llvm/test/Transforms/LoopVectorize/predicated-single-exit.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
; RUN: opt -S < %s -p loop-vectorize -force-vector-width=4 | FileCheck %s
;
@A = global [64 x i8] zeroinitializer
@B = global [64 x i8] zeroinitializer
@C = global [64 x i8] zeroinitializer

define i64 @single_exit_in_conditional_block() {
; CHECK-LABEL: define i64 @single_exit_in_conditional_block() {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
; CHECK-NEXT:    br label %[[BLOCK_A:.*]]
; CHECK:       [[BLOCK_A]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[LOOP_HEADER]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr @A, i64 [[IV]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <4 x i8> [[WIDE_LOAD]], zeroinitializer
; CHECK-NEXT:    [[GEP_B:%.*]] = getelementptr inbounds i8, ptr @B, i64 [[IV]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[GEP_B]], align 1
; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; CHECK-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP1]], <4 x i1> [[TMP3]], <4 x i1> zeroinitializer
; CHECK-NEXT:    [[TMP5:%.*]] = freeze <4 x i1> [[TMP4]]
; CHECK-NEXT:    [[CMP:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
; CHECK-NEXT:    br i1 [[CMP]], label %[[LOOP_END:.*]], label %[[LOOP_LATCH]]
; CHECK:       [[LOOP_LATCH]]:
; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[BLOCK_A]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    br label %[[LOOP_END1:.*]]
; CHECK:       [[LOOP_END]]:
; CHECK-NEXT:    [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP5]], i1 false)
; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[IV]], [[TMP8]]
; CHECK-NEXT:    br label %[[LOOP_END1]]
; CHECK:       [[LOOP_END1]]:
; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[TMP9]], %[[LOOP_END]] ], [ -1, %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret i64 [[RETVAL]]
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
  %gep.A = getelementptr inbounds i8, ptr @A, i64 %iv
  %l.A = load i8, ptr %gep.A, align 1
  %branch.cond = icmp slt i8 %l.A, 0
  br i1 %branch.cond, label %block.a, label %loop.latch

block.a:
  %gep.B = getelementptr inbounds i8, ptr @B, i64 %iv
  %l.B = load i8, ptr %gep.B, align 1
  %cmp = icmp eq i8 %l.A, %l.B
  br i1 %cmp, label %loop.end, label %loop.latch

loop.latch:
  %iv.next = add i64 %iv, 1
  %ec = icmp ne i64 %iv.next, 64
  br i1 %ec, label %loop.header, label %loop.end

loop.end:
  %retval = phi i64 [ %iv, %block.a ], [ -1, %loop.latch ]
  ret i64 %retval
}

define i64 @single_exit_in_conditional_block_with_store() {
; CHECK-LABEL: define i64 @single_exit_in_conditional_block_with_store() {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i8, ptr @A, i64 [[IV]]
; CHECK-NEXT:    [[L_A:%.*]] = load i8, ptr [[GEP_A]], align 1
; CHECK-NEXT:    [[GEP_B:%.*]] = getelementptr inbounds i8, ptr @B, i64 [[IV]]
; CHECK-NEXT:    [[BRANCH_COND:%.*]] = icmp slt i8 [[L_A]], 0
; CHECK-NEXT:    br i1 [[BRANCH_COND]], label %[[BLOCK_A:.*]], label %[[BLOCK_B:.*]]
; CHECK:       [[BLOCK_A]]:
; CHECK-NEXT:    [[L_B:%.*]] = load i8, ptr [[GEP_B]], align 1
; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i8 [[L_A]], [[L_B]]
; CHECK-NEXT:    br i1 [[CMP]], label %[[LOOP_END:.*]], label %[[LOOP_LATCH]]
; CHECK:       [[BLOCK_B]]:
; CHECK-NEXT:    store i8 0, ptr [[GEP_B]], align 1
; CHECK-NEXT:    br label %[[LOOP_LATCH]]
; CHECK:       [[LOOP_LATCH]]:
; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp ne i64 [[IV_NEXT]], 64
; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP_HEADER]], label %[[LOOP_END]]
; CHECK:       [[LOOP_END]]:
; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[IV]], %[[BLOCK_A]] ], [ -1, %[[LOOP_LATCH]] ]
; CHECK-NEXT:    ret i64 [[RETVAL]]
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
  %gep.A = getelementptr inbounds i8, ptr @A, i64 %iv
  %l.A = load i8, ptr %gep.A, align 1
  %gep.B = getelementptr inbounds i8, ptr @B, i64 %iv
  %branch.cond = icmp slt i8 %l.A, 0
  br i1 %branch.cond, label %block.a, label %block.b

block.a:
  %l.B = load i8, ptr %gep.B, align 1
  %cmp = icmp eq i8 %l.A, %l.B
  br i1 %cmp, label %loop.end, label %loop.latch

block.b:
  store i8 0, ptr %gep.B
  br label %loop.latch

loop.latch:
  %iv.next = add i64 %iv, 1
  %ec = icmp ne i64 %iv.next, 64
  br i1 %ec, label %loop.header, label %loop.end

loop.end:
  %retval = phi i64 [ %iv, %block.a ], [ -1, %loop.latch ]
  ret i64 %retval
}


define i64 @single_exit_in_conditional_block2() {
; CHECK-LABEL: define i64 @single_exit_in_conditional_block2() {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
; CHECK-NEXT:    br label %[[BLOCK_A:.*]]
; CHECK:       [[BLOCK_A]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[LOOP_HEADER]] ], [ [[INDEX_NEXT:%.*]], %[[MERGE:.*]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr @A, i64 [[IV]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <4 x i8> [[WIDE_LOAD]], zeroinitializer
; CHECK-NEXT:    [[GEP_B:%.*]] = getelementptr inbounds i8, ptr @B, i64 [[IV]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[GEP_B]], align 1
; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD1]]
; CHECK-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP1]], <4 x i1> [[TMP3]], <4 x i1> zeroinitializer
; CHECK-NEXT:    [[TMP5:%.*]] = freeze <4 x i1> [[TMP4]]
; CHECK-NEXT:    [[CMP:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
; CHECK-NEXT:    br i1 [[CMP]], label %[[LOOP_END:.*]], label %[[MERGE]]
; CHECK:       [[MERGE]]:
; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[BLOCK_A]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    br label %[[LOOP_LATCH:.*]]
; CHECK:       [[LOOP_END]]:
; CHECK-NEXT:    [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP5]], i1 false)
; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[IV]], [[TMP8]]
; CHECK-NEXT:    br label %[[LOOP_LATCH]]
; CHECK:       [[LOOP_LATCH]]:
; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[TMP9]], %[[LOOP_END]] ], [ -1, %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret i64 [[RETVAL]]
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
  %gep.A = getelementptr inbounds i8, ptr @A, i64 %iv
  %l.A = load i8, ptr %gep.A, align 1
  %branch.cond = icmp slt i8 %l.A, 0
  br i1 %branch.cond, label %block.a, label %merge

block.a:
  %gep.B = getelementptr inbounds i8, ptr @B, i64 %iv
  %l.B = load i8, ptr %gep.B, align 1
  %cmp = icmp eq i8 %l.A, %l.B
  br i1 %cmp, label %loop.end, label %merge

merge:
  br label %loop.latch

loop.latch:
  %iv.next = add i64 %iv, 1
  %ec = icmp ne i64 %iv.next, 64
  br i1 %ec, label %loop.header, label %loop.end

loop.end:
  %retval = phi i64 [ %iv, %block.a ], [ -1, %loop.latch ]
  ret i64 %retval
}

; Variant: exit condition defined in header but used in conditional block.
define i64 @exit_cond_defined_in_header() {
; CHECK-LABEL: define i64 @exit_cond_defined_in_header() {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[LOOP_HEADER]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i8, ptr @A, i64 [[IV]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[GEP_A]], align 1
; CHECK-NEXT:    [[TMP1:%.*]] = icmp sgt <4 x i8> [[WIDE_LOAD]], zeroinitializer
; CHECK-NEXT:    [[TMP2:%.*]] = icmp slt <4 x i8> [[WIDE_LOAD]], splat (i8 42)
; CHECK-NEXT:    [[TMP3:%.*]] = select <4 x i1> [[TMP1]], <4 x i1> [[TMP2]], <4 x i1> zeroinitializer
; CHECK-NEXT:    [[TMP4:%.*]] = freeze <4 x i1> [[TMP3]]
; CHECK-NEXT:    [[BRANCH_COND:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP4]])
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
; CHECK-NEXT:    br i1 [[BRANCH_COND]], label %[[BLOCK_A:.*]], label %[[LOOP_LATCH]]
; CHECK:       [[LOOP_LATCH]]:
; CHECK-NEXT:    br i1 [[TMP6]], label %[[LOOP_END:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       [[LOOP_END]]:
; CHECK-NEXT:    br label %[[LOOP_END1:.*]]
; CHECK:       [[BLOCK_A]]:
; CHECK-NEXT:    br label %[[LOOP_END1]]
; CHECK:       [[LOOP_END1]]:
; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ 10, %[[BLOCK_A]] ], [ 63, %[[LOOP_END]] ]
; CHECK-NEXT:    ret i64 [[RETVAL]]
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ %iv.next, %loop.latch ], [ 0, %entry ]
  %gep.A = getelementptr inbounds i8, ptr @A, i64 %iv
  %l.A = load i8, ptr %gep.A, align 1
  %branch.cond = icmp sgt i8 %l.A, 0
  %exit.cond = icmp slt i8 %l.A, 42
  br i1 %branch.cond, label %block.a, label %loop.latch

block.a:
  br i1 %exit.cond, label %loop.end, label %loop.latch

loop.latch:
  %iv.next = add i64 %iv, 1
  %ec = icmp ne i64 %iv.next, 64
  br i1 %ec, label %loop.header, label %loop.end

loop.end:
  %retval = phi i64 [ 10, %block.a ], [ %iv, %loop.latch ]
  ret i64 %retval
}

define i64 @livein_exit_cond_in_conditional(i1 %exit.cond) {
; CHECK-LABEL: define i64 @livein_exit_cond_in_conditional(
; CHECK-SAME: i1 [[EXIT_COND:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[EXIT_COND]], 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:    [[IV:%.*]] = phi i64 [ 0, %[[LOOP_HEADER]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i8, ptr @A, i64 [[IV]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[GEP_A]], align 1
; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <4 x i8> [[WIDE_LOAD]], zeroinitializer
; CHECK-NEXT:    [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x i1> [[BROADCAST_SPLAT]], <4 x i1> zeroinitializer
; CHECK-NEXT:    [[TMP3:%.*]] = freeze <4 x i1> [[TMP2]]
; CHECK-NEXT:    [[BRANCH_COND:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP3]])
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
; CHECK-NEXT:    br i1 [[BRANCH_COND]], label %[[BLOCK_A:.*]], label %[[LOOP_LATCH]]
; CHECK:       [[LOOP_LATCH]]:
; CHECK-NEXT:    br i1 [[TMP5]], label %[[LOOP_END:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK:       [[LOOP_END]]:
; CHECK-NEXT:    br label %[[LOOP_END1:.*]]
; CHECK:       [[BLOCK_A]]:
; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP3]], i1 false)
; CHECK-NEXT:    [[TMP7:%.*]] = add i64 [[IV]], [[TMP6]]
; CHECK-NEXT:    br label %[[LOOP_END1]]
; CHECK:       [[LOOP_END1]]:
; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[TMP7]], %[[BLOCK_A]] ], [ 99, %[[LOOP_END]] ]
; CHECK-NEXT:    ret i64 [[RETVAL]]
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ %iv.next, %loop.latch ], [ 0, %entry ]
  %gep.A = getelementptr inbounds i8, ptr @A, i64 %iv
  %l.A = load i8, ptr %gep.A, align 1
  %branch.cond = icmp slt i8 %l.A, 0
  br i1 %branch.cond, label %block.a, label %loop.latch

block.a:
  br i1 %exit.cond, label %loop.end, label %loop.latch

loop.latch:
  %iv.next = add i64 %iv, 1
  %ec = icmp ne i64 %iv.next, 64
  br i1 %ec, label %loop.header, label %loop.end

loop.end:
  %retval = phi i64 [ %iv, %block.a ], [ 99, %loop.latch ]
  ret i64 %retval
}

define i64 @livein_exit_cond_in_conditional2(i1 %exit.cond) {
; CHECK-LABEL: define i64 @livein_exit_cond_in_conditional2(
; CHECK-SAME: i1 [[EXIT_COND:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[EXIT_COND]], 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:    [[IV:%.*]] = phi i64 [ 0, %[[LOOP_HEADER]] ], [ [[INDEX_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i8, ptr @A, i64 [[IV]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[GEP_A]], align 1
; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <4 x i8> [[WIDE_LOAD]], zeroinitializer
; CHECK-NEXT:    [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x i1> [[BROADCAST_SPLAT]], <4 x i1> zeroinitializer
; CHECK-NEXT:    [[TMP3:%.*]] = freeze <4 x i1> [[TMP2]]
; CHECK-NEXT:    [[BRANCH_COND:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP3]])
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
; CHECK-NEXT:    br i1 [[BRANCH_COND]], label %[[BLOCK_A:.*]], label %[[LOOP_LATCH]]
; CHECK:       [[LOOP_LATCH]]:
; CHECK-NEXT:    br i1 [[TMP5]], label %[[LOOP_END:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK:       [[LOOP_END]]:
; CHECK-NEXT:    br label %[[LOOP_END1:.*]]
; CHECK:       [[BLOCK_A]]:
; CHECK-NEXT:    br label %[[LOOP_END1]]
; CHECK:       [[LOOP_END1]]:
; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ 99, %[[BLOCK_A]] ], [ 63, %[[LOOP_END]] ]
; CHECK-NEXT:    ret i64 [[RETVAL]]
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ %iv.next, %loop.latch ], [ 0, %entry ]
  %gep.A = getelementptr inbounds i8, ptr @A, i64 %iv
  %l.A = load i8, ptr %gep.A, align 1
  %branch.cond = icmp slt i8 %l.A, 0
  br i1 %branch.cond, label %block.a, label %loop.latch

block.a:
  br i1 %exit.cond, label %loop.end, label %loop.latch

loop.latch:
  %iv.next = add i64 %iv, 1
  %ec = icmp ne i64 %iv.next, 64
  br i1 %ec, label %loop.header, label %loop.end

loop.end:
  %retval = phi i64 [ 99, %block.a ], [ %iv, %loop.latch ]
  ret i64 %retval
}

define i64 @diamond_with_join_then_exit() {
; CHECK-LABEL: define i64 @diamond_with_join_then_exit() {
; 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_INTERIM:.*]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr @A, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <4 x i8> [[WIDE_LOAD]], zeroinitializer
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr @C, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
; CHECK-NEXT:    [[TMP3:%.*]] = zext <4 x i8> [[WIDE_LOAD1]] to <4 x i64>
; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr @B, i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
; CHECK-NEXT:    [[TMP5:%.*]] = zext <4 x i8> [[WIDE_LOAD2]] to <4 x i64>
; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP1]], <4 x i64> [[TMP5]], <4 x i64> [[TMP3]]
; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq <4 x i64> [[PREDPHI]], splat (i64 42)
; CHECK-NEXT:    [[TMP7:%.*]] = freeze <4 x i1> [[TMP6]]
; CHECK-NEXT:    [[TMP8:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP7]])
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
; CHECK-NEXT:    br i1 [[TMP8]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[VECTOR_BODY_INTERIM]]
; CHECK:       [[VECTOR_BODY_INTERIM]]:
; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    br label %[[LOOP_END:.*]]
; CHECK:       [[VECTOR_EARLY_EXIT]]:
; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP7]], i1 false)
; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i64> [[PREDPHI]], i64 [[FIRST_ACTIVE_LANE]]
; CHECK-NEXT:    br label %[[LOOP_END]]
; CHECK:       [[LOOP_END]]:
; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[TMP10]], %[[VECTOR_EARLY_EXIT]] ], [ 0, %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret i64 [[RETVAL]]
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ %iv.next, %loop.latch ], [ 0, %entry ]
  %gep.A = getelementptr inbounds i8, ptr @A, i64 %iv
  %l.A = load i8, ptr %gep.A, align 1
  %branch.cond = icmp slt i8 %l.A, 0
  br i1 %branch.cond, label %block.a, label %block.b

block.a:
  %gep.B = getelementptr inbounds i8, ptr @B, i64 %iv
  %l.B = load i8, ptr %gep.B, align 1
  %val.a = zext i8 %l.B to i64
  br label %join

block.b:
  %gep.C = getelementptr inbounds i8, ptr @C, i64 %iv
  %l.C = load i8, ptr %gep.C, align 1
  %val.b = zext i8 %l.C to i64
  br label %join

join:
  %val = phi i64 [ %val.a, %block.a ], [ %val.b, %block.b ]
  %cmp = icmp eq i64 %val, 42
  br i1 %cmp, label %loop.end, label %loop.latch

loop.latch:
  %iv.next = add i64 %iv, 1
  %ec = icmp ne i64 %iv.next, 64
  br i1 %ec, label %loop.header, label %loop.end

loop.end:
  %retval = phi i64 [ %val, %join ], [ 0, %loop.latch ]
  ret i64 %retval
}

define i64 @diamond_with_join_then_exit_and_store() {
; CHECK-LABEL: define i64 @diamond_with_join_then_exit_and_store() {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
; CHECK:       [[LOOP_HEADER]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i8, ptr @A, i64 [[IV]]
; CHECK-NEXT:    [[L_A:%.*]] = load i8, ptr [[GEP_A]], align 1
; CHECK-NEXT:    [[BRANCH_COND:%.*]] = icmp slt i8 [[L_A]], 0
; CHECK-NEXT:    br i1 [[BRANCH_COND]], label %[[BLOCK_A:.*]], label %[[BLOCK_B:.*]]
; CHECK:       [[BLOCK_A]]:
; CHECK-NEXT:    [[GEP_B:%.*]] = getelementptr inbounds i8, ptr @B, i64 [[IV]]
; CHECK-NEXT:    [[L_B:%.*]] = load i8, ptr [[GEP_B]], align 1
; CHECK-NEXT:    [[VAL_A:%.*]] = zext i8 [[L_B]] to i64
; CHECK-NEXT:    br label %[[JOIN:.*]]
; CHECK:       [[BLOCK_B]]:
; CHECK-NEXT:    [[GEP_C:%.*]] = getelementptr inbounds i8, ptr @C, i64 [[IV]]
; CHECK-NEXT:    [[L_C:%.*]] = load i8, ptr [[GEP_C]], align 1
; CHECK-NEXT:    store i8 0, ptr [[GEP_C]], align 1
; CHECK-NEXT:    [[VAL_B:%.*]] = zext i8 [[L_C]] to i64
; CHECK-NEXT:    br label %[[JOIN]]
; CHECK:       [[JOIN]]:
; CHECK-NEXT:    [[VAL:%.*]] = phi i64 [ [[VAL_A]], %[[BLOCK_A]] ], [ [[VAL_B]], %[[BLOCK_B]] ]
; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i64 [[VAL]], 42
; CHECK-NEXT:    br i1 [[CMP]], label %[[LOOP_END:.*]], label %[[LOOP_LATCH]]
; CHECK:       [[LOOP_LATCH]]:
; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp ne i64 [[IV_NEXT]], 64
; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP_HEADER]], label %[[LOOP_END]]
; CHECK:       [[LOOP_END]]:
; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[VAL]], %[[JOIN]] ], [ 0, %[[LOOP_LATCH]] ]
; CHECK-NEXT:    ret i64 [[RETVAL]]
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ %iv.next, %loop.latch ], [ 0, %entry ]
  %gep.A = getelementptr inbounds i8, ptr @A, i64 %iv
  %l.A = load i8, ptr %gep.A, align 1
  %branch.cond = icmp slt i8 %l.A, 0
  br i1 %branch.cond, label %block.a, label %block.b

block.a:
  %gep.B = getelementptr inbounds i8, ptr @B, i64 %iv
  %l.B = load i8, ptr %gep.B, align 1
  %val.a = zext i8 %l.B to i64
  br label %join

block.b:
  %gep.C = getelementptr inbounds i8, ptr @C, i64 %iv
  %l.C = load i8, ptr %gep.C, align 1
  store i8 0, ptr %gep.C
  %val.b = zext i8 %l.C to i64
  br label %join

join:
  %val = phi i64 [ %val.a, %block.a ], [ %val.b, %block.b ]
  %cmp = icmp eq i64 %val, 42
  br i1 %cmp, label %loop.end, label %loop.latch

loop.latch:
  %iv.next = add i64 %iv, 1
  %ec = icmp ne i64 %iv.next, 64
  br i1 %ec, label %loop.header, label %loop.end

loop.end:
  %retval = phi i64 [ %val, %join ], [ 0, %loop.latch ]
  ret i64 %retval
}