# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -O0 -march=aarch64 -run-pass=legalizer %s -o - | FileCheck %s

--- |
  target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
  target triple = "aarch64"

  define void @test_load() { ret void }
  define void @test_store() { ret void }

  define void @store_4xi16(<4 x i16> %v, <4 x i16>* %ptr) {
    store <4 x i16> %v, <4 x i16>* %ptr
    ret void
  }

  define void @store_4xi32(<4 x i32> %v, <4 x i32>* %ptr) {
    store <4 x i32> %v, <4 x i32>* %ptr
    ret void
  }

  define void @store_8xi16(<8 x i16> %v, <8 x i16>* %ptr) {
    store <8 x i16> %v, <8 x i16>* %ptr
    ret void
  }

  define void @store_16xi8(<16 x i8> %v, <16 x i8>* %ptr) {
    store <16 x i8> %v, <16 x i8>* %ptr
    ret void
  }

  define <4 x i16> @load_4xi16(<4 x i16>* %ptr) {
    %res = load <4 x i16>, <4 x i16>* %ptr
    ret <4 x i16> %res
  }

  define <4 x i32> @load_4xi32(<4 x i32>* %ptr) {
    %res = load <4 x i32>, <4 x i32>* %ptr
    ret <4 x i32> %res
  }

  define <8 x i16> @load_8xi16(<8 x i16>* %ptr) {
    %res = load <8 x i16>, <8 x i16>* %ptr
    ret <8 x i16> %res
  }

  define <16 x i8> @load_16xi8(<16 x i8>* %ptr) {
    %res = load <16 x i8>, <16 x i8>* %ptr
    ret <16 x i8> %res
  }

...
---
name:            test_load
body: |
  bb.0:
    liveins: $x0

    ; CHECK-LABEL: name: test_load
    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: [[LOAD:%[0-9]+]]:_(s8) = G_LOAD [[COPY]](p0) :: (load 1)
    ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s8)
    ; CHECK: $w0 = COPY [[ANYEXT]](s32)
    ; CHECK: [[LOAD1:%[0-9]+]]:_(s8) = G_LOAD [[COPY]](p0) :: (load 1)
    ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD1]](s8)
    ; CHECK: $w0 = COPY [[ANYEXT1]](s32)
    ; CHECK: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load 2)
    ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD2]](s16)
    ; CHECK: $w0 = COPY [[ANYEXT2]](s32)
    ; CHECK: $w0 = COPY [[ANYEXT1]](s32)
    ; CHECK: [[LOAD3:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load 8)
    ; CHECK: $x0 = COPY [[LOAD3]](s64)
    ; CHECK: [[LOAD4:%[0-9]+]]:_(p0) = G_LOAD [[COPY]](p0) :: (load 8)
    ; CHECK: [[PTRTOINT:%[0-9]+]]:_(s64) = G_PTRTOINT [[LOAD4]](p0)
    ; CHECK: $x0 = COPY [[PTRTOINT]](s64)
    ; CHECK: [[LOAD5:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load 8)
    ; CHECK: [[BITCAST:%[0-9]+]]:_(s64) = G_BITCAST [[LOAD5]](<2 x s32>)
    ; CHECK: $x0 = COPY [[BITCAST]](s64)
    ; CHECK: [[LOAD6:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load 8, align 16)
    ; CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CHECK: [[GEP:%[0-9]+]]:_(p0) = G_GEP [[COPY]], [[C]](s64)
    ; CHECK: [[LOAD7:%[0-9]+]]:_(s64) = G_LOAD [[GEP]](p0) :: (load 8)
    ; CHECK: [[MV:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[LOAD6]](s64), [[LOAD7]](s64)
    ; CHECK: [[TRUNC:%[0-9]+]]:_(s64) = G_TRUNC [[MV]](s128)
    ; CHECK: $x0 = COPY [[TRUNC]](s64)
    %0:_(p0) = COPY $x0
    %1:_(s1) = G_LOAD %0(p0) :: (load 1)
    %2:_(s32) = G_ANYEXT %1(s1)
    $w0 = COPY %2(s32)
    %3:_(s8) = G_LOAD %0(p0) :: (load 1)
    %4:_(s32) = G_ANYEXT %3(s8)
    $w0 = COPY %4(s32)
    %5:_(s16) = G_LOAD %0(p0) :: (load 2)
    %6:_(s32) = G_ANYEXT %5(s16)
    $w0 = COPY %6(s32)
    %7:_(s32) = G_LOAD %0(p0) :: (load 4)
    $w0 = COPY %4(s32)
    %8:_(s64) = G_LOAD %0(p0) :: (load 8)
    $x0 = COPY %8(s64)
    %9:_(p0) = G_LOAD %0(p0) :: (load 8)
    %10:_(s64) = G_PTRTOINT %9(p0)
    $x0 = COPY %10(s64)
    %11:_(<2 x s32>) = G_LOAD %0(p0) :: (load 8)
    %12:_(s64) = G_BITCAST %11(<2 x s32>)
    $x0 = COPY %12(s64)
    %13:_(s128) = G_LOAD %0(p0) :: (load 16)
    %14:_(s64) = G_TRUNC %13(s128)
    $x0 = COPY %14(s64)
...

---
name:            test_store
body: |
  bb.0:
    liveins: $x0, $w1

    ; CHECK-LABEL: name: test_store
    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $w1
    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
    ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
    ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C]](s32)
    ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[COPY3]]
    ; CHECK: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[AND]](s32)
    ; CHECK: G_STORE [[TRUNC]](s8), [[COPY]](p0) :: (store 1)
    ; CHECK: [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[COPY1]](s32)
    ; CHECK: G_STORE [[TRUNC1]](s8), [[COPY]](p0) :: (store 1)
    ; CHECK: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
    ; CHECK: G_STORE [[TRUNC2]](s16), [[COPY]](p0) :: (store 2)
    ; CHECK: G_STORE [[COPY1]](s32), [[COPY]](p0) :: (store 4)
    ; CHECK: [[PTRTOINT:%[0-9]+]]:_(s64) = G_PTRTOINT [[COPY]](p0)
    ; CHECK: G_STORE [[PTRTOINT]](s64), [[COPY]](p0) :: (store 8)
    ; CHECK: G_STORE [[COPY]](p0), [[COPY]](p0) :: (store 8)
    ; CHECK: [[PTRTOINT1:%[0-9]+]]:_(s64) = G_PTRTOINT [[COPY]](p0)
    ; CHECK: G_STORE [[PTRTOINT1]](s64), [[COPY]](p0) :: (store 8, align 16)
    ; CHECK: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
    ; CHECK: [[GEP:%[0-9]+]]:_(p0) = G_GEP [[COPY]], [[C1]](s64)
    ; CHECK: G_STORE [[PTRTOINT1]](s64), [[GEP]](p0) :: (store 8)
    %0:_(p0) = COPY $x0
    %1:_(s32) = COPY $w1
    %2:_(s1) = G_TRUNC %1(s32)
    G_STORE %2(s1), %0(p0) :: (store 1)
    %3:_(s8) = G_TRUNC %1(s32)
    G_STORE %3(s8), %0(p0) :: (store 1)
    %4:_(s16) = G_TRUNC %1(s32)
    G_STORE %4(s16), %0(p0) :: (store 2)
    G_STORE %1(s32), %0(p0) :: (store 4)
    %5:_(s64) = G_PTRTOINT %0(p0)
    G_STORE %5(s64), %0(p0) :: (store 8)
    G_STORE %0(p0), %0(p0) :: (store 8)
    %6:_(s64) = G_PTRTOINT %0(p0)
    %7:_(s128) = G_MERGE_VALUES %6(s64), %6
    G_STORE %7(s128), %0(p0) :: (store 16)
...
---
name:            store_4xi16
alignment:       2
tracksRegLiveness: true
machineFunctionInfo: {}
body:             |
  bb.1 (%ir-block.0):
    liveins: $d0, $x0

    ; CHECK-LABEL: name: store_4xi16
    ; CHECK: liveins: $d0, $x0
    ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $d0
    ; CHECK: [[COPY1:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: G_STORE [[COPY]](<4 x s16>), [[COPY1]](p0) :: (store 8 into %ir.ptr)
    ; CHECK: RET_ReallyLR
    %0:_(<4 x s16>) = COPY $d0
    %1:_(p0) = COPY $x0
    G_STORE %0(<4 x s16>), %1(p0) :: (store 8 into %ir.ptr)
    RET_ReallyLR

...
---
name:            store_4xi32
alignment:       2
tracksRegLiveness: true
machineFunctionInfo: {}
body:             |
  bb.1 (%ir-block.0):
    liveins: $q0, $x0

    ; CHECK-LABEL: name: store_4xi32
    ; CHECK: liveins: $q0, $x0
    ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
    ; CHECK: [[COPY1:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: G_STORE [[COPY]](<4 x s32>), [[COPY1]](p0) :: (store 16 into %ir.ptr)
    ; CHECK: RET_ReallyLR
    %0:_(<4 x s32>) = COPY $q0
    %1:_(p0) = COPY $x0
    G_STORE %0(<4 x s32>), %1(p0) :: (store 16 into %ir.ptr)
    RET_ReallyLR

...
---
name:            store_8xi16
alignment:       2
tracksRegLiveness: true
machineFunctionInfo: {}
body:             |
  bb.1 (%ir-block.0):
    liveins: $q0, $x0

    ; CHECK-LABEL: name: store_8xi16
    ; CHECK: liveins: $q0, $x0
    ; CHECK: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $q0
    ; CHECK: [[COPY1:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: G_STORE [[COPY]](<8 x s16>), [[COPY1]](p0) :: (store 16 into %ir.ptr)
    ; CHECK: RET_ReallyLR
    %0:_(<8 x s16>) = COPY $q0
    %1:_(p0) = COPY $x0
    G_STORE %0(<8 x s16>), %1(p0) :: (store 16 into %ir.ptr)
    RET_ReallyLR

...
---
name:            store_16xi8
alignment:       2
tracksRegLiveness: true
machineFunctionInfo: {}
body:             |
  bb.1 (%ir-block.0):
    liveins: $q0, $x0

    ; CHECK-LABEL: name: store_16xi8
    ; CHECK: liveins: $q0, $x0
    ; CHECK: [[COPY:%[0-9]+]]:_(<16 x s8>) = COPY $q0
    ; CHECK: [[COPY1:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: G_STORE [[COPY]](<16 x s8>), [[COPY1]](p0) :: (store 16 into %ir.ptr)
    ; CHECK: RET_ReallyLR
    %0:_(<16 x s8>) = COPY $q0
    %1:_(p0) = COPY $x0
    G_STORE %0(<16 x s8>), %1(p0) :: (store 16 into %ir.ptr)
    RET_ReallyLR

...
---
name:            load_4xi16
alignment:       2
tracksRegLiveness: true
machineFunctionInfo: {}
body:             |
  bb.1 (%ir-block.0):
    liveins: $x0

    ; CHECK-LABEL: name: load_4xi16
    ; CHECK: liveins: $x0
    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load 8 from %ir.ptr)
    ; CHECK: $d0 = COPY [[LOAD]](<4 x s16>)
    ; CHECK: RET_ReallyLR implicit $d0
    %0:_(p0) = COPY $x0
    %1:_(<4 x s16>) = G_LOAD %0(p0) :: (load 8 from %ir.ptr)
    $d0 = COPY %1(<4 x s16>)
    RET_ReallyLR implicit $d0

...
---
name:            load_4xi32
alignment:       2
tracksRegLiveness: true
machineFunctionInfo: {}
body:             |
  bb.1 (%ir-block.0):
    liveins: $x0

    ; CHECK-LABEL: name: load_4xi32
    ; CHECK: liveins: $x0
    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.ptr)
    ; CHECK: $q0 = COPY [[LOAD]](<4 x s32>)
    ; CHECK: RET_ReallyLR implicit $q0
    %0:_(p0) = COPY $x0
    %1:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.ptr)
    $q0 = COPY %1(<4 x s32>)
    RET_ReallyLR implicit $q0

...
---
name:            load_8xi16
alignment:       2
tracksRegLiveness: true
machineFunctionInfo: {}
body:             |
  bb.1 (%ir-block.0):
    liveins: $x0

    ; CHECK-LABEL: name: load_8xi16
    ; CHECK: liveins: $x0
    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.ptr)
    ; CHECK: $q0 = COPY [[LOAD]](<8 x s16>)
    ; CHECK: RET_ReallyLR implicit $q0
    %0:_(p0) = COPY $x0
    %1:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.ptr)
    $q0 = COPY %1(<8 x s16>)
    RET_ReallyLR implicit $q0

...
---
name:            load_16xi8
alignment:       2
tracksRegLiveness: true
machineFunctionInfo: {}
body:             |
  bb.1 (%ir-block.0):
    liveins: $x0

    ; CHECK-LABEL: name: load_16xi8
    ; CHECK: liveins: $x0
    ; CHECK: [[COPY:%[0-9]+]]:_(p0) = COPY $x0
    ; CHECK: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.ptr)
    ; CHECK: $q0 = COPY [[LOAD]](<16 x s8>)
    ; CHECK: RET_ReallyLR implicit $q0
    %0:_(p0) = COPY $x0
    %1:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.ptr)
    $q0 = COPY %1(<16 x s8>)
    RET_ReallyLR implicit $q0

...