llvm/test/CodeGen/ARM/vselect_imax.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: opt < %s -passes='print<cost-model>' -mtriple=arm-apple-ios6.0.0 -mcpu=cortex-a8 2>&1 -disable-output | FileCheck %s --check-prefix=COST
; RUN: llc -mtriple=arm-eabi -mattr=+neon %s -o - | FileCheck %s
; Make sure that ARM backend with NEON handles vselect.

define void @vmax_v4i32(ptr %m, <4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: vmax_v4i32:
; CHECK:       @ %bb.0:
; CHECK-NEXT:    add r1, sp, #8
; CHECK-NEXT:    vldr d17, [sp]
; CHECK-NEXT:    vld1.64 {d18, d19}, [r1]
; CHECK-NEXT:    vmov d16, r2, r3
; CHECK-NEXT:    vmax.s32 q8, q8, q9
; CHECK-NEXT:    vst1.64 {d16, d17}, [r0]
; CHECK-NEXT:    mov pc, lr
    %cmpres = icmp sgt <4 x i32> %a, %b
    %maxres = select <4 x i1> %cmpres, <4 x i32> %a,  <4 x i32> %b
    store <4 x i32> %maxres, ptr %m
    ret void
}

%T0_10 = type <16 x i16>
%T1_10 = type <16 x i1>
define void @func_blend10(ptr %loadaddr, ptr %loadaddr2, ptr %blend, ptr %storeaddr) {
; CHECK-LABEL: func_blend10:
; CHECK:       @ %bb.0:
; CHECK-NEXT:    vld1.16 {d16, d17}, [r1:128]!
; CHECK-NEXT:    vld1.16 {d18, d19}, [r0:128]!
; CHECK-NEXT:    vmin.s16 q8, q9, q8
; CHECK-NEXT:    vld1.64 {d20, d21}, [r1:128]
; CHECK-NEXT:    vld1.64 {d18, d19}, [r0:128]
; CHECK-NEXT:    vmin.s16 q9, q9, q10
; CHECK-NEXT:    vst1.16 {d16, d17}, [r3:128]!
; CHECK-NEXT:    vst1.64 {d18, d19}, [r3:128]
; CHECK-NEXT:    mov pc, lr
; COST: func_blend10
; COST: cost of 0 {{.*}} icmp
; COST: cost of 4 {{.*}} select

  %v0 = load %T0_10, ptr %loadaddr
  %v1 = load %T0_10, ptr %loadaddr2
  %c = icmp slt %T0_10 %v0, %v1
  %r = select %T1_10 %c, %T0_10 %v0, %T0_10 %v1
  store %T0_10 %r, ptr %storeaddr
  ret void
}

%T0_14 = type <8 x i32>
%T1_14 = type <8 x i1>
define void @func_blend14(ptr %loadaddr, ptr %loadaddr2, ptr %blend, ptr %storeaddr) {
; CHECK-LABEL: func_blend14:
; CHECK:       @ %bb.0:
; CHECK-NEXT:    vld1.32 {d16, d17}, [r1:128]!
; CHECK-NEXT:    vld1.32 {d18, d19}, [r0:128]!
; CHECK-NEXT:    vmin.s32 q8, q9, q8
; CHECK-NEXT:    vld1.64 {d20, d21}, [r1:128]
; CHECK-NEXT:    vld1.64 {d18, d19}, [r0:128]
; CHECK-NEXT:    vmin.s32 q9, q9, q10
; CHECK-NEXT:    vst1.32 {d16, d17}, [r3:128]!
; CHECK-NEXT:    vst1.64 {d18, d19}, [r3:128]
; CHECK-NEXT:    mov pc, lr
; COST: func_blend14
; COST: cost of 0 {{.*}} icmp
; COST: cost of 4 {{.*}} select
  %v0 = load %T0_14, ptr %loadaddr
  %v1 = load %T0_14, ptr %loadaddr2
  %c = icmp slt %T0_14 %v0, %v1
  %r = select %T1_14 %c, %T0_14 %v0, %T0_14 %v1
  store %T0_14 %r, ptr %storeaddr
  ret void
}

%T0_15 = type <16 x i32>
%T1_15 = type <16 x i1>
define void @func_blend15(ptr %loadaddr, ptr %loadaddr2, ptr %blend, ptr %storeaddr) {
; CHECK-LABEL: func_blend15:
; CHECK:       @ %bb.0:
; CHECK-NEXT:    vld1.32 {d16, d17}, [r1:128]!
; CHECK-NEXT:    vld1.32 {d18, d19}, [r0:128]!
; CHECK-NEXT:    vmin.s32 q8, q9, q8
; CHECK-NEXT:    vld1.32 {d20, d21}, [r1:128]!
; CHECK-NEXT:    vld1.32 {d22, d23}, [r0:128]!
; CHECK-NEXT:    vmin.s32 q10, q11, q10
; CHECK-NEXT:    vld1.32 {d24, d25}, [r1:128]!
; CHECK-NEXT:    vld1.32 {d26, d27}, [r0:128]!
; CHECK-NEXT:    vmin.s32 q12, q13, q12
; CHECK-NEXT:    vld1.64 {d18, d19}, [r1:128]
; CHECK-NEXT:    vld1.64 {d22, d23}, [r0:128]
; CHECK-NEXT:    vmin.s32 q9, q11, q9
; CHECK-NEXT:    vst1.32 {d16, d17}, [r3:128]!
; CHECK-NEXT:    vst1.32 {d20, d21}, [r3:128]!
; CHECK-NEXT:    vst1.32 {d24, d25}, [r3:128]!
; CHECK-NEXT:    vst1.64 {d18, d19}, [r3:128]
; CHECK-NEXT:    mov pc, lr
; COST: func_blend15
; COST: cost of 0 {{.*}} icmp
; COST: cost of 8 {{.*}} select

  %v0 = load %T0_15, ptr %loadaddr
  %v1 = load %T0_15, ptr %loadaddr2
  %c = icmp slt %T0_15 %v0, %v1
  %r = select %T1_15 %c, %T0_15 %v0, %T0_15 %v1
  store %T0_15 %r, ptr %storeaddr
  ret void
}

; We adjusted the cost model of the following selects. When we improve code
; lowering we also need to adjust the cost.
%T0_18 = type <4 x i64>
%T1_18 = type <4 x i1>
define void @func_blend18(ptr %loadaddr, ptr %loadaddr2, ptr %blend, ptr %storeaddr) {
; CHECK-LABEL: func_blend18:
; CHECK:       @ %bb.0:
; CHECK-NEXT:    .save {r4, r5, r6, lr}
; CHECK-NEXT:    push {r4, r5, r6, lr}
; CHECK-NEXT:    vld1.64 {d20, d21}, [r1:128]!
; CHECK-NEXT:    vld1.64 {d22, d23}, [r0:128]!
; CHECK-NEXT:    vmov r4, r6, d21
; CHECK-NEXT:    vld1.64 {d16, d17}, [r1:128]
; CHECK-NEXT:    vld1.64 {d18, d19}, [r0:128]
; CHECK-NEXT:    vmov lr, r12, d16
; CHECK-NEXT:    mov r0, #0
; CHECK-NEXT:    vmov r2, r1, d18
; CHECK-NEXT:    subs r2, r2, lr
; CHECK-NEXT:    vmov r2, r5, d23
; CHECK-NEXT:    sbcs r1, r1, r12
; CHECK-NEXT:    mov r1, #0
; CHECK-NEXT:    movlt r1, #1
; CHECK-NEXT:    cmp r1, #0
; CHECK-NEXT:    mvnne r1, #0
; CHECK-NEXT:    subs r2, r2, r4
; CHECK-NEXT:    sbcs r6, r5, r6
; CHECK-NEXT:    vmov r2, r12, d20
; CHECK-NEXT:    vmov r5, r4, d22
; CHECK-NEXT:    mov r6, #0
; CHECK-NEXT:    movlt r6, #1
; CHECK-NEXT:    cmp r6, #0
; CHECK-NEXT:    mvnne r6, #0
; CHECK-NEXT:    subs r2, r5, r2
; CHECK-NEXT:    sbcs r2, r4, r12
; CHECK-NEXT:    vmov lr, r12, d17
; CHECK-NEXT:    vmov r4, r5, d19
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d24, r2
; CHECK-NEXT:    vdup.32 d25, r6
; CHECK-NEXT:    vbit q10, q11, q12
; CHECK-NEXT:    vdup.32 d22, r1
; CHECK-NEXT:    vst1.64 {d20, d21}, [r3:128]!
; CHECK-NEXT:    subs r4, r4, lr
; CHECK-NEXT:    sbcs r5, r5, r12
; CHECK-NEXT:    movlt r0, #1
; CHECK-NEXT:    cmp r0, #0
; CHECK-NEXT:    mvnne r0, #0
; CHECK-NEXT:    vdup.32 d23, r0
; CHECK-NEXT:    vbit q8, q9, q11
; CHECK-NEXT:    vst1.64 {d16, d17}, [r3:128]
; CHECK-NEXT:    pop {r4, r5, r6, lr}
; CHECK-NEXT:    mov pc, lr
; COST: func_blend18
; COST: cost of 0 {{.*}} icmp
; COST: cost of 21 {{.*}} select
  %v0 = load %T0_18, ptr %loadaddr
  %v1 = load %T0_18, ptr %loadaddr2
  %c = icmp slt %T0_18 %v0, %v1
  %r = select %T1_18 %c, %T0_18 %v0, %T0_18 %v1
  store %T0_18 %r, ptr %storeaddr
  ret void
}

%T0_19 = type <8 x i64>
%T1_19 = type <8 x i1>
define void @func_blend19(ptr %loadaddr, ptr %loadaddr2, ptr %blend, ptr %storeaddr) {
; CHECK-LABEL: func_blend19:
; CHECK:       @ %bb.0:
; CHECK-NEXT:    .save {r4, r5, r6, lr}
; CHECK-NEXT:    push {r4, r5, r6, lr}
; CHECK-NEXT:    vld1.64 {d28, d29}, [r1:128]!
; CHECK-NEXT:    mov r12, #0
; CHECK-NEXT:    vld1.64 {d30, d31}, [r0:128]!
; CHECK-NEXT:    vmov r2, lr, d28
; CHECK-NEXT:    vmov r4, r5, d30
; CHECK-NEXT:    vld1.64 {d16, d17}, [r0:128]!
; CHECK-NEXT:    vld1.64 {d18, d19}, [r1:128]!
; CHECK-NEXT:    vld1.64 {d20, d21}, [r1:128]!
; CHECK-NEXT:    vld1.64 {d26, d27}, [r0:128]!
; CHECK-NEXT:    vld1.64 {d22, d23}, [r1:128]
; CHECK-NEXT:    vld1.64 {d24, d25}, [r0:128]
; CHECK-NEXT:    vmov r0, r1, d23
; CHECK-NEXT:    subs r2, r4, r2
; CHECK-NEXT:    vmov r4, r6, d31
; CHECK-NEXT:    sbcs r2, r5, lr
; CHECK-NEXT:    vmov r5, lr, d29
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d0, r2
; CHECK-NEXT:    subs r4, r4, r5
; CHECK-NEXT:    vmov r2, r5, d25
; CHECK-NEXT:    sbcs r6, r6, lr
; CHECK-NEXT:    mov r6, #0
; CHECK-NEXT:    movlt r6, #1
; CHECK-NEXT:    cmp r6, #0
; CHECK-NEXT:    mvnne r6, #0
; CHECK-NEXT:    vdup.32 d1, r6
; CHECK-NEXT:    vbit q14, q15, q0
; CHECK-NEXT:    vst1.64 {d28, d29}, [r3:128]!
; CHECK-NEXT:    subs r0, r2, r0
; CHECK-NEXT:    sbcs r0, r5, r1
; CHECK-NEXT:    vmov r1, r2, d21
; CHECK-NEXT:    vmov r6, r5, d27
; CHECK-NEXT:    mov r0, #0
; CHECK-NEXT:    movlt r0, #1
; CHECK-NEXT:    cmp r0, #0
; CHECK-NEXT:    mvnne r0, #0
; CHECK-NEXT:    subs r1, r6, r1
; CHECK-NEXT:    sbcs r1, r5, r2
; CHECK-NEXT:    vmov r2, r6, d18
; CHECK-NEXT:    vmov r5, r4, d16
; CHECK-NEXT:    mov r1, #0
; CHECK-NEXT:    movlt r1, #1
; CHECK-NEXT:    cmp r1, #0
; CHECK-NEXT:    mvnne r1, #0
; CHECK-NEXT:    subs r2, r5, r2
; CHECK-NEXT:    sbcs r2, r4, r6
; CHECK-NEXT:    vmov r6, lr, d19
; CHECK-NEXT:    vmov r4, r5, d17
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d30, r2
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    subs r4, r4, r6
; CHECK-NEXT:    sbcs r6, r5, lr
; CHECK-NEXT:    vmov r4, r5, d26
; CHECK-NEXT:    vmov r6, lr, d20
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d31, r2
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    vbif q8, q9, q15
; CHECK-NEXT:    vst1.64 {d16, d17}, [r3:128]!
; CHECK-NEXT:    subs r4, r4, r6
; CHECK-NEXT:    sbcs r6, r5, lr
; CHECK-NEXT:    vmov r4, r5, d24
; CHECK-NEXT:    vmov r6, lr, d22
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d0, r2
; CHECK-NEXT:    vdup.32 d1, r1
; CHECK-NEXT:    vorr q9, q0, q0
; CHECK-NEXT:    vbsl q9, q13, q10
; CHECK-NEXT:    vst1.64 {d18, d19}, [r3:128]!
; CHECK-NEXT:    subs r4, r4, r6
; CHECK-NEXT:    sbcs r6, r5, lr
; CHECK-NEXT:    movlt r12, #1
; CHECK-NEXT:    cmp r12, #0
; CHECK-NEXT:    mvnne r12, #0
; CHECK-NEXT:    vdup.32 d2, r12
; CHECK-NEXT:    vdup.32 d3, r0
; CHECK-NEXT:    vorr q10, q1, q1
; CHECK-NEXT:    vbsl q10, q12, q11
; CHECK-NEXT:    vst1.64 {d20, d21}, [r3:128]
; CHECK-NEXT:    pop {r4, r5, r6, lr}
; CHECK-NEXT:    mov pc, lr
; COST: func_blend19
; COST: cost of 0 {{.*}} icmp
; COST: cost of 54 {{.*}} select
  %v0 = load %T0_19, ptr %loadaddr
  %v1 = load %T0_19, ptr %loadaddr2
  %c = icmp slt %T0_19 %v0, %v1
  %r = select %T1_19 %c, %T0_19 %v0, %T0_19 %v1
  store %T0_19 %r, ptr %storeaddr
  ret void
}

%T0_20 = type <16 x i64>
%T1_20 = type <16 x i1>
define void @func_blend20(ptr %loadaddr, ptr %loadaddr2, ptr %blend, ptr %storeaddr) {
; CHECK-LABEL: func_blend20:
; CHECK:       @ %bb.0:
; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, lr}
; CHECK-NEXT:    push {r4, r5, r6, r7, r8, lr}
; CHECK-NEXT:    .vsave {d8, d9}
; CHECK-NEXT:    vpush {d8, d9}
; CHECK-NEXT:    add lr, r1, #64
; CHECK-NEXT:    vld1.64 {d16, d17}, [r1:128]!
; CHECK-NEXT:    add r8, r0, #64
; CHECK-NEXT:    mov r12, #0
; CHECK-NEXT:    vld1.64 {d26, d27}, [r0:128]!
; CHECK-NEXT:    vmov r4, r5, d16
; CHECK-NEXT:    vmov r6, r7, d26
; CHECK-NEXT:    vld1.64 {d18, d19}, [r8:128]!
; CHECK-NEXT:    vld1.64 {d20, d21}, [lr:128]!
; CHECK-NEXT:    vld1.64 {d22, d23}, [lr:128]!
; CHECK-NEXT:    vld1.64 {d24, d25}, [r8:128]!
; CHECK-NEXT:    vld1.64 {d30, d31}, [r0:128]!
; CHECK-NEXT:    vld1.64 {d4, d5}, [r1:128]!
; CHECK-NEXT:    vld1.64 {d0, d1}, [r0:128]!
; CHECK-NEXT:    vld1.64 {d2, d3}, [r1:128]!
; CHECK-NEXT:    subs r4, r6, r4
; CHECK-NEXT:    sbcs r4, r7, r5
; CHECK-NEXT:    vmov r5, r6, d17
; CHECK-NEXT:    vmov r7, r2, d27
; CHECK-NEXT:    mov r4, #0
; CHECK-NEXT:    movlt r4, #1
; CHECK-NEXT:    cmp r4, #0
; CHECK-NEXT:    mvnne r4, #0
; CHECK-NEXT:    vdup.32 d28, r4
; CHECK-NEXT:    subs r5, r7, r5
; CHECK-NEXT:    sbcs r2, r2, r6
; CHECK-NEXT:    vmov r5, r6, d24
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d29, r2
; CHECK-NEXT:    vmov r2, r4, d22
; CHECK-NEXT:    vbit q8, q13, q14
; CHECK-NEXT:    subs r2, r5, r2
; CHECK-NEXT:    sbcs r2, r6, r4
; CHECK-NEXT:    vmov r4, r5, d23
; CHECK-NEXT:    vmov r6, r7, d25
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d28, r2
; CHECK-NEXT:    subs r4, r6, r4
; CHECK-NEXT:    sbcs r4, r7, r5
; CHECK-NEXT:    vmov r2, r5, d4
; CHECK-NEXT:    vmov r6, r7, d30
; CHECK-NEXT:    mov r4, #0
; CHECK-NEXT:    movlt r4, #1
; CHECK-NEXT:    cmp r4, #0
; CHECK-NEXT:    mvnne r4, #0
; CHECK-NEXT:    vdup.32 d29, r4
; CHECK-NEXT:    mov r4, #0
; CHECK-NEXT:    vbit q11, q12, q14
; CHECK-NEXT:    vld1.64 {d24, d25}, [r1:128]
; CHECK-NEXT:    vld1.64 {d28, d29}, [r0:128]
; CHECK-NEXT:    subs r2, r6, r2
; CHECK-NEXT:    sbcs r2, r7, r5
; CHECK-NEXT:    vmov r6, r7, d31
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d6, r2
; CHECK-NEXT:    vmov r2, r5, d5
; CHECK-NEXT:    subs r2, r6, r2
; CHECK-NEXT:    sbcs r2, r7, r5
; CHECK-NEXT:    vmov r6, r7, d18
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d7, r2
; CHECK-NEXT:    vmov r2, r5, d20
; CHECK-NEXT:    subs r2, r6, r2
; CHECK-NEXT:    sbcs r2, r7, r5
; CHECK-NEXT:    vmov r6, r7, d19
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d26, r2
; CHECK-NEXT:    vmov r2, r5, d21
; CHECK-NEXT:    subs r2, r6, r2
; CHECK-NEXT:    sbcs r2, r7, r5
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d27, r2
; CHECK-NEXT:    vbif q9, q10, q13
; CHECK-NEXT:    vld1.64 {d26, d27}, [r8:128]!
; CHECK-NEXT:    vorr q10, q3, q3
; CHECK-NEXT:    vmov r6, r7, d27
; CHECK-NEXT:    vbsl q10, q15, q2
; CHECK-NEXT:    vld1.64 {d30, d31}, [lr:128]!
; CHECK-NEXT:    vmov r2, r5, d31
; CHECK-NEXT:    vld1.64 {d4, d5}, [r8:128]
; CHECK-NEXT:    vld1.64 {d6, d7}, [lr:128]
; CHECK-NEXT:    subs r2, r6, r2
; CHECK-NEXT:    sbcs r2, r7, r5
; CHECK-NEXT:    vmov r6, r7, d26
; CHECK-NEXT:    vmov r2, r5, d30
; CHECK-NEXT:    movlt r4, #1
; CHECK-NEXT:    cmp r4, #0
; CHECK-NEXT:    mvnne r4, #0
; CHECK-NEXT:    subs r2, r6, r2
; CHECK-NEXT:    sbcs r0, r7, r5
; CHECK-NEXT:    vmov r1, r2, d24
; CHECK-NEXT:    vmov r5, r6, d28
; CHECK-NEXT:    mov r0, #0
; CHECK-NEXT:    movlt r0, #1
; CHECK-NEXT:    cmp r0, #0
; CHECK-NEXT:    mvnne r0, #0
; CHECK-NEXT:    subs r1, r5, r1
; CHECK-NEXT:    sbcs r1, r6, r2
; CHECK-NEXT:    vmov r2, r5, d2
; CHECK-NEXT:    vmov r6, r7, d0
; CHECK-NEXT:    mov r1, #0
; CHECK-NEXT:    movlt r1, #1
; CHECK-NEXT:    cmp r1, #0
; CHECK-NEXT:    mvnne r1, #0
; CHECK-NEXT:    subs r2, r6, r2
; CHECK-NEXT:    sbcs r2, r7, r5
; CHECK-NEXT:    vmov r6, r7, d1
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d8, r2
; CHECK-NEXT:    vmov r2, r5, d3
; CHECK-NEXT:    subs r2, r6, r2
; CHECK-NEXT:    sbcs r2, r7, r5
; CHECK-NEXT:    mov r2, #0
; CHECK-NEXT:    movlt r2, #1
; CHECK-NEXT:    cmp r2, #0
; CHECK-NEXT:    mvnne r2, #0
; CHECK-NEXT:    vdup.32 d9, r2
; CHECK-NEXT:    vmov r2, r7, d25
; CHECK-NEXT:    vbif q0, q1, q4
; CHECK-NEXT:    vdup.32 d2, r1
; CHECK-NEXT:    vmov r1, r6, d29
; CHECK-NEXT:    vdup.32 d8, r0
; CHECK-NEXT:    vdup.32 d9, r4
; CHECK-NEXT:    vmov r5, r4, d7
; CHECK-NEXT:    mov r0, r3
; CHECK-NEXT:    vst1.64 {d16, d17}, [r0:128]!
; CHECK-NEXT:    vorr q8, q4, q4
; CHECK-NEXT:    vst1.64 {d20, d21}, [r0:128]!
; CHECK-NEXT:    vbsl q8, q13, q15
; CHECK-NEXT:    vst1.64 {d0, d1}, [r0:128]!
; CHECK-NEXT:    subs r1, r1, r2
; CHECK-NEXT:    sbcs r1, r6, r7
; CHECK-NEXT:    vmov r7, r6, d4
; CHECK-NEXT:    mov r1, #0
; CHECK-NEXT:    movlt r1, #1
; CHECK-NEXT:    cmp r1, #0
; CHECK-NEXT:    mvnne r1, #0
; CHECK-NEXT:    vdup.32 d3, r1
; CHECK-NEXT:    vmov r1, r2, d6
; CHECK-NEXT:    vbit q12, q14, q1
; CHECK-NEXT:    vst1.64 {d24, d25}, [r0:128]
; CHECK-NEXT:    add r0, r3, #64
; CHECK-NEXT:    vst1.64 {d18, d19}, [r0:128]!
; CHECK-NEXT:    vst1.64 {d22, d23}, [r0:128]!
; CHECK-NEXT:    vst1.64 {d16, d17}, [r0:128]!
; CHECK-NEXT:    subs r1, r7, r1
; CHECK-NEXT:    sbcs r1, r6, r2
; CHECK-NEXT:    vmov r2, r7, d5
; CHECK-NEXT:    mov r1, #0
; CHECK-NEXT:    movlt r1, #1
; CHECK-NEXT:    cmp r1, #0
; CHECK-NEXT:    mvnne r1, #0
; CHECK-NEXT:    vdup.32 d26, r1
; CHECK-NEXT:    subs r2, r2, r5
; CHECK-NEXT:    sbcs r2, r7, r4
; CHECK-NEXT:    movlt r12, #1
; CHECK-NEXT:    cmp r12, #0
; CHECK-NEXT:    mvnne r12, #0
; CHECK-NEXT:    vdup.32 d27, r12
; CHECK-NEXT:    vorr q10, q13, q13
; CHECK-NEXT:    vbsl q10, q2, q3
; CHECK-NEXT:    vst1.64 {d20, d21}, [r0:128]
; CHECK-NEXT:    vpop {d8, d9}
; CHECK-NEXT:    pop {r4, r5, r6, r7, r8, lr}
; CHECK-NEXT:    mov pc, lr
; COST: func_blend20
; COST: cost of 0 {{.*}} icmp
; COST: cost of 108 {{.*}} select
  %v0 = load %T0_20, ptr %loadaddr
  %v1 = load %T0_20, ptr %loadaddr2
  %c = icmp slt %T0_20 %v0, %v1
  %r = select %T1_20 %c, %T0_20 %v0, %T0_20 %v1
  store %T0_20 %r, ptr %storeaddr
  ret void
}