-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
+;
+; SINK-AFTER-LABEL: @PR27246(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
+; SINK-AFTER:       for.cond1.preheader:
+; SINK-AFTER-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
+; SINK-AFTER-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1_LCSSA:%.*]], [[FOR_COND_CLEANUP3]] ]
+; SINK-AFTER-NEXT:    br label [[FOR_COND1:%.*]]
+; SINK-AFTER:       for.cond.cleanup:
+; SINK-AFTER-NEXT:    [[E_1_LCSSA_LCSSA:%.*]] = phi i32 [ [[E_1_LCSSA]], [[FOR_COND_CLEANUP3]] ]
+; SINK-AFTER-NEXT:    ret i32 [[E_1_LCSSA_LCSSA]]
+; SINK-AFTER:       for.cond1:
+; SINK-AFTER-NEXT:    [[E_1:%.*]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
+; SINK-AFTER-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
+; SINK-AFTER-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
+; SINK-AFTER-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
+; SINK-AFTER-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
+; SINK-AFTER:       for.cond.cleanup3:
+; SINK-AFTER-NEXT:    [[E_1_LCSSA]] = phi i32 [ [[E_1]], [[FOR_COND1]] ]
+; SINK-AFTER-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
+;
+; NO-SINK-AFTER-LABEL: @PR27246(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
+; NO-SINK-AFTER:       for.cond1.preheader:
+; NO-SINK-AFTER-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
+; NO-SINK-AFTER-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1_LCSSA:%.*]], [[FOR_COND_CLEANUP3]] ]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_COND1:%.*]]
+; NO-SINK-AFTER:       for.cond.cleanup:
+; NO-SINK-AFTER-NEXT:    [[E_1_LCSSA_LCSSA:%.*]] = phi i32 [ [[E_1_LCSSA]], [[FOR_COND_CLEANUP3]] ]
+; NO-SINK-AFTER-NEXT:    ret i32 [[E_1_LCSSA_LCSSA]]
+; NO-SINK-AFTER:       for.cond1:
+; NO-SINK-AFTER-NEXT:    [[E_1:%.*]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
+; NO-SINK-AFTER-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
+; NO-SINK-AFTER-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
+; NO-SINK-AFTER-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
+; NO-SINK-AFTER:       for.cond.cleanup3:
+; NO-SINK-AFTER-NEXT:    [[E_1_LCSSA]] = phi i32 [ [[E_1]], [[FOR_COND1]] ]
+; NO-SINK-AFTER-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
+;
 entry:
   br label %for.cond1.preheader
 
@@ -289,44 +1959,326 @@ for.cond.cleanup3:
   br i1 %exitcond, label %for.cond.cleanup, label %for.cond1.preheader
 }
 
+;
+define void @PR30183(i32 %pre_load, i32* %a, i32* %b, i64 %n) {
+; CHECK-LABEL: @PR30183(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 6
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], -4
+; CHECK-NEXT:    [[IND_END:%.*]] = shl i64 [[N_VEC]], 1
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[SCALAR_BODY:%.*]]
+; CHECK:       scalar.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
+; CHECK-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @PR30183(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
+; UNROLL-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
+; UNROLL-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 14
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], -8
+; UNROLL-NEXT:    [[IND_END:%.*]] = shl i64 [[N_VEC]], 1
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL:       scalar.body:
+; UNROLL-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
+; UNROLL-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
 ; UNROLL-NO-IC-LABEL: @PR30183(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-IC:       vector.ph:
-; UNROLL-NO-IC:         [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD:%.*]], i32 3
-; UNROLL-NO-IC-NEXT:    br label %vector.body
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD:%.*]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-IC:       vector.body:
-; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
-; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], %vector.ph ], [ [[TMP42:%.*]], %vector.body ]
-; UNROLL-NO-IC:         [[TMP27:%.*]] = load i32, i32* {{.*}}
-; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = load i32, i32* {{.*}}
-; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = load i32, i32* {{.*}}
-; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = load i32, i32* {{.*}}
-; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = insertelement <4 x i32> poison, i32 [[TMP27]], i32 0
-; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = insertelement <4 x i32> [[TMP35]], i32 [[TMP28]], i32 1
-; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = insertelement <4 x i32> [[TMP36]], i32 [[TMP29]], i32 2
-; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = insertelement <4 x i32> [[TMP37]], i32 [[TMP30]], i32 3
-; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = load i32, i32* {{.*}}
-; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = load i32, i32* {{.*}}
-; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = load i32, i32* {{.*}}
-; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = load i32, i32* {{.*}}
-; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = insertelement <4 x i32> poison, i32 [[TMP31]], i32 0
-; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = insertelement <4 x i32> [[TMP39]], i32 [[TMP32]], i32 1
-; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = insertelement <4 x i32> [[TMP40]], i32 [[TMP33]], i32 2
-; UNROLL-NO-IC-NEXT:    [[TMP42]] = insertelement <4 x i32> [[TMP41]], i32 [[TMP34]], i32 3
-; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP38]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[TMP38]], <4 x i32> [[TMP42]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP42:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 6
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i64 [[OFFSET_IDX]], 10
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add i64 [[OFFSET_IDX]], 12
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add i64 [[OFFSET_IDX]], 14
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = add nuw nsw i64 [[TMP3]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = add nuw nsw i64 [[TMP6]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = add nuw nsw i64 [[TMP7]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = add nuw nsw i64 [[TMP8]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = add nuw nsw i64 [[TMP9]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = add nuw nsw i64 [[TMP10]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP11]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP12]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP13]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP14]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP15]]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP16]]
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP17]]
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP19]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP20]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP21]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP22]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = insertelement <4 x i32> poison, i32 [[TMP27]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = insertelement <4 x i32> [[TMP31]], i32 [[TMP28]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = insertelement <4 x i32> [[TMP32]], i32 [[TMP29]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = insertelement <4 x i32> [[TMP33]], i32 [[TMP30]], i32 3
+; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP23]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP24]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP25]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP26]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = insertelement <4 x i32> poison, i32 [[TMP35]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = insertelement <4 x i32> [[TMP39]], i32 [[TMP36]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = insertelement <4 x i32> [[TMP40]], i32 [[TMP37]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP42]] = insertelement <4 x i32> [[TMP41]], i32 [[TMP38]], i32 3
+; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP34]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[TMP34]], <4 x i32> [[TMP42]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; UNROLL-NO-IC:         br i1 {{.*}}, label %middle.block, label %vector.body
+; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP45]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP42]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP42]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-IC:       scalar.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR2:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
+; UNROLL-NO-IC-NEXT:    [[VAR1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I_NEXT]]
+; UNROLL-NO-IC-NEXT:    [[VAR2]] = load i32, i32* [[VAR1]], align 4
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @PR30183(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ [[PRE_LOAD:%.*]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[OFFSET_IDX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[OFFSET_IDX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[INDUCTION]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[INDUCTION1]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP3]]
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 4
+; UNROLL-NO-VF-NEXT:    [[TMP8]] = load i32, i32* [[TMP6]], align 4
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[ENTRY:%.*]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-VF:       scalar.body:
+; UNROLL-NO-VF-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR2:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
+; UNROLL-NO-VF-NEXT:    [[VAR1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I_NEXT]]
+; UNROLL-NO-VF-NEXT:    [[VAR2]] = load i32, i32* [[VAR1]], align 4
+; UNROLL-NO-VF-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
+; UNROLL-NO-VF-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @PR30183(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD:%.*]], i32 3
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 2
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 4
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 6
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[TMP3]], 2
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = add nuw nsw i64 [[TMP6]], 2
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP7]]
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP9]]
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP10]]
+; SINK-AFTER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP11]], align 4
+; SINK-AFTER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP12]], align 4
+; SINK-AFTER-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP13]], align 4
+; SINK-AFTER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP14]], align 4
+; SINK-AFTER-NEXT:    [[TMP19:%.*]] = insertelement <4 x i32> poison, i32 [[TMP15]], i32 0
+; SINK-AFTER-NEXT:    [[TMP20:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP16]], i32 1
+; SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP20]], i32 [[TMP17]], i32 2
+; SINK-AFTER-NEXT:    [[TMP22]] = insertelement <4 x i32> [[TMP21]], i32 [[TMP18]], i32 3
+; SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; SINK-AFTER:       scalar.body:
+; SINK-AFTER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR2:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
+; SINK-AFTER-NEXT:    [[VAR1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I_NEXT]]
+; SINK-AFTER-NEXT:    [[VAR2]] = load i32, i32* [[VAR1]], align 4
+; SINK-AFTER-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
+; SINK-AFTER-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @PR30183(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
+; NO-SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; NO-SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD:%.*]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 6
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[TMP3]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[TMP4]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = add nuw nsw i64 [[TMP5]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = add nuw nsw i64 [[TMP6]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP7]]
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP9]]
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP10]]
+; NO-SINK-AFTER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP11]], align 4
+; NO-SINK-AFTER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP12]], align 4
+; NO-SINK-AFTER-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP13]], align 4
+; NO-SINK-AFTER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP14]], align 4
+; NO-SINK-AFTER-NEXT:    [[TMP19:%.*]] = insertelement <4 x i32> poison, i32 [[TMP15]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP20:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP16]], i32 1
+; NO-SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP20]], i32 [[TMP17]], i32 2
+; NO-SINK-AFTER-NEXT:    [[TMP22]] = insertelement <4 x i32> [[TMP21]], i32 [[TMP18]], i32 3
+; NO-SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; NO-SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; NO-SINK-AFTER:       scalar.body:
+; NO-SINK-AFTER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR2:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
+; NO-SINK-AFTER-NEXT:    [[VAR1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I_NEXT]]
+; NO-SINK-AFTER-NEXT:    [[VAR2]] = load i32, i32* [[VAR1]], align 4
+; NO-SINK-AFTER-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
+; NO-SINK-AFTER-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
 ;
-define void @PR30183(i32 %pre_load, i32* %a, i32* %b, i64 %n) {
 entry:
   br label %scalar.body
 
 scalar.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %scalar.body ]
-  %tmp0 = phi i32 [ %pre_load, %entry ], [ %tmp2, %scalar.body ]
+  %var0 = phi i32 [ %pre_load, %entry ], [ %var2, %scalar.body ]
   %i.next = add nuw nsw i64 %i, 2
-  %tmp1 = getelementptr inbounds i32, i32* %a, i64 %i.next
-  %tmp2 = load i32, i32* %tmp1
+  %var1 = getelementptr inbounds i32, i32* %a, i64 %i.next
+  %var2 = load i32, i32* %var1
   %cond = icmp eq i64 %i.next,%n
   br i1 %cond, label %for.end, label %scalar.body
 
@@ -334,21 +2286,170 @@ for.end:
   ret void
 }
 
+;
+define void @constant_folded_previous_value() {
+; CHECK-LABEL: @constant_folded_previous_value(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    br i1 undef, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 undef, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    br label [[SCALAR_BODY:%.*]]
+; CHECK:       scalar.body:
+; CHECK-NEXT:    br i1 undef, label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @constant_folded_previous_value(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    br i1 undef, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    br i1 undef, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL:       scalar.body:
+; UNROLL-NEXT:    br i1 undef, label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
 ; UNROLL-NO-IC-LABEL: @constant_folded_previous_value(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-IC:       vector.body:
-; UNROLL-NO-IC:         [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, %vector.ph ], [ <i64 1, i64 1, i64 1, i64 1>, %vector.body ]
-; UNROLL-NO-IC:         [[TMP0:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> <i64 1, i64 1, i64 1, i64 1>, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NO-IC-NEXT:     add nuw i64
-; UNROLL-NO-IC:         br i1 {{.*}}, label %middle.block, label %vector.body
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, [[VECTOR_PH]] ], [ <i64 1, i64 1, i64 1, i64 1>, [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> <i64 1, i64 1, i64 1, i64 1>, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 1, [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-IC:       scalar.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR3:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR3]] = add i64 0, 1
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], undef
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @constant_folded_previous_value(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i64 0, 1
+; UNROLL-NO-VF-NEXT:    [[TMP1]] = add i64 0, 1
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP1]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
+; UNROLL-NO-VF:       scalar.body:
+; UNROLL-NO-VF-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR3:%.*]], [[SCALAR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VAR3]] = add i64 0, 1
+; UNROLL-NO-VF-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NO-VF-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], undef
+; UNROLL-NO-VF-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @constant_folded_previous_value(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, [[VECTOR_PH]] ], [ <i64 1, i64 1, i64 1, i64 1>, [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> <i64 1, i64 1, i64 1, i64 1>, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
+; SINK-AFTER-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 1, [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; SINK-AFTER:       scalar.body:
+; SINK-AFTER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR3:%.*]], [[SCALAR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VAR3]] = add i64 0, 1
+; SINK-AFTER-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; SINK-AFTER-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], undef
+; SINK-AFTER-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @constant_folded_previous_value(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, [[VECTOR_PH]] ], [ <i64 1, i64 1, i64 1, i64 1>, [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> <i64 1, i64 1, i64 1, i64 1>, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 1, [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; NO-SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
+; NO-SINK-AFTER:       scalar.body:
+; NO-SINK-AFTER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR3:%.*]], [[SCALAR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VAR3]] = add i64 0, 1
+; NO-SINK-AFTER-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; NO-SINK-AFTER-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], undef
+; NO-SINK-AFTER-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
 ;
-define void @constant_folded_previous_value() {
 entry:
   br label %scalar.body
 
 scalar.body:
   %i = phi i64 [ 0, %entry ], [ %i.next, %scalar.body ]
-  %tmp2 = phi i64 [ 0, %entry ], [ %tmp3, %scalar.body ]
-  %tmp3 = add i64 0, 1
+  %var2 = phi i64 [ 0, %entry ], [ %var3, %scalar.body ]
+  %var3 = add i64 0, 1
   %i.next = add nuw nsw i64 %i, 1
   %cond = icmp eq i64 %i.next, undef
   br i1 %cond, label %for.end, label %scalar.body
@@ -362,31 +2463,230 @@ for.end:
 ; another at the second last index. We need these 2 extracts because
 ; the first order recurrence phi is used outside the loop, so we require the phi
 ; itself and not its update (addx).
-; UNROLL-NO-IC-LABEL: extract_second_last_iteration
-; UNROLL-NO-IC: vector.body
-; UNROLL-NO-IC:   %step.add = add <4 x i32> %vec.ind, <i32 4, i32 4, i32 4, i32 4>
-; UNROLL-NO-IC:   %[[L1:.+]] = add <4 x i32> %vec.ind, %broadcast.splat
-; UNROLL-NO-IC:   %[[L2:.+]] = add <4 x i32> %step.add, %broadcast.splat
-; UNROLL-NO-IC:   %index.next = add nuw i32 %index, 8
-; UNROLL-NO-IC:   icmp eq i32 %index.next, 96
-; UNROLL-NO-IC: middle.block
-; UNROLL-NO-IC:   icmp eq i32 96, 96
-; UNROLL-NO-IC:   %vector.recur.extract = extractelement <4 x i32> %[[L2]], i32 3
-; UNROLL-NO-IC:   %vector.recur.extract.for.phi = extractelement <4 x i32> %[[L2]], i32 2
-; UNROLL-NO-IC: for.end
-; UNROLL-NO-IC:   %val.phi.lcssa = phi i32 [ %scalar.recur, %for.body ], [ %vector.recur.extract.for.phi, %middle.block ]
 ; Check the case when unrolled but not vectorized.
-; UNROLL-NO-VF-LABEL: extract_second_last_iteration
-; UNROLL-NO-VF: vector.body:
-; UNROLL-NO-VF:   %induction = add i32 %index, 0
-; UNROLL-NO-VF:   %induction1 = add i32 %index, 1
-; UNROLL-NO-VF:   %[[L1:.+]] = add i32 %induction, %x
-; UNROLL-NO-VF:   %[[L2:.+]] = add i32 %induction1, %x
-; UNROLL-NO-VF:   %index.next = add nuw i32 %index, 2
-; UNROLL-NO-VF:   icmp eq i32 %index.next, 96
-; UNROLL-NO-VF: for.end:
-; UNROLL-NO-VF:   %val.phi.lcssa = phi i32 [ %scalar.recur, %for.body ], [ %[[L1]], %middle.block ]
 define i32 @extract_second_last_iteration(i32* %cval, i32 %x)  {
+; CHECK-LABEL: @extract_second_last_iteration(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ 2, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP1]] = add i32 [[TMP0]], 4
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
+; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = add i32 [[TMP0]], [[X:%.*]]
+; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    br i1 undef, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ undef, [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
+;
+; UNROLL-LABEL: @extract_second_last_iteration(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP0:%.*]] = phi i32 [ 2, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP1]] = add i32 [[TMP0]], 8
+; UNROLL-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
+; UNROLL-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[TMP3:%.*]] = add i32 [[TMP0]], 4
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = add i32 [[TMP3]], [[X:%.*]]
+; UNROLL-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    br i1 undef, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ undef, [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
+;
+; UNROLL-NO-IC-LABEL: @extract_second_last_iteration(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X:%.*]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[X]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT2]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i32 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i32 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i32 [[INDEX]], 5
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i32 [[INDEX]], 6
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i32 [[INDEX]], 7
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = zext <4 x i32> [[VEC_IND]] to <4 x i64>
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = zext <4 x i32> [[STEP_ADD]] to <4 x i64>
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; UNROLL-NO-IC-NEXT:    [[TMP11]] = add <4 x i32> [[STEP_ADD]], [[BROADCAST_SPLAT3]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP10]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], <i32 4, i32 4, i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, 96
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP11]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP11]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[ADDX:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
+; UNROLL-NO-IC-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
+; UNROLL-NO-IC-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
+; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[SCALAR_RECUR]], [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
+;
+; UNROLL-NO-VF-LABEL: @extract_second_last_iteration(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i32 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = zext i32 [[INDUCTION]] to i64
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = zext i32 [[INDUCTION1]] to i64
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i32 [[INDUCTION]], [[X:%.*]]
+; UNROLL-NO-VF-NEXT:    [[TMP3]] = add i32 [[INDUCTION1]], [[X]]
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, 96
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-VF:       for.body:
+; UNROLL-NO-VF-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[ADDX:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
+; UNROLL-NO-VF-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
+; UNROLL-NO-VF-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
+; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[SCALAR_RECUR]], [[FOR_BODY]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
+;
+; SINK-AFTER-LABEL: @extract_second_last_iteration(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X:%.*]], i32 0
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 1
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i32 [[INDEX]], 2
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i32 [[INDEX]], 3
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = zext <4 x i32> [[VEC_IND]] to <4 x i64>
+; SINK-AFTER-NEXT:    [[TMP5]] = add <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP5]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
+; SINK-AFTER-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, 96
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP5]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP5]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; SINK-AFTER:       for.body:
+; SINK-AFTER-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[ADDX:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
+; SINK-AFTER-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
+; SINK-AFTER-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
+; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
+; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[SCALAR_RECUR]], [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
+;
+; NO-SINK-AFTER-LABEL: @extract_second_last_iteration(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X:%.*]], i32 0
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i32 [[INDEX]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i32 [[INDEX]], 3
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = zext <4 x i32> [[VEC_IND]] to <4 x i64>
+; NO-SINK-AFTER-NEXT:    [[TMP5]] = add <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP5]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, 96
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP5]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP5]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-SINK-AFTER:       for.body:
+; NO-SINK-AFTER-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[ADDX:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
+; NO-SINK-AFTER-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
+; NO-SINK-AFTER-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
+; NO-SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[SCALAR_RECUR]], [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
+;
 entry:
   br label %for.body
 
@@ -420,15 +2720,473 @@ for.end:
 ;   return a;
 ; }
 ;
-; UNROLL-NO-IC-LABEL: @PR33613(
-; UNROLL-NO-IC:     vector.body:
-; UNROLL-NO-IC:       [[VECTOR_RECUR:%.*]] = phi <4 x double>
-; UNROLL-NO-IC:       shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> {{.*}}, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; UNROLL-NO-IC-NEXT:  shufflevector <4 x double> {{.*}}, <4 x double> {{.*}}, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; UNROLL-NO-IC-NOT:   insertelement <4 x double>
-; UNROLL-NO-IC:     middle.block:
 ;
 define i32 @PR33613(double* %b, double %j, i32 %d) {
+; CHECK-LABEL: @PR33613(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x double> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = mul i64 [[INDEX]], 25
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B:%.*]], i64 [[IDXPROM]]
+; CHECK-NEXT:    [[TMP1:%.*]] = or i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 25
+; CHECK-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP4:%.*]] = mul i64 [[TMP3]], 25
+; CHECK-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; CHECK-NEXT:    [[TMP5:%.*]] = or i64 [[INDEX]], 3
+; CHECK-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 25
+; CHECK-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[TMP2]]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP3]], i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP4]], i64 [[TMP6]]
+; CHECK-NEXT:    [[TMP11:%.*]] = load double, double* [[TMP7]], align 8
+; CHECK-NEXT:    [[TMP12:%.*]] = load double, double* [[TMP8]], align 8
+; CHECK-NEXT:    [[TMP13:%.*]] = load double, double* [[TMP9]], align 8
+; CHECK-NEXT:    [[TMP14:%.*]] = load double, double* [[TMP10]], align 8
+; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x double> poison, double [[TMP11]], i32 0
+; CHECK-NEXT:    [[TMP16:%.*]] = insertelement <4 x double> [[TMP15]], double [[TMP12]], i32 1
+; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x double> [[TMP16]], double [[TMP13]], i32 2
+; CHECK-NEXT:    [[TMP18]] = insertelement <4 x double> [[TMP17]], double [[TMP14]], i32 3
+; CHECK-NEXT:    [[TMP19:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[TMP17]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[TMP20:%.*]] = fmul <4 x double> [[TMP19]], [[TMP18]]
+; CHECK-NEXT:    [[TMP21:%.*]] = fcmp une <4 x double> [[TMP20]], zeroinitializer
+; CHECK-NEXT:    [[TMP22:%.*]] = zext <4 x i1> [[TMP21]] to <4 x i32>
+; CHECK-NEXT:    [[TMP23]] = add <4 x i32> [[VEC_PHI]], [[TMP22]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
+; CHECK-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP25:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP23]])
+; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ undef, [[FOR_BODY]] ], [ [[TMP25]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[A_1_LCSSA]]
+; CHECK:       for.body:
+; CHECK-NEXT:    br i1 undef, label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
+;
+; UNROLL-LABEL: @PR33613(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI9:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP48:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x double> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP38:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP0:%.*]] = mul i64 [[INDEX]], 25
+; UNROLL-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B:%.*]], i64 [[IDXPROM]]
+; UNROLL-NEXT:    [[TMP1:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 25
+; UNROLL-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; UNROLL-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 2
+; UNROLL-NEXT:    [[TMP4:%.*]] = mul i64 [[TMP3]], 25
+; UNROLL-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; UNROLL-NEXT:    [[TMP5:%.*]] = or i64 [[INDEX]], 3
+; UNROLL-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 25
+; UNROLL-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; UNROLL-NEXT:    [[TMP7:%.*]] = or i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 25
+; UNROLL-NEXT:    [[NEXT_GEP5:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = or i64 [[INDEX]], 5
+; UNROLL-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 25
+; UNROLL-NEXT:    [[NEXT_GEP6:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; UNROLL-NEXT:    [[TMP11:%.*]] = or i64 [[INDEX]], 6
+; UNROLL-NEXT:    [[TMP12:%.*]] = mul i64 [[TMP11]], 25
+; UNROLL-NEXT:    [[NEXT_GEP7:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; UNROLL-NEXT:    [[TMP13:%.*]] = or i64 [[INDEX]], 7
+; UNROLL-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], 25
+; UNROLL-NEXT:    [[NEXT_GEP8:%.*]] = getelementptr double, double* [[B]], i64 [[IDXPROM]]
+; UNROLL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[TMP0]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[TMP2]]
+; UNROLL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP3]], i64 [[TMP4]]
+; UNROLL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP4]], i64 [[TMP6]]
+; UNROLL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP5]], i64 [[TMP8]]
+; UNROLL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP6]], i64 [[TMP10]]
+; UNROLL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP7]], i64 [[TMP12]]
+; UNROLL-NEXT:    [[TMP22:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP8]], i64 [[TMP14]]
+; UNROLL-NEXT:    [[TMP23:%.*]] = load double, double* [[TMP15]], align 8
+; UNROLL-NEXT:    [[TMP24:%.*]] = load double, double* [[TMP16]], align 8
+; UNROLL-NEXT:    [[TMP25:%.*]] = load double, double* [[TMP17]], align 8
+; UNROLL-NEXT:    [[TMP26:%.*]] = load double, double* [[TMP18]], align 8
+; UNROLL-NEXT:    [[TMP27:%.*]] = insertelement <4 x double> poison, double [[TMP23]], i32 0
+; UNROLL-NEXT:    [[TMP28:%.*]] = insertelement <4 x double> [[TMP27]], double [[TMP24]], i32 1
+; UNROLL-NEXT:    [[TMP29:%.*]] = insertelement <4 x double> [[TMP28]], double [[TMP25]], i32 2
+; UNROLL-NEXT:    [[TMP30:%.*]] = insertelement <4 x double> [[TMP29]], double [[TMP26]], i32 3
+; UNROLL-NEXT:    [[TMP31:%.*]] = load double, double* [[TMP19]], align 8
+; UNROLL-NEXT:    [[TMP32:%.*]] = load double, double* [[TMP20]], align 8
+; UNROLL-NEXT:    [[TMP33:%.*]] = load double, double* [[TMP21]], align 8
+; UNROLL-NEXT:    [[TMP34:%.*]] = load double, double* [[TMP22]], align 8
+; UNROLL-NEXT:    [[TMP35:%.*]] = insertelement <4 x double> poison, double [[TMP31]], i32 0
+; UNROLL-NEXT:    [[TMP36:%.*]] = insertelement <4 x double> [[TMP35]], double [[TMP32]], i32 1
+; UNROLL-NEXT:    [[TMP37:%.*]] = insertelement <4 x double> [[TMP36]], double [[TMP33]], i32 2
+; UNROLL-NEXT:    [[TMP38]] = insertelement <4 x double> [[TMP37]], double [[TMP34]], i32 3
+; UNROLL-NEXT:    [[TMP39:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[TMP29]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP40:%.*]] = shufflevector <4 x double> [[TMP30]], <4 x double> [[TMP37]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP41:%.*]] = fmul <4 x double> [[TMP39]], [[TMP30]]
+; UNROLL-NEXT:    [[TMP42:%.*]] = fmul <4 x double> [[TMP40]], [[TMP38]]
+; UNROLL-NEXT:    [[TMP43:%.*]] = fcmp une <4 x double> [[TMP41]], zeroinitializer
+; UNROLL-NEXT:    [[TMP44:%.*]] = fcmp une <4 x double> [[TMP42]], zeroinitializer
+; UNROLL-NEXT:    [[TMP45:%.*]] = zext <4 x i1> [[TMP43]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP46:%.*]] = zext <4 x i1> [[TMP44]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP47]] = add <4 x i32> [[VEC_PHI]], [[TMP45]]
+; UNROLL-NEXT:    [[TMP48]] = add <4 x i32> [[VEC_PHI9]], [[TMP46]]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP49:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
+; UNROLL-NEXT:    br i1 [[TMP49]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP48]], [[TMP47]]
+; UNROLL-NEXT:    [[TMP50:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; UNROLL-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.cond.cleanup:
+; UNROLL-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ undef, [[FOR_BODY]] ], [ [[TMP50]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i32 [[A_1_LCSSA]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    br i1 undef, label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
+;
+; UNROLL-NO-IC-LABEL: @PR33613(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = getelementptr double, double* [[B:%.*]], i64 256000
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP50:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI9:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP51:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x double> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP41:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 25
+; UNROLL-NO-IC-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B]], i64 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 25
+; UNROLL-NO-IC-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 25
+; UNROLL-NO-IC-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr double, double* [[B]], i64 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 25
+; UNROLL-NO-IC-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr double, double* [[B]], i64 [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 25
+; UNROLL-NO-IC-NEXT:    [[NEXT_GEP5:%.*]] = getelementptr double, double* [[B]], i64 [[TMP9]]
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 5
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 25
+; UNROLL-NO-IC-NEXT:    [[NEXT_GEP6:%.*]] = getelementptr double, double* [[B]], i64 [[TMP11]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 6
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 25
+; UNROLL-NO-IC-NEXT:    [[NEXT_GEP7:%.*]] = getelementptr double, double* [[B]], i64 [[TMP13]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 7
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 25
+; UNROLL-NO-IC-NEXT:    [[NEXT_GEP8:%.*]] = getelementptr double, double* [[B]], i64 [[TMP15]]
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = add i32 [[OFFSET_IDX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP3]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP4]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP5]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP6]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP7]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP8]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = load double, double* [[TMP18]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = load double, double* [[TMP19]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = load double, double* [[TMP20]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = load double, double* [[TMP21]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = insertelement <4 x double> poison, double [[TMP26]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = insertelement <4 x double> [[TMP30]], double [[TMP27]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = insertelement <4 x double> [[TMP31]], double [[TMP28]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = insertelement <4 x double> [[TMP32]], double [[TMP29]], i32 3
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = load double, double* [[TMP22]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = load double, double* [[TMP23]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = load double, double* [[TMP24]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = load double, double* [[TMP25]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = insertelement <4 x double> poison, double [[TMP34]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = insertelement <4 x double> [[TMP38]], double [[TMP35]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = insertelement <4 x double> [[TMP39]], double [[TMP36]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP41]] = insertelement <4 x double> [[TMP40]], double [[TMP37]], i32 3
+; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[TMP33]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = shufflevector <4 x double> [[TMP33]], <4 x double> [[TMP41]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = fmul <4 x double> [[TMP42]], [[TMP33]]
+; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = fmul <4 x double> [[TMP43]], [[TMP41]]
+; UNROLL-NO-IC-NEXT:    [[TMP46:%.*]] = fcmp une <4 x double> [[TMP44]], zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP47:%.*]] = fcmp une <4 x double> [[TMP45]], zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP48:%.*]] = zext <4 x i1> [[TMP46]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP49:%.*]] = zext <4 x i1> [[TMP47]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP50]] = add <4 x i32> [[VEC_PHI]], [[TMP48]]
+; UNROLL-NO-IC-NEXT:    [[TMP51]] = add <4 x i32> [[VEC_PHI9]], [[TMP49]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP52:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP52]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP51]], [[TMP50]]
+; UNROLL-NO-IC-NEXT:    [[TMP53:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10240, 10240
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[TMP41]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x double> [[TMP41]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[J]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi double* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP53]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.cond.cleanup:
+; UNROLL-NO-IC-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ [[A_1:%.*]], [[FOR_BODY]] ], [ [[TMP53]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[A_1_LCSSA]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[B_ADDR_012:%.*]] = phi double* [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[I_011:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC1:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[A_010:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_1]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP54:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 [[IDXPROM]]
+; UNROLL-NO-IC-NEXT:    [[TMP54]] = load double, double* [[ARRAYIDX]], align 8
+; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = fmul double [[SCALAR_RECUR]], [[TMP54]]
+; UNROLL-NO-IC-NEXT:    [[TOBOOL:%.*]] = fcmp une double [[MUL]], 0.000000e+00
+; UNROLL-NO-IC-NEXT:    [[INC:%.*]] = zext i1 [[TOBOOL]] to i32
+; UNROLL-NO-IC-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
+; UNROLL-NO-IC-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
+; UNROLL-NO-IC-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 25
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
+;
+; UNROLL-NO-VF-LABEL: @PR33613(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
+; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = getelementptr double, double* [[B:%.*]], i64 256000
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VEC_PHI4:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi double [ [[J:%.*]], [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 25
+; UNROLL-NO-VF-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B]], i64 [[TMP1]]
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 25
+; UNROLL-NO-VF-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP3]]
+; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION3:%.*]] = add i32 [[OFFSET_IDX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[IDXPROM]]
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[IDXPROM]]
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = load double, double* [[TMP4]], align 8
+; UNROLL-NO-VF-NEXT:    [[TMP7]] = load double, double* [[TMP5]], align 8
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = fmul double [[VECTOR_RECUR]], [[TMP6]]
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = fmul double [[TMP6]], [[TMP7]]
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = fcmp une double [[TMP8]], 0.000000e+00
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = fcmp une double [[TMP9]], 0.000000e+00
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = zext i1 [[TMP10]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = zext i1 [[TMP11]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP14]] = add i32 [[VEC_PHI]], [[TMP12]]
+; UNROLL-NO-VF-NEXT:    [[TMP15]] = add i32 [[VEC_PHI4]], [[TMP13]]
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP15]], [[TMP14]]
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10240, 10240
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[J]], [[ENTRY:%.*]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi double* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-VF:       for.cond.cleanup:
+; UNROLL-NO-VF-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ [[A_1:%.*]], [[FOR_BODY]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    ret i32 [[A_1_LCSSA]]
+; UNROLL-NO-VF:       for.body:
+; UNROLL-NO-VF-NEXT:    [[B_ADDR_012:%.*]] = phi double* [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[I_011:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC1:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[A_010:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_1]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP17:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 [[IDXPROM]]
+; UNROLL-NO-VF-NEXT:    [[TMP17]] = load double, double* [[ARRAYIDX]], align 8
+; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = fmul double [[SCALAR_RECUR]], [[TMP17]]
+; UNROLL-NO-VF-NEXT:    [[TOBOOL:%.*]] = fcmp une double [[MUL]], 0.000000e+00
+; UNROLL-NO-VF-NEXT:    [[INC:%.*]] = zext i1 [[TOBOOL]] to i32
+; UNROLL-NO-VF-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
+; UNROLL-NO-VF-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
+; UNROLL-NO-VF-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 25
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+;
+; SINK-AFTER-LABEL: @PR33613(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
+; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[IND_END:%.*]] = getelementptr double, double* [[B:%.*]], i64 256000
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP25:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x double> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 25
+; SINK-AFTER-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B]], i64 [[TMP1]]
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 1
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 25
+; SINK-AFTER-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP3]]
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 2
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 25
+; SINK-AFTER-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr double, double* [[B]], i64 [[TMP5]]
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 3
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 25
+; SINK-AFTER-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr double, double* [[B]], i64 [[TMP7]]
+; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], 0
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[IDXPROM]]
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[IDXPROM]]
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP3]], i64 [[IDXPROM]]
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP4]], i64 [[IDXPROM]]
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = load double, double* [[TMP9]], align 8
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = load double, double* [[TMP10]], align 8
+; SINK-AFTER-NEXT:    [[TMP15:%.*]] = load double, double* [[TMP11]], align 8
+; SINK-AFTER-NEXT:    [[TMP16:%.*]] = load double, double* [[TMP12]], align 8
+; SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x double> poison, double [[TMP13]], i32 0
+; SINK-AFTER-NEXT:    [[TMP18:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP14]], i32 1
+; SINK-AFTER-NEXT:    [[TMP19:%.*]] = insertelement <4 x double> [[TMP18]], double [[TMP15]], i32 2
+; SINK-AFTER-NEXT:    [[TMP20]] = insertelement <4 x double> [[TMP19]], double [[TMP16]], i32 3
+; SINK-AFTER-NEXT:    [[TMP21:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[TMP20]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP22:%.*]] = fmul <4 x double> [[TMP21]], [[TMP20]]
+; SINK-AFTER-NEXT:    [[TMP23:%.*]] = fcmp une <4 x double> [[TMP22]], zeroinitializer
+; SINK-AFTER-NEXT:    [[TMP24:%.*]] = zext <4 x i1> [[TMP23]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP25]] = add <4 x i32> [[VEC_PHI]], [[TMP24]]
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
+; SINK-AFTER-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[TMP27:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP25]])
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10240, 10240
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[TMP20]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x double> [[TMP20]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[J]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi double* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; SINK-AFTER:       for.cond.cleanup:
+; SINK-AFTER-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ [[A_1:%.*]], [[FOR_BODY]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    ret i32 [[A_1_LCSSA]]
+; SINK-AFTER:       for.body:
+; SINK-AFTER-NEXT:    [[B_ADDR_012:%.*]] = phi double* [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[I_011:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC1:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[A_010:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_1]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 [[IDXPROM]]
+; SINK-AFTER-NEXT:    [[TMP28]] = load double, double* [[ARRAYIDX]], align 8
+; SINK-AFTER-NEXT:    [[MUL:%.*]] = fmul double [[SCALAR_RECUR]], [[TMP28]]
+; SINK-AFTER-NEXT:    [[TOBOOL:%.*]] = fcmp une double [[MUL]], 0.000000e+00
+; SINK-AFTER-NEXT:    [[INC:%.*]] = zext i1 [[TOBOOL]] to i32
+; SINK-AFTER-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
+; SINK-AFTER-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
+; SINK-AFTER-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 25
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
+;
+; NO-SINK-AFTER-LABEL: @PR33613(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
+; NO-SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[IND_END:%.*]] = getelementptr double, double* [[B:%.*]], i64 256000
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP25:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x double> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 25
+; NO-SINK-AFTER-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B]], i64 [[TMP1]]
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 25
+; NO-SINK-AFTER-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP3]]
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 25
+; NO-SINK-AFTER-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr double, double* [[B]], i64 [[TMP5]]
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 3
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 25
+; NO-SINK-AFTER-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr double, double* [[B]], i64 [[TMP7]]
+; NO-SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[IDXPROM]]
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[IDXPROM]]
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP3]], i64 [[IDXPROM]]
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP4]], i64 [[IDXPROM]]
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = load double, double* [[TMP9]], align 8
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = load double, double* [[TMP10]], align 8
+; NO-SINK-AFTER-NEXT:    [[TMP15:%.*]] = load double, double* [[TMP11]], align 8
+; NO-SINK-AFTER-NEXT:    [[TMP16:%.*]] = load double, double* [[TMP12]], align 8
+; NO-SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x double> poison, double [[TMP13]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP18:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP14]], i32 1
+; NO-SINK-AFTER-NEXT:    [[TMP19:%.*]] = insertelement <4 x double> [[TMP18]], double [[TMP15]], i32 2
+; NO-SINK-AFTER-NEXT:    [[TMP20]] = insertelement <4 x double> [[TMP19]], double [[TMP16]], i32 3
+; NO-SINK-AFTER-NEXT:    [[TMP21:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[TMP20]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP22:%.*]] = fmul <4 x double> [[TMP21]], [[TMP20]]
+; NO-SINK-AFTER-NEXT:    [[TMP23:%.*]] = fcmp une <4 x double> [[TMP22]], zeroinitializer
+; NO-SINK-AFTER-NEXT:    [[TMP24:%.*]] = zext <4 x i1> [[TMP23]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP25]] = add <4 x i32> [[VEC_PHI]], [[TMP24]]
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[TMP27:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP25]])
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10240, 10240
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[TMP20]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x double> [[TMP20]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[J]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi double* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-SINK-AFTER:       for.cond.cleanup:
+; NO-SINK-AFTER-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ [[A_1:%.*]], [[FOR_BODY]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    ret i32 [[A_1_LCSSA]]
+; NO-SINK-AFTER:       for.body:
+; NO-SINK-AFTER-NEXT:    [[B_ADDR_012:%.*]] = phi double* [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[I_011:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC1:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[A_010:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_1]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 [[IDXPROM]]
+; NO-SINK-AFTER-NEXT:    [[TMP28]] = load double, double* [[ARRAYIDX]], align 8
+; NO-SINK-AFTER-NEXT:    [[MUL:%.*]] = fmul double [[SCALAR_RECUR]], [[TMP28]]
+; NO-SINK-AFTER-NEXT:    [[TOBOOL:%.*]] = fcmp une double [[MUL]], 0.000000e+00
+; NO-SINK-AFTER-NEXT:    [[INC:%.*]] = zext i1 [[TOBOOL]] to i32
+; NO-SINK-AFTER-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
+; NO-SINK-AFTER-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
+; NO-SINK-AFTER-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 25
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
+;
 entry:
   %idxprom = sext i32 %d to i64
   br label %for.body
@@ -459,17 +3217,428 @@ for.body:
 ;     b[i] = (a[i] * a[i + 1]);
 ; }
 ;
-; SINK-AFTER-LABEL: sink_after
 ; Check that the sext sank after the load in the vector loop.
-; SINK-AFTER: vector.body
-; SINK-AFTER:   %vector.recur = phi <4 x i16> [ %vector.recur.init, %vector.ph ], [ %wide.load, %vector.body ]
-; SINK-AFTER:   %wide.load = load <4 x i16>
-; SINK-AFTER:   %[[VSHUF:.+]] = shufflevector <4 x i16> %vector.recur, <4 x i16> %wide.load, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; SINK-AFTER:   %[[VCONV:.+]] = sext <4 x i16> %[[VSHUF]] to <4 x i32>
-; SINK-AFTER:   %[[VCONV3:.+]] = sext <4 x i16> %wide.load to <4 x i32>
-; SINK-AFTER:   mul nsw <4 x i32> %[[VCONV3]], %[[VCONV]]
 ;
 define void @sink_after(i16* %a, i32* %b, i64 %n) {
+; CHECK-LABEL: @sink_after(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK:       vector.memcheck:
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[N]]
+; CHECK-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; CHECK-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[SCEVGEP5]] to i32*
+; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[TMP1]], [[B]]
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[SCEVGEP]] to i16*
+; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult i16* [[SCEVGEP3]], [[TMP2]]
+; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -4
+; CHECK-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP5]], align 2, !alias.scope !26
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[TMP6]] to <4 x i32>
+; CHECK-NEXT:    [[TMP8:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; CHECK-NEXT:    [[TMP9:%.*]] = mul nsw <4 x i32> [[TMP8]], [[TMP7]]
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <4 x i32>*
+; CHECK-NEXT:    store <4 x i32> [[TMP9]], <4 x i32>* [[TMP11]], align 4, !alias.scope !29, !noalias !26
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP13:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; CHECK-NEXT:    [[TMP13]] = load i16, i16* [[ARRAYIDX2]], align 2
+; CHECK-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP13]] to i32
+; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @sink_after(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL:       vector.memcheck:
+; UNROLL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[N]]
+; UNROLL-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; UNROLL-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; UNROLL-NEXT:    [[TMP1:%.*]] = bitcast i16* [[SCEVGEP5]] to i32*
+; UNROLL-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[TMP1]], [[B]]
+; UNROLL-NEXT:    [[TMP2:%.*]] = bitcast i32* [[SCEVGEP]] to i16*
+; UNROLL-NEXT:    [[BOUND1:%.*]] = icmp ult i16* [[SCEVGEP3]], [[TMP2]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -8
+; UNROLL-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP3]]
+; UNROLL-NEXT:    [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
+; UNROLL-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP5]], align 2, !alias.scope !26
+; UNROLL-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i64 4
+; UNROLL-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP6]] to <4 x i16>*
+; UNROLL-NEXT:    [[WIDE_LOAD7]] = load <4 x i16>, <4 x i16>* [[TMP7]], align 2, !alias.scope !26
+; UNROLL-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP9:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP10:%.*]] = sext <4 x i16> [[TMP8]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP11:%.*]] = sext <4 x i16> [[TMP9]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP12:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP13:%.*]] = sext <4 x i16> [[WIDE_LOAD7]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP14:%.*]] = mul nsw <4 x i32> [[TMP12]], [[TMP10]]
+; UNROLL-NEXT:    [[TMP15:%.*]] = mul nsw <4 x i32> [[TMP13]], [[TMP11]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP16]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> [[TMP14]], <4 x i32>* [[TMP17]], align 4, !alias.scope !29, !noalias !26
+; UNROLL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[TMP16]], i64 4
+; UNROLL-NEXT:    [[TMP19:%.*]] = bitcast i32* [[TMP18]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> [[TMP15]], <4 x i32>* [[TMP19]], align 4, !alias.scope !29, !noalias !26
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD7]], i32 3
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP21:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NEXT:    [[TMP21]] = load i16, i16* [[ARRAYIDX2]], align 2
+; UNROLL-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP21]] to i32
+; UNROLL-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; UNROLL-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @sink_after(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-IC:       vector.memcheck:
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP34:%.*]] = bitcast i16* [[SCEVGEP3]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP56:%.*]] = bitcast i16* [[SCEVGEP5]] to i8*
+; UNROLL-NO-IC-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; UNROLL-NO-IC-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-IC-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, i16* [[TMP5]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = bitcast i16* [[TMP7]] to <4 x i16>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP8]], align 2, !alias.scope !26
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, i16* [[TMP5]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = bitcast i16* [[TMP9]] to <4 x i16>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD7]] = load <4 x i16>, <4 x i16>* [[TMP10]], align 2, !alias.scope !26
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = sext <4 x i16> [[TMP11]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = sext <4 x i16> [[TMP12]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = sext <4 x i16> [[WIDE_LOAD7]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = mul nsw <4 x i32> [[TMP15]], [[TMP13]]
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = mul nsw <4 x i32> [[TMP16]], [[TMP14]]
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP2]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = bitcast i32* [[TMP21]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP17]], <4 x i32>* [[TMP22]], align 4, !alias.scope !29, !noalias !26
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = bitcast i32* [[TMP23]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP18]], <4 x i32>* [[TMP24]], align 4, !alias.scope !29, !noalias !26
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD7]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD7]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP26:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NO-IC-NEXT:    [[TMP26]] = load i16, i16* [[ARRAYIDX2]], align 2
+; UNROLL-NO-IC-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP26]] to i32
+; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @sink_after(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-VF:       vector.memcheck:
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP34:%.*]] = bitcast i16* [[SCEVGEP3]] to i8*
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP56:%.*]] = bitcast i16* [[SCEVGEP5]] to i8*
+; UNROLL-NO-VF-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; UNROLL-NO-VF-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; UNROLL-NO-VF-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-VF-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION7:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDUCTION]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[INDUCTION7]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP1]]
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP2]]
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = load i16, i16* [[TMP3]], align 2, !alias.scope !25
+; UNROLL-NO-VF-NEXT:    [[TMP6]] = load i16, i16* [[TMP4]], align 2, !alias.scope !25
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sext i16 [[VECTOR_RECUR]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP5]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = sext i16 [[TMP5]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = sext i16 [[TMP6]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[TMP9]], [[TMP7]]
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = mul nsw i32 [[TMP10]], [[TMP8]]
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION7]]
+; UNROLL-NO-VF-NEXT:    store i32 [[TMP11]], i32* [[TMP13]], align 4, !alias.scope !28, !noalias !25
+; UNROLL-NO-VF-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4, !alias.scope !28, !noalias !25
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-VF:       for.body:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP16:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NO-VF-NEXT:    [[TMP16]] = load i16, i16* [[ARRAYIDX2]], align 2
+; UNROLL-NO-VF-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP16]] to i32
+; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NO-VF-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @sink_after(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; SINK-AFTER:       vector.memcheck:
+; SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; SINK-AFTER-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; SINK-AFTER-NEXT:    [[SCEVGEP34:%.*]] = bitcast i16* [[SCEVGEP3]] to i8*
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; SINK-AFTER-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; SINK-AFTER-NEXT:    [[SCEVGEP56:%.*]] = bitcast i16* [[SCEVGEP5]] to i8*
+; SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; SINK-AFTER-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP2]]
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 0
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP5]], align 2, !alias.scope !26
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[TMP6]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = mul nsw <4 x i32> [[TMP8]], [[TMP7]]
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP1]]
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP10]], i32 0
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>*
+; SINK-AFTER-NEXT:    store <4 x i32> [[TMP9]], <4 x i32>* [[TMP12]], align 4, !alias.scope !29, !noalias !26
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; SINK-AFTER:       for.body:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP14:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SINK-AFTER-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; SINK-AFTER-NEXT:    [[TMP14]] = load i16, i16* [[ARRAYIDX2]], align 2
+; SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP14]] to i32
+; SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @sink_after(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; NO-SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; NO-SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; NO-SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; NO-SINK-AFTER:       vector.memcheck:
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP34:%.*]] = bitcast i16* [[SCEVGEP3]] to i8*
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP56:%.*]] = bitcast i16* [[SCEVGEP5]] to i8*
+; NO-SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; NO-SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; NO-SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-SINK-AFTER-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP2]]
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
+; NO-SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP5]], align 2, !alias.scope !26
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[TMP6]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = mul nsw <4 x i32> [[TMP8]], [[TMP7]]
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP1]]
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP10]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>*
+; NO-SINK-AFTER-NEXT:    store <4 x i32> [[TMP9]], <4 x i32>* [[TMP12]], align 4, !alias.scope !29, !noalias !26
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-SINK-AFTER:       for.body:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP14:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; NO-SINK-AFTER-NEXT:    [[TMP14]] = load i16, i16* [[ARRAYIDX2]], align 2
+; NO-SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP14]] to i32
+; NO-SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; NO-SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
+;
 entry:
   %.pre = load i16, i16* %a
   br label %for.body
@@ -506,16 +3675,606 @@ for.end:
 ;   }
 ; }
 ;
-; SINK-AFTER-LABEL: @PR34711
 ; Check that the sext sank after the load in the vector loop.
-; SINK-AFTER: vector.body
-; SINK-AFTER:   %vector.recur = phi <4 x i16> [ %vector.recur.init, %vector.ph ], [ {{.*}}, %vector.body ]
-; SINK-AFTER:   %[[VSHUF:.+]] = shufflevector <4 x i16> %vector.recur, <4 x i16> %{{.*}}, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; SINK-AFTER:   %[[VCONV:.+]] = sext <4 x i16> %[[VSHUF]] to <4 x i32>
-; SINK-AFTER:   %[[VCONV3:.+]] = sext <4 x i16> {{.*}} to <4 x i32>
-; SINK-AFTER:   mul nsw <4 x i32> %[[VCONV3]], %[[VCONV]]
 ;
 define void @PR34711([2 x i16]* %a, i32* %b, i32* %c, i64 %n) {
+; CHECK-LABEL: @PR34711(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
+; CHECK-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK:       vector.memcheck:
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[C:%.*]], i64 [[N]]
+; CHECK-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[N]]
+; CHECK-NEXT:    [[SCEVGEP6:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 0, i64 1
+; CHECK-NEXT:    [[SCEVGEP8:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 [[N]], i64 0
+; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[SCEVGEP4]], [[C]]
+; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ugt i32* [[SCEVGEP]], [[B]]
+; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[SCEVGEP8]] to i32*
+; CHECK-NEXT:    [[BOUND010:%.*]] = icmp ugt i32* [[TMP0]], [[C]]
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[SCEVGEP]] to i16*
+; CHECK-NEXT:    [[BOUND111:%.*]] = icmp ult i16* [[SCEVGEP6]], [[TMP1]]
+; CHECK-NEXT:    [[FOUND_CONFLICT12:%.*]] = and i1 [[BOUND010]], [[BOUND111]]
+; CHECK-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT12]]
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[SCEVGEP8]] to i32*
+; CHECK-NEXT:    [[BOUND013:%.*]] = icmp ugt i32* [[TMP2]], [[B]]
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[SCEVGEP4]] to i16*
+; CHECK-NEXT:    [[BOUND114:%.*]] = icmp ult i16* [[SCEVGEP6]], [[TMP3]]
+; CHECK-NEXT:    [[FOUND_CONFLICT15:%.*]] = and i1 [[BOUND013]], [[BOUND114]]
+; CHECK-NEXT:    [[CONFLICT_RDX16:%.*]] = or i1 [[CONFLICT_RDX]], [[FOUND_CONFLICT15]]
+; CHECK-NEXT:    br i1 [[CONFLICT_RDX16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -4
+; CHECK-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = or i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP5:%.*]] = or i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP6:%.*]] = or i64 [[INDEX]], 3
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDEX]], i64 1
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP4]], i64 1
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP5]], i64 1
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP6]], i64 1
+; CHECK-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP7]] to <4 x i32>*
+; CHECK-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP12]], align 4, !alias.scope !33, !noalias !36
+; CHECK-NEXT:    [[TMP13:%.*]] = load i16, i16* [[TMP8]], align 2, !alias.scope !39
+; CHECK-NEXT:    [[TMP14:%.*]] = load i16, i16* [[TMP9]], align 2, !alias.scope !39
+; CHECK-NEXT:    [[TMP15:%.*]] = load i16, i16* [[TMP10]], align 2, !alias.scope !39
+; CHECK-NEXT:    [[TMP16:%.*]] = load i16, i16* [[TMP11]], align 2, !alias.scope !39
+; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x i16> poison, i16 [[TMP13]], i32 0
+; CHECK-NEXT:    [[TMP18:%.*]] = insertelement <4 x i16> [[TMP17]], i16 [[TMP14]], i32 1
+; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x i16> [[TMP18]], i16 [[TMP15]], i32 2
+; CHECK-NEXT:    [[TMP20]] = insertelement <4 x i16> [[TMP19]], i16 [[TMP16]], i32 3
+; CHECK-NEXT:    [[TMP21:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP19]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[TMP22:%.*]] = sext <4 x i16> [[TMP21]] to <4 x i32>
+; CHECK-NEXT:    [[TMP23:%.*]] = sext <4 x i16> [[TMP20]] to <4 x i32>
+; CHECK-NEXT:    [[TMP24:%.*]] = mul nsw <4 x i32> [[TMP23]], [[TMP22]]
+; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP26:%.*]] = bitcast i32* [[TMP25]] to <4 x i32>*
+; CHECK-NEXT:    store <4 x i32> [[TMP24]], <4 x i32>* [[TMP26]], align 4, !alias.scope !40, !noalias !39
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[TMP16]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
+; CHECK-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
+; CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; CHECK-NEXT:    [[TMP28]] = load i16, i16* [[CUR_INDEX]], align 2
+; CHECK-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP28]] to i32
+; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @PR34711(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
+; UNROLL-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL:       vector.memcheck:
+; UNROLL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[C:%.*]], i64 [[N]]
+; UNROLL-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[N]]
+; UNROLL-NEXT:    [[SCEVGEP6:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 0, i64 1
+; UNROLL-NEXT:    [[SCEVGEP8:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 [[N]], i64 0
+; UNROLL-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[SCEVGEP4]], [[C]]
+; UNROLL-NEXT:    [[BOUND1:%.*]] = icmp ugt i32* [[SCEVGEP]], [[B]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NEXT:    [[TMP0:%.*]] = bitcast i16* [[SCEVGEP8]] to i32*
+; UNROLL-NEXT:    [[BOUND010:%.*]] = icmp ugt i32* [[TMP0]], [[C]]
+; UNROLL-NEXT:    [[TMP1:%.*]] = bitcast i32* [[SCEVGEP]] to i16*
+; UNROLL-NEXT:    [[BOUND111:%.*]] = icmp ult i16* [[SCEVGEP6]], [[TMP1]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT12:%.*]] = and i1 [[BOUND010]], [[BOUND111]]
+; UNROLL-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT12]]
+; UNROLL-NEXT:    [[TMP2:%.*]] = bitcast i16* [[SCEVGEP8]] to i32*
+; UNROLL-NEXT:    [[BOUND013:%.*]] = icmp ugt i32* [[TMP2]], [[B]]
+; UNROLL-NEXT:    [[TMP3:%.*]] = bitcast i32* [[SCEVGEP4]] to i16*
+; UNROLL-NEXT:    [[BOUND114:%.*]] = icmp ult i16* [[SCEVGEP6]], [[TMP3]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT15:%.*]] = and i1 [[BOUND013]], [[BOUND114]]
+; UNROLL-NEXT:    [[CONFLICT_RDX16:%.*]] = or i1 [[CONFLICT_RDX]], [[FOUND_CONFLICT15]]
+; UNROLL-NEXT:    br i1 [[CONFLICT_RDX16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -8
+; UNROLL-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP38:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP4:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP5:%.*]] = or i64 [[INDEX]], 2
+; UNROLL-NEXT:    [[TMP6:%.*]] = or i64 [[INDEX]], 3
+; UNROLL-NEXT:    [[TMP7:%.*]] = or i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP8:%.*]] = or i64 [[INDEX]], 5
+; UNROLL-NEXT:    [[TMP9:%.*]] = or i64 [[INDEX]], 6
+; UNROLL-NEXT:    [[TMP10:%.*]] = or i64 [[INDEX]], 7
+; UNROLL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDEX]], i64 1
+; UNROLL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP4]], i64 1
+; UNROLL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP5]], i64 1
+; UNROLL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP6]], i64 1
+; UNROLL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP7]], i64 1
+; UNROLL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP8]], i64 1
+; UNROLL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP9]], i64 1
+; UNROLL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP10]], i64 1
+; UNROLL-NEXT:    [[TMP20:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP20]], align 4, !alias.scope !33, !noalias !36
+; UNROLL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP11]], i64 4
+; UNROLL-NEXT:    [[TMP22:%.*]] = bitcast i32* [[TMP21]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP22]], align 4, !alias.scope !33, !noalias !36
+; UNROLL-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP12]], align 2, !alias.scope !39
+; UNROLL-NEXT:    [[TMP24:%.*]] = load i16, i16* [[TMP13]], align 2, !alias.scope !39
+; UNROLL-NEXT:    [[TMP25:%.*]] = load i16, i16* [[TMP14]], align 2, !alias.scope !39
+; UNROLL-NEXT:    [[TMP26:%.*]] = load i16, i16* [[TMP15]], align 2, !alias.scope !39
+; UNROLL-NEXT:    [[TMP27:%.*]] = insertelement <4 x i16> poison, i16 [[TMP23]], i32 0
+; UNROLL-NEXT:    [[TMP28:%.*]] = insertelement <4 x i16> [[TMP27]], i16 [[TMP24]], i32 1
+; UNROLL-NEXT:    [[TMP29:%.*]] = insertelement <4 x i16> [[TMP28]], i16 [[TMP25]], i32 2
+; UNROLL-NEXT:    [[TMP30:%.*]] = insertelement <4 x i16> [[TMP29]], i16 [[TMP26]], i32 3
+; UNROLL-NEXT:    [[TMP31:%.*]] = load i16, i16* [[TMP16]], align 2, !alias.scope !39
+; UNROLL-NEXT:    [[TMP32:%.*]] = load i16, i16* [[TMP17]], align 2, !alias.scope !39
+; UNROLL-NEXT:    [[TMP33:%.*]] = load i16, i16* [[TMP18]], align 2, !alias.scope !39
+; UNROLL-NEXT:    [[TMP34:%.*]] = load i16, i16* [[TMP19]], align 2, !alias.scope !39
+; UNROLL-NEXT:    [[TMP35:%.*]] = insertelement <4 x i16> poison, i16 [[TMP31]], i32 0
+; UNROLL-NEXT:    [[TMP36:%.*]] = insertelement <4 x i16> [[TMP35]], i16 [[TMP32]], i32 1
+; UNROLL-NEXT:    [[TMP37:%.*]] = insertelement <4 x i16> [[TMP36]], i16 [[TMP33]], i32 2
+; UNROLL-NEXT:    [[TMP38]] = insertelement <4 x i16> [[TMP37]], i16 [[TMP34]], i32 3
+; UNROLL-NEXT:    [[TMP39:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP29]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP40:%.*]] = shufflevector <4 x i16> [[TMP30]], <4 x i16> [[TMP37]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP41:%.*]] = sext <4 x i16> [[TMP39]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP42:%.*]] = sext <4 x i16> [[TMP40]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP43:%.*]] = sext <4 x i16> [[TMP30]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP44:%.*]] = sext <4 x i16> [[TMP38]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP45:%.*]] = mul nsw <4 x i32> [[TMP43]], [[TMP41]]
+; UNROLL-NEXT:    [[TMP46:%.*]] = mul nsw <4 x i32> [[TMP44]], [[TMP42]]
+; UNROLL-NEXT:    [[TMP47:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP48:%.*]] = bitcast i32* [[TMP47]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> [[TMP45]], <4 x i32>* [[TMP48]], align 4, !alias.scope !40, !noalias !39
+; UNROLL-NEXT:    [[TMP49:%.*]] = getelementptr inbounds i32, i32* [[TMP47]], i64 4
+; UNROLL-NEXT:    [[TMP50:%.*]] = bitcast i32* [[TMP49]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> [[TMP46]], <4 x i32>* [[TMP50]], align 4, !alias.scope !40, !noalias !39
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP51:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP51]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[TMP34]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP52:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
+; UNROLL-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
+; UNROLL-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
+; UNROLL-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NEXT:    [[TMP52]] = load i16, i16* [[CUR_INDEX]], align 2
+; UNROLL-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP52]] to i32
+; UNROLL-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; UNROLL-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @PR34711(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[C1:%.*]] = bitcast i32* [[C:%.*]] to i8*
+; UNROLL-NO-IC-NEXT:    [[B3:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; UNROLL-NO-IC-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
+; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-IC:       vector.memcheck:
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[C]], i64 [[N]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP45:%.*]] = bitcast i32* [[SCEVGEP4]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP6:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 0, i64 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP67:%.*]] = bitcast i16* [[SCEVGEP6]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP8:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 [[N]], i64 0
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP89:%.*]] = bitcast i16* [[SCEVGEP8]] to i8*
+; UNROLL-NO-IC-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[C1]], [[SCEVGEP45]]
+; UNROLL-NO-IC-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[B3]], [[SCEVGEP2]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-IC-NEXT:    [[BOUND010:%.*]] = icmp ult i8* [[C1]], [[SCEVGEP89]]
+; UNROLL-NO-IC-NEXT:    [[BOUND111:%.*]] = icmp ult i8* [[SCEVGEP67]], [[SCEVGEP2]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT12:%.*]] = and i1 [[BOUND010]], [[BOUND111]]
+; UNROLL-NO-IC-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT12]]
+; UNROLL-NO-IC-NEXT:    [[BOUND013:%.*]] = icmp ult i8* [[B3]], [[SCEVGEP89]]
+; UNROLL-NO-IC-NEXT:    [[BOUND114:%.*]] = icmp ult i8* [[SCEVGEP67]], [[SCEVGEP45]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT15:%.*]] = and i1 [[BOUND013]], [[BOUND114]]
+; UNROLL-NO-IC-NEXT:    [[CONFLICT_RDX16:%.*]] = or i1 [[CONFLICT_RDX]], [[FOUND_CONFLICT15]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CONFLICT_RDX16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP37:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], 5
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 6
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 7
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP0]], i64 1
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP1]], i64 1
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP2]], i64 1
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP3]], i64 1
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP4]], i64 1
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP5]], i64 1
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP6]], i64 1
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP7]], i64 1
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[TMP8]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = bitcast i32* [[TMP18]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP19]], align 4, !alias.scope !33, !noalias !36
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP8]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP20]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP21]], align 4, !alias.scope !33, !noalias !36
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = load i16, i16* [[TMP10]], align 2, !alias.scope !39
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP11]], align 2, !alias.scope !39
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = load i16, i16* [[TMP12]], align 2, !alias.scope !39
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = load i16, i16* [[TMP13]], align 2, !alias.scope !39
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = insertelement <4 x i16> poison, i16 [[TMP22]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = insertelement <4 x i16> [[TMP26]], i16 [[TMP23]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = insertelement <4 x i16> [[TMP27]], i16 [[TMP24]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = insertelement <4 x i16> [[TMP28]], i16 [[TMP25]], i32 3
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = load i16, i16* [[TMP14]], align 2, !alias.scope !39
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = load i16, i16* [[TMP15]], align 2, !alias.scope !39
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = load i16, i16* [[TMP16]], align 2, !alias.scope !39
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = load i16, i16* [[TMP17]], align 2, !alias.scope !39
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = insertelement <4 x i16> poison, i16 [[TMP30]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = insertelement <4 x i16> [[TMP34]], i16 [[TMP31]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = insertelement <4 x i16> [[TMP35]], i16 [[TMP32]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP37]] = insertelement <4 x i16> [[TMP36]], i16 [[TMP33]], i32 3
+; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP29]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = shufflevector <4 x i16> [[TMP29]], <4 x i16> [[TMP37]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = sext <4 x i16> [[TMP38]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = sext <4 x i16> [[TMP39]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = sext <4 x i16> [[TMP29]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = sext <4 x i16> [[TMP37]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = mul nsw <4 x i32> [[TMP42]], [[TMP40]]
+; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = mul nsw <4 x i32> [[TMP43]], [[TMP41]]
+; UNROLL-NO-IC-NEXT:    [[TMP46:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[TMP47:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP48:%.*]] = getelementptr inbounds i32, i32* [[TMP46]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP49:%.*]] = bitcast i32* [[TMP48]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP44]], <4 x i32>* [[TMP49]], align 4, !alias.scope !40, !noalias !39
+; UNROLL-NO-IC-NEXT:    [[TMP50:%.*]] = getelementptr inbounds i32, i32* [[TMP46]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP51:%.*]] = bitcast i32* [[TMP50]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP45]], <4 x i32>* [[TMP51]], align 4, !alias.scope !40, !noalias !39
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP52:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP52]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP37]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP37]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP53:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
+; UNROLL-NO-IC-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
+; UNROLL-NO-IC-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP53]] = load i16, i16* [[CUR_INDEX]], align 2
+; UNROLL-NO-IC-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP53]] to i32
+; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @PR34711(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[C1:%.*]] = bitcast i32* [[C:%.*]] to i8*
+; UNROLL-NO-VF-NEXT:    [[B3:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; UNROLL-NO-VF-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
+; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-VF:       vector.memcheck:
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[C]], i64 [[N]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP45:%.*]] = bitcast i32* [[SCEVGEP4]] to i8*
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP6:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 0, i64 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP67:%.*]] = bitcast i16* [[SCEVGEP6]] to i8*
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP8:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 [[N]], i64 0
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP89:%.*]] = bitcast i16* [[SCEVGEP8]] to i8*
+; UNROLL-NO-VF-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[C1]], [[SCEVGEP45]]
+; UNROLL-NO-VF-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[B3]], [[SCEVGEP2]]
+; UNROLL-NO-VF-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-VF-NEXT:    [[BOUND010:%.*]] = icmp ult i8* [[C1]], [[SCEVGEP89]]
+; UNROLL-NO-VF-NEXT:    [[BOUND111:%.*]] = icmp ult i8* [[SCEVGEP67]], [[SCEVGEP2]]
+; UNROLL-NO-VF-NEXT:    [[FOUND_CONFLICT12:%.*]] = and i1 [[BOUND010]], [[BOUND111]]
+; UNROLL-NO-VF-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT12]]
+; UNROLL-NO-VF-NEXT:    [[BOUND013:%.*]] = icmp ult i8* [[B3]], [[SCEVGEP89]]
+; UNROLL-NO-VF-NEXT:    [[BOUND114:%.*]] = icmp ult i8* [[SCEVGEP67]], [[SCEVGEP45]]
+; UNROLL-NO-VF-NEXT:    [[FOUND_CONFLICT15:%.*]] = and i1 [[BOUND013]], [[BOUND114]]
+; UNROLL-NO-VF-NEXT:    [[CONFLICT_RDX16:%.*]] = or i1 [[CONFLICT_RDX]], [[FOUND_CONFLICT15]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CONFLICT_RDX16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION17:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDUCTION17]]
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDUCTION]], i64 1
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDUCTION17]], i64 1
+; UNROLL-NO-VF-NEXT:    store i32 7, i32* [[TMP0]], align 4, !alias.scope !32, !noalias !35
+; UNROLL-NO-VF-NEXT:    store i32 7, i32* [[TMP1]], align 4, !alias.scope !32, !noalias !35
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 2, !alias.scope !38
+; UNROLL-NO-VF-NEXT:    [[TMP5]] = load i16, i16* [[TMP3]], align 2, !alias.scope !38
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = sext i16 [[VECTOR_RECUR]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sext i16 [[TMP4]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP4]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = sext i16 [[TMP5]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = mul nsw i32 [[TMP8]], [[TMP6]]
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[TMP9]], [[TMP7]]
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION17]]
+; UNROLL-NO-VF-NEXT:    store i32 [[TMP10]], i32* [[TMP12]], align 4, !alias.scope !39, !noalias !38
+; UNROLL-NO-VF-NEXT:    store i32 [[TMP11]], i32* [[TMP13]], align 4, !alias.scope !39, !noalias !38
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-VF:       for.body:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
+; UNROLL-NO-VF-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
+; UNROLL-NO-VF-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
+; UNROLL-NO-VF-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP15]] = load i16, i16* [[CUR_INDEX]], align 2
+; UNROLL-NO-VF-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP15]] to i32
+; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NO-VF-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @PR34711(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[C1:%.*]] = bitcast i32* [[C:%.*]] to i8*
+; SINK-AFTER-NEXT:    [[B3:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; SINK-AFTER-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
+; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; SINK-AFTER:       vector.memcheck:
+; SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[C]], i64 [[N]]
+; SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; SINK-AFTER-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; SINK-AFTER-NEXT:    [[SCEVGEP45:%.*]] = bitcast i32* [[SCEVGEP4]] to i8*
+; SINK-AFTER-NEXT:    [[SCEVGEP6:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 0, i64 1
+; SINK-AFTER-NEXT:    [[SCEVGEP67:%.*]] = bitcast i16* [[SCEVGEP6]] to i8*
+; SINK-AFTER-NEXT:    [[SCEVGEP8:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 [[N]], i64 0
+; SINK-AFTER-NEXT:    [[SCEVGEP89:%.*]] = bitcast i16* [[SCEVGEP8]] to i8*
+; SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[C1]], [[SCEVGEP45]]
+; SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[B3]], [[SCEVGEP2]]
+; SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; SINK-AFTER-NEXT:    [[BOUND010:%.*]] = icmp ult i8* [[C1]], [[SCEVGEP89]]
+; SINK-AFTER-NEXT:    [[BOUND111:%.*]] = icmp ult i8* [[SCEVGEP67]], [[SCEVGEP2]]
+; SINK-AFTER-NEXT:    [[FOUND_CONFLICT12:%.*]] = and i1 [[BOUND010]], [[BOUND111]]
+; SINK-AFTER-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT12]]
+; SINK-AFTER-NEXT:    [[BOUND013:%.*]] = icmp ult i8* [[B3]], [[SCEVGEP89]]
+; SINK-AFTER-NEXT:    [[BOUND114:%.*]] = icmp ult i8* [[SCEVGEP67]], [[SCEVGEP45]]
+; SINK-AFTER-NEXT:    [[FOUND_CONFLICT15:%.*]] = and i1 [[BOUND013]], [[BOUND114]]
+; SINK-AFTER-NEXT:    [[CONFLICT_RDX16:%.*]] = or i1 [[CONFLICT_RDX]], [[FOUND_CONFLICT15]]
+; SINK-AFTER-NEXT:    br i1 [[CONFLICT_RDX16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 2
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 3
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[TMP0]]
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP0]], i64 1
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP1]], i64 1
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP2]], i64 1
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP3]], i64 1
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <4 x i32>*
+; SINK-AFTER-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP10]], align 4, !alias.scope !33, !noalias !36
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = load i16, i16* [[TMP5]], align 2, !alias.scope !39
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP6]], align 2, !alias.scope !39
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = load i16, i16* [[TMP7]], align 2, !alias.scope !39
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = load i16, i16* [[TMP8]], align 2, !alias.scope !39
+; SINK-AFTER-NEXT:    [[TMP15:%.*]] = insertelement <4 x i16> poison, i16 [[TMP11]], i32 0
+; SINK-AFTER-NEXT:    [[TMP16:%.*]] = insertelement <4 x i16> [[TMP15]], i16 [[TMP12]], i32 1
+; SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x i16> [[TMP16]], i16 [[TMP13]], i32 2
+; SINK-AFTER-NEXT:    [[TMP18]] = insertelement <4 x i16> [[TMP17]], i16 [[TMP14]], i32 3
+; SINK-AFTER-NEXT:    [[TMP19:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP18]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP20:%.*]] = sext <4 x i16> [[TMP19]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP21:%.*]] = sext <4 x i16> [[TMP18]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP22:%.*]] = mul nsw <4 x i32> [[TMP21]], [[TMP20]]
+; SINK-AFTER-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP0]]
+; SINK-AFTER-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i32 0
+; SINK-AFTER-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP24]] to <4 x i32>*
+; SINK-AFTER-NEXT:    store <4 x i32> [[TMP22]], <4 x i32>* [[TMP25]], align 4, !alias.scope !40, !noalias !39
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP18]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP18]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; SINK-AFTER:       for.body:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP27:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
+; SINK-AFTER-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
+; SINK-AFTER-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
+; SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; SINK-AFTER-NEXT:    [[TMP27]] = load i16, i16* [[CUR_INDEX]], align 2
+; SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP27]] to i32
+; SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @PR34711(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[C1:%.*]] = bitcast i32* [[C:%.*]] to i8*
+; NO-SINK-AFTER-NEXT:    [[B3:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; NO-SINK-AFTER-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
+; NO-SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
+; NO-SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; NO-SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; NO-SINK-AFTER:       vector.memcheck:
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[C]], i64 [[N]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP45:%.*]] = bitcast i32* [[SCEVGEP4]] to i8*
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP6:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 0, i64 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP67:%.*]] = bitcast i16* [[SCEVGEP6]] to i8*
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP8:%.*]] = getelementptr [2 x i16], [2 x i16]* [[A]], i64 [[N]], i64 0
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP89:%.*]] = bitcast i16* [[SCEVGEP8]] to i8*
+; NO-SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[C1]], [[SCEVGEP45]]
+; NO-SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[B3]], [[SCEVGEP2]]
+; NO-SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-SINK-AFTER-NEXT:    [[BOUND010:%.*]] = icmp ult i8* [[C1]], [[SCEVGEP89]]
+; NO-SINK-AFTER-NEXT:    [[BOUND111:%.*]] = icmp ult i8* [[SCEVGEP67]], [[SCEVGEP2]]
+; NO-SINK-AFTER-NEXT:    [[FOUND_CONFLICT12:%.*]] = and i1 [[BOUND010]], [[BOUND111]]
+; NO-SINK-AFTER-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT12]]
+; NO-SINK-AFTER-NEXT:    [[BOUND013:%.*]] = icmp ult i8* [[B3]], [[SCEVGEP89]]
+; NO-SINK-AFTER-NEXT:    [[BOUND114:%.*]] = icmp ult i8* [[SCEVGEP67]], [[SCEVGEP45]]
+; NO-SINK-AFTER-NEXT:    [[FOUND_CONFLICT15:%.*]] = and i1 [[BOUND013]], [[BOUND114]]
+; NO-SINK-AFTER-NEXT:    [[CONFLICT_RDX16:%.*]] = or i1 [[CONFLICT_RDX]], [[FOUND_CONFLICT15]]
+; NO-SINK-AFTER-NEXT:    br i1 [[CONFLICT_RDX16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 3
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[TMP0]]
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP0]], i64 1
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP1]], i64 1
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP2]], i64 1
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP3]], i64 1
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <4 x i32>*
+; NO-SINK-AFTER-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP10]], align 4, !alias.scope !33, !noalias !36
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = load i16, i16* [[TMP5]], align 2, !alias.scope !39
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP6]], align 2, !alias.scope !39
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = load i16, i16* [[TMP7]], align 2, !alias.scope !39
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = load i16, i16* [[TMP8]], align 2, !alias.scope !39
+; NO-SINK-AFTER-NEXT:    [[TMP15:%.*]] = insertelement <4 x i16> poison, i16 [[TMP11]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP16:%.*]] = insertelement <4 x i16> [[TMP15]], i16 [[TMP12]], i32 1
+; NO-SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x i16> [[TMP16]], i16 [[TMP13]], i32 2
+; NO-SINK-AFTER-NEXT:    [[TMP18]] = insertelement <4 x i16> [[TMP17]], i16 [[TMP14]], i32 3
+; NO-SINK-AFTER-NEXT:    [[TMP19:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP18]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP20:%.*]] = sext <4 x i16> [[TMP19]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP21:%.*]] = sext <4 x i16> [[TMP18]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP22:%.*]] = mul nsw <4 x i32> [[TMP21]], [[TMP20]]
+; NO-SINK-AFTER-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP0]]
+; NO-SINK-AFTER-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP24]] to <4 x i32>*
+; NO-SINK-AFTER-NEXT:    store <4 x i32> [[TMP22]], <4 x i32>* [[TMP25]], align 4, !alias.scope !40, !noalias !39
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP18]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP18]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-SINK-AFTER:       for.body:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP27:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
+; NO-SINK-AFTER-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
+; NO-SINK-AFTER-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
+; NO-SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; NO-SINK-AFTER-NEXT:    [[TMP27]] = load i16, i16* [[CUR_INDEX]], align 2
+; NO-SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP27]] to i32
+; NO-SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; NO-SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
+;
 entry:
   %pre.index = getelementptr inbounds [2 x i16], [2 x i16]* %a, i64 0, i64 0
   %.pre = load i16, i16* %pre.index
@@ -546,36 +4305,443 @@ for.end:
 ;     b[i] = ((a[i] + 2) * a[i + 1]);
 ; }
 ;
-; SINK-AFTER-LABEL:  @sink_after_with_multiple_users
-; SINK-AFTER-LABEL:  vector.ph:
-; SINK-AFTER-NEXT:    %n.mod.vf = urem i64 %n, 4
-; SINK-AFTER-NEXT:    %n.vec = sub i64 %n, %n.mod.vf
-; SINK-AFTER-NEXT:    %vector.recur.init = insertelement <4 x i16> poison, i16 %.pre, i32 3
-; SINK-AFTER-NEXT:    br label %vector.body
-
-; SINK-AFTER-LABEL:  vector.body:
-; SINK-AFTER-NEXT:    %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; SINK-AFTER-NEXT:    %vector.recur = phi <4 x i16> [ %vector.recur.init, %vector.ph ], [ %wide.load, %vector.body ]
-; SINK-AFTER-NEXT:    %1 = add i64 %index, 0
-; SINK-AFTER-NEXT:    %2 = add nuw nsw i64 %1, 1
-; SINK-AFTER-NEXT:    %3 = getelementptr inbounds i16, i16* %a, i64 %2
-; SINK-AFTER-NEXT:    %4 = getelementptr inbounds i16, i16* %3, i32 0
-; SINK-AFTER-NEXT:    %5 = bitcast i16* %4 to <4 x i16>*
-; SINK-AFTER-NEXT:    %wide.load = load <4 x i16>, <4 x i16>* %5, align 2, !alias.scope !43
-; SINK-AFTER-NEXT:    %6 = shufflevector <4 x i16> %vector.recur, <4 x i16> %wide.load, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; SINK-AFTER-NEXT:    %7 = sext <4 x i16> %6 to <4 x i32>
-; SINK-AFTER-NEXT:    %8 = add nsw <4 x i32> %7, <i32 2, i32 2, i32 2, i32 2>
-; SINK-AFTER-NEXT:    %9 = sext <4 x i16> %wide.load to <4 x i32>
-; SINK-AFTER-NEXT:    %10 = mul nsw <4 x i32> %8, %9
-; SINK-AFTER-NEXT:    %11 = getelementptr inbounds i32, i32* %b, i64 %1
-; SINK-AFTER-NEXT:    %12 = getelementptr inbounds i32, i32* %11, i32 0
-; SINK-AFTER-NEXT:    %13 = bitcast i32* %12 to <4 x i32>*
-; SINK-AFTER-NEXT:    store <4 x i32> %10, <4 x i32>* %13, align 4, !alias.scope !46, !noalias !43
-; SINK-AFTER-NEXT:    %index.next = add nuw i64 %index, 4
-; SINK-AFTER-NEXT:    %14 = icmp eq i64 %index.next, %n.vec
-; SINK-AFTER-NEXT:    br i1 %14, label %middle.block, label %vector.body, !llvm.loop !48
+
 ;
 define void @sink_after_with_multiple_users(i16* %a, i32* %b, i64 %n) {
+; CHECK-LABEL: @sink_after_with_multiple_users(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK:       vector.memcheck:
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[N]]
+; CHECK-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; CHECK-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[SCEVGEP5]] to i32*
+; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[TMP1]], [[B]]
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[SCEVGEP]] to i16*
+; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult i16* [[SCEVGEP3]], [[TMP2]]
+; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -4
+; CHECK-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP5]], align 2, !alias.scope !43
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; CHECK-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[TMP6]] to <4 x i32>
+; CHECK-NEXT:    [[TMP8:%.*]] = add nsw <4 x i32> [[TMP7]], <i32 2, i32 2, i32 2, i32 2>
+; CHECK-NEXT:    [[TMP9:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; CHECK-NEXT:    [[TMP10:%.*]] = mul nsw <4 x i32> [[TMP8]], [[TMP9]]
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>*
+; CHECK-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4, !alias.scope !46, !noalias !43
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP14:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; CHECK-NEXT:    [[TMP14]] = load i16, i16* [[ARRAYIDX2]], align 2
+; CHECK-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP14]] to i32
+; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
+; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @sink_after_with_multiple_users(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL:       vector.memcheck:
+; UNROLL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[N]]
+; UNROLL-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; UNROLL-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; UNROLL-NEXT:    [[TMP1:%.*]] = bitcast i16* [[SCEVGEP5]] to i32*
+; UNROLL-NEXT:    [[BOUND0:%.*]] = icmp ugt i32* [[TMP1]], [[B]]
+; UNROLL-NEXT:    [[TMP2:%.*]] = bitcast i32* [[SCEVGEP]] to i16*
+; UNROLL-NEXT:    [[BOUND1:%.*]] = icmp ult i16* [[SCEVGEP3]], [[TMP2]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -8
+; UNROLL-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP3]]
+; UNROLL-NEXT:    [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
+; UNROLL-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP5]], align 2, !alias.scope !43
+; UNROLL-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i64 4
+; UNROLL-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP6]] to <4 x i16>*
+; UNROLL-NEXT:    [[WIDE_LOAD7]] = load <4 x i16>, <4 x i16>* [[TMP7]], align 2, !alias.scope !43
+; UNROLL-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP9:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP10:%.*]] = sext <4 x i16> [[TMP8]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP11:%.*]] = sext <4 x i16> [[TMP9]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP12:%.*]] = add nsw <4 x i32> [[TMP10]], <i32 2, i32 2, i32 2, i32 2>
+; UNROLL-NEXT:    [[TMP13:%.*]] = add nsw <4 x i32> [[TMP11]], <i32 2, i32 2, i32 2, i32 2>
+; UNROLL-NEXT:    [[TMP14:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP15:%.*]] = sext <4 x i16> [[WIDE_LOAD7]] to <4 x i32>
+; UNROLL-NEXT:    [[TMP16:%.*]] = mul nsw <4 x i32> [[TMP12]], [[TMP14]]
+; UNROLL-NEXT:    [[TMP17:%.*]] = mul nsw <4 x i32> [[TMP13]], [[TMP15]]
+; UNROLL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP19:%.*]] = bitcast i32* [[TMP18]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> [[TMP16]], <4 x i32>* [[TMP19]], align 4, !alias.scope !46, !noalias !43
+; UNROLL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP18]], i64 4
+; UNROLL-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP20]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> [[TMP17]], <4 x i32>* [[TMP21]], align 4, !alias.scope !46, !noalias !43
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD7]], i32 3
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP23:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NEXT:    [[TMP23]] = load i16, i16* [[ARRAYIDX2]], align 2
+; UNROLL-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP23]] to i32
+; UNROLL-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
+; UNROLL-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @sink_after_with_multiple_users(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-IC:       vector.memcheck:
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP34:%.*]] = bitcast i16* [[SCEVGEP3]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP56:%.*]] = bitcast i16* [[SCEVGEP5]] to i8*
+; UNROLL-NO-IC-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; UNROLL-NO-IC-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-IC-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, i16* [[TMP5]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = bitcast i16* [[TMP7]] to <4 x i16>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP8]], align 2, !alias.scope !43
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, i16* [[TMP5]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = bitcast i16* [[TMP9]] to <4 x i16>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD7]] = load <4 x i16>, <4 x i16>* [[TMP10]], align 2, !alias.scope !43
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = sext <4 x i16> [[TMP11]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = sext <4 x i16> [[TMP12]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP13]], <i32 2, i32 2, i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 2, i32 2, i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = sext <4 x i16> [[WIDE_LOAD7]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = mul nsw <4 x i32> [[TMP15]], [[TMP17]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = mul nsw <4 x i32> [[TMP16]], [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP2]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = bitcast i32* [[TMP23]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP19]], <4 x i32>* [[TMP24]], align 4, !alias.scope !46, !noalias !43
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = bitcast i32* [[TMP25]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP20]], <4 x i32>* [[TMP26]], align 4, !alias.scope !46, !noalias !43
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD7]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD7]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NO-IC-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NO-IC-NEXT:    [[TMP28]] = load i16, i16* [[ARRAYIDX2]], align 2
+; UNROLL-NO-IC-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP28]] to i32
+; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @sink_after_with_multiple_users(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-VF:       vector.memcheck:
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP34:%.*]] = bitcast i16* [[SCEVGEP3]] to i8*
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; UNROLL-NO-VF-NEXT:    [[SCEVGEP56:%.*]] = bitcast i16* [[SCEVGEP5]] to i8*
+; UNROLL-NO-VF-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; UNROLL-NO-VF-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; UNROLL-NO-VF-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-VF-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION7:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDUCTION]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[INDUCTION7]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP1]]
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP2]]
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = load i16, i16* [[TMP3]], align 2, !alias.scope !42
+; UNROLL-NO-VF-NEXT:    [[TMP6]] = load i16, i16* [[TMP4]], align 2, !alias.scope !42
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sext i16 [[VECTOR_RECUR]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP5]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = add nsw i32 [[TMP7]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = add nsw i32 [[TMP8]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = sext i16 [[TMP5]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = sext i16 [[TMP6]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = mul nsw i32 [[TMP9]], [[TMP11]]
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = mul nsw i32 [[TMP10]], [[TMP12]]
+; UNROLL-NO-VF-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION7]]
+; UNROLL-NO-VF-NEXT:    store i32 [[TMP13]], i32* [[TMP15]], align 4, !alias.scope !45, !noalias !42
+; UNROLL-NO-VF-NEXT:    store i32 [[TMP14]], i32* [[TMP16]], align 4, !alias.scope !45, !noalias !42
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-VF:       for.body:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; UNROLL-NO-VF-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; UNROLL-NO-VF-NEXT:    [[TMP18]] = load i16, i16* [[ARRAYIDX2]], align 2
+; UNROLL-NO-VF-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP18]] to i32
+; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; UNROLL-NO-VF-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @sink_after_with_multiple_users(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; SINK-AFTER:       vector.memcheck:
+; SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; SINK-AFTER-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; SINK-AFTER-NEXT:    [[SCEVGEP34:%.*]] = bitcast i16* [[SCEVGEP3]] to i8*
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; SINK-AFTER-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; SINK-AFTER-NEXT:    [[SCEVGEP56:%.*]] = bitcast i16* [[SCEVGEP5]] to i8*
+; SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; SINK-AFTER-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP2]]
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 0
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP5]], align 2, !alias.scope !43
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[TMP6]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = add nsw <4 x i32> [[TMP7]], <i32 2, i32 2, i32 2, i32 2>
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = mul nsw <4 x i32> [[TMP8]], [[TMP9]]
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP1]]
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[TMP11]], i32 0
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = bitcast i32* [[TMP12]] to <4 x i32>*
+; SINK-AFTER-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP13]], align 4, !alias.scope !46, !noalias !43
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; SINK-AFTER:       for.body:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; SINK-AFTER-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
+; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SINK-AFTER-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; SINK-AFTER-NEXT:    [[TMP15]] = load i16, i16* [[ARRAYIDX2]], align 2
+; SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP15]] to i32
+; SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
+; SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @sink_after_with_multiple_users(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[B1:%.*]] = bitcast i32* [[B:%.*]] to i8*
+; NO-SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
+; NO-SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; NO-SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; NO-SINK-AFTER:       vector.memcheck:
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP:%.*]] = getelementptr i32, i32* [[B]], i64 [[N]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP2:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i16, i16* [[A]], i64 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP34:%.*]] = bitcast i16* [[SCEVGEP3]] to i8*
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i16, i16* [[A]], i64 [[TMP0]]
+; NO-SINK-AFTER-NEXT:    [[SCEVGEP56:%.*]] = bitcast i16* [[SCEVGEP5]] to i8*
+; NO-SINK-AFTER-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[B1]], [[SCEVGEP56]]
+; NO-SINK-AFTER-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]]
+; NO-SINK-AFTER-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-SINK-AFTER-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP2]]
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
+; NO-SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP5]], align 2, !alias.scope !43
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[TMP6]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = add nsw <4 x i32> [[TMP7]], <i32 2, i32 2, i32 2, i32 2>
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = mul nsw <4 x i32> [[TMP8]], [[TMP9]]
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP1]]
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[TMP11]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = bitcast i32* [[TMP12]] to <4 x i32>*
+; NO-SINK-AFTER-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP13]], align 4, !alias.scope !46, !noalias !43
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_MEMCHECK]] ], [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-SINK-AFTER:       for.body:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
+; NO-SINK-AFTER-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
+; NO-SINK-AFTER-NEXT:    [[TMP15]] = load i16, i16* [[ARRAYIDX2]], align 2
+; NO-SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP15]] to i32
+; NO-SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
+; NO-SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
+;
 entry:
   %.pre = load i16, i16* %a
   br label %for.body
@@ -604,9 +4770,148 @@ for.end:
 ; branches will become users of the condition bit after vectorization and would
 ; need to be sunk if the loop is vectorized.
 define void @do_not_sink_branch(i32 %x, i32* %in, i32* %out, i32 %tc) local_unnamed_addr #0 {
-; NO-SINK-AFTER-LABEL: do_not_sink_branch
-; NO-SINK-AFTER-NOT:   vector.ph:
-; NO-SINK-AFTER:       }
+; CHECK-LABEL: @do_not_sink_branch(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP530:%.*]] = icmp sgt i32 [[TC:%.*]], 0
+; CHECK-NEXT:    br label [[FOR_BODY4:%.*]]
+; CHECK:       for.body4:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
+; CHECK-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
+; CHECK-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
+; CHECK:       cond.true:
+; CHECK-NEXT:    [[TMP0:%.*]] = zext i32 [[INDVARS_IV]] to i64
+; CHECK-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
+; CHECK-NEXT:    br label [[COND_END]]
+; CHECK:       cond.end:
+; CHECK-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[INDVARS_IV]] to i64
+; CHECK-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i64 [[TMP1]]
+; CHECK-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
+; CHECK:       for.end12.loopexit:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @do_not_sink_branch(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[CMP530:%.*]] = icmp sgt i32 [[TC:%.*]], 0
+; UNROLL-NEXT:    br label [[FOR_BODY4:%.*]]
+; UNROLL:       for.body4:
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
+; UNROLL-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
+; UNROLL-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
+; UNROLL:       cond.true:
+; UNROLL-NEXT:    [[TMP0:%.*]] = zext i32 [[INDVARS_IV]] to i64
+; UNROLL-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i64 [[TMP0]]
+; UNROLL-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
+; UNROLL-NEXT:    br label [[COND_END]]
+; UNROLL:       cond.end:
+; UNROLL-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
+; UNROLL-NEXT:    [[TMP1:%.*]] = zext i32 [[INDVARS_IV]] to i64
+; UNROLL-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i64 [[TMP1]]
+; UNROLL-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
+; UNROLL:       for.end12.loopexit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @do_not_sink_branch(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[CMP530:%.*]] = icmp slt i32 0, [[TC:%.*]]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY4:%.*]]
+; UNROLL-NO-IC:       for.body4:
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
+; UNROLL-NO-IC:       cond.true:
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i32 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[COND_END]]
+; UNROLL-NO-IC:       cond.end:
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i32 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
+; UNROLL-NO-IC:       for.end12.loopexit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @do_not_sink_branch(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    [[CMP530:%.*]] = icmp slt i32 0, [[TC:%.*]]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY4:%.*]]
+; UNROLL-NO-VF:       for.body4:
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
+; UNROLL-NO-VF:       cond.true:
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i32 [[INDVARS_IV]]
+; UNROLL-NO-VF-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
+; UNROLL-NO-VF-NEXT:    br label [[COND_END]]
+; UNROLL-NO-VF:       cond.end:
+; UNROLL-NO-VF-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
+; UNROLL-NO-VF-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i32 [[INDVARS_IV]]
+; UNROLL-NO-VF-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
+; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; UNROLL-NO-VF-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
+; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
+; UNROLL-NO-VF:       for.end12.loopexit:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @do_not_sink_branch(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    [[CMP530:%.*]] = icmp slt i32 0, [[TC:%.*]]
+; SINK-AFTER-NEXT:    br label [[FOR_BODY4:%.*]]
+; SINK-AFTER:       for.body4:
+; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
+; SINK-AFTER-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
+; SINK-AFTER-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
+; SINK-AFTER:       cond.true:
+; SINK-AFTER-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i32 [[INDVARS_IV]]
+; SINK-AFTER-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
+; SINK-AFTER-NEXT:    br label [[COND_END]]
+; SINK-AFTER:       cond.end:
+; SINK-AFTER-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
+; SINK-AFTER-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i32 [[INDVARS_IV]]
+; SINK-AFTER-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
+; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; SINK-AFTER-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
+; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
+; SINK-AFTER:       for.end12.loopexit:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @do_not_sink_branch(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    [[CMP530:%.*]] = icmp slt i32 0, [[TC:%.*]]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_BODY4:%.*]]
+; NO-SINK-AFTER:       for.body4:
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
+; NO-SINK-AFTER-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
+; NO-SINK-AFTER:       cond.true:
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i32 [[INDVARS_IV]]
+; NO-SINK-AFTER-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
+; NO-SINK-AFTER-NEXT:    br label [[COND_END]]
+; NO-SINK-AFTER:       cond.end:
+; NO-SINK-AFTER-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
+; NO-SINK-AFTER-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i32 [[INDVARS_IV]]
+; NO-SINK-AFTER-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
+; NO-SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; NO-SINK-AFTER-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
+; NO-SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
+; NO-SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
+; NO-SINK-AFTER:       for.end12.loopexit:
+; NO-SINK-AFTER-NEXT:    ret void
+;
 entry:
   %cmp530 = icmp slt i32 0, %tc
   br label %for.body4
@@ -637,22 +4942,253 @@ for.end12.loopexit:                               ; preds = %cond.end
 ; Dead instructions, like the exit condition are not part of the actual VPlan
 ; and do not need to be sunk. PR44634.
 define void @sink_dead_inst() {
-; SINK-AFTER-LABEL: define void @sink_dead_inst(
-; SINK-AFTER-LABEL: vector.body:                                      ; preds = %vector.body, %vector.ph
-; SINK-AFTER-NEXT:    %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; SINK-AFTER-NEXT:    %vec.ind = phi <4 x i16> [ <i16 -27, i16 -26, i16 -25, i16 -24>, %vector.ph ], [ %vec.ind.next, %vector.body ]
-; SINK-AFTER-NEXT:    %vector.recur = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, %vector.ph ], [ %3, %vector.body ]
-; SINK-AFTER-NEXT:    %vector.recur1 = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 -27>, %vector.ph ], [ %1, %vector.body ]
-; SINK-AFTER-NEXT:    %0 = add <4 x i16> %vec.ind, <i16 1, i16 1, i16 1, i16 1>
-; SINK-AFTER-NEXT:    %1 = zext <4 x i16> %0 to <4 x i32>
-; SINK-AFTER-NEXT:    %2 = shufflevector <4 x i32> %vector.recur1, <4 x i32> %1, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; SINK-AFTER-NEXT:    %3 = add <4 x i16> %0, <i16 5, i16 5, i16 5, i16 5>
-; SINK-AFTER-NEXT:    %4 = shufflevector <4 x i16> %vector.recur, <4 x i16> %3, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; SINK-AFTER-NEXT:    %5 = sub <4 x i16> %4, <i16 10, i16 10, i16 10, i16 10>
-; SINK-AFTER-NEXT:    %index.next = add nuw i32 %index, 4
-; SINK-AFTER-NEXT:    %vec.ind.next = add <4 x i16> %vec.ind, <i16 4, i16 4, i16 4, i16 4>
-; SINK-AFTER-NEXT:    %6 = icmp eq i32 %index.next, 40
-; SINK-AFTER-NEXT:    br i1 %6, label %middle.block, label %vector.body, !llvm.loop !50
+;
+; CHECK-LABEL: @sink_dead_inst(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i16 [ -24, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP1]] = add i16 [[TMP0]], 4
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
+; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP3:%.*]] = or i16 [[TMP0]], 1
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = zext i16 [[TMP3]] to i32
+; CHECK-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT2]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ -27, [[ENTRY]] ], [ 13, [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR5:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT4]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR5]], 15
+; CHECK-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; CHECK-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @sink_dead_inst(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP0:%.*]] = phi i16 [ -24, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP1]] = add i16 [[TMP0]], 8
+; UNROLL-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
+; UNROLL-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[TMP3:%.*]] = or i16 [[TMP0]], 5
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = zext i16 [[TMP3]] to i32
+; UNROLL-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT5:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT3]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ -27, [[ENTRY]] ], [ 13, [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[FOR_COND:%.*]]
+; UNROLL:       for.cond:
+; UNROLL-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
+; UNROLL-NEXT:    [[SCALAR_RECUR6:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT5]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
+; UNROLL-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR6]], 15
+; UNROLL-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; UNROLL-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
+; UNROLL-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @sink_dead_inst(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 -27, i16 -26, i16 -25, i16 -24>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 -27>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add <4 x i16> [[STEP_ADD]], <i16 1, i16 1, i16 1, i16 1>
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = zext <4 x i16> [[TMP0]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP3]] = zext <4 x i16> [[TMP1]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR2]], <4 x i32> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add <4 x i16> [[TMP0]], <i16 5, i16 5, i16 5, i16 5>
+; UNROLL-NO-IC-NEXT:    [[TMP7]] = add <4 x i16> [[TMP1]], <i16 5, i16 5, i16 5, i16 5>
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP6]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = shufflevector <4 x i16> [[TMP6]], <4 x i16> [[TMP7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = sub <4 x i16> [[TMP8]], <i16 10, i16 10, i16 10, i16 10>
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = sub <4 x i16> [[TMP9]], <i16 10, i16 10, i16 10, i16 10>
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], <i16 4, i16 4, i16 4, i16 4>
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 43, 40
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP7]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP7]], i32 2
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI4:%.*]] = extractelement <4 x i32> [[TMP3]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT5:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT3]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 13, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_COND:%.*]]
+; UNROLL-NO-IC:       for.cond:
+; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[REC_1_PREV:%.*]], [[FOR_COND]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR6:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT5]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
+; UNROLL-NO-IC-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[SCALAR_RECUR]], 10
+; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR6]], 15
+; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; UNROLL-NO-IC-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
+; UNROLL-NO-IC-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @sink_dead_inst(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ 0, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR2:%.*]] = phi i32 [ -27, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = trunc i32 [[INDEX]] to i16
+; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = add i16 -27, [[TMP0]]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i16 [[OFFSET_IDX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i16 [[OFFSET_IDX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i16 [[INDUCTION]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i16 [[INDUCTION1]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = zext i16 [[TMP1]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP4]] = zext i16 [[TMP2]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = add i16 [[TMP1]], 5
+; UNROLL-NO-VF-NEXT:    [[TMP6]] = add i16 [[TMP2]], 5
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sub i16 [[VECTOR_RECUR]], 10
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sub i16 [[TMP5]], 10
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], 42
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i32 43, 42
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT3:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[TMP4]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 15, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_COND:%.*]]
+; UNROLL-NO-VF:       for.cond:
+; UNROLL-NO-VF-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[REC_1_PREV:%.*]], [[FOR_COND]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR4:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT3]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
+; UNROLL-NO-VF-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[SCALAR_RECUR]], 10
+; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR4]], 15
+; UNROLL-NO-VF-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; UNROLL-NO-VF-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
+; UNROLL-NO-VF-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP50:![0-9]+]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @sink_dead_inst(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 -27, i16 -26, i16 -25, i16 -24>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 -27>, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
+; SINK-AFTER-NEXT:    [[TMP1]] = zext <4 x i16> [[TMP0]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP3]] = add <4 x i16> [[TMP0]], <i16 5, i16 5, i16 5, i16 5>
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = sub <4 x i16> [[TMP4]], <i16 10, i16 10, i16 10, i16 10>
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
+; SINK-AFTER-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 43, 40
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP3]], i32 2
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x i32> [[TMP1]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI3:%.*]] = extractelement <4 x i32> [[TMP1]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT2]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 13, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    br label [[FOR_COND:%.*]]
+; SINK-AFTER:       for.cond:
+; SINK-AFTER-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[REC_1_PREV:%.*]], [[FOR_COND]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR5:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT4]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
+; SINK-AFTER-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[SCALAR_RECUR]], 10
+; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR5]], 15
+; SINK-AFTER-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; SINK-AFTER-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
+; SINK-AFTER-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
+; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @sink_dead_inst(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 -27, i16 -26, i16 -25, i16 -24>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 -27>, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
+; NO-SINK-AFTER-NEXT:    [[TMP1]] = zext <4 x i16> [[TMP0]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP3]] = add <4 x i16> [[TMP0]], <i16 5, i16 5, i16 5, i16 5>
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = sub <4 x i16> [[TMP4]], <i16 10, i16 10, i16 10, i16 10>
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 43, 40
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP3]], i32 2
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x i32> [[TMP1]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI3:%.*]] = extractelement <4 x i32> [[TMP1]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT2]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 13, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY]] ]
+; NO-SINK-AFTER-NEXT:    br label [[FOR_COND:%.*]]
+; NO-SINK-AFTER:       for.cond:
+; NO-SINK-AFTER-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[REC_1_PREV:%.*]], [[FOR_COND]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR5:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT4]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
+; NO-SINK-AFTER-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[SCALAR_RECUR]], 10
+; NO-SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR5]], 15
+; NO-SINK-AFTER-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; NO-SINK-AFTER-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
+; NO-SINK-AFTER-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
 ;
 entry:
   br label %for.cond
@@ -673,6 +5209,7 @@ for.end:
 }
 
 define i32 @sink_into_replication_region(i32 %y) {
+;
 ; CHECK-LABEL: @sink_into_replication_region(
 ; CHECK-NEXT:  bb:
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
@@ -687,13 +5224,13 @@ define i32 @sink_into_replication_region(i32 %y) {
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE9:%.*]] ]
-; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[PRED_UDIV_CONTINUE9]] ]
-; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[PRED_UDIV_CONTINUE9]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE8:%.*]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[PRED_UDIV_CONTINUE8]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[PRED_UDIV_CONTINUE8]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i32 0
-; CHECK-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT2]], <4 x i32> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[VEC_IV:%.*]] = or <4 x i32> [[BROADCAST_SPLAT3]], <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT1]], <4 x i32> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[VEC_IV:%.*]] = or <4 x i32> [[BROADCAST_SPLAT2]], <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp ule <4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
@@ -704,68 +5241,577 @@ define i32 @sink_into_replication_region(i32 %y) {
 ; CHECK:       pred.udiv.continue:
 ; CHECK-NEXT:    [[TMP6:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP5]], [[PRED_UDIV_IF]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
-; CHECK-NEXT:    br i1 [[TMP7]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]
 ; CHECK:       pred.udiv.if3:
 ; CHECK-NEXT:    [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], -1
 ; CHECK-NEXT:    [[TMP9:%.*]] = udiv i32 219220132, [[TMP8]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = insertelement <4 x i32> [[TMP6]], i32 [[TMP9]], i32 1
-; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
+; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE4]]
 ; CHECK:       pred.udiv.continue4:
-; CHECK-NEXT:    [[TMP11:%.*]] = phi <4 x i32> [ [[TMP6]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP10]], [[PRED_UDIV_IF4]] ]
+; CHECK-NEXT:    [[TMP11:%.*]] = phi <4 x i32> [ [[TMP6]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP10]], [[PRED_UDIV_IF3]] ]
 ; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
-; CHECK-NEXT:    br i1 [[TMP12]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
+; CHECK-NEXT:    br i1 [[TMP12]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]
 ; CHECK:       pred.udiv.if5:
 ; CHECK-NEXT:    [[TMP13:%.*]] = add i32 [[OFFSET_IDX]], -2
 ; CHECK-NEXT:    [[TMP14:%.*]] = udiv i32 219220132, [[TMP13]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP14]], i32 2
-; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
+; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE6]]
 ; CHECK:       pred.udiv.continue6:
-; CHECK-NEXT:    [[TMP16:%.*]] = phi <4 x i32> [ [[TMP11]], [[PRED_UDIV_CONTINUE5]] ], [ [[TMP15]], [[PRED_UDIV_IF6]] ]
+; CHECK-NEXT:    [[TMP16:%.*]] = phi <4 x i32> [ [[TMP11]], [[PRED_UDIV_CONTINUE4]] ], [ [[TMP15]], [[PRED_UDIV_IF5]] ]
 ; CHECK-NEXT:    [[TMP17:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
-; CHECK-NEXT:    br i1 [[TMP17]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9]]
+; CHECK-NEXT:    br i1 [[TMP17]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8]]
 ; CHECK:       pred.udiv.if7:
 ; CHECK-NEXT:    [[TMP18:%.*]] = add i32 [[OFFSET_IDX]], -3
 ; CHECK-NEXT:    [[TMP19:%.*]] = udiv i32 219220132, [[TMP18]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = insertelement <4 x i32> [[TMP16]], i32 [[TMP19]], i32 3
-; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE8]]
 ; CHECK:       pred.udiv.continue8:
-; CHECK-NEXT:    [[TMP21]] = phi <4 x i32> [ [[TMP16]], [[PRED_UDIV_CONTINUE7]] ], [ [[TMP20]], [[PRED_UDIV_IF8]] ]
+; CHECK-NEXT:    [[TMP21]] = phi <4 x i32> [ [[TMP16]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP20]], [[PRED_UDIV_IF7]] ]
 ; CHECK-NEXT:    [[TMP22:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP21]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NEXT:    [[TMP23]] = add <4 x i32> [[VEC_PHI1]], [[TMP22]]
+; CHECK-NEXT:    [[TMP23]] = add <4 x i32> [[VEC_PHI]], [[TMP22]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof !52, [[LOOP46:!llvm.loop !.*]]
+; CHECK-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52:![0-9]+]], !llvm.loop [[LOOP53:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP25:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP23]], <4 x i32> [[VEC_PHI1]]
+; CHECK-NEXT:    [[TMP25:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP23]], <4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP26:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP25]])
 ; CHECK-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[BB2:%.*]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[TMP:%.*]] = phi i32 [ undef, [[BB2]] ], [ [[TMP26]], [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    ret i32 [[TMP]]
+; CHECK-NEXT:    [[VAR:%.*]] = phi i32 [ undef, [[BB2]] ], [ [[TMP26]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[VAR]]
 ; CHECK:       bb2:
-; CHECK-NEXT:    br i1 undef, label [[BB1]], label [[BB2]], !prof !54, [[LOOP48:!llvm.loop !.*]]
+; CHECK-NEXT:    br i1 undef, label [[BB1]], label [[BB2]], !prof [[PROF54:![0-9]+]], !llvm.loop [[LOOP55:![0-9]+]]
+;
+; UNROLL-LABEL: @sink_into_replication_region(
+; UNROLL-NEXT:  bb:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; UNROLL-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; UNROLL-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 7
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i32 [[N_RND_UP]], -8
+; UNROLL-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = add i32 [[TMP1]], -1
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE19:%.*]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP42:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP45:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; UNROLL-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP46:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT3]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NEXT:    [[VEC_IV:%.*]] = or <4 x i32> [[BROADCAST_SPLAT4]], <i32 0, i32 1, i32 2, i32 3>
+; UNROLL-NEXT:    [[VEC_IV5:%.*]] = or <4 x i32> [[BROADCAST_SPLAT4]], <i32 4, i32 5, i32 6, i32 7>
+; UNROLL-NEXT:    [[TMP2:%.*]] = icmp ule <4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT]]
+; UNROLL-NEXT:    [[TMP3:%.*]] = icmp ule <4 x i32> [[VEC_IV5]], [[BROADCAST_SPLAT]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
+; UNROLL-NEXT:    br i1 [[TMP4]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; UNROLL:       pred.udiv.if:
+; UNROLL-NEXT:    [[TMP5:%.*]] = udiv i32 219220132, [[OFFSET_IDX]]
+; UNROLL-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32> poison, i32 [[TMP5]], i32 0
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; UNROLL:       pred.udiv.continue:
+; UNROLL-NEXT:    [[TMP7:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP6]], [[PRED_UDIV_IF]] ]
+; UNROLL-NEXT:    [[TMP8:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
+; UNROLL-NEXT:    br i1 [[TMP8]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
+; UNROLL:       pred.udiv.if6:
+; UNROLL-NEXT:    [[TMP9:%.*]] = add i32 [[OFFSET_IDX]], -1
+; UNROLL-NEXT:    [[TMP10:%.*]] = udiv i32 219220132, [[TMP9]]
+; UNROLL-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> [[TMP7]], i32 [[TMP10]], i32 1
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
+; UNROLL:       pred.udiv.continue7:
+; UNROLL-NEXT:    [[TMP12:%.*]] = phi <4 x i32> [ [[TMP7]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP11]], [[PRED_UDIV_IF6]] ]
+; UNROLL-NEXT:    [[TMP13:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
+; UNROLL-NEXT:    br i1 [[TMP13]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; UNROLL:       pred.udiv.if8:
+; UNROLL-NEXT:    [[TMP14:%.*]] = add i32 [[OFFSET_IDX]], -2
+; UNROLL-NEXT:    [[TMP15:%.*]] = udiv i32 219220132, [[TMP14]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP15]], i32 2
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; UNROLL:       pred.udiv.continue9:
+; UNROLL-NEXT:    [[TMP17:%.*]] = phi <4 x i32> [ [[TMP12]], [[PRED_UDIV_CONTINUE7]] ], [ [[TMP16]], [[PRED_UDIV_IF8]] ]
+; UNROLL-NEXT:    [[TMP18:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
+; UNROLL-NEXT:    br i1 [[TMP18]], label [[PRED_UDIV_IF10:%.*]], label [[PRED_UDIV_CONTINUE11:%.*]]
+; UNROLL:       pred.udiv.if10:
+; UNROLL-NEXT:    [[TMP19:%.*]] = add i32 [[OFFSET_IDX]], -3
+; UNROLL-NEXT:    [[TMP20:%.*]] = udiv i32 219220132, [[TMP19]]
+; UNROLL-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP20]], i32 3
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE11]]
+; UNROLL:       pred.udiv.continue11:
+; UNROLL-NEXT:    [[TMP22:%.*]] = phi <4 x i32> [ [[TMP17]], [[PRED_UDIV_CONTINUE9]] ], [ [[TMP21]], [[PRED_UDIV_IF10]] ]
+; UNROLL-NEXT:    [[TMP23:%.*]] = extractelement <4 x i1> [[TMP3]], i32 0
+; UNROLL-NEXT:    br i1 [[TMP23]], label [[PRED_UDIV_IF12:%.*]], label [[PRED_UDIV_CONTINUE13:%.*]]
+; UNROLL:       pred.udiv.if12:
+; UNROLL-NEXT:    [[TMP24:%.*]] = add i32 [[OFFSET_IDX]], -4
+; UNROLL-NEXT:    [[TMP25:%.*]] = udiv i32 219220132, [[TMP24]]
+; UNROLL-NEXT:    [[TMP26:%.*]] = insertelement <4 x i32> poison, i32 [[TMP25]], i32 0
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE13]]
+; UNROLL:       pred.udiv.continue13:
+; UNROLL-NEXT:    [[TMP27:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE11]] ], [ [[TMP26]], [[PRED_UDIV_IF12]] ]
+; UNROLL-NEXT:    [[TMP28:%.*]] = extractelement <4 x i1> [[TMP3]], i32 1
+; UNROLL-NEXT:    br i1 [[TMP28]], label [[PRED_UDIV_IF14:%.*]], label [[PRED_UDIV_CONTINUE15:%.*]]
+; UNROLL:       pred.udiv.if14:
+; UNROLL-NEXT:    [[TMP29:%.*]] = add i32 [[OFFSET_IDX]], -5
+; UNROLL-NEXT:    [[TMP30:%.*]] = udiv i32 219220132, [[TMP29]]
+; UNROLL-NEXT:    [[TMP31:%.*]] = insertelement <4 x i32> [[TMP27]], i32 [[TMP30]], i32 1
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE15]]
+; UNROLL:       pred.udiv.continue15:
+; UNROLL-NEXT:    [[TMP32:%.*]] = phi <4 x i32> [ [[TMP27]], [[PRED_UDIV_CONTINUE13]] ], [ [[TMP31]], [[PRED_UDIV_IF14]] ]
+; UNROLL-NEXT:    [[TMP33:%.*]] = extractelement <4 x i1> [[TMP3]], i32 2
+; UNROLL-NEXT:    br i1 [[TMP33]], label [[PRED_UDIV_IF16:%.*]], label [[PRED_UDIV_CONTINUE17:%.*]]
+; UNROLL:       pred.udiv.if16:
+; UNROLL-NEXT:    [[TMP34:%.*]] = add i32 [[OFFSET_IDX]], -6
+; UNROLL-NEXT:    [[TMP35:%.*]] = udiv i32 219220132, [[TMP34]]
+; UNROLL-NEXT:    [[TMP36:%.*]] = insertelement <4 x i32> [[TMP32]], i32 [[TMP35]], i32 2
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE17]]
+; UNROLL:       pred.udiv.continue17:
+; UNROLL-NEXT:    [[TMP37:%.*]] = phi <4 x i32> [ [[TMP32]], [[PRED_UDIV_CONTINUE15]] ], [ [[TMP36]], [[PRED_UDIV_IF16]] ]
+; UNROLL-NEXT:    [[TMP38:%.*]] = extractelement <4 x i1> [[TMP3]], i32 3
+; UNROLL-NEXT:    br i1 [[TMP38]], label [[PRED_UDIV_IF18:%.*]], label [[PRED_UDIV_CONTINUE19]]
+; UNROLL:       pred.udiv.if18:
+; UNROLL-NEXT:    [[TMP39:%.*]] = add i32 [[OFFSET_IDX]], -7
+; UNROLL-NEXT:    [[TMP40:%.*]] = udiv i32 219220132, [[TMP39]]
+; UNROLL-NEXT:    [[TMP41:%.*]] = insertelement <4 x i32> [[TMP37]], i32 [[TMP40]], i32 3
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE19]]
+; UNROLL:       pred.udiv.continue19:
+; UNROLL-NEXT:    [[TMP42]] = phi <4 x i32> [ [[TMP37]], [[PRED_UDIV_CONTINUE17]] ], [ [[TMP41]], [[PRED_UDIV_IF18]] ]
+; UNROLL-NEXT:    [[TMP43:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[TMP22]], <4 x i32> [[TMP42]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP45]] = add <4 x i32> [[VEC_PHI]], [[TMP43]]
+; UNROLL-NEXT:    [[TMP46]] = add <4 x i32> [[VEC_PHI2]], [[TMP44]]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP47:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP47]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52:![0-9]+]], !llvm.loop [[LOOP53:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[TMP48:%.*]] = select <4 x i1> [[TMP3]], <4 x i32> [[TMP46]], <4 x i32> [[VEC_PHI2]]
+; UNROLL-NEXT:    [[TMP49:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP45]], <4 x i32> [[VEC_PHI]]
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP48]], [[TMP49]]
+; UNROLL-NEXT:    [[TMP50:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; UNROLL-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[BB2:%.*]]
+; UNROLL:       bb1:
+; UNROLL-NEXT:    [[VAR:%.*]] = phi i32 [ undef, [[BB2]] ], [ [[TMP50]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i32 [[VAR]]
+; UNROLL:       bb2:
+; UNROLL-NEXT:    br i1 undef, label [[BB1]], label [[BB2]], !prof [[PROF54:![0-9]+]], !llvm.loop [[LOOP55:![0-9]+]]
+;
+; UNROLL-NO-IC-LABEL: @sink_into_replication_region(
+; UNROLL-NO-IC-NEXT:  bb:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; UNROLL-NO-IC-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 7
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 -1, i32 -2, i32 -3>
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE19:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP43:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP46:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 -4, i32 -4, i32 -4, i32 -4>
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT3]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[VEC_IV:%.*]] = add <4 x i32> [[BROADCAST_SPLAT4]], <i32 0, i32 1, i32 2, i32 3>
+; UNROLL-NO-IC-NEXT:    [[VEC_IV5:%.*]] = add <4 x i32> [[BROADCAST_SPLAT4]], <i32 4, i32 5, i32 6, i32 7>
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = icmp ule <4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = icmp ule <4 x i32> [[VEC_IV5]], [[BROADCAST_SPLAT]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP4]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if:
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = udiv i32 219220132, [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = insertelement <4 x i32> poison, i32 [[TMP6]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; UNROLL-NO-IC:       pred.udiv.continue:
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UDIV_IF]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP9]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if6:
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add i32 [[OFFSET_IDX]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = udiv i32 219220132, [[TMP10]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP8]], i32 [[TMP11]], i32 1
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
+; UNROLL-NO-IC:       pred.udiv.continue7:
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = phi <4 x i32> [ [[TMP8]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP12]], [[PRED_UDIV_IF6]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP14]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if8:
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = add i32 [[OFFSET_IDX]], -2
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = udiv i32 219220132, [[TMP15]]
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP16]], i32 2
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; UNROLL-NO-IC:       pred.udiv.continue9:
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = phi <4 x i32> [ [[TMP13]], [[PRED_UDIV_CONTINUE7]] ], [ [[TMP17]], [[PRED_UDIV_IF8]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP19]], label [[PRED_UDIV_IF10:%.*]], label [[PRED_UDIV_CONTINUE11:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if10:
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = add i32 [[OFFSET_IDX]], -3
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = udiv i32 219220132, [[TMP20]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP21]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE11]]
+; UNROLL-NO-IC:       pred.udiv.continue11:
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = phi <4 x i32> [ [[TMP18]], [[PRED_UDIV_CONTINUE9]] ], [ [[TMP22]], [[PRED_UDIV_IF10]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = extractelement <4 x i1> [[TMP3]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP24]], label [[PRED_UDIV_IF12:%.*]], label [[PRED_UDIV_CONTINUE13:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if12:
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = add i32 [[OFFSET_IDX]], -4
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = udiv i32 219220132, [[TMP25]]
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = insertelement <4 x i32> poison, i32 [[TMP26]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE13]]
+; UNROLL-NO-IC:       pred.udiv.continue13:
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE11]] ], [ [[TMP27]], [[PRED_UDIV_IF12]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = extractelement <4 x i1> [[TMP3]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP29]], label [[PRED_UDIV_IF14:%.*]], label [[PRED_UDIV_CONTINUE15:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if14:
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = add i32 [[OFFSET_IDX]], -5
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = udiv i32 219220132, [[TMP30]]
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = insertelement <4 x i32> [[TMP28]], i32 [[TMP31]], i32 1
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE15]]
+; UNROLL-NO-IC:       pred.udiv.continue15:
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = phi <4 x i32> [ [[TMP28]], [[PRED_UDIV_CONTINUE13]] ], [ [[TMP32]], [[PRED_UDIV_IF14]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = extractelement <4 x i1> [[TMP3]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP34]], label [[PRED_UDIV_IF16:%.*]], label [[PRED_UDIV_CONTINUE17:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if16:
+; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = add i32 [[OFFSET_IDX]], -6
+; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = udiv i32 219220132, [[TMP35]]
+; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = insertelement <4 x i32> [[TMP33]], i32 [[TMP36]], i32 2
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE17]]
+; UNROLL-NO-IC:       pred.udiv.continue17:
+; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = phi <4 x i32> [ [[TMP33]], [[PRED_UDIV_CONTINUE15]] ], [ [[TMP37]], [[PRED_UDIV_IF16]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = extractelement <4 x i1> [[TMP3]], i32 3
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP39]], label [[PRED_UDIV_IF18:%.*]], label [[PRED_UDIV_CONTINUE19]]
+; UNROLL-NO-IC:       pred.udiv.if18:
+; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = add i32 [[OFFSET_IDX]], -7
+; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = udiv i32 219220132, [[TMP40]]
+; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = insertelement <4 x i32> [[TMP38]], i32 [[TMP41]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE19]]
+; UNROLL-NO-IC:       pred.udiv.continue19:
+; UNROLL-NO-IC-NEXT:    [[TMP43]] = phi <4 x i32> [ [[TMP38]], [[PRED_UDIV_CONTINUE17]] ], [ [[TMP42]], [[PRED_UDIV_IF18]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP23]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = shufflevector <4 x i32> [[TMP23]], <4 x i32> [[TMP43]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP46]] = add <4 x i32> [[VEC_PHI]], [[TMP44]]
+; UNROLL-NO-IC-NEXT:    [[TMP47]] = add <4 x i32> [[VEC_PHI2]], [[TMP45]]
+; UNROLL-NO-IC-NEXT:    [[TMP48:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP46]], <4 x i32> [[VEC_PHI]]
+; UNROLL-NO-IC-NEXT:    [[TMP49:%.*]] = select <4 x i1> [[TMP3]], <4 x i32> [[TMP47]], <4 x i32> [[VEC_PHI2]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], <i32 -4, i32 -4, i32 -4, i32 -4>
+; UNROLL-NO-IC-NEXT:    [[TMP50:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP50]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52:![0-9]+]], !llvm.loop [[LOOP53:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP49]], [[TMP48]]
+; UNROLL-NO-IC-NEXT:    [[TMP51:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP43]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP43]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[BB2:%.*]]
+; UNROLL-NO-IC:       bb1:
+; UNROLL-NO-IC-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[VAR]]
+; UNROLL-NO-IC:       bb2:
+; UNROLL-NO-IC-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
+; UNROLL-NO-IC-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
+; UNROLL-NO-IC-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
+; UNROLL-NO-IC-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
+; UNROLL-NO-IC-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF54:![0-9]+]], !llvm.loop [[LOOP55:![0-9]+]]
+;
+; UNROLL-NO-VF-LABEL: @sink_into_replication_region(
+; UNROLL-NO-VF-NEXT:  bb:
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; UNROLL-NO-VF-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 1
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE5:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[PRED_UDIV_CONTINUE5]] ]
+; UNROLL-NO-VF-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[PRED_UDIV_CONTINUE5]] ]
+; UNROLL-NO-VF-NEXT:    [[VEC_PHI2:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[PRED_UDIV_CONTINUE5]] ]
+; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; UNROLL-NO-VF-NEXT:    [[VEC_IV:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[VEC_IV3:%.*]] = add i32 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = icmp ule i32 [[VEC_IV]], [[TRIP_COUNT_MINUS_1]]
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = icmp ule i32 [[VEC_IV3]], [[TRIP_COUNT_MINUS_1]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; UNROLL-NO-VF:       pred.udiv.if:
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = udiv i32 219220132, [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; UNROLL-NO-VF:       pred.udiv.continue:
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP4]], [[PRED_UDIV_IF]] ]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP3]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5]]
+; UNROLL-NO-VF:       pred.udiv.if4:
+; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i32 [[OFFSET_IDX]], -1
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = udiv i32 219220132, [[INDUCTION1]]
+; UNROLL-NO-VF-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
+; UNROLL-NO-VF:       pred.udiv.continue5:
+; UNROLL-NO-VF-NEXT:    [[TMP7]] = phi i32 [ poison, [[PRED_UDIV_CONTINUE]] ], [ [[TMP6]], [[PRED_UDIV_IF4]] ]
+; UNROLL-NO-VF-NEXT:    [[TMP8]] = add i32 [[VEC_PHI]], [[VECTOR_RECUR]]
+; UNROLL-NO-VF-NEXT:    [[TMP9]] = add i32 [[VEC_PHI2]], [[TMP5]]
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = select i1 [[TMP2]], i32 [[TMP8]], i32 [[VEC_PHI]]
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP3]], i32 [[TMP9]], i32 [[VEC_PHI2]]
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF51:![0-9]+]], !llvm.loop [[LOOP52:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP11]], [[TMP10]]
+; UNROLL-NO-VF-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[BB2:%.*]]
+; UNROLL-NO-VF:       bb1:
+; UNROLL-NO-VF-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    ret i32 [[VAR]]
+; UNROLL-NO-VF:       bb2:
+; UNROLL-NO-VF-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; UNROLL-NO-VF-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NO-VF-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
+; UNROLL-NO-VF-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
+; UNROLL-NO-VF-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
+; UNROLL-NO-VF-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF53:![0-9]+]], !llvm.loop [[LOOP54:![0-9]+]]
+;
+; SINK-AFTER-LABEL: @sink_into_replication_region(
+; SINK-AFTER-NEXT:  bb:
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; SINK-AFTER-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 3
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; SINK-AFTER-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i32 0
+; SINK-AFTER-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; SINK-AFTER-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 -1, i32 -2, i32 -3>
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE8:%.*]] ]
+; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UDIV_CONTINUE8]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[PRED_UDIV_CONTINUE8]] ]
+; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[PRED_UDIV_CONTINUE8]] ]
+; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i32 0
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT1]], <4 x i32> poison, <4 x i32> zeroinitializer
+; SINK-AFTER-NEXT:    [[VEC_IV:%.*]] = add <4 x i32> [[BROADCAST_SPLAT2]], <i32 0, i32 1, i32 2, i32 3>
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp ule <4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT]]
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
+; SINK-AFTER-NEXT:    br i1 [[TMP3]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; SINK-AFTER:       pred.udiv.if:
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], 0
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = udiv i32 219220132, [[TMP4]]
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32> poison, i32 [[TMP5]], i32 0
+; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; SINK-AFTER:       pred.udiv.continue:
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP6]], [[PRED_UDIV_IF]] ]
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
+; SINK-AFTER-NEXT:    br i1 [[TMP8]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]
+; SINK-AFTER:       pred.udiv.if3:
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = add i32 [[OFFSET_IDX]], -1
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = udiv i32 219220132, [[TMP9]]
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> [[TMP7]], i32 [[TMP10]], i32 1
+; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE4]]
+; SINK-AFTER:       pred.udiv.continue4:
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = phi <4 x i32> [ [[TMP7]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP11]], [[PRED_UDIV_IF3]] ]
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]
+; SINK-AFTER:       pred.udiv.if5:
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = add i32 [[OFFSET_IDX]], -2
+; SINK-AFTER-NEXT:    [[TMP15:%.*]] = udiv i32 219220132, [[TMP14]]
+; SINK-AFTER-NEXT:    [[TMP16:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP15]], i32 2
+; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE6]]
+; SINK-AFTER:       pred.udiv.continue6:
+; SINK-AFTER-NEXT:    [[TMP17:%.*]] = phi <4 x i32> [ [[TMP12]], [[PRED_UDIV_CONTINUE4]] ], [ [[TMP16]], [[PRED_UDIV_IF5]] ]
+; SINK-AFTER-NEXT:    [[TMP18:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
+; SINK-AFTER-NEXT:    br i1 [[TMP18]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8]]
+; SINK-AFTER:       pred.udiv.if7:
+; SINK-AFTER-NEXT:    [[TMP19:%.*]] = add i32 [[OFFSET_IDX]], -3
+; SINK-AFTER-NEXT:    [[TMP20:%.*]] = udiv i32 219220132, [[TMP19]]
+; SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP20]], i32 3
+; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE8]]
+; SINK-AFTER:       pred.udiv.continue8:
+; SINK-AFTER-NEXT:    [[TMP22]] = phi <4 x i32> [ [[TMP17]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP21]], [[PRED_UDIV_IF7]] ]
+; SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP24]] = add <4 x i32> [[VEC_PHI]], [[TMP23]]
+; SINK-AFTER-NEXT:    [[TMP25:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP24]], <4 x i32> [[VEC_PHI]]
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 -4, i32 -4, i32 -4, i32 -4>
+; SINK-AFTER-NEXT:    [[TMP26:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52:![0-9]+]], !llvm.loop [[LOOP53:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[TMP27:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP25]])
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
+; SINK-AFTER-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
+; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    br label [[BB2:%.*]]
+; SINK-AFTER:       bb1:
+; SINK-AFTER-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    ret i32 [[VAR]]
+; SINK-AFTER:       bb2:
+; SINK-AFTER-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; SINK-AFTER-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; SINK-AFTER-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
+; SINK-AFTER-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
+; SINK-AFTER-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
+; SINK-AFTER-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
+; SINK-AFTER-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF54:![0-9]+]], !llvm.loop [[LOOP55:![0-9]+]]
+;
+; NO-SINK-AFTER-LABEL: @sink_into_replication_region(
+; NO-SINK-AFTER-NEXT:  bb:
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; NO-SINK-AFTER-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; NO-SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 3
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; NO-SINK-AFTER-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i32 0
+; NO-SINK-AFTER-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; NO-SINK-AFTER-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 -1, i32 -2, i32 -3>
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE8:%.*]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UDIV_CONTINUE8]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[PRED_UDIV_CONTINUE8]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[PRED_UDIV_CONTINUE8]] ]
+; NO-SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i32 0
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT1]], <4 x i32> poison, <4 x i32> zeroinitializer
+; NO-SINK-AFTER-NEXT:    [[VEC_IV:%.*]] = add <4 x i32> [[BROADCAST_SPLAT2]], <i32 0, i32 1, i32 2, i32 3>
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp ule <4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT]]
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP3]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; NO-SINK-AFTER:       pred.udiv.if:
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = udiv i32 219220132, [[TMP4]]
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32> poison, i32 [[TMP5]], i32 0
+; NO-SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; NO-SINK-AFTER:       pred.udiv.continue:
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP6]], [[PRED_UDIV_IF]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP8]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]
+; NO-SINK-AFTER:       pred.udiv.if3:
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = add i32 [[OFFSET_IDX]], -1
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = udiv i32 219220132, [[TMP9]]
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> [[TMP7]], i32 [[TMP10]], i32 1
+; NO-SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE4]]
+; NO-SINK-AFTER:       pred.udiv.continue4:
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = phi <4 x i32> [ [[TMP7]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP11]], [[PRED_UDIV_IF3]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]
+; NO-SINK-AFTER:       pred.udiv.if5:
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = add i32 [[OFFSET_IDX]], -2
+; NO-SINK-AFTER-NEXT:    [[TMP15:%.*]] = udiv i32 219220132, [[TMP14]]
+; NO-SINK-AFTER-NEXT:    [[TMP16:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP15]], i32 2
+; NO-SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE6]]
+; NO-SINK-AFTER:       pred.udiv.continue6:
+; NO-SINK-AFTER-NEXT:    [[TMP17:%.*]] = phi <4 x i32> [ [[TMP12]], [[PRED_UDIV_CONTINUE4]] ], [ [[TMP16]], [[PRED_UDIV_IF5]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP18:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP18]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8]]
+; NO-SINK-AFTER:       pred.udiv.if7:
+; NO-SINK-AFTER-NEXT:    [[TMP19:%.*]] = add i32 [[OFFSET_IDX]], -3
+; NO-SINK-AFTER-NEXT:    [[TMP20:%.*]] = udiv i32 219220132, [[TMP19]]
+; NO-SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP20]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE8]]
+; NO-SINK-AFTER:       pred.udiv.continue8:
+; NO-SINK-AFTER-NEXT:    [[TMP22]] = phi <4 x i32> [ [[TMP17]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP21]], [[PRED_UDIV_IF7]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP24]] = add <4 x i32> [[VEC_PHI]], [[TMP23]]
+; NO-SINK-AFTER-NEXT:    [[TMP25:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP24]], <4 x i32> [[VEC_PHI]]
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 -4, i32 -4, i32 -4, i32 -4>
+; NO-SINK-AFTER-NEXT:    [[TMP26:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52:![0-9]+]], !llvm.loop [[LOOP53:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[TMP27:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP25]])
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[BB2:%.*]]
+; NO-SINK-AFTER:       bb1:
+; NO-SINK-AFTER-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    ret i32 [[VAR]]
+; NO-SINK-AFTER:       bb2:
+; NO-SINK-AFTER-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; NO-SINK-AFTER-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; NO-SINK-AFTER-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
+; NO-SINK-AFTER-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
+; NO-SINK-AFTER-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
+; NO-SINK-AFTER-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
+; NO-SINK-AFTER-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF54:![0-9]+]], !llvm.loop [[LOOP55:![0-9]+]]
 ;
 bb:
   br label %bb2
 
   bb1:                                              ; preds = %bb2
-  %tmp = phi i32 [ %tmp6, %bb2 ]
-  ret i32 %tmp
+  %var = phi i32 [ %var6, %bb2 ]
+  ret i32 %var
 
   bb2:                                              ; preds = %bb2, %bb
-  %tmp3 = phi i32 [ %tmp8, %bb2 ], [ %y, %bb ]
-  %tmp4 = phi i32 [ %tmp7, %bb2 ], [ 0, %bb ]
-  %tmp5 = phi i32 [ %tmp6, %bb2 ], [ 0, %bb ]
-  %tmp6 = add i32 %tmp5, %tmp4
-  %tmp7 = udiv i32 219220132, %tmp3
-  %tmp8 = add nsw i32 %tmp3, -1
-  %tmp9 = icmp slt i32 %tmp3, 2
-  br i1 %tmp9, label %bb1, label %bb2, !prof !2
+  %var3 = phi i32 [ %var8, %bb2 ], [ %y, %bb ]
+  %var4 = phi i32 [ %var7, %bb2 ], [ 0, %bb ]
+  %var5 = phi i32 [ %var6, %bb2 ], [ 0, %bb ]
+  %var6 = add i32 %var5, %var4
+  %var7 = udiv i32 219220132, %var3
+  %var8 = add nsw i32 %var3, -1
+  %var9 = icmp slt i32 %var3, 2
+  br i1 %var9, label %bb1, label %bb2, !prof !2
 }
 
 define i32 @sink_into_replication_region_multiple(i32 *%x, i32 %y) {
+;
 ; CHECK-LABEL: @sink_into_replication_region_multiple(
 ; CHECK-NEXT:  bb:
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
@@ -780,10 +5826,10 @@ define i32 @sink_into_replication_region_multiple(i32 *%x, i32 %y) {
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE16:%.*]] ]
-; CHECK-NEXT:    [[VEC_IND2:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[PRED_STORE_CONTINUE16]] ]
-; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[PRED_STORE_CONTINUE16]] ]
-; CHECK-NEXT:    [[VEC_PHI4:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[PRED_STORE_CONTINUE16]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE15:%.*]] ]
+; CHECK-NEXT:    [[VEC_IND2:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[PRED_STORE_CONTINUE15]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], -1
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[OFFSET_IDX]], -2
@@ -798,31 +5844,31 @@ define i32 @sink_into_replication_region_multiple(i32 *%x, i32 %y) {
 ; CHECK:       pred.udiv.continue:
 ; CHECK-NEXT:    [[TMP9:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_UDIV_IF]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP5]], i32 1
-; CHECK-NEXT:    br i1 [[TMP10]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]
+; CHECK-NEXT:    br i1 [[TMP10]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
 ; CHECK:       pred.udiv.if4:
 ; CHECK-NEXT:    [[TMP11:%.*]] = udiv i32 219220132, [[TMP2]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP9]], i32 [[TMP11]], i32 1
-; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE6]]
+; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
 ; CHECK:       pred.udiv.continue5:
-; CHECK-NEXT:    [[TMP13:%.*]] = phi <4 x i32> [ [[TMP9]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP12]], [[PRED_UDIV_IF5]] ]
+; CHECK-NEXT:    [[TMP13:%.*]] = phi <4 x i32> [ [[TMP9]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP12]], [[PRED_UDIV_IF4]] ]
 ; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <4 x i1> [[TMP5]], i32 2
-; CHECK-NEXT:    br i1 [[TMP14]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8:%.*]]
+; CHECK-NEXT:    br i1 [[TMP14]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
 ; CHECK:       pred.udiv.if6:
 ; CHECK-NEXT:    [[TMP15:%.*]] = udiv i32 219220132, [[TMP3]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP15]], i32 2
-; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE8]]
+; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
 ; CHECK:       pred.udiv.continue7:
-; CHECK-NEXT:    [[TMP17:%.*]] = phi <4 x i32> [ [[TMP13]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP16]], [[PRED_UDIV_IF7]] ]
+; CHECK-NEXT:    [[TMP17:%.*]] = phi <4 x i32> [ [[TMP13]], [[PRED_UDIV_CONTINUE5]] ], [ [[TMP16]], [[PRED_UDIV_IF6]] ]
 ; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <4 x i1> [[TMP5]], i32 3
-; CHECK-NEXT:    br i1 [[TMP18]], label [[PRED_UDIV_IF9:%.*]], label [[PRED_UDIV_CONTINUE10:%.*]]
+; CHECK-NEXT:    br i1 [[TMP18]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
 ; CHECK:       pred.udiv.if8:
 ; CHECK-NEXT:    [[TMP19:%.*]] = udiv i32 219220132, [[TMP4]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP19]], i32 3
-; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE10]]
+; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
 ; CHECK:       pred.udiv.continue9:
-; CHECK-NEXT:    [[TMP21]] = phi <4 x i32> [ [[TMP17]], [[PRED_UDIV_CONTINUE8]] ], [ [[TMP20]], [[PRED_UDIV_IF9]] ]
+; CHECK-NEXT:    [[TMP21]] = phi <4 x i32> [ [[TMP17]], [[PRED_UDIV_CONTINUE7]] ], [ [[TMP20]], [[PRED_UDIV_IF8]] ]
 ; CHECK-NEXT:    [[TMP22:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP21]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NEXT:    [[TMP23]] = add <4 x i32> [[VEC_PHI4]], [[TMP22]]
+; CHECK-NEXT:    [[TMP23]] = add <4 x i32> [[VEC_PHI]], [[TMP22]]
 ; CHECK-NEXT:    [[TMP24:%.*]] = extractelement <4 x i1> [[TMP5]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP24]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
 ; CHECK:       pred.store.if:
@@ -832,86 +5878,1102 @@ define i32 @sink_into_replication_region_multiple(i32 *%x, i32 %y) {
 ; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE]]
 ; CHECK:       pred.store.continue:
 ; CHECK-NEXT:    [[TMP27:%.*]] = extractelement <4 x i1> [[TMP5]], i32 1
-; CHECK-NEXT:    br i1 [[TMP27]], label [[PRED_STORE_IF11:%.*]], label [[PRED_STORE_CONTINUE12:%.*]]
+; CHECK-NEXT:    br i1 [[TMP27]], label [[PRED_STORE_IF10:%.*]], label [[PRED_STORE_CONTINUE11:%.*]]
 ; CHECK:       pred.store.if10:
 ; CHECK-NEXT:    [[TMP28:%.*]] = or i32 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP29]]
 ; CHECK-NEXT:    store i32 [[TMP2]], i32* [[TMP30]], align 4
-; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE12]]
+; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE11]]
 ; CHECK:       pred.store.continue11:
 ; CHECK-NEXT:    [[TMP31:%.*]] = extractelement <4 x i1> [[TMP5]], i32 2
-; CHECK-NEXT:    br i1 [[TMP31]], label [[PRED_STORE_IF13:%.*]], label [[PRED_STORE_CONTINUE14:%.*]]
+; CHECK-NEXT:    br i1 [[TMP31]], label [[PRED_STORE_IF12:%.*]], label [[PRED_STORE_CONTINUE13:%.*]]
 ; CHECK:       pred.store.if12:
 ; CHECK-NEXT:    [[TMP32:%.*]] = or i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP33:%.*]] = sext i32 [[TMP32]] to i64
 ; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP33]]
 ; CHECK-NEXT:    store i32 [[TMP3]], i32* [[TMP34]], align 4
-; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE14]]
+; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE13]]
 ; CHECK:       pred.store.continue13:
 ; CHECK-NEXT:    [[TMP35:%.*]] = extractelement <4 x i1> [[TMP5]], i32 3
-; CHECK-NEXT:    br i1 [[TMP35]], label [[PRED_STORE_IF15:%.*]], label [[PRED_STORE_CONTINUE16]]
+; CHECK-NEXT:    br i1 [[TMP35]], label [[PRED_STORE_IF14:%.*]], label [[PRED_STORE_CONTINUE15]]
 ; CHECK:       pred.store.if14:
 ; CHECK-NEXT:    [[TMP36:%.*]] = or i32 [[INDEX]], 3
 ; CHECK-NEXT:    [[TMP37:%.*]] = sext i32 [[TMP36]] to i64
 ; CHECK-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP37]]
 ; CHECK-NEXT:    store i32 [[TMP4]], i32* [[TMP38]], align 4
-; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE16]]
+; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE15]]
 ; CHECK:       pred.store.continue15:
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT3]] = add <4 x i32> [[VEC_IND2]], <i32 4, i32 4, i32 4, i32 4>
 ; CHECK-NEXT:    [[TMP39:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP39]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof !52, [[LOOP49:!llvm.loop !.*]]
+; CHECK-NEXT:    br i1 [[TMP39]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52]], !llvm.loop [[LOOP56:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP40:%.*]] = select <4 x i1> [[TMP5]], <4 x i32> [[TMP23]], <4 x i32> [[VEC_PHI4]]
+; CHECK-NEXT:    [[TMP40:%.*]] = select <4 x i1> [[TMP5]], <4 x i32> [[TMP23]], <4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP41:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP40]])
 ; CHECK-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[BB2:%.*]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[TMP:%.*]] = phi i32 [ undef, [[BB2]] ], [ [[TMP41]], [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    ret i32 [[TMP]]
+; CHECK-NEXT:    [[VAR:%.*]] = phi i32 [ undef, [[BB2]] ], [ [[TMP41]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[VAR]]
 ; CHECK:       bb2:
-; CHECK-NEXT:    br i1 undef, label [[BB1]], label [[BB2]], !prof !54, [[LOOP50:!llvm.loop !.*]]
+; CHECK-NEXT:    br i1 undef, label [[BB1]], label [[BB2]], !prof [[PROF54]], !llvm.loop [[LOOP57:![0-9]+]]
+;
+; UNROLL-LABEL: @sink_into_replication_region_multiple(
+; UNROLL-NEXT:  bb:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; UNROLL-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; UNROLL-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 7
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i32 [[N_RND_UP]], -8
+; UNROLL-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = add i32 [[TMP1]], -1
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE35:%.*]] ]
+; UNROLL-NEXT:    [[VEC_IND3:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP42:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP45:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP46:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; UNROLL-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], -1
+; UNROLL-NEXT:    [[TMP3:%.*]] = add i32 [[OFFSET_IDX]], -2
+; UNROLL-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], -3
+; UNROLL-NEXT:    [[TMP5:%.*]] = add i32 [[OFFSET_IDX]], -4
+; UNROLL-NEXT:    [[TMP6:%.*]] = add i32 [[OFFSET_IDX]], -5
+; UNROLL-NEXT:    [[TMP7:%.*]] = add i32 [[OFFSET_IDX]], -6
+; UNROLL-NEXT:    [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], -7
+; UNROLL-NEXT:    [[STEP_ADD4:%.*]] = add <4 x i32> [[VEC_IND3]], <i32 4, i32 4, i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP9:%.*]] = icmp ule <4 x i32> [[VEC_IND3]], [[BROADCAST_SPLAT]]
+; UNROLL-NEXT:    [[TMP10:%.*]] = icmp ule <4 x i32> [[STEP_ADD4]], [[BROADCAST_SPLAT]]
+; UNROLL-NEXT:    [[TMP11:%.*]] = extractelement <4 x i1> [[TMP9]], i32 0
+; UNROLL-NEXT:    br i1 [[TMP11]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; UNROLL:       pred.udiv.if:
+; UNROLL-NEXT:    [[TMP12:%.*]] = udiv i32 219220132, [[OFFSET_IDX]]
+; UNROLL-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> poison, i32 [[TMP12]], i32 0
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; UNROLL:       pred.udiv.continue:
+; UNROLL-NEXT:    [[TMP14:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP13]], [[PRED_UDIV_IF]] ]
+; UNROLL-NEXT:    [[TMP15:%.*]] = extractelement <4 x i1> [[TMP9]], i32 1
+; UNROLL-NEXT:    br i1 [[TMP15]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; UNROLL:       pred.udiv.if8:
+; UNROLL-NEXT:    [[TMP16:%.*]] = udiv i32 219220132, [[TMP2]]
+; UNROLL-NEXT:    [[TMP17:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP16]], i32 1
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; UNROLL:       pred.udiv.continue9:
+; UNROLL-NEXT:    [[TMP18:%.*]] = phi <4 x i32> [ [[TMP14]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP17]], [[PRED_UDIV_IF8]] ]
+; UNROLL-NEXT:    [[TMP19:%.*]] = extractelement <4 x i1> [[TMP9]], i32 2
+; UNROLL-NEXT:    br i1 [[TMP19]], label [[PRED_UDIV_IF10:%.*]], label [[PRED_UDIV_CONTINUE11:%.*]]
+; UNROLL:       pred.udiv.if10:
+; UNROLL-NEXT:    [[TMP20:%.*]] = udiv i32 219220132, [[TMP3]]
+; UNROLL-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP20]], i32 2
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE11]]
+; UNROLL:       pred.udiv.continue11:
+; UNROLL-NEXT:    [[TMP22:%.*]] = phi <4 x i32> [ [[TMP18]], [[PRED_UDIV_CONTINUE9]] ], [ [[TMP21]], [[PRED_UDIV_IF10]] ]
+; UNROLL-NEXT:    [[TMP23:%.*]] = extractelement <4 x i1> [[TMP9]], i32 3
+; UNROLL-NEXT:    br i1 [[TMP23]], label [[PRED_UDIV_IF12:%.*]], label [[PRED_UDIV_CONTINUE13:%.*]]
+; UNROLL:       pred.udiv.if12:
+; UNROLL-NEXT:    [[TMP24:%.*]] = udiv i32 219220132, [[TMP4]]
+; UNROLL-NEXT:    [[TMP25:%.*]] = insertelement <4 x i32> [[TMP22]], i32 [[TMP24]], i32 3
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE13]]
+; UNROLL:       pred.udiv.continue13:
+; UNROLL-NEXT:    [[TMP26:%.*]] = phi <4 x i32> [ [[TMP22]], [[PRED_UDIV_CONTINUE11]] ], [ [[TMP25]], [[PRED_UDIV_IF12]] ]
+; UNROLL-NEXT:    [[TMP27:%.*]] = extractelement <4 x i1> [[TMP10]], i32 0
+; UNROLL-NEXT:    br i1 [[TMP27]], label [[PRED_UDIV_IF14:%.*]], label [[PRED_UDIV_CONTINUE15:%.*]]
+; UNROLL:       pred.udiv.if14:
+; UNROLL-NEXT:    [[TMP28:%.*]] = udiv i32 219220132, [[TMP5]]
+; UNROLL-NEXT:    [[TMP29:%.*]] = insertelement <4 x i32> poison, i32 [[TMP28]], i32 0
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE15]]
+; UNROLL:       pred.udiv.continue15:
+; UNROLL-NEXT:    [[TMP30:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE13]] ], [ [[TMP29]], [[PRED_UDIV_IF14]] ]
+; UNROLL-NEXT:    [[TMP31:%.*]] = extractelement <4 x i1> [[TMP10]], i32 1
+; UNROLL-NEXT:    br i1 [[TMP31]], label [[PRED_UDIV_IF16:%.*]], label [[PRED_UDIV_CONTINUE17:%.*]]
+; UNROLL:       pred.udiv.if16:
+; UNROLL-NEXT:    [[TMP32:%.*]] = udiv i32 219220132, [[TMP6]]
+; UNROLL-NEXT:    [[TMP33:%.*]] = insertelement <4 x i32> [[TMP30]], i32 [[TMP32]], i32 1
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE17]]
+; UNROLL:       pred.udiv.continue17:
+; UNROLL-NEXT:    [[TMP34:%.*]] = phi <4 x i32> [ [[TMP30]], [[PRED_UDIV_CONTINUE15]] ], [ [[TMP33]], [[PRED_UDIV_IF16]] ]
+; UNROLL-NEXT:    [[TMP35:%.*]] = extractelement <4 x i1> [[TMP10]], i32 2
+; UNROLL-NEXT:    br i1 [[TMP35]], label [[PRED_UDIV_IF18:%.*]], label [[PRED_UDIV_CONTINUE19:%.*]]
+; UNROLL:       pred.udiv.if18:
+; UNROLL-NEXT:    [[TMP36:%.*]] = udiv i32 219220132, [[TMP7]]
+; UNROLL-NEXT:    [[TMP37:%.*]] = insertelement <4 x i32> [[TMP34]], i32 [[TMP36]], i32 2
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE19]]
+; UNROLL:       pred.udiv.continue19:
+; UNROLL-NEXT:    [[TMP38:%.*]] = phi <4 x i32> [ [[TMP34]], [[PRED_UDIV_CONTINUE17]] ], [ [[TMP37]], [[PRED_UDIV_IF18]] ]
+; UNROLL-NEXT:    [[TMP39:%.*]] = extractelement <4 x i1> [[TMP10]], i32 3
+; UNROLL-NEXT:    br i1 [[TMP39]], label [[PRED_UDIV_IF20:%.*]], label [[PRED_UDIV_CONTINUE21:%.*]]
+; UNROLL:       pred.udiv.if20:
+; UNROLL-NEXT:    [[TMP40:%.*]] = udiv i32 219220132, [[TMP8]]
+; UNROLL-NEXT:    [[TMP41:%.*]] = insertelement <4 x i32> [[TMP38]], i32 [[TMP40]], i32 3
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE21]]
+; UNROLL:       pred.udiv.continue21:
+; UNROLL-NEXT:    [[TMP42]] = phi <4 x i32> [ [[TMP38]], [[PRED_UDIV_CONTINUE19]] ], [ [[TMP41]], [[PRED_UDIV_IF20]] ]
+; UNROLL-NEXT:    [[TMP43:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP26]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[TMP26]], <4 x i32> [[TMP42]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NEXT:    [[TMP45]] = add <4 x i32> [[VEC_PHI]], [[TMP43]]
+; UNROLL-NEXT:    [[TMP46]] = add <4 x i32> [[VEC_PHI7]], [[TMP44]]
+; UNROLL-NEXT:    [[TMP47:%.*]] = extractelement <4 x i1> [[TMP9]], i32 0
+; UNROLL-NEXT:    br i1 [[TMP47]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; UNROLL:       pred.store.if:
+; UNROLL-NEXT:    [[TMP48:%.*]] = sext i32 [[INDEX]] to i64
+; UNROLL-NEXT:    [[TMP49:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i64 [[TMP48]]
+; UNROLL-NEXT:    store i32 [[OFFSET_IDX]], i32* [[TMP49]], align 4
+; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; UNROLL:       pred.store.continue:
+; UNROLL-NEXT:    [[TMP50:%.*]] = extractelement <4 x i1> [[TMP9]], i32 1
+; UNROLL-NEXT:    br i1 [[TMP50]], label [[PRED_STORE_IF22:%.*]], label [[PRED_STORE_CONTINUE23:%.*]]
+; UNROLL:       pred.store.if22:
+; UNROLL-NEXT:    [[TMP51:%.*]] = or i32 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP52:%.*]] = sext i32 [[TMP51]] to i64
+; UNROLL-NEXT:    [[TMP53:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP52]]
+; UNROLL-NEXT:    store i32 [[TMP2]], i32* [[TMP53]], align 4
+; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE23]]
+; UNROLL:       pred.store.continue23:
+; UNROLL-NEXT:    [[TMP54:%.*]] = extractelement <4 x i1> [[TMP9]], i32 2
+; UNROLL-NEXT:    br i1 [[TMP54]], label [[PRED_STORE_IF24:%.*]], label [[PRED_STORE_CONTINUE25:%.*]]
+; UNROLL:       pred.store.if24:
+; UNROLL-NEXT:    [[TMP55:%.*]] = or i32 [[INDEX]], 2
+; UNROLL-NEXT:    [[TMP56:%.*]] = sext i32 [[TMP55]] to i64
+; UNROLL-NEXT:    [[TMP57:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP56]]
+; UNROLL-NEXT:    store i32 [[TMP3]], i32* [[TMP57]], align 4
+; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE25]]
+; UNROLL:       pred.store.continue25:
+; UNROLL-NEXT:    [[TMP58:%.*]] = extractelement <4 x i1> [[TMP9]], i32 3
+; UNROLL-NEXT:    br i1 [[TMP58]], label [[PRED_STORE_IF26:%.*]], label [[PRED_STORE_CONTINUE27:%.*]]
+; UNROLL:       pred.store.if26:
+; UNROLL-NEXT:    [[TMP59:%.*]] = or i32 [[INDEX]], 3
+; UNROLL-NEXT:    [[TMP60:%.*]] = sext i32 [[TMP59]] to i64
+; UNROLL-NEXT:    [[TMP61:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP60]]
+; UNROLL-NEXT:    store i32 [[TMP4]], i32* [[TMP61]], align 4
+; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE27]]
+; UNROLL:       pred.store.continue27:
+; UNROLL-NEXT:    [[TMP62:%.*]] = extractelement <4 x i1> [[TMP10]], i32 0
+; UNROLL-NEXT:    br i1 [[TMP62]], label [[PRED_STORE_IF28:%.*]], label [[PRED_STORE_CONTINUE29:%.*]]
+; UNROLL:       pred.store.if28:
+; UNROLL-NEXT:    [[TMP63:%.*]] = or i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP64:%.*]] = sext i32 [[TMP63]] to i64
+; UNROLL-NEXT:    [[TMP65:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP64]]
+; UNROLL-NEXT:    store i32 [[TMP5]], i32* [[TMP65]], align 4
+; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE29]]
+; UNROLL:       pred.store.continue29:
+; UNROLL-NEXT:    [[TMP66:%.*]] = extractelement <4 x i1> [[TMP10]], i32 1
+; UNROLL-NEXT:    br i1 [[TMP66]], label [[PRED_STORE_IF30:%.*]], label [[PRED_STORE_CONTINUE31:%.*]]
+; UNROLL:       pred.store.if30:
+; UNROLL-NEXT:    [[TMP67:%.*]] = or i32 [[INDEX]], 5
+; UNROLL-NEXT:    [[TMP68:%.*]] = sext i32 [[TMP67]] to i64
+; UNROLL-NEXT:    [[TMP69:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP68]]
+; UNROLL-NEXT:    store i32 [[TMP6]], i32* [[TMP69]], align 4
+; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE31]]
+; UNROLL:       pred.store.continue31:
+; UNROLL-NEXT:    [[TMP70:%.*]] = extractelement <4 x i1> [[TMP10]], i32 2
+; UNROLL-NEXT:    br i1 [[TMP70]], label [[PRED_STORE_IF32:%.*]], label [[PRED_STORE_CONTINUE33:%.*]]
+; UNROLL:       pred.store.if32:
+; UNROLL-NEXT:    [[TMP71:%.*]] = or i32 [[INDEX]], 6
+; UNROLL-NEXT:    [[TMP72:%.*]] = sext i32 [[TMP71]] to i64
+; UNROLL-NEXT:    [[TMP73:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP72]]
+; UNROLL-NEXT:    store i32 [[TMP7]], i32* [[TMP73]], align 4
+; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE33]]
+; UNROLL:       pred.store.continue33:
+; UNROLL-NEXT:    [[TMP74:%.*]] = extractelement <4 x i1> [[TMP10]], i32 3
+; UNROLL-NEXT:    br i1 [[TMP74]], label [[PRED_STORE_IF34:%.*]], label [[PRED_STORE_CONTINUE35]]
+; UNROLL:       pred.store.if34:
+; UNROLL-NEXT:    [[TMP75:%.*]] = or i32 [[INDEX]], 7
+; UNROLL-NEXT:    [[TMP76:%.*]] = sext i32 [[TMP75]] to i64
+; UNROLL-NEXT:    [[TMP77:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP76]]
+; UNROLL-NEXT:    store i32 [[TMP8]], i32* [[TMP77]], align 4
+; UNROLL-NEXT:    br label [[PRED_STORE_CONTINUE35]]
+; UNROLL:       pred.store.continue35:
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 8
+; UNROLL-NEXT:    [[VEC_IND_NEXT6]] = add <4 x i32> [[VEC_IND3]], <i32 8, i32 8, i32 8, i32 8>
+; UNROLL-NEXT:    [[TMP78:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP78]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52]], !llvm.loop [[LOOP56:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[TMP79:%.*]] = select <4 x i1> [[TMP10]], <4 x i32> [[TMP46]], <4 x i32> [[VEC_PHI7]]
+; UNROLL-NEXT:    [[TMP80:%.*]] = select <4 x i1> [[TMP9]], <4 x i32> [[TMP45]], <4 x i32> [[VEC_PHI]]
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP79]], [[TMP80]]
+; UNROLL-NEXT:    [[TMP81:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; UNROLL-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[BB2:%.*]]
+; UNROLL:       bb1:
+; UNROLL-NEXT:    [[VAR:%.*]] = phi i32 [ undef, [[BB2]] ], [ [[TMP81]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i32 [[VAR]]
+; UNROLL:       bb2:
+; UNROLL-NEXT:    br i1 undef, label [[BB1]], label [[BB2]], !prof [[PROF54]], !llvm.loop [[LOOP57:![0-9]+]]
+;
+; UNROLL-NO-IC-LABEL: @sink_into_replication_region_multiple(
+; UNROLL-NO-IC-NEXT:  bb:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; UNROLL-NO-IC-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 7
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 -1, i32 -2, i32 -3>
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE35:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND3:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP43:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP46:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[PRED_STORE_CONTINUE35]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 -4, i32 -4, i32 -4, i32 -4>
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i32 [[OFFSET_IDX]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], -2
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i32 [[OFFSET_IDX]], -3
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i32 [[OFFSET_IDX]], -4
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i32 [[OFFSET_IDX]], -5
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], -6
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add i32 [[OFFSET_IDX]], -7
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD4:%.*]] = add <4 x i32> [[VEC_IND3]], <i32 4, i32 4, i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = icmp ule <4 x i32> [[VEC_IND3]], [[BROADCAST_SPLAT]]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = icmp ule <4 x i32> [[STEP_ADD4]], [[BROADCAST_SPLAT]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = extractelement <4 x i1> [[TMP10]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP12]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if:
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = udiv i32 219220132, [[TMP2]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> poison, i32 [[TMP13]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; UNROLL-NO-IC:       pred.udiv.continue:
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP14]], [[PRED_UDIV_IF]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = extractelement <4 x i1> [[TMP10]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP16]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if8:
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = udiv i32 219220132, [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = insertelement <4 x i32> [[TMP15]], i32 [[TMP17]], i32 1
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; UNROLL-NO-IC:       pred.udiv.continue9:
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = phi <4 x i32> [ [[TMP15]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP18]], [[PRED_UDIV_IF8]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = extractelement <4 x i1> [[TMP10]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP20]], label [[PRED_UDIV_IF10:%.*]], label [[PRED_UDIV_CONTINUE11:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if10:
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = udiv i32 219220132, [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP21]], i32 2
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE11]]
+; UNROLL-NO-IC:       pred.udiv.continue11:
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = phi <4 x i32> [ [[TMP19]], [[PRED_UDIV_CONTINUE9]] ], [ [[TMP22]], [[PRED_UDIV_IF10]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = extractelement <4 x i1> [[TMP10]], i32 3
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP24]], label [[PRED_UDIV_IF12:%.*]], label [[PRED_UDIV_CONTINUE13:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if12:
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = udiv i32 219220132, [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = insertelement <4 x i32> [[TMP23]], i32 [[TMP25]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE13]]
+; UNROLL-NO-IC:       pred.udiv.continue13:
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = phi <4 x i32> [ [[TMP23]], [[PRED_UDIV_CONTINUE11]] ], [ [[TMP26]], [[PRED_UDIV_IF12]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = extractelement <4 x i1> [[TMP11]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP28]], label [[PRED_UDIV_IF14:%.*]], label [[PRED_UDIV_CONTINUE15:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if14:
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = udiv i32 219220132, [[TMP6]]
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = insertelement <4 x i32> poison, i32 [[TMP29]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE15]]
+; UNROLL-NO-IC:       pred.udiv.continue15:
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE13]] ], [ [[TMP30]], [[PRED_UDIV_IF14]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = extractelement <4 x i1> [[TMP11]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP32]], label [[PRED_UDIV_IF16:%.*]], label [[PRED_UDIV_CONTINUE17:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if16:
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = udiv i32 219220132, [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = insertelement <4 x i32> [[TMP31]], i32 [[TMP33]], i32 1
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE17]]
+; UNROLL-NO-IC:       pred.udiv.continue17:
+; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = phi <4 x i32> [ [[TMP31]], [[PRED_UDIV_CONTINUE15]] ], [ [[TMP34]], [[PRED_UDIV_IF16]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = extractelement <4 x i1> [[TMP11]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP36]], label [[PRED_UDIV_IF18:%.*]], label [[PRED_UDIV_CONTINUE19:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if18:
+; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = udiv i32 219220132, [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = insertelement <4 x i32> [[TMP35]], i32 [[TMP37]], i32 2
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE19]]
+; UNROLL-NO-IC:       pred.udiv.continue19:
+; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = phi <4 x i32> [ [[TMP35]], [[PRED_UDIV_CONTINUE17]] ], [ [[TMP38]], [[PRED_UDIV_IF18]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = extractelement <4 x i1> [[TMP11]], i32 3
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP40]], label [[PRED_UDIV_IF20:%.*]], label [[PRED_UDIV_CONTINUE21:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if20:
+; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = udiv i32 219220132, [[TMP9]]
+; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = insertelement <4 x i32> [[TMP39]], i32 [[TMP41]], i32 3
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE21]]
+; UNROLL-NO-IC:       pred.udiv.continue21:
+; UNROLL-NO-IC-NEXT:    [[TMP43]] = phi <4 x i32> [ [[TMP39]], [[PRED_UDIV_CONTINUE19]] ], [ [[TMP42]], [[PRED_UDIV_IF20]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP27]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = shufflevector <4 x i32> [[TMP27]], <4 x i32> [[TMP43]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP46]] = add <4 x i32> [[VEC_PHI]], [[TMP44]]
+; UNROLL-NO-IC-NEXT:    [[TMP47]] = add <4 x i32> [[VEC_PHI7]], [[TMP45]]
+; UNROLL-NO-IC-NEXT:    [[TMP48:%.*]] = extractelement <4 x i1> [[TMP10]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP48]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; UNROLL-NO-IC:       pred.store.if:
+; UNROLL-NO-IC-NEXT:    [[TMP49:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP50:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i32 [[TMP49]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP2]], i32* [[TMP50]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; UNROLL-NO-IC:       pred.store.continue:
+; UNROLL-NO-IC-NEXT:    [[TMP51:%.*]] = extractelement <4 x i1> [[TMP10]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP51]], label [[PRED_STORE_IF22:%.*]], label [[PRED_STORE_CONTINUE23:%.*]]
+; UNROLL-NO-IC:       pred.store.if22:
+; UNROLL-NO-IC-NEXT:    [[TMP52:%.*]] = add i32 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP53:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP52]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP3]], i32* [[TMP53]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE23]]
+; UNROLL-NO-IC:       pred.store.continue23:
+; UNROLL-NO-IC-NEXT:    [[TMP54:%.*]] = extractelement <4 x i1> [[TMP10]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP54]], label [[PRED_STORE_IF24:%.*]], label [[PRED_STORE_CONTINUE25:%.*]]
+; UNROLL-NO-IC:       pred.store.if24:
+; UNROLL-NO-IC-NEXT:    [[TMP55:%.*]] = add i32 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP56:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP55]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP4]], i32* [[TMP56]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE25]]
+; UNROLL-NO-IC:       pred.store.continue25:
+; UNROLL-NO-IC-NEXT:    [[TMP57:%.*]] = extractelement <4 x i1> [[TMP10]], i32 3
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP57]], label [[PRED_STORE_IF26:%.*]], label [[PRED_STORE_CONTINUE27:%.*]]
+; UNROLL-NO-IC:       pred.store.if26:
+; UNROLL-NO-IC-NEXT:    [[TMP58:%.*]] = add i32 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP59:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP58]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP5]], i32* [[TMP59]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE27]]
+; UNROLL-NO-IC:       pred.store.continue27:
+; UNROLL-NO-IC-NEXT:    [[TMP60:%.*]] = extractelement <4 x i1> [[TMP11]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP60]], label [[PRED_STORE_IF28:%.*]], label [[PRED_STORE_CONTINUE29:%.*]]
+; UNROLL-NO-IC:       pred.store.if28:
+; UNROLL-NO-IC-NEXT:    [[TMP61:%.*]] = add i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP62:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP61]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP6]], i32* [[TMP62]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE29]]
+; UNROLL-NO-IC:       pred.store.continue29:
+; UNROLL-NO-IC-NEXT:    [[TMP63:%.*]] = extractelement <4 x i1> [[TMP11]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP63]], label [[PRED_STORE_IF30:%.*]], label [[PRED_STORE_CONTINUE31:%.*]]
+; UNROLL-NO-IC:       pred.store.if30:
+; UNROLL-NO-IC-NEXT:    [[TMP64:%.*]] = add i32 [[INDEX]], 5
+; UNROLL-NO-IC-NEXT:    [[TMP65:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP64]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP7]], i32* [[TMP65]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE31]]
+; UNROLL-NO-IC:       pred.store.continue31:
+; UNROLL-NO-IC-NEXT:    [[TMP66:%.*]] = extractelement <4 x i1> [[TMP11]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP66]], label [[PRED_STORE_IF32:%.*]], label [[PRED_STORE_CONTINUE33:%.*]]
+; UNROLL-NO-IC:       pred.store.if32:
+; UNROLL-NO-IC-NEXT:    [[TMP67:%.*]] = add i32 [[INDEX]], 6
+; UNROLL-NO-IC-NEXT:    [[TMP68:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP67]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP8]], i32* [[TMP68]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE33]]
+; UNROLL-NO-IC:       pred.store.continue33:
+; UNROLL-NO-IC-NEXT:    [[TMP69:%.*]] = extractelement <4 x i1> [[TMP11]], i32 3
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP69]], label [[PRED_STORE_IF34:%.*]], label [[PRED_STORE_CONTINUE35]]
+; UNROLL-NO-IC:       pred.store.if34:
+; UNROLL-NO-IC-NEXT:    [[TMP70:%.*]] = add i32 [[INDEX]], 7
+; UNROLL-NO-IC-NEXT:    [[TMP71:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP70]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP9]], i32* [[TMP71]], align 4
+; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE35]]
+; UNROLL-NO-IC:       pred.store.continue35:
+; UNROLL-NO-IC-NEXT:    [[TMP72:%.*]] = select <4 x i1> [[TMP10]], <4 x i32> [[TMP46]], <4 x i32> [[VEC_PHI]]
+; UNROLL-NO-IC-NEXT:    [[TMP73:%.*]] = select <4 x i1> [[TMP11]], <4 x i32> [[TMP47]], <4 x i32> [[VEC_PHI7]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], <i32 -4, i32 -4, i32 -4, i32 -4>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT6]] = add <4 x i32> [[STEP_ADD4]], <i32 4, i32 4, i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP74:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP74]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52]], !llvm.loop [[LOOP56:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP73]], [[TMP72]]
+; UNROLL-NO-IC-NEXT:    [[TMP75:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP43]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP43]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[BB]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP75]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[BB2:%.*]]
+; UNROLL-NO-IC:       bb1:
+; UNROLL-NO-IC-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP75]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[VAR]]
+; UNROLL-NO-IC:       bb2:
+; UNROLL-NO-IC-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[G:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[IV]]
+; UNROLL-NO-IC-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
+; UNROLL-NO-IC-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
+; UNROLL-NO-IC-NEXT:    store i32 [[VAR3]], i32* [[G]], align 4
+; UNROLL-NO-IC-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
+; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add nsw i32 [[IV]], 1
+; UNROLL-NO-IC-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
+; UNROLL-NO-IC-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF54]], !llvm.loop [[LOOP57:![0-9]+]]
+;
+; UNROLL-NO-VF-LABEL: @sink_into_replication_region_multiple(
+; UNROLL-NO-VF-NEXT:  bb:
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; UNROLL-NO-VF-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 1
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE9:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[PRED_STORE_CONTINUE9]] ]
+; UNROLL-NO-VF-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[PRED_STORE_CONTINUE9]] ]
+; UNROLL-NO-VF-NEXT:    [[VEC_PHI5:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[PRED_STORE_CONTINUE9]] ]
+; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION2:%.*]] = add i32 [[OFFSET_IDX]], -1
+; UNROLL-NO-VF-NEXT:    [[INDUCTION3:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION4:%.*]] = add i32 [[INDEX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = icmp ule i32 [[INDUCTION3]], [[TRIP_COUNT_MINUS_1]]
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = icmp ule i32 [[INDUCTION4]], [[TRIP_COUNT_MINUS_1]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; UNROLL-NO-VF:       pred.udiv.if:
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = udiv i32 219220132, [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; UNROLL-NO-VF:       pred.udiv.continue:
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP4]], [[PRED_UDIV_IF]] ]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP3]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
+; UNROLL-NO-VF:       pred.udiv.if6:
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = udiv i32 219220132, [[INDUCTION2]]
+; UNROLL-NO-VF-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
+; UNROLL-NO-VF:       pred.udiv.continue7:
+; UNROLL-NO-VF-NEXT:    [[TMP7]] = phi i32 [ poison, [[PRED_UDIV_CONTINUE]] ], [ [[TMP6]], [[PRED_UDIV_IF6]] ]
+; UNROLL-NO-VF-NEXT:    [[TMP8]] = add i32 [[VEC_PHI]], [[VECTOR_RECUR]]
+; UNROLL-NO-VF-NEXT:    [[TMP9]] = add i32 [[VEC_PHI5]], [[TMP5]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; UNROLL-NO-VF:       pred.store.if:
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i32 [[INDUCTION3]]
+; UNROLL-NO-VF-NEXT:    store i32 [[INDUCTION]], i32* [[TMP10]], align 4
+; UNROLL-NO-VF-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; UNROLL-NO-VF:       pred.store.continue:
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP3]], label [[PRED_STORE_IF8:%.*]], label [[PRED_STORE_CONTINUE9]]
+; UNROLL-NO-VF:       pred.store.if8:
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[INDUCTION4]]
+; UNROLL-NO-VF-NEXT:    store i32 [[INDUCTION2]], i32* [[TMP11]], align 4
+; UNROLL-NO-VF-NEXT:    br label [[PRED_STORE_CONTINUE9]]
+; UNROLL-NO-VF:       pred.store.continue9:
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = select i1 [[TMP2]], i32 [[TMP8]], i32 [[VEC_PHI]]
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = select i1 [[TMP3]], i32 [[TMP9]], i32 [[VEC_PHI5]]
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF51]], !llvm.loop [[LOOP55:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP13]], [[TMP12]]
+; UNROLL-NO-VF-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[BB]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    br label [[BB2:%.*]]
+; UNROLL-NO-VF:       bb1:
+; UNROLL-NO-VF-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    ret i32 [[VAR]]
+; UNROLL-NO-VF:       bb2:
+; UNROLL-NO-VF-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-VF-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; UNROLL-NO-VF-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NO-VF-NEXT:    [[G:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[IV]]
+; UNROLL-NO-VF-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
+; UNROLL-NO-VF-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
+; UNROLL-NO-VF-NEXT:    store i32 [[VAR3]], i32* [[G]], align 4
+; UNROLL-NO-VF-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
+; UNROLL-NO-VF-NEXT:    [[IV_NEXT]] = add nsw i32 [[IV]], 1
+; UNROLL-NO-VF-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF53]], !llvm.loop [[LOOP56:![0-9]+]]
+;
+; SINK-AFTER-LABEL: @sink_into_replication_region_multiple(
+; SINK-AFTER-NEXT:  bb:
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; SINK-AFTER-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 3
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
+; SINK-AFTER-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
+; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; SINK-AFTER-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i32 0
+; SINK-AFTER-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; SINK-AFTER-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 -1, i32 -2, i32 -3>
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE15:%.*]] ]
+; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; SINK-AFTER-NEXT:    [[VEC_IND2:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], 0
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i32 [[OFFSET_IDX]], -1
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], -2
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = add i32 [[OFFSET_IDX]], -3
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = icmp ule <4 x i32> [[VEC_IND2]], [[BROADCAST_SPLAT]]
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
+; SINK-AFTER-NEXT:    br i1 [[TMP7]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; SINK-AFTER:       pred.udiv.if:
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = udiv i32 219220132, [[TMP2]]
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = insertelement <4 x i32> poison, i32 [[TMP8]], i32 0
+; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; SINK-AFTER:       pred.udiv.continue:
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UDIV_IF]] ]
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = extractelement <4 x i1> [[TMP6]], i32 1
+; SINK-AFTER-NEXT:    br i1 [[TMP11]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
+; SINK-AFTER:       pred.udiv.if4:
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = udiv i32 219220132, [[TMP3]]
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP10]], i32 [[TMP12]], i32 1
+; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
+; SINK-AFTER:       pred.udiv.continue5:
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = phi <4 x i32> [ [[TMP10]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF4]] ]
+; SINK-AFTER-NEXT:    [[TMP15:%.*]] = extractelement <4 x i1> [[TMP6]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[TMP15]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
+; SINK-AFTER:       pred.udiv.if6:
+; SINK-AFTER-NEXT:    [[TMP16:%.*]] = udiv i32 219220132, [[TMP4]]
+; SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP16]], i32 2
+; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
+; SINK-AFTER:       pred.udiv.continue7:
+; SINK-AFTER-NEXT:    [[TMP18:%.*]] = phi <4 x i32> [ [[TMP14]], [[PRED_UDIV_CONTINUE5]] ], [ [[TMP17]], [[PRED_UDIV_IF6]] ]
+; SINK-AFTER-NEXT:    [[TMP19:%.*]] = extractelement <4 x i1> [[TMP6]], i32 3
+; SINK-AFTER-NEXT:    br i1 [[TMP19]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; SINK-AFTER:       pred.udiv.if8:
+; SINK-AFTER-NEXT:    [[TMP20:%.*]] = udiv i32 219220132, [[TMP5]]
+; SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP20]], i32 3
+; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; SINK-AFTER:       pred.udiv.continue9:
+; SINK-AFTER-NEXT:    [[TMP22]] = phi <4 x i32> [ [[TMP18]], [[PRED_UDIV_CONTINUE7]] ], [ [[TMP21]], [[PRED_UDIV_IF8]] ]
+; SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP24]] = add <4 x i32> [[VEC_PHI]], [[TMP23]]
+; SINK-AFTER-NEXT:    [[TMP25:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
+; SINK-AFTER-NEXT:    br i1 [[TMP25]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; SINK-AFTER:       pred.store.if:
+; SINK-AFTER-NEXT:    [[TMP26:%.*]] = add i32 [[INDEX]], 0
+; SINK-AFTER-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i32 [[TMP26]]
+; SINK-AFTER-NEXT:    store i32 [[TMP2]], i32* [[TMP27]], align 4
+; SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; SINK-AFTER:       pred.store.continue:
+; SINK-AFTER-NEXT:    [[TMP28:%.*]] = extractelement <4 x i1> [[TMP6]], i32 1
+; SINK-AFTER-NEXT:    br i1 [[TMP28]], label [[PRED_STORE_IF10:%.*]], label [[PRED_STORE_CONTINUE11:%.*]]
+; SINK-AFTER:       pred.store.if10:
+; SINK-AFTER-NEXT:    [[TMP29:%.*]] = add i32 [[INDEX]], 1
+; SINK-AFTER-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP29]]
+; SINK-AFTER-NEXT:    store i32 [[TMP3]], i32* [[TMP30]], align 4
+; SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE11]]
+; SINK-AFTER:       pred.store.continue11:
+; SINK-AFTER-NEXT:    [[TMP31:%.*]] = extractelement <4 x i1> [[TMP6]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[TMP31]], label [[PRED_STORE_IF12:%.*]], label [[PRED_STORE_CONTINUE13:%.*]]
+; SINK-AFTER:       pred.store.if12:
+; SINK-AFTER-NEXT:    [[TMP32:%.*]] = add i32 [[INDEX]], 2
+; SINK-AFTER-NEXT:    [[TMP33:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP32]]
+; SINK-AFTER-NEXT:    store i32 [[TMP4]], i32* [[TMP33]], align 4
+; SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE13]]
+; SINK-AFTER:       pred.store.continue13:
+; SINK-AFTER-NEXT:    [[TMP34:%.*]] = extractelement <4 x i1> [[TMP6]], i32 3
+; SINK-AFTER-NEXT:    br i1 [[TMP34]], label [[PRED_STORE_IF14:%.*]], label [[PRED_STORE_CONTINUE15]]
+; SINK-AFTER:       pred.store.if14:
+; SINK-AFTER-NEXT:    [[TMP35:%.*]] = add i32 [[INDEX]], 3
+; SINK-AFTER-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP35]]
+; SINK-AFTER-NEXT:    store i32 [[TMP5]], i32* [[TMP36]], align 4
+; SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE15]]
+; SINK-AFTER:       pred.store.continue15:
+; SINK-AFTER-NEXT:    [[TMP37:%.*]] = select <4 x i1> [[TMP6]], <4 x i32> [[TMP24]], <4 x i32> [[VEC_PHI]]
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 -4, i32 -4, i32 -4, i32 -4>
+; SINK-AFTER-NEXT:    [[VEC_IND_NEXT3]] = add <4 x i32> [[VEC_IND2]], <i32 4, i32 4, i32 4, i32 4>
+; SINK-AFTER-NEXT:    [[TMP38:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP38]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52]], !llvm.loop [[LOOP56:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[TMP39:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP37]])
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
+; SINK-AFTER-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[BB]] ]
+; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP39]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    br label [[BB2:%.*]]
+; SINK-AFTER:       bb1:
+; SINK-AFTER-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP39]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    ret i32 [[VAR]]
+; SINK-AFTER:       bb2:
+; SINK-AFTER-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; SINK-AFTER-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; SINK-AFTER-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; SINK-AFTER-NEXT:    [[G:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[IV]]
+; SINK-AFTER-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
+; SINK-AFTER-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
+; SINK-AFTER-NEXT:    store i32 [[VAR3]], i32* [[G]], align 4
+; SINK-AFTER-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
+; SINK-AFTER-NEXT:    [[IV_NEXT]] = add nsw i32 [[IV]], 1
+; SINK-AFTER-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
+; SINK-AFTER-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF54]], !llvm.loop [[LOOP57:![0-9]+]]
+;
+; NO-SINK-AFTER-LABEL: @sink_into_replication_region_multiple(
+; NO-SINK-AFTER-NEXT:  bb:
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
+; NO-SINK-AFTER-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
+; NO-SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 3
+; NO-SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 4
+; NO-SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
+; NO-SINK-AFTER-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
+; NO-SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; NO-SINK-AFTER-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i32 0
+; NO-SINK-AFTER-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; NO-SINK-AFTER-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 -1, i32 -2, i32 -3>
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE15:%.*]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_IND2:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[PRED_STORE_CONTINUE15]] ]
+; NO-SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i32 [[OFFSET_IDX]], -1
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], -2
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = add i32 [[OFFSET_IDX]], -3
+; NO-SINK-AFTER-NEXT:    [[TMP6:%.*]] = icmp ule <4 x i32> [[VEC_IND2]], [[BROADCAST_SPLAT]]
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP7]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; NO-SINK-AFTER:       pred.udiv.if:
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = udiv i32 219220132, [[TMP2]]
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = insertelement <4 x i32> poison, i32 [[TMP8]], i32 0
+; NO-SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; NO-SINK-AFTER:       pred.udiv.continue:
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UDIV_IF]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = extractelement <4 x i1> [[TMP6]], i32 1
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP11]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
+; NO-SINK-AFTER:       pred.udiv.if4:
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = udiv i32 219220132, [[TMP3]]
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP10]], i32 [[TMP12]], i32 1
+; NO-SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
+; NO-SINK-AFTER:       pred.udiv.continue5:
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = phi <4 x i32> [ [[TMP10]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF4]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP15:%.*]] = extractelement <4 x i1> [[TMP6]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP15]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
+; NO-SINK-AFTER:       pred.udiv.if6:
+; NO-SINK-AFTER-NEXT:    [[TMP16:%.*]] = udiv i32 219220132, [[TMP4]]
+; NO-SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP16]], i32 2
+; NO-SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
+; NO-SINK-AFTER:       pred.udiv.continue7:
+; NO-SINK-AFTER-NEXT:    [[TMP18:%.*]] = phi <4 x i32> [ [[TMP14]], [[PRED_UDIV_CONTINUE5]] ], [ [[TMP17]], [[PRED_UDIV_IF6]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP19:%.*]] = extractelement <4 x i1> [[TMP6]], i32 3
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP19]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; NO-SINK-AFTER:       pred.udiv.if8:
+; NO-SINK-AFTER-NEXT:    [[TMP20:%.*]] = udiv i32 219220132, [[TMP5]]
+; NO-SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP20]], i32 3
+; NO-SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; NO-SINK-AFTER:       pred.udiv.continue9:
+; NO-SINK-AFTER-NEXT:    [[TMP22]] = phi <4 x i32> [ [[TMP18]], [[PRED_UDIV_CONTINUE7]] ], [ [[TMP21]], [[PRED_UDIV_IF8]] ]
+; NO-SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP24]] = add <4 x i32> [[VEC_PHI]], [[TMP23]]
+; NO-SINK-AFTER-NEXT:    [[TMP25:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP25]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; NO-SINK-AFTER:       pred.store.if:
+; NO-SINK-AFTER-NEXT:    [[TMP26:%.*]] = add i32 [[INDEX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i32 [[TMP26]]
+; NO-SINK-AFTER-NEXT:    store i32 [[TMP2]], i32* [[TMP27]], align 4
+; NO-SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; NO-SINK-AFTER:       pred.store.continue:
+; NO-SINK-AFTER-NEXT:    [[TMP28:%.*]] = extractelement <4 x i1> [[TMP6]], i32 1
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP28]], label [[PRED_STORE_IF10:%.*]], label [[PRED_STORE_CONTINUE11:%.*]]
+; NO-SINK-AFTER:       pred.store.if10:
+; NO-SINK-AFTER-NEXT:    [[TMP29:%.*]] = add i32 [[INDEX]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP29]]
+; NO-SINK-AFTER-NEXT:    store i32 [[TMP3]], i32* [[TMP30]], align 4
+; NO-SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE11]]
+; NO-SINK-AFTER:       pred.store.continue11:
+; NO-SINK-AFTER-NEXT:    [[TMP31:%.*]] = extractelement <4 x i1> [[TMP6]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP31]], label [[PRED_STORE_IF12:%.*]], label [[PRED_STORE_CONTINUE13:%.*]]
+; NO-SINK-AFTER:       pred.store.if12:
+; NO-SINK-AFTER-NEXT:    [[TMP32:%.*]] = add i32 [[INDEX]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP33:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP32]]
+; NO-SINK-AFTER-NEXT:    store i32 [[TMP4]], i32* [[TMP33]], align 4
+; NO-SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE13]]
+; NO-SINK-AFTER:       pred.store.continue13:
+; NO-SINK-AFTER-NEXT:    [[TMP34:%.*]] = extractelement <4 x i1> [[TMP6]], i32 3
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP34]], label [[PRED_STORE_IF14:%.*]], label [[PRED_STORE_CONTINUE15]]
+; NO-SINK-AFTER:       pred.store.if14:
+; NO-SINK-AFTER-NEXT:    [[TMP35:%.*]] = add i32 [[INDEX]], 3
+; NO-SINK-AFTER-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP35]]
+; NO-SINK-AFTER-NEXT:    store i32 [[TMP5]], i32* [[TMP36]], align 4
+; NO-SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE15]]
+; NO-SINK-AFTER:       pred.store.continue15:
+; NO-SINK-AFTER-NEXT:    [[TMP37:%.*]] = select <4 x i1> [[TMP6]], <4 x i32> [[TMP24]], <4 x i32> [[VEC_PHI]]
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 -4, i32 -4, i32 -4, i32 -4>
+; NO-SINK-AFTER-NEXT:    [[VEC_IND_NEXT3]] = add <4 x i32> [[VEC_IND2]], <i32 4, i32 4, i32 4, i32 4>
+; NO-SINK-AFTER-NEXT:    [[TMP38:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP38]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF52]], !llvm.loop [[LOOP56:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[TMP39:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP37]])
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[BB]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP39]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    br label [[BB2:%.*]]
+; NO-SINK-AFTER:       bb1:
+; NO-SINK-AFTER-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP39]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    ret i32 [[VAR]]
+; NO-SINK-AFTER:       bb2:
+; NO-SINK-AFTER-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; NO-SINK-AFTER-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; NO-SINK-AFTER-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; NO-SINK-AFTER-NEXT:    [[G:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[IV]]
+; NO-SINK-AFTER-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
+; NO-SINK-AFTER-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
+; NO-SINK-AFTER-NEXT:    store i32 [[VAR3]], i32* [[G]], align 4
+; NO-SINK-AFTER-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
+; NO-SINK-AFTER-NEXT:    [[IV_NEXT]] = add nsw i32 [[IV]], 1
+; NO-SINK-AFTER-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
+; NO-SINK-AFTER-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF54]], !llvm.loop [[LOOP57:![0-9]+]]
 ;
 bb:
   br label %bb2
 
   bb1:                                              ; preds = %bb2
-  %tmp = phi i32 [ %tmp6, %bb2 ]
-  ret i32 %tmp
+  %var = phi i32 [ %var6, %bb2 ]
+  ret i32 %var
 
   bb2:                                              ; preds = %bb2, %bb
-  %tmp3 = phi i32 [ %tmp8, %bb2 ], [ %y, %bb ]
+  %var3 = phi i32 [ %var8, %bb2 ], [ %y, %bb ]
   %iv = phi i32 [ %iv.next, %bb2 ], [ 0, %bb ]
-  %tmp4 = phi i32 [ %tmp7, %bb2 ], [ 0, %bb ]
-  %tmp5 = phi i32 [ %tmp6, %bb2 ], [ 0, %bb ]
+  %var4 = phi i32 [ %var7, %bb2 ], [ 0, %bb ]
+  %var5 = phi i32 [ %var6, %bb2 ], [ 0, %bb ]
   %g = getelementptr inbounds i32, i32* %x, i32 %iv
-  %tmp6 = add i32 %tmp5, %tmp4
-  %tmp7 = udiv i32 219220132, %tmp3
-  store i32 %tmp3, i32* %g, align 4
-  %tmp8 = add nsw i32 %tmp3, -1
+  %var6 = add i32 %var5, %var4
+  %var7 = udiv i32 219220132, %var3
+  store i32 %var3, i32* %g, align 4
+  %var8 = add nsw i32 %var3, -1
   %iv.next = add nsw i32 %iv, 1
-  %tmp9 = icmp slt i32 %tmp3, 2
-  br i1 %tmp9, label %bb1, label %bb2, !prof !2
+  %var9 = icmp slt i32 %var3, 2
+  br i1 %var9, label %bb1, label %bb2, !prof !2
 }
 
 ; %vec.dead will be marked as dead instruction in the vector loop and no recipe
 ; will be created for it. Make sure a valid sink target is used.
 define void @sink_after_dead_inst(i32* %A.ptr) {
-; CHECK-LABEL: @sink_after_dead_inst
-; CHECK-LABEL: vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %vector.ph ], [ [[INDEX_NEXT]], %vector.body ]
+; CHECK-LABEL: @sink_after_dead_inst(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = zext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[SEXT:%.*]] = shl i64 [[OFFSET_IDX]], 48
-; CHECK-NEXT:    [[SHIFT:%.*]] = ashr exact i64 [[SEXT]], 48
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, i32* %A.ptr, i64 [[SHIFT]]
-; CHECK-NEXT:    [[CAST:%.*]] = bitcast i32* [[GEP]]  to <4 x i32>*
-; CHECK-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[CAST]], align 4
-; CHECK-NEXT:    [[INDEX_NEXT:%.*]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
-; CHECK-NEXT:    br i1 [[EC]], label %middle.block, label %vector.body
-
+; CHECK-NEXT:    [[TMP0:%.*]] = ashr exact i64 [[SEXT]], 48
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <4 x i32>*
+; CHECK-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP2]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
+; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP58:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    br i1 undef, label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP59:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; UNROLL-LABEL: @sink_after_dead_inst(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = zext i32 [[INDEX]] to i64
+; UNROLL-NEXT:    [[SEXT:%.*]] = shl i64 [[OFFSET_IDX]], 48
+; UNROLL-NEXT:    [[TMP0:%.*]] = ashr exact i64 [[SEXT]], 48
+; UNROLL-NEXT:    [[TMP1:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i64 [[TMP0]]
+; UNROLL-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP2]], align 4
+; UNROLL-NEXT:    [[TMP3:%.*]] = getelementptr i32, i32* [[TMP1]], i64 4
+; UNROLL-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <4 x i32>*
+; UNROLL-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP4]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; UNROLL-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
+; UNROLL-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP58:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[LOOP:%.*]]
+; UNROLL:       loop:
+; UNROLL-NEXT:    br i1 undef, label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP59:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @sink_after_dead_inst(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 0, i16 1, i16 2, i16 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i16 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i16 [[OFFSET_IDX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i16 [[OFFSET_IDX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i16 [[OFFSET_IDX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i16 [[OFFSET_IDX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i16 [[OFFSET_IDX]], 5
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i16 [[OFFSET_IDX]], 6
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i16 [[OFFSET_IDX]], 7
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add <4 x i16> [[STEP_ADD]], <i16 1, i16 1, i16 1, i16 1>
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = or <4 x i16> [[TMP8]], [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = or <4 x i16> [[TMP9]], [[TMP9]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = zext <4 x i16> [[TMP10]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP13]] = zext <4 x i16> [[TMP11]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP12]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = shufflevector <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = icmp eq <4 x i32> [[TMP14]], <i32 15, i32 15, i32 15, i32 15>
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = icmp eq <4 x i32> [[TMP15]], <i32 15, i32 15, i32 15, i32 15>
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = icmp eq <4 x i1> [[TMP16]], <i1 true, i1 true, i1 true, i1 true>
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = icmp eq <4 x i1> [[TMP17]], <i1 true, i1 true, i1 true, i1 true>
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = and <4 x i1> [[TMP16]], [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = and <4 x i1> [[TMP17]], [[TMP19]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = zext <4 x i1> [[TMP20]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = zext <4 x i1> [[TMP21]] to <4 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i16 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr i32, i32* [[TMP24]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = bitcast i32* [[TMP26]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP27]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = getelementptr i32, i32* [[TMP24]], i32 4
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = bitcast i32* [[TMP28]] to <4 x i32>*
+; UNROLL-NO-IC-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP29]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], <i16 4, i16 4, i16 4, i16 4>
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP58:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 16, 16
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP13]], i32 3
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP13]], i32 2
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
+; UNROLL-NO-IC:       loop:
+; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[FOR_PREV:%.*]], [[LOOP]] ]
+; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR]], 15
+; UNROLL-NO-IC-NEXT:    [[C:%.*]] = icmp eq i1 [[CMP]], true
+; UNROLL-NO-IC-NEXT:    [[VEC_DEAD:%.*]] = and i1 [[C]], true
+; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; UNROLL-NO-IC-NEXT:    [[B1:%.*]] = or i16 [[IV_NEXT]], [[IV_NEXT]]
+; UNROLL-NO-IC-NEXT:    [[B3:%.*]] = and i1 [[CMP]], [[C]]
+; UNROLL-NO-IC-NEXT:    [[FOR_PREV]] = zext i16 [[B1]] to i32
+; UNROLL-NO-IC-NEXT:    [[EXT:%.*]] = zext i1 [[B3]] to i32
+; UNROLL-NO-IC-NEXT:    [[A_GEP:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[IV]]
+; UNROLL-NO-IC-NEXT:    store i32 0, i32* [[A_GEP]], align 4
+; UNROLL-NO-IC-NEXT:    br i1 [[VEC_DEAD]], label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP59:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; UNROLL-NO-VF-LABEL: @sink_after_dead_inst(
+; UNROLL-NO-VF-NEXT:  entry:
+; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i16 [[OFFSET_IDX]], 0
+; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i16 [[OFFSET_IDX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i16 [[INDUCTION]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i16 [[INDUCTION1]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = or i16 [[TMP0]], [[TMP0]]
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = or i16 [[TMP1]], [[TMP1]]
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = zext i16 [[TMP2]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP5]] = zext i16 [[TMP3]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[VECTOR_RECUR]], 15
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[TMP4]], 15
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = icmp eq i1 [[TMP6]], true
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = icmp eq i1 [[TMP7]], true
+; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = and i1 [[TMP6]], [[TMP8]]
+; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = and i1 [[TMP7]], [[TMP9]]
+; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = zext i1 [[TMP10]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = zext i1 [[TMP11]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i16 [[INDUCTION]]
+; UNROLL-NO-VF-NEXT:    [[TMP15:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[INDUCTION1]]
+; UNROLL-NO-VF-NEXT:    store i32 0, i32* [[TMP14]], align 4
+; UNROLL-NO-VF-NEXT:    store i32 0, i32* [[TMP15]], align 4
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP57:![0-9]+]]
+; UNROLL-NO-VF:       middle.block:
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i32 16, 16
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF:       scalar.ph:
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    br label [[LOOP:%.*]]
+; UNROLL-NO-VF:       loop:
+; UNROLL-NO-VF-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[FOR_PREV:%.*]], [[LOOP]] ]
+; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR]], 15
+; UNROLL-NO-VF-NEXT:    [[C:%.*]] = icmp eq i1 [[CMP]], true
+; UNROLL-NO-VF-NEXT:    [[VEC_DEAD:%.*]] = and i1 [[C]], true
+; UNROLL-NO-VF-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; UNROLL-NO-VF-NEXT:    [[B1:%.*]] = or i16 [[IV_NEXT]], [[IV_NEXT]]
+; UNROLL-NO-VF-NEXT:    [[B3:%.*]] = and i1 [[CMP]], [[C]]
+; UNROLL-NO-VF-NEXT:    [[FOR_PREV]] = zext i16 [[B1]] to i32
+; UNROLL-NO-VF-NEXT:    [[EXT:%.*]] = zext i1 [[B3]] to i32
+; UNROLL-NO-VF-NEXT:    [[A_GEP:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[IV]]
+; UNROLL-NO-VF-NEXT:    store i32 0, i32* [[A_GEP]], align 4
+; UNROLL-NO-VF-NEXT:    br i1 [[VEC_DEAD]], label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP58:![0-9]+]]
+; UNROLL-NO-VF:       for.end:
+; UNROLL-NO-VF-NEXT:    ret void
+;
+; SINK-AFTER-LABEL: @sink_after_dead_inst(
+; SINK-AFTER-NEXT:  entry:
+; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; SINK-AFTER:       vector.body:
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 0, i16 1, i16 2, i16 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i16 [[OFFSET_IDX]], 0
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i16 [[OFFSET_IDX]], 1
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i16 [[OFFSET_IDX]], 2
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i16 [[OFFSET_IDX]], 3
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = or <4 x i16> [[TMP4]], [[TMP4]]
+; SINK-AFTER-NEXT:    [[TMP6]] = zext <4 x i16> [[TMP5]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP6]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; SINK-AFTER-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[TMP7]], <i32 15, i32 15, i32 15, i32 15>
+; SINK-AFTER-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i1> [[TMP8]], <i1 true, i1 true, i1 true, i1 true>
+; SINK-AFTER-NEXT:    [[TMP10:%.*]] = and <4 x i1> [[TMP8]], [[TMP9]]
+; SINK-AFTER-NEXT:    [[TMP11:%.*]] = zext <4 x i1> [[TMP10]] to <4 x i32>
+; SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i16 [[TMP0]]
+; SINK-AFTER-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[TMP12]], i32 0
+; SINK-AFTER-NEXT:    [[TMP14:%.*]] = bitcast i32* [[TMP13]] to <4 x i32>*
+; SINK-AFTER-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP14]], align 4
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
+; SINK-AFTER-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
+; SINK-AFTER-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP58:![0-9]+]]
+; SINK-AFTER:       middle.block:
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 16, 16
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP6]], i32 3
+; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP6]], i32 2
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER:       scalar.ph:
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    br label [[LOOP:%.*]]
+; SINK-AFTER:       loop:
+; SINK-AFTER-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[FOR_PREV:%.*]], [[LOOP]] ]
+; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR]], 15
+; SINK-AFTER-NEXT:    [[C:%.*]] = icmp eq i1 [[CMP]], true
+; SINK-AFTER-NEXT:    [[VEC_DEAD:%.*]] = and i1 [[C]], true
+; SINK-AFTER-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; SINK-AFTER-NEXT:    [[B1:%.*]] = or i16 [[IV_NEXT]], [[IV_NEXT]]
+; SINK-AFTER-NEXT:    [[B3:%.*]] = and i1 [[CMP]], [[C]]
+; SINK-AFTER-NEXT:    [[FOR_PREV]] = zext i16 [[B1]] to i32
+; SINK-AFTER-NEXT:    [[EXT:%.*]] = zext i1 [[B3]] to i32
+; SINK-AFTER-NEXT:    [[A_GEP:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[IV]]
+; SINK-AFTER-NEXT:    store i32 0, i32* [[A_GEP]], align 4
+; SINK-AFTER-NEXT:    br i1 [[VEC_DEAD]], label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP59:![0-9]+]]
+; SINK-AFTER:       for.end:
+; SINK-AFTER-NEXT:    ret void
+;
+; NO-SINK-AFTER-LABEL: @sink_after_dead_inst(
+; NO-SINK-AFTER-NEXT:  entry:
+; NO-SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-SINK-AFTER:       vector.ph:
+; NO-SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-SINK-AFTER:       vector.body:
+; NO-SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 0, i16 1, i16 2, i16 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
+; NO-SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; NO-SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i16 [[OFFSET_IDX]], 0
+; NO-SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i16 [[OFFSET_IDX]], 1
+; NO-SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i16 [[OFFSET_IDX]], 2
+; NO-SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i16 [[OFFSET_IDX]], 3
+; NO-SINK-AFTER-NEXT:    [[TMP4:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
+; NO-SINK-AFTER-NEXT:    [[TMP5:%.*]] = or <4 x i16> [[TMP4]], [[TMP4]]
+; NO-SINK-AFTER-NEXT:    [[TMP6]] = zext <4 x i16> [[TMP5]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP6]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; NO-SINK-AFTER-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[TMP7]], <i32 15, i32 15, i32 15, i32 15>
+; NO-SINK-AFTER-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i1> [[TMP8]], <i1 true, i1 true, i1 true, i1 true>
+; NO-SINK-AFTER-NEXT:    [[TMP10:%.*]] = and <4 x i1> [[TMP8]], [[TMP9]]
+; NO-SINK-AFTER-NEXT:    [[TMP11:%.*]] = zext <4 x i1> [[TMP10]] to <4 x i32>
+; NO-SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i16 [[TMP0]]
+; NO-SINK-AFTER-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[TMP12]], i32 0
+; NO-SINK-AFTER-NEXT:    [[TMP14:%.*]] = bitcast i32* [[TMP13]] to <4 x i32>*
+; NO-SINK-AFTER-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP14]], align 4
+; NO-SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; NO-SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
+; NO-SINK-AFTER-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
+; NO-SINK-AFTER-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP58:![0-9]+]]
+; NO-SINK-AFTER:       middle.block:
+; NO-SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 16, 16
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP6]], i32 3
+; NO-SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP6]], i32 2
+; NO-SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-SINK-AFTER:       scalar.ph:
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; NO-SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; NO-SINK-AFTER-NEXT:    br label [[LOOP:%.*]]
+; NO-SINK-AFTER:       loop:
+; NO-SINK-AFTER-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; NO-SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[FOR_PREV:%.*]], [[LOOP]] ]
+; NO-SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR]], 15
+; NO-SINK-AFTER-NEXT:    [[C:%.*]] = icmp eq i1 [[CMP]], true
+; NO-SINK-AFTER-NEXT:    [[VEC_DEAD:%.*]] = and i1 [[C]], true
+; NO-SINK-AFTER-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; NO-SINK-AFTER-NEXT:    [[B1:%.*]] = or i16 [[IV_NEXT]], [[IV_NEXT]]
+; NO-SINK-AFTER-NEXT:    [[B3:%.*]] = and i1 [[CMP]], [[C]]
+; NO-SINK-AFTER-NEXT:    [[FOR_PREV]] = zext i16 [[B1]] to i32
+; NO-SINK-AFTER-NEXT:    [[EXT:%.*]] = zext i1 [[B3]] to i32
+; NO-SINK-AFTER-NEXT:    [[A_GEP:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[IV]]
+; NO-SINK-AFTER-NEXT:    store i32 0, i32* [[A_GEP]], align 4
+; NO-SINK-AFTER-NEXT:    br i1 [[VEC_DEAD]], label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP59:![0-9]+]]
+; NO-SINK-AFTER:       for.end:
+; NO-SINK-AFTER-NEXT:    ret void
+;
 entry:
   br label %loop
 

diff  --git a/llvm/test/Transforms/LoopVectorize/float-induction.ll b/llvm/test/Transforms/LoopVectorize/float-induction.ll
index b777fe21666f..86ec6f5d4190 100644
--- a/llvm/test/Transforms/LoopVectorize/float-induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/float-induction.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s  -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -dce -instcombine -S | FileCheck --check-prefix VEC4_INTERL1 %s
 ; RUN: opt < %s  -loop-vectorize -force-vector-interleave=2 -force-vector-width=4 -dce -instcombine -S | FileCheck --check-prefix VEC4_INTERL2 %s
 ; RUN: opt < %s  -loop-vectorize -force-vector-interleave=2 -force-vector-width=1 -dce -instcombine -S | FileCheck --check-prefix VEC1_INTERL2 %s
@@ -13,72 +14,235 @@
 ;  }
 ;}
 
+
+
+
+define void @fp_iv_loop1_fast_FMF(float %init, float* noalias nocapture %A, i32 %N) {
 ; VEC4_INTERL1-LABEL: @fp_iv_loop1_fast_FMF(
+; VEC4_INTERL1-NEXT:  entry:
+; VEC4_INTERL1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC4_INTERL1-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; VEC4_INTERL1:       for.body.lr.ph:
+; VEC4_INTERL1-NEXT:    [[FPINC:%.*]] = load float, float* @fp_inc, align 4
+; VEC4_INTERL1-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC4_INTERL1-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC4_INTERL1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
+; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
-; VEC4_INTERL1:         [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float %init, i32 0
+; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
+; VEC4_INTERL1-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[TMP3:%.*]] = fmul fast float [[FPINC]], [[CAST_CRD]]
+; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP3]]
+; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i32 0
 ; VEC4_INTERL1-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <4 x float> poison, float %fpinc, i32 0
+; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <4 x float> poison, float [[FPINC]], i32 0
 ; VEC4_INTERL1-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT2]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL1-NEXT:    [[TMP5:%.*]] = fmul fast <4 x float> [[DOTSPLAT3]], <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
-; VEC4_INTERL1-NEXT:    [[INDUCTION4:%.*]] = fsub fast <4 x float> [[DOTSPLAT]], [[TMP5]]
-; VEC4_INTERL1-NEXT:    [[TMP6:%.*]] = fmul fast float %fpinc, 4.000000e+00
-; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <4 x float> poison, float [[TMP6]], i32 0
-; VEC4_INTERL1-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT5]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL1-NEXT:    br label %vector.body
+; VEC4_INTERL1-NEXT:    [[TMP4:%.*]] = fmul fast <4 x float> [[DOTSPLAT3]], <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
+; VEC4_INTERL1-NEXT:    [[INDUCTION:%.*]] = fsub fast <4 x float> [[DOTSPLAT]], [[TMP4]]
+; VEC4_INTERL1-NEXT:    [[TMP5:%.*]] = fmul fast float [[FPINC]], 4.000000e+00
+; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT4:%.*]] = insertelement <4 x float> poison, float [[TMP5]], i32 0
+; VEC4_INTERL1-NEXT:    [[DOTSPLAT5:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT4]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL1-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC4_INTERL1:       vector.body:
-; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ [[INDUCTION4]], %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDEX]]
-; VEC4_INTERL1-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP8]] to <4 x float>*
-; VEC4_INTERL1-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP9]], align 4
+; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL1-NEXT:    [[TMP7:%.*]] = bitcast float* [[TMP6]] to <4 x float>*
+; VEC4_INTERL1-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP7]], align 4
 ; VEC4_INTERL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT]] = fsub fast <4 x float> [[VEC_IND]], [[DOTSPLAT6]]
-; VEC4_INTERL1:         br i1 {{.*}}, label %middle.block, label %vector.body
-
+; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT]] = fsub fast <4 x float> [[VEC_IND]], [[DOTSPLAT5]]
+; VEC4_INTERL1-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VEC4_INTERL1:       middle.block:
+; VEC4_INTERL1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL1:       scalar.ph:
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL1-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL1:       for.body:
+; VEC4_INTERL1-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[X_05:%.*]] = phi float [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC4_INTERL1-NEXT:    store float [[X_05]], float* [[ARRAYIDX]], align 4
+; VEC4_INTERL1-NEXT:    [[ADD]] = fsub fast float [[X_05]], [[FPINC]]
+; VEC4_INTERL1-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC4_INTERL1-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC4_INTERL1-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC4_INTERL1-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; VEC4_INTERL1:       for.end.loopexit:
+; VEC4_INTERL1-NEXT:    br label [[FOR_END]]
+; VEC4_INTERL1:       for.end:
+; VEC4_INTERL1-NEXT:    ret void
+;
 ; VEC4_INTERL2-LABEL: @fp_iv_loop1_fast_FMF(
+; VEC4_INTERL2-NEXT:  entry:
+; VEC4_INTERL2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC4_INTERL2-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; VEC4_INTERL2:       for.body.lr.ph:
+; VEC4_INTERL2-NEXT:    [[FPINC:%.*]] = load float, float* @fp_inc, align 4
+; VEC4_INTERL2-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC4_INTERL2-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC4_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
+; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL2:       vector.ph:
-; VEC4_INTERL2:         [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float %init, i32 0
+; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
+; VEC4_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[FPINC]], [[CAST_CRD]]
+; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP3]]
+; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i32 0
 ; VEC4_INTERL2-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT3:%.*]] = insertelement <4 x float> poison, float %fpinc, i32 0
-; VEC4_INTERL2-NEXT:    [[DOTSPLAT4:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT3]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast <4 x float> [[DOTSPLAT4]], <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
-; VEC4_INTERL2-NEXT:    [[INDUCTION5:%.*]] = fsub fast <4 x float> [[DOTSPLAT]], [[TMP5]]
-; VEC4_INTERL2-NEXT:    [[TMP6:%.*]] = fmul fast float %fpinc, 4.000000e+00
-; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT6:%.*]] = insertelement <4 x float> poison, float [[TMP6]], i32 0
-; VEC4_INTERL2-NEXT:    [[DOTSPLAT7:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT6]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL2-NEXT:    br label %vector.body
+; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <4 x float> poison, float [[FPINC]], i32 0
+; VEC4_INTERL2-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT2]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL2-NEXT:    [[TMP4:%.*]] = fmul fast <4 x float> [[DOTSPLAT3]], <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
+; VEC4_INTERL2-NEXT:    [[INDUCTION:%.*]] = fsub fast <4 x float> [[DOTSPLAT]], [[TMP4]]
+; VEC4_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[FPINC]], 4.000000e+00
+; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT4:%.*]] = insertelement <4 x float> poison, float [[TMP5]], i32 0
+; VEC4_INTERL2-NEXT:    [[DOTSPLAT5:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT4]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC4_INTERL2:       vector.body:
-; VEC4_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL2-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ [[INDUCTION5]], %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL2-NEXT:    [[STEP_ADD:%.*]] = fsub fast <4 x float> [[VEC_IND]], [[DOTSPLAT7]]
-; VEC4_INTERL2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDEX]]
-; VEC4_INTERL2-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to <4 x float>*
-; VEC4_INTERL2-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP10]], align 4
-; VEC4_INTERL2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, float* [[TMP9]], i64 4
-; VEC4_INTERL2-NEXT:    [[TMP12:%.*]] = bitcast float* [[TMP11]] to <4 x float>*
-; VEC4_INTERL2-NEXT:    store <4 x float> [[STEP_ADD]], <4 x float>* [[TMP12]], align 4
+; VEC4_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[STEP_ADD:%.*]] = fsub fast <4 x float> [[VEC_IND]], [[DOTSPLAT5]]
+; VEC4_INTERL2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL2-NEXT:    [[TMP7:%.*]] = bitcast float* [[TMP6]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP7]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 4
+; VEC4_INTERL2-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP8]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[STEP_ADD]], <4 x float>* [[TMP9]], align 4
 ; VEC4_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; VEC4_INTERL2-NEXT:    [[VEC_IND_NEXT]] = fsub fast <4 x float> [[STEP_ADD]], [[DOTSPLAT7]]
-; VEC4_INTERL2:         br i1 {{.*}}, label %middle.block, label %vector.body
-
+; VEC4_INTERL2-NEXT:    [[VEC_IND_NEXT]] = fsub fast <4 x float> [[STEP_ADD]], [[DOTSPLAT5]]
+; VEC4_INTERL2-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VEC4_INTERL2:       middle.block:
+; VEC4_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL2:       scalar.ph:
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL2:       for.body:
+; VEC4_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[X_05:%.*]] = phi float [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC4_INTERL2-NEXT:    store float [[X_05]], float* [[ARRAYIDX]], align 4
+; VEC4_INTERL2-NEXT:    [[ADD]] = fsub fast float [[X_05]], [[FPINC]]
+; VEC4_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC4_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC4_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC4_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; VEC4_INTERL2:       for.end.loopexit:
+; VEC4_INTERL2-NEXT:    br label [[FOR_END]]
+; VEC4_INTERL2:       for.end:
+; VEC4_INTERL2-NEXT:    ret void
+;
 ; VEC1_INTERL2-LABEL: @fp_iv_loop1_fast_FMF(
+; VEC1_INTERL2-NEXT:  entry:
+; VEC1_INTERL2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC1_INTERL2-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; VEC1_INTERL2:       for.body.lr.ph:
+; VEC1_INTERL2-NEXT:    [[FPINC:%.*]] = load float, float* @fp_inc, align 4
+; VEC1_INTERL2-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC1_INTERL2-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC1_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC1_INTERL2:       vector.ph:
-; VEC1_INTERL2:         br label %vector.body
+; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[FPINC]], [[CAST_CRD]]
+; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP3]]
+; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC1_INTERL2:       vector.body:
-; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
+; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
-; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = sitofp i64 [[INDEX]] to float
-; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = fmul fast float %fpinc, [[TMP6]]
-; VEC1_INTERL2-NEXT:    [[FP_OFFSET_IDX:%.*]] = fsub fast float %init, [[TMP7]]
-; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = fsub fast float [[FP_OFFSET_IDX]], %fpinc
-; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDEX]]
-; VEC1_INTERL2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDUCTION2]]
-; VEC1_INTERL2-NEXT:    store float [[FP_OFFSET_IDX]], float* [[TMP9]], align 4
-; VEC1_INTERL2-NEXT:    store float [[TMP8]], float* [[TMP10]], align 4
+; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[FPINC]], [[TMP4]]
+; VEC1_INTERL2-NEXT:    [[OFFSET_IDX:%.*]] = fsub fast float [[INIT]], [[TMP5]]
+; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fsub fast float [[OFFSET_IDX]], [[FPINC]]
+; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
+; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX]], float* [[TMP7]], align 4
+; VEC1_INTERL2-NEXT:    store float [[TMP6]], float* [[TMP8]], align 4
 ; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VEC1_INTERL2:         br i1 {{.*}}, label %middle.block, label %vector.body
-
-define void @fp_iv_loop1_fast_FMF(float %init, float* noalias nocapture %A, i32 %N) {
+; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VEC1_INTERL2:       middle.block:
+; VEC1_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC1_INTERL2:       scalar.ph:
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_LR_PH]] ]
+; VEC1_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC1_INTERL2:       for.body:
+; VEC1_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; VEC1_INTERL2-NEXT:    [[X_05:%.*]] = phi float [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
+; VEC1_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC1_INTERL2-NEXT:    store float [[X_05]], float* [[ARRAYIDX]], align 4
+; VEC1_INTERL2-NEXT:    [[ADD]] = fsub fast float [[X_05]], [[FPINC]]
+; VEC1_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC1_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC1_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC1_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; VEC1_INTERL2:       for.end.loopexit:
+; VEC1_INTERL2-NEXT:    br label [[FOR_END]]
+; VEC1_INTERL2:       for.end:
+; VEC1_INTERL2-NEXT:    ret void
+;
+; VEC2_INTERL1_PRED_STORE-LABEL: @fp_iv_loop1_fast_FMF(
+; VEC2_INTERL1_PRED_STORE-NEXT:  entry:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; VEC2_INTERL1_PRED_STORE:       for.body.lr.ph:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[FPINC:%.*]] = load float, float* @fp_inc, align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.ph:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP3:%.*]] = fmul fast float [[FPINC]], [[CAST_CRD]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP3]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <2 x float> poison, float [[FPINC]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT2]], <2 x float> poison, <2 x i32> zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP4:%.*]] = fmul fast <2 x float> [[DOTSPLAT3]], <float 0.000000e+00, float 1.000000e+00>
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDUCTION:%.*]] = fsub fast <2 x float> [[DOTSPLAT]], [[TMP4]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP5:%.*]] = fmul fast float [[FPINC]], 2.000000e+00
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT4:%.*]] = insertelement <2 x float> poison, float [[TMP5]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLAT5:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT4]], <2 x float> poison, <2 x i32> zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND:%.*]] = phi <2 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP7:%.*]] = bitcast float* [[TMP6]] to <2 x float>*
+; VEC2_INTERL1_PRED_STORE-NEXT:    store <2 x float> [[VEC_IND]], <2 x float>* [[TMP7]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND_NEXT]] = fsub fast <2 x float> [[VEC_IND]], [[DOTSPLAT5]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       middle.block:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP_N]], label [[FOR_END]], label [[FOR_BODY]]
+; VEC2_INTERL1_PRED_STORE:       for.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[X_05:%.*]] = phi float [ [[ADD:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_LR_PH]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[X_05]], float* [[ARRAYIDX]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ADD]] = fsub fast float [[X_05]], [[FPINC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       for.end:
+; VEC2_INTERL1_PRED_STORE-NEXT:    ret void
+;
 entry:
   %cmp4 = icmp sgt i32 %N, 0
   br i1 %cmp4, label %for.body.lr.ph, label %for.end
@@ -109,6 +273,10 @@ for.end:                                          ; preds = %for.end.loopexit, %
 ; 'fast' spontaneously - FMF should propagate from the original IR.
 
 define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i32 %N) {
+;
+;
+;
+;
 ; VEC4_INTERL1-LABEL: @fp_iv_loop1_reassoc_FMF(
 ; VEC4_INTERL1-NEXT:  entry:
 ; VEC4_INTERL1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
@@ -144,7 +312,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i
 ; VEC4_INTERL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT]] = fsub reassoc <4 x float> [[VEC_IND]], [[DOTSPLAT5]]
 ; VEC4_INTERL1-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; VEC4_INTERL1-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP4:!llvm.loop !.*]]
+; VEC4_INTERL1-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VEC4_INTERL1:       middle.block:
 ; VEC4_INTERL1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
 ; VEC4_INTERL1-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
@@ -161,7 +329,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i
 ; VEC4_INTERL1-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; VEC4_INTERL1-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
 ; VEC4_INTERL1-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
-; VEC4_INTERL1-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], [[LOOP5:!llvm.loop !.*]]
+; VEC4_INTERL1-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; VEC4_INTERL1:       for.end.loopexit:
 ; VEC4_INTERL1-NEXT:    br label [[FOR_END]]
 ; VEC4_INTERL1:       for.end:
@@ -206,7 +374,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i
 ; VEC4_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; VEC4_INTERL2-NEXT:    [[VEC_IND_NEXT]] = fsub reassoc <4 x float> [[STEP_ADD]], [[DOTSPLAT5]]
 ; VEC4_INTERL2-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; VEC4_INTERL2-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP4:!llvm.loop !.*]]
+; VEC4_INTERL2-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VEC4_INTERL2:       middle.block:
 ; VEC4_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
 ; VEC4_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
@@ -223,7 +391,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i
 ; VEC4_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; VEC4_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
 ; VEC4_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
-; VEC4_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], [[LOOP5:!llvm.loop !.*]]
+; VEC4_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; VEC4_INTERL2:       for.end.loopexit:
 ; VEC4_INTERL2-NEXT:    br label [[FOR_END]]
 ; VEC4_INTERL2:       for.end:
@@ -261,7 +429,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i
 ; VEC1_INTERL2-NEXT:    store float [[TMP8]], float* [[TMP10]], align 4
 ; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC1_INTERL2-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; VEC1_INTERL2-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VEC1_INTERL2-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; VEC1_INTERL2:       middle.block:
 ; VEC1_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
 ; VEC1_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
@@ -272,13 +440,13 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i
 ; VEC1_INTERL2:       for.body:
 ; VEC1_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; VEC1_INTERL2-NEXT:    [[X_05:%.*]] = phi float [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
-; VEC1_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDVARS_IV]]
+; VEC1_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
 ; VEC1_INTERL2-NEXT:    store float [[X_05]], float* [[ARRAYIDX]], align 4
 ; VEC1_INTERL2-NEXT:    [[ADD]] = fsub reassoc float [[X_05]], [[FPINC]]
 ; VEC1_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; VEC1_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
 ; VEC1_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
-; VEC1_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT:%.*]], label [[FOR_BODY]]
+; VEC1_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VEC1_INTERL2:       for.end.loopexit:
 ; VEC1_INTERL2-NEXT:    br label [[FOR_END]]
 ; VEC1_INTERL2:       for.end:
@@ -319,7 +487,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND_NEXT]] = fsub reassoc <2 x float> [[VEC_IND]], [[DOTSPLAT5]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP4:!llvm.loop !.*]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VEC2_INTERL1_PRED_STORE:       middle.block:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP_N]], label [[FOR_END]], label [[FOR_BODY]]
@@ -332,7 +500,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, float* noalias nocapture %A, i
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
-; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], [[LOOP5:!llvm.loop !.*]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; VEC2_INTERL1_PRED_STORE:       for.end:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    ret void
 ;
@@ -370,108 +538,594 @@ for.end:                                          ; preds = %for.end.loopexit, %
 ;  }
 ;}
 
-; VEC4_INTERL1-LABEL: @fp_iv_loop2(
-; VEC4_INTERL1:       vector.ph:
-; VEC4_INTERL1:         [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float %init, i32 0
-; VEC4_INTERL1-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL1-NEXT:    [[INDUCTION2:%.*]] = fadd fast <4 x float> [[DOTSPLAT]], <float 0.000000e+00, float 5.000000e-01, float 1.000000e+00, float 1.500000e+00>
-; VEC4_INTERL1-NEXT:    br label %vector.body
-; VEC4_INTERL1:       vector.body:
-; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ [[INDUCTION2]], %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDEX]]
-; VEC4_INTERL1-NEXT:    [[TMP8:%.*]] = bitcast float* [[TMP7]] to <4 x float>*
-; VEC4_INTERL1-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP8]], align 4
-; VEC4_INTERL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT]] = fadd fast <4 x float> [[VEC_IND]], <float 2.000000e+00, float 2.000000e+00, float 2.000000e+00, float 2.000000e+00>
-; VEC4_INTERL1:         br i1 {{.*}}, label %middle.block, label %vector.body
 
 define void @fp_iv_loop2(float %init, float* noalias nocapture %A, i32 %N) #0 {
-entry:
-  %cmp4 = icmp sgt i32 %N, 0
-  br i1 %cmp4, label %for.body.preheader, label %for.end
-
-for.body.preheader:                               ; preds = %entry
-  br label %for.body
-
-for.body:                                         ; preds = %for.body.preheader, %for.body
-  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
-  %x.06 = phi float [ %conv1, %for.body ], [ %init, %for.body.preheader ]
-  %arrayidx = getelementptr inbounds float, float* %A, i64 %indvars.iv
-  store float %x.06, float* %arrayidx, align 4
-  %conv1 = fadd fast float %x.06, 5.000000e-01
-  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
-  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
-  %exitcond = icmp eq i32 %lftr.wideiv, %N
-  br i1 %exitcond, label %for.end.loopexit, label %for.body
-
-for.end.loopexit:                                 ; preds = %for.body
-  br label %for.end
-
-for.end:                                          ; preds = %for.end.loopexit, %entry
-  ret void
-}
-
-;void fp_iv_loop3(float init, float * __restrict__ A, float * __restrict__ B, float * __restrict__ C, int N) {
-;  int i = 0;
-;  float x = init;
-;  float y = 0.1;
-;  for (; i < N; ++i) {
-;    A[i] = x;
-;    x += fp_inc;
-;    y -= 0.5;
-;    B[i] = x + y;
-;    C[i] = y;
-;  }
-;}
-
-; VEC4_INTERL1-LABEL: @fp_iv_loop3(
-; VEC4_INTERL1:       for.body.lr.ph:
-; VEC4_INTERL1:         [[TMP0:%.*]] = load float, float* @fp_inc, align 4
+; VEC4_INTERL1-LABEL: @fp_iv_loop2(
+; VEC4_INTERL1-NEXT:  entry:
+; VEC4_INTERL1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC4_INTERL1-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; VEC4_INTERL1:       for.body.preheader:
+; VEC4_INTERL1-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC4_INTERL1-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC4_INTERL1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
+; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
-; VEC4_INTERL1:         [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float %init, i32 0
+; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
+; VEC4_INTERL1-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
+; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
+; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i32 0
 ; VEC4_INTERL1-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
-; VEC4_INTERL1-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT5]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL1-NEXT:    [[TMP7:%.*]] = fmul fast <4 x float> [[DOTSPLAT6]], <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
-; VEC4_INTERL1-NEXT:    [[INDUCTION7:%.*]] = fadd fast <4 x float> [[DOTSPLAT]], [[TMP7]]
-; VEC4_INTERL1-NEXT:    [[TMP8:%.*]] = fmul fast float [[TMP0]], 4.000000e+00
-; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT8:%.*]] = insertelement <4 x float> poison, float [[TMP8]], i32 0
-; VEC4_INTERL1-NEXT:    [[DOTSPLAT9:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT8]], <4 x float> poison, <4 x i32> zeroinitializer
-; VEC4_INTERL1-NEXT:    [[BROADCAST_SPLATINSERT12:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
-; VEC4_INTERL1-NEXT:    [[BROADCAST_SPLAT13:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT12]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL1-NEXT:    [[INDUCTION:%.*]] = fadd fast <4 x float> [[DOTSPLAT]], <float 0.000000e+00, float 5.000000e-01, float 1.000000e+00, float 1.500000e+00>
 ; VEC4_INTERL1-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC4_INTERL1:       vector.body:
-; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ <float 0x3FB99999A0000000, float 0xBFD99999A0000000, float 0xBFECCCCCC0000000, float 0xBFF6666660000000>, %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[VEC_IND10:%.*]] = phi <4 x float> [ [[INDUCTION7]], %vector.ph ], [ [[VEC_IND_NEXT11:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDEX]]
-; VEC4_INTERL1-NEXT:    [[TMP13:%.*]] = bitcast float* [[TMP12]] to <4 x float>*
-; VEC4_INTERL1-NEXT:    store <4 x float> [[VEC_IND10]], <4 x float>* [[TMP13]], align 4
-; VEC4_INTERL1-NEXT:    [[TMP14:%.*]] = fadd fast <4 x float> [[VEC_IND10]], [[BROADCAST_SPLAT13]]
-; VEC4_INTERL1-NEXT:    [[TMP15:%.*]] = fadd fast <4 x float> [[VEC_IND]], <float -5.000000e-01, float -5.000000e-01, float -5.000000e-01, float -5.000000e-01>
-; VEC4_INTERL1-NEXT:    [[TMP16:%.*]] = fadd fast <4 x float> [[TMP15]], [[TMP14]]
-; VEC4_INTERL1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, float* %B, i64 [[INDEX]]
-; VEC4_INTERL1-NEXT:    [[TMP18:%.*]] = bitcast float* [[TMP17]] to <4 x float>*
-; VEC4_INTERL1-NEXT:    store <4 x float> [[TMP16]], <4 x float>* [[TMP18]], align 4
-; VEC4_INTERL1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, float* %C, i64 [[INDEX]]
-; VEC4_INTERL1-NEXT:    [[TMP20:%.*]] = bitcast float* [[TMP19]] to <4 x float>*
-; VEC4_INTERL1-NEXT:    store <4 x float> [[TMP15]], <4 x float>* [[TMP20]], align 4
+; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL1-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP4]] to <4 x float>*
+; VEC4_INTERL1-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP5]], align 4
 ; VEC4_INTERL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT]] = fadd fast <4 x float> [[VEC_IND]], <float -2.000000e+00, float -2.000000e+00, float -2.000000e+00, float -2.000000e+00>
-; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT11]] = fadd fast <4 x float> [[VEC_IND10]], [[DOTSPLAT9]]
-; VEC4_INTERL1:         br i1 {{.*}}, label %middle.block, label %vector.body
-
-define void @fp_iv_loop3(float %init, float* noalias nocapture %A, float* noalias nocapture %B, float* noalias nocapture %C, i32 %N) {
-entry:
-  %cmp9 = icmp sgt i32 %N, 0
-  br i1 %cmp9, label %for.body.lr.ph, label %for.end
-
-for.body.lr.ph:                                   ; preds = %entry
-  %0 = load float, float* @fp_inc, align 4
-  br label %for.body
-
-for.body:                                         ; preds = %for.body, %for.body.lr.ph
+; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT]] = fadd fast <4 x float> [[VEC_IND]], <float 2.000000e+00, float 2.000000e+00, float 2.000000e+00, float 2.000000e+00>
+; VEC4_INTERL1-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VEC4_INTERL1:       middle.block:
+; VEC4_INTERL1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL1:       scalar.ph:
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_PREHEADER]] ]
+; VEC4_INTERL1-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL1:       for.body:
+; VEC4_INTERL1-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[X_06:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC4_INTERL1-NEXT:    store float [[X_06]], float* [[ARRAYIDX]], align 4
+; VEC4_INTERL1-NEXT:    [[CONV1]] = fadd fast float [[X_06]], 5.000000e-01
+; VEC4_INTERL1-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC4_INTERL1-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC4_INTERL1-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC4_INTERL1-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; VEC4_INTERL1:       for.end.loopexit:
+; VEC4_INTERL1-NEXT:    br label [[FOR_END]]
+; VEC4_INTERL1:       for.end:
+; VEC4_INTERL1-NEXT:    ret void
+;
+; VEC4_INTERL2-LABEL: @fp_iv_loop2(
+; VEC4_INTERL2-NEXT:  entry:
+; VEC4_INTERL2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC4_INTERL2-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; VEC4_INTERL2:       for.body.preheader:
+; VEC4_INTERL2-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC4_INTERL2-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC4_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
+; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC4_INTERL2:       vector.ph:
+; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
+; VEC4_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
+; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
+; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i32 0
+; VEC4_INTERL2-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL2-NEXT:    [[INDUCTION:%.*]] = fadd fast <4 x float> [[DOTSPLAT]], <float 0.000000e+00, float 5.000000e-01, float 1.000000e+00, float 1.500000e+00>
+; VEC4_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC4_INTERL2:       vector.body:
+; VEC4_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[STEP_ADD:%.*]] = fadd fast <4 x float> [[VEC_IND]], <float 2.000000e+00, float 2.000000e+00, float 2.000000e+00, float 2.000000e+00>
+; VEC4_INTERL2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL2-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP4]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP5]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP4]], i64 4
+; VEC4_INTERL2-NEXT:    [[TMP7:%.*]] = bitcast float* [[TMP6]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[STEP_ADD]], <4 x float>* [[TMP7]], align 4
+; VEC4_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; VEC4_INTERL2-NEXT:    [[VEC_IND_NEXT]] = fadd fast <4 x float> [[VEC_IND]], <float 4.000000e+00, float 4.000000e+00, float 4.000000e+00, float 4.000000e+00>
+; VEC4_INTERL2-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VEC4_INTERL2:       middle.block:
+; VEC4_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL2:       scalar.ph:
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_PREHEADER]] ]
+; VEC4_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL2:       for.body:
+; VEC4_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[X_06:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC4_INTERL2-NEXT:    store float [[X_06]], float* [[ARRAYIDX]], align 4
+; VEC4_INTERL2-NEXT:    [[CONV1]] = fadd fast float [[X_06]], 5.000000e-01
+; VEC4_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC4_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC4_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC4_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; VEC4_INTERL2:       for.end.loopexit:
+; VEC4_INTERL2-NEXT:    br label [[FOR_END]]
+; VEC4_INTERL2:       for.end:
+; VEC4_INTERL2-NEXT:    ret void
+;
+; VEC1_INTERL2-LABEL: @fp_iv_loop2(
+; VEC1_INTERL2-NEXT:  entry:
+; VEC1_INTERL2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC1_INTERL2-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; VEC1_INTERL2:       for.body.preheader:
+; VEC1_INTERL2-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC1_INTERL2-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC1_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC1_INTERL2:       vector.ph:
+; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
+; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
+; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC1_INTERL2:       vector.body:
+; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
+; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP4]], 5.000000e-01
+; VEC1_INTERL2-NEXT:    [[OFFSET_IDX:%.*]] = fadd fast float [[TMP5]], [[INIT]]
+; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fadd fast float [[OFFSET_IDX]], 5.000000e-01
+; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
+; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX]], float* [[TMP7]], align 4
+; VEC1_INTERL2-NEXT:    store float [[TMP6]], float* [[TMP8]], align 4
+; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; VEC1_INTERL2:       middle.block:
+; VEC1_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC1_INTERL2:       scalar.ph:
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_PREHEADER]] ]
+; VEC1_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC1_INTERL2:       for.body:
+; VEC1_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[X_06:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC1_INTERL2-NEXT:    store float [[X_06]], float* [[ARRAYIDX]], align 4
+; VEC1_INTERL2-NEXT:    [[CONV1]] = fadd fast float [[X_06]], 5.000000e-01
+; VEC1_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC1_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC1_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC1_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VEC1_INTERL2:       for.end.loopexit:
+; VEC1_INTERL2-NEXT:    br label [[FOR_END]]
+; VEC1_INTERL2:       for.end:
+; VEC1_INTERL2-NEXT:    ret void
+;
+; VEC2_INTERL1_PRED_STORE-LABEL: @fp_iv_loop2(
+; VEC2_INTERL1_PRED_STORE-NEXT:  entry:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; VEC2_INTERL1_PRED_STORE:       for.body.preheader:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.ph:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDUCTION:%.*]] = fadd fast <2 x float> [[DOTSPLAT]], <float 0.000000e+00, float 5.000000e-01>
+; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND:%.*]] = phi <2 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP4]] to <2 x float>*
+; VEC2_INTERL1_PRED_STORE-NEXT:    store <2 x float> [[VEC_IND]], <2 x float>* [[TMP5]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND_NEXT]] = fadd fast <2 x float> [[VEC_IND]], <float 1.000000e+00, float 1.000000e+00>
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       middle.block:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP_N]], label [[FOR_END]], label [[FOR_BODY]]
+; VEC2_INTERL1_PRED_STORE:       for.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[X_06:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_PREHEADER]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[X_06]], float* [[ARRAYIDX]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CONV1]] = fadd fast float [[X_06]], 5.000000e-01
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       for.end:
+; VEC2_INTERL1_PRED_STORE-NEXT:    ret void
+;
+entry:
+  %cmp4 = icmp sgt i32 %N, 0
+  br i1 %cmp4, label %for.body.preheader, label %for.end
+
+for.body.preheader:                               ; preds = %entry
+  br label %for.body
+
+for.body:                                         ; preds = %for.body.preheader, %for.body
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
+  %x.06 = phi float [ %conv1, %for.body ], [ %init, %for.body.preheader ]
+  %arrayidx = getelementptr inbounds float, float* %A, i64 %indvars.iv
+  store float %x.06, float* %arrayidx, align 4
+  %conv1 = fadd fast float %x.06, 5.000000e-01
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, %N
+  br i1 %exitcond, label %for.end.loopexit, label %for.body
+
+for.end.loopexit:                                 ; preds = %for.body
+  br label %for.end
+
+for.end:                                          ; preds = %for.end.loopexit, %entry
+  ret void
+}
+
+;void fp_iv_loop3(float init, float * __restrict__ A, float * __restrict__ B, float * __restrict__ C, int N) {
+;  int i = 0;
+;  float x = init;
+;  float y = 0.1;
+;  for (; i < N; ++i) {
+;    A[i] = x;
+;    x += fp_inc;
+;    y -= 0.5;
+;    B[i] = x + y;
+;    C[i] = y;
+;  }
+;}
+
+
+define void @fp_iv_loop3(float %init, float* noalias nocapture %A, float* noalias nocapture %B, float* noalias nocapture %C, i32 %N) {
+; VEC4_INTERL1-LABEL: @fp_iv_loop3(
+; VEC4_INTERL1-NEXT:  entry:
+; VEC4_INTERL1-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC4_INTERL1-NEXT:    br i1 [[CMP9]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; VEC4_INTERL1:       for.body.lr.ph:
+; VEC4_INTERL1-NEXT:    [[TMP0:%.*]] = load float, float* @fp_inc, align 4
+; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -1
+; VEC4_INTERL1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; VEC4_INTERL1-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; VEC4_INTERL1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP1]], 3
+; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC4_INTERL1:       vector.ph:
+; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP3]], 8589934588
+; VEC4_INTERL1-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[TMP4:%.*]] = fmul fast float [[CAST_CRD]], -5.000000e-01
+; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP4]], 0x3FB99999A0000000
+; VEC4_INTERL1-NEXT:    [[CAST_CRD3:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP0]], [[CAST_CRD3]]
+; VEC4_INTERL1-NEXT:    [[IND_END4:%.*]] = fadd fast float [[TMP5]], [[INIT:%.*]]
+; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i32 0
+; VEC4_INTERL1-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
+; VEC4_INTERL1-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT5]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL1-NEXT:    [[TMP6:%.*]] = fmul fast <4 x float> [[DOTSPLAT6]], <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
+; VEC4_INTERL1-NEXT:    [[INDUCTION:%.*]] = fadd fast <4 x float> [[DOTSPLAT]], [[TMP6]]
+; VEC4_INTERL1-NEXT:    [[TMP7:%.*]] = fmul fast float [[TMP0]], 4.000000e+00
+; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT7:%.*]] = insertelement <4 x float> poison, float [[TMP7]], i32 0
+; VEC4_INTERL1-NEXT:    [[DOTSPLAT8:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT7]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL1-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
+; VEC4_INTERL1-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL1-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC4_INTERL1:       vector.body:
+; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ <float 0x3FB99999A0000000, float 0xBFD99999A0000000, float 0xBFECCCCCC0000000, float 0xBFF6666660000000>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[VEC_IND9:%.*]] = phi <4 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT10:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL1-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP8]] to <4 x float>*
+; VEC4_INTERL1-NEXT:    store <4 x float> [[VEC_IND9]], <4 x float>* [[TMP9]], align 4
+; VEC4_INTERL1-NEXT:    [[TMP10:%.*]] = fadd fast <4 x float> [[VEC_IND9]], [[BROADCAST_SPLAT]]
+; VEC4_INTERL1-NEXT:    [[TMP11:%.*]] = fadd fast <4 x float> [[VEC_IND]], <float -5.000000e-01, float -5.000000e-01, float -5.000000e-01, float -5.000000e-01>
+; VEC4_INTERL1-NEXT:    [[TMP12:%.*]] = fadd fast <4 x float> [[TMP11]], [[TMP10]]
+; VEC4_INTERL1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[INDEX]]
+; VEC4_INTERL1-NEXT:    [[TMP14:%.*]] = bitcast float* [[TMP13]] to <4 x float>*
+; VEC4_INTERL1-NEXT:    store <4 x float> [[TMP12]], <4 x float>* [[TMP14]], align 4
+; VEC4_INTERL1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, float* [[C:%.*]], i64 [[INDEX]]
+; VEC4_INTERL1-NEXT:    [[TMP16:%.*]] = bitcast float* [[TMP15]] to <4 x float>*
+; VEC4_INTERL1-NEXT:    store <4 x float> [[TMP11]], <4 x float>* [[TMP16]], align 4
+; VEC4_INTERL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT]] = fadd fast <4 x float> [[VEC_IND]], <float -2.000000e+00, float -2.000000e+00, float -2.000000e+00, float -2.000000e+00>
+; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT10]] = fadd fast <4 x float> [[VEC_IND9]], [[DOTSPLAT8]]
+; VEC4_INTERL1-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; VEC4_INTERL1:       middle.block:
+; VEC4_INTERL1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL1:       scalar.ph:
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0x3FB99999A0000000, [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi float [ [[IND_END4]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL1-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL1:       for.body:
+; VEC4_INTERL1-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[Y_012:%.*]] = phi float [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[CONV1:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[X_011:%.*]] = phi float [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC4_INTERL1-NEXT:    store float [[X_011]], float* [[ARRAYIDX]], align 4
+; VEC4_INTERL1-NEXT:    [[ADD]] = fadd fast float [[X_011]], [[TMP0]]
+; VEC4_INTERL1-NEXT:    [[CONV1]] = fadd fast float [[Y_012]], -5.000000e-01
+; VEC4_INTERL1-NEXT:    [[ADD2:%.*]] = fadd fast float [[CONV1]], [[ADD]]
+; VEC4_INTERL1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDVARS_IV]]
+; VEC4_INTERL1-NEXT:    store float [[ADD2]], float* [[ARRAYIDX4]], align 4
+; VEC4_INTERL1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[INDVARS_IV]]
+; VEC4_INTERL1-NEXT:    store float [[CONV1]], float* [[ARRAYIDX6]], align 4
+; VEC4_INTERL1-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC4_INTERL1-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC4_INTERL1-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC4_INTERL1-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; VEC4_INTERL1:       for.end.loopexit:
+; VEC4_INTERL1-NEXT:    br label [[FOR_END]]
+; VEC4_INTERL1:       for.end:
+; VEC4_INTERL1-NEXT:    ret void
+;
+; VEC4_INTERL2-LABEL: @fp_iv_loop3(
+; VEC4_INTERL2-NEXT:  entry:
+; VEC4_INTERL2-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC4_INTERL2-NEXT:    br i1 [[CMP9]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; VEC4_INTERL2:       for.body.lr.ph:
+; VEC4_INTERL2-NEXT:    [[TMP0:%.*]] = load float, float* @fp_inc, align 4
+; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -1
+; VEC4_INTERL2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; VEC4_INTERL2-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; VEC4_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP1]], 7
+; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC4_INTERL2:       vector.ph:
+; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP3]], 8589934584
+; VEC4_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[TMP4:%.*]] = fmul fast float [[CAST_CRD]], -5.000000e-01
+; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP4]], 0x3FB99999A0000000
+; VEC4_INTERL2-NEXT:    [[CAST_CRD3:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP0]], [[CAST_CRD3]]
+; VEC4_INTERL2-NEXT:    [[IND_END4:%.*]] = fadd fast float [[TMP5]], [[INIT:%.*]]
+; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i32 0
+; VEC4_INTERL2-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT6:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
+; VEC4_INTERL2-NEXT:    [[DOTSPLAT7:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT6]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL2-NEXT:    [[TMP6:%.*]] = fmul fast <4 x float> [[DOTSPLAT7]], <float 0.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
+; VEC4_INTERL2-NEXT:    [[INDUCTION:%.*]] = fadd fast <4 x float> [[DOTSPLAT]], [[TMP6]]
+; VEC4_INTERL2-NEXT:    [[TMP7:%.*]] = fmul fast float [[TMP0]], 4.000000e+00
+; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT8:%.*]] = insertelement <4 x float> poison, float [[TMP7]], i32 0
+; VEC4_INTERL2-NEXT:    [[DOTSPLAT9:%.*]] = shufflevector <4 x float> [[DOTSPLATINSERT8]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL2-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
+; VEC4_INTERL2-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL2-NEXT:    [[BROADCAST_SPLATINSERT14:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
+; VEC4_INTERL2-NEXT:    [[BROADCAST_SPLAT15:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT14]], <4 x float> poison, <4 x i32> zeroinitializer
+; VEC4_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC4_INTERL2:       vector.body:
+; VEC4_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ <float 0x3FB99999A0000000, float 0xBFD99999A0000000, float 0xBFECCCCCC0000000, float 0xBFF6666660000000>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[VEC_IND10:%.*]] = phi <4 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT13:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[STEP_ADD11:%.*]] = fadd fast <4 x float> [[VEC_IND10]], [[DOTSPLAT9]]
+; VEC4_INTERL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL2-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP8]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[VEC_IND10]], <4 x float>* [[TMP9]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 4
+; VEC4_INTERL2-NEXT:    [[TMP11:%.*]] = bitcast float* [[TMP10]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[STEP_ADD11]], <4 x float>* [[TMP11]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP12:%.*]] = fadd fast <4 x float> [[VEC_IND10]], [[BROADCAST_SPLAT]]
+; VEC4_INTERL2-NEXT:    [[TMP13:%.*]] = fadd fast <4 x float> [[STEP_ADD11]], [[BROADCAST_SPLAT15]]
+; VEC4_INTERL2-NEXT:    [[TMP14:%.*]] = fadd fast <4 x float> [[VEC_IND]], <float -5.000000e-01, float -5.000000e-01, float -5.000000e-01, float -5.000000e-01>
+; VEC4_INTERL2-NEXT:    [[TMP15:%.*]] = fadd fast <4 x float> [[VEC_IND]], <float -2.500000e+00, float -2.500000e+00, float -2.500000e+00, float -2.500000e+00>
+; VEC4_INTERL2-NEXT:    [[TMP16:%.*]] = fadd fast <4 x float> [[TMP14]], [[TMP12]]
+; VEC4_INTERL2-NEXT:    [[TMP17:%.*]] = fadd fast <4 x float> [[TMP15]], [[TMP13]]
+; VEC4_INTERL2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[INDEX]]
+; VEC4_INTERL2-NEXT:    [[TMP19:%.*]] = bitcast float* [[TMP18]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[TMP16]], <4 x float>* [[TMP19]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[TMP18]], i64 4
+; VEC4_INTERL2-NEXT:    [[TMP21:%.*]] = bitcast float* [[TMP20]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[TMP17]], <4 x float>* [[TMP21]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[C:%.*]], i64 [[INDEX]]
+; VEC4_INTERL2-NEXT:    [[TMP23:%.*]] = bitcast float* [[TMP22]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[TMP14]], <4 x float>* [[TMP23]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 4
+; VEC4_INTERL2-NEXT:    [[TMP25:%.*]] = bitcast float* [[TMP24]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[TMP15]], <4 x float>* [[TMP25]], align 4
+; VEC4_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; VEC4_INTERL2-NEXT:    [[VEC_IND_NEXT]] = fadd fast <4 x float> [[VEC_IND]], <float -4.000000e+00, float -4.000000e+00, float -4.000000e+00, float -4.000000e+00>
+; VEC4_INTERL2-NEXT:    [[VEC_IND_NEXT13]] = fadd fast <4 x float> [[STEP_ADD11]], [[DOTSPLAT9]]
+; VEC4_INTERL2-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; VEC4_INTERL2:       middle.block:
+; VEC4_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL2:       scalar.ph:
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0x3FB99999A0000000, [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi float [ [[IND_END4]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_LR_PH]] ]
+; VEC4_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL2:       for.body:
+; VEC4_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[Y_012:%.*]] = phi float [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[CONV1:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[X_011:%.*]] = phi float [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC4_INTERL2-NEXT:    store float [[X_011]], float* [[ARRAYIDX]], align 4
+; VEC4_INTERL2-NEXT:    [[ADD]] = fadd fast float [[X_011]], [[TMP0]]
+; VEC4_INTERL2-NEXT:    [[CONV1]] = fadd fast float [[Y_012]], -5.000000e-01
+; VEC4_INTERL2-NEXT:    [[ADD2:%.*]] = fadd fast float [[CONV1]], [[ADD]]
+; VEC4_INTERL2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDVARS_IV]]
+; VEC4_INTERL2-NEXT:    store float [[ADD2]], float* [[ARRAYIDX4]], align 4
+; VEC4_INTERL2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[INDVARS_IV]]
+; VEC4_INTERL2-NEXT:    store float [[CONV1]], float* [[ARRAYIDX6]], align 4
+; VEC4_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC4_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC4_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC4_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; VEC4_INTERL2:       for.end.loopexit:
+; VEC4_INTERL2-NEXT:    br label [[FOR_END]]
+; VEC4_INTERL2:       for.end:
+; VEC4_INTERL2-NEXT:    ret void
+;
+; VEC1_INTERL2-LABEL: @fp_iv_loop3(
+; VEC1_INTERL2-NEXT:  entry:
+; VEC1_INTERL2-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC1_INTERL2-NEXT:    br i1 [[CMP9]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; VEC1_INTERL2:       for.body.lr.ph:
+; VEC1_INTERL2-NEXT:    [[TMP0:%.*]] = load float, float* @fp_inc, align 4
+; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -1
+; VEC1_INTERL2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; VEC1_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP1]], 0
+; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC1_INTERL2:       vector.ph:
+; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP3]], 8589934590
+; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = fmul fast float [[CAST_CRD]], -5.000000e-01
+; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP4]], 0x3FB99999A0000000
+; VEC1_INTERL2-NEXT:    [[CAST_CRD3:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP0]], [[CAST_CRD3]]
+; VEC1_INTERL2-NEXT:    [[IND_END4:%.*]] = fadd fast float [[TMP5]], [[INIT:%.*]]
+; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC1_INTERL2:       vector.body:
+; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC1_INTERL2-NEXT:    [[INDUCTION5:%.*]] = or i64 [[INDEX]], 1
+; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = fmul fast float [[TMP6]], -5.000000e-01
+; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = fmul fast float [[TMP0]], [[TMP8]]
+; VEC1_INTERL2-NEXT:    [[OFFSET_IDX8:%.*]] = fadd fast float [[TMP9]], [[INIT]]
+; VEC1_INTERL2-NEXT:    [[TMP10:%.*]] = fadd fast float [[OFFSET_IDX8]], [[TMP0]]
+; VEC1_INTERL2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC1_INTERL2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION5]]
+; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX8]], float* [[TMP11]], align 4
+; VEC1_INTERL2-NEXT:    store float [[TMP10]], float* [[TMP12]], align 4
+; VEC1_INTERL2-NEXT:    [[TMP13:%.*]] = fadd fast float [[OFFSET_IDX8]], [[TMP0]]
+; VEC1_INTERL2-NEXT:    [[TMP14:%.*]] = fadd fast float [[TMP10]], [[TMP0]]
+; VEC1_INTERL2-NEXT:    [[TMP15:%.*]] = fadd fast float [[TMP7]], 0xBFD99999A0000000
+; VEC1_INTERL2-NEXT:    [[TMP16:%.*]] = fadd fast float [[TMP7]], 0xBFECCCCCC0000000
+; VEC1_INTERL2-NEXT:    [[TMP17:%.*]] = fadd fast float [[TMP15]], [[TMP13]]
+; VEC1_INTERL2-NEXT:    [[TMP18:%.*]] = fadd fast float [[TMP16]], [[TMP14]]
+; VEC1_INTERL2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[INDEX]]
+; VEC1_INTERL2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDUCTION5]]
+; VEC1_INTERL2-NEXT:    store float [[TMP17]], float* [[TMP19]], align 4
+; VEC1_INTERL2-NEXT:    store float [[TMP18]], float* [[TMP20]], align 4
+; VEC1_INTERL2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, float* [[C:%.*]], i64 [[INDEX]]
+; VEC1_INTERL2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[INDUCTION5]]
+; VEC1_INTERL2-NEXT:    store float [[TMP15]], float* [[TMP21]], align 4
+; VEC1_INTERL2-NEXT:    store float [[TMP16]], float* [[TMP22]], align 4
+; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VEC1_INTERL2-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; VEC1_INTERL2:       middle.block:
+; VEC1_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC1_INTERL2:       scalar.ph:
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0x3FB99999A0000000, [[FOR_BODY_LR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi float [ [[IND_END4]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_LR_PH]] ]
+; VEC1_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC1_INTERL2:       for.body:
+; VEC1_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; VEC1_INTERL2-NEXT:    [[Y_012:%.*]] = phi float [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[CONV1:%.*]], [[FOR_BODY]] ]
+; VEC1_INTERL2-NEXT:    [[X_011:%.*]] = phi float [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
+; VEC1_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC1_INTERL2-NEXT:    store float [[X_011]], float* [[ARRAYIDX]], align 4
+; VEC1_INTERL2-NEXT:    [[ADD]] = fadd fast float [[X_011]], [[TMP0]]
+; VEC1_INTERL2-NEXT:    [[CONV1]] = fadd fast float [[Y_012]], -5.000000e-01
+; VEC1_INTERL2-NEXT:    [[ADD2:%.*]] = fadd fast float [[CONV1]], [[ADD]]
+; VEC1_INTERL2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDVARS_IV]]
+; VEC1_INTERL2-NEXT:    store float [[ADD2]], float* [[ARRAYIDX4]], align 4
+; VEC1_INTERL2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[INDVARS_IV]]
+; VEC1_INTERL2-NEXT:    store float [[CONV1]], float* [[ARRAYIDX6]], align 4
+; VEC1_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC1_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC1_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC1_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; VEC1_INTERL2:       for.end.loopexit:
+; VEC1_INTERL2-NEXT:    br label [[FOR_END]]
+; VEC1_INTERL2:       for.end:
+; VEC1_INTERL2-NEXT:    ret void
+;
+; VEC2_INTERL1_PRED_STORE-LABEL: @fp_iv_loop3(
+; VEC2_INTERL1_PRED_STORE-NEXT:  entry:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP9]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; VEC2_INTERL1_PRED_STORE:       for.body.lr.ph:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP0:%.*]] = load float, float* @fp_inc, align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP1]], 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.ph:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP3]], 8589934590
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP4:%.*]] = fmul fast float [[CAST_CRD]], -5.000000e-01
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP4]], 0x3FB99999A0000000
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CAST_CRD3:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP0]], [[CAST_CRD3]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END4:%.*]] = fadd fast float [[TMP5]], [[INIT:%.*]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT5]], <2 x float> poison, <2 x i32> zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP6:%.*]] = fmul fast <2 x float> [[DOTSPLAT6]], <float 0.000000e+00, float 1.000000e+00>
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDUCTION:%.*]] = fadd fast <2 x float> [[DOTSPLAT]], [[TMP6]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP7:%.*]] = fmul fast float [[TMP0]], 2.000000e+00
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT7:%.*]] = insertelement <2 x float> poison, float [[TMP7]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLAT8:%.*]] = shufflevector <2 x float> [[DOTSPLATINSERT7]], <2 x float> poison, <2 x i32> zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND:%.*]] = phi <2 x float> [ <float 0x3FB99999A0000000, float 0xBFD99999A0000000>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND9:%.*]] = phi <2 x float> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT10:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP8]] to <2 x float>*
+; VEC2_INTERL1_PRED_STORE-NEXT:    store <2 x float> [[VEC_IND9]], <2 x float>* [[TMP9]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP10:%.*]] = fadd fast <2 x float> [[VEC_IND9]], [[BROADCAST_SPLAT]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP11:%.*]] = fadd fast <2 x float> [[VEC_IND]], <float -5.000000e-01, float -5.000000e-01>
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP12:%.*]] = fadd fast <2 x float> [[TMP11]], [[TMP10]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[INDEX]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP14:%.*]] = bitcast float* [[TMP13]] to <2 x float>*
+; VEC2_INTERL1_PRED_STORE-NEXT:    store <2 x float> [[TMP12]], <2 x float>* [[TMP14]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, float* [[C:%.*]], i64 [[INDEX]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP16:%.*]] = bitcast float* [[TMP15]] to <2 x float>*
+; VEC2_INTERL1_PRED_STORE-NEXT:    store <2 x float> [[TMP11]], <2 x float>* [[TMP16]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND_NEXT]] = fadd fast <2 x float> [[VEC_IND]], <float -1.000000e+00, float -1.000000e+00>
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND_NEXT10]] = fadd fast <2 x float> [[VEC_IND9]], [[DOTSPLAT8]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       middle.block:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP_N]], label [[FOR_END]], label [[FOR_BODY]]
+; VEC2_INTERL1_PRED_STORE:       for.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[Y_012:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0x3FB99999A0000000, [[FOR_BODY_LR_PH]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[X_011:%.*]] = phi float [ [[ADD:%.*]], [[FOR_BODY]] ], [ [[IND_END4]], [[MIDDLE_BLOCK]] ], [ [[INIT]], [[FOR_BODY_LR_PH]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[X_011]], float* [[ARRAYIDX]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ADD]] = fadd fast float [[X_011]], [[TMP0]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CONV1]] = fadd fast float [[Y_012]], -5.000000e-01
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ADD2:%.*]] = fadd fast float [[CONV1]], [[ADD]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[INDVARS_IV]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[ADD2]], float* [[ARRAYIDX4]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[C]], i64 [[INDVARS_IV]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[CONV1]], float* [[ARRAYIDX6]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       for.end:
+; VEC2_INTERL1_PRED_STORE-NEXT:    ret void
+;
+entry:
+  %cmp9 = icmp sgt i32 %N, 0
+  br i1 %cmp9, label %for.body.lr.ph, label %for.end
+
+for.body.lr.ph:                                   ; preds = %entry
+  %0 = load float, float* @fp_inc, align 4
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %for.body.lr.ph
   %indvars.iv = phi i64 [ 0, %for.body.lr.ph ], [ %indvars.iv.next, %for.body ]
   %y.012 = phi float [ 0x3FB99999A0000000, %for.body.lr.ph ], [ %conv1, %for.body ]
   %x.011 = phi float [ %init, %for.body.lr.ph ], [ %add, %for.body ]
@@ -505,20 +1159,202 @@ for.end:
 ;  }
 ;}
 
+
+define void @fp_iv_loop4(float* noalias nocapture %A, i32 %N) {
 ; VEC4_INTERL1-LABEL: @fp_iv_loop4(
+; VEC4_INTERL1-NEXT:  entry:
+; VEC4_INTERL1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC4_INTERL1-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; VEC4_INTERL1:       for.body.preheader:
+; VEC4_INTERL1-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC4_INTERL1-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC4_INTERL1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
+; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
-; VEC4_INTERL1:         br label %vector.body
+; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
+; VEC4_INTERL1-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
+; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], 1.000000e+00
+; VEC4_INTERL1-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC4_INTERL1:       vector.body:
-; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ <float 1.000000e+00, float 1.500000e+00, float 2.000000e+00, float 2.500000e+00>, %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
-; VEC4_INTERL1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDEX]]
-; VEC4_INTERL1-NEXT:    [[TMP8:%.*]] = bitcast float* [[TMP7]] to <4 x float>*
-; VEC4_INTERL1-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP8]], align 4
+; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ <float 1.000000e+00, float 1.500000e+00, float 2.000000e+00, float 2.500000e+00>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL1-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP4]] to <4 x float>*
+; VEC4_INTERL1-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP5]], align 4
 ; VEC4_INTERL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VEC4_INTERL1-NEXT:    [[VEC_IND_NEXT]] = fadd fast <4 x float> [[VEC_IND]], <float 2.000000e+00, float 2.000000e+00, float 2.000000e+00, float 2.000000e+00>
-; VEC4_INTERL1:         br i1 {{.*}}, label %middle.block, label %vector.body
-
-define void @fp_iv_loop4(float* noalias nocapture %A, i32 %N) {
+; VEC4_INTERL1-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; VEC4_INTERL1:       middle.block:
+; VEC4_INTERL1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL1:       scalar.ph:
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[FOR_BODY_PREHEADER]] ]
+; VEC4_INTERL1-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL1:       for.body:
+; VEC4_INTERL1-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[X_06:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC4_INTERL1-NEXT:    store float [[X_06]], float* [[ARRAYIDX]], align 4
+; VEC4_INTERL1-NEXT:    [[CONV1]] = fadd fast float [[X_06]], 5.000000e-01
+; VEC4_INTERL1-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC4_INTERL1-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC4_INTERL1-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC4_INTERL1-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; VEC4_INTERL1:       for.end.loopexit:
+; VEC4_INTERL1-NEXT:    br label [[FOR_END]]
+; VEC4_INTERL1:       for.end:
+; VEC4_INTERL1-NEXT:    ret void
+;
+; VEC4_INTERL2-LABEL: @fp_iv_loop4(
+; VEC4_INTERL2-NEXT:  entry:
+; VEC4_INTERL2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC4_INTERL2-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; VEC4_INTERL2:       for.body.preheader:
+; VEC4_INTERL2-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC4_INTERL2-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC4_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
+; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC4_INTERL2:       vector.ph:
+; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
+; VEC4_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
+; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], 1.000000e+00
+; VEC4_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC4_INTERL2:       vector.body:
+; VEC4_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[VEC_IND:%.*]] = phi <4 x float> [ <float 1.000000e+00, float 1.500000e+00, float 2.000000e+00, float 2.500000e+00>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC4_INTERL2-NEXT:    [[STEP_ADD:%.*]] = fadd fast <4 x float> [[VEC_IND]], <float 2.000000e+00, float 2.000000e+00, float 2.000000e+00, float 2.000000e+00>
+; VEC4_INTERL2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL2-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP4]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[VEC_IND]], <4 x float>* [[TMP5]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP4]], i64 4
+; VEC4_INTERL2-NEXT:    [[TMP7:%.*]] = bitcast float* [[TMP6]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    store <4 x float> [[STEP_ADD]], <4 x float>* [[TMP7]], align 4
+; VEC4_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; VEC4_INTERL2-NEXT:    [[VEC_IND_NEXT]] = fadd fast <4 x float> [[VEC_IND]], <float 4.000000e+00, float 4.000000e+00, float 4.000000e+00, float 4.000000e+00>
+; VEC4_INTERL2-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; VEC4_INTERL2:       middle.block:
+; VEC4_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL2:       scalar.ph:
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[FOR_BODY_PREHEADER]] ]
+; VEC4_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL2:       for.body:
+; VEC4_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[X_06:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC4_INTERL2-NEXT:    store float [[X_06]], float* [[ARRAYIDX]], align 4
+; VEC4_INTERL2-NEXT:    [[CONV1]] = fadd fast float [[X_06]], 5.000000e-01
+; VEC4_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC4_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC4_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC4_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; VEC4_INTERL2:       for.end.loopexit:
+; VEC4_INTERL2-NEXT:    br label [[FOR_END]]
+; VEC4_INTERL2:       for.end:
+; VEC4_INTERL2-NEXT:    ret void
+;
+; VEC1_INTERL2-LABEL: @fp_iv_loop4(
+; VEC1_INTERL2-NEXT:  entry:
+; VEC1_INTERL2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC1_INTERL2-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; VEC1_INTERL2:       for.body.preheader:
+; VEC1_INTERL2-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC1_INTERL2-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC1_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC1_INTERL2:       vector.ph:
+; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
+; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], 1.000000e+00
+; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC1_INTERL2:       vector.body:
+; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
+; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fmul fast float [[TMP4]], 5.000000e-01
+; VEC1_INTERL2-NEXT:    [[OFFSET_IDX:%.*]] = fadd fast float [[TMP5]], 1.000000e+00
+; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fadd fast float [[TMP5]], 1.500000e+00
+; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
+; VEC1_INTERL2-NEXT:    store float [[OFFSET_IDX]], float* [[TMP7]], align 4
+; VEC1_INTERL2-NEXT:    store float [[TMP6]], float* [[TMP8]], align 4
+; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VEC1_INTERL2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; VEC1_INTERL2:       middle.block:
+; VEC1_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; VEC1_INTERL2:       scalar.ph:
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[FOR_BODY_PREHEADER]] ]
+; VEC1_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC1_INTERL2:       for.body:
+; VEC1_INTERL2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[X_06:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC1_INTERL2-NEXT:    store float [[X_06]], float* [[ARRAYIDX]], align 4
+; VEC1_INTERL2-NEXT:    [[CONV1]] = fadd fast float [[X_06]], 5.000000e-01
+; VEC1_INTERL2-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC1_INTERL2-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC1_INTERL2-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC1_INTERL2-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; VEC1_INTERL2:       for.end.loopexit:
+; VEC1_INTERL2-NEXT:    br label [[FOR_END]]
+; VEC1_INTERL2:       for.end:
+; VEC1_INTERL2-NEXT:    ret void
+;
+; VEC2_INTERL1_PRED_STORE-LABEL: @fp_iv_loop4(
+; VEC2_INTERL1_PRED_STORE-NEXT:  entry:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[N:%.*]], 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP4]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
+; VEC2_INTERL1_PRED_STORE:       for.body.preheader:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.ph:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP3:%.*]] = fmul fast float [[CAST_CRD]], 5.000000e-01
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP3]], 1.000000e+00
+; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND:%.*]] = phi <2 x float> [ <float 1.000000e+00, float 1.500000e+00>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP4]] to <2 x float>*
+; VEC2_INTERL1_PRED_STORE-NEXT:    store <2 x float> [[VEC_IND]], <2 x float>* [[TMP5]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VEC_IND_NEXT]] = fadd fast <2 x float> [[VEC_IND]], <float 1.000000e+00, float 1.000000e+00>
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       middle.block:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP_N]], label [[FOR_END]], label [[FOR_BODY]]
+; VEC2_INTERL1_PRED_STORE:       for.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[X_06:%.*]] = phi float [ [[CONV1:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[FOR_BODY_PREHEADER]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[X_06]], float* [[ARRAYIDX]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CONV1]] = fadd fast float [[X_06]], 5.000000e-01
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       for.end:
+; VEC2_INTERL1_PRED_STORE-NEXT:    ret void
+;
 entry:
   %cmp4 = icmp sgt i32 %N, 0
   br i1 %cmp4, label %for.body.preheader, label %for.end
@@ -544,47 +1380,331 @@ for.end:                                          ; preds = %for.end.loopexit, %
   ret void
 }
 
+
+define void @non_primary_iv_float_scalar(float* %A, i64 %N) {
+; VEC4_INTERL1-LABEL: @non_primary_iv_float_scalar(
+; VEC4_INTERL1-NEXT:  entry:
+; VEC4_INTERL1-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; VEC4_INTERL1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
+; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC4_INTERL1:       vector.ph:
+; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804
+; VEC4_INTERL1-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC4_INTERL1:       vector.body:
+; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE7:%.*]] ]
+; VEC4_INTERL1-NEXT:    [[TMP0:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL1-NEXT:    [[TMP2:%.*]] = bitcast float* [[TMP1]] to <4 x float>*
+; VEC4_INTERL1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP2]], align 4
+; VEC4_INTERL1-NEXT:    [[TMP3:%.*]] = fcmp fast oeq <4 x float> [[WIDE_LOAD]], zeroinitializer
+; VEC4_INTERL1-NEXT:    [[TMP4:%.*]] = extractelement <4 x i1> [[TMP3]], i32 0
+; VEC4_INTERL1-NEXT:    br i1 [[TMP4]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; VEC4_INTERL1:       pred.store.if:
+; VEC4_INTERL1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDEX]]
+; VEC4_INTERL1-NEXT:    store float [[TMP0]], float* [[TMP5]], align 4
+; VEC4_INTERL1-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; VEC4_INTERL1:       pred.store.continue:
+; VEC4_INTERL1-NEXT:    [[TMP6:%.*]] = extractelement <4 x i1> [[TMP3]], i32 1
+; VEC4_INTERL1-NEXT:    br i1 [[TMP6]], label [[PRED_STORE_IF2:%.*]], label [[PRED_STORE_CONTINUE3:%.*]]
+; VEC4_INTERL1:       pred.store.if2:
+; VEC4_INTERL1-NEXT:    [[TMP7:%.*]] = fadd fast float [[TMP0]], 1.000000e+00
+; VEC4_INTERL1-NEXT:    [[TMP8:%.*]] = or i64 [[INDEX]], 1
+; VEC4_INTERL1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP8]]
+; VEC4_INTERL1-NEXT:    store float [[TMP7]], float* [[TMP9]], align 4
+; VEC4_INTERL1-NEXT:    br label [[PRED_STORE_CONTINUE3]]
+; VEC4_INTERL1:       pred.store.continue3:
+; VEC4_INTERL1-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP3]], i32 2
+; VEC4_INTERL1-NEXT:    br i1 [[TMP10]], label [[PRED_STORE_IF4:%.*]], label [[PRED_STORE_CONTINUE5:%.*]]
+; VEC4_INTERL1:       pred.store.if4:
+; VEC4_INTERL1-NEXT:    [[TMP11:%.*]] = fadd fast float [[TMP0]], 2.000000e+00
+; VEC4_INTERL1-NEXT:    [[TMP12:%.*]] = or i64 [[INDEX]], 2
+; VEC4_INTERL1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]]
+; VEC4_INTERL1-NEXT:    store float [[TMP11]], float* [[TMP13]], align 4
+; VEC4_INTERL1-NEXT:    br label [[PRED_STORE_CONTINUE5]]
+; VEC4_INTERL1:       pred.store.continue5:
+; VEC4_INTERL1-NEXT:    [[TMP14:%.*]] = extractelement <4 x i1> [[TMP3]], i32 3
+; VEC4_INTERL1-NEXT:    br i1 [[TMP14]], label [[PRED_STORE_IF6:%.*]], label [[PRED_STORE_CONTINUE7]]
+; VEC4_INTERL1:       pred.store.if6:
+; VEC4_INTERL1-NEXT:    [[TMP15:%.*]] = fadd fast float [[TMP0]], 3.000000e+00
+; VEC4_INTERL1-NEXT:    [[TMP16:%.*]] = or i64 [[INDEX]], 3
+; VEC4_INTERL1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP16]]
+; VEC4_INTERL1-NEXT:    store float [[TMP15]], float* [[TMP17]], align 4
+; VEC4_INTERL1-NEXT:    br label [[PRED_STORE_CONTINUE7]]
+; VEC4_INTERL1:       pred.store.continue7:
+; VEC4_INTERL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; VEC4_INTERL1-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; VEC4_INTERL1:       middle.block:
+; VEC4_INTERL1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; VEC4_INTERL1-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL1:       scalar.ph:
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; VEC4_INTERL1-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[CAST_CRD]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; VEC4_INTERL1-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL1:       for.body:
+; VEC4_INTERL1-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[J:%.*]] = phi float [ [[J_NEXT:%.*]], [[FOR_INC]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC4_INTERL1-NEXT:    [[VAR0:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; VEC4_INTERL1-NEXT:    [[VAR1:%.*]] = load float, float* [[VAR0]], align 4
+; VEC4_INTERL1-NEXT:    [[VAR2:%.*]] = fcmp fast oeq float [[VAR1]], 0.000000e+00
+; VEC4_INTERL1-NEXT:    br i1 [[VAR2]], label [[IF_PRED:%.*]], label [[FOR_INC]]
+; VEC4_INTERL1:       if.pred:
+; VEC4_INTERL1-NEXT:    store float [[J]], float* [[VAR0]], align 4
+; VEC4_INTERL1-NEXT:    br label [[FOR_INC]]
+; VEC4_INTERL1:       for.inc:
+; VEC4_INTERL1-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; VEC4_INTERL1-NEXT:    [[J_NEXT]] = fadd fast float [[J]], 1.000000e+00
+; VEC4_INTERL1-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; VEC4_INTERL1-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP13:![0-9]+]]
+; VEC4_INTERL1:       for.end:
+; VEC4_INTERL1-NEXT:    ret void
+;
+; VEC4_INTERL2-LABEL: @non_primary_iv_float_scalar(
+; VEC4_INTERL2-NEXT:  entry:
+; VEC4_INTERL2-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; VEC4_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 8
+; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC4_INTERL2:       vector.ph:
+; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775800
+; VEC4_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC4_INTERL2:       vector.body:
+; VEC4_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE17:%.*]] ]
+; VEC4_INTERL2-NEXT:    [[TMP0:%.*]] = or i64 [[INDEX]], 4
+; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC4_INTERL2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC4_INTERL2-NEXT:    [[TMP3:%.*]] = bitcast float* [[TMP2]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP3]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 4
+; VEC4_INTERL2-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP4]] to <4 x float>*
+; VEC4_INTERL2-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x float>, <4 x float>* [[TMP5]], align 4
+; VEC4_INTERL2-NEXT:    [[TMP6:%.*]] = fcmp fast oeq <4 x float> [[WIDE_LOAD]], zeroinitializer
+; VEC4_INTERL2-NEXT:    [[TMP7:%.*]] = fcmp fast oeq <4 x float> [[WIDE_LOAD3]], zeroinitializer
+; VEC4_INTERL2-NEXT:    [[TMP8:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
+; VEC4_INTERL2-NEXT:    br i1 [[TMP8]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; VEC4_INTERL2:       pred.store.if:
+; VEC4_INTERL2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDEX]]
+; VEC4_INTERL2-NEXT:    store float [[TMP1]], float* [[TMP9]], align 4
+; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; VEC4_INTERL2:       pred.store.continue:
+; VEC4_INTERL2-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP6]], i32 1
+; VEC4_INTERL2-NEXT:    br i1 [[TMP10]], label [[PRED_STORE_IF4:%.*]], label [[PRED_STORE_CONTINUE5:%.*]]
+; VEC4_INTERL2:       pred.store.if4:
+; VEC4_INTERL2-NEXT:    [[TMP11:%.*]] = fadd fast float [[TMP1]], 1.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP12:%.*]] = or i64 [[INDEX]], 1
+; VEC4_INTERL2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]]
+; VEC4_INTERL2-NEXT:    store float [[TMP11]], float* [[TMP13]], align 4
+; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE5]]
+; VEC4_INTERL2:       pred.store.continue5:
+; VEC4_INTERL2-NEXT:    [[TMP14:%.*]] = extractelement <4 x i1> [[TMP6]], i32 2
+; VEC4_INTERL2-NEXT:    br i1 [[TMP14]], label [[PRED_STORE_IF6:%.*]], label [[PRED_STORE_CONTINUE7:%.*]]
+; VEC4_INTERL2:       pred.store.if6:
+; VEC4_INTERL2-NEXT:    [[TMP15:%.*]] = fadd fast float [[TMP1]], 2.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP16:%.*]] = or i64 [[INDEX]], 2
+; VEC4_INTERL2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP16]]
+; VEC4_INTERL2-NEXT:    store float [[TMP15]], float* [[TMP17]], align 4
+; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE7]]
+; VEC4_INTERL2:       pred.store.continue7:
+; VEC4_INTERL2-NEXT:    [[TMP18:%.*]] = extractelement <4 x i1> [[TMP6]], i32 3
+; VEC4_INTERL2-NEXT:    br i1 [[TMP18]], label [[PRED_STORE_IF8:%.*]], label [[PRED_STORE_CONTINUE9:%.*]]
+; VEC4_INTERL2:       pred.store.if8:
+; VEC4_INTERL2-NEXT:    [[TMP19:%.*]] = fadd fast float [[TMP1]], 3.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP20:%.*]] = or i64 [[INDEX]], 3
+; VEC4_INTERL2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP20]]
+; VEC4_INTERL2-NEXT:    store float [[TMP19]], float* [[TMP21]], align 4
+; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE9]]
+; VEC4_INTERL2:       pred.store.continue9:
+; VEC4_INTERL2-NEXT:    [[TMP22:%.*]] = extractelement <4 x i1> [[TMP7]], i32 0
+; VEC4_INTERL2-NEXT:    br i1 [[TMP22]], label [[PRED_STORE_IF10:%.*]], label [[PRED_STORE_CONTINUE11:%.*]]
+; VEC4_INTERL2:       pred.store.if10:
+; VEC4_INTERL2-NEXT:    [[TMP23:%.*]] = fadd fast float [[TMP1]], 4.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP0]]
+; VEC4_INTERL2-NEXT:    store float [[TMP23]], float* [[TMP24]], align 4
+; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE11]]
+; VEC4_INTERL2:       pred.store.continue11:
+; VEC4_INTERL2-NEXT:    [[TMP25:%.*]] = extractelement <4 x i1> [[TMP7]], i32 1
+; VEC4_INTERL2-NEXT:    br i1 [[TMP25]], label [[PRED_STORE_IF12:%.*]], label [[PRED_STORE_CONTINUE13:%.*]]
+; VEC4_INTERL2:       pred.store.if12:
+; VEC4_INTERL2-NEXT:    [[TMP26:%.*]] = fadd fast float [[TMP1]], 5.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP27:%.*]] = or i64 [[INDEX]], 5
+; VEC4_INTERL2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP27]]
+; VEC4_INTERL2-NEXT:    store float [[TMP26]], float* [[TMP28]], align 4
+; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE13]]
+; VEC4_INTERL2:       pred.store.continue13:
+; VEC4_INTERL2-NEXT:    [[TMP29:%.*]] = extractelement <4 x i1> [[TMP7]], i32 2
+; VEC4_INTERL2-NEXT:    br i1 [[TMP29]], label [[PRED_STORE_IF14:%.*]], label [[PRED_STORE_CONTINUE15:%.*]]
+; VEC4_INTERL2:       pred.store.if14:
+; VEC4_INTERL2-NEXT:    [[TMP30:%.*]] = fadd fast float [[TMP1]], 6.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP31:%.*]] = or i64 [[INDEX]], 6
+; VEC4_INTERL2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP31]]
+; VEC4_INTERL2-NEXT:    store float [[TMP30]], float* [[TMP32]], align 4
+; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE15]]
+; VEC4_INTERL2:       pred.store.continue15:
+; VEC4_INTERL2-NEXT:    [[TMP33:%.*]] = extractelement <4 x i1> [[TMP7]], i32 3
+; VEC4_INTERL2-NEXT:    br i1 [[TMP33]], label [[PRED_STORE_IF16:%.*]], label [[PRED_STORE_CONTINUE17]]
+; VEC4_INTERL2:       pred.store.if16:
+; VEC4_INTERL2-NEXT:    [[TMP34:%.*]] = fadd fast float [[TMP1]], 7.000000e+00
+; VEC4_INTERL2-NEXT:    [[TMP35:%.*]] = or i64 [[INDEX]], 7
+; VEC4_INTERL2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP35]]
+; VEC4_INTERL2-NEXT:    store float [[TMP34]], float* [[TMP36]], align 4
+; VEC4_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE17]]
+; VEC4_INTERL2:       pred.store.continue17:
+; VEC4_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; VEC4_INTERL2-NEXT:    [[TMP37:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[TMP37]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; VEC4_INTERL2:       middle.block:
+; VEC4_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; VEC4_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL2:       scalar.ph:
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; VEC4_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[CAST_CRD]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; VEC4_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC4_INTERL2:       for.body:
+; VEC4_INTERL2-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[J:%.*]] = phi float [ [[J_NEXT:%.*]], [[FOR_INC]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC4_INTERL2-NEXT:    [[VAR0:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; VEC4_INTERL2-NEXT:    [[VAR1:%.*]] = load float, float* [[VAR0]], align 4
+; VEC4_INTERL2-NEXT:    [[VAR2:%.*]] = fcmp fast oeq float [[VAR1]], 0.000000e+00
+; VEC4_INTERL2-NEXT:    br i1 [[VAR2]], label [[IF_PRED:%.*]], label [[FOR_INC]]
+; VEC4_INTERL2:       if.pred:
+; VEC4_INTERL2-NEXT:    store float [[J]], float* [[VAR0]], align 4
+; VEC4_INTERL2-NEXT:    br label [[FOR_INC]]
+; VEC4_INTERL2:       for.inc:
+; VEC4_INTERL2-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; VEC4_INTERL2-NEXT:    [[J_NEXT]] = fadd fast float [[J]], 1.000000e+00
+; VEC4_INTERL2-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; VEC4_INTERL2-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP13:![0-9]+]]
+; VEC4_INTERL2:       for.end:
+; VEC4_INTERL2-NEXT:    ret void
+;
+; VEC1_INTERL2-LABEL: @non_primary_iv_float_scalar(
+; VEC1_INTERL2-NEXT:  entry:
+; VEC1_INTERL2-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; VEC1_INTERL2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
+; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; VEC1_INTERL2:       vector.ph:
+; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
+; VEC1_INTERL2-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; VEC1_INTERL2:       vector.body:
+; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE5:%.*]] ]
+; VEC1_INTERL2-NEXT:    [[INDUCTION2:%.*]] = or i64 [[INDEX]], 1
+; VEC1_INTERL2-NEXT:    [[TMP0:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC1_INTERL2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDUCTION2]]
+; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = load float, float* [[TMP1]], align 4
+; VEC1_INTERL2-NEXT:    [[TMP4:%.*]] = load float, float* [[TMP2]], align 4
+; VEC1_INTERL2-NEXT:    [[TMP5:%.*]] = fcmp fast oeq float [[TMP3]], 0.000000e+00
+; VEC1_INTERL2-NEXT:    [[TMP6:%.*]] = fcmp fast oeq float [[TMP4]], 0.000000e+00
+; VEC1_INTERL2-NEXT:    br i1 [[TMP5]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; VEC1_INTERL2:       pred.store.if:
+; VEC1_INTERL2-NEXT:    store float [[TMP0]], float* [[TMP1]], align 4
+; VEC1_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; VEC1_INTERL2:       pred.store.continue:
+; VEC1_INTERL2-NEXT:    br i1 [[TMP6]], label [[PRED_STORE_IF4:%.*]], label [[PRED_STORE_CONTINUE5]]
+; VEC1_INTERL2:       pred.store.if4:
+; VEC1_INTERL2-NEXT:    [[TMP7:%.*]] = fadd fast float [[TMP0]], 1.000000e+00
+; VEC1_INTERL2-NEXT:    store float [[TMP7]], float* [[TMP2]], align 4
+; VEC1_INTERL2-NEXT:    br label [[PRED_STORE_CONTINUE5]]
+; VEC1_INTERL2:       pred.store.continue5:
+; VEC1_INTERL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VEC1_INTERL2-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; VEC1_INTERL2:       middle.block:
+; VEC1_INTERL2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; VEC1_INTERL2-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; VEC1_INTERL2:       scalar.ph:
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; VEC1_INTERL2-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi float [ [[CAST_CRD]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; VEC1_INTERL2-NEXT:    br label [[FOR_BODY:%.*]]
+; VEC1_INTERL2:       for.body:
+; VEC1_INTERL2-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[J:%.*]] = phi float [ [[J_NEXT:%.*]], [[FOR_INC]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; VEC1_INTERL2-NEXT:    [[VAR0:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; VEC1_INTERL2-NEXT:    [[VAR1:%.*]] = load float, float* [[VAR0]], align 4
+; VEC1_INTERL2-NEXT:    [[VAR2:%.*]] = fcmp fast oeq float [[VAR1]], 0.000000e+00
+; VEC1_INTERL2-NEXT:    br i1 [[VAR2]], label [[IF_PRED:%.*]], label [[FOR_INC]]
+; VEC1_INTERL2:       if.pred:
+; VEC1_INTERL2-NEXT:    store float [[J]], float* [[VAR0]], align 4
+; VEC1_INTERL2-NEXT:    br label [[FOR_INC]]
+; VEC1_INTERL2:       for.inc:
+; VEC1_INTERL2-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; VEC1_INTERL2-NEXT:    [[J_NEXT]] = fadd fast float [[J]], 1.000000e+00
+; VEC1_INTERL2-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; VEC1_INTERL2-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
+; VEC1_INTERL2:       for.end:
+; VEC1_INTERL2-NEXT:    ret void
+;
 ; VEC2_INTERL1_PRED_STORE-LABEL: @non_primary_iv_float_scalar(
+; VEC2_INTERL1_PRED_STORE-NEXT:  entry:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
+; VEC2_INTERL1_PRED_STORE:       vector.ph:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CAST_CRD:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC2_INTERL1_PRED_STORE:       vector.body:
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE7:.*]] ]
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = sitofp i64 [[INDEX]] to float
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDEX]]
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP3:%.*]] = bitcast float* [[TMP2]] to <2 x float>*
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP3]], align 4
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP4:%.*]] = fcmp fast oeq <2 x float> [[WIDE_LOAD]], zeroinitializer
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
-; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP5]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
-; VEC2_INTERL1_PRED_STORE:       [[PRED_STORE_IF]]:
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[GEP0:%.*]] = getelementptr inbounds float, float* %A, i64 [[INDEX]]
-; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[TMP1]], float* [[GEP0]], align 4
-; VEC2_INTERL1_PRED_STORE-NEXT:    br label %[[PRED_STORE_CONTINUE]]
-; VEC2_INTERL1_PRED_STORE:       [[PRED_STORE_CONTINUE]]:
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP4]], i32 1
-; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP8]], label %[[PRED_STORE_IF6:.*]], label %[[PRED_STORE_CONTINUE7]]
-; VEC2_INTERL1_PRED_STORE:       [[PRED_STORE_IF6]]:
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP9:%.*]] = fadd fast float [[TMP1]], 1.000000e+00
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP10:%.*]] = or i64 [[INDEX]], 1
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, float* %A, i64 [[TMP10]]
-; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[TMP9]], float* [[TMP11]], align 4
-; VEC2_INTERL1_PRED_STORE-NEXT:    br label %[[PRED_STORE_CONTINUE7]]
-; VEC2_INTERL1_PRED_STORE:       [[PRED_STORE_CONTINUE7]]:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE3:%.*]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP0:%.*]] = sitofp i64 [[INDEX]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[INDEX]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP2:%.*]] = bitcast float* [[TMP1]] to <2 x float>*
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP2]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP3:%.*]] = fcmp fast oeq <2 x float> [[WIDE_LOAD]], zeroinitializer
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP4]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; VEC2_INTERL1_PRED_STORE:       pred.store.if:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDEX]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[TMP0]], float* [[TMP5]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; VEC2_INTERL1_PRED_STORE:       pred.store.continue:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP6:%.*]] = extractelement <2 x i1> [[TMP3]], i32 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP6]], label [[PRED_STORE_IF2:%.*]], label [[PRED_STORE_CONTINUE3]]
+; VEC2_INTERL1_PRED_STORE:       pred.store.if2:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP7:%.*]] = fadd fast float [[TMP0]], 1.000000e+00
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP8:%.*]] = or i64 [[INDEX]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP8]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[TMP7]], float* [[TMP9]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[PRED_STORE_CONTINUE3]]
+; VEC2_INTERL1_PRED_STORE:       pred.store.continue3:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VEC2_INTERL1_PRED_STORE:         br i1 {{.*}}, label %middle.block, label %vector.body
-
-define void @non_primary_iv_float_scalar(float* %A, i64 %N) {
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       middle.block:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; VEC2_INTERL1_PRED_STORE:       for.body:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[J:%.*]] = phi float [ [[J_NEXT:%.*]], [[FOR_INC]] ], [ [[CAST_CRD]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VAR0:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VAR1:%.*]] = load float, float* [[VAR0]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[VAR2:%.*]] = fcmp fast oeq float [[VAR1]], 0.000000e+00
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[VAR2]], label [[IF_PRED:%.*]], label [[FOR_INC]]
+; VEC2_INTERL1_PRED_STORE:       if.pred:
+; VEC2_INTERL1_PRED_STORE-NEXT:    store float [[J]], float* [[VAR0]], align 4
+; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[FOR_INC]]
+; VEC2_INTERL1_PRED_STORE:       for.inc:
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[J_NEXT]] = fadd fast float [[J]], 1.000000e+00
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP13:![0-9]+]]
+; VEC2_INTERL1_PRED_STORE:       for.end:
+; VEC2_INTERL1_PRED_STORE-NEXT:    ret void
+;
 entry:
   br label %for.body
 
 for.body:
   %i = phi i64 [ %i.next, %for.inc ], [ 0, %entry ]
   %j = phi float [ %j.next, %for.inc ], [ 0.0, %entry ]
-  %tmp0 = getelementptr inbounds float, float* %A, i64 %i
-  %tmp1 = load float, float* %tmp0, align 4
-  %tmp2 = fcmp fast oeq float %tmp1, 0.0
-  br i1 %tmp2, label %if.pred, label %for.inc
+  %var0 = getelementptr inbounds float, float* %A, i64 %i
+  %var1 = load float, float* %var0, align 4
+  %var2 = fcmp fast oeq float %var1, 0.0
+  br i1 %var2, label %if.pred, label %for.inc
 
 if.pred:
-  store float %j, float* %tmp0, align 4
+  store float %j, float* %var0, align 4
   br label %for.inc
 
 for.inc:

diff  --git a/llvm/test/Transforms/LoopVectorize/induction.ll b/llvm/test/Transforms/LoopVectorize/induction.ll
index 75326216c6d7..f649b6ad026f 100644
--- a/llvm/test/Transforms/LoopVectorize/induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/induction.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=2 -S | FileCheck %s
 ; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=2 -instcombine -S | FileCheck %s --check-prefix=IND
 ; RUN: opt < %s -loop-vectorize -force-vector-interleave=2 -force-vector-width=2 -instcombine -S | FileCheck %s --check-prefix=UNROLL
@@ -8,19 +9,237 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
 
 ; Make sure that we can handle multiple integer induction variables.
 ;
+define void @multi_int_induction(i32* %A, i32 %N) {
 ; CHECK-LABEL: @multi_int_induction(
+; CHECK-NEXT:  for.body.lr.ph:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; CHECK-NEXT:    [[IND_END:%.*]] = add i32 190, [[CAST_CRD]]
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK-NEXT:    %vec.ind = phi <2 x i32> [ <i32 190, i32 191>, %vector.ph ], [ %vec.ind.next, %vector.body ]
-; CHECK:         [[TMP3:%.*]] = add i64 %index, 0
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* %A, i64 [[TMP3]]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP3]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <2 x i32>*
-; CHECK-NEXT:    store <2 x i32> %vec.ind, <2 x i32>* [[TMP6]], align 4
-; CHECK:         %index.next = add nuw i64 %index, 2
-; CHECK-NEXT:    %vec.ind.next = add <2 x i32> %vec.ind, <i32 2, i32 2>
-; CHECK:         br i1 {{.*}}, label %middle.block, label %vector.body
-define void @multi_int_induction(i32* %A, i32 %N) {
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP6]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
+; CHECK-NEXT:    [[INC]] = add nsw i32 [[COUNT_09]], 1
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp ne i32 [[LFTR_WIDEIV]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP2:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @multi_int_induction(
+; IND-NEXT:  for.body.lr.ph:
+; IND-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; IND-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; IND-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; IND-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; IND-NEXT:    [[IND_END:%.*]] = add i32 [[CAST_CRD]], 190
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
+; IND-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP4]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; IND-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
+; IND-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; IND-NEXT:    [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; IND-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; IND-NEXT:    store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
+; IND-NEXT:    [[INC]] = add nsw i32 [[COUNT_09]], 1
+; IND-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
+; IND-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; IND-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; IND-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @multi_int_induction(
+; UNROLL-NEXT:  for.body.lr.ph:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
+; UNROLL-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; UNROLL-NEXT:    [[IND_END:%.*]] = add i32 [[CAST_CRD]], 190
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; UNROLL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP4]], align 4
+; UNROLL-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i64 2
+; UNROLL-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP6]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; UNROLL-NEXT:    store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
+; UNROLL-NEXT:    [[INC]] = add nsw i32 [[COUNT_09]], 1
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; UNROLL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @multi_int_induction(
+; UNROLL-NO-IC-NEXT:  for.body.lr.ph:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = add i32 190, [[CAST_CRD]]
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP8]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP10]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
+; UNROLL-NO-IC-NEXT:    [[INC]] = add nsw i32 [[COUNT_09]], 1
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp ne i32 [[LFTR_WIDEIV]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP2:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @multi_int_induction(
+; INTERLEAVE-NEXT:  for.body.lr.ph:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
+; INTERLEAVE-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; INTERLEAVE-NEXT:    [[IND_END:%.*]] = add i32 [[CAST_CRD]], 190
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 190, i32 191, i32 192, i32 193>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP4]], align 4
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i64 4
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP6]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
+; INTERLEAVE-NEXT:    store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
+; INTERLEAVE-NEXT:    [[INC]] = add nsw i32 [[COUNT_09]], 1
+; INTERLEAVE-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
+; INTERLEAVE-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; INTERLEAVE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    ret void
+;
 for.body.lr.ph:
   br label %for.body
 
@@ -44,16 +263,367 @@ for.end:
 ; create in the loop vectorizer we need to perform some form of redundancy
 ; elimination to get rid of multiple uses.
 
-; IND-LABEL: scalar_use
 
-; IND:     br label %vector.body
-; IND:     vector.body:
 ;   Vectorized induction variable.
-; IND-NOT:  insertelement <2 x i64>
-; IND-NOT:  shufflevector <2 x i64>
-; IND:     br {{.*}}, label %vector.body
 
 define void @scalar_use(float* %a, float %b, i64 %offset, i64 %offset2, i64 %n) {
+; CHECK-LABEL: @scalar_use(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK:       vector.memcheck:
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
+; CHECK-NEXT:    [[SCEVGEP1:%.*]] = bitcast float* [[SCEVGEP]] to i8*
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
+; CHECK-NEXT:    [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
+; CHECK-NEXT:    [[SCEVGEP23:%.*]] = bitcast float* [[SCEVGEP2]] to i8*
+; CHECK-NEXT:    [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
+; CHECK-NEXT:    [[SCEVGEP45:%.*]] = bitcast float* [[SCEVGEP4]] to i8*
+; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
+; CHECK-NEXT:    [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
+; CHECK-NEXT:    [[SCEVGEP67:%.*]] = bitcast float* [[SCEVGEP6]] to i8*
+; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP67]]
+; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP45]], [[SCEVGEP23]]
+; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[TMP2]], [[OFFSET]]
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[TMP4]], i32 0
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP5]] to <2 x float>*
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP6]], align 4, !alias.scope !4, !noalias !7
+; CHECK-NEXT:    [[TMP7:%.*]] = add i64 [[TMP2]], [[OFFSET2]]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 0
+; CHECK-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP9]] to <2 x float>*
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x float>, <2 x float>* [[TMP10]], align 4, !alias.scope !7
+; CHECK-NEXT:    [[TMP11:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD8]]
+; CHECK-NEXT:    [[TMP12:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP11]]
+; CHECK-NEXT:    [[TMP13:%.*]] = bitcast float* [[TMP5]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP12]], <2 x float>* [[TMP13]], align 4, !alias.scope !4, !noalias !7
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
+; CHECK-NEXT:    [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
+; CHECK-NEXT:    [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; CHECK-NEXT:    [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
+; CHECK-NEXT:    [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; CHECK-NEXT:    [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; CHECK-NEXT:    [[M:%.*]] = fmul fast float [[B]], [[L2]]
+; CHECK-NEXT:    [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; CHECK-NEXT:    store float [[AD]], float* [[ARR_IDX]], align 4
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK:       loopexit:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @scalar_use(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; IND:       vector.memcheck:
+; IND-NEXT:    [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
+; IND-NEXT:    [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
+; IND-NEXT:    [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
+; IND-NEXT:    [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
+; IND-NEXT:    [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
+; IND-NEXT:    [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
+; IND-NEXT:    [[BOUND0:%.*]] = icmp ult float* [[SCEVGEP]], [[SCEVGEP6]]
+; IND-NEXT:    [[BOUND1:%.*]] = icmp ult float* [[SCEVGEP4]], [[SCEVGEP2]]
+; IND-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; IND-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -2
+; IND-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i32 0
+; IND-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], [[OFFSET]]
+; IND-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP2]]
+; IND-NEXT:    [[TMP4:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
+; IND-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP4]], align 4, !alias.scope !4, !noalias !7
+; IND-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
+; IND-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
+; IND-NEXT:    [[TMP7:%.*]] = bitcast float* [[TMP6]] to <2 x float>*
+; IND-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x float>, <2 x float>* [[TMP7]], align 4, !alias.scope !7
+; IND-NEXT:    [[TMP8:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD8]]
+; IND-NEXT:    [[TMP9:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP8]]
+; IND-NEXT:    [[TMP10:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
+; IND-NEXT:    store <2 x float> [[TMP9]], <2 x float>* [[TMP10]], align 4, !alias.scope !4, !noalias !7
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; IND-NEXT:    [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
+; IND-NEXT:    [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
+; IND-NEXT:    [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; IND-NEXT:    [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
+; IND-NEXT:    [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; IND-NEXT:    [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; IND-NEXT:    [[M:%.*]] = fmul fast float [[L2]], [[B]]
+; IND-NEXT:    [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; IND-NEXT:    store float [[AD]], float* [[ARR_IDX]], align 4
+; IND-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IND-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IND-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; IND:       loopexit:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @scalar_use(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL:       vector.memcheck:
+; UNROLL-NEXT:    [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
+; UNROLL-NEXT:    [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
+; UNROLL-NEXT:    [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
+; UNROLL-NEXT:    [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
+; UNROLL-NEXT:    [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
+; UNROLL-NEXT:    [[BOUND0:%.*]] = icmp ult float* [[SCEVGEP]], [[SCEVGEP6]]
+; UNROLL-NEXT:    [[BOUND1:%.*]] = icmp ult float* [[SCEVGEP4]], [[SCEVGEP2]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -4
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT11:%.*]] = insertelement <2 x float> poison, float [[B]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT12:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT11]], <2 x float> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], [[OFFSET]]
+; UNROLL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP2]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
+; UNROLL-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP4]], align 4, !alias.scope !4, !noalias !7
+; UNROLL-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[TMP3]], i64 2
+; UNROLL-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP5]] to <2 x float>*
+; UNROLL-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x float>, <2 x float>* [[TMP6]], align 4, !alias.scope !4, !noalias !7
+; UNROLL-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP7]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP8]] to <2 x float>*
+; UNROLL-NEXT:    [[WIDE_LOAD9:%.*]] = load <2 x float>, <2 x float>* [[TMP9]], align 4, !alias.scope !7
+; UNROLL-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 2
+; UNROLL-NEXT:    [[TMP11:%.*]] = bitcast float* [[TMP10]] to <2 x float>*
+; UNROLL-NEXT:    [[WIDE_LOAD10:%.*]] = load <2 x float>, <2 x float>* [[TMP11]], align 4, !alias.scope !7
+; UNROLL-NEXT:    [[TMP12:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD9]]
+; UNROLL-NEXT:    [[TMP13:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT12]], [[WIDE_LOAD10]]
+; UNROLL-NEXT:    [[TMP14:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP12]]
+; UNROLL-NEXT:    [[TMP15:%.*]] = fadd fast <2 x float> [[WIDE_LOAD8]], [[TMP13]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
+; UNROLL-NEXT:    store <2 x float> [[TMP14]], <2 x float>* [[TMP16]], align 4, !alias.scope !4, !noalias !7
+; UNROLL-NEXT:    [[TMP17:%.*]] = bitcast float* [[TMP5]] to <2 x float>*
+; UNROLL-NEXT:    store <2 x float> [[TMP15]], <2 x float>* [[TMP17]], align 4, !alias.scope !4, !noalias !7
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
+; UNROLL-NEXT:    [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
+; UNROLL-NEXT:    [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; UNROLL-NEXT:    [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
+; UNROLL-NEXT:    [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; UNROLL-NEXT:    [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; UNROLL-NEXT:    [[M:%.*]] = fmul fast float [[L2]], [[B]]
+; UNROLL-NEXT:    [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; UNROLL-NEXT:    store float [[AD]], float* [[ARR_IDX]], align 4
+; UNROLL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; UNROLL:       loopexit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @scalar_use(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-IC:       vector.memcheck:
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP1:%.*]] = bitcast float* [[SCEVGEP]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP23:%.*]] = bitcast float* [[SCEVGEP2]] to i8*
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP45:%.*]] = bitcast float* [[SCEVGEP4]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP67:%.*]] = bitcast float* [[SCEVGEP6]] to i8*
+; UNROLL-NO-IC-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP67]]
+; UNROLL-NO-IC-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP45]], [[SCEVGEP23]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-IC-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT11:%.*]] = insertelement <2 x float> poison, float [[B]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT12:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT11]], <2 x float> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[TMP2]], [[OFFSET]]
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i64 [[TMP3]], [[OFFSET]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP6]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP8]] to <2 x float>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP9]], align 4, !alias.scope !4, !noalias !7
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP6]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = bitcast float* [[TMP10]] to <2 x float>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x float>, <2 x float>* [[TMP11]], align 4, !alias.scope !4, !noalias !7
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add i64 [[TMP2]], [[OFFSET2]]
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = add i64 [[TMP3]], [[OFFSET2]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]]
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP13]]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP14]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = bitcast float* [[TMP16]] to <2 x float>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD9:%.*]] = load <2 x float>, <2 x float>* [[TMP17]], align 4, !alias.scope !7
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[TMP14]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = bitcast float* [[TMP18]] to <2 x float>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD10:%.*]] = load <2 x float>, <2 x float>* [[TMP19]], align 4, !alias.scope !7
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD9]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT12]], [[WIDE_LOAD10]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP20]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = fadd fast <2 x float> [[WIDE_LOAD8]], [[TMP21]]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = bitcast float* [[TMP8]] to <2 x float>*
+; UNROLL-NO-IC-NEXT:    store <2 x float> [[TMP22]], <2 x float>* [[TMP24]], align 4, !alias.scope !4, !noalias !7
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = bitcast float* [[TMP10]] to <2 x float>*
+; UNROLL-NO-IC-NEXT:    store <2 x float> [[TMP23]], <2 x float>* [[TMP25]], align 4, !alias.scope !4, !noalias !7
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
+; UNROLL-NO-IC-NEXT:    [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
+; UNROLL-NO-IC-NEXT:    [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; UNROLL-NO-IC-NEXT:    [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
+; UNROLL-NO-IC-NEXT:    [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; UNROLL-NO-IC-NEXT:    [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; UNROLL-NO-IC-NEXT:    [[M:%.*]] = fmul fast float [[B]], [[L2]]
+; UNROLL-NO-IC-NEXT:    [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; UNROLL-NO-IC-NEXT:    store float [[AD]], float* [[ARR_IDX]], align 4
+; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; UNROLL-NO-IC:       loopexit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @scalar_use(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; INTERLEAVE:       vector.memcheck:
+; INTERLEAVE-NEXT:    [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
+; INTERLEAVE-NEXT:    [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
+; INTERLEAVE-NEXT:    [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
+; INTERLEAVE-NEXT:    [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
+; INTERLEAVE-NEXT:    [[BOUND0:%.*]] = icmp ult float* [[SCEVGEP]], [[SCEVGEP6]]
+; INTERLEAVE-NEXT:    [[BOUND1:%.*]] = icmp ult float* [[SCEVGEP4]], [[SCEVGEP2]]
+; INTERLEAVE-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; INTERLEAVE-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -8
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[B:%.*]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT11:%.*]] = insertelement <4 x float> poison, float [[B]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLAT12:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT11]], <4 x float> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], [[OFFSET]]
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP2]]
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = bitcast float* [[TMP3]] to <4 x float>*
+; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 4, !alias.scope !4, !noalias !7
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[TMP3]], i64 4
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP5]] to <4 x float>*
+; INTERLEAVE-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP6]], align 4, !alias.scope !4, !noalias !7
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP7]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP8]] to <4 x float>*
+; INTERLEAVE-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, <4 x float>* [[TMP9]], align 4, !alias.scope !7
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 4
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = bitcast float* [[TMP10]] to <4 x float>*
+; INTERLEAVE-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP11]], align 4, !alias.scope !7
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = fmul fast <4 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD9]]
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = fmul fast <4 x float> [[BROADCAST_SPLAT12]], [[WIDE_LOAD10]]
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[TMP12]]
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[TMP13]]
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = bitcast float* [[TMP3]] to <4 x float>*
+; INTERLEAVE-NEXT:    store <4 x float> [[TMP14]], <4 x float>* [[TMP16]], align 4, !alias.scope !4, !noalias !7
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = bitcast float* [[TMP5]] to <4 x float>*
+; INTERLEAVE-NEXT:    store <4 x float> [[TMP15]], <4 x float>* [[TMP17]], align 4, !alias.scope !4, !noalias !7
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
+; INTERLEAVE-NEXT:    [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
+; INTERLEAVE-NEXT:    [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
+; INTERLEAVE-NEXT:    [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
+; INTERLEAVE-NEXT:    [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
+; INTERLEAVE-NEXT:    [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
+; INTERLEAVE-NEXT:    [[M:%.*]] = fmul fast float [[L2]], [[B]]
+; INTERLEAVE-NEXT:    [[AD:%.*]] = fadd fast float [[L1]], [[M]]
+; INTERLEAVE-NEXT:    store float [[AD]], float* [[ARR_IDX]], align 4
+; INTERLEAVE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; INTERLEAVE-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; INTERLEAVE:       loopexit:
+; INTERLEAVE-NEXT:    ret void
+;
 entry:
   br label %for.body
 
@@ -82,34 +652,232 @@ loopexit:
 ; for (int i = 0; i < n; ++i)
 ;   sum += a[i];
 ;
-; CHECK-LABEL: @scalarize_induction_variable_01(
-; CHECK: vector.body:
-; CHECK:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK:   %[[i0:.+]] = add i64 %index, 0
-; CHECK:   getelementptr inbounds i64, i64* %a, i64 %[[i0]]
 ;
-; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_01(
-; UNROLL-NO-IC: vector.body:
-; UNROLL-NO-IC:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; UNROLL-NO-IC:   %[[i0:.+]] = add i64 %index, 0
-; UNROLL-NO-IC:   %[[i2:.+]] = add i64 %index, 2
-; UNROLL-NO-IC:   getelementptr inbounds i64, i64* %a, i64 %[[i0]]
-; UNROLL-NO-IC:   getelementptr inbounds i64, i64* %a, i64 %[[i2]]
 ;
-; IND-LABEL: @scalarize_induction_variable_01(
-; IND:     vector.body:
-; IND:       %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; IND-NOT:   add i64 {{.*}}, 2
-; IND:       getelementptr inbounds i64, i64* %a, i64 %index
 ;
-; UNROLL-LABEL: @scalarize_induction_variable_01(
-; UNROLL:     vector.body:
-; UNROLL:       %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; UNROLL-NOT:   add i64 {{.*}}, 4
-; UNROLL:       %[[g1:.+]] = getelementptr inbounds i64, i64* %a, i64 %index
-; UNROLL:       getelementptr inbounds i64, i64* %[[g1]], i64 2
 
 define i64 @scalarize_induction_variable_01(i64 *%a, i64 %n) {
+; CHECK-LABEL: @scalarize_induction_variable_01(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[TMP2]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP3]], align 8
+; CHECK-NEXT:    [[TMP4]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP4]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[SUM:%.*]] = phi i64 [ [[TMP9:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
+; CHECK-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
+; CHECK-NEXT:    [[TMP9]] = add i64 [[TMP8]], [[SUM]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; CHECK-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    [[TMP10:%.*]] = phi i64 [ [[TMP9]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[TMP10]]
+;
+; IND-LABEL: @scalarize_induction_variable_01(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[INDEX]]
+; IND-NEXT:    [[TMP1:%.*]] = bitcast i64* [[TMP0]] to <2 x i64>*
+; IND-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP1]], align 8
+; IND-NEXT:    [[TMP2]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP2]])
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; IND-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP4]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[SUM:%.*]] = phi i64 [ [[TMP7:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
+; IND-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
+; IND-NEXT:    [[TMP7]] = add i64 [[TMP6]], [[SUM]]
+; IND-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; IND-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; IND-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    [[TMP8:%.*]] = phi i64 [ [[TMP7]], [[FOR_BODY]] ], [ [[TMP4]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    ret i64 [[TMP8]]
+;
+; UNROLL-LABEL: @scalarize_induction_variable_01(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP1:%.*]] = bitcast i64* [[TMP0]] to <2 x i64>*
+; UNROLL-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP1]], align 8
+; UNROLL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[TMP0]], i64 2
+; UNROLL-NEXT:    [[TMP3:%.*]] = bitcast i64* [[TMP2]] to <2 x i64>*
+; UNROLL-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i64>, <2 x i64>* [[TMP3]], align 8
+; UNROLL-NEXT:    [[TMP4]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
+; UNROLL-NEXT:    [[TMP5]] = add <2 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP5]], [[TMP4]]
+; UNROLL-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[SUM:%.*]] = phi i64 [ [[TMP10:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = load i64, i64* [[TMP8]], align 8
+; UNROLL-NEXT:    [[TMP10]] = add i64 [[TMP9]], [[SUM]]
+; UNROLL-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    [[TMP11:%.*]] = phi i64 [ [[TMP10]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i64 [[TMP11]]
+;
+; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_01(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, i64* [[TMP2]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = bitcast i64* [[TMP4]] to <2 x i64>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP5]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, i64* [[TMP2]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = bitcast i64* [[TMP6]] to <2 x i64>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i64>, <2 x i64>* [[TMP7]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP8]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
+; UNROLL-NO-IC-NEXT:    [[TMP9]] = add <2 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP9]], [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[SUM:%.*]] = phi i64 [ [[TMP14:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = load i64, i64* [[TMP12]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP14]] = add i64 [[TMP13]], [[SUM]]
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = phi i64 [ [[TMP14]], [[FOR_BODY]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i64 [[TMP15]]
+;
+; INTERLEAVE-LABEL: @scalarize_induction_variable_01(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775800
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = bitcast i64* [[TMP0]] to <4 x i64>*
+; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, <4 x i64>* [[TMP1]], align 8
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[TMP0]], i64 4
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = bitcast i64* [[TMP2]] to <4 x i64>*
+; INTERLEAVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, <4 x i64>* [[TMP3]], align 8
+; INTERLEAVE-NEXT:    [[TMP4]] = add <4 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
+; INTERLEAVE-NEXT:    [[TMP5]] = add <4 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[BIN_RDX:%.*]] = add <4 x i64> [[TMP5]], [[TMP4]]
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[BIN_RDX]])
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; INTERLEAVE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[SUM:%.*]] = phi i64 [ [[TMP10:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = load i64, i64* [[TMP8]], align 8
+; INTERLEAVE-NEXT:    [[TMP10]] = add i64 [[TMP9]], [[SUM]]
+; INTERLEAVE-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = phi i64 [ [[TMP10]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    ret i64 [[TMP11]]
+;
 entry:
   br label %for.body
 
@@ -135,55 +903,363 @@ for.end:
 ; for (int i ; 0; i < n; i += 8)
 ;   s += (a[i] + b[i] + 1.0f);
 ;
-; CHECK-LABEL: @scalarize_induction_variable_02(
-; CHECK: vector.body:
-; CHECK:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK:   %offset.idx = mul i64 %index, 8
-; CHECK:   %[[i0:.+]] = add i64 %offset.idx, 0
-; CHECK:   %[[i1:.+]] = add i64 %offset.idx, 8
-; CHECK:   getelementptr inbounds float, float* %a, i64 %[[i0]]
-; CHECK:   getelementptr inbounds float, float* %a, i64 %[[i1]]
-; CHECK:   getelementptr inbounds float, float* %b, i64 %[[i0]]
-; CHECK:   getelementptr inbounds float, float* %b, i64 %[[i1]]
 ;
-; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_02(
-; UNROLL-NO-IC: vector.body:
-; UNROLL-NO-IC:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; UNROLL-NO-IC:   %offset.idx = mul i64 %index, 8
-; UNROLL-NO-IC:   %[[i0:.+]] = add i64 %offset.idx, 0
-; UNROLL-NO-IC:   %[[i1:.+]] = add i64 %offset.idx, 8
-; UNROLL-NO-IC:   %[[i2:.+]] = add i64 %offset.idx, 16
-; UNROLL-NO-IC:   %[[i3:.+]] = add i64 %offset.idx, 24
-; UNROLL-NO-IC:   getelementptr inbounds float, float* %a, i64 %[[i0]]
-; UNROLL-NO-IC:   getelementptr inbounds float, float* %a, i64 %[[i1]]
-; UNROLL-NO-IC:   getelementptr inbounds float, float* %a, i64 %[[i2]]
-; UNROLL-NO-IC:   getelementptr inbounds float, float* %a, i64 %[[i3]]
-; UNROLL-NO-IC:   getelementptr inbounds float, float* %b, i64 %[[i0]]
-; UNROLL-NO-IC:   getelementptr inbounds float, float* %b, i64 %[[i1]]
-; UNROLL-NO-IC:   getelementptr inbounds float, float* %b, i64 %[[i2]]
-; UNROLL-NO-IC:   getelementptr inbounds float, float* %b, i64 %[[i3]]
 ;
-; IND-LABEL: @scalarize_induction_variable_02(
-; IND: vector.body:
-; IND:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; IND:   %[[i0:.+]] = shl i64 %index, 3
-; IND:   %[[i1:.+]] = or i64 %[[i0]], 8
-; IND:   getelementptr inbounds float, float* %a, i64 %[[i0]]
-; IND:   getelementptr inbounds float, float* %a, i64 %[[i1]]
 ;
-; UNROLL-LABEL: @scalarize_induction_variable_02(
-; UNROLL: vector.body:
-; UNROLL:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; UNROLL:   %[[i0:.+]] = shl i64 %index, 3
-; UNROLL:   %[[i1:.+]] = or i64 %[[i0]], 8
-; UNROLL:   %[[i2:.+]] = or i64 %[[i0]], 16
-; UNROLL:   %[[i3:.+]] = or i64 %[[i0]], 24
-; UNROLL:   getelementptr inbounds float, float* %a, i64 %[[i0]]
-; UNROLL:   getelementptr inbounds float, float* %a, i64 %[[i1]]
-; UNROLL:   getelementptr inbounds float, float* %a, i64 %[[i2]]
-; UNROLL:   getelementptr inbounds float, float* %a, i64 %[[i3]]
 
 define float @scalarize_induction_variable_02(float* %a, float* %b, i64 %n) {
+; CHECK-LABEL: @scalarize_induction_variable_02(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
+; CHECK-NEXT:    [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 8
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 8
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP3]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP5]], align 4
+; CHECK-NEXT:    [[TMP8:%.*]] = load float, float* [[TMP6]], align 4
+; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP7]], i32 0
+; CHECK-NEXT:    [[TMP10:%.*]] = insertelement <2 x float> [[TMP9]], float [[TMP8]], i32 1
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP3]]
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP13:%.*]] = load float, float* [[TMP11]], align 4
+; CHECK-NEXT:    [[TMP14:%.*]] = load float, float* [[TMP12]], align 4
+; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <2 x float> poison, float [[TMP13]], i32 0
+; CHECK-NEXT:    [[TMP16:%.*]] = insertelement <2 x float> [[TMP15]], float [[TMP14]], i32 1
+; CHECK-NEXT:    [[TMP17:%.*]] = fadd fast <2 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00>
+; CHECK-NEXT:    [[TMP18:%.*]] = fadd fast <2 x float> [[TMP17]], [[TMP10]]
+; CHECK-NEXT:    [[TMP19]] = fadd fast <2 x float> [[TMP18]], [[TMP16]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP21:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float -0.000000e+00, <2 x float> [[TMP19]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; CHECK-NEXT:    [[TMP23:%.*]] = load float, float* [[TMP22]], align 4
+; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
+; CHECK-NEXT:    [[TMP25:%.*]] = load float, float* [[TMP24]], align 4
+; CHECK-NEXT:    [[TMP26:%.*]] = fadd fast float [[S]], 1.000000e+00
+; CHECK-NEXT:    [[TMP27:%.*]] = fadd fast float [[TMP26]], [[TMP23]]
+; CHECK-NEXT:    [[TMP28]] = fadd fast float [[TMP27]], [[TMP25]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 8
+; CHECK-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    [[S_LCSSA:%.*]] = phi float [ [[TMP28]], [[FOR_BODY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret float [[S_LCSSA]]
+;
+; IND-LABEL: @scalarize_induction_variable_02(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
+; IND-NEXT:    [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
+; IND-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
+; IND-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 8
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 4611686018427387902
+; IND-NEXT:    [[IND_END:%.*]] = shl i64 [[N_VEC]], 3
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3
+; IND-NEXT:    [[TMP3:%.*]] = or i64 [[OFFSET_IDX]], 8
+; IND-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[OFFSET_IDX]]
+; IND-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP3]]
+; IND-NEXT:    [[TMP6:%.*]] = load float, float* [[TMP4]], align 4
+; IND-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP5]], align 4
+; IND-NEXT:    [[TMP8:%.*]] = insertelement <2 x float> poison, float [[TMP6]], i32 0
+; IND-NEXT:    [[TMP9:%.*]] = insertelement <2 x float> [[TMP8]], float [[TMP7]], i32 1
+; IND-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[OFFSET_IDX]]
+; IND-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP3]]
+; IND-NEXT:    [[TMP12:%.*]] = load float, float* [[TMP10]], align 4
+; IND-NEXT:    [[TMP13:%.*]] = load float, float* [[TMP11]], align 4
+; IND-NEXT:    [[TMP14:%.*]] = insertelement <2 x float> poison, float [[TMP12]], i32 0
+; IND-NEXT:    [[TMP15:%.*]] = insertelement <2 x float> [[TMP14]], float [[TMP13]], i32 1
+; IND-NEXT:    [[TMP16:%.*]] = fadd fast <2 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00>
+; IND-NEXT:    [[TMP17:%.*]] = fadd fast <2 x float> [[TMP16]], [[TMP9]]
+; IND-NEXT:    [[TMP18]] = fadd fast <2 x float> [[TMP17]], [[TMP15]]
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[TMP20:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float -0.000000e+00, <2 x float> [[TMP18]])
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; IND-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP20]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
+; IND-NEXT:    [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP27:%.*]], [[FOR_BODY]] ]
+; IND-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; IND-NEXT:    [[TMP22:%.*]] = load float, float* [[TMP21]], align 4
+; IND-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
+; IND-NEXT:    [[TMP24:%.*]] = load float, float* [[TMP23]], align 4
+; IND-NEXT:    [[TMP25:%.*]] = fadd fast float [[S]], 1.000000e+00
+; IND-NEXT:    [[TMP26:%.*]] = fadd fast float [[TMP25]], [[TMP22]]
+; IND-NEXT:    [[TMP27]] = fadd fast float [[TMP26]], [[TMP24]]
+; IND-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 8
+; IND-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; IND-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    [[S_LCSSA:%.*]] = phi float [ [[TMP27]], [[FOR_BODY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    ret float [[S_LCSSA]]
+;
+; UNROLL-LABEL: @scalarize_induction_variable_02(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
+; UNROLL-NEXT:    [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
+; UNROLL-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
+; UNROLL-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 24
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 4611686018427387900
+; UNROLL-NEXT:    [[IND_END:%.*]] = shl i64 [[N_VEC]], 3
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP34:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3
+; UNROLL-NEXT:    [[TMP3:%.*]] = or i64 [[OFFSET_IDX]], 8
+; UNROLL-NEXT:    [[TMP4:%.*]] = or i64 [[OFFSET_IDX]], 16
+; UNROLL-NEXT:    [[TMP5:%.*]] = or i64 [[OFFSET_IDX]], 24
+; UNROLL-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[OFFSET_IDX]]
+; UNROLL-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP3]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
+; UNROLL-NEXT:    [[TMP10:%.*]] = load float, float* [[TMP6]], align 4
+; UNROLL-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP7]], align 4
+; UNROLL-NEXT:    [[TMP12:%.*]] = insertelement <2 x float> poison, float [[TMP10]], i32 0
+; UNROLL-NEXT:    [[TMP13:%.*]] = insertelement <2 x float> [[TMP12]], float [[TMP11]], i32 1
+; UNROLL-NEXT:    [[TMP14:%.*]] = load float, float* [[TMP8]], align 4
+; UNROLL-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP9]], align 4
+; UNROLL-NEXT:    [[TMP16:%.*]] = insertelement <2 x float> poison, float [[TMP14]], i32 0
+; UNROLL-NEXT:    [[TMP17:%.*]] = insertelement <2 x float> [[TMP16]], float [[TMP15]], i32 1
+; UNROLL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[OFFSET_IDX]]
+; UNROLL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP3]]
+; UNROLL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP4]]
+; UNROLL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP5]]
+; UNROLL-NEXT:    [[TMP22:%.*]] = load float, float* [[TMP18]], align 4
+; UNROLL-NEXT:    [[TMP23:%.*]] = load float, float* [[TMP19]], align 4
+; UNROLL-NEXT:    [[TMP24:%.*]] = insertelement <2 x float> poison, float [[TMP22]], i32 0
+; UNROLL-NEXT:    [[TMP25:%.*]] = insertelement <2 x float> [[TMP24]], float [[TMP23]], i32 1
+; UNROLL-NEXT:    [[TMP26:%.*]] = load float, float* [[TMP20]], align 4
+; UNROLL-NEXT:    [[TMP27:%.*]] = load float, float* [[TMP21]], align 4
+; UNROLL-NEXT:    [[TMP28:%.*]] = insertelement <2 x float> poison, float [[TMP26]], i32 0
+; UNROLL-NEXT:    [[TMP29:%.*]] = insertelement <2 x float> [[TMP28]], float [[TMP27]], i32 1
+; UNROLL-NEXT:    [[TMP30:%.*]] = fadd fast <2 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00>
+; UNROLL-NEXT:    [[TMP31:%.*]] = fadd fast <2 x float> [[VEC_PHI1]], <float 1.000000e+00, float 1.000000e+00>
+; UNROLL-NEXT:    [[TMP32:%.*]] = fadd fast <2 x float> [[TMP30]], [[TMP13]]
+; UNROLL-NEXT:    [[TMP33:%.*]] = fadd fast <2 x float> [[TMP31]], [[TMP17]]
+; UNROLL-NEXT:    [[TMP34]] = fadd fast <2 x float> [[TMP32]], [[TMP25]]
+; UNROLL-NEXT:    [[TMP35]] = fadd fast <2 x float> [[TMP33]], [[TMP29]]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = fadd fast <2 x float> [[TMP35]], [[TMP34]]
+; UNROLL-NEXT:    [[TMP37:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float -0.000000e+00, <2 x float> [[BIN_RDX]])
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP37]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP44:%.*]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; UNROLL-NEXT:    [[TMP39:%.*]] = load float, float* [[TMP38]], align 4
+; UNROLL-NEXT:    [[TMP40:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
+; UNROLL-NEXT:    [[TMP41:%.*]] = load float, float* [[TMP40]], align 4
+; UNROLL-NEXT:    [[TMP42:%.*]] = fadd fast float [[S]], 1.000000e+00
+; UNROLL-NEXT:    [[TMP43:%.*]] = fadd fast float [[TMP42]], [[TMP39]]
+; UNROLL-NEXT:    [[TMP44]] = fadd fast float [[TMP43]], [[TMP41]]
+; UNROLL-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 8
+; UNROLL-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    [[S_LCSSA:%.*]] = phi float [ [[TMP44]], [[FOR_BODY]] ], [ [[TMP37]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret float [[S_LCSSA]]
+;
+; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_02(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 8
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP36:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 16
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 24
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP6]]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP7]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = load float, float* [[TMP8]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = insertelement <2 x float> poison, float [[TMP11]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = insertelement <2 x float> [[TMP13]], float [[TMP12]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP9]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = load float, float* [[TMP10]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = insertelement <2 x float> poison, float [[TMP15]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = insertelement <2 x float> [[TMP17]], float [[TMP16]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP6]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = load float, float* [[TMP19]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = load float, float* [[TMP20]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = insertelement <2 x float> poison, float [[TMP23]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = insertelement <2 x float> [[TMP25]], float [[TMP24]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = load float, float* [[TMP21]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = load float, float* [[TMP22]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = insertelement <2 x float> poison, float [[TMP27]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = insertelement <2 x float> [[TMP29]], float [[TMP28]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = fadd fast <2 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00>
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = fadd fast <2 x float> [[VEC_PHI1]], <float 1.000000e+00, float 1.000000e+00>
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = fadd fast <2 x float> [[TMP31]], [[TMP14]]
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = fadd fast <2 x float> [[TMP32]], [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[TMP35]] = fadd fast <2 x float> [[TMP33]], [[TMP26]]
+; UNROLL-NO-IC-NEXT:    [[TMP36]] = fadd fast <2 x float> [[TMP34]], [[TMP30]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP37]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = fadd fast <2 x float> [[TMP36]], [[TMP35]]
+; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float -0.000000e+00, <2 x float> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP38]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP45:%.*]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = load float, float* [[TMP39]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
+; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = load float, float* [[TMP41]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = fadd fast float [[S]], 1.000000e+00
+; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = fadd fast float [[TMP43]], [[TMP40]]
+; UNROLL-NO-IC-NEXT:    [[TMP45]] = fadd fast float [[TMP44]], [[TMP42]]
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 8
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    [[S_LCSSA:%.*]] = phi float [ [[TMP45]], [[FOR_BODY]] ], [ [[TMP38]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret float [[S_LCSSA]]
+;
+; INTERLEAVE-LABEL: @scalarize_induction_variable_02(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 64
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[TMP2]], 7
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i64 8, i64 [[N_MOD_VF]]
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = sub nsw i64 [[TMP2]], [[TMP4]]
+; INTERLEAVE-NEXT:    [[IND_END:%.*]] = shl i64 [[N_VEC]], 3
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = or i64 [[OFFSET_IDX]], 32
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[OFFSET_IDX]]
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = bitcast float* [[TMP6]] to <32 x float>*
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP7]] to <32 x float>*
+; INTERLEAVE-NEXT:    [[WIDE_VEC:%.*]] = load <32 x float>, <32 x float>* [[TMP8]], align 4
+; INTERLEAVE-NEXT:    [[WIDE_VEC2:%.*]] = load <32 x float>, <32 x float>* [[TMP9]], align 4
+; INTERLEAVE-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <32 x float> [[WIDE_VEC]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
+; INTERLEAVE-NEXT:    [[STRIDED_VEC3:%.*]] = shufflevector <32 x float> [[WIDE_VEC2]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[OFFSET_IDX]]
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP5]]
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = bitcast float* [[TMP10]] to <32 x float>*
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = bitcast float* [[TMP11]] to <32 x float>*
+; INTERLEAVE-NEXT:    [[WIDE_VEC4:%.*]] = load <32 x float>, <32 x float>* [[TMP12]], align 4
+; INTERLEAVE-NEXT:    [[WIDE_VEC5:%.*]] = load <32 x float>, <32 x float>* [[TMP13]], align 4
+; INTERLEAVE-NEXT:    [[STRIDED_VEC6:%.*]] = shufflevector <32 x float> [[WIDE_VEC4]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
+; INTERLEAVE-NEXT:    [[STRIDED_VEC7:%.*]] = shufflevector <32 x float> [[WIDE_VEC5]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = fadd fast <4 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = fadd fast <4 x float> [[VEC_PHI1]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = fadd fast <4 x float> [[TMP14]], [[STRIDED_VEC]]
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = fadd fast <4 x float> [[TMP15]], [[STRIDED_VEC3]]
+; INTERLEAVE-NEXT:    [[TMP18]] = fadd fast <4 x float> [[TMP16]], [[STRIDED_VEC6]]
+; INTERLEAVE-NEXT:    [[TMP19]] = fadd fast <4 x float> [[TMP17]], [[STRIDED_VEC7]]
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x float> [[TMP19]], [[TMP18]]
+; INTERLEAVE-NEXT:    [[TMP21:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float -0.000000e+00, <4 x float> [[BIN_RDX]])
+; INTERLEAVE-NEXT:    br label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; INTERLEAVE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP21]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = load float, float* [[TMP22]], align 4
+; INTERLEAVE-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
+; INTERLEAVE-NEXT:    [[TMP25:%.*]] = load float, float* [[TMP24]], align 4
+; INTERLEAVE-NEXT:    [[TMP26:%.*]] = fadd fast float [[S]], 1.000000e+00
+; INTERLEAVE-NEXT:    [[TMP27:%.*]] = fadd fast float [[TMP26]], [[TMP23]]
+; INTERLEAVE-NEXT:    [[TMP28]] = fadd fast float [[TMP27]], [[TMP25]]
+; INTERLEAVE-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 8
+; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP14:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    ret float [[TMP28]]
+;
 entry:
   br label %for.body
 
@@ -212,27 +1288,299 @@ for.end:
 ; for (int i = 0; i < n; ++i)
 ;   a[i].f ^= y;
 ;
-; INTERLEAVE-LABEL: @scalarize_induction_variable_03(
-; INTERLEAVE: vector.body:
-; INTERLEAVE:   %[[i0:.+]] = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; INTERLEAVE:   %[[i1:.+]] = or i64 %[[i0]], 1
-; INTERLEAVE:   %[[i2:.+]] = or i64 %[[i0]], 2
-; INTERLEAVE:   %[[i3:.+]] = or i64 %[[i0]], 3
-; INTERLEAVE:   %[[i4:.+]] = or i64 %[[i0]], 4
-; INTERLEAVE:   %[[i5:.+]] = or i64 %[[i0]], 5
-; INTERLEAVE:   %[[i6:.+]] = or i64 %[[i0]], 6
-; INTERLEAVE:   %[[i7:.+]] = or i64 %[[i0]], 7
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i0]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i1]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i2]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i3]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i4]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i5]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i6]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i7]], i32 1
 
 %pair.i32 = type { i32, i32 }
 define void @scalarize_induction_variable_03(%pair.i32 *%p, i32 %y, i64 %n) {
+; CHECK-LABEL: @scalarize_induction_variable_03(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[TMP0]], i32 1
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP1]], i32 1
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 8
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 8
+; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i32 0
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP5]], i32 1
+; CHECK-NEXT:    [[TMP8:%.*]] = xor <2 x i32> [[TMP7]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i32> [[TMP8]], i32 0
+; CHECK-NEXT:    store i32 [[TMP9]], i32* [[TMP2]], align 8
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x i32> [[TMP8]], i32 1
+; CHECK-NEXT:    store i32 [[TMP10]], i32* [[TMP3]], align 8
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; CHECK-NEXT:    [[TMP12:%.*]] = load i32, i32* [[F]], align 8
+; CHECK-NEXT:    [[TMP13:%.*]] = xor i32 [[TMP12]], [[Y]]
+; CHECK-NEXT:    store i32 [[TMP13]], i32* [[F]], align 8
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; CHECK-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @scalarize_induction_variable_03(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
+; IND-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i32 0
+; IND-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP0:%.*]] = or i64 [[INDEX]], 1
+; IND-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[INDEX]], i32 1
+; IND-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP0]], i32 1
+; IND-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 8
+; IND-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 8
+; IND-NEXT:    [[TMP5:%.*]] = insertelement <2 x i32> poison, i32 [[TMP3]], i32 0
+; IND-NEXT:    [[TMP6:%.*]] = insertelement <2 x i32> [[TMP5]], i32 [[TMP4]], i32 1
+; IND-NEXT:    [[TMP7:%.*]] = xor <2 x i32> [[TMP6]], [[BROADCAST_SPLAT]]
+; IND-NEXT:    [[TMP8:%.*]] = extractelement <2 x i32> [[TMP7]], i32 0
+; IND-NEXT:    store i32 [[TMP8]], i32* [[TMP1]], align 8
+; IND-NEXT:    [[TMP9:%.*]] = extractelement <2 x i32> [[TMP7]], i32 1
+; IND-NEXT:    store i32 [[TMP9]], i32* [[TMP2]], align 8
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; IND-NEXT:    [[TMP11:%.*]] = load i32, i32* [[F]], align 8
+; IND-NEXT:    [[TMP12:%.*]] = xor i32 [[TMP11]], [[Y]]
+; IND-NEXT:    store i32 [[TMP12]], i32* [[F]], align 8
+; IND-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; IND-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; IND-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @scalarize_induction_variable_03(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <2 x i32> poison, i32 [[Y]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT1]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP0:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP1:%.*]] = or i64 [[INDEX]], 2
+; UNROLL-NEXT:    [[TMP2:%.*]] = or i64 [[INDEX]], 3
+; UNROLL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[INDEX]], i32 1
+; UNROLL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP0]], i32 1
+; UNROLL-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP1]], i32 1
+; UNROLL-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP2]], i32 1
+; UNROLL-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP3]], align 8
+; UNROLL-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP4]], align 8
+; UNROLL-NEXT:    [[TMP9:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i32 0
+; UNROLL-NEXT:    [[TMP10:%.*]] = insertelement <2 x i32> [[TMP9]], i32 [[TMP8]], i32 1
+; UNROLL-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP5]], align 8
+; UNROLL-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP6]], align 8
+; UNROLL-NEXT:    [[TMP13:%.*]] = insertelement <2 x i32> poison, i32 [[TMP11]], i32 0
+; UNROLL-NEXT:    [[TMP14:%.*]] = insertelement <2 x i32> [[TMP13]], i32 [[TMP12]], i32 1
+; UNROLL-NEXT:    [[TMP15:%.*]] = xor <2 x i32> [[TMP10]], [[BROADCAST_SPLAT]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = xor <2 x i32> [[TMP14]], [[BROADCAST_SPLAT2]]
+; UNROLL-NEXT:    [[TMP17:%.*]] = extractelement <2 x i32> [[TMP15]], i32 0
+; UNROLL-NEXT:    store i32 [[TMP17]], i32* [[TMP3]], align 8
+; UNROLL-NEXT:    [[TMP18:%.*]] = extractelement <2 x i32> [[TMP15]], i32 1
+; UNROLL-NEXT:    store i32 [[TMP18]], i32* [[TMP4]], align 8
+; UNROLL-NEXT:    [[TMP19:%.*]] = extractelement <2 x i32> [[TMP16]], i32 0
+; UNROLL-NEXT:    store i32 [[TMP19]], i32* [[TMP5]], align 8
+; UNROLL-NEXT:    [[TMP20:%.*]] = extractelement <2 x i32> [[TMP16]], i32 1
+; UNROLL-NEXT:    store i32 [[TMP20]], i32* [[TMP6]], align 8
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; UNROLL-NEXT:    [[TMP22:%.*]] = load i32, i32* [[F]], align 8
+; UNROLL-NEXT:    [[TMP23:%.*]] = xor i32 [[TMP22]], [[Y]]
+; UNROLL-NEXT:    store i32 [[TMP23]], i32* [[F]], align 8
+; UNROLL-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_03(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <2 x i32> poison, i32 [[Y]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT1]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[TMP0]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP1]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP2]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP3]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP4]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP5]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = insertelement <2 x i32> poison, i32 [[TMP8]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = insertelement <2 x i32> [[TMP10]], i32 [[TMP9]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP6]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP7]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = insertelement <2 x i32> poison, i32 [[TMP12]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = insertelement <2 x i32> [[TMP14]], i32 [[TMP13]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = xor <2 x i32> [[TMP11]], [[BROADCAST_SPLAT]]
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = xor <2 x i32> [[TMP15]], [[BROADCAST_SPLAT2]]
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = extractelement <2 x i32> [[TMP16]], i32 0
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP18]], i32* [[TMP4]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = extractelement <2 x i32> [[TMP16]], i32 1
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP19]], i32* [[TMP5]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = extractelement <2 x i32> [[TMP17]], i32 0
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP20]], i32* [[TMP6]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = extractelement <2 x i32> [[TMP17]], i32 1
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP21]], i32* [[TMP7]], align 8
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = load i32, i32* [[F]], align 8
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = xor i32 [[TMP23]], [[Y]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP24]], i32* [[F]], align 8
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @scalarize_induction_variable_03(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 9
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[SMAX]], 7
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = select i1 [[TMP0]], i64 8, i64 [[N_MOD_VF]]
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = sub nsw i64 [[SMAX]], [[TMP1]]
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y:%.*]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT3]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = or i64 [[INDEX]], 1
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 2
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = or i64 [[INDEX]], 3
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = or i64 [[INDEX]], 4
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = or i64 [[INDEX]], 5
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = or i64 [[INDEX]], 6
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = or i64 [[INDEX]], 7
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[INDEX]], i32 1
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP2]], i32 1
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP3]], i32 1
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP4]], i32 1
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]], i32 1
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP6]], i32 1
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP7]], i32 1
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP8]], i32 1
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP9]] to <8 x i32>*
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = bitcast i32* [[TMP13]] to <8 x i32>*
+; INTERLEAVE-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i32>, <8 x i32>* [[TMP17]], align 8
+; INTERLEAVE-NEXT:    [[WIDE_VEC1:%.*]] = load <8 x i32>, <8 x i32>* [[TMP18]], align 8
+; INTERLEAVE-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; INTERLEAVE-NEXT:    [[STRIDED_VEC2:%.*]] = shufflevector <8 x i32> [[WIDE_VEC1]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = xor <4 x i32> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]
+; INTERLEAVE-NEXT:    [[TMP20:%.*]] = xor <4 x i32> [[STRIDED_VEC2]], [[BROADCAST_SPLAT4]]
+; INTERLEAVE-NEXT:    [[TMP21:%.*]] = extractelement <4 x i32> [[TMP19]], i32 0
+; INTERLEAVE-NEXT:    store i32 [[TMP21]], i32* [[TMP9]], align 8
+; INTERLEAVE-NEXT:    [[TMP22:%.*]] = extractelement <4 x i32> [[TMP19]], i32 1
+; INTERLEAVE-NEXT:    store i32 [[TMP22]], i32* [[TMP10]], align 8
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = extractelement <4 x i32> [[TMP19]], i32 2
+; INTERLEAVE-NEXT:    store i32 [[TMP23]], i32* [[TMP11]], align 8
+; INTERLEAVE-NEXT:    [[TMP24:%.*]] = extractelement <4 x i32> [[TMP19]], i32 3
+; INTERLEAVE-NEXT:    store i32 [[TMP24]], i32* [[TMP12]], align 8
+; INTERLEAVE-NEXT:    [[TMP25:%.*]] = extractelement <4 x i32> [[TMP20]], i32 0
+; INTERLEAVE-NEXT:    store i32 [[TMP25]], i32* [[TMP13]], align 8
+; INTERLEAVE-NEXT:    [[TMP26:%.*]] = extractelement <4 x i32> [[TMP20]], i32 1
+; INTERLEAVE-NEXT:    store i32 [[TMP26]], i32* [[TMP14]], align 8
+; INTERLEAVE-NEXT:    [[TMP27:%.*]] = extractelement <4 x i32> [[TMP20]], i32 2
+; INTERLEAVE-NEXT:    store i32 [[TMP27]], i32* [[TMP15]], align 8
+; INTERLEAVE-NEXT:    [[TMP28:%.*]] = extractelement <4 x i32> [[TMP20]], i32 3
+; INTERLEAVE-NEXT:    store i32 [[TMP28]], i32* [[TMP16]], align 8
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    br label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; INTERLEAVE-NEXT:    [[TMP30:%.*]] = load i32, i32* [[F]], align 8
+; INTERLEAVE-NEXT:    [[TMP31:%.*]] = xor i32 [[TMP30]], [[Y]]
+; INTERLEAVE-NEXT:    store i32 [[TMP31]], i32* [[F]], align 8
+; INTERLEAVE-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP16:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    ret void
+;
 entry:
   br label %for.body
 
@@ -256,26 +1604,385 @@ for.end:
 ; for (int i = 0; i < n; ++i)
 ;   p[i].f = a[i * 4]
 ;
-; INTERLEAVE-LABEL: @scalarize_induction_variable_04(
-; INTERLEAVE: vector.body:
-; INTERLEAVE:   %[[i0:.+]] = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; INTERLEAVE:   %[[i1:.+]] = or i64 %[[i0]], 1
-; INTERLEAVE:   %[[i2:.+]] = or i64 %[[i0]], 2
-; INTERLEAVE:   %[[i3:.+]] = or i64 %[[i0]], 3
-; INTERLEAVE:   %[[i4:.+]] = or i64 %[[i0]], 4
-; INTERLEAVE:   %[[i5:.+]] = or i64 %[[i0]], 5
-; INTERLEAVE:   %[[i6:.+]] = or i64 %[[i0]], 6
-; INTERLEAVE:   %[[i7:.+]] = or i64 %[[i0]], 7
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i0]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i1]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i2]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i3]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i4]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i5]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i6]], i32 1
-; INTERLEAVE:   getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %[[i7]], i32 1
 
 define void @scalarize_induction_variable_04(i32* %a, %pair.i32* %p, i32 %n) {
+; CHECK-LABEL: @scalarize_induction_variable_04(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A4:%.*]] = bitcast i32* [[A:%.*]] to i8*
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK:       vector.memcheck:
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
+; CHECK-NEXT:    [[SCEVGEP1:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; CHECK-NEXT:    [[SCEVGEP2:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]]
+; CHECK-NEXT:    [[SCEVGEP23:%.*]] = bitcast %pair.i32* [[SCEVGEP2]] to i8*
+; CHECK-NEXT:    [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
+; CHECK-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[TMP6]], 1
+; CHECK-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP7]]
+; CHECK-NEXT:    [[SCEVGEP56:%.*]] = bitcast i32* [[SCEVGEP5]] to i8*
+; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP56]]
+; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[A4]], [[SCEVGEP23]]
+; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP10:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <2 x i64> [[TMP10]], i32 0
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP11]]
+; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <2 x i64> [[TMP10]], i32 1
+; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP13]]
+; CHECK-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 1, !alias.scope !17
+; CHECK-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP14]], align 1, !alias.scope !17
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP8]], i32 1
+; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP9]], i32 1
+; CHECK-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 1, !alias.scope !20, !noalias !17
+; CHECK-NEXT:    store i32 [[TMP16]], i32* [[TMP18]], align 1, !alias.scope !20, !noalias !17
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[TMP20:%.*]] = shl nsw i64 [[I]], 2
+; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
+; CHECK-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 1
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; CHECK-NEXT:    store i32 [[TMP22]], i32* [[TMP23]], align 1
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; CHECK-NEXT:    [[TMP24:%.*]] = trunc i64 [[I_NEXT]] to i32
+; CHECK-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP24]], [[N]]
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @scalarize_induction_variable_04(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; IND-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; IND-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; IND:       vector.memcheck:
+; IND-NEXT:    [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
+; IND-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; IND-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; IND-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; IND-NEXT:    [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
+; IND-NEXT:    [[TMP7:%.*]] = or i64 [[TMP6]], 1
+; IND-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A:%.*]], i64 [[TMP7]]
+; IND-NEXT:    [[BOUND0:%.*]] = icmp ult i32* [[SCEVGEP]], [[SCEVGEP5]]
+; IND-NEXT:    [[TMP8:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]], i32 0
+; IND-NEXT:    [[BOUND1:%.*]] = icmp ugt i32* [[TMP8]], [[A]]
+; IND-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; IND-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP9:%.*]] = or i64 [[INDEX]], 1
+; IND-NEXT:    [[TMP10:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; IND-NEXT:    [[TMP11:%.*]] = extractelement <2 x i64> [[TMP10]], i32 0
+; IND-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP11]]
+; IND-NEXT:    [[TMP13:%.*]] = extractelement <2 x i64> [[TMP10]], i32 1
+; IND-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP13]]
+; IND-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 1, !alias.scope !17
+; IND-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP14]], align 1, !alias.scope !17
+; IND-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[INDEX]], i32 1
+; IND-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP9]], i32 1
+; IND-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 1, !alias.scope !20, !noalias !17
+; IND-NEXT:    store i32 [[TMP16]], i32* [[TMP18]], align 1, !alias.scope !20, !noalias !17
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; IND-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[TMP20:%.*]] = shl nsw i64 [[I]], 2
+; IND-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
+; IND-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 1
+; IND-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; IND-NEXT:    store i32 [[TMP22]], i32* [[TMP23]], align 1
+; IND-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; IND-NEXT:    [[TMP24:%.*]] = trunc i64 [[I_NEXT]] to i32
+; IND-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP24]], [[N]]
+; IND-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @scalarize_induction_variable_04(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL:       vector.memcheck:
+; UNROLL-NEXT:    [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
+; UNROLL-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; UNROLL-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; UNROLL-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; UNROLL-NEXT:    [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
+; UNROLL-NEXT:    [[TMP7:%.*]] = or i64 [[TMP6]], 1
+; UNROLL-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A:%.*]], i64 [[TMP7]]
+; UNROLL-NEXT:    [[BOUND0:%.*]] = icmp ult i32* [[SCEVGEP]], [[SCEVGEP5]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]], i32 0
+; UNROLL-NEXT:    [[BOUND1:%.*]] = icmp ugt i32* [[TMP8]], [[A]]
+; UNROLL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP9:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP10:%.*]] = or i64 [[INDEX]], 2
+; UNROLL-NEXT:    [[TMP11:%.*]] = or i64 [[INDEX]], 3
+; UNROLL-NEXT:    [[TMP12:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; UNROLL-NEXT:    [[STEP_ADD:%.*]] = shl <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; UNROLL-NEXT:    [[TMP13:%.*]] = add <2 x i64> [[STEP_ADD]], <i64 8, i64 8>
+; UNROLL-NEXT:    [[TMP14:%.*]] = extractelement <2 x i64> [[TMP12]], i32 0
+; UNROLL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP14]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = extractelement <2 x i64> [[TMP12]], i32 1
+; UNROLL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP16]]
+; UNROLL-NEXT:    [[TMP18:%.*]] = extractelement <2 x i64> [[TMP13]], i32 0
+; UNROLL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
+; UNROLL-NEXT:    [[TMP20:%.*]] = extractelement <2 x i64> [[TMP13]], i32 1
+; UNROLL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
+; UNROLL-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP15]], align 1, !alias.scope !17
+; UNROLL-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP17]], align 1, !alias.scope !17
+; UNROLL-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP19]], align 1, !alias.scope !17
+; UNROLL-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP21]], align 1, !alias.scope !17
+; UNROLL-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[INDEX]], i32 1
+; UNROLL-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP9]], i32 1
+; UNROLL-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP10]], i32 1
+; UNROLL-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP11]], i32 1
+; UNROLL-NEXT:    store i32 [[TMP22]], i32* [[TMP26]], align 1, !alias.scope !20, !noalias !17
+; UNROLL-NEXT:    store i32 [[TMP23]], i32* [[TMP27]], align 1, !alias.scope !20, !noalias !17
+; UNROLL-NEXT:    store i32 [[TMP24]], i32* [[TMP28]], align 1, !alias.scope !20, !noalias !17
+; UNROLL-NEXT:    store i32 [[TMP25]], i32* [[TMP29]], align 1, !alias.scope !20, !noalias !17
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 4, i64 4>
+; UNROLL-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[TMP31:%.*]] = shl nsw i64 [[I]], 2
+; UNROLL-NEXT:    [[TMP32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP31]]
+; UNROLL-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 1
+; UNROLL-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; UNROLL-NEXT:    store i32 [[TMP33]], i32* [[TMP34]], align 1
+; UNROLL-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NEXT:    [[TMP35:%.*]] = trunc i64 [[I_NEXT]] to i32
+; UNROLL-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP35]], [[N]]
+; UNROLL-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_04(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[A4:%.*]] = bitcast i32* [[A:%.*]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; UNROLL-NO-IC:       vector.memcheck:
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP1:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP2:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP23:%.*]] = bitcast %pair.i32* [[SCEVGEP2]] to i8*
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[TMP6]], 1
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[SCEVGEP56:%.*]] = bitcast i32* [[SCEVGEP5]] to i8*
+; UNROLL-NO-IC-NEXT:    [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP56]]
+; UNROLL-NO-IC-NEXT:    [[BOUND1:%.*]] = icmp ult i8* [[A4]], [[SCEVGEP23]]
+; UNROLL-NO-IC-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; UNROLL-NO-IC-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = shl nsw <2 x i64> [[STEP_ADD]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = extractelement <2 x i64> [[TMP12]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP14]]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = extractelement <2 x i64> [[TMP12]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP16]]
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = extractelement <2 x i64> [[TMP13]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = extractelement <2 x i64> [[TMP13]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP15]], align 1, !alias.scope !17
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP17]], align 1, !alias.scope !17
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP19]], align 1, !alias.scope !17
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP21]], align 1, !alias.scope !17
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP8]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP9]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP10]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP11]], i32 1
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP22]], i32* [[TMP26]], align 1, !alias.scope !20, !noalias !17
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP23]], i32* [[TMP27]], align 1, !alias.scope !20, !noalias !17
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP24]], i32* [[TMP28]], align 1, !alias.scope !20, !noalias !17
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP25]], i32* [[TMP29]], align 1, !alias.scope !20, !noalias !17
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = shl nsw i64 [[I]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP31]]
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 1
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; UNROLL-NO-IC-NEXT:    store i32 [[TMP33]], i32* [[TMP34]], align 1
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = trunc i64 [[I_NEXT]] to i32
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP35]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @scalarize_induction_variable_04(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
+; INTERLEAVE:       vector.memcheck:
+; INTERLEAVE-NEXT:    [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = add i32 [[N]], -1
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = or i64 [[TMP6]], 1
+; INTERLEAVE-NEXT:    [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A:%.*]], i64 [[TMP7]]
+; INTERLEAVE-NEXT:    [[BOUND0:%.*]] = icmp ult i32* [[SCEVGEP]], [[SCEVGEP5]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]], i32 0
+; INTERLEAVE-NEXT:    [[BOUND1:%.*]] = icmp ugt i32* [[TMP8]], [[A]]
+; INTERLEAVE-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; INTERLEAVE-NEXT:    br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[TMP2]], 7
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i64 8, i64 [[N_MOD_VF]]
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = sub nsw i64 [[TMP2]], [[TMP10]]
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = or i64 [[INDEX]], 1
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = or i64 [[INDEX]], 2
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = or i64 [[INDEX]], 3
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = or i64 [[INDEX]], 4
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = or i64 [[INDEX]], 5
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = or i64 [[INDEX]], 6
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = or i64 [[INDEX]], 7
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = shl nsw i64 [[INDEX]], 2
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = shl nsw i64 [[TMP14]], 2
+; INTERLEAVE-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
+; INTERLEAVE-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP19]]
+; INTERLEAVE-NEXT:    [[TMP22:%.*]] = bitcast i32* [[TMP20]] to <16 x i32>*
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = bitcast i32* [[TMP21]] to <16 x i32>*
+; INTERLEAVE-NEXT:    [[WIDE_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[TMP22]], align 1
+; INTERLEAVE-NEXT:    [[WIDE_VEC7:%.*]] = load <16 x i32>, <16 x i32>* [[TMP23]], align 1
+; INTERLEAVE-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[INDEX]], i32 1
+; INTERLEAVE-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP11]], i32 1
+; INTERLEAVE-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP12]], i32 1
+; INTERLEAVE-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP13]], i32 1
+; INTERLEAVE-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP14]], i32 1
+; INTERLEAVE-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP15]], i32 1
+; INTERLEAVE-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP16]], i32 1
+; INTERLEAVE-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP17]], i32 1
+; INTERLEAVE-NEXT:    [[TMP32:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i32 0
+; INTERLEAVE-NEXT:    store i32 [[TMP32]], i32* [[TMP24]], align 1, !alias.scope !17, !noalias !20
+; INTERLEAVE-NEXT:    [[TMP33:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i32 4
+; INTERLEAVE-NEXT:    store i32 [[TMP33]], i32* [[TMP25]], align 1, !alias.scope !17, !noalias !20
+; INTERLEAVE-NEXT:    [[TMP34:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i32 8
+; INTERLEAVE-NEXT:    store i32 [[TMP34]], i32* [[TMP26]], align 1, !alias.scope !17, !noalias !20
+; INTERLEAVE-NEXT:    [[TMP35:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i32 12
+; INTERLEAVE-NEXT:    store i32 [[TMP35]], i32* [[TMP27]], align 1, !alias.scope !17, !noalias !20
+; INTERLEAVE-NEXT:    [[TMP36:%.*]] = extractelement <16 x i32> [[WIDE_VEC7]], i32 0
+; INTERLEAVE-NEXT:    store i32 [[TMP36]], i32* [[TMP28]], align 1, !alias.scope !17, !noalias !20
+; INTERLEAVE-NEXT:    [[TMP37:%.*]] = extractelement <16 x i32> [[WIDE_VEC7]], i32 4
+; INTERLEAVE-NEXT:    store i32 [[TMP37]], i32* [[TMP29]], align 1, !alias.scope !17, !noalias !20
+; INTERLEAVE-NEXT:    [[TMP38:%.*]] = extractelement <16 x i32> [[WIDE_VEC7]], i32 8
+; INTERLEAVE-NEXT:    store i32 [[TMP38]], i32* [[TMP30]], align 1, !alias.scope !17, !noalias !20
+; INTERLEAVE-NEXT:    [[TMP39:%.*]] = extractelement <16 x i32> [[WIDE_VEC7]], i32 12
+; INTERLEAVE-NEXT:    store i32 [[TMP39]], i32* [[TMP31]], align 1, !alias.scope !17, !noalias !20
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP40:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP40]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    br label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[TMP41:%.*]] = shl nsw i64 [[I]], 2
+; INTERLEAVE-NEXT:    [[TMP42:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP41]]
+; INTERLEAVE-NEXT:    [[TMP43:%.*]] = load i32, i32* [[TMP42]], align 1
+; INTERLEAVE-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
+; INTERLEAVE-NEXT:    store i32 [[TMP43]], i32* [[TMP44]], align 1
+; INTERLEAVE-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; INTERLEAVE-NEXT:    [[TMP45:%.*]] = trunc i64 [[I_NEXT]] to i32
+; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP45]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    ret void
+;
 entry:
   br label %for.body
 
@@ -307,89 +2014,509 @@ for.end:
 ;   sum += x;
 ; }
 ;
+;
+;
+;
+
+define i32 @scalarize_induction_variable_05(i32* %a, i32 %x, i1 %c, i32 %n) {
 ; CHECK-LABEL: @scalarize_induction_variable_05(
-; CHECK: vector.body:
-; CHECK:   %index = phi i32 [ 0, %vector.ph ], [ %index.next, %pred.udiv.continue{{[0-9]+}} ]
-; CHECK:   %[[I0:.+]] = add i32 %index, 0
-; CHECK:   getelementptr inbounds i32, i32* %a, i32 %[[I0]]
-; CHECK: pred.udiv.if:
-; CHECK:   udiv i32 {{.*}}, %[[I0]]
-; CHECK: pred.udiv.if{{[0-9]+}}:
-; CHECK:   %[[I1:.+]] = add i32 %index, 1
-; CHECK:   udiv i32 {{.*}}, %[[I1]]
-;
-; UNROLL-NO_IC-LABEL: @scalarize_induction_variable_05(
-; UNROLL-NO-IC: vector.body:
-; UNROLL-NO-IC:   %index = phi i32 [ 0, %vector.ph ], [ %index.next, %pred.udiv.continue{{[0-9]+}} ]
-; UNROLL-NO-IC:   %[[I0:.+]] = add i32 %index, 0
-; UNROLL-NO-IC:   %[[I2:.+]] = add i32 %index, 2
-; UNROLL-NO-IC:   getelementptr inbounds i32, i32* %a, i32 %[[I0]]
-; UNROLL-NO-IC:   getelementptr inbounds i32, i32* %a, i32 %[[I2]]
-; UNROLL-NO-IC: pred.udiv.if:
-; UNROLL-NO-IC:   udiv i32 {{.*}}, %[[I0]]
-; UNROLL-NO-IC: pred.udiv.if{{[0-9]+}}:
-; UNROLL-NO-IC:   %[[I1:.+]] = add i32 %index, 1
-; UNROLL-NO-IC:   udiv i32 {{.*}}, %[[I1]]
-; UNROLL-NO-IC: pred.udiv.if{{[0-9]+}}:
-; UNROLL-NO-IC:   udiv i32 {{.*}}, %[[I2]]
-; UNROLL-NO-IC: pred.udiv.if{{[0-9]+}}:
-; UNROLL-NO-IC:   %[[I3:.+]] = add i32 %index, 3
-; UNROLL-NO-IC:   udiv i32 {{.*}}, %[[I3]]
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[SMAX]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[SMAX]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT]], <2 x i1> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UDIV_CONTINUE2]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[PRED_UDIV_CONTINUE2]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP3]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 0
+; CHECK-NEXT:    br i1 [[TMP4]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; CHECK:       pred.udiv.if:
+; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
+; CHECK-NEXT:    [[TMP6:%.*]] = udiv i32 [[TMP5]], [[TMP0]]
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP6]], i32 0
+; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; CHECK:       pred.udiv.continue:
+; CHECK-NEXT:    [[TMP8:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UDIV_IF]] ]
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 1
+; CHECK-NEXT:    br i1 [[TMP9]], label [[PRED_UDIV_IF1:%.*]], label [[PRED_UDIV_CONTINUE2]]
+; CHECK:       pred.udiv.if1:
+; CHECK-NEXT:    [[TMP10:%.*]] = add i32 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1
+; CHECK-NEXT:    [[TMP12:%.*]] = udiv i32 [[TMP11]], [[TMP10]]
+; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <2 x i32> [[TMP8]], i32 [[TMP12]], i32 1
+; CHECK-NEXT:    br label [[PRED_UDIV_CONTINUE2]]
+; CHECK:       pred.udiv.continue2:
+; CHECK-NEXT:    [[TMP14:%.*]] = phi <2 x i32> [ [[TMP8]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF1]] ]
+; CHECK-NEXT:    [[TMP15:%.*]] = xor <2 x i1> [[BROADCAST_SPLAT]], <i1 true, i1 true>
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP15]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP14]]
+; CHECK-NEXT:    [[TMP16]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP18:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP16]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
+; CHECK-NEXT:    [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
+; CHECK-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[I]]
+; CHECK-NEXT:    [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
+; CHECK-NEXT:    br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
+; CHECK-NEXT:    br label [[IF_END]]
+; CHECK:       if.end:
+; CHECK-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[VAR4]] = add i32 [[VAR3]], [[SUM]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i32 [[I]], 1
+; CHECK-NEXT:    [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[VAR5]]
 ;
 ; IND-LABEL: @scalarize_induction_variable_05(
-; IND: vector.body:
-; IND:   %index = phi i32 [ 0, %vector.ph ], [ %index.next, %pred.udiv.continue{{[0-9]+}} ]
-; IND:   %[[E0:.+]] = sext i32 %index to i64
-; IND:   getelementptr inbounds i32, i32* %a, i64 %[[E0]]
-; IND: pred.udiv.if:
-; IND:   udiv i32 {{.*}}, %index
-; IND: pred.udiv.if{{[0-9]+}}:
-; IND:   %[[I1:.+]] = or i32 %index, 1
-; IND:   udiv i32 {{.*}}, %[[I1]]
+; IND-NEXT:  entry:
+; IND-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i32 [[SMAX]], 2147483646
+; IND-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i32 0
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ]
+; IND-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[PRED_UDIV_CONTINUE2]] ]
+; IND-NEXT:    [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
+; IND-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
+; IND-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <2 x i32>*
+; IND-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP2]], align 4
+; IND-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; IND:       pred.udiv.if:
+; IND-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
+; IND-NEXT:    [[TMP4:%.*]] = udiv i32 [[TMP3]], [[INDEX]]
+; IND-NEXT:    [[TMP5:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i32 0
+; IND-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; IND:       pred.udiv.continue:
+; IND-NEXT:    [[TMP6:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP5]], [[PRED_UDIV_IF]] ]
+; IND-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF1:%.*]], label [[PRED_UDIV_CONTINUE2]]
+; IND:       pred.udiv.if1:
+; IND-NEXT:    [[TMP7:%.*]] = or i32 [[INDEX]], 1
+; IND-NEXT:    [[TMP8:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1
+; IND-NEXT:    [[TMP9:%.*]] = udiv i32 [[TMP8]], [[TMP7]]
+; IND-NEXT:    [[TMP10:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP9]], i32 1
+; IND-NEXT:    br label [[PRED_UDIV_CONTINUE2]]
+; IND:       pred.udiv.continue2:
+; IND-NEXT:    [[TMP11:%.*]] = phi <2 x i32> [ [[TMP6]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP10]], [[PRED_UDIV_IF1]] ]
+; IND-NEXT:    [[TMP12:%.*]] = xor <2 x i1> [[BROADCAST_SPLATINSERT]], <i1 true, i1 poison>
+; IND-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i1> [[TMP12]], <2 x i1> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP13]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP11]]
+; IND-NEXT:    [[TMP14]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP14]])
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; IND-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP16]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
+; IND-NEXT:    [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
+; IND-NEXT:    [[TMP17:%.*]] = zext i32 [[I]] to i64
+; IND-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP17]]
+; IND-NEXT:    [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
+; IND-NEXT:    br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
+; IND:       if.then:
+; IND-NEXT:    [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
+; IND-NEXT:    br label [[IF_END]]
+; IND:       if.end:
+; IND-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
+; IND-NEXT:    [[VAR4]] = add i32 [[VAR3]], [[SUM]]
+; IND-NEXT:    [[I_NEXT]] = add nuw nsw i32 [[I]], 1
+; IND-NEXT:    [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; IND-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP16]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    ret i32 [[VAR5]]
 ;
 ; UNROLL-LABEL: @scalarize_induction_variable_05(
-; UNROLL: vector.body:
-; UNROLL:   %index = phi i32 [ 0, %vector.ph ], [ %index.next, %pred.udiv.continue{{[0-9]+}} ]
-; UNROLL:   %[[I2:.+]] = or i32 %index, 2
-; UNROLL:   %[[E0:.+]] = sext i32 %index to i64
-; UNROLL:   %[[G0:.+]] = getelementptr inbounds i32, i32* %a, i64 %[[E0]]
-; UNROLL:   getelementptr inbounds i32, i32* %[[G0]], i64 2
-; UNROLL: pred.udiv.if:
-; UNROLL:   udiv i32 {{.*}}, %index
-; UNROLL: pred.udiv.if{{[0-9]+}}:
-; UNROLL:   %[[I1:.+]] = or i32 %index, 1
-; UNROLL:   udiv i32 {{.*}}, %[[I1]]
-; UNROLL: pred.udiv.if{{[0-9]+}}:
-; UNROLL:   udiv i32 {{.*}}, %[[I2]]
-; UNROLL: pred.udiv.if{{[0-9]+}}:
-; UNROLL:   %[[I3:.+]] = or i32 %index, 3
-; UNROLL:   udiv i32 {{.*}}, %[[I3]]
-
-define i32 @scalarize_induction_variable_05(i32* %a, i32 %x, i1 %c, i32 %n) {
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i32 [[SMAX]], 2147483644
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT6:%.*]] = insertelement <2 x i1> poison, i1 [[C]], i32 0
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE11:%.*]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[PRED_UDIV_CONTINUE11]] ]
+; UNROLL-NEXT:    [[VEC_PHI2:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[PRED_UDIV_CONTINUE11]] ]
+; UNROLL-NEXT:    [[TMP0:%.*]] = or i32 [[INDEX]], 2
+; UNROLL-NEXT:    [[TMP1:%.*]] = sext i32 [[INDEX]] to i64
+; UNROLL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
+; UNROLL-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
+; UNROLL-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP3]], align 4
+; UNROLL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 2
+; UNROLL-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
+; UNROLL-NEXT:    [[WIDE_LOAD3:%.*]] = load <2 x i32>, <2 x i32>* [[TMP5]], align 4
+; UNROLL-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; UNROLL:       pred.udiv.if:
+; UNROLL-NEXT:    [[TMP6:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
+; UNROLL-NEXT:    [[TMP7:%.*]] = udiv i32 [[TMP6]], [[INDEX]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i32 0
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; UNROLL:       pred.udiv.continue:
+; UNROLL-NEXT:    [[TMP9:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_UDIV_IF]] ]
+; UNROLL-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
+; UNROLL:       pred.udiv.if4:
+; UNROLL-NEXT:    [[TMP10:%.*]] = or i32 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP11:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1
+; UNROLL-NEXT:    [[TMP12:%.*]] = udiv i32 [[TMP11]], [[TMP10]]
+; UNROLL-NEXT:    [[TMP13:%.*]] = insertelement <2 x i32> [[TMP9]], i32 [[TMP12]], i32 1
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
+; UNROLL:       pred.udiv.continue5:
+; UNROLL-NEXT:    [[TMP14:%.*]] = phi <2 x i32> [ [[TMP9]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF4]] ]
+; UNROLL-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; UNROLL:       pred.udiv.if8:
+; UNROLL-NEXT:    [[TMP15:%.*]] = extractelement <2 x i32> [[WIDE_LOAD3]], i32 0
+; UNROLL-NEXT:    [[TMP16:%.*]] = udiv i32 [[TMP15]], [[TMP0]]
+; UNROLL-NEXT:    [[TMP17:%.*]] = insertelement <2 x i32> poison, i32 [[TMP16]], i32 0
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; UNROLL:       pred.udiv.continue9:
+; UNROLL-NEXT:    [[TMP18:%.*]] = phi <2 x i32> [ poison, [[PRED_UDIV_CONTINUE5]] ], [ [[TMP17]], [[PRED_UDIV_IF8]] ]
+; UNROLL-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF10:%.*]], label [[PRED_UDIV_CONTINUE11]]
+; UNROLL:       pred.udiv.if10:
+; UNROLL-NEXT:    [[TMP19:%.*]] = or i32 [[INDEX]], 3
+; UNROLL-NEXT:    [[TMP20:%.*]] = extractelement <2 x i32> [[WIDE_LOAD3]], i32 1
+; UNROLL-NEXT:    [[TMP21:%.*]] = udiv i32 [[TMP20]], [[TMP19]]
+; UNROLL-NEXT:    [[TMP22:%.*]] = insertelement <2 x i32> [[TMP18]], i32 [[TMP21]], i32 1
+; UNROLL-NEXT:    br label [[PRED_UDIV_CONTINUE11]]
+; UNROLL:       pred.udiv.continue11:
+; UNROLL-NEXT:    [[TMP23:%.*]] = phi <2 x i32> [ [[TMP18]], [[PRED_UDIV_CONTINUE9]] ], [ [[TMP22]], [[PRED_UDIV_IF10]] ]
+; UNROLL-NEXT:    [[TMP24:%.*]] = xor <2 x i1> [[BROADCAST_SPLATINSERT]], <i1 true, i1 poison>
+; UNROLL-NEXT:    [[TMP25:%.*]] = shufflevector <2 x i1> [[TMP24]], <2 x i1> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[TMP26:%.*]] = xor <2 x i1> [[BROADCAST_SPLATINSERT6]], <i1 true, i1 poison>
+; UNROLL-NEXT:    [[TMP27:%.*]] = shufflevector <2 x i1> [[TMP26]], <2 x i1> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP25]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP14]]
+; UNROLL-NEXT:    [[PREDPHI12:%.*]] = select <2 x i1> [[TMP27]], <2 x i32> [[WIDE_LOAD3]], <2 x i32> [[TMP23]]
+; UNROLL-NEXT:    [[TMP28]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]
+; UNROLL-NEXT:    [[TMP29]] = add <2 x i32> [[PREDPHI12]], [[VEC_PHI2]]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = add <2 x i32> [[TMP29]], [[TMP28]]
+; UNROLL-NEXT:    [[TMP31:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP31]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
+; UNROLL-NEXT:    [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
+; UNROLL-NEXT:    [[TMP32:%.*]] = zext i32 [[I]] to i64
+; UNROLL-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP32]]
+; UNROLL-NEXT:    [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
+; UNROLL-NEXT:    br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
+; UNROLL:       if.then:
+; UNROLL-NEXT:    [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
+; UNROLL-NEXT:    br label [[IF_END]]
+; UNROLL:       if.end:
+; UNROLL-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
+; UNROLL-NEXT:    [[VAR4]] = add i32 [[VAR3]], [[SUM]]
+; UNROLL-NEXT:    [[I_NEXT]] = add nuw nsw i32 [[I]], 1
+; UNROLL-NEXT:    [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i32 [[VAR5]]
+;
+; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_05(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[SMAX]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[SMAX]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT]], <2 x i1> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT6:%.*]] = insertelement <2 x i1> poison, i1 [[C]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT7:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT6]], <2 x i1> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE11:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UDIV_CONTINUE11]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP32:%.*]], [[PRED_UDIV_CONTINUE11]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI2:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP33:%.*]], [[PRED_UDIV_CONTINUE11]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP5]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD3:%.*]] = load <2 x i32>, <2 x i32>* [[TMP7]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP8]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if:
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = udiv i32 [[TMP9]], [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = insertelement <2 x i32> poison, i32 [[TMP10]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; UNROLL-NO-IC:       pred.udiv.continue:
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP11]], [[PRED_UDIV_IF]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP13]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if4:
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = add i32 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = udiv i32 [[TMP15]], [[TMP14]]
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = insertelement <2 x i32> [[TMP12]], i32 [[TMP16]], i32 1
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
+; UNROLL-NO-IC:       pred.udiv.continue5:
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = phi <2 x i32> [ [[TMP12]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP17]], [[PRED_UDIV_IF4]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT7]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP19]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; UNROLL-NO-IC:       pred.udiv.if8:
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = extractelement <2 x i32> [[WIDE_LOAD3]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = udiv i32 [[TMP20]], [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = insertelement <2 x i32> poison, i32 [[TMP21]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; UNROLL-NO-IC:       pred.udiv.continue9:
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = phi <2 x i32> [ poison, [[PRED_UDIV_CONTINUE5]] ], [ [[TMP22]], [[PRED_UDIV_IF8]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT7]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP24]], label [[PRED_UDIV_IF10:%.*]], label [[PRED_UDIV_CONTINUE11]]
+; UNROLL-NO-IC:       pred.udiv.if10:
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = add i32 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = extractelement <2 x i32> [[WIDE_LOAD3]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = udiv i32 [[TMP26]], [[TMP25]]
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = insertelement <2 x i32> [[TMP23]], i32 [[TMP27]], i32 1
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE11]]
+; UNROLL-NO-IC:       pred.udiv.continue11:
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = phi <2 x i32> [ [[TMP23]], [[PRED_UDIV_CONTINUE9]] ], [ [[TMP28]], [[PRED_UDIV_IF10]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = xor <2 x i1> [[BROADCAST_SPLAT]], <i1 true, i1 true>
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = xor <2 x i1> [[BROADCAST_SPLAT7]], <i1 true, i1 true>
+; UNROLL-NO-IC-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP30]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[PREDPHI12:%.*]] = select <2 x i1> [[TMP31]], <2 x i32> [[WIDE_LOAD3]], <2 x i32> [[TMP29]]
+; UNROLL-NO-IC-NEXT:    [[TMP32]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]
+; UNROLL-NO-IC-NEXT:    [[TMP33]] = add <2 x i32> [[PREDPHI12]], [[VEC_PHI2]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP34]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <2 x i32> [[TMP33]], [[TMP32]]
+; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP35]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[I]]
+; UNROLL-NO-IC-NEXT:    [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
+; UNROLL-NO-IC-NEXT:    br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
+; UNROLL-NO-IC:       if.then:
+; UNROLL-NO-IC-NEXT:    [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
+; UNROLL-NO-IC-NEXT:    br label [[IF_END]]
+; UNROLL-NO-IC:       if.end:
+; UNROLL-NO-IC-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR4]] = add i32 [[VAR3]], [[SUM]]
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i32 [[I]], 1
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP35]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[VAR5]]
+;
+; INTERLEAVE-LABEL: @scalarize_induction_variable_05(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i32 [[SMAX]], 2147483640
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C:%.*]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT10:%.*]] = insertelement <4 x i1> poison, i1 [[C]], i32 0
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE19:%.*]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP48:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[PRED_UDIV_CONTINUE19]] ]
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = or i32 [[INDEX]], 4
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = sext i32 [[INDEX]] to i64
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <4 x i32>*
+; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP3]], align 4
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 4
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <4 x i32>*
+; INTERLEAVE-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i32>, <4 x i32>* [[TMP5]], align 4
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
+; INTERLEAVE:       pred.udiv.if:
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 0
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = udiv i32 [[TMP6]], [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = insertelement <4 x i32> poison, i32 [[TMP7]], i32 0
+; INTERLEAVE-NEXT:    br label [[PRED_UDIV_CONTINUE]]
+; INTERLEAVE:       pred.udiv.continue:
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_UDIV_IF]] ]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
+; INTERLEAVE:       pred.udiv.if4:
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = or i32 [[INDEX]], 1
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 1
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = udiv i32 [[TMP11]], [[TMP10]]
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP9]], i32 [[TMP12]], i32 1
+; INTERLEAVE-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
+; INTERLEAVE:       pred.udiv.continue5:
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = phi <4 x i32> [ [[TMP9]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF4]] ]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
+; INTERLEAVE:       pred.udiv.if6:
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = or i32 [[INDEX]], 2
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 2
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = udiv i32 [[TMP16]], [[TMP15]]
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP17]], i32 2
+; INTERLEAVE-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
+; INTERLEAVE:       pred.udiv.continue7:
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = phi <4 x i32> [ [[TMP14]], [[PRED_UDIV_CONTINUE5]] ], [ [[TMP18]], [[PRED_UDIV_IF6]] ]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
+; INTERLEAVE:       pred.udiv.if8:
+; INTERLEAVE-NEXT:    [[TMP20:%.*]] = or i32 [[INDEX]], 3
+; INTERLEAVE-NEXT:    [[TMP21:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
+; INTERLEAVE-NEXT:    [[TMP22:%.*]] = udiv i32 [[TMP21]], [[TMP20]]
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP22]], i32 3
+; INTERLEAVE-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
+; INTERLEAVE:       pred.udiv.continue9:
+; INTERLEAVE-NEXT:    [[TMP24:%.*]] = phi <4 x i32> [ [[TMP19]], [[PRED_UDIV_CONTINUE7]] ], [ [[TMP23]], [[PRED_UDIV_IF8]] ]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF12:%.*]], label [[PRED_UDIV_CONTINUE13:%.*]]
+; INTERLEAVE:       pred.udiv.if12:
+; INTERLEAVE-NEXT:    [[TMP25:%.*]] = extractelement <4 x i32> [[WIDE_LOAD3]], i32 0
+; INTERLEAVE-NEXT:    [[TMP26:%.*]] = udiv i32 [[TMP25]], [[TMP0]]
+; INTERLEAVE-NEXT:    [[TMP27:%.*]] = insertelement <4 x i32> poison, i32 [[TMP26]], i32 0
+; INTERLEAVE-NEXT:    br label [[PRED_UDIV_CONTINUE13]]
+; INTERLEAVE:       pred.udiv.continue13:
+; INTERLEAVE-NEXT:    [[TMP28:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE9]] ], [ [[TMP27]], [[PRED_UDIV_IF12]] ]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF14:%.*]], label [[PRED_UDIV_CONTINUE15:%.*]]
+; INTERLEAVE:       pred.udiv.if14:
+; INTERLEAVE-NEXT:    [[TMP29:%.*]] = or i32 [[INDEX]], 5
+; INTERLEAVE-NEXT:    [[TMP30:%.*]] = extractelement <4 x i32> [[WIDE_LOAD3]], i32 1
+; INTERLEAVE-NEXT:    [[TMP31:%.*]] = udiv i32 [[TMP30]], [[TMP29]]
+; INTERLEAVE-NEXT:    [[TMP32:%.*]] = insertelement <4 x i32> [[TMP28]], i32 [[TMP31]], i32 1
+; INTERLEAVE-NEXT:    br label [[PRED_UDIV_CONTINUE15]]
+; INTERLEAVE:       pred.udiv.continue15:
+; INTERLEAVE-NEXT:    [[TMP33:%.*]] = phi <4 x i32> [ [[TMP28]], [[PRED_UDIV_CONTINUE13]] ], [ [[TMP32]], [[PRED_UDIV_IF14]] ]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF16:%.*]], label [[PRED_UDIV_CONTINUE17:%.*]]
+; INTERLEAVE:       pred.udiv.if16:
+; INTERLEAVE-NEXT:    [[TMP34:%.*]] = or i32 [[INDEX]], 6
+; INTERLEAVE-NEXT:    [[TMP35:%.*]] = extractelement <4 x i32> [[WIDE_LOAD3]], i32 2
+; INTERLEAVE-NEXT:    [[TMP36:%.*]] = udiv i32 [[TMP35]], [[TMP34]]
+; INTERLEAVE-NEXT:    [[TMP37:%.*]] = insertelement <4 x i32> [[TMP33]], i32 [[TMP36]], i32 2
+; INTERLEAVE-NEXT:    br label [[PRED_UDIV_CONTINUE17]]
+; INTERLEAVE:       pred.udiv.continue17:
+; INTERLEAVE-NEXT:    [[TMP38:%.*]] = phi <4 x i32> [ [[TMP33]], [[PRED_UDIV_CONTINUE15]] ], [ [[TMP37]], [[PRED_UDIV_IF16]] ]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF18:%.*]], label [[PRED_UDIV_CONTINUE19]]
+; INTERLEAVE:       pred.udiv.if18:
+; INTERLEAVE-NEXT:    [[TMP39:%.*]] = or i32 [[INDEX]], 7
+; INTERLEAVE-NEXT:    [[TMP40:%.*]] = extractelement <4 x i32> [[WIDE_LOAD3]], i32 3
+; INTERLEAVE-NEXT:    [[TMP41:%.*]] = udiv i32 [[TMP40]], [[TMP39]]
+; INTERLEAVE-NEXT:    [[TMP42:%.*]] = insertelement <4 x i32> [[TMP38]], i32 [[TMP41]], i32 3
+; INTERLEAVE-NEXT:    br label [[PRED_UDIV_CONTINUE19]]
+; INTERLEAVE:       pred.udiv.continue19:
+; INTERLEAVE-NEXT:    [[TMP43:%.*]] = phi <4 x i32> [ [[TMP38]], [[PRED_UDIV_CONTINUE17]] ], [ [[TMP42]], [[PRED_UDIV_IF18]] ]
+; INTERLEAVE-NEXT:    [[TMP44:%.*]] = xor <4 x i1> [[BROADCAST_SPLATINSERT]], <i1 true, i1 poison, i1 poison, i1 poison>
+; INTERLEAVE-NEXT:    [[TMP45:%.*]] = shufflevector <4 x i1> [[TMP44]], <4 x i1> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[TMP46:%.*]] = xor <4 x i1> [[BROADCAST_SPLATINSERT10]], <i1 true, i1 poison, i1 poison, i1 poison>
+; INTERLEAVE-NEXT:    [[TMP47:%.*]] = shufflevector <4 x i1> [[TMP46]], <4 x i1> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP45]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[TMP24]]
+; INTERLEAVE-NEXT:    [[PREDPHI20:%.*]] = select <4 x i1> [[TMP47]], <4 x i32> [[WIDE_LOAD3]], <4 x i32> [[TMP43]]
+; INTERLEAVE-NEXT:    [[TMP48]] = add <4 x i32> [[PREDPHI]], [[VEC_PHI]]
+; INTERLEAVE-NEXT:    [[TMP49]] = add <4 x i32> [[PREDPHI20]], [[VEC_PHI2]]
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP50:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP50]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP49]], [[TMP48]]
+; INTERLEAVE-NEXT:    [[TMP51:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; INTERLEAVE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP51]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
+; INTERLEAVE-NEXT:    [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
+; INTERLEAVE-NEXT:    [[TMP52:%.*]] = zext i32 [[I]] to i64
+; INTERLEAVE-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP52]]
+; INTERLEAVE-NEXT:    [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
+; INTERLEAVE:       if.then:
+; INTERLEAVE-NEXT:    [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
+; INTERLEAVE-NEXT:    br label [[IF_END]]
+; INTERLEAVE:       if.end:
+; INTERLEAVE-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VAR4]] = add i32 [[VAR3]], [[SUM]]
+; INTERLEAVE-NEXT:    [[I_NEXT]] = add nuw nsw i32 [[I]], 1
+; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    ret i32 [[VAR5]]
+;
 entry:
   br label %for.body
 
 for.body:
   %i = phi i32 [ 0, %entry ], [ %i.next, %if.end ]
-  %sum = phi i32 [ 0, %entry ], [ %tmp4, %if.end ]
-  %tmp0 = getelementptr inbounds i32, i32* %a, i32 %i
-  %tmp1 = load i32, i32* %tmp0, align 4
+  %sum = phi i32 [ 0, %entry ], [ %var4, %if.end ]
+  %var0 = getelementptr inbounds i32, i32* %a, i32 %i
+  %var1 = load i32, i32* %var0, align 4
   br i1 %c, label %if.then, label %if.end
 
 if.then:
-  %tmp2 = udiv i32 %tmp1, %i
+  %var2 = udiv i32 %var1, %i
   br label %if.end
 
 if.end:
-  %tmp3 = phi i32 [ %tmp2, %if.then ], [ %tmp1, %for.body ]
-  %tmp4 = add i32 %tmp3, %sum
+  %var3 = phi i32 [ %var2, %if.then ], [ %var1, %for.body ]
+  %var4 = add i32 %var3, %sum
   %i.next = add nuw nsw i32 %i, 1
   %cond = icmp slt i32 %i.next, %n
   br i1 %cond, label %for.body, label %for.end
 
 for.end:
-  %tmp5  = phi i32 [ %tmp4, %if.end ]
-  ret i32 %tmp5
+  %var5  = phi i32 [ %var4, %if.end ]
+  ret i32 %var5
 }
 
 ; Ensure we generate both a vector and a scalar induction variable. In this
@@ -397,46 +2524,315 @@ for.end:
 ; vectorized (trunc) as well as an instruction that will remain in scalar form
 ; (gepelementptr).
 ;
-; CHECK-LABEL: @iv_vector_and_scalar_users(
-; CHECK: vector.body:
-; CHECK:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK:   [[VEC_IV_1:%.+]] = phi <2 x i64> [ <i64 0, i64 1>, %vector.ph ], [ [[VEC_IV_1_NEXT:%.+]], %vector.body ]
-; CHECK:   [[VEC_IV_2:%.+]] = phi <2 x i32> [ <i32 0, i32 1>, %vector.ph ], [ [[VEC_IV_2_NEXT:%.+]], %vector.body ]
-; CHECK:   %[[i0:.+]] = add i64 %index, 0
-; CHECK:   %[[i1:.+]] = add i64 %index, 1
-; CHECK:   getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %[[i0]], i32 1
-; CHECK:   getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %[[i1]], i32 1
-; CHECK:   %index.next = add nuw i64 %index, 2
-; CHECK:   [[VEC_IV_1_NEXT]] = add <2 x i64> [[VEC_IV_1]], <i64 2, i64 2>
-; CHECK:   [[VEC_IV_2_NEXT]] = add <2 x i32> [[VEC_IV_2]], <i32 2, i32 2>
 ;
-; IND-LABEL: @iv_vector_and_scalar_users(
-; IND: vector.body:
-; IND:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; IND:   [[VEC_IV:%.+]] = phi <2 x i32> [ <i32 0, i32 1>, %vector.ph ], [ [[VEC_IV_NEXT:%.+]], %vector.body ]
-; IND:   %[[i1:.+]] = or i64 %index, 1
-; IND:   getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %index, i32 1
-; IND:   getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %[[i1]], i32 1
-; IND:   %index.next = add nuw i64 %index, 2
-; IND:   [[VEC_IV_NEXT]] = add <2 x i32> [[VEC_IV]], <i32 2, i32 2>
 ;
-; UNROLL-LABEL: @iv_vector_and_scalar_users(
-; UNROLL: vector.body:
-; UNROLL:   %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; UNROLL:   [[VEC_IV:%.+]] = phi <2 x i32> [ <i32 0, i32 1>, %vector.ph ], [ [[VEC_IV_NEXT:%.+]], %vector.body ]
-; UNROLL:   %[[i1:.+]] = or i64 %index, 1
-; UNROLL:   %[[i2:.+]] = or i64 %index, 2
-; UNROLL:   %[[i3:.+]] = or i64 %index, 3
-; UNROLL:   %[[add:.+]]= add <2 x i32> %[[splat:.+]], <i32 2, i32 poison>
-; UNROLL:   getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %index, i32 1
-; UNROLL:   getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %[[i1]], i32 1
-; UNROLL:   getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %[[i2]], i32 1
-; UNROLL:   getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %[[i3]], i32 1
-; UNROLL:   %index.next = add nuw i64 %index, 4
-; UNROLL:   [[VEC_IV_NEXT]] = add <2 x i32> [[VEC_IV]], <i32 4, i32 4>
 
 %pair.i16 = type { i16, i16 }
 define void @iv_vector_and_scalar_users(%pair.i16* %p, i32 %a, i32 %n) {
+; CHECK-LABEL: @iv_vector_and_scalar_users(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND1:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP5:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND1]]
+; CHECK-NEXT:    [[TMP6:%.*]] = trunc <2 x i32> [[TMP5]] to <2 x i16>
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[TMP3]], i32 1
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP4]], i32 1
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i16> [[TMP6]], i32 0
+; CHECK-NEXT:    store i16 [[TMP9]], i16* [[TMP7]], align 2
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x i16> [[TMP6]], i32 1
+; CHECK-NEXT:    store i16 [[TMP10]], i16* [[TMP8]], align 2
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i32> [[VEC_IND1]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[TMP12:%.*]] = trunc i64 [[I]] to i32
+; CHECK-NEXT:    [[TMP13:%.*]] = add i32 [[A]], [[TMP12]]
+; CHECK-NEXT:    [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
+; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
+; CHECK-NEXT:    store i16 [[TMP14]], i16* [[TMP15]], align 2
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; CHECK-NEXT:    [[TMP16:%.*]] = trunc i64 [[I_NEXT]] to i32
+; CHECK-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP16]], [[N]]
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @iv_vector_and_scalar_users(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; IND-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; IND-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
+; IND-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i32 0
+; IND-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND1:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 1
+; IND-NEXT:    [[TMP4:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND1]]
+; IND-NEXT:    [[TMP5:%.*]] = trunc <2 x i32> [[TMP4]] to <2 x i16>
+; IND-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[INDEX]], i32 1
+; IND-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP3]], i32 1
+; IND-NEXT:    [[TMP8:%.*]] = extractelement <2 x i16> [[TMP5]], i32 0
+; IND-NEXT:    store i16 [[TMP8]], i16* [[TMP6]], align 2
+; IND-NEXT:    [[TMP9:%.*]] = extractelement <2 x i16> [[TMP5]], i32 1
+; IND-NEXT:    store i16 [[TMP9]], i16* [[TMP7]], align 2
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i32> [[VEC_IND1]], <i32 2, i32 2>
+; IND-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[TMP11:%.*]] = trunc i64 [[I]] to i32
+; IND-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], [[A]]
+; IND-NEXT:    [[TMP13:%.*]] = trunc i32 [[TMP12]] to i16
+; IND-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
+; IND-NEXT:    store i16 [[TMP13]], i16* [[TMP14]], align 2
+; IND-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; IND-NEXT:    [[TMP15:%.*]] = trunc i64 [[I_NEXT]] to i32
+; IND-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP15]], [[N]]
+; IND-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @iv_vector_and_scalar_users(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT6:%.*]] = insertelement <2 x i32> poison, i32 [[A]], i32 0
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND2:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 1
+; UNROLL-NEXT:    [[TMP4:%.*]] = or i64 [[INDEX]], 2
+; UNROLL-NEXT:    [[TMP5:%.*]] = or i64 [[INDEX]], 3
+; UNROLL-NEXT:    [[TMP6:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND2]]
+; UNROLL-NEXT:    [[TMP7:%.*]] = add <2 x i32> [[BROADCAST_SPLATINSERT6]], <i32 2, i32 poison>
+; UNROLL-NEXT:    [[TMP8:%.*]] = shufflevector <2 x i32> [[TMP7]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[TMP9:%.*]] = add <2 x i32> [[TMP8]], [[VEC_IND2]]
+; UNROLL-NEXT:    [[TMP10:%.*]] = trunc <2 x i32> [[TMP6]] to <2 x i16>
+; UNROLL-NEXT:    [[TMP11:%.*]] = trunc <2 x i32> [[TMP9]] to <2 x i16>
+; UNROLL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[INDEX]], i32 1
+; UNROLL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP3]], i32 1
+; UNROLL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP4]], i32 1
+; UNROLL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP5]], i32 1
+; UNROLL-NEXT:    [[TMP16:%.*]] = extractelement <2 x i16> [[TMP10]], i32 0
+; UNROLL-NEXT:    store i16 [[TMP16]], i16* [[TMP12]], align 2
+; UNROLL-NEXT:    [[TMP17:%.*]] = extractelement <2 x i16> [[TMP10]], i32 1
+; UNROLL-NEXT:    store i16 [[TMP17]], i16* [[TMP13]], align 2
+; UNROLL-NEXT:    [[TMP18:%.*]] = extractelement <2 x i16> [[TMP11]], i32 0
+; UNROLL-NEXT:    store i16 [[TMP18]], i16* [[TMP14]], align 2
+; UNROLL-NEXT:    [[TMP19:%.*]] = extractelement <2 x i16> [[TMP11]], i32 1
+; UNROLL-NEXT:    store i16 [[TMP19]], i16* [[TMP15]], align 2
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT5]] = add <2 x i32> [[VEC_IND2]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[TMP21:%.*]] = trunc i64 [[I]] to i32
+; UNROLL-NEXT:    [[TMP22:%.*]] = add i32 [[TMP21]], [[A]]
+; UNROLL-NEXT:    [[TMP23:%.*]] = trunc i32 [[TMP22]] to i16
+; UNROLL-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
+; UNROLL-NEXT:    store i16 [[TMP23]], i16* [[TMP24]], align 2
+; UNROLL-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NEXT:    [[TMP25:%.*]] = trunc i64 [[I_NEXT]] to i32
+; UNROLL-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP25]], [[N]]
+; UNROLL-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @iv_vector_and_scalar_users(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT6:%.*]] = insertelement <2 x i32> poison, i32 [[A]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT7:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT6]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND2:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD3:%.*]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND2]]
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add <2 x i32> [[BROADCAST_SPLAT7]], [[STEP_ADD3]]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = trunc <2 x i32> [[TMP7]] to <2 x i16>
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = trunc <2 x i32> [[TMP8]] to <2 x i16>
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[TMP3]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP4]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP5]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP6]], i32 1
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = extractelement <2 x i16> [[TMP9]], i32 0
+; UNROLL-NO-IC-NEXT:    store i16 [[TMP15]], i16* [[TMP11]], align 2
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = extractelement <2 x i16> [[TMP9]], i32 1
+; UNROLL-NO-IC-NEXT:    store i16 [[TMP16]], i16* [[TMP12]], align 2
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = extractelement <2 x i16> [[TMP10]], i32 0
+; UNROLL-NO-IC-NEXT:    store i16 [[TMP17]], i16* [[TMP13]], align 2
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = extractelement <2 x i16> [[TMP10]], i32 1
+; UNROLL-NO-IC-NEXT:    store i16 [[TMP18]], i16* [[TMP14]], align 2
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT5]] = add <2 x i32> [[STEP_ADD3]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = trunc i64 [[I]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = add i32 [[A]], [[TMP20]]
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = trunc i32 [[TMP21]] to i16
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
+; UNROLL-NO-IC-NEXT:    store i16 [[TMP22]], i16* [[TMP23]], align 2
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = trunc i64 [[I_NEXT]] to i32
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP24]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @iv_vector_and_scalar_users(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[A:%.*]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT6:%.*]] = insertelement <4 x i32> poison, i32 [[A]], i32 0
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND2:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = or i64 [[INDEX]], 1
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = or i64 [[INDEX]], 2
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = or i64 [[INDEX]], 3
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = or i64 [[INDEX]], 4
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = or i64 [[INDEX]], 5
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = or i64 [[INDEX]], 6
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = or i64 [[INDEX]], 7
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[BROADCAST_SPLAT]], [[VEC_IND2]]
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = add <4 x i32> [[BROADCAST_SPLATINSERT6]], <i32 4, i32 poison, i32 poison, i32 poison>
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32> [[TMP11]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = add <4 x i32> [[TMP12]], [[VEC_IND2]]
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = trunc <4 x i32> [[TMP10]] to <4 x i16>
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = trunc <4 x i32> [[TMP13]] to <4 x i16>
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[INDEX]], i32 1
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP3]], i32 1
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP4]], i32 1
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP5]], i32 1
+; INTERLEAVE-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP6]], i32 1
+; INTERLEAVE-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP7]], i32 1
+; INTERLEAVE-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP8]], i32 1
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP9]], i32 1
+; INTERLEAVE-NEXT:    [[TMP24:%.*]] = extractelement <4 x i16> [[TMP14]], i32 0
+; INTERLEAVE-NEXT:    store i16 [[TMP24]], i16* [[TMP16]], align 2
+; INTERLEAVE-NEXT:    [[TMP25:%.*]] = extractelement <4 x i16> [[TMP14]], i32 1
+; INTERLEAVE-NEXT:    store i16 [[TMP25]], i16* [[TMP17]], align 2
+; INTERLEAVE-NEXT:    [[TMP26:%.*]] = extractelement <4 x i16> [[TMP14]], i32 2
+; INTERLEAVE-NEXT:    store i16 [[TMP26]], i16* [[TMP18]], align 2
+; INTERLEAVE-NEXT:    [[TMP27:%.*]] = extractelement <4 x i16> [[TMP14]], i32 3
+; INTERLEAVE-NEXT:    store i16 [[TMP27]], i16* [[TMP19]], align 2
+; INTERLEAVE-NEXT:    [[TMP28:%.*]] = extractelement <4 x i16> [[TMP15]], i32 0
+; INTERLEAVE-NEXT:    store i16 [[TMP28]], i16* [[TMP20]], align 2
+; INTERLEAVE-NEXT:    [[TMP29:%.*]] = extractelement <4 x i16> [[TMP15]], i32 1
+; INTERLEAVE-NEXT:    store i16 [[TMP29]], i16* [[TMP21]], align 2
+; INTERLEAVE-NEXT:    [[TMP30:%.*]] = extractelement <4 x i16> [[TMP15]], i32 2
+; INTERLEAVE-NEXT:    store i16 [[TMP30]], i16* [[TMP22]], align 2
+; INTERLEAVE-NEXT:    [[TMP31:%.*]] = extractelement <4 x i16> [[TMP15]], i32 3
+; INTERLEAVE-NEXT:    store i16 [[TMP31]], i16* [[TMP23]], align 2
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT5]] = add <4 x i32> [[VEC_IND2]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[TMP33:%.*]] = trunc i64 [[I]] to i32
+; INTERLEAVE-NEXT:    [[TMP34:%.*]] = add i32 [[TMP33]], [[A]]
+; INTERLEAVE-NEXT:    [[TMP35:%.*]] = trunc i32 [[TMP34]] to i16
+; INTERLEAVE-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
+; INTERLEAVE-NEXT:    store i16 [[TMP35]], i16* [[TMP36]], align 2
+; INTERLEAVE-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; INTERLEAVE-NEXT:    [[TMP37:%.*]] = trunc i64 [[I_NEXT]] to i32
+; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp eq i32 [[TMP37]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    ret void
+;
 entry:
   br label %for.body
 
@@ -462,9 +2858,131 @@ for.end:
 ; to 0.
 ; PR17532
 
-; CHECK-LABEL: i8_loop
-; CHECK: icmp eq i32 {{.*}}, 256
 define i32 @i8_loop() nounwind readnone ssp uwtable {
+; CHECK-LABEL: @i8_loop(
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i32 [[INDEX]] to i8
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i8 0, [[TMP1]]
+; CHECK-NEXT:    [[TMP2:%.*]] = add i8 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP3]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP3]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 256, 256
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[TMP10:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[TMP0:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[TMP0]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[TMP6:%.*]]
+; CHECK:       6:
+; CHECK-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP7:%.*]], [[TMP6]] ]
+; CHECK-NEXT:    [[B_0:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP8:%.*]], [[TMP6]] ]
+; CHECK-NEXT:    [[TMP7]] = and i32 [[A_0]], 4
+; CHECK-NEXT:    [[TMP8]] = add i8 [[B_0]], -1
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i8 [[TMP8]], 0
+; CHECK-NEXT:    br i1 [[TMP9]], label [[TMP10]], label [[TMP6]], !llvm.loop [[LOOP29:![0-9]+]]
+; CHECK:       10:
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP7]], [[TMP6]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[DOTLCSSA]]
+;
+; IND-LABEL: @i8_loop(
+; IND-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
+; IND-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    br i1 true, label [[TMP3:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    br label [[TMP2:%.*]]
+; IND:       2:
+; IND-NEXT:    br i1 undef, label [[TMP3]], label [[TMP2]], !llvm.loop [[LOOP29:![0-9]+]]
+; IND:       3:
+; IND-NEXT:    ret i32 0
+;
+; UNROLL-LABEL: @i8_loop(
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
+; UNROLL-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    br i1 true, label [[TMP3:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[TMP2:%.*]]
+; UNROLL:       2:
+; UNROLL-NEXT:    br i1 undef, label [[TMP3]], label [[TMP2]], !llvm.loop [[LOOP29:![0-9]+]]
+; UNROLL:       3:
+; UNROLL-NEXT:    ret i32 0
+;
+; UNROLL-NO-IC-LABEL: @i8_loop(
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ <i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = trunc i32 [[INDEX]] to i8
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = sub i8 0, [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i8 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i8 [[OFFSET_IDX]], -2
+; UNROLL-NO-IC-NEXT:    [[TMP4]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP5]] = and <2 x i32> [[VEC_PHI1]], <i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = and <2 x i32> [[TMP5]], [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 256, 256
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[TMP12:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[TMP0:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[TMP0]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[TMP8:%.*]]
+; UNROLL-NO-IC:       8:
+; UNROLL-NO-IC-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP9:%.*]], [[TMP8]] ]
+; UNROLL-NO-IC-NEXT:    [[B_0:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP10:%.*]], [[TMP8]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP9]] = and i32 [[A_0]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP10]] = add i8 [[B_0]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = icmp eq i8 [[TMP10]], 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP11]], label [[TMP12]], label [[TMP8]], !llvm.loop [[LOOP29:![0-9]+]]
+; UNROLL-NO-IC:       12:
+; UNROLL-NO-IC-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP9]], [[TMP8]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[DOTLCSSA]]
+;
+; INTERLEAVE-LABEL: @i8_loop(
+; INTERLEAVE-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
+; INTERLEAVE-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    br i1 true, label [[TMP3:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    br label [[TMP2:%.*]]
+; INTERLEAVE:       2:
+; INTERLEAVE-NEXT:    br i1 undef, label [[TMP3]], label [[TMP2]], !llvm.loop [[LOOP29:![0-9]+]]
+; INTERLEAVE:       3:
+; INTERLEAVE-NEXT:    ret i32 0
+;
   br label %1
 
 ; <label>:1                                       ; preds = %1, %0
@@ -479,10 +2997,132 @@ define i32 @i8_loop() nounwind readnone ssp uwtable {
   ret i32 %2
 }
 
-; CHECK-LABEL: i16_loop
-; CHECK: icmp eq i32 {{.*}}, 65536
 
 define i32 @i16_loop() nounwind readnone ssp uwtable {
+; CHECK-LABEL: @i16_loop(
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i32 [[INDEX]] to i16
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i16 0, [[TMP1]]
+; CHECK-NEXT:    [[TMP2:%.*]] = add i16 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP3]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP3]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 65536, 65536
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[TMP10:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[TMP0:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[TMP0]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[TMP6:%.*]]
+; CHECK:       6:
+; CHECK-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP7:%.*]], [[TMP6]] ]
+; CHECK-NEXT:    [[B_0:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP8:%.*]], [[TMP6]] ]
+; CHECK-NEXT:    [[TMP7]] = and i32 [[A_0]], 4
+; CHECK-NEXT:    [[TMP8]] = add i16 [[B_0]], -1
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i16 [[TMP8]], 0
+; CHECK-NEXT:    br i1 [[TMP9]], label [[TMP10]], label [[TMP6]], !llvm.loop [[LOOP31:![0-9]+]]
+; CHECK:       10:
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP7]], [[TMP6]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[DOTLCSSA]]
+;
+; IND-LABEL: @i16_loop(
+; IND-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
+; IND-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    br i1 true, label [[TMP3:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    br label [[TMP2:%.*]]
+; IND:       2:
+; IND-NEXT:    br i1 undef, label [[TMP3]], label [[TMP2]], !llvm.loop [[LOOP31:![0-9]+]]
+; IND:       3:
+; IND-NEXT:    ret i32 0
+;
+; UNROLL-LABEL: @i16_loop(
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
+; UNROLL-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    br i1 true, label [[TMP3:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[TMP2:%.*]]
+; UNROLL:       2:
+; UNROLL-NEXT:    br i1 undef, label [[TMP3]], label [[TMP2]], !llvm.loop [[LOOP31:![0-9]+]]
+; UNROLL:       3:
+; UNROLL-NEXT:    ret i32 0
+;
+; UNROLL-NO-IC-LABEL: @i16_loop(
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ <i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = trunc i32 [[INDEX]] to i16
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = sub i16 0, [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i16 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i16 [[OFFSET_IDX]], -2
+; UNROLL-NO-IC-NEXT:    [[TMP4]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP5]] = and <2 x i32> [[VEC_PHI1]], <i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = and <2 x i32> [[TMP5]], [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 65536, 65536
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[TMP12:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[TMP0:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[TMP0]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[TMP8:%.*]]
+; UNROLL-NO-IC:       8:
+; UNROLL-NO-IC-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP9:%.*]], [[TMP8]] ]
+; UNROLL-NO-IC-NEXT:    [[B_0:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP10:%.*]], [[TMP8]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP9]] = and i32 [[A_0]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP10]] = add i16 [[B_0]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = icmp eq i16 [[TMP10]], 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP11]], label [[TMP12]], label [[TMP8]], !llvm.loop [[LOOP31:![0-9]+]]
+; UNROLL-NO-IC:       12:
+; UNROLL-NO-IC-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP9]], [[TMP8]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[DOTLCSSA]]
+;
+; INTERLEAVE-LABEL: @i16_loop(
+; INTERLEAVE-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
+; INTERLEAVE-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    br i1 true, label [[TMP3:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    br label [[TMP2:%.*]]
+; INTERLEAVE:       2:
+; INTERLEAVE-NEXT:    br i1 undef, label [[TMP3]], label [[TMP2]], !llvm.loop [[LOOP31:![0-9]+]]
+; INTERLEAVE:       3:
+; INTERLEAVE-NEXT:    ret i32 0
+;
   br label %1
 
 ; <label>:1                                       ; preds = %1, %0
@@ -500,17 +3140,131 @@ define i32 @i16_loop() nounwind readnone ssp uwtable {
 ; This loop has a backedge taken count of i32_max. We need to check for this
 ; condition and branch directly to the scalar loop.
 
-; CHECK-LABEL: max_i32_backedgetaken
-; CHECK:  br i1 true, label %scalar.ph, label %vector.ph
 
-; CHECK: middle.block:
-; CHECK:  %[[v9:.+]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32>
-; CHECK: scalar.ph:
-; CHECK:  %bc.resume.val = phi i32 [ 0, %middle.block ], [ 0, %[[v0:.+]] ]
-; CHECK:  %bc.merge.rdx = phi i32 [ 1, %[[v0:.+]] ], [ %[[v9]], %middle.block ]
 
 define i32 @max_i32_backedgetaken() nounwind readnone ssp uwtable {
-
+; CHECK-LABEL: @max_i32_backedgetaken(
+; CHECK-NEXT:    br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 0, [[INDEX]]
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP2]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP2]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 0, 0
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[TMP9:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[TMP0:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[TMP0]] ], [ [[TMP4]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[TMP5:%.*]]
+; CHECK:       5:
+; CHECK-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP6:%.*]], [[TMP5]] ]
+; CHECK-NEXT:    [[B_0:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP7:%.*]], [[TMP5]] ]
+; CHECK-NEXT:    [[TMP6]] = and i32 [[A_0]], 4
+; CHECK-NEXT:    [[TMP7]] = add i32 [[B_0]], -1
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
+; CHECK-NEXT:    br i1 [[TMP8]], label [[TMP9]], label [[TMP5]], !llvm.loop [[LOOP33:![0-9]+]]
+; CHECK:       9:
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP6]], [[TMP5]] ], [ [[TMP4]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[DOTLCSSA]]
+;
+; IND-LABEL: @max_i32_backedgetaken(
+; IND-NEXT:    br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    br i1 undef, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    br i1 undef, label [[TMP4:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    br label [[TMP1:%.*]]
+; IND:       1:
+; IND-NEXT:    [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH]] ], [ [[TMP2:%.*]], [[TMP1]] ]
+; IND-NEXT:    [[TMP2]] = add i32 [[B_0]], -1
+; IND-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[TMP2]], 0
+; IND-NEXT:    br i1 [[TMP3]], label [[TMP4]], label [[TMP1]], !llvm.loop [[LOOP33:![0-9]+]]
+; IND:       4:
+; IND-NEXT:    ret i32 0
+;
+; UNROLL-LABEL: @max_i32_backedgetaken(
+; UNROLL-NEXT:    br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    br i1 undef, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    br i1 undef, label [[TMP4:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[TMP1:%.*]]
+; UNROLL:       1:
+; UNROLL-NEXT:    [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH]] ], [ [[TMP2:%.*]], [[TMP1]] ]
+; UNROLL-NEXT:    [[TMP2]] = add i32 [[B_0]], -1
+; UNROLL-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[TMP2]], 0
+; UNROLL-NEXT:    br i1 [[TMP3]], label [[TMP4]], label [[TMP1]], !llvm.loop [[LOOP33:![0-9]+]]
+; UNROLL:       4:
+; UNROLL-NEXT:    ret i32 0
+;
+; UNROLL-NO-IC-LABEL: @max_i32_backedgetaken(
+; UNROLL-NO-IC-NEXT:    br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ <i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 0, [[INDEX]]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], -2
+; UNROLL-NO-IC-NEXT:    [[TMP3]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP4]] = and <2 x i32> [[VEC_PHI1]], <i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = and <2 x i32> [[TMP4]], [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 0, 0
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[TMP11:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[TMP0:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[TMP0]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[TMP7:%.*]]
+; UNROLL-NO-IC:       7:
+; UNROLL-NO-IC-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP8:%.*]], [[TMP7]] ]
+; UNROLL-NO-IC-NEXT:    [[B_0:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP9:%.*]], [[TMP7]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP8]] = and i32 [[A_0]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP9]] = add i32 [[B_0]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP10]], label [[TMP11]], label [[TMP7]], !llvm.loop [[LOOP33:![0-9]+]]
+; UNROLL-NO-IC:       11:
+; UNROLL-NO-IC-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP8]], [[TMP7]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[DOTLCSSA]]
+;
+; INTERLEAVE-LABEL: @max_i32_backedgetaken(
+; INTERLEAVE-NEXT:    br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    br i1 undef, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    br i1 undef, label [[TMP4:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    br label [[TMP1:%.*]]
+; INTERLEAVE:       1:
+; INTERLEAVE-NEXT:    [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH]] ], [ [[TMP2:%.*]], [[TMP1]] ]
+; INTERLEAVE-NEXT:    [[TMP2]] = add i32 [[B_0]], -1
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[TMP2]], 0
+; INTERLEAVE-NEXT:    br i1 [[TMP3]], label [[TMP4]], label [[TMP1]], !llvm.loop [[LOOP33:![0-9]+]]
+; INTERLEAVE:       4:
+; INTERLEAVE-NEXT:    ret i32 0
+;
   br label %1
 
 ; <label>:1                                       ; preds = %1, %0
@@ -528,18 +3282,244 @@ define i32 @max_i32_backedgetaken() nounwind readnone ssp uwtable {
 ; When generating the overflow check we must sure that the induction start value
 ; is defined before the branch to the scalar preheader.
 
-; CHECK-LABEL: testoverflowcheck
-; CHECK: entry
-; CHECK: %[[LOAD:.*]] = load i8
-; CHECK: br
 
-; CHECK: scalar.ph
-; CHECK: phi i8 [ %{{.*}}, %middle.block ], [ %[[LOAD]], %entry ]
 
 @e = global i8 1, align 1
 @d = common global i32 0, align 4
 @c = common global i32 0, align 4
 define i32 @testoverflowcheck() {
+; CHECK-LABEL: @testoverflowcheck(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* @d, align 4
+; CHECK-NEXT:    [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = sub i8 -1, [[DOTPR_I]]
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP3]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP3]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; CHECK-NEXT:    [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_CRD]]
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x i32> <i32 -1, i32 -1>, i32 [[C_PROMOTED_I]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP5:%.*]] = trunc i32 [[INDEX]] to i8
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[DOTPR_I]], [[TMP5]]
+; CHECK-NEXT:    [[TMP6:%.*]] = add i8 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP7]] = and <2 x i32> [[BROADCAST_SPLAT]], [[VEC_PHI]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP7]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[C_PROMOTED_I]], [[ENTRY]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[COND_END_I:%.*]]
+; CHECK:       cond.end.i:
+; CHECK-NEXT:    [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
+; CHECK-NEXT:    [[AND3_I:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[AND_I:%.*]], [[COND_END_I]] ]
+; CHECK-NEXT:    [[AND_I]] = and i32 [[TMP0]], [[AND3_I]]
+; CHECK-NEXT:    [[INC_I]] = add i8 [[INC4_I]], 1
+; CHECK-NEXT:    [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
+; CHECK-NEXT:    br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
+; CHECK:       loopexit:
+; CHECK-NEXT:    [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[AND_I_LCSSA]]
+;
+; IND-LABEL: @testoverflowcheck(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
+; IND-NEXT:    [[TMP0:%.*]] = load i32, i32* @d, align 4
+; IND-NEXT:    [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
+; IND-NEXT:    [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -1
+; IND-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
+; IND-NEXT:    [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp eq i8 [[DOTPR_I]], -1
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP3]], 510
+; IND-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; IND-NEXT:    [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_CRD]]
+; IND-NEXT:    [[TMP4:%.*]] = insertelement <2 x i32> <i32 poison, i32 -1>, i32 [[C_PROMOTED_I]], i32 0
+; IND-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
+; IND-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[TMP6:%.*]] = and <2 x i32> [[TMP4]], [[BROADCAST_SPLAT]]
+; IND-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP6]])
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
+; IND-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]
+; IND-NEXT:    br label [[COND_END_I:%.*]]
+; IND:       cond.end.i:
+; IND-NEXT:    [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
+; IND-NEXT:    [[TMP8:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]
+; IND-NEXT:    [[INC_I]] = add i8 [[INC4_I]], 1
+; IND-NEXT:    [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
+; IND-NEXT:    br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
+; IND:       loopexit:
+; IND-NEXT:    [[AND_I_LCSSA:%.*]] = phi i32 [ [[TMP8]], [[COND_END_I]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    ret i32 [[AND_I_LCSSA]]
+;
+; UNROLL-LABEL: @testoverflowcheck(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
+; UNROLL-NEXT:    [[TMP0:%.*]] = load i32, i32* @d, align 4
+; UNROLL-NEXT:    [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
+; UNROLL-NEXT:    [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -1
+; UNROLL-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
+; UNROLL-NEXT:    [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ugt i8 [[DOTPR_I]], -4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP3]], 508
+; UNROLL-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; UNROLL-NEXT:    [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_CRD]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = insertelement <2 x i32> <i32 poison, i32 -1>, i32 [[C_PROMOTED_I]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[TMP6:%.*]] = and <2 x i32> [[BROADCAST_SPLATINSERT2]], [[BROADCAST_SPLATINSERT]]
+; UNROLL-NEXT:    [[TMP7:%.*]] = shufflevector <2 x i32> [[TMP6]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = and <2 x i32> [[TMP7]], [[TMP4]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
+; UNROLL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]
+; UNROLL-NEXT:    br label [[COND_END_I:%.*]]
+; UNROLL:       cond.end.i:
+; UNROLL-NEXT:    [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
+; UNROLL-NEXT:    [[TMP9:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]
+; UNROLL-NEXT:    [[INC_I]] = add i8 [[INC4_I]], 1
+; UNROLL-NEXT:    [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
+; UNROLL-NEXT:    br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
+; UNROLL:       loopexit:
+; UNROLL-NEXT:    [[AND_I_LCSSA:%.*]] = phi i32 [ [[TMP9]], [[COND_END_I]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i32 [[AND_I_LCSSA]]
+;
+; UNROLL-NO-IC-LABEL: @testoverflowcheck(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = load i32, i32* @d, align 4
+; UNROLL-NO-IC-NEXT:    [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = sub i8 -1, [[DOTPR_I]]
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP3]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP3]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_CRD]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = insertelement <2 x i32> <i32 -1, i32 -1>, i32 [[C_PROMOTED_I]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ <i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = trunc i32 [[INDEX]] to i8
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[DOTPR_I]], [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i8 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i8 [[OFFSET_IDX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP8]] = and <2 x i32> [[BROADCAST_SPLAT]], [[VEC_PHI]]
+; UNROLL-NO-IC-NEXT:    [[TMP9]] = and <2 x i32> [[BROADCAST_SPLAT3]], [[VEC_PHI1]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = and <2 x i32> [[TMP9]], [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[C_PROMOTED_I]], [[ENTRY]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[COND_END_I:%.*]]
+; UNROLL-NO-IC:       cond.end.i:
+; UNROLL-NO-IC-NEXT:    [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
+; UNROLL-NO-IC-NEXT:    [[AND3_I:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[AND_I:%.*]], [[COND_END_I]] ]
+; UNROLL-NO-IC-NEXT:    [[AND_I]] = and i32 [[TMP0]], [[AND3_I]]
+; UNROLL-NO-IC-NEXT:    [[INC_I]] = add i8 [[INC4_I]], 1
+; UNROLL-NO-IC-NEXT:    [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
+; UNROLL-NO-IC:       loopexit:
+; UNROLL-NO-IC-NEXT:    [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[AND_I_LCSSA]]
+;
+; INTERLEAVE-LABEL: @testoverflowcheck(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = load i32, i32* @d, align 4
+; INTERLEAVE-NEXT:    [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -1
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ugt i8 [[DOTPR_I]], -8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP3]], 504
+; INTERLEAVE-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; INTERLEAVE-NEXT:    [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_CRD]]
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = insertelement <4 x i32> <i32 poison, i32 -1, i32 -1, i32 -1>, i32 [[C_PROMOTED_I]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP0]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP0]], i32 0
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = and <4 x i32> [[BROADCAST_SPLATINSERT2]], [[BROADCAST_SPLATINSERT]]
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[BIN_RDX:%.*]] = and <4 x i32> [[TMP7]], [[TMP4]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = call i32 @llvm.vector.reduce.and.v4i32(<4 x i32> [[BIN_RDX]])
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
+; INTERLEAVE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]
+; INTERLEAVE-NEXT:    br label [[COND_END_I:%.*]]
+; INTERLEAVE:       cond.end.i:
+; INTERLEAVE-NEXT:    [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]
+; INTERLEAVE-NEXT:    [[INC_I]] = add i8 [[INC4_I]], 1
+; INTERLEAVE-NEXT:    [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
+; INTERLEAVE-NEXT:    br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
+; INTERLEAVE:       loopexit:
+; INTERLEAVE-NEXT:    [[AND_I_LCSSA:%.*]] = phi i32 [ [[TMP9]], [[COND_END_I]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    ret i32 [[AND_I_LCSSA]]
+;
 entry:
   %.pr.i = load i8, i8* @e, align 1
   %0 = load i32, i32* @d, align 4
@@ -563,21 +3543,411 @@ loopexit:
 ; we need to convert the SCEV expression into an AddRecExpr.
 ; The expression gets converted to {zext i8 %t to i32,+,1}.
 
-; CHECK-LABEL: wrappingindvars1
-; CHECK-LABEL: vector.scevcheck
-; CHECK-LABEL: vector.ph
-; CHECK: %[[START:.*]] = add <2 x i32> %{{.*}}, <i32 0, i32 1>
-; CHECK-LABEL: vector.body
-; CHECK: %[[PHI:.*]] = phi <2 x i32> [ %[[START]], %vector.ph ], [ %[[STEP:.*]], %vector.body ]
-; CHECK: %[[STEP]] = add <2 x i32> %[[PHI]], <i32 2, i32 2>
 define void @wrappingindvars1(i8 %t, i32 %len, i32 *%A) {
- entry:
+; CHECK-LABEL: @wrappingindvars1(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[ST:%.*]] = zext i8 [[T:%.*]] to i16
+; CHECK-NEXT:    [[EXT:%.*]] = zext i8 [[T]] to i32
+; CHECK-NEXT:    [[ECMP:%.*]] = icmp ult i16 [[ST]], 42
+; CHECK-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; CHECK:       vector.scevcheck:
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; CHECK-NEXT:    [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = sub i8 [[T]], [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ugt i8 [[TMP3]], [[T]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp ult i8 [[TMP2]], [[T]]
+; CHECK-NEXT:    [[TMP6:%.*]] = select i1 false, i1 [[TMP4]], i1 [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
+; CHECK-NEXT:    [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = or i1 false, [[TMP8]]
+; CHECK-NEXT:    [[TMP10:%.*]] = trunc i32 [[LEN]] to i8
+; CHECK-NEXT:    [[TMP11:%.*]] = add i8 [[T]], [[TMP10]]
+; CHECK-NEXT:    [[TMP12:%.*]] = sub i8 [[T]], [[TMP10]]
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp sgt i8 [[TMP12]], [[T]]
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp slt i8 [[TMP11]], [[T]]
+; CHECK-NEXT:    [[TMP15:%.*]] = select i1 false, i1 [[TMP13]], i1 [[TMP14]]
+; CHECK-NEXT:    [[TMP16:%.*]] = icmp ugt i32 [[LEN]], 255
+; CHECK-NEXT:    [[TMP17:%.*]] = or i1 [[TMP15]], [[TMP16]]
+; CHECK-NEXT:    [[TMP18:%.*]] = or i1 [[TMP9]], [[TMP17]]
+; CHECK-NEXT:    br i1 [[TMP18]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; CHECK-NEXT:    [[IND_END:%.*]] = add i8 [[T]], [[CAST_CRD]]
+; CHECK-NEXT:    [[IND_END3:%.*]] = add i32 [[EXT]], [[N_VEC]]
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i8> poison, i8 [[T]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i8> [[DOTSPLATINSERT]], <2 x i8> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i8> [[DOTSPLAT]], <i8 0, i8 1>
+; CHECK-NEXT:    [[DOTSPLATINSERT4:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT5:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT4]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION6:%.*]] = add <2 x i32> [[DOTSPLAT5]], <i32 0, i32 1>
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i8> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND7:%.*]] = phi <2 x i32> [ [[INDUCTION6]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT8:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP19:%.*]] = trunc i32 [[INDEX]] to i8
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[T]], [[TMP19]]
+; CHECK-NEXT:    [[TMP20:%.*]] = add i8 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP21:%.*]] = add i8 [[OFFSET_IDX]], 1
+; CHECK-NEXT:    [[TMP22:%.*]] = add i32 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i8 [[TMP20]]
+; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i32 0
+; CHECK-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP24]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND7]], <2 x i32>* [[TMP25]], align 4
+; CHECK-NEXT:    [[TMP26:%.*]] = add <2 x i8> [[VEC_IND]], <i8 1, i8 1>
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i8> [[VEC_IND]], <i8 2, i8 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT8]] = add <2 x i32> [[VEC_IND7]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP27:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END3]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[IDX]]
+; CHECK-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; CHECK-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; CHECK-NEXT:    [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
+; CHECK-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; CHECK-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; CHECK-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @wrappingindvars1(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
+; IND-NEXT:    [[ECMP:%.*]] = icmp ult i8 [[T]], 42
+; IND-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; IND:       loop.preheader:
+; IND-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; IND:       vector.scevcheck:
+; IND-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; IND-NEXT:    [[TMP2:%.*]] = xor i8 [[T]], -1
+; IND-NEXT:    [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
+; IND-NEXT:    [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
+; IND-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
+; IND-NEXT:    [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
+; IND-NEXT:    [[TMP7:%.*]] = add i8 [[TMP6]], [[T]]
+; IND-NEXT:    [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
+; IND-NEXT:    [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
+; IND-NEXT:    [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
+; IND-NEXT:    [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
+; IND-NEXT:    br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -2
+; IND-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; IND-NEXT:    [[IND_END:%.*]] = add i8 [[CAST_CRD]], [[T]]
+; IND-NEXT:    [[IND_END3:%.*]] = add i32 [[N_VEC]], [[EXT]]
+; IND-NEXT:    [[DOTSPLATINSERT4:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i32 0
+; IND-NEXT:    [[DOTSPLAT5:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT4]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    [[INDUCTION6:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT5]], <i32 0, i32 1>
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND7:%.*]] = phi <2 x i32> [ [[INDUCTION6]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT8:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP12:%.*]] = trunc i32 [[INDEX]] to i8
+; IND-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[TMP12]], [[T]]
+; IND-NEXT:    [[TMP13:%.*]] = sext i8 [[OFFSET_IDX]] to i64
+; IND-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP13]]
+; IND-NEXT:    [[TMP15:%.*]] = bitcast i32* [[TMP14]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[VEC_IND7]], <2 x i32>* [[TMP15]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT8]] = add <2 x i32> [[VEC_IND7]], <i32 2, i32 2>
+; IND-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; IND-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; IND-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END3]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
+; IND-NEXT:    br label [[LOOP:%.*]]
+; IND:       loop:
+; IND-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[TMP17:%.*]] = sext i8 [[IDX]] to i64
+; IND-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP17]]
+; IND-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; IND-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; IND-NEXT:    [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
+; IND-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; IND-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; IND-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
+; IND:       exit.loopexit:
+; IND-NEXT:    br label [[EXIT]]
+; IND:       exit:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @wrappingindvars1(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
+; UNROLL-NEXT:    [[ECMP:%.*]] = icmp ult i8 [[T]], 42
+; UNROLL-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; UNROLL:       loop.preheader:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; UNROLL:       vector.scevcheck:
+; UNROLL-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; UNROLL-NEXT:    [[TMP2:%.*]] = xor i8 [[T]], -1
+; UNROLL-NEXT:    [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
+; UNROLL-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
+; UNROLL-NEXT:    [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
+; UNROLL-NEXT:    [[TMP7:%.*]] = add i8 [[TMP6]], [[T]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
+; UNROLL-NEXT:    [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
+; UNROLL-NEXT:    [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
+; UNROLL-NEXT:    br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -4
+; UNROLL-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; UNROLL-NEXT:    [[IND_END:%.*]] = add i8 [[CAST_CRD]], [[T]]
+; UNROLL-NEXT:    [[IND_END3:%.*]] = add i32 [[N_VEC]], [[EXT]]
+; UNROLL-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i32 0
+; UNROLL-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT5]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[INDUCTION7:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT6]], <i32 0, i32 1>
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND8:%.*]] = phi <2 x i32> [ [[INDUCTION7]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP12:%.*]] = trunc i32 [[INDEX]] to i8
+; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[TMP12]], [[T]]
+; UNROLL-NEXT:    [[STEP_ADD9:%.*]] = add <2 x i32> [[VEC_IND8]], <i32 2, i32 2>
+; UNROLL-NEXT:    [[TMP13:%.*]] = sext i8 [[OFFSET_IDX]] to i64
+; UNROLL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP13]]
+; UNROLL-NEXT:    [[TMP15:%.*]] = bitcast i32* [[TMP14]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[VEC_IND8]], <2 x i32>* [[TMP15]], align 4
+; UNROLL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP14]], i64 2
+; UNROLL-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP16]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[STEP_ADD9]], <2 x i32>* [[TMP17]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT11]] = add <2 x i32> [[VEC_IND8]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END3]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NEXT:    br label [[LOOP:%.*]]
+; UNROLL:       loop:
+; UNROLL-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[TMP19:%.*]] = sext i8 [[IDX]] to i64
+; UNROLL-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP19]]
+; UNROLL-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; UNROLL-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; UNROLL-NEXT:    [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
+; UNROLL-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; UNROLL-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; UNROLL-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
+; UNROLL:       exit.loopexit:
+; UNROLL-NEXT:    br label [[EXIT]]
+; UNROLL:       exit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @wrappingindvars1(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[ST:%.*]] = zext i8 [[T:%.*]] to i16
+; UNROLL-NO-IC-NEXT:    [[EXT:%.*]] = zext i8 [[T]] to i32
+; UNROLL-NO-IC-NEXT:    [[ECMP:%.*]] = icmp ult i16 [[ST]], 42
+; UNROLL-NO-IC-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; UNROLL-NO-IC:       loop.preheader:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; UNROLL-NO-IC:       vector.scevcheck:
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = sub i8 [[T]], [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = icmp ugt i8 [[TMP3]], [[T]]
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = icmp ult i8 [[TMP2]], [[T]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = select i1 false, i1 [[TMP4]], i1 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = or i1 false, [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = trunc i32 [[LEN]] to i8
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = add i8 [[T]], [[TMP10]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = sub i8 [[T]], [[TMP10]]
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = icmp sgt i8 [[TMP12]], [[T]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = icmp slt i8 [[TMP11]], [[T]]
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = select i1 false, i1 [[TMP13]], i1 [[TMP14]]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = icmp ugt i32 [[LEN]], 255
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = or i1 [[TMP15]], [[TMP16]]
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = or i1 [[TMP9]], [[TMP17]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP18]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = add i8 [[T]], [[CAST_CRD]]
+; UNROLL-NO-IC-NEXT:    [[IND_END3:%.*]] = add i32 [[EXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i8> poison, i8 [[T]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i8> [[DOTSPLATINSERT]], <2 x i8> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[INDUCTION:%.*]] = add <2 x i8> [[DOTSPLAT]], <i8 0, i8 1>
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT5]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[INDUCTION7:%.*]] = add <2 x i32> [[DOTSPLAT6]], <i32 0, i32 1>
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i8> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND8:%.*]] = phi <2 x i32> [ [[INDUCTION7]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i8> [[VEC_IND]], <i8 2, i8 2>
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = trunc i32 [[INDEX]] to i8
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[T]], [[TMP19]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = add i8 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = add i8 [[OFFSET_IDX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = add i8 [[OFFSET_IDX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = add i8 [[OFFSET_IDX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = add i32 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD9:%.*]] = add <2 x i32> [[VEC_IND8]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i8 [[TMP20]]
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[TMP22]]
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = bitcast i32* [[TMP28]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND8]], <2 x i32>* [[TMP29]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[TMP26]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = bitcast i32* [[TMP30]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD9]], <2 x i32>* [[TMP31]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = add <2 x i8> [[VEC_IND]], <i8 1, i8 1>
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = add <2 x i8> [[STEP_ADD]], <i8 1, i8 1>
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i8> [[STEP_ADD]], <i8 2, i8 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT11]] = add <2 x i32> [[STEP_ADD9]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP34]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END3]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
+; UNROLL-NO-IC:       loop:
+; UNROLL-NO-IC-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[IDX]]
+; UNROLL-NO-IC-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; UNROLL-NO-IC-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; UNROLL-NO-IC-NEXT:    [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
+; UNROLL-NO-IC-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; UNROLL-NO-IC-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; UNROLL-NO-IC-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
+; UNROLL-NO-IC:       exit.loopexit:
+; UNROLL-NO-IC-NEXT:    br label [[EXIT]]
+; UNROLL-NO-IC:       exit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @wrappingindvars1(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
+; INTERLEAVE-NEXT:    [[ECMP:%.*]] = icmp ult i8 [[T]], 42
+; INTERLEAVE-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; INTERLEAVE:       loop.preheader:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; INTERLEAVE:       vector.scevcheck:
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = xor i8 [[T]], -1
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = add i8 [[TMP6]], [[T]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
+; INTERLEAVE-NEXT:    br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -8
+; INTERLEAVE-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; INTERLEAVE-NEXT:    [[IND_END:%.*]] = add i8 [[CAST_CRD]], [[T]]
+; INTERLEAVE-NEXT:    [[IND_END3:%.*]] = add i32 [[N_VEC]], [[EXT]]
+; INTERLEAVE-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <4 x i32> poison, i32 [[EXT]], i32 0
+; INTERLEAVE-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT5]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[INDUCTION7:%.*]] = add nuw nsw <4 x i32> [[DOTSPLAT6]], <i32 0, i32 1, i32 2, i32 3>
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND8:%.*]] = phi <4 x i32> [ [[INDUCTION7]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = trunc i32 [[INDEX]] to i8
+; INTERLEAVE-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[TMP12]], [[T]]
+; INTERLEAVE-NEXT:    [[STEP_ADD9:%.*]] = add <4 x i32> [[VEC_IND8]], <i32 4, i32 4, i32 4, i32 4>
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = sext i8 [[OFFSET_IDX]] to i64
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP13]]
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = bitcast i32* [[TMP14]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[VEC_IND8]], <4 x i32>* [[TMP15]], align 4
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP14]], i64 4
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP16]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[STEP_ADD9]], <4 x i32>* [[TMP17]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT11]] = add <4 x i32> [[VEC_IND8]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END3]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
+; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
+; INTERLEAVE:       loop:
+; INTERLEAVE-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = sext i8 [[IDX]] to i64
+; INTERLEAVE-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP19]]
+; INTERLEAVE-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; INTERLEAVE-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; INTERLEAVE-NEXT:    [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
+; INTERLEAVE-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; INTERLEAVE-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
+; INTERLEAVE:       exit.loopexit:
+; INTERLEAVE-NEXT:    br label [[EXIT]]
+; INTERLEAVE:       exit:
+; INTERLEAVE-NEXT:    ret void
+;
+  entry:
   %st = zext i8 %t to i16
   %ext = zext i8 %t to i32
   %ecmp = icmp ult i16 %st, 42
   br i1 %ecmp, label %loop, label %exit
 
- loop:
+  loop:
 
   %idx = phi i8 [ %t, %entry ], [ %idx.inc, %loop ]
   %idx.b = phi i32 [ 0, %entry ], [ %idx.b.inc, %loop ]
@@ -593,7 +3963,7 @@ define void @wrappingindvars1(i8 %t, i32 %len, i32 *%A) {
   %c = icmp ult i32 %idx.b, %len
   br i1 %c, label %loop, label %exit
 
- exit:
+  exit:
   ret void
 }
 
@@ -601,15 +3971,422 @@ define void @wrappingindvars1(i8 %t, i32 %len, i32 *%A) {
 ; In order to recognize %sphi as an induction PHI and vectorize this loop,
 ; we need to convert the SCEV expression into an AddRecExpr.
 ; The expression gets converted to ({4 * (zext %t to i32),+,4}).
-; CHECK-LABEL: wrappingindvars2
-; CHECK-LABEL: vector.scevcheck
-; CHECK-LABEL: vector.ph
-; CHECK: %[[START:.*]] = add <2 x i32> %{{.*}}, <i32 0, i32 4>
-; CHECK-LABEL: vector.body
-; CHECK: %[[PHI:.*]] = phi <2 x i32> [ %[[START]], %vector.ph ], [ %[[STEP:.*]], %vector.body ]
-; CHECK: %[[STEP]] = add <2 x i32> %[[PHI]], <i32 8, i32 8>
 define void @wrappingindvars2(i8 %t, i32 %len, i32 *%A) {
-
+; CHECK-LABEL: @wrappingindvars2(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[ST:%.*]] = zext i8 [[T:%.*]] to i16
+; CHECK-NEXT:    [[EXT:%.*]] = zext i8 [[T]] to i32
+; CHECK-NEXT:    [[EXT_MUL:%.*]] = mul i32 [[EXT]], 4
+; CHECK-NEXT:    [[ECMP:%.*]] = icmp ult i16 [[ST]], 42
+; CHECK-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; CHECK:       vector.scevcheck:
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; CHECK-NEXT:    [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = sub i8 [[T]], [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ugt i8 [[TMP3]], [[T]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp ult i8 [[TMP2]], [[T]]
+; CHECK-NEXT:    [[TMP6:%.*]] = select i1 false, i1 [[TMP4]], i1 [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
+; CHECK-NEXT:    [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = or i1 false, [[TMP8]]
+; CHECK-NEXT:    [[TMP10:%.*]] = trunc i32 [[LEN]] to i8
+; CHECK-NEXT:    [[TMP11:%.*]] = add i8 [[T]], [[TMP10]]
+; CHECK-NEXT:    [[TMP12:%.*]] = sub i8 [[T]], [[TMP10]]
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp sgt i8 [[TMP12]], [[T]]
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp slt i8 [[TMP11]], [[T]]
+; CHECK-NEXT:    [[TMP15:%.*]] = select i1 false, i1 [[TMP13]], i1 [[TMP14]]
+; CHECK-NEXT:    [[TMP16:%.*]] = icmp ugt i32 [[LEN]], 255
+; CHECK-NEXT:    [[TMP17:%.*]] = or i1 [[TMP15]], [[TMP16]]
+; CHECK-NEXT:    [[TMP18:%.*]] = or i1 [[TMP9]], [[TMP17]]
+; CHECK-NEXT:    br i1 [[TMP18]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; CHECK-NEXT:    [[IND_END:%.*]] = add i8 [[T]], [[CAST_CRD]]
+; CHECK-NEXT:    [[TMP19:%.*]] = mul i32 [[N_VEC]], 4
+; CHECK-NEXT:    [[IND_END2:%.*]] = add i32 [[EXT_MUL]], [[TMP19]]
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i8> poison, i8 [[T]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i8> [[DOTSPLATINSERT]], <2 x i8> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i8> [[DOTSPLAT]], <i8 0, i8 1>
+; CHECK-NEXT:    [[DOTSPLATINSERT4:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT5:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT4]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION6:%.*]] = add <2 x i32> [[DOTSPLAT5]], <i32 0, i32 4>
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i8> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND7:%.*]] = phi <2 x i32> [ [[INDUCTION6]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT8:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP20:%.*]] = trunc i32 [[INDEX]] to i8
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[T]], [[TMP20]]
+; CHECK-NEXT:    [[TMP21:%.*]] = add i8 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP22:%.*]] = add i8 [[OFFSET_IDX]], 1
+; CHECK-NEXT:    [[TMP23:%.*]] = add i32 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i8 [[TMP21]]
+; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[TMP24]], i32 0
+; CHECK-NEXT:    [[TMP26:%.*]] = bitcast i32* [[TMP25]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND7]], <2 x i32>* [[TMP26]], align 4
+; CHECK-NEXT:    [[TMP27:%.*]] = add <2 x i8> [[VEC_IND]], <i8 1, i8 1>
+; CHECK-NEXT:    [[TMP28:%.*]] = zext <2 x i8> [[TMP27]] to <2 x i32>
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i8> [[VEC_IND]], <i8 2, i8 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT8]] = add <2 x i32> [[VEC_IND7]], <i32 8, i32 8>
+; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[IDX]]
+; CHECK-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; CHECK-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; CHECK-NEXT:    [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
+; CHECK-NEXT:    [[MUL]] = mul i32 [[IDX_INC_EXT]], 4
+; CHECK-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; CHECK-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; CHECK-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @wrappingindvars2(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
+; IND-NEXT:    [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 2
+; IND-NEXT:    [[ECMP:%.*]] = icmp ult i8 [[T]], 42
+; IND-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; IND:       loop.preheader:
+; IND-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; IND:       vector.scevcheck:
+; IND-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; IND-NEXT:    [[TMP2:%.*]] = xor i8 [[T]], -1
+; IND-NEXT:    [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
+; IND-NEXT:    [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
+; IND-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
+; IND-NEXT:    [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
+; IND-NEXT:    [[TMP7:%.*]] = add i8 [[TMP6]], [[T]]
+; IND-NEXT:    [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
+; IND-NEXT:    [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
+; IND-NEXT:    [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
+; IND-NEXT:    [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
+; IND-NEXT:    br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -2
+; IND-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; IND-NEXT:    [[IND_END:%.*]] = add i8 [[CAST_CRD]], [[T]]
+; IND-NEXT:    [[TMP12:%.*]] = add i32 [[N_VEC]], [[EXT]]
+; IND-NEXT:    [[IND_END2:%.*]] = shl i32 [[TMP12]], 2
+; IND-NEXT:    [[DOTSPLATINSERT4:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i32 0
+; IND-NEXT:    [[DOTSPLAT5:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT4]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    [[INDUCTION6:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT5]], <i32 0, i32 4>
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND7:%.*]] = phi <2 x i32> [ [[INDUCTION6]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT8:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP13:%.*]] = trunc i32 [[INDEX]] to i8
+; IND-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[TMP13]], [[T]]
+; IND-NEXT:    [[TMP14:%.*]] = sext i8 [[OFFSET_IDX]] to i64
+; IND-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP14]]
+; IND-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[VEC_IND7]], <2 x i32>* [[TMP16]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT8]] = add <2 x i32> [[VEC_IND7]], <i32 8, i32 8>
+; IND-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; IND-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
+; IND-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; IND-NEXT:    br label [[LOOP:%.*]]
+; IND:       loop:
+; IND-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[TMP18:%.*]] = sext i8 [[IDX]] to i64
+; IND-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
+; IND-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; IND-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; IND-NEXT:    [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
+; IND-NEXT:    [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 2
+; IND-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; IND-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; IND-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
+; IND:       exit.loopexit:
+; IND-NEXT:    br label [[EXIT]]
+; IND:       exit:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @wrappingindvars2(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
+; UNROLL-NEXT:    [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 2
+; UNROLL-NEXT:    [[ECMP:%.*]] = icmp ult i8 [[T]], 42
+; UNROLL-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; UNROLL:       loop.preheader:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; UNROLL:       vector.scevcheck:
+; UNROLL-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; UNROLL-NEXT:    [[TMP2:%.*]] = xor i8 [[T]], -1
+; UNROLL-NEXT:    [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
+; UNROLL-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
+; UNROLL-NEXT:    [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
+; UNROLL-NEXT:    [[TMP7:%.*]] = add i8 [[TMP6]], [[T]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
+; UNROLL-NEXT:    [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
+; UNROLL-NEXT:    [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
+; UNROLL-NEXT:    br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -4
+; UNROLL-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; UNROLL-NEXT:    [[IND_END:%.*]] = add i8 [[CAST_CRD]], [[T]]
+; UNROLL-NEXT:    [[TMP12:%.*]] = add i32 [[N_VEC]], [[EXT]]
+; UNROLL-NEXT:    [[IND_END2:%.*]] = shl i32 [[TMP12]], 2
+; UNROLL-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i32 0
+; UNROLL-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT5]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[INDUCTION7:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT6]], <i32 0, i32 4>
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND8:%.*]] = phi <2 x i32> [ [[INDUCTION7]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[TMP13:%.*]] = trunc i32 [[INDEX]] to i8
+; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[TMP13]], [[T]]
+; UNROLL-NEXT:    [[STEP_ADD9:%.*]] = add <2 x i32> [[VEC_IND8]], <i32 8, i32 8>
+; UNROLL-NEXT:    [[TMP14:%.*]] = sext i8 [[OFFSET_IDX]] to i64
+; UNROLL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP14]]
+; UNROLL-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[VEC_IND8]], <2 x i32>* [[TMP16]], align 4
+; UNROLL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i64 2
+; UNROLL-NEXT:    [[TMP18:%.*]] = bitcast i32* [[TMP17]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[STEP_ADD9]], <2 x i32>* [[TMP18]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT11]] = add <2 x i32> [[VEC_IND8]], <i32 16, i32 16>
+; UNROLL-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NEXT:    br label [[LOOP:%.*]]
+; UNROLL:       loop:
+; UNROLL-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[TMP20:%.*]] = sext i8 [[IDX]] to i64
+; UNROLL-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
+; UNROLL-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; UNROLL-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; UNROLL-NEXT:    [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
+; UNROLL-NEXT:    [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 2
+; UNROLL-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; UNROLL-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; UNROLL-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
+; UNROLL:       exit.loopexit:
+; UNROLL-NEXT:    br label [[EXIT]]
+; UNROLL:       exit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @wrappingindvars2(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[ST:%.*]] = zext i8 [[T:%.*]] to i16
+; UNROLL-NO-IC-NEXT:    [[EXT:%.*]] = zext i8 [[T]] to i32
+; UNROLL-NO-IC-NEXT:    [[EXT_MUL:%.*]] = mul i32 [[EXT]], 4
+; UNROLL-NO-IC-NEXT:    [[ECMP:%.*]] = icmp ult i16 [[ST]], 42
+; UNROLL-NO-IC-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; UNROLL-NO-IC:       loop.preheader:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; UNROLL-NO-IC:       vector.scevcheck:
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = sub i8 [[T]], [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = icmp ugt i8 [[TMP3]], [[T]]
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = icmp ult i8 [[TMP2]], [[T]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = select i1 false, i1 [[TMP4]], i1 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = or i1 false, [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = trunc i32 [[LEN]] to i8
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = add i8 [[T]], [[TMP10]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = sub i8 [[T]], [[TMP10]]
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = icmp sgt i8 [[TMP12]], [[T]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = icmp slt i8 [[TMP11]], [[T]]
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = select i1 false, i1 [[TMP13]], i1 [[TMP14]]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = icmp ugt i32 [[LEN]], 255
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = or i1 [[TMP15]], [[TMP16]]
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = or i1 [[TMP9]], [[TMP17]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP18]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = add i8 [[T]], [[CAST_CRD]]
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = mul i32 [[N_VEC]], 4
+; UNROLL-NO-IC-NEXT:    [[IND_END2:%.*]] = add i32 [[EXT_MUL]], [[TMP19]]
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i8> poison, i8 [[T]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i8> [[DOTSPLATINSERT]], <2 x i8> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[INDUCTION:%.*]] = add <2 x i8> [[DOTSPLAT]], <i8 0, i8 1>
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT5]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[INDUCTION7:%.*]] = add <2 x i32> [[DOTSPLAT6]], <i32 0, i32 4>
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i8> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND8:%.*]] = phi <2 x i32> [ [[INDUCTION7]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i8> [[VEC_IND]], <i8 2, i8 2>
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = trunc i32 [[INDEX]] to i8
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[T]], [[TMP20]]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = add i8 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = add i8 [[OFFSET_IDX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = add i8 [[OFFSET_IDX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = add i8 [[OFFSET_IDX]], 3
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD9:%.*]] = add <2 x i32> [[VEC_IND8]], <i32 8, i32 8>
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = add i32 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i8 [[TMP21]]
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[TMP23]]
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, i32* [[TMP27]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = bitcast i32* [[TMP29]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND8]], <2 x i32>* [[TMP30]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, i32* [[TMP27]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = bitcast i32* [[TMP31]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD9]], <2 x i32>* [[TMP32]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = add <2 x i8> [[VEC_IND]], <i8 1, i8 1>
+; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = add <2 x i8> [[STEP_ADD]], <i8 1, i8 1>
+; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = zext <2 x i8> [[TMP33]] to <2 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = zext <2 x i8> [[TMP34]] to <2 x i32>
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i8> [[STEP_ADD]], <i8 2, i8 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT11]] = add <2 x i32> [[STEP_ADD9]], <i32 8, i32 8>
+; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP37]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
+; UNROLL-NO-IC:       loop:
+; UNROLL-NO-IC-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[IDX]]
+; UNROLL-NO-IC-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; UNROLL-NO-IC-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; UNROLL-NO-IC-NEXT:    [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
+; UNROLL-NO-IC-NEXT:    [[MUL]] = mul i32 [[IDX_INC_EXT]], 4
+; UNROLL-NO-IC-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; UNROLL-NO-IC-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; UNROLL-NO-IC-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
+; UNROLL-NO-IC:       exit.loopexit:
+; UNROLL-NO-IC-NEXT:    br label [[EXIT]]
+; UNROLL-NO-IC:       exit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @wrappingindvars2(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
+; INTERLEAVE-NEXT:    [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 2
+; INTERLEAVE-NEXT:    [[ECMP:%.*]] = icmp ult i8 [[T]], 42
+; INTERLEAVE-NEXT:    br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
+; INTERLEAVE:       loop.preheader:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; INTERLEAVE:       vector.scevcheck:
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = xor i8 [[T]], -1
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = add i8 [[TMP6]], [[T]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
+; INTERLEAVE-NEXT:    br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -8
+; INTERLEAVE-NEXT:    [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
+; INTERLEAVE-NEXT:    [[IND_END:%.*]] = add i8 [[CAST_CRD]], [[T]]
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = add i32 [[N_VEC]], [[EXT]]
+; INTERLEAVE-NEXT:    [[IND_END2:%.*]] = shl i32 [[TMP12]], 2
+; INTERLEAVE-NEXT:    [[DOTSPLATINSERT5:%.*]] = insertelement <4 x i32> poison, i32 [[EXT_MUL]], i32 0
+; INTERLEAVE-NEXT:    [[DOTSPLAT6:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT5]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[INDUCTION7:%.*]] = add nuw nsw <4 x i32> [[DOTSPLAT6]], <i32 0, i32 4, i32 8, i32 12>
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND8:%.*]] = phi <4 x i32> [ [[INDUCTION7]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = trunc i32 [[INDEX]] to i8
+; INTERLEAVE-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[TMP13]], [[T]]
+; INTERLEAVE-NEXT:    [[STEP_ADD9:%.*]] = add <4 x i32> [[VEC_IND8]], <i32 16, i32 16, i32 16, i32 16>
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = sext i8 [[OFFSET_IDX]] to i64
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP14]]
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[VEC_IND8]], <4 x i32>* [[TMP16]], align 4
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i64 4
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = bitcast i32* [[TMP17]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[STEP_ADD9]], <4 x i32>* [[TMP18]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT11]] = add <4 x i32> [[VEC_IND8]], <i32 32, i32 32, i32 32, i32 32>
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
+; INTERLEAVE:       loop:
+; INTERLEAVE-NEXT:    [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[TMP20:%.*]] = sext i8 [[IDX]] to i64
+; INTERLEAVE-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
+; INTERLEAVE-NEXT:    store i32 [[SPHI]], i32* [[PTR]], align 4
+; INTERLEAVE-NEXT:    [[IDX_INC]] = add i8 [[IDX]], 1
+; INTERLEAVE-NEXT:    [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
+; INTERLEAVE-NEXT:    [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 2
+; INTERLEAVE-NEXT:    [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
+; INTERLEAVE-NEXT:    [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
+; INTERLEAVE-NEXT:    br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
+; INTERLEAVE:       exit.loopexit:
+; INTERLEAVE-NEXT:    br label [[EXIT]]
+; INTERLEAVE:       exit:
+; INTERLEAVE-NEXT:    ret void
+;
 entry:
   %st = zext i8 %t to i16
   %ext = zext i8 %t to i32
@@ -618,7 +4395,7 @@ entry:
   %ecmp = icmp ult i16 %st, 42
   br i1 %ecmp, label %loop, label %exit
 
- loop:
+  loop:
 
   %idx = phi i8 [ %t, %entry ], [ %idx.inc, %loop ]
   %sphi = phi i32 [ %ext.mul, %entry ], [%mul, %loop]
@@ -635,31 +4412,208 @@ entry:
   %c = icmp ult i32 %idx.b, %len
   br i1 %c, label %loop, label %exit
 
- exit:
+  exit:
   ret void
 }
 
 ; Check that we generate vectorized IVs in the pre-header
 ; instead of widening the scalar IV inside the loop, when
 ; we know how to do that.
-; IND-LABEL: veciv
-; IND: vector.body:
-; IND: %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; IND: %vec.ind = phi <2 x i32> [ <i32 0, i32 1>, %vector.ph ], [ %vec.ind.next, %vector.body ]
-; IND: %index.next = add nuw i32 %index, 2
-; IND: %vec.ind.next = add <2 x i32> %vec.ind, <i32 2, i32 2>
-; IND: %[[CMP:.*]] = icmp eq i32 %index.next
-; IND: br i1 %[[CMP]]
-; UNROLL-LABEL: veciv
-; UNROLL: vector.body:
-; UNROLL: %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; UNROLL: %vec.ind = phi <2 x i32> [ <i32 0, i32 1>, %vector.ph ], [ %vec.ind.next, %vector.body ]
-; UNROLL: %step.add = add <2 x i32> %vec.ind, <i32 2, i32 2>
-; UNROLL: %index.next = add nuw i32 %index, 4
-; UNROLL: %vec.ind.next = add <2 x i32> %vec.ind, <i32 4, i32 4>
-; UNROLL: %[[CMP:.*]] = icmp eq i32 %index.next
-; UNROLL: br i1 %[[CMP]]
 define void @veciv(i32* nocapture %a, i32 %start, i32 %k) {
+; CHECK-LABEL: @veciv(
+; CHECK-NEXT:  for.body.preheader:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP4]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[INDVARS_IV]]
+; CHECK-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @veciv(
+; IND-NEXT:  for.body.preheader:
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i32 [[K]], -2
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
+; IND-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
+; IND-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP2]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; IND-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N_VEC]], [[K]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[TMP4:%.*]] = sext i32 [[INDVARS_IV]] to i64
+; IND-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
+; IND-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; IND-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; IND-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; IND-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; IND:       exit:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @veciv(
+; UNROLL-NEXT:  for.body.preheader:
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i32 [[K]], -4
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; UNROLL-NEXT:    [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
+; UNROLL-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
+; UNROLL-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP2]], align 4
+; UNROLL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 2
+; UNROLL-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP4]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N_VEC]], [[K]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[TMP6:%.*]] = sext i32 [[INDVARS_IV]] to i64
+; UNROLL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP6]]
+; UNROLL-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; UNROLL:       exit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @veciv(
+; UNROLL-NO-IC-NEXT:  for.body.preheader:
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[K]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i32 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i32 [[INDEX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP2]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP7]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = bitcast i32* [[TMP8]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP9]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; UNROLL-NO-IC:       exit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @veciv(
+; INTERLEAVE-NEXT:  for.body.preheader:
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i32 [[K]], -8
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP2]], align 4
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP4]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N_VEC]], [[K]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = sext i32 [[INDVARS_IV]] to i64
+; INTERLEAVE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP6]]
+; INTERLEAVE-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; INTERLEAVE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; INTERLEAVE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; INTERLEAVE:       exit:
+; INTERLEAVE-NEXT:    ret void
+;
 for.body.preheader:
   br label %for.body
 
@@ -675,15 +4629,255 @@ exit:
   ret void
 }
 
-; IND-LABEL: trunciv
-; IND: vector.body:
-; IND: %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; IND: %[[VECIND:.*]] = phi <2 x i32> [ <i32 0, i32 1>, %vector.ph ], [ %[[STEPADD:.*]], %vector.body ]
-; IND: %index.next = add nuw i64 %index, 2
-; IND: %[[STEPADD]] = add <2 x i32> %[[VECIND]], <i32 2, i32 2>
-; IND: %[[CMP:.*]] = icmp eq i64 %index.next
-; IND: br i1 %[[CMP]]
 define void @trunciv(i32* nocapture %a, i32 %start, i64 %k) {
+; CHECK-LABEL: @trunciv(
+; CHECK-NEXT:  for.body.preheader:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; CHECK:       vector.scevcheck:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[K]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[TMP0]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = add i32 0, [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = sub i32 0, [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp slt i32 [[TMP2]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = select i1 false, i1 [[TMP4]], i1 [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp ugt i64 [[TMP0]], 4294967295
+; CHECK-NEXT:    [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = or i1 false, [[TMP8]]
+; CHECK-NEXT:    br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[K]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[K]], [[N_MOD_VF]]
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND1:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP10:%.*]] = trunc i64 [[INDEX]] to i32
+; CHECK-NEXT:    [[TMP11:%.*]] = add i32 [[TMP10]], 0
+; CHECK-NEXT:    [[TMP12:%.*]] = add i32 [[TMP10]], 1
+; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP11]]
+; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP13]], i32 0
+; CHECK-NEXT:    [[TMP15:%.*]] = bitcast i32* [[TMP14]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND1]], <2 x i32>* [[TMP15]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i32> [[VEC_IND1]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TRUNC_IV]]
+; CHECK-NEXT:    store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @trunciv(
+; IND-NEXT:  for.body.preheader:
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; IND:       vector.scevcheck:
+; IND-NEXT:    [[TMP0:%.*]] = add i64 [[K]], -2147483649
+; IND-NEXT:    [[TMP1:%.*]] = icmp ult i64 [[TMP0]], -2147483648
+; IND-NEXT:    br i1 [[TMP1]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[K]], -2
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND1:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[SEXT:%.*]] = shl i64 [[INDEX]], 32
+; IND-NEXT:    [[TMP2:%.*]] = ashr exact i64 [[SEXT]], 32
+; IND-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP2]]
+; IND-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[VEC_IND1]], <2 x i32>* [[TMP4]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i32> [[VEC_IND1]], <i32 2, i32 2>
+; IND-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[K]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; IND-NEXT:    [[SEXT3:%.*]] = shl i64 [[INDVARS_IV]], 32
+; IND-NEXT:    [[TMP6:%.*]] = ashr exact i64 [[SEXT3]], 32
+; IND-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP6]]
+; IND-NEXT:    store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
+; IND-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; IND-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
+; IND-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
+; IND:       exit:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @trunciv(
+; UNROLL-NEXT:  for.body.preheader:
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; UNROLL:       vector.scevcheck:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i64 [[K]], -2147483649
+; UNROLL-NEXT:    [[TMP1:%.*]] = icmp ult i64 [[TMP0]], -2147483648
+; UNROLL-NEXT:    br i1 [[TMP1]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[K]], -4
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND2:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[STEP_ADD3:%.*]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
+; UNROLL-NEXT:    [[SEXT:%.*]] = shl i64 [[INDEX]], 32
+; UNROLL-NEXT:    [[TMP2:%.*]] = ashr exact i64 [[SEXT]], 32
+; UNROLL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP2]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[VEC_IND2]], <2 x i32>* [[TMP4]], align 4
+; UNROLL-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i64 2
+; UNROLL-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[STEP_ADD3]], <2 x i32>* [[TMP6]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT5]] = add <2 x i32> [[VEC_IND2]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[K]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; UNROLL-NEXT:    [[SEXT6:%.*]] = shl i64 [[INDVARS_IV]], 32
+; UNROLL-NEXT:    [[TMP8:%.*]] = ashr exact i64 [[SEXT6]], 32
+; UNROLL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
+; UNROLL-NEXT:    store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
+; UNROLL:       exit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @trunciv(
+; UNROLL-NO-IC-NEXT:  for.body.preheader:
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; UNROLL-NO-IC:       vector.scevcheck:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[K]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = trunc i64 [[TMP0]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i32 0, [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = sub i32 0, [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP3]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = icmp slt i32 [[TMP2]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = select i1 false, i1 [[TMP4]], i1 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = icmp ugt i64 [[TMP0]], 4294967295
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = or i1 false, [[TMP8]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[K]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[K]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND2:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD3:%.*]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = trunc i64 [[INDEX]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = add i32 [[TMP10]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add i32 [[TMP10]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = add i32 [[TMP10]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = add i32 [[TMP10]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP11]]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP13]]
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = bitcast i32* [[TMP17]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND2]], <2 x i32>* [[TMP18]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = bitcast i32* [[TMP19]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD3]], <2 x i32>* [[TMP20]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT5]] = add <2 x i32> [[STEP_ADD3]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TRUNC_IV]]
+; UNROLL-NO-IC-NEXT:    store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
+; UNROLL-NO-IC:       exit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @trunciv(
+; INTERLEAVE-NEXT:  for.body.preheader:
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; INTERLEAVE:       vector.scevcheck:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add i64 [[K]], -2147483649
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = icmp ult i64 [[TMP0]], -2147483648
+; INTERLEAVE-NEXT:    br i1 [[TMP1]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i64 [[K]], -8
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND2:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[STEP_ADD3:%.*]] = add <4 x i32> [[VEC_IND2]], <i32 4, i32 4, i32 4, i32 4>
+; INTERLEAVE-NEXT:    [[SEXT:%.*]] = shl i64 [[INDEX]], 32
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = ashr exact i64 [[SEXT]], 32
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP2]]
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[VEC_IND2]], <4 x i32>* [[TMP4]], align 4
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i64 4
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[STEP_ADD3]], <4 x i32>* [[TMP6]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT5]] = add <4 x i32> [[VEC_IND2]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[K]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; INTERLEAVE-NEXT:    [[SEXT6:%.*]] = shl i64 [[INDVARS_IV]], 32
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = ashr exact i64 [[SEXT6]], 32
+; INTERLEAVE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
+; INTERLEAVE-NEXT:    store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
+; INTERLEAVE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; INTERLEAVE-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
+; INTERLEAVE:       exit:
+; INTERLEAVE-NEXT:    ret void
+;
 for.body.preheader:
   br label %for.body
 
@@ -700,65 +4894,233 @@ exit:
   ret void
 }
 
+;
+;
+define void @nonprimary(i32* nocapture %a, i32 %start, i32 %i, i32 %k) {
 ; CHECK-LABEL: @nonprimary(
-; CHECK: vector.ph:
-; CHECK:   %[[INSERT:.*]] = insertelement <2 x i32> poison, i32 %i, i32 0
-; CHECK:   %[[SPLAT:.*]] = shufflevector <2 x i32> %[[INSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
-; CHECK:   %[[START:.*]] = add <2 x i32> %[[SPLAT]], <i32 0, i32 1>
-; CHECK: vector.body:
-; CHECK:   %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK:   %vec.ind = phi <2 x i32> [ %[[START]], %vector.ph ], [ %vec.ind.next, %vector.body ]
-; CHECK:   %offset.idx = add i32 %i, %index
-; CHECK:   %[[A1:.*]] = add i32 %offset.idx, 0
-; CHECK:   %[[G1:.*]] = getelementptr inbounds i32, i32* %a, i32 %[[A1]]
-; CHECK:   %[[G3:.*]] = getelementptr inbounds i32, i32* %[[G1]], i32 0
-; CHECK:   %[[B1:.*]] = bitcast i32* %[[G3]] to <2 x i32>*
-; CHECK:   store <2 x i32> %vec.ind, <2 x i32>* %[[B1]]
-; CHECK:   %index.next = add nuw i32 %index, 2
-; CHECK:   %vec.ind.next = add <2 x i32> %vec.ind, <i32 2, i32 2>
-; CHECK:   %[[CMP:.*]] = icmp eq i32 %index.next, %n.vec
-; CHECK:   br i1 %[[CMP]]
+; CHECK-NEXT:  for.body.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], 1
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i32 0
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP5]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[INDVARS_IV]]
+; CHECK-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
 ;
 ; IND-LABEL: @nonprimary(
-; IND: vector.ph:
-; IND:   %[[INSERT:.*]] = insertelement <2 x i32> poison, i32 %i, i32 0
-; IND:   %[[SPLAT:.*]] = shufflevector <2 x i32> %[[INSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
-; IND:   %[[START:.*]] = add <2 x i32> %[[SPLAT]], <i32 0, i32 1>
-; IND: vector.body:
-; IND:   %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; IND:   %vec.ind = phi <2 x i32> [ %[[START]], %vector.ph ], [ %vec.ind.next, %vector.body ]
-; IND:   %[[A1:.*]] = add i32 %index, %i
-; IND:   %[[S1:.*]] = sext i32 %[[A1]] to i64
-; IND:   %[[G1:.*]] = getelementptr inbounds i32, i32* %a, i64 %[[S1]]
-; IND:   %[[B1:.*]] = bitcast i32* %[[G1]] to <2 x i32>*
-; IND:   store <2 x i32> %vec.ind, <2 x i32>* %[[B1]]
-; IND:   %index.next = add nuw i32 %index, 2
-; IND:   %vec.ind.next = add <2 x i32> %vec.ind, <i32 2, i32 2>
-; IND:   %[[CMP:.*]] = icmp eq i32 %index.next, %n.vec
-; IND:   br i1 %[[CMP]]
+; IND-NEXT:  for.body.preheader:
+; IND-NEXT:    [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -2
+; IND-NEXT:    [[IND_END:%.*]] = add i32 [[N_VEC]], [[I]]
+; IND-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i32 0
+; IND-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[INDEX]], [[I]]
+; IND-NEXT:    [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i64
+; IND-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
+; IND-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP3]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; IND-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[TMP5:%.*]] = sext i32 [[INDVARS_IV]] to i64
+; IND-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP5]]
+; IND-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; IND-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; IND-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; IND-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
+; IND:       exit:
+; IND-NEXT:    ret void
 ;
 ; UNROLL-LABEL: @nonprimary(
-; UNROLL: vector.ph:
-; UNROLL:   %[[INSERT:.*]] = insertelement <2 x i32> poison, i32 %i, i32 0
-; UNROLL:   %[[SPLAT:.*]] = shufflevector <2 x i32> %[[INSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
-; UNROLL:   %[[START:.*]] = add <2 x i32> %[[SPLAT]], <i32 0, i32 1>
-; UNROLL: vector.body:
-; UNROLL:   %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; UNROLL:   %vec.ind = phi <2 x i32> [ %[[START]], %vector.ph ], [ %vec.ind.next, %vector.body ]
-; UNROLL:   %step.add = add <2 x i32> %vec.ind, <i32 2, i32 2>
-; UNROLL:   %[[A1:.*]] = add i32 %index, %i
-; UNROLL:   %[[S1:.*]] = sext i32 %[[A1]] to i64
-; UNROLL:   %[[G1:.*]] = getelementptr inbounds i32, i32* %a, i64 %[[S1]]
-; UNROLL:   %[[B1:.*]] = bitcast i32* %[[G1]] to <2 x i32>*
-; UNROLL:   store <2 x i32> %vec.ind, <2 x i32>* %[[B1]]
-; UNROLL:   %[[G2:.*]] = getelementptr inbounds i32, i32* %[[G1]], i64 2
-; UNROLL:   %[[B2:.*]] = bitcast i32* %[[G2]] to <2 x i32>*
-; UNROLL:   store <2 x i32> %step.add, <2 x i32>* %[[B2]]
-; UNROLL:   %index.next = add nuw i32 %index, 4
-; UNROLL:   %vec.ind.next = add <2 x i32> %vec.ind, <i32 4, i32 4>
-; UNROLL:   %[[CMP:.*]] = icmp eq i32 %index.next, %n.vec
-; UNROLL:   br i1 %[[CMP]]
-define void @nonprimary(i32* nocapture %a, i32 %start, i32 %i, i32 %k) {
+; UNROLL-NEXT:  for.body.preheader:
+; UNROLL-NEXT:    [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -4
+; UNROLL-NEXT:    [[IND_END:%.*]] = add i32 [[N_VEC]], [[I]]
+; UNROLL-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i32 0
+; UNROLL-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; UNROLL-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[INDEX]], [[I]]
+; UNROLL-NEXT:    [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i64
+; UNROLL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
+; UNROLL-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP3]], align 4
+; UNROLL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 2
+; UNROLL-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP5]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[TMP7:%.*]] = sext i32 [[INDVARS_IV]] to i64
+; UNROLL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
+; UNROLL-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
+; UNROLL:       exit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @nonprimary(
+; UNROLL-NO-IC-NEXT:  for.body.preheader:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i32 [[OFFSET_IDX]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP8]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP10]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[INDVARS_IV]]
+; UNROLL-NO-IC-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
+; UNROLL-NO-IC:       exit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @nonprimary(
+; INTERLEAVE-NEXT:  for.body.preheader:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP0]], -8
+; INTERLEAVE-NEXT:    [[IND_END:%.*]] = add i32 [[N_VEC]], [[I]]
+; INTERLEAVE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[I]], i32 0
+; INTERLEAVE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
+; INTERLEAVE-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[INDEX]], [[I]]
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i64
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP3]], align 4
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 4
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP5]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = sext i32 [[INDVARS_IV]] to i64
+; INTERLEAVE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
+; INTERLEAVE-NEXT:    store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
+; INTERLEAVE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
+; INTERLEAVE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
+; INTERLEAVE:       exit:
+; INTERLEAVE-NEXT:    ret void
+;
 for.body.preheader:
   br label %for.body
 
@@ -774,28 +5136,236 @@ exit:
   ret void
 }
 
+define void @non_primary_iv_trunc(i32* %a, i64 %n) {
 ; CHECK-LABEL: @non_primary_iv_trunc(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK:         [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
-; CHECK:         [[TMP3:%.*]] = add i64 %index, 0
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* %a, i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <2 x i32>*
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP6]], align 4
-; CHECK-NEXT:    %index.next = add nuw i64 %index, 2
-; CHECK:         [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
-; CHECK:         br i1 {{.*}}, label %middle.block, label %vector.body
-define void @non_primary_iv_trunc(i32* %a, i64 %n) {
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND2:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND2]], <2 x i32>* [[TMP3]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 4, i64 4>
+; CHECK-NEXT:    [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], <i32 4, i32 4>
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
+; CHECK-NEXT:    [[VAR1:%.*]] = trunc i64 [[J]] to i32
+; CHECK-NEXT:    store i32 [[VAR1]], i32* [[VAR0]], align 4
+; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; CHECK-NEXT:    [[J_NEXT]] = add nuw nsw i64 [[J]], 2
+; CHECK-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @non_primary_iv_trunc(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
+; IND-NEXT:    [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 1
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND2:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
+; IND-NEXT:    [[TMP1:%.*]] = bitcast i32* [[TMP0]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[VEC_IND2]], <2 x i32>* [[TMP1]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], <i32 4, i32 4>
+; IND-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; IND-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; IND-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
+; IND-NEXT:    [[VAR1:%.*]] = trunc i64 [[J]] to i32
+; IND-NEXT:    store i32 [[VAR1]], i32* [[VAR0]], align 4
+; IND-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; IND-NEXT:    [[J_NEXT]] = add nuw nsw i64 [[J]], 2
+; IND-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; IND-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @non_primary_iv_trunc(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804
+; UNROLL-NEXT:    [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 1
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[STEP_ADD4:%.*]] = add <2 x i32> [[VEC_IND3]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP1:%.*]] = bitcast i32* [[TMP0]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[VEC_IND3]], <2 x i32>* [[TMP1]], align 4
+; UNROLL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 2
+; UNROLL-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[STEP_ADD4]], <2 x i32>* [[TMP3]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT6]] = add <2 x i32> [[VEC_IND3]], <i32 8, i32 8>
+; UNROLL-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
+; UNROLL-NEXT:    [[VAR1:%.*]] = trunc i64 [[J]] to i32
+; UNROLL-NEXT:    store i32 [[VAR1]], i32* [[VAR0]], align 4
+; UNROLL-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NEXT:    [[J_NEXT]] = add nuw nsw i64 [[J]], 2
+; UNROLL-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @non_primary_iv_trunc(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], <i64 4, i64 4>
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD4:%.*]] = add <2 x i32> [[VEC_IND3]], <i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND3]], <2 x i32>* [[TMP5]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD4]], <2 x i32>* [[TMP7]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], <i64 4, i64 4>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT6]] = add <2 x i32> [[STEP_ADD4]], <i32 4, i32 4>
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
+; UNROLL-NO-IC-NEXT:    [[VAR1:%.*]] = trunc i64 [[J]] to i32
+; UNROLL-NO-IC-NEXT:    store i32 [[VAR1]], i32* [[VAR0]], align 4
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; UNROLL-NO-IC-NEXT:    [[J_NEXT]] = add nuw nsw i64 [[J]], 2
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @non_primary_iv_trunc(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775800
+; INTERLEAVE-NEXT:    [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 1
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND3:%.*]] = phi <4 x i32> [ <i32 0, i32 2, i32 4, i32 6>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[STEP_ADD4:%.*]] = add <4 x i32> [[VEC_IND3]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = bitcast i32* [[TMP0]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[VEC_IND3]], <4 x i32>* [[TMP1]], align 4
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 4
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[STEP_ADD4]], <4 x i32>* [[TMP3]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT6]] = add <4 x i32> [[VEC_IND3]], <i32 16, i32 16, i32 16, i32 16>
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
+; INTERLEAVE-NEXT:    [[VAR1:%.*]] = trunc i64 [[J]] to i32
+; INTERLEAVE-NEXT:    store i32 [[VAR1]], i32* [[VAR0]], align 4
+; INTERLEAVE-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
+; INTERLEAVE-NEXT:    [[J_NEXT]] = add nuw nsw i64 [[J]], 2
+; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    ret void
+;
 entry:
   br label %for.body
 
 for.body:
   %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]
   %j = phi i64 [ %j.next, %for.body ], [ 0, %entry ]
-  %tmp0 = getelementptr inbounds i32, i32* %a, i64 %i
-  %tmp1 = trunc i64 %j to i32
-  store i32 %tmp1, i32* %tmp0, align 4
+  %var0 = getelementptr inbounds i32, i32* %a, i64 %i
+  %var1 = trunc i64 %j to i32
+  store i32 %var1, i32* %var0, align 4
   %i.next = add nuw nsw i64 %i, 1
   %j.next = add nuw nsw i64 %j, 2
   %cond = icmp slt i64 %i.next, %n
@@ -806,48 +5376,459 @@ for.end:
 }
 
 ; PR32419. Ensure we transform truncated non-primary induction variables. In
-; the test case below we replace %tmp1 with a new induction variable. Because
+; the test case below we replace %var1 with a new induction variable. Because
 ; the truncated value is non-primary, we must compute an offset from the
 ; primary induction variable.
 ;
+;
+define i32 @PR32419(i32 %a, i16 %b) {
 ; CHECK-LABEL: @PR32419(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> zeroinitializer, i32 [[A:%.*]], i32 0
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %[[PRED_UREM_CONTINUE4:.*]] ]
-; CHECK:         [[OFFSET_IDX:%.*]] = add i32 -20, [[INDEX]]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE4:%.*]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 -20, i32 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE4]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[PRED_UREM_CONTINUE4]] ]
+; CHECK-NEXT:    [[VEC_IND1:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], [[PRED_UREM_CONTINUE4]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 -20, [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc i32 [[OFFSET_IDX]] to i16
-; CHECK:         [[TMP8:%.*]] = add i16 [[TMP1]], 0
-; CHECK-NEXT:    [[TMP9:%.*]] = urem i16 %b, [[TMP8]]
-; CHECK:         [[TMP15:%.*]] = add i16 [[TMP1]], 1
-; CHECK-NEXT:    [[TMP16:%.*]] = urem i16 %b, [[TMP15]]
-; CHECK:       [[PRED_UREM_CONTINUE4]]:
-; CHECK:         br i1 {{.*}}, label %middle.block, label %vector.body
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND1]], zeroinitializer
+; CHECK-NEXT:    [[TMP3:%.*]] = xor <2 x i1> [[TMP2]], <i1 true, i1 true>
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
+; CHECK-NEXT:    br i1 [[TMP4]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
+; CHECK:       pred.urem.if:
+; CHECK-NEXT:    [[TMP5:%.*]] = add i16 [[TMP1]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = urem i16 [[B:%.*]], [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x i16> poison, i16 [[TMP6]], i32 0
+; CHECK-NEXT:    br label [[PRED_UREM_CONTINUE]]
+; CHECK:       pred.urem.continue:
+; CHECK-NEXT:    [[TMP8:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UREM_IF]] ]
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i1> [[TMP3]], i32 1
+; CHECK-NEXT:    br i1 [[TMP9]], label [[PRED_UREM_IF3:%.*]], label [[PRED_UREM_CONTINUE4]]
+; CHECK:       pred.urem.if3:
+; CHECK-NEXT:    [[TMP10:%.*]] = add i16 [[TMP1]], 1
+; CHECK-NEXT:    [[TMP11:%.*]] = urem i16 [[B]], [[TMP10]]
+; CHECK-NEXT:    [[TMP12:%.*]] = insertelement <2 x i16> [[TMP8]], i16 [[TMP11]], i32 1
+; CHECK-NEXT:    br label [[PRED_UREM_CONTINUE4]]
+; CHECK:       pred.urem.continue4:
+; CHECK-NEXT:    [[TMP13:%.*]] = phi <2 x i16> [ [[TMP8]], [[PRED_UREM_CONTINUE]] ], [ [[TMP12]], [[PRED_UREM_IF3]] ]
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> zeroinitializer, <2 x i16> [[TMP13]]
+; CHECK-NEXT:    [[TMP14:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>
+; CHECK-NEXT:    [[TMP15]] = or <2 x i32> [[VEC_PHI]], [[TMP14]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i16> [[VEC_IND1]], <i16 2, i16 2>
+; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20
+; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[TMP15]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 20, 20
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ -20, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[A]], [[ENTRY]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
+; CHECK-NEXT:    [[VAR0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR6:%.*]], [[FOR_INC]] ]
+; CHECK-NEXT:    [[VAR1:%.*]] = trunc i32 [[I]] to i16
+; CHECK-NEXT:    [[VAR2:%.*]] = icmp eq i16 [[VAR1]], 0
+; CHECK-NEXT:    br i1 [[VAR2]], label [[FOR_INC]], label [[FOR_COND:%.*]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[VAR3:%.*]] = urem i16 [[B]], [[VAR1]]
+; CHECK-NEXT:    br label [[FOR_INC]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[VAR4:%.*]] = phi i16 [ [[VAR3]], [[FOR_COND]] ], [ 0, [[FOR_BODY]] ]
+; CHECK-NEXT:    [[VAR5:%.*]] = sext i16 [[VAR4]] to i32
+; CHECK-NEXT:    [[VAR6]] = or i32 [[VAR0]], [[VAR5]]
+; CHECK-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; CHECK-NEXT:    [[COND:%.*]] = icmp eq i32 [[I_NEXT]], 0
+; CHECK-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    [[VAR7:%.*]] = phi i32 [ [[VAR6]], [[FOR_INC]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[VAR7]]
+;
+; IND-LABEL: @PR32419(
+; IND-NEXT:  entry:
+; IND-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> <i32 poison, i32 0>, i32 [[A:%.*]], i32 0
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE4:%.*]] ]
+; IND-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[PRED_UREM_CONTINUE4]] ]
+; IND-NEXT:    [[VEC_IND1:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], [[PRED_UREM_CONTINUE4]] ]
+; IND-NEXT:    [[TMP1:%.*]] = trunc i32 [[INDEX]] to i16
+; IND-NEXT:    [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND1]], zeroinitializer
+; IND-NEXT:    [[TMP3:%.*]] = xor <2 x i1> [[TMP2]], <i1 true, i1 true>
+; IND-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
+; IND-NEXT:    br i1 [[TMP4]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
+; IND:       pred.urem.if:
+; IND-NEXT:    [[TMP5:%.*]] = add i16 [[TMP1]], -20
+; IND-NEXT:    [[TMP6:%.*]] = urem i16 [[B:%.*]], [[TMP5]]
+; IND-NEXT:    [[TMP7:%.*]] = insertelement <2 x i16> poison, i16 [[TMP6]], i32 0
+; IND-NEXT:    br label [[PRED_UREM_CONTINUE]]
+; IND:       pred.urem.continue:
+; IND-NEXT:    [[TMP8:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UREM_IF]] ]
+; IND-NEXT:    [[TMP9:%.*]] = extractelement <2 x i1> [[TMP3]], i32 1
+; IND-NEXT:    br i1 [[TMP9]], label [[PRED_UREM_IF3:%.*]], label [[PRED_UREM_CONTINUE4]]
+; IND:       pred.urem.if3:
+; IND-NEXT:    [[TMP10:%.*]] = add i16 [[TMP1]], -19
+; IND-NEXT:    [[TMP11:%.*]] = urem i16 [[B]], [[TMP10]]
+; IND-NEXT:    [[TMP12:%.*]] = insertelement <2 x i16> [[TMP8]], i16 [[TMP11]], i32 1
+; IND-NEXT:    br label [[PRED_UREM_CONTINUE4]]
+; IND:       pred.urem.continue4:
+; IND-NEXT:    [[TMP13:%.*]] = phi <2 x i16> [ [[TMP8]], [[PRED_UREM_CONTINUE]] ], [ [[TMP12]], [[PRED_UREM_IF3]] ]
+; IND-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> zeroinitializer, <2 x i16> [[TMP13]]
+; IND-NEXT:    [[TMP14:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>
+; IND-NEXT:    [[TMP15]] = or <2 x i32> [[VEC_PHI]], [[TMP14]]
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i16> [[VEC_IND1]], <i16 2, i16 2>
+; IND-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20
+; IND-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[TMP15]])
+; IND-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    br label [[FOR_BODY:%.*]]
+; IND:       for.body:
+; IND-NEXT:    br i1 undef, label [[FOR_INC:%.*]], label [[FOR_COND:%.*]]
+; IND:       for.cond:
+; IND-NEXT:    br label [[FOR_INC]]
+; IND:       for.inc:
+; IND-NEXT:    br i1 undef, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; IND:       for.end:
+; IND-NEXT:    [[VAR7:%.*]] = phi i32 [ undef, [[FOR_INC]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    ret i32 [[VAR7]]
+;
+; UNROLL-LABEL: @PR32419(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> <i32 poison, i32 0>, i32 [[A:%.*]], i32 0
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE12:%.*]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[PRED_UREM_CONTINUE12]] ]
+; UNROLL-NEXT:    [[VEC_PHI2:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[PRED_UREM_CONTINUE12]] ]
+; UNROLL-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[PRED_UREM_CONTINUE12]] ]
+; UNROLL-NEXT:    [[TMP1:%.*]] = trunc i32 [[INDEX]] to i16
+; UNROLL-NEXT:    [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND3]], zeroinitializer
+; UNROLL-NEXT:    [[TMP3:%.*]] = icmp eq <2 x i16> [[VEC_IND3]], <i16 -2, i16 -2>
+; UNROLL-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP2]], <i1 true, i1 true>
+; UNROLL-NEXT:    [[TMP5:%.*]] = xor <2 x i1> [[TMP3]], <i1 true, i1 true>
+; UNROLL-NEXT:    [[TMP6:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
+; UNROLL-NEXT:    br i1 [[TMP6]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
+; UNROLL:       pred.urem.if:
+; UNROLL-NEXT:    [[TMP7:%.*]] = add i16 [[TMP1]], -20
+; UNROLL-NEXT:    [[TMP8:%.*]] = urem i16 [[B:%.*]], [[TMP7]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = insertelement <2 x i16> poison, i16 [[TMP8]], i32 0
+; UNROLL-NEXT:    br label [[PRED_UREM_CONTINUE]]
+; UNROLL:       pred.urem.continue:
+; UNROLL-NEXT:    [[TMP10:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UREM_IF]] ]
+; UNROLL-NEXT:    [[TMP11:%.*]] = extractelement <2 x i1> [[TMP4]], i32 1
+; UNROLL-NEXT:    br i1 [[TMP11]], label [[PRED_UREM_IF7:%.*]], label [[PRED_UREM_CONTINUE8:%.*]]
+; UNROLL:       pred.urem.if7:
+; UNROLL-NEXT:    [[TMP12:%.*]] = add i16 [[TMP1]], -19
+; UNROLL-NEXT:    [[TMP13:%.*]] = urem i16 [[B]], [[TMP12]]
+; UNROLL-NEXT:    [[TMP14:%.*]] = insertelement <2 x i16> [[TMP10]], i16 [[TMP13]], i32 1
+; UNROLL-NEXT:    br label [[PRED_UREM_CONTINUE8]]
+; UNROLL:       pred.urem.continue8:
+; UNROLL-NEXT:    [[TMP15:%.*]] = phi <2 x i16> [ [[TMP10]], [[PRED_UREM_CONTINUE]] ], [ [[TMP14]], [[PRED_UREM_IF7]] ]
+; UNROLL-NEXT:    [[TMP16:%.*]] = extractelement <2 x i1> [[TMP5]], i32 0
+; UNROLL-NEXT:    br i1 [[TMP16]], label [[PRED_UREM_IF9:%.*]], label [[PRED_UREM_CONTINUE10:%.*]]
+; UNROLL:       pred.urem.if9:
+; UNROLL-NEXT:    [[TMP17:%.*]] = add i16 [[TMP1]], -18
+; UNROLL-NEXT:    [[TMP18:%.*]] = urem i16 [[B]], [[TMP17]]
+; UNROLL-NEXT:    [[TMP19:%.*]] = insertelement <2 x i16> poison, i16 [[TMP18]], i32 0
+; UNROLL-NEXT:    br label [[PRED_UREM_CONTINUE10]]
+; UNROLL:       pred.urem.continue10:
+; UNROLL-NEXT:    [[TMP20:%.*]] = phi <2 x i16> [ poison, [[PRED_UREM_CONTINUE8]] ], [ [[TMP19]], [[PRED_UREM_IF9]] ]
+; UNROLL-NEXT:    [[TMP21:%.*]] = extractelement <2 x i1> [[TMP5]], i32 1
+; UNROLL-NEXT:    br i1 [[TMP21]], label [[PRED_UREM_IF11:%.*]], label [[PRED_UREM_CONTINUE12]]
+; UNROLL:       pred.urem.if11:
+; UNROLL-NEXT:    [[TMP22:%.*]] = add i16 [[TMP1]], -17
+; UNROLL-NEXT:    [[TMP23:%.*]] = urem i16 [[B]], [[TMP22]]
+; UNROLL-NEXT:    [[TMP24:%.*]] = insertelement <2 x i16> [[TMP20]], i16 [[TMP23]], i32 1
+; UNROLL-NEXT:    br label [[PRED_UREM_CONTINUE12]]
+; UNROLL:       pred.urem.continue12:
+; UNROLL-NEXT:    [[TMP25:%.*]] = phi <2 x i16> [ [[TMP20]], [[PRED_UREM_CONTINUE10]] ], [ [[TMP24]], [[PRED_UREM_IF11]] ]
+; UNROLL-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> zeroinitializer, <2 x i16> [[TMP15]]
+; UNROLL-NEXT:    [[PREDPHI13:%.*]] = select <2 x i1> [[TMP3]], <2 x i16> zeroinitializer, <2 x i16> [[TMP25]]
+; UNROLL-NEXT:    [[TMP26:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>
+; UNROLL-NEXT:    [[TMP27:%.*]] = sext <2 x i16> [[PREDPHI13]] to <2 x i32>
+; UNROLL-NEXT:    [[TMP28]] = or <2 x i32> [[VEC_PHI]], [[TMP26]]
+; UNROLL-NEXT:    [[TMP29]] = or <2 x i32> [[VEC_PHI2]], [[TMP27]]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT6]] = add <2 x i16> [[VEC_IND3]], <i16 4, i16 4>
+; UNROLL-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20
+; UNROLL-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = or <2 x i32> [[TMP29]], [[TMP28]]
+; UNROLL-NEXT:    [[TMP31:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL:       for.body:
+; UNROLL-NEXT:    br i1 undef, label [[FOR_INC:%.*]], label [[FOR_COND:%.*]]
+; UNROLL:       for.cond:
+; UNROLL-NEXT:    br label [[FOR_INC]]
+; UNROLL:       for.inc:
+; UNROLL-NEXT:    br i1 undef, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; UNROLL:       for.end:
+; UNROLL-NEXT:    [[VAR7:%.*]] = phi i32 [ undef, [[FOR_INC]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i32 [[VAR7]]
+;
+; UNROLL-NO-IC-LABEL: @PR32419(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> zeroinitializer, i32 [[A:%.*]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE12:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 -20, i32 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE12]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[PRED_UREM_CONTINUE12]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI2:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[PRED_UREM_CONTINUE12]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[PRED_UREM_CONTINUE12]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD4:%.*]] = add <2 x i16> [[VEC_IND3]], <i16 2, i16 2>
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i32 -20, [[INDEX]]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = trunc i32 [[OFFSET_IDX]] to i16
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND3]], zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = icmp eq <2 x i16> [[STEP_ADD4]], zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP2]], <i1 true, i1 true>
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = xor <2 x i1> [[TMP3]], <i1 true, i1 true>
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP6]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
+; UNROLL-NO-IC:       pred.urem.if:
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i16 [[TMP1]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = urem i16 [[B:%.*]], [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = insertelement <2 x i16> poison, i16 [[TMP8]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UREM_CONTINUE]]
+; UNROLL-NO-IC:       pred.urem.continue:
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UREM_IF]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = extractelement <2 x i1> [[TMP4]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP11]], label [[PRED_UREM_IF7:%.*]], label [[PRED_UREM_CONTINUE8:%.*]]
+; UNROLL-NO-IC:       pred.urem.if7:
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add i16 [[TMP1]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = urem i16 [[B]], [[TMP12]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = insertelement <2 x i16> [[TMP10]], i16 [[TMP13]], i32 1
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UREM_CONTINUE8]]
+; UNROLL-NO-IC:       pred.urem.continue8:
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = phi <2 x i16> [ [[TMP10]], [[PRED_UREM_CONTINUE]] ], [ [[TMP14]], [[PRED_UREM_IF7]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = extractelement <2 x i1> [[TMP5]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP16]], label [[PRED_UREM_IF9:%.*]], label [[PRED_UREM_CONTINUE10:%.*]]
+; UNROLL-NO-IC:       pred.urem.if9:
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = add i16 [[TMP1]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = urem i16 [[B]], [[TMP17]]
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = insertelement <2 x i16> poison, i16 [[TMP18]], i32 0
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UREM_CONTINUE10]]
+; UNROLL-NO-IC:       pred.urem.continue10:
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = phi <2 x i16> [ poison, [[PRED_UREM_CONTINUE8]] ], [ [[TMP19]], [[PRED_UREM_IF9]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = extractelement <2 x i1> [[TMP5]], i32 1
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP21]], label [[PRED_UREM_IF11:%.*]], label [[PRED_UREM_CONTINUE12]]
+; UNROLL-NO-IC:       pred.urem.if11:
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = add i16 [[TMP1]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = urem i16 [[B]], [[TMP22]]
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = insertelement <2 x i16> [[TMP20]], i16 [[TMP23]], i32 1
+; UNROLL-NO-IC-NEXT:    br label [[PRED_UREM_CONTINUE12]]
+; UNROLL-NO-IC:       pred.urem.continue12:
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = phi <2 x i16> [ [[TMP20]], [[PRED_UREM_CONTINUE10]] ], [ [[TMP24]], [[PRED_UREM_IF11]] ]
+; UNROLL-NO-IC-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> zeroinitializer, <2 x i16> [[TMP15]]
+; UNROLL-NO-IC-NEXT:    [[PREDPHI13:%.*]] = select <2 x i1> [[TMP3]], <2 x i16> zeroinitializer, <2 x i16> [[TMP25]]
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = sext <2 x i16> [[PREDPHI13]] to <2 x i32>
+; UNROLL-NO-IC-NEXT:    [[TMP28]] = or <2 x i32> [[VEC_PHI]], [[TMP26]]
+; UNROLL-NO-IC-NEXT:    [[TMP29]] = or <2 x i32> [[VEC_PHI2]], [[TMP27]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT6]] = add <2 x i16> [[STEP_ADD4]], <i16 2, i16 2>
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = or <2 x i32> [[TMP29]], [[TMP28]]
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 20, 20
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ -20, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[A]], [[ENTRY]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
+; UNROLL-NO-IC:       for.body:
+; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR6:%.*]], [[FOR_INC]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR1:%.*]] = trunc i32 [[I]] to i16
+; UNROLL-NO-IC-NEXT:    [[VAR2:%.*]] = icmp eq i16 [[VAR1]], 0
+; UNROLL-NO-IC-NEXT:    br i1 [[VAR2]], label [[FOR_INC]], label [[FOR_COND:%.*]]
+; UNROLL-NO-IC:       for.cond:
+; UNROLL-NO-IC-NEXT:    [[VAR3:%.*]] = urem i16 [[B]], [[VAR1]]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_INC]]
+; UNROLL-NO-IC:       for.inc:
+; UNROLL-NO-IC-NEXT:    [[VAR4:%.*]] = phi i16 [ [[VAR3]], [[FOR_COND]] ], [ 0, [[FOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VAR5:%.*]] = sext i16 [[VAR4]] to i32
+; UNROLL-NO-IC-NEXT:    [[VAR6]] = or i32 [[VAR0]], [[VAR5]]
+; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp eq i32 [[I_NEXT]], 0
+; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; UNROLL-NO-IC:       for.end:
+; UNROLL-NO-IC-NEXT:    [[VAR7:%.*]] = phi i32 [ [[VAR6]], [[FOR_INC]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i32 [[VAR7]]
+;
+; INTERLEAVE-LABEL: @PR32419(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = insertelement <4 x i32> <i32 poison, i32 0, i32 0, i32 0>, i32 [[A:%.*]], i32 0
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE20:%.*]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP48:%.*]], [[PRED_UREM_CONTINUE20]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[PRED_UREM_CONTINUE20]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND3:%.*]] = phi <4 x i16> [ <i16 -20, i16 -19, i16 -18, i16 -17>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[PRED_UREM_CONTINUE20]] ]
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = trunc i32 [[INDEX]] to i16
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = icmp eq <4 x i16> [[VEC_IND3]], zeroinitializer
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i16> [[VEC_IND3]], <i16 -4, i16 -4, i16 -4, i16 -4>
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = xor <4 x i1> [[TMP2]], <i1 true, i1 true, i1 true, i1 true>
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = xor <4 x i1> [[TMP3]], <i1 true, i1 true, i1 true, i1 true>
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = extractelement <4 x i1> [[TMP4]], i32 0
+; INTERLEAVE-NEXT:    br i1 [[TMP6]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
+; INTERLEAVE:       pred.urem.if:
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = add i16 [[TMP1]], -20
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = urem i16 [[B:%.*]], [[TMP7]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = insertelement <4 x i16> poison, i16 [[TMP8]], i32 0
+; INTERLEAVE-NEXT:    br label [[PRED_UREM_CONTINUE]]
+; INTERLEAVE:       pred.urem.continue:
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = phi <4 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UREM_IF]] ]
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = extractelement <4 x i1> [[TMP4]], i32 1
+; INTERLEAVE-NEXT:    br i1 [[TMP11]], label [[PRED_UREM_IF7:%.*]], label [[PRED_UREM_CONTINUE8:%.*]]
+; INTERLEAVE:       pred.urem.if7:
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = add i16 [[TMP1]], -19
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = urem i16 [[B]], [[TMP12]]
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = insertelement <4 x i16> [[TMP10]], i16 [[TMP13]], i32 1
+; INTERLEAVE-NEXT:    br label [[PRED_UREM_CONTINUE8]]
+; INTERLEAVE:       pred.urem.continue8:
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = phi <4 x i16> [ [[TMP10]], [[PRED_UREM_CONTINUE]] ], [ [[TMP14]], [[PRED_UREM_IF7]] ]
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = extractelement <4 x i1> [[TMP4]], i32 2
+; INTERLEAVE-NEXT:    br i1 [[TMP16]], label [[PRED_UREM_IF9:%.*]], label [[PRED_UREM_CONTINUE10:%.*]]
+; INTERLEAVE:       pred.urem.if9:
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = add i16 [[TMP1]], -18
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = urem i16 [[B]], [[TMP17]]
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = insertelement <4 x i16> [[TMP15]], i16 [[TMP18]], i32 2
+; INTERLEAVE-NEXT:    br label [[PRED_UREM_CONTINUE10]]
+; INTERLEAVE:       pred.urem.continue10:
+; INTERLEAVE-NEXT:    [[TMP20:%.*]] = phi <4 x i16> [ [[TMP15]], [[PRED_UREM_CONTINUE8]] ], [ [[TMP19]], [[PRED_UREM_IF9]] ]
+; INTERLEAVE-NEXT:    [[TMP21:%.*]] = extractelement <4 x i1> [[TMP4]], i32 3
+; INTERLEAVE-NEXT:    br i1 [[TMP21]], label [[PRED_UREM_IF11:%.*]], label [[PRED_UREM_CONTINUE12:%.*]]
+; INTERLEAVE:       pred.urem.if11:
+; INTERLEAVE-NEXT:    [[TMP22:%.*]] = add i16 [[TMP1]], -17
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = urem i16 [[B]], [[TMP22]]
+; INTERLEAVE-NEXT:    [[TMP24:%.*]] = insertelement <4 x i16> [[TMP20]], i16 [[TMP23]], i32 3
+; INTERLEAVE-NEXT:    br label [[PRED_UREM_CONTINUE12]]
+; INTERLEAVE:       pred.urem.continue12:
+; INTERLEAVE-NEXT:    [[TMP25:%.*]] = phi <4 x i16> [ [[TMP20]], [[PRED_UREM_CONTINUE10]] ], [ [[TMP24]], [[PRED_UREM_IF11]] ]
+; INTERLEAVE-NEXT:    [[TMP26:%.*]] = extractelement <4 x i1> [[TMP5]], i32 0
+; INTERLEAVE-NEXT:    br i1 [[TMP26]], label [[PRED_UREM_IF13:%.*]], label [[PRED_UREM_CONTINUE14:%.*]]
+; INTERLEAVE:       pred.urem.if13:
+; INTERLEAVE-NEXT:    [[TMP27:%.*]] = add i16 [[TMP1]], -16
+; INTERLEAVE-NEXT:    [[TMP28:%.*]] = urem i16 [[B]], [[TMP27]]
+; INTERLEAVE-NEXT:    [[TMP29:%.*]] = insertelement <4 x i16> poison, i16 [[TMP28]], i32 0
+; INTERLEAVE-NEXT:    br label [[PRED_UREM_CONTINUE14]]
+; INTERLEAVE:       pred.urem.continue14:
+; INTERLEAVE-NEXT:    [[TMP30:%.*]] = phi <4 x i16> [ poison, [[PRED_UREM_CONTINUE12]] ], [ [[TMP29]], [[PRED_UREM_IF13]] ]
+; INTERLEAVE-NEXT:    [[TMP31:%.*]] = extractelement <4 x i1> [[TMP5]], i32 1
+; INTERLEAVE-NEXT:    br i1 [[TMP31]], label [[PRED_UREM_IF15:%.*]], label [[PRED_UREM_CONTINUE16:%.*]]
+; INTERLEAVE:       pred.urem.if15:
+; INTERLEAVE-NEXT:    [[TMP32:%.*]] = add i16 [[TMP1]], -15
+; INTERLEAVE-NEXT:    [[TMP33:%.*]] = urem i16 [[B]], [[TMP32]]
+; INTERLEAVE-NEXT:    [[TMP34:%.*]] = insertelement <4 x i16> [[TMP30]], i16 [[TMP33]], i32 1
+; INTERLEAVE-NEXT:    br label [[PRED_UREM_CONTINUE16]]
+; INTERLEAVE:       pred.urem.continue16:
+; INTERLEAVE-NEXT:    [[TMP35:%.*]] = phi <4 x i16> [ [[TMP30]], [[PRED_UREM_CONTINUE14]] ], [ [[TMP34]], [[PRED_UREM_IF15]] ]
+; INTERLEAVE-NEXT:    [[TMP36:%.*]] = extractelement <4 x i1> [[TMP5]], i32 2
+; INTERLEAVE-NEXT:    br i1 [[TMP36]], label [[PRED_UREM_IF17:%.*]], label [[PRED_UREM_CONTINUE18:%.*]]
+; INTERLEAVE:       pred.urem.if17:
+; INTERLEAVE-NEXT:    [[TMP37:%.*]] = add i16 [[TMP1]], -14
+; INTERLEAVE-NEXT:    [[TMP38:%.*]] = urem i16 [[B]], [[TMP37]]
+; INTERLEAVE-NEXT:    [[TMP39:%.*]] = insertelement <4 x i16> [[TMP35]], i16 [[TMP38]], i32 2
+; INTERLEAVE-NEXT:    br label [[PRED_UREM_CONTINUE18]]
+; INTERLEAVE:       pred.urem.continue18:
+; INTERLEAVE-NEXT:    [[TMP40:%.*]] = phi <4 x i16> [ [[TMP35]], [[PRED_UREM_CONTINUE16]] ], [ [[TMP39]], [[PRED_UREM_IF17]] ]
+; INTERLEAVE-NEXT:    [[TMP41:%.*]] = extractelement <4 x i1> [[TMP5]], i32 3
+; INTERLEAVE-NEXT:    br i1 [[TMP41]], label [[PRED_UREM_IF19:%.*]], label [[PRED_UREM_CONTINUE20]]
+; INTERLEAVE:       pred.urem.if19:
+; INTERLEAVE-NEXT:    [[TMP42:%.*]] = add i16 [[TMP1]], -13
+; INTERLEAVE-NEXT:    [[TMP43:%.*]] = urem i16 [[B]], [[TMP42]]
+; INTERLEAVE-NEXT:    [[TMP44:%.*]] = insertelement <4 x i16> [[TMP40]], i16 [[TMP43]], i32 3
+; INTERLEAVE-NEXT:    br label [[PRED_UREM_CONTINUE20]]
+; INTERLEAVE:       pred.urem.continue20:
+; INTERLEAVE-NEXT:    [[TMP45:%.*]] = phi <4 x i16> [ [[TMP40]], [[PRED_UREM_CONTINUE18]] ], [ [[TMP44]], [[PRED_UREM_IF19]] ]
+; INTERLEAVE-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x i16> zeroinitializer, <4 x i16> [[TMP25]]
+; INTERLEAVE-NEXT:    [[PREDPHI21:%.*]] = select <4 x i1> [[TMP3]], <4 x i16> zeroinitializer, <4 x i16> [[TMP45]]
+; INTERLEAVE-NEXT:    [[TMP46:%.*]] = sext <4 x i16> [[PREDPHI]] to <4 x i32>
+; INTERLEAVE-NEXT:    [[TMP47:%.*]] = sext <4 x i16> [[PREDPHI21]] to <4 x i32>
+; INTERLEAVE-NEXT:    [[TMP48]] = or <4 x i32> [[VEC_PHI]], [[TMP46]]
+; INTERLEAVE-NEXT:    [[TMP49]] = or <4 x i32> [[VEC_PHI2]], [[TMP47]]
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT6]] = add <4 x i16> [[VEC_IND3]], <i16 8, i16 8, i16 8, i16 8>
+; INTERLEAVE-NEXT:    [[TMP50:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
+; INTERLEAVE-NEXT:    br i1 [[TMP50]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[BIN_RDX:%.*]] = or <4 x i32> [[TMP49]], [[TMP48]]
+; INTERLEAVE-NEXT:    [[TMP51:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[BIN_RDX]])
+; INTERLEAVE-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ -4, [[MIDDLE_BLOCK]] ], [ -20, [[ENTRY:%.*]] ]
+; INTERLEAVE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP51]], [[MIDDLE_BLOCK]] ], [ [[A]], [[ENTRY]] ]
+; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
+; INTERLEAVE:       for.body:
+; INTERLEAVE-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
+; INTERLEAVE-NEXT:    [[VAR0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR6:%.*]], [[FOR_INC]] ]
+; INTERLEAVE-NEXT:    [[VAR1:%.*]] = trunc i32 [[I]] to i16
+; INTERLEAVE-NEXT:    [[VAR2:%.*]] = icmp eq i16 [[VAR1]], 0
+; INTERLEAVE-NEXT:    br i1 [[VAR2]], label [[FOR_INC]], label [[FOR_COND:%.*]]
+; INTERLEAVE:       for.cond:
+; INTERLEAVE-NEXT:    [[VAR3:%.*]] = urem i16 [[B]], [[VAR1]]
+; INTERLEAVE-NEXT:    br label [[FOR_INC]]
+; INTERLEAVE:       for.inc:
+; INTERLEAVE-NEXT:    [[VAR4:%.*]] = phi i16 [ [[VAR3]], [[FOR_COND]] ], [ 0, [[FOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VAR5:%.*]] = sext i16 [[VAR4]] to i32
+; INTERLEAVE-NEXT:    [[VAR6]] = or i32 [[VAR0]], [[VAR5]]
+; INTERLEAVE-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
+; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp eq i32 [[I_NEXT]], 0
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; INTERLEAVE:       for.end:
+; INTERLEAVE-NEXT:    [[VAR7:%.*]] = phi i32 [ [[VAR6]], [[FOR_INC]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    ret i32 [[VAR7]]
 ;
-define i32 @PR32419(i32 %a, i16 %b) {
 entry:
   br label %for.body
 
 for.body:
   %i = phi i32 [ -20, %entry ], [ %i.next, %for.inc ]
-  %tmp0 = phi i32 [ %a, %entry ], [ %tmp6, %for.inc ]
-  %tmp1 = trunc i32 %i to i16
-  %tmp2 = icmp eq i16 %tmp1, 0
-  br i1 %tmp2, label %for.inc, label %for.cond
+  %var0 = phi i32 [ %a, %entry ], [ %var6, %for.inc ]
+  %var1 = trunc i32 %i to i16
+  %var2 = icmp eq i16 %var1, 0
+  br i1 %var2, label %for.inc, label %for.cond
 
 for.cond:
-  %tmp3 = urem i16 %b, %tmp1
+  %var3 = urem i16 %b, %var1
   br label %for.inc
 
 for.inc:
-  %tmp4 = phi i16 [ %tmp3, %for.cond ], [ 0, %for.body ]
-  %tmp5 = sext i16 %tmp4 to i32
-  %tmp6 = or i32 %tmp0, %tmp5
+  %var4 = phi i16 [ %var3, %for.cond ], [ 0, %for.body ]
+  %var5 = sext i16 %var4 to i32
+  %var6 = or i32 %var0, %var5
   %i.next = add nsw i32 %i, 1
   %cond = icmp eq i32 %i.next, 0
   br i1 %cond, label %for.end, label %for.body
 
 for.end:
-  %tmp7 = phi i32 [ %tmp6, %for.inc ]
-  ret i32 %tmp7
+  %var7 = phi i32 [ %var6, %for.inc ]
+  ret i32 %var7
 }
 
 ; Ensure that the shuffle vector for first order recurrence is inserted
@@ -856,16 +5837,383 @@ for.end:
 ; This also ensures the first-order recurrence splice recipe is placed
 ; correctly if it is fed by an induction.
 define i64 @trunc_with_first_order_recurrence() {
-; CHECK-LABEL: trunc_with_first_order_recurrence
-; CHECK-LABEL: vector.body:
-; CHECK-NEXT:    %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-; CHECK-NEXT:    %vec.phi = phi <2 x i64>
-; CHECK-NEXT:    [[VEC_IV_1:%.+]] = phi <2 x i64> [ <i64 1, i64 2>, %vector.ph ], [ [[VEC_IV_1_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:    [[VEC_IV_2:%.+]] = phi <2 x i32> [ <i32 1, i32 2>, %vector.ph ], [ [[VEC_IV_2_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:    [[VEC_RECUR:%.+]] = phi <2 x i32> [ <i32 poison, i32 42>, %vector.ph ], [ [[VEC_IV_3:%.+]], %vector.body ]
-; CHECK-NEXT:    [[VEC_IV_3]] = phi <2 x i32> [ <i32 1, i32 2>, %vector.ph ], [ [[VEC_IV_3_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:    [[VEC_IV_4:%.+]] = phi <2 x i32> [ <i32 1, i32 2>, %vector.ph ], [ [[VEC_IV_4_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:    shufflevector <2 x i32> [[VEC_RECUR]], <2 x i32> [[VEC_IV_3]], <2 x i32> <i32 1, i32 2>
+; CHECK-LABEL: @trunc_with_first_order_recurrence(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 1, i64 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND2:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[VEC_IND4:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND4]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND6:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND4]], <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT:    [[TMP1:%.*]] = mul <2 x i32> [[VEC_IND2]], [[VEC_IND4]]
+; CHECK-NEXT:    [[TMP2:%.*]] = add <2 x i32> [[TMP1]], <i32 42, i32 42>
+; CHECK-NEXT:    [[TMP3:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND4]]
+; CHECK-NEXT:    [[TMP4:%.*]] = add <2 x i32> [[TMP3]], [[TMP2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = sext <2 x i32> [[TMP4]] to <2 x i64>
+; CHECK-NEXT:    [[TMP6:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = shl <2 x i32> [[VEC_IND6]], <i32 1, i32 1>
+; CHECK-NEXT:    [[TMP8:%.*]] = add <2 x i32> [[TMP2]], [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>
+; CHECK-NEXT:    [[TMP10]] = add <2 x i64> [[TMP6]], [[TMP9]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT5]] = add <2 x i32> [[VEC_IND4]], <i32 2, i32 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT7]] = add <2 x i32> [[VEC_IND6]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP10]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 113, 112
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND4]], i32 1
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[VEC_IND4]], i32 0
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[DOTLCSSA]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[C6]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[C8:%.*]] = mul i32 [[X]], [[C6]]
+; CHECK-NEXT:    [[C9:%.*]] = add i32 [[C8]], 42
+; CHECK-NEXT:    [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
+; CHECK-NEXT:    [[C11:%.*]] = add i32 [[C10]], [[C9]]
+; CHECK-NEXT:    [[C12:%.*]] = sext i32 [[C11]] to i64
+; CHECK-NEXT:    [[C13:%.*]] = add i64 [[C5]], [[C12]]
+; CHECK-NEXT:    [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
+; CHECK-NEXT:    [[C15:%.*]] = add i32 [[C9]], [[C14]]
+; CHECK-NEXT:    [[C16:%.*]] = sext i32 [[C15]] to i64
+; CHECK-NEXT:    [[C23]] = add i64 [[C13]], [[C16]]
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
+; CHECK-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
+; CHECK-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
+;
+; IND-LABEL: @trunc_with_first_order_recurrence(
+; IND-NEXT:  entry:
+; IND-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND2:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[VEC_IND4:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND4]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND6:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND4]], <2 x i32> <i32 1, i32 2>
+; IND-NEXT:    [[TMP1:%.*]] = mul <2 x i32> [[VEC_IND2]], [[VEC_IND4]]
+; IND-NEXT:    [[TMP2:%.*]] = add <2 x i32> [[TMP1]], <i32 42, i32 42>
+; IND-NEXT:    [[TMP3:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND4]]
+; IND-NEXT:    [[TMP4:%.*]] = add <2 x i32> [[TMP3]], [[TMP2]]
+; IND-NEXT:    [[TMP5:%.*]] = sext <2 x i32> [[TMP4]] to <2 x i64>
+; IND-NEXT:    [[TMP6:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP5]]
+; IND-NEXT:    [[TMP7:%.*]] = shl <2 x i32> [[VEC_IND6]], <i32 1, i32 1>
+; IND-NEXT:    [[TMP8:%.*]] = add <2 x i32> [[TMP2]], [[TMP7]]
+; IND-NEXT:    [[TMP9:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>
+; IND-NEXT:    [[TMP10]] = add <2 x i64> [[TMP6]], [[TMP9]]
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
+; IND-NEXT:    [[VEC_IND_NEXT5]] = add <2 x i32> [[VEC_IND4]], <i32 2, i32 2>
+; IND-NEXT:    [[VEC_IND_NEXT7]] = add <2 x i32> [[VEC_IND6]], <i32 2, i32 2>
+; IND-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
+; IND-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP10]])
+; IND-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND4]], i32 1
+; IND-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    br label [[LOOP:%.*]]
+; IND:       exit:
+; IND-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    ret i64 [[DOTLCSSA]]
+; IND:       loop:
+; IND-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[C6]] = trunc i64 [[INDVARS_IV]] to i32
+; IND-NEXT:    [[C8:%.*]] = mul i32 [[X]], [[C6]]
+; IND-NEXT:    [[C9:%.*]] = add i32 [[C8]], 42
+; IND-NEXT:    [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
+; IND-NEXT:    [[C11:%.*]] = add i32 [[C10]], [[C9]]
+; IND-NEXT:    [[C12:%.*]] = sext i32 [[C11]] to i64
+; IND-NEXT:    [[C13:%.*]] = add i64 [[C5]], [[C12]]
+; IND-NEXT:    [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; IND-NEXT:    [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
+; IND-NEXT:    [[C15:%.*]] = add i32 [[C9]], [[C14]]
+; IND-NEXT:    [[C16:%.*]] = sext i32 [[C15]] to i64
+; IND-NEXT:    [[C23]] = add i64 [[C13]], [[C16]]
+; IND-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; IND-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
+; IND-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
+; IND-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
+;
+; UNROLL-LABEL: @trunc_with_first_order_recurrence(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_PHI2:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND4:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD9:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND8:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND12:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT15:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[STEP_ADD5:%.*]] = add <2 x i32> [[VEC_IND4]], <i32 2, i32 2>
+; UNROLL-NEXT:    [[STEP_ADD9]] = add <2 x i32> [[VEC_IND8]], <i32 2, i32 2>
+; UNROLL-NEXT:    [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND8]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND8]], <2 x i32> [[STEP_ADD9]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NEXT:    [[TMP2:%.*]] = mul <2 x i32> [[VEC_IND4]], [[VEC_IND8]]
+; UNROLL-NEXT:    [[TMP3:%.*]] = mul <2 x i32> [[STEP_ADD5]], [[STEP_ADD9]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = add <2 x i32> [[TMP2]], <i32 42, i32 42>
+; UNROLL-NEXT:    [[TMP5:%.*]] = add <2 x i32> [[TMP3]], <i32 42, i32 42>
+; UNROLL-NEXT:    [[TMP6:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND8]]
+; UNROLL-NEXT:    [[TMP7:%.*]] = add <2 x i32> [[TMP1]], [[STEP_ADD9]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = add <2 x i32> [[TMP6]], [[TMP4]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = add <2 x i32> [[TMP7]], [[TMP5]]
+; UNROLL-NEXT:    [[TMP10:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>
+; UNROLL-NEXT:    [[TMP11:%.*]] = sext <2 x i32> [[TMP9]] to <2 x i64>
+; UNROLL-NEXT:    [[TMP12:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP10]]
+; UNROLL-NEXT:    [[TMP13:%.*]] = add <2 x i64> [[VEC_PHI2]], [[TMP11]]
+; UNROLL-NEXT:    [[TMP14:%.*]] = shl <2 x i32> [[VEC_IND12]], <i32 1, i32 1>
+; UNROLL-NEXT:    [[STEP_ADD13:%.*]] = shl <2 x i32> [[VEC_IND12]], <i32 1, i32 1>
+; UNROLL-NEXT:    [[TMP15:%.*]] = add <2 x i32> [[STEP_ADD13]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP16:%.*]] = add <2 x i32> [[TMP4]], [[TMP14]]
+; UNROLL-NEXT:    [[TMP17:%.*]] = add <2 x i32> [[TMP5]], [[TMP15]]
+; UNROLL-NEXT:    [[TMP18:%.*]] = sext <2 x i32> [[TMP16]] to <2 x i64>
+; UNROLL-NEXT:    [[TMP19:%.*]] = sext <2 x i32> [[TMP17]] to <2 x i64>
+; UNROLL-NEXT:    [[TMP20]] = add <2 x i64> [[TMP12]], [[TMP18]]
+; UNROLL-NEXT:    [[TMP21]] = add <2 x i64> [[TMP13]], [[TMP19]]
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT7]] = add <2 x i32> [[VEC_IND4]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[VEC_IND_NEXT11]] = add <2 x i32> [[VEC_IND8]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[VEC_IND_NEXT15]] = add <2 x i32> [[VEC_IND12]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
+; UNROLL-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP21]], [[TMP20]]
+; UNROLL-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD9]], i32 1
+; UNROLL-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[LOOP:%.*]]
+; UNROLL:       exit:
+; UNROLL-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    ret i64 [[DOTLCSSA]]
+; UNROLL:       loop:
+; UNROLL-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[C6]] = trunc i64 [[INDVARS_IV]] to i32
+; UNROLL-NEXT:    [[C8:%.*]] = mul i32 [[X]], [[C6]]
+; UNROLL-NEXT:    [[C9:%.*]] = add i32 [[C8]], 42
+; UNROLL-NEXT:    [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
+; UNROLL-NEXT:    [[C11:%.*]] = add i32 [[C10]], [[C9]]
+; UNROLL-NEXT:    [[C12:%.*]] = sext i32 [[C11]] to i64
+; UNROLL-NEXT:    [[C13:%.*]] = add i64 [[C5]], [[C12]]
+; UNROLL-NEXT:    [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; UNROLL-NEXT:    [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
+; UNROLL-NEXT:    [[C15:%.*]] = add i32 [[C9]], [[C14]]
+; UNROLL-NEXT:    [[C16:%.*]] = sext i32 [[C15]] to i64
+; UNROLL-NEXT:    [[C23]] = add i64 [[C13]], [[C16]]
+; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
+; UNROLL-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
+; UNROLL-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
+;
+; UNROLL-NO-IC-LABEL: @trunc_with_first_order_recurrence(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_PHI2:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 1, i64 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND4:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD9:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND8:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND12:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT15:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD5:%.*]] = add <2 x i32> [[VEC_IND4]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD9]] = add <2 x i32> [[VEC_IND8]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD13:%.*]] = add <2 x i32> [[VEC_IND12]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND8]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND8]], <2 x i32> [[STEP_ADD9]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = mul <2 x i32> [[VEC_IND4]], [[VEC_IND8]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = mul <2 x i32> [[STEP_ADD5]], [[STEP_ADD9]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add <2 x i32> [[TMP2]], <i32 42, i32 42>
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add <2 x i32> [[TMP3]], <i32 42, i32 42>
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND8]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add <2 x i32> [[TMP1]], [[STEP_ADD9]]
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add <2 x i32> [[TMP6]], [[TMP4]]
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add <2 x i32> [[TMP7]], [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = sext <2 x i32> [[TMP9]] to <2 x i64>
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP10]]
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = add <2 x i64> [[VEC_PHI2]], [[TMP11]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = shl <2 x i32> [[VEC_IND12]], <i32 1, i32 1>
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = shl <2 x i32> [[STEP_ADD13]], <i32 1, i32 1>
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = add <2 x i32> [[TMP4]], [[TMP14]]
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = add <2 x i32> [[TMP5]], [[TMP15]]
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = sext <2 x i32> [[TMP16]] to <2 x i64>
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = sext <2 x i32> [[TMP17]] to <2 x i64>
+; UNROLL-NO-IC-NEXT:    [[TMP20]] = add <2 x i64> [[TMP12]], [[TMP18]]
+; UNROLL-NO-IC-NEXT:    [[TMP21]] = add <2 x i64> [[TMP13]], [[TMP19]]
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT7]] = add <2 x i32> [[STEP_ADD5]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT11]] = add <2 x i32> [[STEP_ADD9]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT15]] = add <2 x i32> [[STEP_ADD13]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP21]], [[TMP20]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 113, 112
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD9]], i32 1
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[STEP_ADD9]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
+; UNROLL-NO-IC:       exit:
+; UNROLL-NO-IC-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    ret i64 [[DOTLCSSA]]
+; UNROLL-NO-IC:       loop:
+; UNROLL-NO-IC-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; UNROLL-NO-IC-NEXT:    [[C6]] = trunc i64 [[INDVARS_IV]] to i32
+; UNROLL-NO-IC-NEXT:    [[C8:%.*]] = mul i32 [[X]], [[C6]]
+; UNROLL-NO-IC-NEXT:    [[C9:%.*]] = add i32 [[C8]], 42
+; UNROLL-NO-IC-NEXT:    [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
+; UNROLL-NO-IC-NEXT:    [[C11:%.*]] = add i32 [[C10]], [[C9]]
+; UNROLL-NO-IC-NEXT:    [[C12:%.*]] = sext i32 [[C11]] to i64
+; UNROLL-NO-IC-NEXT:    [[C13:%.*]] = add i64 [[C5]], [[C12]]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; UNROLL-NO-IC-NEXT:    [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
+; UNROLL-NO-IC-NEXT:    [[C15:%.*]] = add i32 [[C9]], [[C14]]
+; UNROLL-NO-IC-NEXT:    [[C16:%.*]] = sext i32 [[C15]] to i64
+; UNROLL-NO-IC-NEXT:    [[C23]] = add i64 [[C13]], [[C16]]
+; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
+; UNROLL-NO-IC-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
+;
+; INTERLEAVE-LABEL: @trunc_with_first_order_recurrence(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND4:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD9:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND8:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND12:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT15:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[STEP_ADD5:%.*]] = add <4 x i32> [[VEC_IND4]], <i32 4, i32 4, i32 4, i32 4>
+; INTERLEAVE-NEXT:    [[STEP_ADD9]] = add <4 x i32> [[VEC_IND8]], <i32 4, i32 4, i32 4, i32 4>
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[VEC_IND8]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[VEC_IND8]], <4 x i32> [[STEP_ADD9]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = mul <4 x i32> [[VEC_IND4]], [[VEC_IND8]]
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = mul <4 x i32> [[STEP_ADD5]], [[STEP_ADD9]]
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = add <4 x i32> [[TMP2]], <i32 42, i32 42, i32 42, i32 42>
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP3]], <i32 42, i32 42, i32 42, i32 42>
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = add <4 x i32> [[TMP0]], [[VEC_IND8]]
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = add <4 x i32> [[TMP1]], [[STEP_ADD9]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = add <4 x i32> [[TMP6]], [[TMP4]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = add <4 x i32> [[TMP7]], [[TMP5]]
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = sext <4 x i32> [[TMP8]] to <4 x i64>
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = sext <4 x i32> [[TMP9]] to <4 x i64>
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = add <4 x i64> [[VEC_PHI]], [[TMP10]]
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = add <4 x i64> [[VEC_PHI2]], [[TMP11]]
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = shl <4 x i32> [[VEC_IND12]], <i32 1, i32 1, i32 1, i32 1>
+; INTERLEAVE-NEXT:    [[STEP_ADD13:%.*]] = shl <4 x i32> [[VEC_IND12]], <i32 1, i32 1, i32 1, i32 1>
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = add <4 x i32> [[STEP_ADD13]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = add <4 x i32> [[TMP4]], [[TMP14]]
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = add <4 x i32> [[TMP5]], [[TMP15]]
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = sext <4 x i32> [[TMP16]] to <4 x i64>
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = sext <4 x i32> [[TMP17]] to <4 x i64>
+; INTERLEAVE-NEXT:    [[TMP20]] = add <4 x i64> [[TMP12]], [[TMP18]]
+; INTERLEAVE-NEXT:    [[TMP21]] = add <4 x i64> [[TMP13]], [[TMP19]]
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT7]] = add <4 x i32> [[VEC_IND4]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT11]] = add <4 x i32> [[VEC_IND8]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT15]] = add <4 x i32> [[VEC_IND12]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
+; INTERLEAVE-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[BIN_RDX:%.*]] = add <4 x i64> [[TMP21]], [[TMP20]]
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[BIN_RDX]])
+; INTERLEAVE-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD9]], i32 3
+; INTERLEAVE-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
+; INTERLEAVE:       exit:
+; INTERLEAVE-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    ret i64 [[DOTLCSSA]]
+; INTERLEAVE:       loop:
+; INTERLEAVE-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[C6]] = trunc i64 [[INDVARS_IV]] to i32
+; INTERLEAVE-NEXT:    [[C8:%.*]] = mul i32 [[X]], [[C6]]
+; INTERLEAVE-NEXT:    [[C9:%.*]] = add i32 [[C8]], 42
+; INTERLEAVE-NEXT:    [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
+; INTERLEAVE-NEXT:    [[C11:%.*]] = add i32 [[C10]], [[C9]]
+; INTERLEAVE-NEXT:    [[C12:%.*]] = sext i32 [[C11]] to i64
+; INTERLEAVE-NEXT:    [[C13:%.*]] = add i64 [[C5]], [[C12]]
+; INTERLEAVE-NEXT:    [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; INTERLEAVE-NEXT:    [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
+; INTERLEAVE-NEXT:    [[C15:%.*]] = add i32 [[C9]], [[C14]]
+; INTERLEAVE-NEXT:    [[C16:%.*]] = sext i32 [[C15]] to i64
+; INTERLEAVE-NEXT:    [[C23]] = add i64 [[C13]], [[C16]]
+; INTERLEAVE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; INTERLEAVE-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
+; INTERLEAVE-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
+;
 entry:
   br label %loop
 
@@ -899,18 +6247,270 @@ loop:                                         ; preds = %loop, %entry
 
 ; Test case for PR52460.
 define void @pr52460_first_order_recurrence_truncated_iv(i32* noalias %src, i32* %dst) {
-; CHECK-LABEL: vector.body:
-; CHECK-NEXT:  [[MAIN_IV:%.+]] = phi i64 [ 0, %vector.ph ], [ [[MAIN_IV_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:  [[VEC_IV_64:%.+]] = phi <2 x i64> [ <i64 0, i64 1>, %vector.ph ], [ [[VEC_IV_64_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:  [[VEC_RECUR:%.+]] = phi <2 x i32> [ <i32 poison, i32 0>, %vector.ph ], [ [[VEC_IV_32:%.+]], %vector.body ]
-; CHECK-NEXT:  [[VEC_IV_32]] = phi <2 x i32> [ <i32 0, i32 1>, %vector.ph ], [ [[VEC_IV_32_NEXT:%.+]], %vector.body ]
-; CHECK:       [[RECUR_SHUFFLE:%.+]] = shufflevector <2 x i32> [[VEC_RECUR]], <2 x i32> [[VEC_IV_32]], <2 x i32> <i32 1, i32 2>
-; CHECK:       mul nsw <2 x i32> {{.+}}, [[RECUR_SHUFFLE]]
-; CHECK:       [[ADD:%.+]] = add <2 x i32> [[VEC_IV_32]]
-; CHECK:       store <2 x i32> [[ADD]], <2 x i32>*
-; CHECK-NEXT:  [[MAIN_IV_NEXT]] = add nuw i64 [[MAIN_IV]], 2
-; CHECK-NEXT:  [[VEC_IV_64_NEXT]] = add <2 x i64> [[VEC_IV_64]], <i64 2, i64 2>
-; CHECK-NEXT:  [[VEC_IV_32_NEXT]] = add <2 x i32> [[VEC_IV_32]], <i32 2, i32 2>
+;
+; CHECK-LABEL: @pr52460_first_order_recurrence_truncated_iv(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND2:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND2]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDEX]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[TMP1]], 0
+; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[TMP1]], 1
+; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND2]], <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, i32* [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP5]], i32 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP6:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP4]]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[DST:%.*]], i32 [[TMP2]]
+; CHECK-NEXT:    [[TMP8:%.*]] = add <2 x i32> [[VEC_IND2]], [[TMP6]]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i32, i32* [[TMP7]], i32 0
+; CHECK-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP8]], <2 x i32>* [[TMP10]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, 100
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i32 1
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i32 0
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[TRUNC_IV:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[TRUNC_IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_TRUNC:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[LV:%.*]] = load i32, i32* [[SRC]], align 4
+; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[LV]], [[SCALAR_RECUR]]
+; CHECK-NEXT:    [[TRUNC_IV_NEXT]] = add i32 [[TRUNC_IV]], 1
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[IV_TRUNC]] = trunc i64 [[IV]] to i32
+; CHECK-NEXT:    [[DST_GEP:%.*]] = getelementptr i32, i32* [[DST]], i32 [[IV_TRUNC]]
+; CHECK-NEXT:    [[ADD:%.*]] = add i32 [[IV_TRUNC]], [[MUL]]
+; CHECK-NEXT:    store i32 [[ADD]], i32* [[DST_GEP]], align 4
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[TRUNC_IV_NEXT]], 100
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @pr52460_first_order_recurrence_truncated_iv(
+; IND-NEXT:  entry:
+; IND-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND2:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND2]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND2]], <2 x i32> <i32 1, i32 2>
+; IND-NEXT:    [[TMP1:%.*]] = load i32, i32* [[SRC:%.*]], align 4
+; IND-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP1]], i32 0
+; IND-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    [[TMP2:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP0]]
+; IND-NEXT:    [[SEXT:%.*]] = shl i64 [[INDEX]], 32
+; IND-NEXT:    [[TMP3:%.*]] = ashr exact i64 [[SEXT]], 32
+; IND-NEXT:    [[TMP4:%.*]] = getelementptr i32, i32* [[DST:%.*]], i64 [[TMP3]]
+; IND-NEXT:    [[TMP5:%.*]] = add <2 x i32> [[VEC_IND2]], [[TMP2]]
+; IND-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[TMP5]], <2 x i32>* [[TMP6]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
+; IND-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
+; IND-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    br label [[LOOP:%.*]]
+; IND:       loop:
+; IND-NEXT:    br i1 undef, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
+; IND:       exit:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @pr52460_first_order_recurrence_truncated_iv(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD4:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[STEP_ADD4]] = add <2 x i32> [[VEC_IND3]], <i32 2, i32 2>
+; UNROLL-NEXT:    [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND3]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND3]], <2 x i32> [[STEP_ADD4]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NEXT:    [[TMP2:%.*]] = load i32, i32* [[SRC:%.*]], align 4
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP2]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[BROADCAST_SPLATINSERT7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP2]], i32 0
+; UNROLL-NEXT:    [[BROADCAST_SPLAT8:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT7]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[TMP3:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP0]]
+; UNROLL-NEXT:    [[TMP4:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT8]], [[TMP1]]
+; UNROLL-NEXT:    [[SEXT:%.*]] = shl i64 [[INDEX]], 32
+; UNROLL-NEXT:    [[TMP5:%.*]] = ashr exact i64 [[SEXT]], 32
+; UNROLL-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[DST:%.*]], i64 [[TMP5]]
+; UNROLL-NEXT:    [[TMP7:%.*]] = add <2 x i32> [[VEC_IND3]], [[TMP3]]
+; UNROLL-NEXT:    [[TMP8:%.*]] = add <2 x i32> [[STEP_ADD4]], [[TMP4]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[TMP7]], <2 x i32>* [[TMP9]], align 4
+; UNROLL-NEXT:    [[TMP10:%.*]] = getelementptr i32, i32* [[TMP6]], i64 2
+; UNROLL-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[TMP8]], <2 x i32>* [[TMP11]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT6]] = add <2 x i32> [[VEC_IND3]], <i32 4, i32 4>
+; UNROLL-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
+; UNROLL-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    br label [[LOOP:%.*]]
+; UNROLL:       loop:
+; UNROLL-NEXT:    br i1 undef, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
+; UNROLL:       exit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @pr52460_first_order_recurrence_truncated_iv(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD4:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[OFFSET_IDX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[OFFSET_IDX]], 2
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD4]] = add <2 x i32> [[VEC_IND3]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = trunc i64 [[INDEX]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i32 [[TMP2]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i32 [[TMP2]], 2
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i32 [[TMP2]], 3
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND3]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = shufflevector <2 x i32> [[VEC_IND3]], <2 x i32> [[STEP_ADD4]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = load i32, i32* [[SRC:%.*]], align 4
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP9]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = load i32, i32* [[SRC]], align 4
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP10]], i32 0
+; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT8:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT7]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP7]]
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT8]], [[TMP8]]
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr i32, i32* [[DST:%.*]], i32 [[TMP3]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr i32, i32* [[DST]], i32 [[TMP5]]
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = add <2 x i32> [[VEC_IND3]], [[TMP11]]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = add <2 x i32> [[STEP_ADD4]], [[TMP12]]
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr i32, i32* [[TMP13]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = bitcast i32* [[TMP17]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP15]], <2 x i32>* [[TMP18]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr i32, i32* [[TMP13]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = bitcast i32* [[TMP19]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP16]], <2 x i32>* [[TMP20]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], <i64 2, i64 2>
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT6]] = add <2 x i32> [[STEP_ADD4]], <i32 2, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, 100
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD4]], i32 1
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[STEP_ADD4]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
+; UNROLL-NO-IC:       loop:
+; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; UNROLL-NO-IC-NEXT:    [[TRUNC_IV:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[TRUNC_IV_NEXT:%.*]], [[LOOP]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_TRUNC:%.*]], [[LOOP]] ]
+; UNROLL-NO-IC-NEXT:    [[LV:%.*]] = load i32, i32* [[SRC]], align 4
+; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = mul nsw i32 [[LV]], [[SCALAR_RECUR]]
+; UNROLL-NO-IC-NEXT:    [[TRUNC_IV_NEXT]] = add i32 [[TRUNC_IV]], 1
+; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; UNROLL-NO-IC-NEXT:    [[IV_TRUNC]] = trunc i64 [[IV]] to i32
+; UNROLL-NO-IC-NEXT:    [[DST_GEP:%.*]] = getelementptr i32, i32* [[DST]], i32 [[IV_TRUNC]]
+; UNROLL-NO-IC-NEXT:    [[ADD:%.*]] = add i32 [[IV_TRUNC]], [[MUL]]
+; UNROLL-NO-IC-NEXT:    store i32 [[ADD]], i32* [[DST_GEP]], align 4
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[TRUNC_IV_NEXT]], 100
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
+; UNROLL-NO-IC:       exit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @pr52460_first_order_recurrence_truncated_iv(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD4:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND3:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[STEP_ADD4]] = add <4 x i32> [[VEC_IND3]], <i32 4, i32 4, i32 4, i32 4>
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[VEC_IND3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[VEC_IND3]], <4 x i32> [[STEP_ADD4]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = load i32, i32* [[SRC:%.*]], align 4
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT7:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i32 0
+; INTERLEAVE-NEXT:    [[BROADCAST_SPLAT8:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT7]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = mul nsw <4 x i32> [[BROADCAST_SPLAT]], [[TMP0]]
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = mul nsw <4 x i32> [[BROADCAST_SPLAT8]], [[TMP1]]
+; INTERLEAVE-NEXT:    [[SEXT:%.*]] = shl i64 [[INDEX]], 32
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = ashr exact i64 [[SEXT]], 32
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[DST:%.*]], i64 [[TMP5]]
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = add <4 x i32> [[VEC_IND3]], [[TMP3]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = add <4 x i32> [[STEP_ADD4]], [[TMP4]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = bitcast i32* [[TMP6]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[TMP7]], <4 x i32>* [[TMP9]], align 4
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr i32, i32* [[TMP6]], i64 4
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[TMP8]], <4 x i32>* [[TMP11]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT6]] = add <4 x i32> [[VEC_IND3]], <i32 8, i32 8, i32 8, i32 8>
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96
+; INTERLEAVE-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD4]], i32 3
+; INTERLEAVE-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ 96, [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ 96, [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
+; INTERLEAVE:       loop:
+; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; INTERLEAVE-NEXT:    [[TRUNC_IV:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[TRUNC_IV_NEXT:%.*]], [[LOOP]] ]
+; INTERLEAVE-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_TRUNC:%.*]], [[LOOP]] ]
+; INTERLEAVE-NEXT:    [[LV:%.*]] = load i32, i32* [[SRC]], align 4
+; INTERLEAVE-NEXT:    [[MUL:%.*]] = mul nsw i32 [[LV]], [[SCALAR_RECUR]]
+; INTERLEAVE-NEXT:    [[TRUNC_IV_NEXT]] = add i32 [[TRUNC_IV]], 1
+; INTERLEAVE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; INTERLEAVE-NEXT:    [[IV_TRUNC]] = trunc i64 [[IV]] to i32
+; INTERLEAVE-NEXT:    [[SEXT9:%.*]] = shl i64 [[IV]], 32
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = ashr exact i64 [[SEXT9]], 32
+; INTERLEAVE-NEXT:    [[DST_GEP:%.*]] = getelementptr i32, i32* [[DST]], i64 [[TMP13]]
+; INTERLEAVE-NEXT:    [[ADD:%.*]] = add i32 [[MUL]], [[IV_TRUNC]]
+; INTERLEAVE-NEXT:    store i32 [[ADD]], i32* [[DST_GEP]], align 4
+; INTERLEAVE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[TRUNC_IV_NEXT]], 100
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
+; INTERLEAVE:       exit:
+; INTERLEAVE-NEXT:    ret void
 ;
 entry:
   br label %loop
@@ -938,19 +6538,404 @@ exit:
 ; predicates generated for the vector loop. They should be removed in the
 ; vector loop.
 define void @test_optimized_cast_induction_feeding_first_order_recurrence(i64 %n, i32 %step, i32* %ptr) {
+;
 ; CHECK-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
-; CHECK-LABEL: vector.body:
-; CHECK-NEXT:    [[MAIN_IV:%.+]] = phi i64 [ 0, %vector.ph ], [ [[MAIN_IV_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:    [[VEC_RECUR:%.+]] = phi <2 x i32> [ <i32 poison, i32 0>, %vector.ph ], [ [[VEC_IV:%.+]], %vector.body ]
-; CHECK-NEXT:    [[VEC_IV]] = phi <2 x i32> [ %induction, %vector.ph ], [ [[VEC_IV_NEXT:%.+]], %vector.body ]
-; CHECK-NEXT:    [[MAIN_IV_0:%.+]] = add i64 [[MAIN_IV]], 0
-; CHECK-NEXT:    [[RECUR_SHUFFLE:%.+]] = shufflevector <2 x i32> [[VEC_RECUR]], <2 x i32> [[VEC_IV]], <2 x i32> <i32 1, i32 2>
-; CHECK-NEXT:    [[GEP0:%.+]] = getelementptr inbounds i32, i32* %ptr, i64 [[MAIN_IV_0]]
-; CHECK-NEXT:    [[GEP1:%.+]] = getelementptr inbounds i32, i32* [[GEP0]], i32 0
-; CHECK-NEXT:    [[GEP_CAST:%.+]] = bitcast i32* [[GEP1]] to <2 x i32>*
-; CHECK-NEXT:    store <2 x i32> [[RECUR_SHUFFLE]], <2 x i32>* [[GEP_CAST]], align 4
-; CHECK-NEXT:    [[MAIN_IV_NEXT]] = add nuw i64 [[MAIN_IV]], 2
-; CHECK-NEXT:    [[VEC_IV_NEXT]] = add <2 x i32> [[VEC_IV]],
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; CHECK:       vector.scevcheck:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
+; CHECK-NEXT:    [[TMP2:%.*]] = sub i8 0, [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 0
+; CHECK-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]
+; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[TMP0]] to i8
+; CHECK-NEXT:    [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])
+; CHECK-NEXT:    [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
+; CHECK-NEXT:    [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
+; CHECK-NEXT:    [[TMP6:%.*]] = add i8 0, [[MUL_RESULT]]
+; CHECK-NEXT:    [[TMP7:%.*]] = sub i8 0, [[MUL_RESULT]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp sgt i8 [[TMP7]], 0
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp slt i8 [[TMP6]], 0
+; CHECK-NEXT:    [[TMP10:%.*]] = select i1 [[TMP3]], i1 [[TMP8]], i1 [[TMP9]]
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], 255
+; CHECK-NEXT:    [[TMP12:%.*]] = icmp ne i8 [[TMP1]], 0
+; CHECK-NEXT:    [[TMP13:%.*]] = and i1 [[TMP11]], [[TMP12]]
+; CHECK-NEXT:    [[TMP14:%.*]] = or i1 [[TMP10]], [[TMP13]]
+; CHECK-NEXT:    [[TMP15:%.*]] = or i1 [[TMP14]], [[MUL_OVERFLOW]]
+; CHECK-NEXT:    [[TMP16:%.*]] = or i1 false, [[TMP15]]
+; CHECK-NEXT:    [[TMP17:%.*]] = sext i8 [[TMP1]] to i32
+; CHECK-NEXT:    [[IDENT_CHECK:%.*]] = icmp ne i32 [[STEP]], [[TMP17]]
+; CHECK-NEXT:    [[TMP18:%.*]] = or i1 [[TMP16]], [[IDENT_CHECK]]
+; CHECK-NEXT:    br i1 [[TMP18]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; CHECK-NEXT:    [[IND_END:%.*]] = mul i32 [[CAST_CRD]], [[STEP]]
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP19:%.*]] = mul <2 x i32> <i32 0, i32 1>, [[DOTSPLAT]]
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> zeroinitializer, [[TMP19]]
+; CHECK-NEXT:    [[TMP20:%.*]] = mul i32 [[STEP]], 2
+; CHECK-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP20]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP22:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[TMP21]]
+; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i32 0
+; CHECK-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP24]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP22]], <2 x i32>* [[TMP25]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
+; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 1
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 0
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
+; CHECK-NEXT:    [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
+; CHECK-NEXT:    store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
+; CHECK-NEXT:    [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
+; CHECK-NEXT:    [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+; IND-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
+; IND-NEXT:  entry:
+; IND-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
+; IND-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; IND:       vector.scevcheck:
+; IND-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
+; IND-NEXT:    [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
+; IND-NEXT:    [[TMP2:%.*]] = sub i8 0, [[TMP1]]
+; IND-NEXT:    [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 0
+; IND-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]
+; IND-NEXT:    [[TMP5:%.*]] = trunc i64 [[TMP0]] to i8
+; IND-NEXT:    [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])
+; IND-NEXT:    [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
+; IND-NEXT:    [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
+; IND-NEXT:    [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -128
+; IND-NEXT:    [[TMP7:%.*]] = icmp slt i8 [[MUL_RESULT]], 0
+; IND-NEXT:    [[TMP8:%.*]] = select i1 [[TMP3]], i1 [[TMP6]], i1 [[TMP7]]
+; IND-NEXT:    [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 255
+; IND-NEXT:    [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 0
+; IND-NEXT:    [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]
+; IND-NEXT:    [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]
+; IND-NEXT:    [[TMP13:%.*]] = or i1 [[TMP12]], [[MUL_OVERFLOW]]
+; IND-NEXT:    [[TMP14:%.*]] = add i32 [[STEP]], -128
+; IND-NEXT:    [[TMP15:%.*]] = icmp ult i32 [[TMP14]], -256
+; IND-NEXT:    [[TMP16:%.*]] = or i1 [[TMP13]], [[TMP15]]
+; IND-NEXT:    br i1 [[TMP16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; IND:       vector.ph:
+; IND-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -2
+; IND-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; IND-NEXT:    [[IND_END:%.*]] = mul i32 [[CAST_CRD]], [[STEP]]
+; IND-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i32 0
+; IND-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    [[TMP17:%.*]] = mul nuw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
+; IND-NEXT:    [[TMP18:%.*]] = shl i32 [[STEP]], 1
+; IND-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP18]], i32 0
+; IND-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
+; IND-NEXT:    br label [[VECTOR_BODY:%.*]]
+; IND:       vector.body:
+; IND-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[VEC_IND]] = phi <2 x i32> [ [[TMP17]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IND-NEXT:    [[TMP19:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
+; IND-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[INDEX]]
+; IND-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP20]] to <2 x i32>*
+; IND-NEXT:    store <2 x i32> [[TMP19]], <2 x i32>* [[TMP21]], align 4
+; IND-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; IND-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
+; IND-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IND-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
+; IND:       middle.block:
+; IND-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; IND-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 1
+; IND-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND:       scalar.ph:
+; IND-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
+; IND-NEXT:    br label [[LOOP:%.*]]
+; IND:       loop:
+; IND-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
+; IND-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
+; IND-NEXT:    [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
+; IND-NEXT:    [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
+; IND-NEXT:    [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
+; IND-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
+; IND-NEXT:    store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
+; IND-NEXT:    [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
+; IND-NEXT:    [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
+; IND-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
+; IND-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
+; IND:       exit:
+; IND-NEXT:    ret void
+;
+; UNROLL-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
+; UNROLL-NEXT:  entry:
+; UNROLL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; UNROLL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; UNROLL:       vector.scevcheck:
+; UNROLL-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
+; UNROLL-NEXT:    [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
+; UNROLL-NEXT:    [[TMP2:%.*]] = sub i8 0, [[TMP1]]
+; UNROLL-NEXT:    [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 0
+; UNROLL-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]
+; UNROLL-NEXT:    [[TMP5:%.*]] = trunc i64 [[TMP0]] to i8
+; UNROLL-NEXT:    [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])
+; UNROLL-NEXT:    [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
+; UNROLL-NEXT:    [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
+; UNROLL-NEXT:    [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -128
+; UNROLL-NEXT:    [[TMP7:%.*]] = icmp slt i8 [[MUL_RESULT]], 0
+; UNROLL-NEXT:    [[TMP8:%.*]] = select i1 [[TMP3]], i1 [[TMP6]], i1 [[TMP7]]
+; UNROLL-NEXT:    [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 255
+; UNROLL-NEXT:    [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 0
+; UNROLL-NEXT:    [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]
+; UNROLL-NEXT:    [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]
+; UNROLL-NEXT:    [[TMP13:%.*]] = or i1 [[TMP12]], [[MUL_OVERFLOW]]
+; UNROLL-NEXT:    [[TMP14:%.*]] = add i32 [[STEP]], -128
+; UNROLL-NEXT:    [[TMP15:%.*]] = icmp ult i32 [[TMP14]], -256
+; UNROLL-NEXT:    [[TMP16:%.*]] = or i1 [[TMP13]], [[TMP15]]
+; UNROLL-NEXT:    br i1 [[TMP16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL:       vector.ph:
+; UNROLL-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -4
+; UNROLL-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; UNROLL-NEXT:    [[IND_END:%.*]] = mul i32 [[CAST_CRD]], [[STEP]]
+; UNROLL-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i32 0
+; UNROLL-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    [[TMP17:%.*]] = mul nuw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
+; UNROLL-NEXT:    [[TMP18:%.*]] = shl i32 [[STEP]], 1
+; UNROLL-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP18]], i32 0
+; UNROLL-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL:       vector.body:
+; UNROLL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ [[TMP17]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NEXT:    [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
+; UNROLL-NEXT:    [[TMP19:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NEXT:    [[TMP20:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[INDEX]]
+; UNROLL-NEXT:    [[TMP22:%.*]] = bitcast i32* [[TMP21]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[TMP19]], <2 x i32>* [[TMP22]], align 4
+; UNROLL-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i64 2
+; UNROLL-NEXT:    [[TMP24:%.*]] = bitcast i32* [[TMP23]] to <2 x i32>*
+; UNROLL-NEXT:    store <2 x i32> [[TMP20]], <2 x i32>* [[TMP24]], align 4
+; UNROLL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], [[DOTSPLAT3]]
+; UNROLL-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NEXT:    br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
+; UNROLL:       middle.block:
+; UNROLL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 1
+; UNROLL-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL:       scalar.ph:
+; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NEXT:    br label [[LOOP:%.*]]
+; UNROLL:       loop:
+; UNROLL-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
+; UNROLL-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
+; UNROLL-NEXT:    [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
+; UNROLL-NEXT:    [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
+; UNROLL-NEXT:    [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
+; UNROLL-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
+; UNROLL-NEXT:    store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
+; UNROLL-NEXT:    [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
+; UNROLL-NEXT:    [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
+; UNROLL-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
+; UNROLL-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
+; UNROLL:       exit:
+; UNROLL-NEXT:    ret void
+;
+; UNROLL-NO-IC-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
+; UNROLL-NO-IC-NEXT:  entry:
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; UNROLL-NO-IC:       vector.scevcheck:
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = sub i8 0, [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = trunc i64 [[TMP0]] to i8
+; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])
+; UNROLL-NO-IC-NEXT:    [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
+; UNROLL-NO-IC-NEXT:    [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i8 0, [[MUL_RESULT]]
+; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = sub i8 0, [[MUL_RESULT]]
+; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = icmp sgt i8 [[TMP7]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = icmp slt i8 [[TMP6]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = select i1 [[TMP3]], i1 [[TMP8]], i1 [[TMP9]]
+; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], 255
+; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = icmp ne i8 [[TMP1]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = and i1 [[TMP11]], [[TMP12]]
+; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = or i1 [[TMP10]], [[TMP13]]
+; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = or i1 [[TMP14]], [[MUL_OVERFLOW]]
+; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = or i1 false, [[TMP15]]
+; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = sext i8 [[TMP1]] to i32
+; UNROLL-NO-IC-NEXT:    [[IDENT_CHECK:%.*]] = icmp ne i32 [[STEP]], [[TMP17]]
+; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = or i1 [[TMP16]], [[IDENT_CHECK]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP18]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = mul i32 [[CAST_CRD]], [[STEP]]
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = mul <2 x i32> <i32 0, i32 1>, [[DOTSPLAT]]
+; UNROLL-NO-IC-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> zeroinitializer, [[TMP19]]
+; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = mul i32 [[STEP]], 2
+; UNROLL-NO-IC-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP20]], i32 0
+; UNROLL-NO-IC-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
+; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
+; UNROLL-NO-IC:       vector.body:
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], 0
+; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = add i64 [[INDEX]], 2
+; UNROLL-NO-IC-NEXT:    [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
+; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>
+; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[TMP21]]
+; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[TMP22]]
+; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, i32* [[TMP25]], i32 0
+; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = bitcast i32* [[TMP27]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP23]], <2 x i32>* [[TMP28]], align 4
+; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, i32* [[TMP25]], i32 2
+; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = bitcast i32* [[TMP29]] to <2 x i32>*
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP24]], <2 x i32>* [[TMP30]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], [[DOTSPLAT3]]
+; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
+; UNROLL-NO-IC:       middle.block:
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 1
+; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 0
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC:       scalar.ph:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
+; UNROLL-NO-IC:       loop:
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
+; UNROLL-NO-IC-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
+; UNROLL-NO-IC-NEXT:    [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
+; UNROLL-NO-IC-NEXT:    [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
+; UNROLL-NO-IC-NEXT:    [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
+; UNROLL-NO-IC-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
+; UNROLL-NO-IC-NEXT:    store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
+; UNROLL-NO-IC-NEXT:    [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
+; UNROLL-NO-IC-NEXT:    [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
+; UNROLL-NO-IC:       exit:
+; UNROLL-NO-IC-NEXT:    ret void
+;
+; INTERLEAVE-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
+; INTERLEAVE-NEXT:  entry:
+; INTERLEAVE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; INTERLEAVE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; INTERLEAVE:       vector.scevcheck:
+; INTERLEAVE-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
+; INTERLEAVE-NEXT:    [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
+; INTERLEAVE-NEXT:    [[TMP2:%.*]] = sub i8 0, [[TMP1]]
+; INTERLEAVE-NEXT:    [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 0
+; INTERLEAVE-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = trunc i64 [[TMP0]] to i8
+; INTERLEAVE-NEXT:    [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])
+; INTERLEAVE-NEXT:    [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
+; INTERLEAVE-NEXT:    [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -128
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = icmp slt i8 [[MUL_RESULT]], 0
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP3]], i1 [[TMP6]], i1 [[TMP7]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 255
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 0
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = or i1 [[TMP12]], [[MUL_OVERFLOW]]
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = add i32 [[STEP]], -128
+; INTERLEAVE-NEXT:    [[TMP15:%.*]] = icmp ult i32 [[TMP14]], -256
+; INTERLEAVE-NEXT:    [[TMP16:%.*]] = or i1 [[TMP13]], [[TMP15]]
+; INTERLEAVE-NEXT:    br i1 [[TMP16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE:       vector.ph:
+; INTERLEAVE-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -8
+; INTERLEAVE-NEXT:    [[CAST_CRD:%.*]] = trunc i64 [[N_VEC]] to i32
+; INTERLEAVE-NEXT:    [[IND_END:%.*]] = mul i32 [[CAST_CRD]], [[STEP]]
+; INTERLEAVE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[STEP]], i32 0
+; INTERLEAVE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    [[TMP17:%.*]] = mul <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>
+; INTERLEAVE-NEXT:    [[TMP18:%.*]] = shl i32 [[STEP]], 2
+; INTERLEAVE-NEXT:    [[DOTSPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP18]], i32 0
+; INTERLEAVE-NEXT:    [[DOTSPLAT3:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT2]], <4 x i32> poison, <4 x i32> zeroinitializer
+; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE:       vector.body:
+; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[TMP17]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[STEP_ADD]] = add <4 x i32> [[VEC_IND]], [[DOTSPLAT3]]
+; INTERLEAVE-NEXT:    [[TMP19:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[VEC_IND]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; INTERLEAVE-NEXT:    [[TMP20:%.*]] = shufflevector <4 x i32> [[VEC_IND]], <4 x i32> [[STEP_ADD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+; INTERLEAVE-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP22:%.*]] = bitcast i32* [[TMP21]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[TMP19]], <4 x i32>* [[TMP22]], align 4
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i64 4
+; INTERLEAVE-NEXT:    [[TMP24:%.*]] = bitcast i32* [[TMP23]] to <4 x i32>*
+; INTERLEAVE-NEXT:    store <4 x i32> [[TMP20]], <4 x i32>* [[TMP24]], align 4
+; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; INTERLEAVE-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], [[DOTSPLAT3]]
+; INTERLEAVE-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-NEXT:    br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
+; INTERLEAVE:       middle.block:
+; INTERLEAVE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; INTERLEAVE-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i32 3
+; INTERLEAVE-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE:       scalar.ph:
+; INTERLEAVE-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
+; INTERLEAVE:       loop:
+; INTERLEAVE-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
+; INTERLEAVE-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
+; INTERLEAVE-NEXT:    [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
+; INTERLEAVE-NEXT:    [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
+; INTERLEAVE-NEXT:    [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
+; INTERLEAVE-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
+; INTERLEAVE-NEXT:    store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
+; INTERLEAVE-NEXT:    [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
+; INTERLEAVE-NEXT:    [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
+; INTERLEAVE-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
+; INTERLEAVE:       exit:
+; INTERLEAVE-NEXT:    ret void
 ;
 entry:
   br label %loop