test/Transforms/LoopVectorize/find-last-ptr-induction.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s

target datalayout = "e-p1:32:32"

; Pointer induction stepping over the array that is loaded from.
define ptr @find_last_ptr_induction(ptr %a, i64 %n) {
; CHECK-LABEL: define ptr @find_last_ptr_induction(
; CHECK-SAME: ptr [[A:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[TMP0:%.*]] = and i64 [[N]], 3
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[TMP0]]
; CHECK-NEXT:    [[TMP1:%.*]] = shl i64 [[N_VEC]], 2
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]]
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], %[[VECTOR_PH]] ], [ [[PTR_IND:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x ptr> [ splat (ptr null), %[[VECTOR_PH]] ], [ [[TMP9:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP3:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP8:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VECTOR_GEP:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 0, i64 4, i64 8, i64 12>
; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x ptr> [[VECTOR_GEP]], i64 0
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
; CHECK-NEXT:    [[TMP5:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], splat (i32 3)
; CHECK-NEXT:    [[TMP6:%.*]] = freeze <4 x i1> [[TMP5]]
; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
; CHECK-NEXT:    [[TMP8]] = select i1 [[TMP7]], <4 x i1> [[TMP5]], <4 x i1> [[TMP3]]
; CHECK-NEXT:    [[TMP9]] = select i1 [[TMP7]], <4 x ptr> [[VECTOR_GEP]], <4 x ptr> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 16
; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP11:%.*]] = call ptr @llvm.experimental.vector.extract.last.active.v4p0(<4 x ptr> [[TMP9]], <4 x i1> [[TMP8]], ptr null)
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[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, %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[TMP2]], %[[MIDDLE_BLOCK]] ], [ [[A]], %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi ptr [ [[TMP11]], %[[MIDDLE_BLOCK]] ], [ null, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[P:%.*]] = phi ptr [ [[BC_RESUME_VAL1]], %[[SCALAR_PH]] ], [ [[P_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[LAST:%.*]] = phi ptr [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[LAST_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[L:%.*]] = load i32, ptr [[P]], align 4
; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[L]], 3
; CHECK-NEXT:    [[LAST_NEXT]] = select i1 [[CMP]], ptr [[P]], ptr [[LAST]]
; CHECK-NEXT:    [[P_NEXT]] = getelementptr inbounds i32, ptr [[P]], i64 1
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    [[R:%.*]] = phi ptr [ [[LAST_NEXT]], %[[LOOP]] ], [ [[TMP11]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret ptr [[R]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %p = phi ptr [ %a, %entry ], [ %p.next, %loop ]
  %last = phi ptr [ null, %entry ], [ %last.next, %loop ]
  %l = load i32, ptr %p
  %cmp = icmp sgt i32 %l, 3
  %last.next = select i1 %cmp, ptr %p, ptr %last
  %p.next = getelementptr inbounds i32, ptr %p, i64 1
  %iv.next = add nuw nsw i64 %iv, 1
  %ec = icmp eq i64 %iv.next, %n
  br i1 %ec, label %exit, label %loop

exit:
  %r = phi ptr [ %last.next, %loop ]
  ret ptr %r
}

; Pointer induction and loaded array are independent; non-null start value.
define ptr @find_last_ptr_induction_separate_array(ptr %a, ptr %b, ptr %init, i64 %n) {
; CHECK-LABEL: define ptr @find_last_ptr_induction_separate_array(
; CHECK-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[INIT:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[TMP0:%.*]] = and i64 [[N]], 3
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[TMP0]]
; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[N_VEC]], 24
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP1]]
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x ptr> poison, ptr [[INIT]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x ptr> [[BROADCAST_SPLATINSERT]], <4 x ptr> poison, <4 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:    [[POINTER_PHI:%.*]] = phi ptr [ [[B]], %[[VECTOR_PH]] ], [ [[PTR_IND:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x ptr> [ [[BROADCAST_SPLAT]], %[[VECTOR_PH]] ], [ [[TMP9:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP3:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP8:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VECTOR_GEP:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 0, i64 24, i64 48, i64 72>
; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
; CHECK-NEXT:    [[TMP5:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], splat (i32 3)
; CHECK-NEXT:    [[TMP6:%.*]] = freeze <4 x i1> [[TMP5]]
; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
; CHECK-NEXT:    [[TMP8]] = select i1 [[TMP7]], <4 x i1> [[TMP5]], <4 x i1> [[TMP3]]
; CHECK-NEXT:    [[TMP9]] = select i1 [[TMP7]], <4 x ptr> [[VECTOR_GEP]], <4 x ptr> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 96
; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <4 x ptr> [[BROADCAST_SPLAT]], i64 0
; CHECK-NEXT:    [[TMP12:%.*]] = call ptr @llvm.experimental.vector.extract.last.active.v4p0(<4 x ptr> [[TMP9]], <4 x i1> [[TMP8]], ptr [[TMP11]])
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[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, %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[TMP2]], %[[MIDDLE_BLOCK]] ], [ [[B]], %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi ptr [ [[TMP12]], %[[MIDDLE_BLOCK]] ], [ [[INIT]], %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[P:%.*]] = phi ptr [ [[BC_RESUME_VAL1]], %[[SCALAR_PH]] ], [ [[P_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[LAST:%.*]] = phi ptr [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[LAST_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; CHECK-NEXT:    [[L:%.*]] = load i32, ptr [[GEP]], align 4
; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[L]], 3
; CHECK-NEXT:    [[LAST_NEXT]] = select i1 [[CMP]], ptr [[P]], ptr [[LAST]]
; CHECK-NEXT:    [[P_NEXT]] = getelementptr inbounds i8, ptr [[P]], i64 24
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    [[R:%.*]] = phi ptr [ [[LAST_NEXT]], %[[LOOP]] ], [ [[TMP12]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret ptr [[R]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %p = phi ptr [ %b, %entry ], [ %p.next, %loop ]
  %last = phi ptr [ %init, %entry ], [ %last.next, %loop ]
  %gep = getelementptr inbounds i32, ptr %a, i64 %iv
  %l = load i32, ptr %gep
  %cmp = icmp sgt i32 %l, 3
  %last.next = select i1 %cmp, ptr %p, ptr %last
  %p.next = getelementptr inbounds i8, ptr %p, i64 24
  %iv.next = add nuw nsw i64 %iv, 1
  %ec = icmp eq i64 %iv.next, %n
  br i1 %ec, label %exit, label %loop

exit:
  %r = phi ptr [ %last.next, %loop ]
  ret ptr %r
}

; Selected pointer is not a pointer induction.
define ptr @find_last_ptr_load(ptr %a, i64 %n) {
; CHECK-LABEL: define ptr @find_last_ptr_load(
; CHECK-SAME: ptr [[A:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[TMP0:%.*]] = and i64 [[N]], 3
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[TMP0]]
; 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 ptr> [ splat (ptr null), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP1:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x ptr>, ptr [[TMP2]], align 8
; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne <4 x ptr> [[WIDE_LOAD]], splat (ptr null)
; CHECK-NEXT:    [[TMP4:%.*]] = freeze <4 x i1> [[TMP3]]
; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP4]])
; CHECK-NEXT:    [[TMP6]] = select i1 [[TMP5]], <4 x i1> [[TMP3]], <4 x i1> [[TMP1]]
; CHECK-NEXT:    [[TMP7]] = select i1 [[TMP5]], <4 x ptr> [[WIDE_LOAD]], <4 x ptr> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP9:%.*]] = call ptr @llvm.experimental.vector.extract.last.active.v4p0(<4 x ptr> [[TMP7]], <4 x i1> [[TMP6]], ptr null)
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[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, %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi ptr [ [[TMP9]], %[[MIDDLE_BLOCK]] ], [ null, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[LAST:%.*]] = phi ptr [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[LAST_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]]
; CHECK-NEXT:    [[L:%.*]] = load ptr, ptr [[GEP]], align 8
; CHECK-NEXT:    [[CMP:%.*]] = icmp ne ptr [[L]], null
; CHECK-NEXT:    [[LAST_NEXT]] = select i1 [[CMP]], ptr [[L]], ptr [[LAST]]
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    [[R:%.*]] = phi ptr [ [[LAST_NEXT]], %[[LOOP]] ], [ [[TMP9]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret ptr [[R]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %last = phi ptr [ null, %entry ], [ %last.next, %loop ]
  %gep = getelementptr inbounds ptr, ptr %a, i64 %iv
  %l = load ptr, ptr %gep
  %cmp = icmp ne ptr %l, null
  %last.next = select i1 %cmp, ptr %l, ptr %last
  %iv.next = add nuw nsw i64 %iv, 1
  %ec = icmp eq i64 %iv.next, %n
  br i1 %ec, label %exit, label %loop

exit:
  %r = phi ptr [ %last.next, %loop ]
  ret ptr %r
}

; Pointer index paired with a min/max reduction (argmax).
define ptr @argmax_ptr_index(ptr %a, i32 %init, i64 %n) {
; CHECK-LABEL: define ptr @argmax_ptr_index(
; CHECK-SAME: ptr [[A:%.*]], i32 [[INIT:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 1, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[P:%.*]] = phi ptr [ [[A]], %[[ENTRY]] ], [ [[P_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[MAX_VAL:%.*]] = phi i32 [ [[INIT]], %[[ENTRY]] ], [ [[MAX_VAL_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[MAX_PTR:%.*]] = phi ptr [ null, %[[ENTRY]] ], [ [[MAX_PTR_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; CHECK-NEXT:    [[L:%.*]] = load i32, ptr [[GEP]], align 4
; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[L]], [[MAX_VAL]]
; CHECK-NEXT:    [[MAX_PTR_NEXT]] = select i1 [[CMP]], ptr [[P]], ptr [[MAX_PTR]]
; CHECK-NEXT:    [[MAX_VAL_NEXT]] = tail call i32 @llvm.smax.i32(i32 [[L]], i32 [[MAX_VAL]])
; CHECK-NEXT:    [[P_NEXT]] = getelementptr inbounds i32, ptr [[P]], i64 1
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    [[R:%.*]] = phi ptr [ [[MAX_PTR_NEXT]], %[[LOOP]] ]
; CHECK-NEXT:    ret ptr [[R]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 1, %entry ], [ %iv.next, %loop ]
  %p = phi ptr [ %a, %entry ], [ %p.next, %loop ]
  %max.val = phi i32 [ %init, %entry ], [ %max.val.next, %loop ]
  %max.ptr = phi ptr [ null, %entry ], [ %max.ptr.next, %loop ]
  %gep = getelementptr inbounds i32, ptr %a, i64 %iv
  %l = load i32, ptr %gep
  %cmp = icmp sgt i32 %l, %max.val
  %max.ptr.next = select i1 %cmp, ptr %p, ptr %max.ptr
  %max.val.next = tail call i32 @llvm.smax.i32(i32 %l, i32 %max.val)
  %p.next = getelementptr inbounds i32, ptr %p, i64 1
  %iv.next = add nuw nsw i64 %iv, 1
  %ec = icmp eq i64 %iv.next, %n
  br i1 %ec, label %exit, label %loop

exit:
  %r = phi ptr [ %max.ptr.next, %loop ]
  ret ptr %r
}

; Pointer induction step is not the canonical IV type.
define ptr addrspace(1) @find_last_ptr_induction_narrow_step(ptr addrspace(1) %a, i64 %n) {
; CHECK-LABEL: define ptr addrspace(1) @find_last_ptr_induction_narrow_step(
; CHECK-SAME: ptr addrspace(1) [[A:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[TMP0:%.*]] = and i64 [[N]], 3
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[TMP0]]
; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[N_VEC]] to i32
; CHECK-NEXT:    [[TMP2:%.*]] = shl i32 [[TMP1]], 2
; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr addrspace(1) [[A]], i32 [[TMP2]]
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr addrspace(1) [ [[A]], %[[VECTOR_PH]] ], [ [[PTR_IND:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x ptr addrspace(1)> [ splat (ptr addrspace(1) null), %[[VECTOR_PH]] ], [ [[TMP10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP9:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VECTOR_GEP:%.*]] = getelementptr i8, ptr addrspace(1) [[POINTER_PHI]], <4 x i32> <i32 0, i32 4, i32 8, i32 12>
; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <4 x ptr addrspace(1)> [[VECTOR_GEP]], i64 0
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr addrspace(1) [[TMP5]], align 4
; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], splat (i32 3)
; CHECK-NEXT:    [[TMP7:%.*]] = freeze <4 x i1> [[TMP6]]
; CHECK-NEXT:    [[TMP8:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP7]])
; CHECK-NEXT:    [[TMP9]] = select i1 [[TMP8]], <4 x i1> [[TMP6]], <4 x i1> [[TMP4]]
; CHECK-NEXT:    [[TMP10]] = select i1 [[TMP8]], <4 x ptr addrspace(1)> [[VECTOR_GEP]], <4 x ptr addrspace(1)> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr addrspace(1) [[POINTER_PHI]], i32 16
; 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-NEXT:    [[TMP12:%.*]] = call ptr addrspace(1) @llvm.experimental.vector.extract.last.active.v4p1(<4 x ptr addrspace(1)> [[TMP10]], <4 x i1> [[TMP9]], ptr addrspace(1) null)
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[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, %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr addrspace(1) [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ [[A]], %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi ptr addrspace(1) [ [[TMP12]], %[[MIDDLE_BLOCK]] ], [ null, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[PTR_IV:%.*]] = phi ptr addrspace(1) [ [[BC_RESUME_VAL1]], %[[SCALAR_PH]] ], [ [[PTR_IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[LAST:%.*]] = phi ptr addrspace(1) [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[LAST_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[L:%.*]] = load i32, ptr addrspace(1) [[PTR_IV]], align 4
; CHECK-NEXT:    [[C:%.*]] = icmp eq i32 [[L]], 3
; CHECK-NEXT:    [[LAST_NEXT]] = select i1 [[C]], ptr addrspace(1) [[PTR_IV]], ptr addrspace(1) [[LAST]]
; CHECK-NEXT:    [[PTR_IV_NEXT]] = getelementptr inbounds i32, ptr addrspace(1) [[PTR_IV]], i32 1
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    [[R:%.*]] = phi ptr addrspace(1) [ [[LAST_NEXT]], %[[LOOP]] ], [ [[TMP12]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret ptr addrspace(1) [[R]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %ptr.iv = phi ptr addrspace(1) [ %a, %entry ], [ %ptr.iv.next, %loop ]
  %last = phi ptr addrspace(1) [ null, %entry ], [ %last.next, %loop ]
  %l = load i32, ptr addrspace(1) %ptr.iv, align 4
  %c = icmp eq i32 %l, 3
  %last.next = select i1 %c, ptr addrspace(1) %ptr.iv, ptr addrspace(1) %last
  %ptr.iv.next = getelementptr inbounds i32, ptr addrspace(1) %ptr.iv, i32 1
  %iv.next = add nuw nsw i64 %iv, 1
  %ec = icmp eq i64 %iv.next, %n
  br i1 %ec, label %exit, label %loop

exit:
  %r = phi ptr addrspace(1) [ %last.next, %loop ]
  ret ptr addrspace(1) %r
}