; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
; REQUIRES: asserts
; RUN: opt -S < %s -p loop-vectorize -debug-only=loop-vectorize -enable-early-exit-vectorization -force-vector-width=4 -disable-output 2>&1 | FileCheck %s

declare void @init_mem(ptr, i64);

; == SOME LEGAL EXAMPLES ==

; The form of the induction variables requires SCEV predicates.
define i32 @diff_exit_block_needs_scev_check(i32 %end) {
; CHECK-LABEL: LV: Checking a loop in 'diff_exit_block_needs_scev_check'
; CHECK:       Found an early exit loop with symbolic max backedge taken count: (-1 + (1 umax (zext i10 (trunc i32 %end to i10) to i32)))<nsw>
; CHECK-NEXT:  LV: We can vectorize this loop!
; CHECK-NOT:   LV: Not vectorizing:
entry:
  %p1 = alloca [1024 x i32]
  %p2 = alloca [1024 x i32]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  %end.clamped = and i32 %end, 1023
  br label %for.body

for.body:
  %ind = phi i8 [ %ind.next, %for.inc ], [ 0, %entry ]
  %gep.ind = phi i64 [ %gep.ind.next, %for.inc ], [ 0, %entry ]
  %arrayidx1 = getelementptr inbounds i32, ptr %p1, i64 %gep.ind
  %0 = load i32, ptr %arrayidx1, align 4
  %arrayidx2 = getelementptr inbounds i32, ptr %p2, i64 %gep.ind
  %1 = load i32, ptr %arrayidx2, align 4
  %cmp.early = icmp eq i32 %0, %1
  br i1 %cmp.early, label %found, label %for.inc

for.inc:
  %ind.next = add i8 %ind, 1
  %conv = zext i8 %ind.next to i32
  %gep.ind.next = add i64 %gep.ind, 1
  %cmp = icmp ult i32 %conv, %end.clamped
  br i1 %cmp, label %for.body, label %exit

found:
  ret i32 1

exit:
  ret i32 0
}


define i64 @same_exit_block_pre_inc_use1() {
; CHECK-LABEL: LV: Checking a loop in 'same_exit_block_pre_inc_use1'
; CHECK:       LV: Found an early exit loop with symbolic max backedge taken count: 63
; CHECK-NEXT:  LV: We can vectorize this loop!
; CHECK:  LV: Not vectorizing: Some exit values in loop with uncountable exit not supported yet.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %cmp3 = icmp eq i8 %ld1, %ld2
  br i1 %cmp3, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @loop_contains_safe_call() {
; CHECK-LABEL: LV: Checking a loop in 'loop_contains_safe_call'
; CHECK:       LV: Found an early exit loop with symbolic max backedge taken count: 63
; CHECK-NEXT:  LV: We can vectorize this loop!
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds float, ptr %p1, i64 %index
  %ld1 = load float, ptr %arrayidx, align 1
  %sqrt = tail call fast float @llvm.sqrt.f32(float %ld1)
  %cmp = fcmp fast ult float %sqrt, 3.0e+00
  br i1 %cmp, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @loop_contains_safe_div() {
; CHECK-LABEL: LV: Checking a loop in 'loop_contains_safe_div'
; CHECK:       LV: Found an early exit loop with symbolic max backedge taken count: 63
; CHECK-NEXT:  LV: We can vectorize this loop!
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i32, ptr %p1, i64 %index
  %ld1 = load i32, ptr %arrayidx, align 1
  %div = udiv i32 %ld1, 20000
  %cmp = icmp eq i32 %div, 1
  br i1 %cmp, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @loop_contains_load_after_early_exit(ptr dereferenceable(1024) align(8) %p2) {
; CHECK-LABEL: LV: Checking a loop in 'loop_contains_load_after_early_exit'
; CHECK:       LV: Found an early exit loop with symbolic max backedge taken count: 63
; CHECK-NEXT:  LV: We can vectorize this loop!
; CHECK:       LV: Not vectorizing: Some exit values in loop with uncountable exit not supported yet.
entry:
  %p1 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i32, ptr %p1, i64 %index
  %ld1 = load i32, ptr %arrayidx, align 1
  %cmp = icmp eq i32 %ld1, 1
  br i1 %cmp, label %loop.inc, label %loop.end

loop.inc:
  %arrayidx2 = getelementptr inbounds i64, ptr %p2, i64 %index
  %ld2 = load i64, ptr %arrayidx2, align 8
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ %ld2, %loop.inc ]
  ret i64 %retval
}


define i64 @one_uncountable_two_countable_same_exit_phi_of_consts() {
; CHECK-LABEL: LV: Checking a loop in 'one_uncountable_two_countable_same_exit_phi_of_consts'
; CHECK:       LV: Found an early exit loop with symbolic max backedge taken count: 61
; CHECK-NEXT:  LV: We can vectorize this loop!
; CHECK-NEXT:  LV: Not vectorizing: Auto-vectorization of early exit loops requiring a scalar epilogue is unsupported.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %cmp1 = icmp ne i64 %index, 64
  br i1 %cmp1, label %search, label %loop.end

search:
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %cmp3 = icmp eq i8 %ld1, %ld2
  br i1 %cmp3, label %loop.end, label %loop.inc

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 128
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ 0, %loop ], [ 1, %search ], [ 0, %loop.inc ]
  ret i64 %retval
}


; == SOME ILLEGAL EXAMPLES ==


define i64 @same_exit_block_pre_inc_use1_too_small_allocas() {
; CHECK-LABEL: LV: Checking a loop in 'same_exit_block_pre_inc_use1_too_small_allocas'
; CHECK:       LV: Not vectorizing: Loop may fault.
entry:
  %p1 = alloca [42 x i8]
  %p2 = alloca [42 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %cmp3 = icmp eq i8 %ld1, %ld2
  br i1 %cmp3, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @same_exit_block_pre_inc_use1_too_small_deref_ptrs(ptr dereferenceable(42) %p1, ptr dereferenceable(42) %p2) {
; CHECK-LABEL: LV: Checking a loop in 'same_exit_block_pre_inc_use1_too_small_deref_ptrs'
; CHECK:       LV: Not vectorizing: Loop may fault.
entry:
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %cmp3 = icmp eq i8 %ld1, %ld2
  br i1 %cmp3, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @same_exit_block_pre_inc_use1_unknown_ptrs(ptr %p1, ptr %p2) {
; CHECK-LABEL: LV: Checking a loop in 'same_exit_block_pre_inc_use1_unknown_ptrs'
; CHECK:       LV: Not vectorizing: Loop may fault.
entry:
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %cmp3 = icmp eq i8 %ld1, %ld2
  br i1 %cmp3, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


; The early exit (i.e. unknown exit-not-taken count) is the latch - we don't
; support this yet.
define i64 @uncountable_exit_on_last_block() {
; CHECK-LABEL: LV: Checking a loop in 'uncountable_exit_on_last_block'
; CHECK:       LV: Not vectorizing: Early exit is not the latch predecessor.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %search ], [ 3, %entry ]
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %search, label %loop.end

search:
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %cmp3 = icmp eq i8 %ld1, %ld2
  br i1 %cmp3, label %loop.end, label %loop

loop.end:
  %retval = phi i64 [ 64, %loop ], [ %index, %search ]
  ret i64 %retval
}


; We don't currently support multiple uncountable early exits.
define i64 @multiple_uncountable_exits() {
; CHECK-LABEL: LV: Checking a loop in 'multiple_uncountable_exits'
; CHECK:       LV: Not vectorizing: Loop has too many uncountable exits.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %search1

search1:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %cmp1 = icmp eq i8 %ld1, %ld2
  br i1 %cmp1, label %loop.end, label %search2

search2:
  %cmp2 = icmp ult i8 %ld1, 34
  br i1 %cmp2, label %loop.end, label %loop.inc

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %search1, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %search1 ], [ 100, %search2 ], [ 43, %loop.inc ]
  ret i64 %retval
}


define i64 @uncountable_exit_infinite_loop() {
; CHECK-LABEL: LV: Checking a loop in 'uncountable_exit_infinite_loop'
; CHECK:       LV: Not vectorizing: Cannot vectorize uncountable loop.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %cmp3 = icmp eq i8 %ld1, %ld2
  br i1 %cmp3, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br label %loop

loop.end:
  %retval = phi i64 [ %index, %loop ]
  ret i64 %retval
}


define i64 @loop_contains_unsafe_call() {
; CHECK-LABEL: LV: Checking a loop in 'loop_contains_unsafe_call'
; CHECK:       LV: Not vectorizing: Early exit loop contains operations that cannot be speculatively executed.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i32, ptr %p1, i64 %index
  %ld1 = load i32, ptr %arrayidx, align 1
  %bad_call = call i32 @foo(i32 %ld1) #0
  %cmp = icmp eq i32 %bad_call, 34
  br i1 %cmp, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @loop_contains_unsafe_div() {
; CHECK-LABEL: LV: Checking a loop in 'loop_contains_unsafe_div'
; CHECK:       LV: Not vectorizing: Early exit loop contains operations that cannot be speculatively executed.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i32, ptr %arrayidx, align 1
  %div = udiv i32 20000, %ld1
  %cmp = icmp eq i32 %div, 1
  br i1 %cmp, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @loop_contains_store(ptr %dest) {
; CHECK-LABEL: LV: Checking a loop in 'loop_contains_store'
; CHECK:       LV: Not vectorizing: Writes to memory unsupported in early exit loops
entry:
  %p1 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i32, ptr %p1, i64 %index
  %ld1 = load i32, ptr %arrayidx, align 1
  %arrayidx2 = getelementptr inbounds i32, ptr %dest, i64 %index
  store i32 %ld1, ptr %arrayidx2, align 4
  %cmp = icmp eq i32 %ld1, 1
  br i1 %cmp, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @uncountable_exit_in_conditional_block(ptr %mask) {
; CHECK-LABEL: LV: Checking a loop in 'uncountable_exit_in_conditional_block'
; CHECK:       LV: Not vectorizing: Early exit is not the latch predecessor.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx1 = getelementptr inbounds i8, ptr %mask, i64 %index
  %ld1 = load i8, ptr %arrayidx1, align 1
  %cmp1 = icmp ne i8 %ld1, 0
  br i1 %cmp1, label %loop.search, label %loop.inc

loop.search:
  %arrayidx2 = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld2 = load i8, ptr %arrayidx2, align 1
  %arrayidx3 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld3 = load i8, ptr %arrayidx3, align 1
  %cmp2 = icmp eq i8 %ld2, %ld3
  br i1 %cmp2, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %retval = phi i64 [ %index, %loop.search ], [ 67, %loop.inc ]
  ret i64 %retval
}


define i64 @same_exit_block_pre_inc_use1_with_reduction() {
; CHECK-LABEL: LV: Checking a loop in 'same_exit_block_pre_inc_use1_with_reduction'
; CHECK:       LV: Not vectorizing: Found reductions or recurrences in early-exit loop.
entry:
  %p1 = alloca [1024 x i8]
  %p2 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  call void @init_mem(ptr %p2, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %red = phi i64 [ %red.next, %loop.inc ], [ 0, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index
  %ld2 = load i8, ptr %arrayidx1, align 1
  %ld2.zext = zext i8 %ld2 to i64
  %red.next = add i64 %red, %ld2.zext
  %cmp3 = icmp eq i8 %ld1, %ld2
  br i1 %cmp3, label %loop.inc, label %loop.end

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.end:
  %final.ind = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  %retval = add i64 %red.next, %final.ind
  ret i64 %retval
}


define i64 @uncountable_exit_has_multiple_outside_successors() {
; CHECK-LABEL: LV: Checking a loop in 'uncountable_exit_has_multiple_outside_successors'
; CHECK:       LV: Not vectorizing: Loop contains an unsupported switch
entry:
  %p1 = alloca [1024 x i8]
  call void @init_mem(ptr %p1, i64 1024)
  br label %loop

loop:
  %index = phi i64 [ %index.next, %loop.inc ], [ 3, %entry ]
  %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index
  %ld1 = load i8, ptr %arrayidx, align 1
  switch i8 %ld1, label %loop.inc [
  i8 2, label %loop.end
  i8 3, label %loop.surprise
  ]

loop.inc:
  %index.next = add i64 %index, 1
  %exitcond = icmp ne i64 %index.next, 67
  br i1 %exitcond, label %loop, label %loop.end

loop.surprise:
  ret i64 3

loop.end:
  %retval = phi i64 [ %index, %loop ], [ 67, %loop.inc ]
  ret i64 %retval
}


declare i32 @foo(i32) readonly
declare <vscale x 4 x i32> @foo_vec(<vscale x 4 x i32>)

attributes #0 = { "vector-function-abi-variant"="_ZGVsNxv_foo(foo_vec)" }