# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -march=amdgcn -mcpu=hawaii -mattr=+flat-for-global -run-pass=regbankselect %s -verify-machineinstrs -o - | FileCheck %s

--- |
  define amdgpu_kernel void @load_constant(i32 addrspace(4)* %ptr0) {
    ret void
  }

  define amdgpu_kernel void @load_constant_volatile(i32 addrspace(4)* %ptr0) {
    ret void
  }

  define amdgpu_kernel void @load_global_uniform_invariant(i32 addrspace(1)* %ptr1) {
    %tmp0 = load i32, i32 addrspace(1)* %ptr1
    ret void
  }

  define amdgpu_kernel void @load_global_uniform_noclobber(i32 addrspace(1)* %ptr1) {
    %tmp0 = load i32, i32 addrspace(1)* %ptr1, !amdgpu.noclobber !0
    ret void
  }

  define amdgpu_kernel void @load_global_uniform_variant(i32 addrspace(1)* %ptr1) {
    %tmp0 = load i32, i32 addrspace(1)* %ptr1
    ret void
  }

  define amdgpu_kernel void @load_global_uniform_volatile_invariant(i32 addrspace(1)* %ptr1) {
    %tmp0 = load i32, i32 addrspace(1)* %ptr1
    ret void
  }

  define amdgpu_kernel void @load_global_uniform_atomic_invariant(i32 addrspace(1)* %ptr1) {
    %tmp0 = load i32, i32 addrspace(1)* %ptr1
    ret void
  }

  define amdgpu_kernel void @load_global_non_uniform(i32 addrspace(1)* %ptr2) {
    %tmp0 = call i32 @llvm.amdgcn.workitem.id.x() #0
    %tmp1 = getelementptr i32, i32 addrspace(1)* %ptr2, i32 %tmp0
    %tmp2 = load i32, i32 addrspace(1)* %tmp1
    ret void
  }

  define void @non_power_of_2() { ret void }

  define amdgpu_kernel void @load_constant_v4i16_from_8_align8(<3 x i16> addrspace(4)* %ptr0) {
    ret void
  }

  declare i32 @llvm.amdgcn.workitem.id.x() #0
  attributes #0 = { nounwind readnone }
  !0 = !{}

...
---
name: load_constant
legalized: true

body: |
  bb.0:
    liveins: $sgpr0_sgpr1
    ; CHECK-LABEL: name: load_constant
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
    ; CHECK: [[LOAD:%[0-9]+]]:sgpr(s32) = G_LOAD [[COPY]](p4) :: (load 4 from %ir.ptr0, addrspace 4)
    %0:_(p4) = COPY $sgpr0_sgpr1
    %1:_(s32) = G_LOAD %0 :: (load 4 from %ir.ptr0)
...

---
name: load_constant_volatile
legalized: true

body: |
  bb.0:
    liveins: $sgpr0_sgpr1
    ; CHECK-LABEL: name: load_constant_volatile
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
    ; CHECK: [[LOAD:%[0-9]+]]:sgpr(s32) = G_LOAD [[COPY]](p4) :: (volatile load 4 from %ir.ptr0, addrspace 4)
    %0:_(p4) = COPY $sgpr0_sgpr1
    %1:_(s32) = G_LOAD %0 :: (volatile load 4 from %ir.ptr0)
...

---
name: load_global_uniform_invariant
legalized: true

body: |
  bb.0:
    liveins: $sgpr0_sgpr1
    ; CHECK-LABEL: name: load_global_uniform_invariant
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
    ; CHECK: [[LOAD:%[0-9]+]]:sgpr(s32) = G_LOAD [[COPY]](p1) :: (invariant load 4 from %ir.ptr1, addrspace 1)
    %0:_(p1) = COPY $sgpr0_sgpr1
    %1:_(s32) = G_LOAD %0 :: (invariant load 4 from %ir.ptr1)
...

---
name: load_global_uniform_noclobber
legalized: true

body: |
  bb.0:
    liveins: $sgpr0_sgpr1
    ; CHECK-LABEL: name: load_global_uniform_noclobber
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
    ; CHECK: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
    ; CHECK: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (load 4 from %ir.ptr1, addrspace 1)
    %0:_(p1) = COPY $sgpr0_sgpr1
    %1:_(s32) = G_LOAD %0 :: (load 4 from %ir.ptr1)
...

---
name: load_global_uniform_variant
legalized: true

body: |
  bb.0:
    liveins: $sgpr0_sgpr1
    ; CHECK-LABEL: name: load_global_uniform_variant
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
    ; CHECK: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
    ; CHECK: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (load 4 from %ir.ptr1, addrspace 1)
    %0:_(p1) = COPY $sgpr0_sgpr1
    %1:_(s32) = G_LOAD %0 :: (load 4 from %ir.ptr1)
...

---
name: load_global_uniform_volatile_invariant
legalized: true

body: |
  bb.0:
    liveins: $sgpr0_sgpr1
    ; CHECK-LABEL: name: load_global_uniform_volatile_invariant
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
    ; CHECK: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
    ; CHECK: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (volatile invariant load 4 from %ir.ptr1, addrspace 1)
    %0:_(p1) = COPY $sgpr0_sgpr1
    %1:_(s32) = G_LOAD %0 :: (volatile invariant load 4 from %ir.ptr1)
...

---
name: load_global_uniform_atomic_invariant
legalized: true

body: |
  bb.0:
    liveins: $sgpr0_sgpr1
    ; CHECK-LABEL: name: load_global_uniform_atomic_invariant
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
    ; CHECK: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
    ; CHECK: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (invariant load acquire 4 from %ir.ptr1, addrspace 1)
    %0:_(p1) = COPY $sgpr0_sgpr1
    %1:_(s32) = G_LOAD %0 :: (invariant load acquire 4 from %ir.ptr1)
...

---
name: load_global_non_uniform
legalized: true

body: |
  bb.0:
    liveins: $sgpr0_sgpr1
    ; CHECK-LABEL: name: load_global_non_uniform
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p1) = COPY $sgpr0_sgpr1
    ; CHECK: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY [[COPY]](p1)
    ; CHECK: [[LOAD:%[0-9]+]]:vgpr(s32) = G_LOAD [[COPY1]](p1) :: (load 4 from %ir.tmp1, addrspace 1)
    %0:_(p1) = COPY $sgpr0_sgpr1
    %1:_(s32) = G_LOAD %0 :: (load 4 from %ir.tmp1)
...

---
name: non_power_of_2
legalized: true

body: |
  bb.0:
    ; CHECK-LABEL: name: non_power_of_2
    ; CHECK: [[DEF:%[0-9]+]]:sgpr(s448) = G_IMPLICIT_DEF
    ; CHECK: [[EXTRACT:%[0-9]+]]:sgpr(s32) = G_EXTRACT [[DEF]](s448), 0
    ; CHECK: $sgpr0 = COPY [[EXTRACT]](s32)
    ; CHECK: SI_RETURN_TO_EPILOG $sgpr0
    %0:_(s448) = G_IMPLICIT_DEF
    %1:_(s32) = G_EXTRACT %0:_(s448), 0
    $sgpr0 = COPY %1:_(s32)
    SI_RETURN_TO_EPILOG $sgpr0
...

---
name: load_constant_v4i16_from_8_align8
legalized: true

body: |
  bb.0:
    ; CHECK-LABEL: name: load_constant_v4i16_from_8_align8
    ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
    ; CHECK: [[LOAD:%[0-9]+]]:sgpr(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load 8 from %ir.ptr0, addrspace 4)
    %0:_(p4) = COPY $sgpr0_sgpr1
    %1:_(<4 x s16>) = G_LOAD %0 :: (load 8 from %ir.ptr0, align 8, addrspace 4)

...