; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -verify-machineinstrs < %s -mtriple=aarch64-none-linux-gnu | FileCheck %s

declare void @callee_stack0()
declare void @callee_stack8([8 x i64], i64)
declare void @callee_stack16([8 x i64], i64, i64)

define dso_local void @caller_to0_from0() nounwind {
; CHECK-LABEL: caller_to0_from0:
; CHECK:       // %bb.0:
; CHECK-NEXT:    b callee_stack0
  tail call void @callee_stack0()
  ret void
}

define dso_local void @caller_to0_from8([8 x i64], i64) nounwind{
; CHECK-LABEL: caller_to0_from8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    b callee_stack0

  tail call void @callee_stack0()
  ret void
}

define dso_local void @caller_to8_from0() {
; CHECK-LABEL: caller_to8_from0:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sub sp, sp, #32
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-NEXT:    .cfi_offset w30, -16
; CHECK-NEXT:    mov w8, #42
; CHECK-NEXT:    str x8, [sp]
; CHECK-NEXT:    bl callee_stack8
; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-NEXT:    add sp, sp, #32
; CHECK-NEXT:    ret

; Caller isn't going to clean up any extra stack we allocate, so it
; can't be a tail call.
  tail call void @callee_stack8([8 x i64] undef, i64 42)
  ret void
}

define dso_local void @caller_to8_from8([8 x i64], i64 %a) {
; CHECK-LABEL: caller_to8_from8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, #42
; CHECK-NEXT:    str x8, [sp]
; CHECK-NEXT:    b callee_stack8

; This should reuse our stack area for the 42
  tail call void @callee_stack8([8 x i64] undef, i64 42)
  ret void
}

define dso_local void @caller_to16_from8([8 x i64], i64 %a) {
; CHECK-LABEL: caller_to16_from8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sub sp, sp, #32
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-NEXT:    .cfi_offset w30, -16
; CHECK-NEXT:    bl callee_stack16
; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-NEXT:    add sp, sp, #32
; CHECK-NEXT:    ret

; Shouldn't be a tail call: we can't use SP+8 because our caller might
; have something there. This may sound obvious but implementation does
; some funky aligning.
  tail call void @callee_stack16([8 x i64] undef, i64 undef, i64 undef)
  ret void
}

define dso_local void @caller_to8_from24([8 x i64], i64 %a, i64 %b, i64 %c) {
; CHECK-LABEL: caller_to8_from24:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, #42
; CHECK-NEXT:    str x8, [sp]
; CHECK-NEXT:    b callee_stack8

; Reuse our area, putting "42" at incoming sp
  tail call void @callee_stack8([8 x i64] undef, i64 42)
  ret void
}

define dso_local void @caller_to16_from16([8 x i64], i64 %a, i64 %b) {
; CHECK-LABEL: caller_to16_from16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldp x8, x9, [sp]
; CHECK-NEXT:    stp x9, x8, [sp]
; CHECK-NEXT:    b callee_stack16

; Here we want to make sure that both loads happen before the stores:
; otherwise either %a or %b will be wrongly clobbered.
  tail call void @callee_stack16([8 x i64] undef, i64 %b, i64 %a)
  ret void


}

@func = dso_local global void(i32)* null

define dso_local void @indirect_tail() {
; CHECK-LABEL: indirect_tail:
; CHECK:       // %bb.0:
; CHECK-NEXT:    adrp x8, func
; CHECK-NEXT:    mov w0, #42
; CHECK-NEXT:    ldr x1, [x8, :lo12:func]
; CHECK-NEXT:    br x1

  %fptr = load void(i32)*, void(i32)** @func
  tail call void %fptr(i32 42)
  ret void
}