; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx1250 -mattr=+load-store-opt,+unaligned-access-mode < %s | FileCheck -enable-var-scope -check-prefixes=GFX1250,GFX1250-UNALIGNED %s

; FIXME: We don't get cases where the address was an SGPR because we
; get a copy to the address register for each one.

@lds = addrspace(3) global [512 x float] poison, align 4
@lds.f64 = addrspace(3) global [512 x double] poison, align 8

define amdgpu_kernel void @simple_read2_f32(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f32:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v2, 0xffc, v0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:8
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 4
  %add.x = add nsw i32 %x.i, 8
  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %add.x
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 4
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @simple_read2_f32_max_offset(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f32_max_offset:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v2, 0xffc, v0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:255
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 4
  %add.x = add nsw i32 %x.i, 255
  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %add.x
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 4
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @simple_read2_f32_too_far(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f32_too_far:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v0, 0xffc, v0
; GFX1250-NEXT:    ds_load_b32 v1, v0
; GFX1250-NEXT:    ds_load_b32 v2, v0 offset:1028
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v1, v1, v2
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 4
  %add.x = add nsw i32 %x.i, 257
  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %add.x
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 4
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @simple_read2_f32_x2(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f32_x2:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v4, 0xffc, v0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v4 offset1:8
; GFX1250-NEXT:    ds_load_2addr_b32 v[2:3], v4 offset0:11 offset1:27
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_dual_add_f32 v0, v0, v1 :: v_dual_add_f32 v1, v2, v3
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v4, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %idx.0 = add nsw i32 %tid.x, 0
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.0
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 4

  %idx.1 = add nsw i32 %tid.x, 8
  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.1
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 4
  %sum.0 = fadd float %val0, %val1

  %idx.2 = add nsw i32 %tid.x, 11
  %arrayidx2 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.2
  %val2 = load float, ptr addrspace(3) %arrayidx2, align 4

  %idx.3 = add nsw i32 %tid.x, 27
  %arrayidx3 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.3
  %val3 = load float, ptr addrspace(3) %arrayidx3, align 4
  %sum.1 = fadd float %val2, %val3

  %sum = fadd float %sum.0, %sum.1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %idx.0
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

; Make sure there is an instruction between the two sets of reads.
define amdgpu_kernel void @simple_read2_f32_x2_barrier(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f32_x2_barrier:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v4, 0xffc, v0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v4 offset1:8
; GFX1250-NEXT:    s_barrier_signal -1
; GFX1250-NEXT:    s_barrier_wait -1
; GFX1250-NEXT:    ds_load_2addr_b32 v[2:3], v4 offset0:11 offset1:27
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_dual_add_f32 v0, v0, v1 :: v_dual_add_f32 v1, v2, v3
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v4, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %idx.0 = add nsw i32 %tid.x, 0
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.0
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 4

  %idx.1 = add nsw i32 %tid.x, 8
  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.1
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 4
  %sum.0 = fadd float %val0, %val1

  call void @llvm.amdgcn.s.barrier() #2

  %idx.2 = add nsw i32 %tid.x, 11
  %arrayidx2 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.2
  %val2 = load float, ptr addrspace(3) %arrayidx2, align 4

  %idx.3 = add nsw i32 %tid.x, 27
  %arrayidx3 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.3
  %val3 = load float, ptr addrspace(3) %arrayidx3, align 4
  %sum.1 = fadd float %val2, %val3

  %sum = fadd float %sum.0, %sum.1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %idx.0
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

; For some reason adding something to the base address for the first
; element results in only folding the inner pair.
define amdgpu_kernel void @simple_read2_f32_x2_nonzero_base(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f32_x2_nonzero_base:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v4, 0xffc, v0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v4 offset0:2 offset1:8
; GFX1250-NEXT:    ds_load_2addr_b32 v[2:3], v4 offset0:11 offset1:27
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_dual_add_f32 v0, v0, v1 :: v_dual_add_f32 v1, v2, v3
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v4, v0, s[0:1] offset:8
; GFX1250-NEXT:    s_endpgm
  %tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %idx.0 = add nsw i32 %tid.x, 2
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.0
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 4

  %idx.1 = add nsw i32 %tid.x, 8
  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.1
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 4
  %sum.0 = fadd float %val0, %val1

  %idx.2 = add nsw i32 %tid.x, 11
  %arrayidx2 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.2
  %val2 = load float, ptr addrspace(3) %arrayidx2, align 4

  %idx.3 = add nsw i32 %tid.x, 27
  %arrayidx3 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %idx.3
  %val3 = load float, ptr addrspace(3) %arrayidx3, align 4
  %sum.1 = fadd float %val2, %val3

  %sum = fadd float %sum.0, %sum.1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %idx.0
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

; Be careful of vectors of pointers. We don't know if the 2 pointers
; in the vectors are really the same base, so this is not safe to
; merge.
; Base pointers come from different subregister of same super
; register. We can't safely merge this.
define amdgpu_kernel void @read2_ptr_is_subreg_arg_f32(ptr addrspace(1) %out, <2 x ptr addrspace(3)> %lds.ptr) #0 {
; GFX1250-LABEL: read2_ptr_is_subreg_arg_f32:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b128 s[0:3], s[4:5], 0x0
; GFX1250-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s3
; GFX1250-NEXT:    ds_load_b32 v1, v1 offset:32
; GFX1250-NEXT:    ds_load_b32 v2, v2
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v1, v1, v2
; GFX1250-NEXT:    global_store_b32 v0, v1, s[0:1] scale_offset
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %index.0 = insertelement <2 x i32> poison, i32 %x.i, i32 0
  %index.1 = insertelement <2 x i32> %index.0, i32 8, i32 0
  %gep = getelementptr inbounds float, <2 x ptr addrspace(3)> %lds.ptr, <2 x i32> %index.1
  %gep.0 = extractelement <2 x ptr addrspace(3)> %gep, i32 0
  %gep.1 = extractelement <2 x ptr addrspace(3)> %gep, i32 1
  %val0 = load float, ptr addrspace(3) %gep.0, align 4
  %val1 = load float, ptr addrspace(3) %gep.1, align 4
  %add.x = add nsw i32 %x.i, 8
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

; Apply a constant scalar offset after the pointer vector extract.  We
; are rejecting merges that have the same, constant 0 offset, so make
; sure we are really rejecting it because of the different
; subregisters.
define amdgpu_kernel void @read2_ptr_is_subreg_arg_offset_f32(ptr addrspace(1) %out, <2 x ptr addrspace(3)> %lds.ptr) #0 {
; GFX1250-LABEL: read2_ptr_is_subreg_arg_offset_f32:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b128 s[0:3], s[4:5], 0x0
; GFX1250-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_dual_mov_b32 v1, s2 :: v_dual_mov_b32 v2, s3
; GFX1250-NEXT:    ds_load_b32 v1, v1 offset:32
; GFX1250-NEXT:    ds_load_b32 v2, v2 offset:32
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v1, v1, v2
; GFX1250-NEXT:    global_store_b32 v0, v1, s[0:1] scale_offset
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %index.0 = insertelement <2 x i32> poison, i32 %x.i, i32 0
  %index.1 = insertelement <2 x i32> %index.0, i32 8, i32 0
  %gep = getelementptr inbounds float, <2 x ptr addrspace(3)> %lds.ptr, <2 x i32> %index.1
  %gep.0 = extractelement <2 x ptr addrspace(3)> %gep, i32 0
  %gep.1 = extractelement <2 x ptr addrspace(3)> %gep, i32 1

  ; Apply an additional offset after the vector that will be more obviously folded.
  %gep.1.offset = getelementptr float, ptr addrspace(3) %gep.1, i32 8

  %val0 = load float, ptr addrspace(3) %gep.0, align 4
  %val1 = load float, ptr addrspace(3) %gep.1.offset, align 4
  %add.x = add nsw i32 %x.i, 8
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @read2_ptr_is_subreg_f32(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: read2_ptr_is_subreg_f32:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v2, 0xffc, v0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:8
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %ptr.0 = insertelement <2 x ptr addrspace(3)> poison, ptr addrspace(3) @lds, i32 0
  %ptr.1 = insertelement <2 x ptr addrspace(3)> %ptr.0, ptr addrspace(3) @lds, i32 1
  %x.i.v.0 = insertelement <2 x i32> poison, i32 %x.i, i32 0
  %x.i.v.1 = insertelement <2 x i32> %x.i.v.0, i32 %x.i, i32 1
  %idx = add <2 x i32> %x.i.v.1, <i32 0, i32 8>
  %gep = getelementptr inbounds [512 x float], <2 x ptr addrspace(3)> %ptr.1, <2 x i32> <i32 0, i32 0>, <2 x i32> %idx
  %gep.0 = extractelement <2 x ptr addrspace(3)> %gep, i32 0
  %gep.1 = extractelement <2 x ptr addrspace(3)> %gep, i32 1
  %val0 = load float, ptr addrspace(3) %gep.0, align 4
  %val1 = load float, ptr addrspace(3) %gep.1, align 4
  %add.x = add nsw i32 %x.i, 8
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @simple_read2_f32_volatile_0(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f32_volatile_0:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v0, 0xffc, v0
; GFX1250-NEXT:    ds_load_b32 v1, v0
; GFX1250-NEXT:    ds_load_b32 v2, v0 offset:32
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v1, v1, v2
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
  %val0 = load volatile float, ptr addrspace(3) %arrayidx0, align 4
  %add.x = add nsw i32 %x.i, 8
  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %add.x
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 4
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @simple_read2_f32_volatile_1(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f32_volatile_1:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v0, 0xffc, v0
; GFX1250-NEXT:    ds_load_b32 v1, v0
; GFX1250-NEXT:    ds_load_b32 v2, v0 offset:32
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v1, v1, v2
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v0, v1, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 4
  %add.x = add nsw i32 %x.i, 8
  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %add.x
  %val1 = load volatile float, ptr addrspace(3) %arrayidx1, align 4
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

; Can't fold since not correctly aligned.
define amdgpu_kernel void @unaligned_read2_f32(ptr addrspace(1) %out, ptr addrspace(3) %lds) #0 {
; GFX1250-LABEL: unaligned_read2_f32:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b96 s[0:2], s[4:5], 0x0
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v2, 0xffc, v0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_add_nc_u32_e32 v0, s2, v2
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v0 offset1:8
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds float, ptr addrspace(3) %lds, i32 %x.i
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 1
  %add.x = add nsw i32 %x.i, 8
  %arrayidx1 = getelementptr inbounds float, ptr addrspace(3) %lds, i32 %add.x
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 1
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @unaligned_offset_read2_f32(ptr addrspace(1) %out, ptr addrspace(3) %lds) #0 {
; GFX1250-LABEL: unaligned_offset_read2_f32:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b96 s[0:2], s[4:5], 0x0
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v2, 0xffc, v0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_add_nc_u32_e32 v0, s2, v2
; GFX1250-NEXT:    ds_load_b64 v[0:1], v0 offset:5
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %base = getelementptr inbounds float, ptr addrspace(3) %lds, i32 %x.i
  %addr0.i8 = getelementptr inbounds i8, ptr addrspace(3) %base, i32 5
  %val0 = load float, ptr addrspace(3) %addr0.i8, align 1
  %addr1.i8 = getelementptr inbounds i8, ptr addrspace(3) %base, i32 9
  %val1 = load float, ptr addrspace(3) %addr1.i8, align 1
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @misaligned_2_simple_read2_f32(ptr addrspace(1) %out, ptr addrspace(3) %lds) #0 {
; GFX1250-LABEL: misaligned_2_simple_read2_f32:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b96 s[0:2], s[4:5], 0x0
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v2, 0xffc, v0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_add_nc_u32_e32 v0, s2, v2
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v0 offset1:8
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds float, ptr addrspace(3) %lds, i32 %x.i
  %val0 = load float, ptr addrspace(3) %arrayidx0, align 2
  %add.x = add nsw i32 %x.i, 8
  %arrayidx1 = getelementptr inbounds float, ptr addrspace(3) %lds, i32 %add.x
  %val1 = load float, ptr addrspace(3) %arrayidx1, align 2
  %sum = fadd float %val0, %val1
  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

define amdgpu_kernel void @simple_read2_f64(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f64:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v4, 0x1ff8, v0
; GFX1250-NEXT:    ds_load_2addr_b64 v[0:3], v4 offset1:8
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f64_e32 v[0:1], v[0:1], v[2:3]
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b64 v4, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x double], ptr addrspace(3) @lds.f64, i32 0, i32 %x.i
  %val0 = load double, ptr addrspace(3) %arrayidx0, align 8
  %add.x = add nsw i32 %x.i, 8
  %arrayidx1 = getelementptr inbounds [512 x double], ptr addrspace(3) @lds.f64, i32 0, i32 %add.x
  %val1 = load double, ptr addrspace(3) %arrayidx1, align 8
  %sum = fadd double %val0, %val1
  %out.gep = getelementptr inbounds double, ptr addrspace(1) %out, i32 %x.i
  store double %sum, ptr addrspace(1) %out.gep, align 8
  ret void
}

define amdgpu_kernel void @simple_read2_f64_max_offset(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f64_max_offset:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v4, 0x1ff8, v0
; GFX1250-NEXT:    ds_load_2addr_b64 v[0:3], v4 offset1:255
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f64_e32 v[0:1], v[0:1], v[2:3]
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b64 v4, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x double], ptr addrspace(3) @lds.f64, i32 0, i32 %x.i
  %val0 = load double, ptr addrspace(3) %arrayidx0, align 8
  %add.x = add nsw i32 %x.i, 255
  %arrayidx1 = getelementptr inbounds [512 x double], ptr addrspace(3) @lds.f64, i32 0, i32 %add.x
  %val1 = load double, ptr addrspace(3) %arrayidx1, align 8
  %sum = fadd double %val0, %val1
  %out.gep = getelementptr inbounds double, ptr addrspace(1) %out, i32 %x.i
  store double %sum, ptr addrspace(1) %out.gep, align 8
  ret void
}

define amdgpu_kernel void @simple_read2_f64_too_far(ptr addrspace(1) %out) #0 {
; GFX1250-LABEL: simple_read2_f64_too_far:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v4, 0x1ff8, v0
; GFX1250-NEXT:    ds_load_b64 v[0:1], v4
; GFX1250-NEXT:    ds_load_b64 v[2:3], v4 offset:2056
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f64_e32 v[0:1], v[0:1], v[2:3]
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b64 v4, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x double], ptr addrspace(3) @lds.f64, i32 0, i32 %x.i
  %val0 = load double, ptr addrspace(3) %arrayidx0, align 8
  %add.x = add nsw i32 %x.i, 257
  %arrayidx1 = getelementptr inbounds [512 x double], ptr addrspace(3) @lds.f64, i32 0, i32 %add.x
  %val1 = load double, ptr addrspace(3) %arrayidx1, align 8
  %sum = fadd double %val0, %val1
  %out.gep = getelementptr inbounds double, ptr addrspace(1) %out, i32 %x.i
  store double %sum, ptr addrspace(1) %out.gep, align 8
  ret void
}

; Alignment only 4
define amdgpu_kernel void @misaligned_read2_f64(ptr addrspace(1) %out, ptr addrspace(3) %lds) #0 {
; GFX1250-LABEL: misaligned_read2_f64:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b96 s[0:2], s[4:5], 0x0
; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v4, 0x1ff8, v0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_add_nc_u32_e32 v2, s2, v4
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:1
; GFX1250-NEXT:    ds_load_2addr_b32 v[2:3], v2 offset0:14 offset1:15
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f64_e32 v[0:1], v[0:1], v[2:3]
; GFX1250-NEXT:    global_store_b64 v4, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds double, ptr addrspace(3) %lds, i32 %x.i
  %val0 = load double, ptr addrspace(3) %arrayidx0, align 4
  %add.x = add nsw i32 %x.i, 7
  %arrayidx1 = getelementptr inbounds double, ptr addrspace(3) %lds, i32 %add.x
  %val1 = load double, ptr addrspace(3) %arrayidx1, align 4
  %sum = fadd double %val0, %val1
  %out.gep = getelementptr inbounds double, ptr addrspace(1) %out, i32 %x.i
  store double %sum, ptr addrspace(1) %out.gep, align 4
  ret void
}

@foo = addrspace(3) global [4 x i32] poison, align 4

define amdgpu_kernel void @load_constant_adjacent_offsets(ptr addrspace(1) %out) {
; GFX1250-LABEL: load_constant_adjacent_offsets:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    ds_load_b64 v[0:1], v2
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_nc_u32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %val0 = load i32, ptr addrspace(3) @foo, align 4
  %val1 = load i32, ptr addrspace(3) getelementptr inbounds ([4 x i32], ptr addrspace(3) @foo, i32 0, i32 1), align 4
  %sum = add i32 %val0, %val1
  store i32 %sum, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @load_constant_disjoint_offsets(ptr addrspace(1) %out) {
; GFX1250-LABEL: load_constant_disjoint_offsets:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:2
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_nc_u32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %val0 = load i32, ptr addrspace(3) @foo, align 4
  %val1 = load i32, ptr addrspace(3) getelementptr inbounds ([4 x i32], ptr addrspace(3) @foo, i32 0, i32 2), align 4
  %sum = add i32 %val0, %val1
  store i32 %sum, ptr addrspace(1) %out, align 4
  ret void
}

@bar = addrspace(3) global [4 x i64] poison, align 4

define amdgpu_kernel void @load_misaligned64_constant_offsets(ptr addrspace(1) %out) {
; GFX1250-LABEL: load_misaligned64_constant_offsets:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_mov_b32_e32 v4, 0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    ds_load_b128 v[0:3], v4
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_nc_u64_e32 v[0:1], v[0:1], v[2:3]
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b64 v4, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
  %val0 = load i64, ptr addrspace(3) @bar, align 4
  %val1 = load i64, ptr addrspace(3) getelementptr inbounds ([4 x i64], ptr addrspace(3) @bar, i32 0, i32 1), align 4
  %sum = add i64 %val0, %val1
  store i64 %sum, ptr addrspace(1) %out, align 8
  ret void
}

@bar.large = addrspace(3) global [4096 x i64] poison, align 4

define amdgpu_kernel void @load_misaligned64_constant_large_offsets(ptr addrspace(1) %out) {
; GFX1250-LABEL: load_misaligned64_constant_large_offsets:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    v_mov_b32_e32 v4, 0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    ds_load_b64 v[0:1], v4 offset:16384
; GFX1250-NEXT:    ds_load_b64 v[2:3], v4 offset:32760
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_nc_u64_e32 v[0:1], v[0:1], v[2:3]
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b64 v4, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
  %val0 = load i64, ptr addrspace(3) getelementptr inbounds ([4096 x i64], ptr addrspace(3) @bar.large, i32 0, i32 2048), align 4
  %val1 = load i64, ptr addrspace(3) getelementptr inbounds ([4096 x i64], ptr addrspace(3) @bar.large, i32 0, i32 4095), align 4
  %sum = add i64 %val0, %val1
  store i64 %sum, ptr addrspace(1) %out, align 8
  ret void
}

@sgemm.lA = internal unnamed_addr addrspace(3) global [264 x float] poison, align 4
@sgemm.lB = internal unnamed_addr addrspace(3) global [776 x float] poison, align 4

define amdgpu_kernel void @sgemm_inner_loop_read2_sequence(ptr addrspace(1) %C, i32 %lda, i32 %ldb) #0 {
; GFX1250-LABEL: sgemm_inner_loop_read2_sequence:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_bfe_u32 s0, ttmp6, 0x4000c
; GFX1250-NEXT:    s_and_b32 s1, ttmp6, 15
; GFX1250-NEXT:    s_add_co_i32 s0, s0, 1
; GFX1250-NEXT:    s_getreg_b32 s2, hwreg(HW_REG_IB_STS2, 6, 4)
; GFX1250-NEXT:    s_mul_i32 s0, ttmp9, s0
; GFX1250-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
; GFX1250-NEXT:    s_add_co_i32 s1, s1, s0
; GFX1250-NEXT:    s_cmp_eq_u32 s2, 0
; GFX1250-NEXT:    s_cselect_b32 s0, ttmp9, s1
; GFX1250-NEXT:    s_lshl_b32 s0, s0, 2
; GFX1250-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX1250-NEXT:    s_add_co_i32 s1, s0, 0xc20
; GFX1250-NEXT:    s_addk_co_i32 s0, 0xc60
; GFX1250-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v4, s0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    ds_load_2addr_b32 v[2:3], v1 offset1:1
; GFX1250-NEXT:    ds_load_2addr_b32 v[4:5], v4 offset1:1
; GFX1250-NEXT:    s_wait_dscnt 0x1
; GFX1250-NEXT:    v_dual_lshrrev_b32 v0, 8, v0 :: v_dual_add_f32 v2, v2, v3
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_and_b32_e32 v8, 0xffc, v0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v8 offset1:1
; GFX1250-NEXT:    ds_load_2addr_b32 v[6:7], v8 offset0:32 offset1:33
; GFX1250-NEXT:    s_wait_dscnt 0x2
; GFX1250-NEXT:    v_add_f32_e32 v2, v2, v4
; GFX1250-NEXT:    v_add_f32_e32 v4, v2, v5
; GFX1250-NEXT:    ds_load_2addr_b32 v[2:3], v8 offset0:64 offset1:65
; GFX1250-NEXT:    s_wait_dscnt 0x2
; GFX1250-NEXT:    v_add_f32_e32 v0, v4, v0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_dual_add_f32 v0, v0, v1 :: v_dual_mov_b32 v1, 0
; GFX1250-NEXT:    s_wait_dscnt 0x1
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v6
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v7
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_add_f32_e32 v0, v0, v3
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v1, v0, s[0:1]
; GFX1250-NEXT:    s_endpgm
  %x.i = tail call i32 @llvm.amdgcn.workgroup.id.x() #1
  %y.i = tail call i32 @llvm.amdgcn.workitem.id.y() #1
  %arrayidx44 = getelementptr inbounds [264 x float], ptr addrspace(3) @sgemm.lA, i32 0, i32 %x.i
  %tmp16 = load float, ptr addrspace(3) %arrayidx44, align 4
  %add47 = add nsw i32 %x.i, 1
  %arrayidx48 = getelementptr inbounds [264 x float], ptr addrspace(3) @sgemm.lA, i32 0, i32 %add47
  %tmp17 = load float, ptr addrspace(3) %arrayidx48, align 4
  %add51 = add nsw i32 %x.i, 16
  %arrayidx52 = getelementptr inbounds [264 x float], ptr addrspace(3) @sgemm.lA, i32 0, i32 %add51
  %tmp18 = load float, ptr addrspace(3) %arrayidx52, align 4
  %add55 = add nsw i32 %x.i, 17
  %arrayidx56 = getelementptr inbounds [264 x float], ptr addrspace(3) @sgemm.lA, i32 0, i32 %add55
  %tmp19 = load float, ptr addrspace(3) %arrayidx56, align 4
  %arrayidx60 = getelementptr inbounds [776 x float], ptr addrspace(3) @sgemm.lB, i32 0, i32 %y.i
  %tmp20 = load float, ptr addrspace(3) %arrayidx60, align 4
  %add63 = add nsw i32 %y.i, 1
  %arrayidx64 = getelementptr inbounds [776 x float], ptr addrspace(3) @sgemm.lB, i32 0, i32 %add63
  %tmp21 = load float, ptr addrspace(3) %arrayidx64, align 4
  %add67 = add nsw i32 %y.i, 32
  %arrayidx68 = getelementptr inbounds [776 x float], ptr addrspace(3) @sgemm.lB, i32 0, i32 %add67
  %tmp22 = load float, ptr addrspace(3) %arrayidx68, align 4
  %add71 = add nsw i32 %y.i, 33
  %arrayidx72 = getelementptr inbounds [776 x float], ptr addrspace(3) @sgemm.lB, i32 0, i32 %add71
  %tmp23 = load float, ptr addrspace(3) %arrayidx72, align 4
  %add75 = add nsw i32 %y.i, 64
  %arrayidx76 = getelementptr inbounds [776 x float], ptr addrspace(3) @sgemm.lB, i32 0, i32 %add75
  %tmp24 = load float, ptr addrspace(3) %arrayidx76, align 4
  %add79 = add nsw i32 %y.i, 65
  %arrayidx80 = getelementptr inbounds [776 x float], ptr addrspace(3) @sgemm.lB, i32 0, i32 %add79
  %tmp25 = load float, ptr addrspace(3) %arrayidx80, align 4
  %sum.0 = fadd float %tmp16, %tmp17
  %sum.1 = fadd float %sum.0, %tmp18
  %sum.2 = fadd float %sum.1, %tmp19
  %sum.3 = fadd float %sum.2, %tmp20
  %sum.4 = fadd float %sum.3, %tmp21
  %sum.5 = fadd float %sum.4, %tmp22
  %sum.6 = fadd float %sum.5, %tmp23
  %sum.7 = fadd float %sum.6, %tmp24
  %sum.8 = fadd float %sum.7, %tmp25
  store float %sum.8, ptr addrspace(1) %C, align 4
  ret void
}

define amdgpu_kernel void @misaligned_read2_v2i32(ptr addrspace(1) %out, ptr addrspace(3) %in) #0 {
; GFX1250-LABEL: misaligned_read2_v2i32:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b96 s[0:2], s[4:5], 0x0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v0, s2
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v0 offset1:1
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
  %load = load <2 x i32>, ptr addrspace(3) %in, align 4
  store <2 x i32> %load, ptr addrspace(1) %out, align 8
  ret void
}

define amdgpu_kernel void @misaligned_read2_i64(ptr addrspace(1) %out, ptr addrspace(3) %in) #0 {
; GFX1250-LABEL: misaligned_read2_i64:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b96 s[0:2], s[4:5], 0x0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v0, s2
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v0 offset1:1
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
  %load = load i64, ptr addrspace(3) %in, align 4
  store i64 %load, ptr addrspace(1) %out, align 8
  ret void
}

define amdgpu_kernel void @ds_read_diff_base_interleaving(
; GFX1250-LABEL: ds_read_diff_base_interleaving:
; GFX1250:       ; %bb.0: ; %bb
; GFX1250-NEXT:    s_load_b128 s[0:3], s[4:5], 0x8
; GFX1250-NEXT:    v_dual_lshrrev_b32 v1, 6, v0 :: v_dual_lshlrev_b32 v0, 2, v0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX1250-NEXT:    v_and_b32_e32 v1, 0x3ff0, v1
; GFX1250-NEXT:    v_and_b32_e32 v0, 0xffc, v0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; GFX1250-NEXT:    v_dual_add_nc_u32 v2, s0, v1 :: v_dual_add_nc_u32 v3, s1, v0
; GFX1250-NEXT:    v_dual_add_nc_u32 v4, s2, v1 :: v_dual_add_nc_u32 v6, s3, v0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:1
; GFX1250-NEXT:    ds_load_2addr_b32 v[2:3], v3 offset1:4
; GFX1250-NEXT:    ds_load_2addr_b32 v[4:5], v4 offset1:1
; GFX1250-NEXT:    ds_load_2addr_b32 v[6:7], v6 offset1:4
; GFX1250-NEXT:    s_wait_dscnt 0x2
; GFX1250-NEXT:    v_dual_mul_f32 v0, v0, v2 :: v_dual_mul_f32 v1, v1, v3
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_dual_mul_f32 v2, v4, v6 :: v_dual_add_f32 v0, 2.0, v0
; GFX1250-NEXT:    v_dual_sub_f32 v0, v0, v2 :: v_dual_mov_b32 v2, 0
; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1250-NEXT:    v_dual_sub_f32 v0, v0, v1 :: v_dual_mul_f32 v1, v5, v7
; GFX1250-NEXT:    v_sub_f32_e32 v0, v0, v1
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b32 v2, v0, s[0:1] offset:40
; GFX1250-NEXT:    s_endpgm
  ptr addrspace(1) nocapture %arg,
  ptr addrspace(3) %arg1,
  ptr addrspace(3) %arg2,
  ptr addrspace(3) %arg3,
  ptr addrspace(3) %arg4) #1 {
bb:
  %tmp = getelementptr float, ptr addrspace(1) %arg, i64 10
  %tmp5 = tail call i32 @llvm.amdgcn.workitem.id.x() #2
  %tmp6 = tail call i32 @llvm.amdgcn.workitem.id.y() #2
  %tmp7 = getelementptr [4 x [4 x float]], ptr addrspace(3) %arg1, i32 0, i32 %tmp6, i32 0
  %tmp8 = getelementptr [4 x [4 x float]], ptr addrspace(3) %arg2, i32 0, i32 0, i32 %tmp5
  %tmp9 = getelementptr [4 x [4 x float]], ptr addrspace(3) %arg3, i32 0, i32 %tmp6, i32 0
  %tmp10 = getelementptr [4 x [4 x float]], ptr addrspace(3) %arg4, i32 0, i32 0, i32 %tmp5
  %tmp11 = getelementptr [4 x [4 x float]], ptr addrspace(3) %arg1, i32 0, i32 %tmp6, i32 1
  %tmp12 = getelementptr [4 x [4 x float]], ptr addrspace(3) %arg2, i32 0, i32 1, i32 %tmp5
  %tmp13 = getelementptr [4 x [4 x float]], ptr addrspace(3) %arg3, i32 0, i32 %tmp6, i32 1
  %tmp14 = getelementptr [4 x [4 x float]], ptr addrspace(3) %arg4, i32 0, i32 1, i32 %tmp5
  %tmp15 = load float, ptr addrspace(3) %tmp7
  %tmp16 = load float, ptr addrspace(3) %tmp8
  %tmp17 = fmul float %tmp15, %tmp16
  %tmp18 = fadd float 2.000000e+00, %tmp17
  %tmp19 = load float, ptr addrspace(3) %tmp9
  %tmp20 = load float, ptr addrspace(3) %tmp10
  %tmp21 = fmul float %tmp19, %tmp20
  %tmp22 = fsub float %tmp18, %tmp21
  %tmp23 = load float, ptr addrspace(3) %tmp11
  %tmp24 = load float, ptr addrspace(3) %tmp12
  %tmp25 = fmul float %tmp23, %tmp24
  %tmp26 = fsub float %tmp22, %tmp25
  %tmp27 = load float, ptr addrspace(3) %tmp13
  %tmp28 = load float, ptr addrspace(3) %tmp14
  %tmp29 = fmul float %tmp27, %tmp28
  %tmp30 = fsub float %tmp26, %tmp29
  store float %tmp30, ptr addrspace(1) %tmp
  ret void
}

define amdgpu_kernel void @ds_read_call_read(ptr addrspace(1) %out, ptr addrspace(3) %arg) {
; GFX1250-LABEL: ds_read_call_read:
; GFX1250:       ; %bb.0:
; GFX1250-NEXT:    s_load_b96 s[36:38], s[4:5], 0x0
; GFX1250-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
; GFX1250-NEXT:    v_dual_mov_b32 v42, 0 :: v_dual_mov_b32 v31, v0
; GFX1250-NEXT:    s_mov_b64 s[10:11], s[6:7]
; GFX1250-NEXT:    s_add_nc_u64 s[8:9], s[4:5], 12
; GFX1250-NEXT:    s_mov_b64 s[12:13], void_func_void@abs64
; GFX1250-NEXT:    s_wait_xcnt 0x0
; GFX1250-NEXT:    s_mov_b64 s[4:5], s[0:1]
; GFX1250-NEXT:    s_mov_b64 s[6:7], s[2:3]
; GFX1250-NEXT:    s_mov_b32 s32, 0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    v_lshl_add_u32 v40, v1, 2, s38
; GFX1250-NEXT:    ds_load_b32 v41, v40
; GFX1250-NEXT:    s_swap_pc_i64 s[30:31], s[12:13]
; GFX1250-NEXT:    ds_load_b32 v0, v40 offset:4
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    v_add_nc_u32_e32 v0, v41, v0
; GFX1250-NEXT:    global_store_b32 v42, v0, s[36:37]
; GFX1250-NEXT:    s_endpgm
  %x = call i32 @llvm.amdgcn.workitem.id.x()
  %arrayidx0 = getelementptr i32, ptr addrspace(3) %arg, i32 %x
  %arrayidx1 = getelementptr i32, ptr addrspace(3) %arrayidx0, i32 1
  %v0 = load i32, ptr addrspace(3) %arrayidx0, align 4
  call void @void_func_void()
  %v1 = load i32, ptr addrspace(3) %arrayidx1, align 4
  %r = add i32 %v0, %v1
  store i32 %r, ptr addrspace(1) %out, align 4
  ret void
}

@v2i32_align1 = internal addrspace(3) global [100 x <2 x i32>] poison, align 1

define amdgpu_kernel void @read2_v2i32_align1_odd_offset(ptr addrspace(1) %out) {
; GFX1250-LABEL: read2_v2i32_align1_odd_offset:
; GFX1250:       ; %bb.0: ; %entry
; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
; GFX1250-NEXT:    ds_load_b64 v[0:1], v2 offset:65
; GFX1250-NEXT:    s_wait_dscnt 0x0
; GFX1250-NEXT:    s_wait_kmcnt 0x0
; GFX1250-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
; GFX1250-NEXT:    s_endpgm
entry:
  %load = load <2 x i32>, ptr addrspace(3) getelementptr (i8, ptr addrspace(3) @v2i32_align1, i32 65), align 1
  store <2 x i32> %load, ptr addrspace(1) %out
  ret void
}

declare void @void_func_void() #3

declare i32 @llvm.amdgcn.workgroup.id.x() #1
declare i32 @llvm.amdgcn.workgroup.id.y() #1
declare i32 @llvm.amdgcn.workitem.id.x() #1
declare i32 @llvm.amdgcn.workitem.id.y() #1

declare void @llvm.amdgcn.s.barrier() #2

attributes #0 = { nounwind }
attributes #1 = { nounwind readnone speculatable }
attributes #2 = { convergent nounwind }
attributes #3 = { nounwind noinline }

;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX1250-UNALIGNED: {{.*}}