[AMDGPU] Add some exhaustive ds read/write alignment tests

[hsmhsm]

Summary

PS: Submitting on behalf of Jay.

llvm/test/CodeGen/AMDGPU/ds-alignment.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx900 -mattr=-unaligned-access-mode < %s | FileCheck %s -check-prefixes=GCN,ALIGNED,ALIGNED-SDAG
+; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx900 -mattr=-unaligned-access-mode < %s | FileCheck %s -check-prefixes=GCN,ALIGNED,ALIGNED-GISEL
+; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx900 -mattr=+unaligned-access-mode < %s | FileCheck %s -check-prefixes=GCN,UNALIGNED
+; RUN: llc -global-isel=1 -march=amdgcn -mcpu=gfx900 -mattr=+unaligned-access-mode < %s | FileCheck %s -check-prefixes=GCN,UNALIGNED
+
+define amdgpu_kernel void @ds1align1(i8 addrspace(3)* %in, i8 addrspace(3)* %out) {
+; GCN-LABEL: ds1align1:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    ds_read_u8 v0, v0
+; GCN-NEXT:    v_mov_b32_e32 v1, s1
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    ds_write_b8 v1, v0
+; GCN-NEXT:    s_endpgm
+  %val = load i8, i8 addrspace(3)* %in, align 1
+  store i8 %val, i8 addrspace(3)* %out, align 1
+  ret void
+}
+
+define amdgpu_kernel void @ds2align1(i16 addrspace(3)* %in, i16 addrspace(3)* %out) {
+; ALIGNED-LABEL: ds2align1:
+; ALIGNED:       ; %bb.0:
+; ALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-NEXT:    ds_read_u8 v1, v0
+; ALIGNED-NEXT:    ds_read_u8 v0, v0 offset:1
+; ALIGNED-NEXT:    v_mov_b32_e32 v2, s1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-NEXT:    ds_write_b8 v2, v1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-NEXT:    ds_write_b8 v2, v0 offset:1
+; ALIGNED-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds2align1:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_u16 v0, v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v1, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b16 v1, v0
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load i16, i16 addrspace(3)* %in, align 1
+  store i16 %val, i16 addrspace(3)* %out, align 1
+  ret void
+}
+
+define amdgpu_kernel void @ds2align2(i16 addrspace(3)* %in, i16 addrspace(3)* %out) {
+; GCN-LABEL: ds2align2:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    ds_read_u16 v0, v0
+; GCN-NEXT:    v_mov_b32_e32 v1, s1
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    ds_write_b16 v1, v0
+; GCN-NEXT:    s_endpgm
+  %val = load i16, i16 addrspace(3)* %in, align 2
+  store i16 %val, i16 addrspace(3)* %out, align 2
+  ret void
+}
+
+define amdgpu_kernel void @ds4align1(i32 addrspace(3)* %in, i32 addrspace(3)* %out) {
+; ALIGNED-LABEL: ds4align1:
+; ALIGNED:       ; %bb.0:
+; ALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-NEXT:    ds_read_u8 v1, v0
+; ALIGNED-NEXT:    ds_read_u8 v2, v0 offset:1
+; ALIGNED-NEXT:    ds_read_u8 v3, v0 offset:2
+; ALIGNED-NEXT:    ds_read_u8 v0, v0 offset:3
+; ALIGNED-NEXT:    v_mov_b32_e32 v4, s1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-NEXT:    ds_write_b8 v4, v1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-NEXT:    ds_write_b8 v4, v2 offset:1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-NEXT:    ds_write_b8 v4, v3 offset:2
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-NEXT:    ds_write_b8 v4, v0 offset:3
+; ALIGNED-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds4align1:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b32 v0, v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v1, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b32 v1, v0
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load i32, i32 addrspace(3)* %in, align 1
+  store i32 %val, i32 addrspace(3)* %out, align 1
+  ret void
+}
+
+define amdgpu_kernel void @ds4align2(i32 addrspace(3)* %in, i32 addrspace(3)* %out) {
+; ALIGNED-LABEL: ds4align2:
+; ALIGNED:       ; %bb.0:
+; ALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-NEXT:    ds_read_u16 v1, v0
+; ALIGNED-NEXT:    ds_read_u16 v0, v0 offset:2
+; ALIGNED-NEXT:    v_mov_b32_e32 v2, s1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-NEXT:    ds_write_b16 v2, v1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-NEXT:    ds_write_b16 v2, v0 offset:2
+; ALIGNED-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds4align2:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b32 v0, v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v1, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b32 v1, v0
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load i32, i32 addrspace(3)* %in, align 2
+  store i32 %val, i32 addrspace(3)* %out, align 2
+  ret void
+}
+
+define amdgpu_kernel void @ds4align4(i32 addrspace(3)* %in, i32 addrspace(3)* %out) {
+; GCN-LABEL: ds4align4:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    ds_read_b32 v0, v0
+; GCN-NEXT:    v_mov_b32_e32 v1, s1
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    ds_write_b32 v1, v0
+; GCN-NEXT:    s_endpgm
+  %val = load i32, i32 addrspace(3)* %in, align 4
+  store i32 %val, i32 addrspace(3)* %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @ds8align1(<2 x i32> addrspace(3)* %in, <2 x i32> addrspace(3)* %out) {
+; ALIGNED-SDAG-LABEL: ds8align1:
+; ALIGNED-SDAG:       ; %bb.0:
+; ALIGNED-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v2, v0
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v3, v0 offset:1
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v4, v0 offset:2
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v5, v0 offset:3
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v6, v0 offset:4
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v7, v0 offset:5
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v8, v0 offset:6
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v0, v0 offset:7
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v1, v4 offset:2
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v1, v5 offset:3
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v1, v2
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v1, v3 offset:1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v1, v8 offset:6
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v1, v0 offset:7
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v1, v6 offset:4
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v1, v7 offset:5
+; ALIGNED-SDAG-NEXT:    s_endpgm
+;
+; ALIGNED-GISEL-LABEL: ds8align1:
+; ALIGNED-GISEL:       ; %bb.0:
+; ALIGNED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v2, v0
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v3, v0 offset:1
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v4, v0 offset:2
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v5, v0 offset:3
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v6, v0 offset:4
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v7, v0 offset:5
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v8, v0 offset:6
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v0, v0 offset:7
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v1, v2
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v1, v3 offset:1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v1, v4 offset:2
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v1, v5 offset:3
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v1, v6 offset:4
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v1, v7 offset:5
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v1, v8 offset:6
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v1, v0 offset:7
+; ALIGNED-GISEL-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds8align1:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read2_b32 v[0:1], v0 offset1:1
+; UNALIGNED-NEXT:    v_mov_b32_e32 v2, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write2_b32 v2, v0, v1 offset1:1
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <2 x i32>, <2 x i32> addrspace(3)* %in, align 1
+  store <2 x i32> %val, <2 x i32> addrspace(3)* %out, align 1
+  ret void
+}
+
+define amdgpu_kernel void @ds8align2(<2 x i32> addrspace(3)* %in, <2 x i32> addrspace(3)* %out) {
+; ALIGNED-SDAG-LABEL: ds8align2:
+; ALIGNED-SDAG:       ; %bb.0:
+; ALIGNED-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v1, v0
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v2, v0 offset:2
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v3, v0 offset:4
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v0, v0 offset:6
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v4, s1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(2)
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v4, v2 offset:2
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v4, v1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(2)
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v4, v0 offset:6
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v4, v3 offset:4
+; ALIGNED-SDAG-NEXT:    s_endpgm
+;
+; ALIGNED-GISEL-LABEL: ds8align2:
+; ALIGNED-GISEL:       ; %bb.0:
+; ALIGNED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v1, v0
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v2, v0 offset:2
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v3, v0 offset:4
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v0, v0 offset:6
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v4, s1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v2 offset:2
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v3 offset:4
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v0 offset:6
+; ALIGNED-GISEL-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds8align2:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read2_b32 v[0:1], v0 offset1:1
+; UNALIGNED-NEXT:    v_mov_b32_e32 v2, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write2_b32 v2, v0, v1 offset1:1
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <2 x i32>, <2 x i32> addrspace(3)* %in, align 2
+  store <2 x i32> %val, <2 x i32> addrspace(3)* %out, align 2
+  ret void
+}
+
+define amdgpu_kernel void @ds8align4(<2 x i32> addrspace(3)* %in, <2 x i32> addrspace(3)* %out) {
+; GCN-LABEL: ds8align4:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    ds_read2_b32 v[0:1], v0 offset1:1
+; GCN-NEXT:    v_mov_b32_e32 v2, s1
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    ds_write2_b32 v2, v0, v1 offset1:1
+; GCN-NEXT:    s_endpgm
+  %val = load <2 x i32>, <2 x i32> addrspace(3)* %in, align 4
+  store <2 x i32> %val, <2 x i32> addrspace(3)* %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @ds8align8(<2 x i32> addrspace(3)* %in, <2 x i32> addrspace(3)* %out) {
+; GCN-LABEL: ds8align8:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    ds_read_b64 v[0:1], v0
+; GCN-NEXT:    v_mov_b32_e32 v2, s1
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    ds_write_b64 v2, v[0:1]
+; GCN-NEXT:    s_endpgm
+  %val = load <2 x i32>, <2 x i32> addrspace(3)* %in, align 8
+  store <2 x i32> %val, <2 x i32> addrspace(3)* %out, align 8
+  ret void
+}
+
+define amdgpu_kernel void @ds12align1(<3 x i32> addrspace(3)* %in, <3 x i32> addrspace(3)* %out) {
+; ALIGNED-SDAG-LABEL: ds12align1:
+; ALIGNED-SDAG:       ; %bb.0:
+; ALIGNED-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v1, v0
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v2, v0 offset:1
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v3, v0 offset:2
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v4, v0 offset:3
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v5, v0 offset:4
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v6, v0 offset:5
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v7, v0 offset:6
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v8, v0 offset:7
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v9, v0 offset:8
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v10, v0 offset:9
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v11, v0 offset:10
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v0, v0 offset:11
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v12, s1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v9 offset:8
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v10 offset:9
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v3 offset:2
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v4 offset:3
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v1
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v2 offset:1
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v5 offset:4
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v6 offset:5
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v7 offset:6
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v8 offset:7
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(11)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v11 offset:10
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(11)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v12, v0 offset:11
+; ALIGNED-SDAG-NEXT:    s_endpgm
+;
+; ALIGNED-GISEL-LABEL: ds12align1:
+; ALIGNED-GISEL:       ; %bb.0:
+; ALIGNED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-GISEL-NEXT:    s_mov_b32 s3, 8
+; ALIGNED-GISEL-NEXT:    s_movk_i32 s2, 0xff
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v2, 0xff
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v3, s0
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v0, v3
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v1, v3 offset:1
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v4, v3 offset:2
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v5, v3 offset:3
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v6, v3 offset:4
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v7, v3 offset:5
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v8, v3 offset:6
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v9, v3 offset:7
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(6)
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_sdwa v1, s3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; ALIGNED-GISEL-NEXT:    v_and_or_b32 v0, v0, s2, v1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(5)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v1, s2, v4
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(4)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v4, s2, v5
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v4, 24, v4
+; ALIGNED-GISEL-NEXT:    v_or3_b32 v0, v0, v1, v4
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(2)
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_sdwa v1, s3, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v4, v8, v2
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v5, v9, v2
+; ALIGNED-GISEL-NEXT:    v_and_or_b32 v1, v6, s2, v1
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v5, 24, v5
+; ALIGNED-GISEL-NEXT:    v_or3_b32 v1, v1, v4, v5
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v4, v3 offset:8
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v5, v3 offset:9
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v6, v3 offset:10
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v3, v3 offset:11
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v7, 8
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(2)
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_sdwa v5, v7, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; ALIGNED-GISEL-NEXT:    v_and_or_b32 v4, v4, v2, v5
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v5, v6, v2
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v2, v3, v2
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v5, 16, v5
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 24, v2
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v6, s1
+; ALIGNED-GISEL-NEXT:    v_or3_b32 v2, v4, v5, v2
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v5, 24, v0
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v0
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v3 offset:1
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v4 offset:2
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v5 offset:3
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v0, 8, v1
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v3, 16, v1
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v4, 24, v1
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v1 offset:4
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v0 offset:5
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v3 offset:6
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v4 offset:7
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v0, 8, v2
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v1, 16, v2
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v3, 24, v2
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v2 offset:8
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v0 offset:9
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v1 offset:10
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v6, v3 offset:11
+; ALIGNED-GISEL-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds12align1:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b96 v[0:2], v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v3, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b96 v3, v[0:2]
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <3 x i32>, <3 x i32> addrspace(3)* %in, align 1
+  store <3 x i32> %val, <3 x i32> addrspace(3)* %out, align 1
+  ret void
+}
+
+define amdgpu_kernel void @ds12align2(<3 x i32> addrspace(3)* %in, <3 x i32> addrspace(3)* %out) {
+; ALIGNED-SDAG-LABEL: ds12align2:
+; ALIGNED-SDAG:       ; %bb.0:
+; ALIGNED-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v2, v0
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v3, v0 offset:2
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v4, v0 offset:4
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v5, v0 offset:6
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v6, v0 offset:8
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v0, v0 offset:10
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v6 offset:8
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v3 offset:2
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v2
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v4 offset:4
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v5 offset:6
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v0 offset:10
+; ALIGNED-SDAG-NEXT:    s_endpgm
+;
+; ALIGNED-GISEL-LABEL: ds12align2:
+; ALIGNED-GISEL:       ; %bb.0:
+; ALIGNED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-GISEL-NEXT:    s_mov_b32 s2, 0xffff
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v1, v0
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v2, v0 offset:2
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v3, v0 offset:4
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v4, v0 offset:6
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v5, v0 offset:8
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v6, v0 offset:10
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(4)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v0, s2, v2
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(2)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v2, s2, v4
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; ALIGNED-GISEL-NEXT:    v_and_or_b32 v0, v1, s2, v0
+; ALIGNED-GISEL-NEXT:    v_and_or_b32 v1, v3, s2, v2
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_and_b32_e32 v2, s2, v6
+; ALIGNED-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v4, s1
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v3, 16, v0
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v0
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v3 offset:2
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v0, 16, v1
+; ALIGNED-GISEL-NEXT:    v_and_or_b32 v2, v5, s2, v2
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v1 offset:4
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v0 offset:6
+; ALIGNED-GISEL-NEXT:    v_lshrrev_b32_e32 v0, 16, v2
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v2 offset:8
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v4, v0 offset:10
+; ALIGNED-GISEL-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds12align2:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b96 v[0:2], v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v3, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b96 v3, v[0:2]
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <3 x i32>, <3 x i32> addrspace(3)* %in, align 2
+  store <3 x i32> %val, <3 x i32> addrspace(3)* %out, align 2
+  ret void
+}
+
+define amdgpu_kernel void @ds12align4(<3 x i32> addrspace(3)* %in, <3 x i32> addrspace(3)* %out) {
+; ALIGNED-LABEL: ds12align4:
+; ALIGNED:       ; %bb.0:
+; ALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-NEXT:    v_mov_b32_e32 v2, s0
+; ALIGNED-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
+; ALIGNED-NEXT:    ds_read_b32 v2, v2 offset:8
+; ALIGNED-NEXT:    v_mov_b32_e32 v3, s1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-NEXT:    ds_write2_b32 v3, v0, v1 offset1:1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-NEXT:    ds_write_b32 v3, v2 offset:8
+; ALIGNED-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds12align4:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b96 v[0:2], v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v3, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b96 v3, v[0:2]
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <3 x i32>, <3 x i32> addrspace(3)* %in, align 4
+  store <3 x i32> %val, <3 x i32> addrspace(3)* %out, align 4
+  ret void
+}
+
+; TODO: Why does the ALIGNED-SDAG code use ds_write_b64 but not ds_read_b64?
+define amdgpu_kernel void @ds12align8(<3 x i32> addrspace(3)* %in, <3 x i32> addrspace(3)* %out) {
+; ALIGNED-SDAG-LABEL: ds12align8:
+; ALIGNED-SDAG:       ; %bb.0:
+; ALIGNED-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; ALIGNED-SDAG-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
+; ALIGNED-SDAG-NEXT:    ds_read_b32 v2, v2 offset:8
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    ds_write_b32 v3, v2 offset:8
+; ALIGNED-SDAG-NEXT:    ds_write_b64 v3, v[0:1]
+; ALIGNED-SDAG-NEXT:    s_endpgm
+;
+; ALIGNED-GISEL-LABEL: ds12align8:
+; ALIGNED-GISEL:       ; %bb.0:
+; ALIGNED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; ALIGNED-GISEL-NEXT:    ds_read_b64 v[0:1], v2
+; ALIGNED-GISEL-NEXT:    ds_read_b32 v2, v2 offset:8
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-GISEL-NEXT:    ds_write_b64 v3, v[0:1]
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-GISEL-NEXT:    ds_write_b32 v3, v2 offset:8
+; ALIGNED-GISEL-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds12align8:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b96 v[0:2], v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v3, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b96 v3, v[0:2]
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <3 x i32>, <3 x i32> addrspace(3)* %in, align 8
+  store <3 x i32> %val, <3 x i32> addrspace(3)* %out, align 8
+  ret void
+}
+
+define amdgpu_kernel void @ds12align16(<3 x i32> addrspace(3)* %in, <3 x i32> addrspace(3)* %out) {
+; GCN-LABEL: ds12align16:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    ds_read_b96 v[0:2], v0
+; GCN-NEXT:    v_mov_b32_e32 v3, s1
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    ds_write_b96 v3, v[0:2]
+; GCN-NEXT:    s_endpgm
+  %val = load <3 x i32>, <3 x i32> addrspace(3)* %in, align 16
+  store <3 x i32> %val, <3 x i32> addrspace(3)* %out, align 16
+  ret void
+}
+
+define amdgpu_kernel void @ds16align1(<4 x i32> addrspace(3)* %in, <4 x i32> addrspace(3)* %out) {
+; ALIGNED-SDAG-LABEL: ds16align1:
+; ALIGNED-SDAG:       ; %bb.0:
+; ALIGNED-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v1, v0
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v2, v0 offset:1
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v3, v0 offset:2
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v4, v0 offset:3
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v5, v0 offset:4
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v6, v0 offset:5
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v7, v0 offset:6
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v8, v0 offset:7
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v9, v0 offset:8
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v10, v0 offset:9
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v11, v0 offset:10
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v12, v0 offset:11
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v13, v0 offset:12
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v14, v0 offset:13
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v15, v0 offset:14
+; ALIGNED-SDAG-NEXT:    ds_read_u8 v0, v0 offset:15
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v16, s1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v13 offset:12
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(3)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v14 offset:13
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v3 offset:2
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v4 offset:3
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v1
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v2 offset:1
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v5 offset:4
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v6 offset:5
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v9 offset:8
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v10 offset:9
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v7 offset:6
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v8 offset:7
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v11 offset:10
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v12 offset:11
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v15 offset:14
+; ALIGNED-SDAG-NEXT:    ds_write_b8 v16, v0 offset:15
+; ALIGNED-SDAG-NEXT:    s_endpgm
+;
+; ALIGNED-GISEL-LABEL: ds16align1:
+; ALIGNED-GISEL:       ; %bb.0:
+; ALIGNED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v1, v0
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v2, v0 offset:1
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v3, v0 offset:2
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v4, v0 offset:3
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v5, v0 offset:4
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v6, v0 offset:5
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v7, v0 offset:6
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v8, v0 offset:7
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v9, v0 offset:8
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v10, v0 offset:9
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v11, v0 offset:10
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v12, v0 offset:11
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v13, v0 offset:12
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v14, v0 offset:13
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v15, v0 offset:14
+; ALIGNED-GISEL-NEXT:    ds_read_u8 v0, v0 offset:15
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v16, s1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v1
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v2 offset:1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v3 offset:2
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v4 offset:3
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v5 offset:4
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v6 offset:5
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v7 offset:6
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v8 offset:7
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v9 offset:8
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v10 offset:9
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v11 offset:10
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v12 offset:11
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v13 offset:12
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v14 offset:13
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(14)
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v15 offset:14
+; ALIGNED-GISEL-NEXT:    ds_write_b8 v16, v0 offset:15
+; ALIGNED-GISEL-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds16align1:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b128 v[0:3], v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v4, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b128 v4, v[0:3]
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <4 x i32>, <4 x i32> addrspace(3)* %in, align 1
+  store <4 x i32> %val, <4 x i32> addrspace(3)* %out, align 1
+  ret void
+}
+
+define amdgpu_kernel void @ds16align2(<4 x i32> addrspace(3)* %in, <4 x i32> addrspace(3)* %out) {
+; ALIGNED-SDAG-LABEL: ds16align2:
+; ALIGNED-SDAG:       ; %bb.0:
+; ALIGNED-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v2, v0
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v3, v0 offset:2
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v4, v0 offset:4
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v5, v0 offset:6
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v6, v0 offset:8
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v7, v0 offset:10
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v8, v0 offset:12
+; ALIGNED-SDAG-NEXT:    ds_read_u16 v0, v0 offset:14
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v8 offset:12
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v3 offset:2
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v2
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v4 offset:4
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v6 offset:8
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v5 offset:6
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v7 offset:10
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-SDAG-NEXT:    ds_write_b16 v1, v0 offset:14
+; ALIGNED-SDAG-NEXT:    s_endpgm
+;
+; ALIGNED-GISEL-LABEL: ds16align2:
+; ALIGNED-GISEL:       ; %bb.0:
+; ALIGNED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v2, v0
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v3, v0 offset:2
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v4, v0 offset:4
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v5, v0 offset:6
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v6, v0 offset:8
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v7, v0 offset:10
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v8, v0 offset:12
+; ALIGNED-GISEL-NEXT:    ds_read_u16 v0, v0 offset:14
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v1, v2
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v1, v3 offset:2
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v1, v4 offset:4
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v1, v5 offset:6
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v1, v6 offset:8
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v1, v7 offset:10
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v1, v8 offset:12
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(7)
+; ALIGNED-GISEL-NEXT:    ds_write_b16 v1, v0 offset:14
+; ALIGNED-GISEL-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds16align2:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b128 v[0:3], v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v4, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b128 v4, v[0:3]
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <4 x i32>, <4 x i32> addrspace(3)* %in, align 2
+  store <4 x i32> %val, <4 x i32> addrspace(3)* %out, align 2
+  ret void
+}
+
+define amdgpu_kernel void @ds16align4(<4 x i32> addrspace(3)* %in, <4 x i32> addrspace(3)* %out) {
+; ALIGNED-SDAG-LABEL: ds16align4:
+; ALIGNED-SDAG:       ; %bb.0:
+; ALIGNED-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; ALIGNED-SDAG-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
+; ALIGNED-SDAG-NEXT:    ds_read2_b32 v[2:3], v2 offset0:2 offset1:3
+; ALIGNED-SDAG-NEXT:    v_mov_b32_e32 v4, s1
+; ALIGNED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-SDAG-NEXT:    ds_write2_b32 v4, v2, v3 offset0:2 offset1:3
+; ALIGNED-SDAG-NEXT:    ds_write2_b32 v4, v0, v1 offset1:1
+; ALIGNED-SDAG-NEXT:    s_endpgm
+;
+; ALIGNED-GISEL-LABEL: ds16align4:
+; ALIGNED-GISEL:       ; %bb.0:
+; ALIGNED-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; ALIGNED-GISEL-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
+; ALIGNED-GISEL-NEXT:    ds_read2_b32 v[2:3], v2 offset0:2 offset1:3
+; ALIGNED-GISEL-NEXT:    v_mov_b32_e32 v4, s1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-GISEL-NEXT:    ds_write2_b32 v4, v0, v1 offset1:1
+; ALIGNED-GISEL-NEXT:    s_waitcnt lgkmcnt(1)
+; ALIGNED-GISEL-NEXT:    ds_write2_b32 v4, v2, v3 offset0:2 offset1:3
+; ALIGNED-GISEL-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds16align4:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b128 v[0:3], v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v4, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b128 v4, v[0:3]
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <4 x i32>, <4 x i32> addrspace(3)* %in, align 4
+  store <4 x i32> %val, <4 x i32> addrspace(3)* %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @ds16align8(<4 x i32> addrspace(3)* %in, <4 x i32> addrspace(3)* %out) {
+; ALIGNED-LABEL: ds16align8:
+; ALIGNED:       ; %bb.0:
+; ALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; ALIGNED-NEXT:    ds_read2_b64 v[0:3], v0 offset1:1
+; ALIGNED-NEXT:    v_mov_b32_e32 v4, s1
+; ALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; ALIGNED-NEXT:    ds_write2_b64 v4, v[0:1], v[2:3] offset1:1
+; ALIGNED-NEXT:    s_endpgm
+;
+; UNALIGNED-LABEL: ds16align8:
+; UNALIGNED:       ; %bb.0:
+; UNALIGNED-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    v_mov_b32_e32 v0, s0
+; UNALIGNED-NEXT:    ds_read_b128 v[0:3], v0
+; UNALIGNED-NEXT:    v_mov_b32_e32 v4, s1
+; UNALIGNED-NEXT:    s_waitcnt lgkmcnt(0)
+; UNALIGNED-NEXT:    ds_write_b128 v4, v[0:3]
+; UNALIGNED-NEXT:    s_endpgm
+  %val = load <4 x i32>, <4 x i32> addrspace(3)* %in, align 8
+  store <4 x i32> %val, <4 x i32> addrspace(3)* %out, align 8
+  ret void
+}
+
+define amdgpu_kernel void @ds16align16(<4 x i32> addrspace(3)* %in, <4 x i32> addrspace(3)* %out) {
+; GCN-LABEL: ds16align16:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-NEXT:    ds_read_b128 v[0:3], v0
+; GCN-NEXT:    v_mov_b32_e32 v4, s1
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    ds_write_b128 v4, v[0:3]
+; GCN-NEXT:    s_endpgm
+  %val = load <4 x i32>, <4 x i32> addrspace(3)* %in, align 16
+  store <4 x i32> %val, <4 x i32> addrspace(3)* %out, align 16
+  ret void
+}