llvm/test/Transforms/LoopVectorize/find-last.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^loop(\.header)?:" --version 6
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -force-target-supports-masked-memory-ops -S %s | FileCheck %s
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -force-target-supports-masked-memory-ops -prefer-inloop-reductions -S %s | FileCheck %s

; Test vectorization of a "find last" pattern where a scalar value is
; conditionally updated based on comparing elements from two arrays.
define i32 @find_last_with_select(ptr noalias %a, ptr noalias %b) {
; CHECK-LABEL: define i32 @find_last_with_select(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br 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 <4 x i32> [ splat (i32 1), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; 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> [[TMP4]], <4 x i1> [[TMP0]]
; CHECK-NEXT:    [[TMP7]] = select i1 [[TMP5]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP7]], <4 x i1> [[TMP6]], i32 1)
; CHECK-NEXT:    br label %[[EXIT:.*]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret i32 [[TMP9]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %rdx = phi i32 [ 1, %entry ], [ %sel, %loop ]
  %gep.a = getelementptr inbounds i32, ptr %a, i64 %iv
  %load.a = load i32, ptr %gep.a, align 4
  %gep.b = getelementptr inbounds i32, ptr %b, i64 %iv
  %load.b = load i32, ptr %gep.b, align 4
  %cmp = icmp slt i32 %load.a, %load.b
  %sel = select i1 %cmp, i32 %rdx, i32 %load.a
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp eq i64 %iv.next, 500
  br i1 %exitcond, label %exit, label %loop

exit:
  ret i32 %sel
}

; Keep when a[i] > b[i] (signed), i.e., update when a[i] <= b[i].
define i32 @find_last_sgt_inverted(ptr noalias %a, ptr noalias %b) {
; CHECK-LABEL: define i32 @find_last_sgt_inverted(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br 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 <4 x i32> [ splat (i32 1), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT:    [[TMP3:%.*]] = icmp sle <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; 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> [[TMP4]], <4 x i1> [[TMP0]]
; CHECK-NEXT:    [[TMP7]] = select i1 [[TMP5]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP7]], <4 x i1> [[TMP6]], i32 1)
; CHECK-NEXT:    br label %[[EXIT:.*]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret i32 [[TMP9]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %rdx = phi i32 [ 1, %entry ], [ %sel, %loop ]
  %gep.a = getelementptr inbounds i32, ptr %a, i64 %iv
  %load.a = load i32, ptr %gep.a, align 4
  %gep.b = getelementptr inbounds i32, ptr %b, i64 %iv
  %load.b = load i32, ptr %gep.b, align 4
  %cmp = icmp sgt i32 %load.a, %load.b
  %sel = select i1 %cmp, i32 %rdx, i32 %load.a
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp eq i64 %iv.next, 500
  br i1 %exitcond, label %exit, label %loop

exit:
  ret i32 %sel
}

; Keep when a[i] < b[i] (unsigned), i.e., update when a[i] >= b[i].
define i32 @find_last_ult_inverted(ptr noalias %a, ptr noalias %b) {
; CHECK-LABEL: define i32 @find_last_ult_inverted(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br 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 <4 x i32> [ splat (i32 1), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT:    [[TMP3:%.*]] = icmp uge <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; 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> [[TMP4]], <4 x i1> [[TMP0]]
; CHECK-NEXT:    [[TMP7]] = select i1 [[TMP5]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP7]], <4 x i1> [[TMP6]], i32 1)
; CHECK-NEXT:    br label %[[EXIT:.*]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret i32 [[TMP9]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %rdx = phi i32 [ 1, %entry ], [ %sel, %loop ]
  %gep.a = getelementptr inbounds i32, ptr %a, i64 %iv
  %load.a = load i32, ptr %gep.a, align 4
  %gep.b = getelementptr inbounds i32, ptr %b, i64 %iv
  %load.b = load i32, ptr %gep.b, align 4
  %cmp = icmp ult i32 %load.a, %load.b
  %sel = select i1 %cmp, i32 %rdx, i32 %load.a
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp eq i64 %iv.next, 500
  br i1 %exitcond, label %exit, label %loop

exit:
  ret i32 %sel
}

; Keep when a[i] != b[i], i.e., update when a[i] == b[i].
define i32 @find_last_ne_inverted(ptr noalias %a, ptr noalias %b) {
; CHECK-LABEL: define i32 @find_last_ne_inverted(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br 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 <4 x i32> [ splat (i32 1), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; 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> [[TMP4]], <4 x i1> [[TMP0]]
; CHECK-NEXT:    [[TMP7]] = select i1 [[TMP5]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP7]], <4 x i1> [[TMP6]], i32 1)
; CHECK-NEXT:    br label %[[EXIT:.*]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret i32 [[TMP9]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %rdx = phi i32 [ 1, %entry ], [ %sel, %loop ]
  %gep.a = getelementptr inbounds i32, ptr %a, i64 %iv
  %load.a = load i32, ptr %gep.a, align 4
  %gep.b = getelementptr inbounds i32, ptr %b, i64 %iv
  %load.b = load i32, ptr %gep.b, align 4
  %cmp = icmp ne i32 %load.a, %load.b
  %sel = select i1 %cmp, i32 %rdx, i32 %load.a
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp eq i64 %iv.next, 500
  br i1 %exitcond, label %exit, label %loop

exit:
  ret i32 %sel
}

define i32 @loop_inv_select_condition_issue_182152(i32 %iv.start, i32 %a, i32 %b) {
; CHECK-LABEL: define i32 @loop_inv_select_condition_issue_182152(
; CHECK-SAME: i32 [[IV_START:%.*]], i32 [[A:%.*]], i32 [[B:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[INV_COND:%.*]] = icmp eq i32 [[A]], [[B]]
; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.smax.i32(i32 [[IV_START]], i32 1)
; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[TMP0]], 1
; CHECK-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[IV_START]]
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP2]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[TMP3:%.*]] = and i32 [[TMP2]], 3
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP2]], [[TMP3]]
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[INV_COND]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP4:%.*]] = add i32 [[IV_START]], [[N_VEC]]
; CHECK-NEXT:    [[TMP5:%.*]] = xor <4 x i1> [[BROADCAST_SPLAT]], splat (i1 true)
; CHECK-NEXT:    [[TMP6:%.*]] = freeze <4 x i1> [[TMP5]]
; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[IV_START]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT1]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[BROADCAST_SPLAT2]], <i32 0, i32 1, i32 2, i32 3>
; 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 <4 x i32> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP8:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP9:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP9]] = select i1 [[TMP7]], <4 x i1> [[TMP6]], <4 x i1> [[TMP8]]
; CHECK-NEXT:    [[TMP10]] = select i1 [[TMP7]], <4 x i32> [[VEC_IND]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.vector.extract.last.active.v4i32(<4 x i32> [[TMP10]], <4 x i1> [[TMP9]], i32 0)
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP2]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[TMP4]], %[[MIDDLE_BLOCK]] ], [ [[IV_START]], %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP12]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
;
entry:
  %inv_cond = icmp eq i32 %a, %b
  br label %loop

loop:
  %iv = phi i32 [ %iv.start, %entry ], [ %iv.next, %loop ]
  %rdx = phi i32 [ 0, %entry ], [ %select, %loop ]
  %select = select i1 %inv_cond, i32 %rdx, i32 %iv
  %iv.next = add i32 %iv, 1
  %exit.cond = icmp sgt i32 %iv, 0
  br i1 %exit.cond, label %exit, label %loop

exit:
  ret i32 %select
}

define ptr @loop_inv_select_condition_issue_185682(i32 %a, i32 %b) {
; CHECK-LABEL: define ptr @loop_inv_select_condition_issue_185682(
; CHECK-SAME: i32 [[A:%.*]], i32 [[B:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    [[INV_COND:%.*]] = icmp eq i32 [[A]], [[B]]
; CHECK-NEXT:    br label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[INV_COND]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP0:%.*]] = xor <4 x i1> [[BROADCAST_SPLAT]], splat (i1 true)
; CHECK-NEXT:    [[TMP1:%.*]] = freeze <4 x i1> [[TMP0]]
; CHECK-NEXT:    [[TMP2:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP1]])
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], <4 x i1> [[TMP1]], <4 x i1> zeroinitializer
; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP4:%.*]] = call ptr @llvm.experimental.vector.extract.last.active.v4p0(<4 x ptr> splat (ptr null), <4 x i1> [[TMP3]], ptr null)
; CHECK-NEXT:    br label %[[EXIT:.*]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret ptr [[TMP4]]
;
entry:
  %inv.cond = icmp eq i32 %a, %b
  br label %loop

loop:
  %rdx = phi ptr [ null, %entry ], [ %select, %loop ]
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %select = select i1 %inv.cond, ptr %rdx, ptr null
  %iv.next = add i64 %iv, 1
  %exit.cond = icmp eq i64 %iv.next, 4
  br i1 %exit.cond, label %exit, label %loop

exit:
  ret ptr %select
}

define i32 @find_last_iv_negated_mask(i1 %c, i32 %n) {
; CHECK-LABEL: define i32 @find_last_iv_negated_mask(
; CHECK-SAME: i1 [[C:%.*]], i32 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[TMP0:%.*]] = and i32 [[N]], 3
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[N]], [[TMP0]]
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 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_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP1:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP1]] = select i1 [[C]], <4 x i32> [[VEC_IND]], <4 x i32> [[VEC_PHI]]
; CHECK-NEXT:    [[TMP2]] = or <4 x i1> [[VEC_PHI1]], [[BROADCAST_SPLAT]]
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> [[TMP1]])
; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP2]])
; CHECK-NEXT:    [[TMP6:%.*]] = freeze i1 [[TMP5]]
; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i32 [[TMP4]], i32 -1
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], [[EXIT:label %.*]], 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 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
;
entry:
  br label %loop

loop:
  %iv = phi i32 [ 0, %entry ], [ %iv.next, %latch ]
  %idx = phi i32 [ -1, %entry ], [ %idx.next, %latch ]
  br i1 %c, label %if.then, label %latch

if.then:
  br label %latch

latch:
  %idx.next = phi i32 [ %iv, %if.then ], [ %idx, %loop ]
  %iv.next = add i32 %iv, 1
  %ec = icmp eq i32 %iv.next, %n
  br i1 %ec, label %exit, label %loop

exit:
  ret i32 %idx.next
}

; The data value %p is itself a phi that may yield %rdx.
define i32 @find_last_nested_phi_in_latch(ptr noalias %a, ptr noalias %b, i64 %n) {
; CHECK-LABEL: define i32 @find_last_nested_phi_in_latch(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
  %rdx = phi i32 [ -1, %entry ], [ %rdx.next, %loop.latch ]
  %gep.b = getelementptr i32, ptr %b, i64 %iv
  %ld.b = load i32, ptr %gep.b
  %c.outer = icmp sgt i32 %ld.b, 0
  br i1 %c.outer, label %outer.then, label %loop.latch

outer.then:
  %gep.a = getelementptr i32, ptr %a, i64 %iv
  %ld.a = load i32, ptr %gep.a
  %c.inner = icmp slt i32 %ld.a, 0
  br i1 %c.inner, label %inner.then, label %inner.merge

inner.then:
  br label %inner.merge

inner.merge:
  %p = phi i32 [ %ld.a, %inner.then ], [ %rdx, %outer.then ]
  br label %loop.latch

loop.latch:
  %rdx.next = phi i32 [ %p, %inner.merge ], [ %rdx, %loop.header ]
  %iv.next = add i64 %iv, 1
  %ec = icmp eq i64 %iv.next, %n
  br i1 %ec, label %exit, label %loop.header

exit:
  ret i32 %rdx.next
}

; Same as above, but the outer merge is not in the latch.
define i32 @find_last_nested_phi(ptr noalias %a, ptr noalias %b, i64 %n) {
; CHECK-LABEL: define i32 @find_last_nested_phi(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
  %rdx = phi i32 [ -1, %entry ], [ %rdx.next, %loop.latch ]
  %gep.b = getelementptr i32, ptr %b, i64 %iv
  %ld.b = load i32, ptr %gep.b
  %c.outer = icmp sgt i32 %ld.b, 0
  br i1 %c.outer, label %outer.then, label %outer.merge

outer.then:
  %gep.a = getelementptr i32, ptr %a, i64 %iv
  %ld.a = load i32, ptr %gep.a
  %c.inner = icmp slt i32 %ld.a, 0
  br i1 %c.inner, label %inner.then, label %inner.merge

inner.then:
  br label %inner.merge

inner.merge:
  %p = phi i32 [ %ld.a, %inner.then ], [ %rdx, %outer.then ]
  br label %outer.merge

outer.merge:
  %o = phi i32 [ %p, %inner.merge ], [ %rdx, %loop.header ]
  br label %loop.latch

loop.latch:
  %rdx.next = phi i32 [ %o, %outer.merge ]
  %iv.next = add i64 %iv, 1
  %ec = icmp eq i64 %iv.next, %n
  br i1 %ec, label %exit, label %loop.header

exit:
  ret i32 %rdx.next
}