# REQUIRES: asserts
# RUN: llc --mtriple=hexagon %s -run-pass=pipeliner -debug-only=pipeliner \
# RUN: -window-sched=force -filetype=null -verify-machineinstrs 2>&1 \
# RUN: | FileCheck %s

# CHECK-NOT: PSEUDO_PROBE
# CHECK: Best window offset is {{[0-9]+}} and Best II is {{[0-9]+}}.

--- |
  define void @exp_approx_top_six(i32 %N, ptr noalias %x, ptr noalias %y) #0 {
  entry:
    %is_zero = icmp eq i32 %N, 0
    br i1 %is_zero, label %exit, label %loop_header

  loop_header:
    %vec_one = tail call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 1065353216)
    %vec_half = tail call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 1056964608)
    %vec_sixth = tail call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 1042983595)
    %vec_24th = tail call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 1026206379)
    %vec_120th = tail call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 1007192201)
    %vec_720th = tail call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 985008993)
    br label %loop_body

  exit:
    ret void

  loop_body:
    %lsr.iv1 = phi ptr [ %cgep3, %loop_body ], [ %x, %loop_header ]
    %lsr.iv = phi ptr [ %cgep, %loop_body ], [ %y, %loop_header ]
    %index = phi i32 [ 0, %loop_header ], [ %index_next, %loop_body ]
    %vec_input = load <32 x i32>, ptr %lsr.iv1, align 128
    %vec_input_pow_2 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_input, <32 x i32> %vec_input)
    %vec_input_pow_3 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_input_pow_2, <32 x i32> %vec_input)
    %vec_input_pow_4 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_input_pow_3, <32 x i32> %vec_input)
    %vec_input_pow_5 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_input_pow_4, <32 x i32> %vec_input)
    %vec_input_pow_6 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_input_pow_5, <32 x i32> %vec_input)
    %vec_exp_approx_1 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_half, <32 x i32> %vec_input_pow_2)
    %vec_exp_approx_2 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_sixth, <32 x i32> %vec_input_pow_3)
    %vec_exp_approx_3 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_24th, <32 x i32> %vec_input_pow_4)
    %vec_exp_approx_4 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_120th, <32 x i32> %vec_input_pow_5)
    %vec_exp_approx_5 = tail call <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32> %vec_720th, <32 x i32> %vec_input_pow_6)
    %vec_exp_sum_1 = tail call <32 x i32> @llvm.hexagon.V6.vaddw.128B(<32 x i32> %vec_one, <32 x i32> %vec_input)
    %vec_exp_sum_2 = tail call <32 x i32> @llvm.hexagon.V6.vaddw.128B(<32 x i32> %vec_exp_sum_1, <32 x i32> %vec_exp_approx_1)
    %vec_exp_sum_3 = tail call <32 x i32> @llvm.hexagon.V6.vaddw.128B(<32 x i32> %vec_exp_sum_2, <32 x i32> %vec_exp_approx_2)
    %vec_exp_sum_4 = tail call <32 x i32> @llvm.hexagon.V6.vaddw.128B(<32 x i32> %vec_exp_sum_3, <32 x i32> %vec_exp_approx_3)
    %vec_exp_sum_5 = tail call <32 x i32> @llvm.hexagon.V6.vaddw.128B(<32 x i32> %vec_exp_sum_4, <32 x i32> %vec_exp_approx_4)
    %vec_exp_result = tail call <32 x i32> @llvm.hexagon.V6.vaddw.128B(<32 x i32> %vec_exp_sum_5, <32 x i32> %vec_exp_approx_5)
    store <32 x i32> %vec_exp_result, ptr %lsr.iv, align 128
    %index_next = add nuw i32 %index, 32
    %loop_cond = icmp ult i32 %index_next, %N
    %cgep = getelementptr i8, ptr %lsr.iv, i32 128
    %cgep3 = getelementptr i8, ptr %lsr.iv1, i32 128
    br i1 %loop_cond, label %loop_body, label %exit
  }

  declare <32 x i32> @llvm.hexagon.V6.vmpyowh.rnd.128B(<32 x i32>, <32 x i32>)
  declare <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32)
  declare <32 x i32> @llvm.hexagon.V6.vaddw.128B(<32 x i32>, <32 x i32>)

  attributes #0 = { "target-features"="+hvx-length128b,+hvxv69,+v66,-long-calls" }
...
---
name:            exp_approx_top_six
tracksRegLiveness: true
body:             |
  bb.0.entry:
    successors: %bb.2(0x30000000), %bb.1(0x50000000)
    liveins: $r0, $r1, $r2

    %0:intregs = COPY $r2
    %1:intregs = COPY $r1
    %2:intregs = COPY $r0
    %3:predregs = C2_cmpeqi %2, 0
    J2_jumpt killed %3, %bb.2, implicit-def dead $pc
    J2_jump %bb.1, implicit-def dead $pc

  bb.1.loop_header:
    successors: %bb.3(0x80000000)

    %4:intregs = A2_tfrsi 1065353216
    %5:hvxvr = V6_lvsplatw killed %4
    %6:intregs = A2_tfrsi 1056964608
    %7:hvxvr = V6_lvsplatw killed %6
    %8:intregs = A2_tfrsi 1042983595
    %9:hvxvr = V6_lvsplatw killed %8
    %10:intregs = A2_tfrsi 1026206379
    %11:hvxvr = V6_lvsplatw killed %10
    %12:intregs = A2_tfrsi 1007192201
    %13:hvxvr = V6_lvsplatw killed %12
    %14:intregs = A2_tfrsi 985008993
    %15:hvxvr = V6_lvsplatw killed %14
    %16:intregs = A2_addi %2, 31
    %17:intregs = S2_lsr_i_r %16, 5
    %18:intregs = COPY %17
    J2_loop0r %bb.3, %18, implicit-def $lc0, implicit-def $sa0, implicit-def $usr
    J2_jump %bb.3, implicit-def dead $pc

  bb.2.exit:
    PS_jmpret $r31, implicit-def dead $pc

  bb.3.loop_body (machine-block-address-taken):
    successors: %bb.3(0x7c000000), %bb.2(0x04000000)

    %19:intregs = PHI %1, %bb.1, %20, %bb.3
    %21:intregs = PHI %0, %bb.1, %22, %bb.3
    %23:hvxvr, %20:intregs = V6_vL32b_pi %19, 128 :: (load (s1024) from %ir.lsr.iv1)
    %24:hvxvr = V6_vmpyowh_rnd %23, %23
    %25:hvxvr = V6_vmpyowh_rnd %24, %23
    %26:hvxvr = V6_vmpyowh_rnd %25, %23
    %27:hvxvr = V6_vmpyowh_rnd %26, %23
    %28:hvxvr = V6_vmpyowh_rnd %27, %23
    %29:hvxvr = V6_vmpyowh_rnd %7, %24
    %30:hvxvr = V6_vmpyowh_rnd %9, %25
    %31:hvxvr = V6_vmpyowh_rnd %11, %26
    %32:hvxvr = V6_vmpyowh_rnd %13, %27
    ; To check the meta MI within loop kernel.
    PSEUDO_PROBE 128, 1, 0, 0
    %33:hvxvr = V6_vmpyowh_rnd %15, killed %28
    %34:hvxvr = V6_vaddw %5, %23
    %35:hvxvr = V6_vaddw killed %34, killed %29
    %36:hvxvr = V6_vaddw killed %35, killed %30
    %37:hvxvr = V6_vaddw killed %36, killed %31
    %38:hvxvr = V6_vaddw killed %37, killed %32
    %39:hvxvr = V6_vaddw killed %38, killed %33
    %22:intregs = V6_vS32b_pi %21, 128, killed %39 :: (store (s1024) into %ir.lsr.iv)
    ENDLOOP0 %bb.3, implicit-def $pc, implicit-def $lc0, implicit $sa0, implicit $lc0
    J2_jump %bb.2, implicit-def dead $pc
...