[AMDGPU] Autogenerate checks in kernel-args.ll

[foad]

llvm/test/CodeGen/AMDGPU/kernel-args.ll

1

I removed -allow-deprecated-dag-overlap and -enable-var-scope because I didn't know what they were for.

2

I removed -mattr=-flat-for-global because I assumed it was only added in the first place to avoid churn in the tests when some compiler default was changed.

3

I removed --amdhsa-code-object-version=2 because it seemed to bloat the checks with a load of directives starting with .amd_kernel_code_t.

[lebedev.ri]

Thanks. Non-AMDGPU-specific part this looks fine to me.
Not sure if @arsenm/etc want to stamp this, or if this needs review at all.
For context, this makes the diff in D106447 much more non-unreadable/misleading.

llvm/test/CodeGen/AMDGPU/kernel-args.ll

1

These don't really matter for autogenerated check lines

llvm/test/CodeGen/AMDGPU/kernel-args.ll

-; RUN: llc < %s -march=amdgcn -verify-machineinstrs | FileCheck -allow-deprecated-dag-overlap -enable-var-scope --check-prefixes=SI,GCN,MESA-GCN,FUNC %s
-; RUN: llc < %s -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs | FileCheck -allow-deprecated-dag-overlap -enable-var-scope -check-prefixes=VI,GCN,MESA-VI,MESA-GCN,FUNC %s
-; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=gfx900 --amdhsa-code-object-version=2 -verify-machineinstrs | FileCheck -allow-deprecated-dag-overlap -enable-var-scope -check-prefixes=VI,GCN,HSA-GFX9,FUNC %s
-; RUN: llc < %s -march=r600 -mcpu=redwood -verify-machineinstrs | FileCheck -allow-deprecated-dag-overlap -enable-var-scope -check-prefixes=EG,EGCM,FUNC %s
-; RUN: llc < %s -march=r600 -mcpu=cayman -verify-machineinstrs | FileCheck -allow-deprecated-dag-overlap -enable-var-scope --check-prefixes=CM,EGCM,FUNC %s
-
-; FUNC-LABEL: {{^}}i8_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; SI: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0xb
-; MESA-VI: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x2c
-; MESA-GCN: s_and_b32 s{{[0-9]+}}, [[VAL]], 0xff
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -march=amdgcn -verify-machineinstrs | FileCheck -check-prefixes=SI %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefixes=VI %s
+; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=gfx900 -verify-machineinstrs | FileCheck -check-prefixes=GFX9 %s
+; RUN: llc < %s -march=r600 -mcpu=redwood -verify-machineinstrs | FileCheck -check-prefixes=EGCM,EG %s
+; RUN: llc < %s -march=r600 -mcpu=cayman -verify-machineinstrs | FileCheck -check-prefixes=EGCM,CM %s
 
-; HSA-GFX9: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x8
-; HSA-GFX9: s_and_b32 s{{[0-9]+}}, [[VAL]], 0xff
-
-
-; EGCM: VTX_READ_8{{.*}} #3
-; EGCM: KC0[2].Y
 define amdgpu_kernel void @i8_arg(i32 addrspace(1)* nocapture %out, i8 %in) nounwind {
+; SI-LABEL: i8_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_and_b32 s4, s2, 0xff
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i8_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_and_b32 s0, s0, 0xff
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i8_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_and_b32 s2, s2, 0xff
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i8_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 1, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: i8_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 1, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = zext i8 %in to i32
   store i32 %ext, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i8_zext_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-; SI: s_load_dword s{{[0-9]}}, s[0:1], 0xb
-; MESA-VI: s_load_dword s{{[0-9]}}, s[0:1], 0x2c
-
-; HSA-GFX9: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x8
-; HSA-GFX9: s_and_b32 s{{[0-9]+}}, [[VAL]], 0xff
-
-
-; EG: BFE_INT   T0.X, T0.X, 0.0, literal.x,
-; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
-; EG-NEXT: 8(1.121039e-44), 2(2.802597e-45)
-
-; CM: BFE_INT * T0.X, T0.X, 0.0, literal.x,
-; CM-NEXT: 8(1.121039e-44), 0(0.000000e+00)
-; CM-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
-; CM-NEXT:  2(2.802597e-45), 0(0.000000e+00)
-define amdgpu_kernel void @i8_zext_arg(i32 addrspace(1)* nocapture %out, i8 zeroext %in) nounwind {
+define amdgpu_kernel void @i8_zext_arg(i32 addrspace(1)* nocapture %out, i8 zeroext %in) nounwind {
+; SI-LABEL: i8_zext_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_and_b32 s4, s2, 0xff
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i8_zext_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_and_b32 s0, s0, 0xff
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i8_zext_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_and_b32 s2, s2, 0xff
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i8_zext_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     BFE_INT T0.X, T0.X, 0.0, literal.x,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.y,
+; EG-NEXT:    8(1.121039e-44), 2(2.802597e-45)
+;
+; CM-LABEL: i8_zext_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 3, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     BFE_INT * T0.X, T0.X, 0.0, literal.x,
+; CM-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = zext i8 %in to i32
   store i32 %ext, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i8_sext_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-; SI: s_load_dword s{{[0-9]}}, s[0:1], 0xb
-
-; MESA-VI: s_load_dword s{{[0-9]}}, s[0:1], 0x2c
-
-; HSA-GFX9: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x8
-; HSA-GFX9: s_sext_i32_i8 s{{[0-9]+}}, [[VAL]]
-; HSA-GFX9: global_store_dword
-
-
-; EG: BFE_INT   T0.X, T0.X, 0.0, literal.x,
-; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
-; EG-NEXT: 8(1.121039e-44), 2(2.802597e-45)
-
-; CM: BFE_INT * T0.X, T0.X, 0.0, literal.x,
-; CM-NEXT: 8(1.121039e-44), 0(0.000000e+00)
-; CM-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
-; CM-NEXT: 2(2.802597e-45), 0(0.000000e+00)
-define amdgpu_kernel void @i8_sext_arg(i32 addrspace(1)* nocapture %out, i8 signext %in) nounwind {
+define amdgpu_kernel void @i8_sext_arg(i32 addrspace(1)* nocapture %out, i8 signext %in) nounwind {
+; SI-LABEL: i8_sext_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_sext_i32_i8 s4, s2
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i8_sext_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_sext_i32_i8 s0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i8_sext_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_sext_i32_i8 s2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i8_sext_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     BFE_INT T0.X, T0.X, 0.0, literal.x,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.y,
+; EG-NEXT:    8(1.121039e-44), 2(2.802597e-45)
+;
+; CM-LABEL: i8_sext_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 3, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     BFE_INT * T0.X, T0.X, 0.0, literal.x,
+; CM-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = sext i8 %in to i32
   store i32 %ext, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i16_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; SI: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0xb
-
-; MESA-VI: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x2c
-; MESA-GCN: s_and_b32 s{{[0-9]+}}, [[VAL]], 0xff
-
-; HSA-GFX9: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x8
-; HSA-GFX9: s_and_b32 s{{[0-9]+}}, [[VAL]], 0xffff{{$}}
-; HSA-GFX9: global_store_dword
-
-; EGCM: VTX_READ_16
-; EGCM: KC0[2].Y
 define amdgpu_kernel void @i16_arg(i32 addrspace(1)* nocapture %out, i16 %in) nounwind {
+; SI-LABEL: i16_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_and_b32 s4, s2, 0xffff
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i16_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_and_b32 s0, s0, 0xffff
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i16_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_and_b32 s2, s2, 0xffff
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i16_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 1, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_16 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: i16_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 1, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_16 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = zext i16 %in to i32
   store i32 %ext, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i16_zext_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; SI: s_load_dword s{{[0-9]}}, s[0:1], 0xb
-; MESA-VI: s_load_dword s{{[0-9]}}, s[0:1], 0x2c
-
-; HSA-GFX9: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x8
-; HSA-GFX9: s_and_b32 s{{[0-9]+}}, [[VAL]], 0xffff{{$}}
-; HSA-GFX9: global_store_dword
-
-; EG: BFE_INT   T0.X, T0.X, 0.0, literal.x,
-; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
-; EG-NEXT: 16(2.242078e-44), 2(2.802597e-45)
-
-; CM: BFE_INT * T0.X, T0.X, 0.0, literal.x,
-; CM-NEXT: 16(2.242078e-44), 0(0.000000e+00)
-; CM-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
-; CM-NEXT: 2(2.802597e-45), 0(0.000000e+00)
-define amdgpu_kernel void @i16_zext_arg(i32 addrspace(1)* nocapture %out, i16 zeroext %in) nounwind {
+define amdgpu_kernel void @i16_zext_arg(i32 addrspace(1)* nocapture %out, i16 zeroext %in) nounwind {
+; SI-LABEL: i16_zext_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_and_b32 s4, s2, 0xffff
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i16_zext_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_and_b32 s0, s0, 0xffff
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i16_zext_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_and_b32 s2, s2, 0xffff
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i16_zext_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_16 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     BFE_INT T0.X, T0.X, 0.0, literal.x,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.y,
+; EG-NEXT:    16(2.242078e-44), 2(2.802597e-45)
+;
+; CM-LABEL: i16_zext_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 3, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_16 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     BFE_INT * T0.X, T0.X, 0.0, literal.x,
+; CM-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = zext i16 %in to i32
   store i32 %ext, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i16_sext_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; SI: s_load_dword s{{[0-9]}}, s[0:1], 0xb
-; MESA-VI: s_load_dword s{{[0-9]}}, s[0:1], 0x2c
-
-
-; HSA-GFX9: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x8
-; HSA-GFX9: s_sext_i32_i16 s{{[0-9]+}}, [[VAL]]
-; HSA-GFX9: global_store_dword
-
-; EG: BFE_INT   T0.X, T0.X, 0.0, literal.x,
-; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
-; EG-NEXT: 16(2.242078e-44), 2(2.802597e-45)
-
-; CM: BFE_INT * T0.X, T0.X, 0.0, literal.x,
-; CM-NEXT: 16(2.242078e-44), 0(0.000000e+00)
-; CM-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
-; CM-NEXT: 2(2.802597e-45), 0(0.000000e+00)
-define amdgpu_kernel void @i16_sext_arg(i32 addrspace(1)* nocapture %out, i16 signext %in) nounwind {
+define amdgpu_kernel void @i16_sext_arg(i32 addrspace(1)* nocapture %out, i16 signext %in) nounwind {
+; SI-LABEL: i16_sext_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_sext_i32_i16 s4, s2
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i16_sext_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_sext_i32_i16 s0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i16_sext_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_sext_i32_i16 s2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i16_sext_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_16 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     BFE_INT T0.X, T0.X, 0.0, literal.x,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.y,
+; EG-NEXT:    16(2.242078e-44), 2(2.802597e-45)
+;
+; CM-LABEL: i16_sext_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 3, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_16 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     BFE_INT * T0.X, T0.X, 0.0, literal.x,
+; CM-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = sext i16 %in to i32
   store i32 %ext, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM: T{{[0-9]\.[XYZW]}}, KC0[2].Z
-; SI: s_load_dword s{{[0-9]}}, s[0:1], 0xb
-; MESA-VI: s_load_dword s{{[0-9]}}, s[0:1], 0x2c
-; HSA-GFX9: s_load_dword s{{[0-9]}}, s[4:5], 0x8
 define amdgpu_kernel void @i32_arg(i32 addrspace(1)* nocapture %out, i32 %in) nounwind {
+; SI-LABEL: i32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T1.X, KC0[2].Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: i32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 2, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T0.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     LSHR * T0.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T1.X, KC0[2].Z,
 entry:
   store i32 %in, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}f32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM: T{{[0-9]\.[XYZW]}}, KC0[2].Z
-; SI: s_load_dword s{{[0-9]}}, s[0:1], 0xb
-; MESA-VI: s_load_dword s{{[0-9]}}, s[0:1], 0x2c
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0x8
 define amdgpu_kernel void @f32_arg(float addrspace(1)* nocapture %out, float %in) nounwind {
+; SI-LABEL: f32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: f32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: f32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: f32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 2, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     LSHR T0.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T1.X, KC0[2].Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: f32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 2, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T0.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     LSHR * T0.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T1.X, KC0[2].Z,
 entry:
   store float %in, float addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v2i8_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-
-; GCN: s_load_dword s
-; GCN-NOT: {{buffer|flat|global}}_load_
 define amdgpu_kernel void @v2i8_arg(<2 x i8> addrspace(1)* %out, <2 x i8> %in) {
+; SI-LABEL: v2i8_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v2i8_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_short v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v2i8_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_short v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v2i8_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @10, KC0[], KC1[]
+; EG-NEXT:    TEX 1 @6
+; EG-NEXT:    ALU 15, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T4.XW, T5.X
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T5.X, T4.X, 41, #3
+; EG-NEXT:     VTX_READ_8 T4.X, T4.X, 40, #3
+; EG-NEXT:    ALU clause starting at 10:
+; EG-NEXT:     MOV * T4.X, 0.0,
+; EG-NEXT:    ALU clause starting at 11:
+; EG-NEXT:     LSHL T0.W, T5.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T4.X, literal.y,
+; EG-NEXT:    8(1.121039e-44), 255(3.573311e-43)
+; EG-NEXT:     AND_INT T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T0.W, PS, literal.x,
+; EG-NEXT:     LSHL * T1.W, PV.W, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; EG-NEXT:     LSHL T4.X, PV.W, PS,
+; EG-NEXT:     LSHL * T4.W, literal.x, PS,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T4.Y, 0.0,
+; EG-NEXT:     MOV * T4.Z, 0.0,
+; EG-NEXT:     LSHR * T5.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v2i8_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 0, @10, KC0[], KC1[]
+; CM-NEXT:    TEX 1 @6
+; CM-NEXT:    ALU 15, @11, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T4.XW, T5.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T5.X, T4.X, 41, #3
+; CM-NEXT:     VTX_READ_8 T4.X, T4.X, 40, #3
+; CM-NEXT:    ALU clause starting at 10:
+; CM-NEXT:     MOV * T4.X, 0.0,
+; CM-NEXT:    ALU clause starting at 11:
+; CM-NEXT:     LSHL T0.Z, T5.X, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T4.X, literal.y, BS:VEC_120/SCL_212
+; CM-NEXT:    8(1.121039e-44), 255(3.573311e-43)
+; CM-NEXT:     AND_INT T1.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, PV.W, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.Z, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; CM-NEXT:     LSHL T4.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T4.W, literal.x, PV.W,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T4.Y, 0.0,
+; CM-NEXT:     MOV * T4.Z, 0.0,
+; CM-NEXT:     LSHR * T5.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <2 x i8> %in, <2 x i8> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}v2i16_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-
-; SI: s_load_dword s{{[0-9]+}}, s[0:1], 0xb
-; MESA-VI: s_load_dword s{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, 0x2c
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, 0x8
 define amdgpu_kernel void @v2i16_arg(<2 x i16> addrspace(1)* %out, <2 x i16> %in) {
+; SI-LABEL: v2i16_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v2i16_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v2i16_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v2i16_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @10, KC0[], KC1[]
+; EG-NEXT:    TEX 1 @6
+; EG-NEXT:    ALU 5, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T4.X, T5.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_16 T5.X, T4.X, 42, #3
+; EG-NEXT:     VTX_READ_16 T4.X, T4.X, 40, #3
+; EG-NEXT:    ALU clause starting at 10:
+; EG-NEXT:     MOV * T4.X, 0.0,
+; EG-NEXT:    ALU clause starting at 11:
+; EG-NEXT:     LSHL T0.W, T5.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T4.X, literal.y,
+; EG-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT T4.X, PV.W, PS,
+; EG-NEXT:     LSHR * T5.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v2i16_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 0, @10, KC0[], KC1[]
+; CM-NEXT:    TEX 1 @6
+; CM-NEXT:    ALU 5, @11, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T4.X, T5.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_16 T5.X, T4.X, 42, #3
+; CM-NEXT:     VTX_READ_16 T4.X, T4.X, 40, #3
+; CM-NEXT:    ALU clause starting at 10:
+; CM-NEXT:     MOV * T4.X, 0.0,
+; CM-NEXT:    ALU clause starting at 11:
+; CM-NEXT:     LSHL T0.Z, T5.X, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T4.X, literal.y, BS:VEC_120/SCL_212
+; CM-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T4.X, PV.Z, PV.W,
+; CM-NEXT:     LSHR * T5.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <2 x i16> %in, <2 x i16> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}v2i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 16
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[2].W
-; SI: s_load_dwordx2 s{{\[[0-9]:[0-9]\]}}, s[0:1], 0xb
-; MESA-VI: s_load_dwordx2 s{{\[[0-9]:[0-9]\]}}, s[0:1], 0x2c
-; HSA-GFX9: s_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x8
 define amdgpu_kernel void @v2i32_arg(<2 x i32> addrspace(1)* nocapture %out, <2 x i32> %in) nounwind {
+; SI-LABEL: v2i32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v2i32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v2i32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v2i32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.Y, KC0[3].X,
+; EG-NEXT:     MOV T0.X, KC0[2].W,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v2i32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     MOV * T0.Y, KC0[3].X,
+; CM-NEXT:     MOV * T0.X, KC0[2].W,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <2 x i32> %in, <2 x i32> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v2f32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 16
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[2].W
-; SI: s_load_dwordx2 s{{\[[0-9]:[0-9]\]}}, s[0:1], 0xb
-; MESA-VI: s_load_dwordx2 s{{\[[0-9]:[0-9]\]}}, s[0:1], 0x2c
-; HSA-GFX9: s_load_dwordx2 s{{\[[0-9]:[0-9]\]}}, s[4:5], 0x8
 define amdgpu_kernel void @v2f32_arg(<2 x float> addrspace(1)* nocapture %out, <2 x float> %in) nounwind {
+; SI-LABEL: v2f32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v2f32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v2f32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v2f32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.Y, KC0[3].X,
+; EG-NEXT:     MOV T0.X, KC0[2].W,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v2f32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     MOV * T0.Y, KC0[3].X,
+; CM-NEXT:     MOV * T0.X, KC0[2].W,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <2 x float> %in, <2 x float> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v3i8_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM-DAG: VTX_READ_8 T{{[0-9]}}.X, T{{[0-9]}}.X, 40
-; EGCM-DAG: VTX_READ_8 T{{[0-9]}}.X, T{{[0-9]}}.X, 41
-; EGCM-DAG: VTX_READ_8 T{{[0-9]}}.X, T{{[0-9]}}.X, 42
-
-; SI: s_load_dword s{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, 0xb
-
-; VI-MESA: s_load_dword s{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, 0x2c
-; VI-HSA: s_load_dword s{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, 0x8
 define amdgpu_kernel void @v3i8_arg(<3 x i8> addrspace(1)* nocapture %out, <3 x i8> %in) nounwind {
+; SI-LABEL: v3i8_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_lshr_b32 s5, s4, 16
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_byte v1, off, s[0:3], 0 offset:2
+; SI-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v3i8_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_lshr_b32 s1, s0, 16
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    s_add_u32 s0, s2, 2
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_addc_u32 s1, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    flat_store_byte v[2:3], v5
+; VI-NEXT:    flat_store_short v[0:1], v4
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v3i8_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_byte_d16_hi v0, v1, s[0:1] offset:2
+; GFX9-NEXT:    global_store_short v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v3i8_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @12, KC0[], KC1[]
+; EG-NEXT:    TEX 2 @6
+; EG-NEXT:    ALU 28, @13, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T4.XW, T7.X
+; EG-NEXT:    MEM_RAT MSKOR T5.XW, T6.X
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T5.X, T4.X, 41, #3
+; EG-NEXT:     VTX_READ_8 T6.X, T4.X, 42, #3
+; EG-NEXT:     VTX_READ_8 T4.X, T4.X, 40, #3
+; EG-NEXT:    ALU clause starting at 12:
+; EG-NEXT:     MOV * T4.X, 0.0,
+; EG-NEXT:    ALU clause starting at 13:
+; EG-NEXT:     LSHL T0.W, T5.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T4.X, literal.y,
+; EG-NEXT:    8(1.121039e-44), 255(3.573311e-43)
+; EG-NEXT:     AND_INT T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T0.W, PS, literal.x,
+; EG-NEXT:     LSHL * T1.W, PV.W, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; EG-NEXT:     LSHL T4.X, PV.W, PS,
+; EG-NEXT:     LSHL * T4.W, literal.x, PS,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T4.Y, 0.0,
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T1.W, PV.W, literal.x,
+; EG-NEXT:     AND_INT * T2.W, T6.X, literal.y,
+; EG-NEXT:    3(4.203895e-45), 255(3.573311e-43)
+; EG-NEXT:     LSHL * T1.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T5.X, T2.W, PV.W,
+; EG-NEXT:     LSHL * T5.W, literal.x, PV.W,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     MOV T5.Y, 0.0,
+; EG-NEXT:     MOV T4.Z, 0.0,
+; EG-NEXT:     MOV * T5.Z, 0.0,
+; EG-NEXT:     LSHR T6.X, T0.W, literal.x,
+; EG-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v3i8_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 0, @12, KC0[], KC1[]
+; CM-NEXT:    TEX 2 @6
+; CM-NEXT:    ALU 29, @13, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T4.XW, T7.X
+; CM-NEXT:    MEM_RAT MSKOR T5.XW, T6.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T5.X, T4.X, 41, #3
+; CM-NEXT:     VTX_READ_8 T6.X, T4.X, 42, #3
+; CM-NEXT:     VTX_READ_8 T4.X, T4.X, 40, #3
+; CM-NEXT:    ALU clause starting at 12:
+; CM-NEXT:     MOV * T4.X, 0.0,
+; CM-NEXT:    ALU clause starting at 13:
+; CM-NEXT:     LSHL T0.Z, T5.X, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T4.X, literal.y, BS:VEC_120/SCL_212
+; CM-NEXT:    8(1.121039e-44), 255(3.573311e-43)
+; CM-NEXT:     AND_INT T1.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, PV.W, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.Z, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; CM-NEXT:     LSHL T4.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T4.W, literal.x, PV.W,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T4.Y, 0.0,
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT * T1.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T6.X, literal.x,
+; CM-NEXT:     LSHL * T1.W, PV.W, literal.y,
+; CM-NEXT:    255(3.573311e-43), 3(4.203895e-45)
+; CM-NEXT:     LSHL T5.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T5.W, literal.x, PV.W,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     MOV T5.Y, 0.0,
+; CM-NEXT:     MOV * T4.Z, 0.0,
+; CM-NEXT:     MOV * T5.Z, 0.0,
+; CM-NEXT:     LSHR * T6.X, T0.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <3 x i8> %in, <3 x i8> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v3i16_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 16
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM-DAG: VTX_READ_16 T{{[0-9]}}.X, T{{[0-9]}}.X, 44
-; EGCM-DAG: VTX_READ_16 T{{[0-9]}}.X, T{{[0-9]}}.X, 46
-; EGCM-DAG: VTX_READ_16 T{{[0-9]}}.X, T{{[0-9]}}.X, 48
-
-; SI: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0xb
-
-; VI-HSA: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x8
-; VI-MESA: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x2c
 define amdgpu_kernel void @v3i16_arg(<3 x i16> addrspace(1)* nocapture %out, <3 x i16> %in) nounwind {
+; SI-LABEL: v3i16_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s5
+; SI-NEXT:    buffer_store_short v0, off, s[0:3], 0 offset:4
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v3i16_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_add_u32 s4, s2, 4
+; VI-NEXT:    s_addc_u32 s5, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v5, s0
+; VI-NEXT:    flat_store_short v[2:3], v4
+; VI-NEXT:    flat_store_dword v[0:1], v5
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v3i16_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    global_store_short v0, v1, s[0:1] offset:4
+; GFX9-NEXT:    global_store_dword v0, v2, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v3i16_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @12, KC0[], KC1[]
+; EG-NEXT:    TEX 2 @6
+; EG-NEXT:    ALU 19, @13, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T6.X, T7.X, 0
+; EG-NEXT:    MEM_RAT MSKOR T5.XW, T8.X
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_16 T6.X, T5.X, 44, #3
+; EG-NEXT:     VTX_READ_16 T7.X, T5.X, 46, #3
+; EG-NEXT:     VTX_READ_16 T5.X, T5.X, 48, #3
+; EG-NEXT:    ALU clause starting at 12:
+; EG-NEXT:     MOV * T5.X, 0.0,
+; EG-NEXT:    ALU clause starting at 13:
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T1.W, PV.W, literal.x,
+; EG-NEXT:     AND_INT * T2.W, T5.X, literal.y,
+; EG-NEXT:    3(4.203895e-45), 65535(9.183409e-41)
+; EG-NEXT:     LSHL * T1.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T5.X, T2.W, PV.W,
+; EG-NEXT:     LSHL * T5.W, literal.x, PV.W,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T5.Y, 0.0,
+; EG-NEXT:     MOV * T5.Z, 0.0,
+; EG-NEXT:     LSHR T8.X, T0.W, literal.x,
+; EG-NEXT:     LSHL T0.W, T7.X, literal.y,
+; EG-NEXT:     AND_INT * T1.W, T6.X, literal.z,
+; EG-NEXT:    2(2.802597e-45), 16(2.242078e-44)
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     OR_INT T6.X, PV.W, PS,
+; EG-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v3i16_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 0, @12, KC0[], KC1[]
+; CM-NEXT:    TEX 2 @6
+; CM-NEXT:    ALU 19, @13, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T5.XW, T8.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T6.X, T7.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_16 T6.X, T5.X, 44, #3
+; CM-NEXT:     VTX_READ_16 T7.X, T5.X, 46, #3
+; CM-NEXT:     VTX_READ_16 T5.X, T5.X, 48, #3
+; CM-NEXT:    ALU clause starting at 12:
+; CM-NEXT:     MOV * T5.X, 0.0,
+; CM-NEXT:    ALU clause starting at 13:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT * T1.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T5.X, literal.x,
+; CM-NEXT:     LSHL * T1.W, PV.W, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; CM-NEXT:     LSHL T5.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T5.W, literal.x, PV.W,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T5.Y, 0.0,
+; CM-NEXT:     MOV * T5.Z, 0.0,
+; CM-NEXT:     LSHL T0.Z, T7.X, literal.x,
+; CM-NEXT:     AND_INT * T1.W, T6.X, literal.y, BS:VEC_120/SCL_212
+; CM-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T6.X, PV.Z, PV.W,
+; CM-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T8.X, T0.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <3 x i16> %in, <3 x i16> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v3i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 32
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].W
-; SI: s_load_dwordx4 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0xd
-; MESA-VI: s_load_dwordx4 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x34
-; HSA-GFX9: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x10
 define amdgpu_kernel void @v3i32_arg(<3 x i32> addrspace(1)* nocapture %out, <3 x i32> %in) nounwind {
+; SI-LABEL: v3i32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s6
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:8
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v3i32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v3, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s5
+; VI-NEXT:    flat_store_dwordx3 v[3:4], v[0:2]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v3i32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    global_store_dwordx3 v3, v[0:2], s[6:7]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v3i32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 8, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.X, T2.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.Y, KC0[3].Z,
+; EG-NEXT:     MOV T0.X, KC0[3].Y,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T2.X, PV.W, literal.x,
+; EG-NEXT:     MOV * T3.X, KC0[3].W,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v3i32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 8, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2, T3.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T0.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T0.X, PV.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV T1.X, KC0[3].W,
+; CM-NEXT:     MOV * T2.Y, KC0[3].Z,
+; CM-NEXT:     MOV * T2.X, KC0[3].Y,
+; CM-NEXT:     LSHR * T3.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <3 x i32> %in, <3 x i32> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v3f32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 32
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].W
-; SI: s_load_dwordx4 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0xd
-; MESA-VI: s_load_dwordx4 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x34
-; HSA-GFX9: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x10
 define amdgpu_kernel void @v3f32_arg(<3 x float> addrspace(1)* nocapture %out, <3 x float> %in) nounwind {
+; SI-LABEL: v3f32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s6
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:8
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v3f32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v3, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s5
+; VI-NEXT:    flat_store_dwordx3 v[3:4], v[0:2]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v3f32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    global_store_dwordx3 v3, v[0:2], s[6:7]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v3f32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 8, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.X, T2.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.Y, KC0[3].Z,
+; EG-NEXT:     MOV T0.X, KC0[3].Y,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T2.X, PV.W, literal.x,
+; EG-NEXT:     MOV * T3.X, KC0[3].W,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v3f32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 8, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2, T3.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T0.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T0.X, PV.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV T1.X, KC0[3].W,
+; CM-NEXT:     MOV * T2.Y, KC0[3].Z,
+; CM-NEXT:     MOV * T2.X, KC0[3].Y,
+; CM-NEXT:     LSHR * T3.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <3 x float> %in, <3 x float> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v4i8_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-
-; GCN-DAG: s_load_dwordx2 s
-; GCN-DAG: s_load_dword s
 define amdgpu_kernel void @v4i8_arg(<4 x i8> addrspace(1)* %out, <4 x i8> %in) {
+; SI-LABEL: v4i8_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v4i8_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v4i8_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v4i8_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @14, KC0[], KC1[]
+; EG-NEXT:    TEX 3 @6
+; EG-NEXT:    ALU 15, @15, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T4.X, T5.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T5.X, T4.X, 42, #3
+; EG-NEXT:     VTX_READ_8 T6.X, T4.X, 40, #3
+; EG-NEXT:     VTX_READ_8 T7.X, T4.X, 43, #3
+; EG-NEXT:     VTX_READ_8 T4.X, T4.X, 41, #3
+; EG-NEXT:    ALU clause starting at 14:
+; EG-NEXT:     MOV * T4.X, 0.0,
+; EG-NEXT:    ALU clause starting at 15:
+; EG-NEXT:     AND_INT * T0.W, T5.X, literal.x,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T0.Z, T4.X, literal.x,
+; EG-NEXT:     LSHL T0.W, PV.W, literal.y,
+; EG-NEXT:     LSHL * T1.W, T7.X, literal.z,
+; EG-NEXT:    255(3.573311e-43), 16(2.242078e-44)
+; EG-NEXT:    24(3.363116e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT T0.W, PS, PV.W,
+; EG-NEXT:     LSHL * T1.W, PV.Z, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT T0.W, PV.W, PS,
+; EG-NEXT:     AND_INT * T1.W, T6.X, literal.x,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     OR_INT T4.X, PV.W, PS,
+; EG-NEXT:     LSHR * T5.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v4i8_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 0, @14, KC0[], KC1[]
+; CM-NEXT:    TEX 3 @6
+; CM-NEXT:    ALU 15, @15, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T4.X, T5.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T5.X, T4.X, 42, #3
+; CM-NEXT:     VTX_READ_8 T6.X, T4.X, 40, #3
+; CM-NEXT:     VTX_READ_8 T7.X, T4.X, 43, #3
+; CM-NEXT:     VTX_READ_8 T4.X, T4.X, 41, #3
+; CM-NEXT:    ALU clause starting at 14:
+; CM-NEXT:     MOV * T4.X, 0.0,
+; CM-NEXT:    ALU clause starting at 15:
+; CM-NEXT:     AND_INT * T0.W, T5.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Y, T4.X, literal.x,
+; CM-NEXT:     LSHL T0.Z, PV.W, literal.y,
+; CM-NEXT:     LSHL * T0.W, T7.X, literal.z, BS:VEC_120/SCL_212
+; CM-NEXT:    255(3.573311e-43), 16(2.242078e-44)
+; CM-NEXT:    24(3.363116e-44), 0(0.000000e+00)
+; CM-NEXT:     OR_INT T0.Z, PV.W, PV.Z,
+; CM-NEXT:     LSHL * T0.W, PV.Y, literal.x,
+; CM-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; CM-NEXT:     OR_INT T0.Z, PV.Z, PV.W,
+; CM-NEXT:     AND_INT * T0.W, T6.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     OR_INT * T4.X, PV.Z, PV.W,
+; CM-NEXT:     LSHR * T5.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <4 x i8> %in, <4 x i8> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}v4i16_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 16
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-
-; SI-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0xb
-; SI-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x9
-
-; MESA-VI-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x24
-; MESA-VI-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x2c
-
-
-; MESA-VI-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x24
-; MESA-VI-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x2c
-
-; HSA-GFX9-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x0
-; HSA-GFX9-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x8
 define amdgpu_kernel void @v4i16_arg(<4 x i16> addrspace(1)* %out, <4 x i16> %in) {
+; SI-LABEL: v4i16_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v4i16_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v4i16_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v4i16_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 1, @20, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @12
+; EG-NEXT:    ALU 5, @22, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @14
+; EG-NEXT:    ALU 5, @28, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @16
+; EG-NEXT:    ALU 5, @34, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @18
+; EG-NEXT:    ALU 7, @40, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.XY, T6.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 12:
+; EG-NEXT:     VTX_READ_16 T6.X, T5.X, 50, #3
+; EG-NEXT:    Fetch clause starting at 14:
+; EG-NEXT:     VTX_READ_16 T6.X, T5.X, 48, #3
+; EG-NEXT:    Fetch clause starting at 16:
+; EG-NEXT:     VTX_READ_16 T6.X, T5.X, 46, #3
+; EG-NEXT:    Fetch clause starting at 18:
+; EG-NEXT:     VTX_READ_16 T5.X, T5.X, 44, #3
+; EG-NEXT:    ALU clause starting at 20:
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:     MOV * T5.X, 0.0,
+; EG-NEXT:    ALU clause starting at 22:
+; EG-NEXT:     LSHL T0.W, T6.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PS, PV.W,
+; EG-NEXT:     MOV * T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, PV.X,
+; EG-NEXT:    ALU clause starting at 28:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T6.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 34:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T6.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV * T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, PV.X,
+; EG-NEXT:    ALU clause starting at 40:
+; EG-NEXT:     LSHR T6.X, KC0[2].Y, literal.x,
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.y,
+; EG-NEXT:     AND_INT * T1.W, T5.X, literal.z,
+; EG-NEXT:    2(2.802597e-45), -65536(nan)
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T5.X, PV.W, PS,
+; EG-NEXT:     MOV T2.X, PV.X,
+; EG-NEXT:     MOV * T5.Y, T3.X,
+;
+; CM-LABEL: v4i16_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 1, @20, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @12
+; CM-NEXT:    ALU 5, @22, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @14
+; CM-NEXT:    ALU 5, @28, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @16
+; CM-NEXT:    ALU 5, @34, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @18
+; CM-NEXT:    ALU 7, @40, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T5, T6.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 12:
+; CM-NEXT:     VTX_READ_16 T6.X, T5.X, 50, #3
+; CM-NEXT:    Fetch clause starting at 14:
+; CM-NEXT:     VTX_READ_16 T6.X, T5.X, 48, #3
+; CM-NEXT:    Fetch clause starting at 16:
+; CM-NEXT:     VTX_READ_16 T6.X, T5.X, 46, #3
+; CM-NEXT:    Fetch clause starting at 18:
+; CM-NEXT:     VTX_READ_16 T5.X, T5.X, 44, #3
+; CM-NEXT:    ALU clause starting at 20:
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:     MOV * T5.X, 0.0,
+; CM-NEXT:    ALU clause starting at 22:
+; CM-NEXT:     LSHL T0.Z, T6.X, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T0.Y, literal.y,
+; CM-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.W, PV.Z,
+; CM-NEXT:     MOV * T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, PV.X,
+; CM-NEXT:    ALU clause starting at 28:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T6.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 34:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T6.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV * T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, PV.X,
+; CM-NEXT:    ALU clause starting at 40:
+; CM-NEXT:     LSHR T6.X, KC0[2].Y, literal.x,
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.y,
+; CM-NEXT:     AND_INT * T0.W, T5.X, literal.z,
+; CM-NEXT:    2(2.802597e-45), -65536(nan)
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     OR_INT * T5.X, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.X,
+; CM-NEXT:     MOV * T5.Y, T3.X,
 entry:
   store <4 x i16> %in, <4 x i16> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}v4i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 32
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].X
-
-; SI: s_load_dwordx4 s{{\[[0-9]:[0-9]\]}}, s[0:1], 0xd
-; MESA-VI: s_load_dwordx4 s{{\[[0-9]:[0-9]\]}}, s[0:1], 0x34
-; HSA-GFX9: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x10
 define amdgpu_kernel void @v4i32_arg(<4 x i32> addrspace(1)* nocapture %out, <4 x i32> %in) nounwind {
+; SI-LABEL: v4i32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v4i32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v4i32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v4i32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 5, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.W, KC0[4].X,
+; EG-NEXT:     MOV * T0.Z, KC0[3].W,
+; EG-NEXT:     MOV * T0.Y, KC0[3].Z,
+; EG-NEXT:     MOV T0.X, KC0[3].Y,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v4i32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 5, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     MOV * T0.W, KC0[4].X,
+; CM-NEXT:     MOV * T0.Z, KC0[3].W,
+; CM-NEXT:     MOV * T0.Y, KC0[3].Z,
+; CM-NEXT:     MOV * T0.X, KC0[3].Y,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <4 x i32> %in, <4 x i32> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v4f32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 32
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[3].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].X
-; SI: s_load_dwordx4 s{{\[[0-9]:[0-9]\]}}, s[0:1], 0xd
-; MESA-VI: s_load_dwordx4 s{{\[[0-9]:[0-9]\]}}, s[0:1], 0x34
-; HSA-GFX9: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x10
 define amdgpu_kernel void @v4f32_arg(<4 x float> addrspace(1)* nocapture %out, <4 x float> %in) nounwind {
+; SI-LABEL: v4f32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v4f32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v4f32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v4f32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 5, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.W, KC0[4].X,
+; EG-NEXT:     MOV * T0.Z, KC0[3].W,
+; EG-NEXT:     MOV * T0.Y, KC0[3].Z,
+; EG-NEXT:     MOV T0.X, KC0[3].Y,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v4f32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 5, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     MOV * T0.W, KC0[4].X,
+; CM-NEXT:     MOV * T0.Z, KC0[3].W,
+; CM-NEXT:     MOV * T0.Y, KC0[3].Z,
+; CM-NEXT:     MOV * T0.X, KC0[3].Y,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <4 x float> %in, <4 x float> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v5i8_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 16
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM-DAG: VTX_READ_8 T{{[0-9]}}.X, T{{[0-9]}}.X, 46
-; EGCM-DAG: VTX_READ_8 T{{[0-9]}}.X, T{{[0-9]}}.X, 46
-; EGCM-DAG: VTX_READ_8 T{{[0-9]}}.X, T{{[0-9]}}.X, 46
-
-; SI: s_load_dword s{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, 0xb
-
-; VI-MESA: s_load_dword s{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, 0x2c
-; VI-HSA: s_load_dword s{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, 0x8
 define amdgpu_kernel void @v5i8_arg(<5 x i8> addrspace(1)* nocapture %out, <5 x i8> %in) nounwind {
+; SI-LABEL: v5i8_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xc
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
+; SI-NEXT:    s_load_dword s0, s[0:1], 0xb
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s2
+; SI-NEXT:    buffer_store_byte v0, off, s[4:7], 0 offset:4
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v5i8_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; VI-NEXT:    s_load_dword s1, s[0:1], 0x30
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_add_u32 s0, s2, 4
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s1
+; VI-NEXT:    s_addc_u32 s1, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_byte v[2:3], v4
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v5i8_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    global_store_byte v0, v1, s[0:1] offset:4
+; GFX9-NEXT:    global_store_dword v0, v2, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v5i8_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @16, KC0[], KC1[]
+; EG-NEXT:    TEX 4 @6
+; EG-NEXT:    ALU 28, @17, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T5.XW, T8.X
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T6.X, T7.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T6.X, T5.X, 44, #3
+; EG-NEXT:     VTX_READ_8 T7.X, T5.X, 47, #3
+; EG-NEXT:     VTX_READ_8 T8.X, T5.X, 45, #3
+; EG-NEXT:     VTX_READ_8 T9.X, T5.X, 46, #3
+; EG-NEXT:     VTX_READ_8 T5.X, T5.X, 48, #3
+; EG-NEXT:    ALU clause starting at 16:
+; EG-NEXT:     MOV * T5.X, 0.0,
+; EG-NEXT:    ALU clause starting at 17:
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T1.W, PV.W, literal.x,
+; EG-NEXT:     AND_INT * T2.W, T5.X, literal.y,
+; EG-NEXT:    3(4.203895e-45), 255(3.573311e-43)
+; EG-NEXT:     LSHL * T1.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T5.X, T2.W, PV.W,
+; EG-NEXT:     LSHL * T5.W, literal.x, PV.W,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     MOV T5.Y, 0.0,
+; EG-NEXT:     MOV T5.Z, 0.0,
+; EG-NEXT:     AND_INT T1.W, T9.X, literal.x,
+; EG-NEXT:     AND_INT * T0.Z, T8.X, literal.x,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     LSHL T1.W, PV.W, literal.x,
+; EG-NEXT:     LSHL * T2.W, T7.X, literal.y,
+; EG-NEXT:    16(2.242078e-44), 24(3.363116e-44)
+; EG-NEXT:     OR_INT T1.W, PS, PV.W,
+; EG-NEXT:     LSHL * T2.W, T0.Z, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT T1.W, PV.W, PS,
+; EG-NEXT:     AND_INT * T2.W, T6.X, literal.x,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     OR_INT T6.X, PV.W, PS,
+; EG-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T8.X, T0.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v5i8_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 0, @16, KC0[], KC1[]
+; CM-NEXT:    TEX 4 @6
+; CM-NEXT:    ALU 28, @17, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T6.X, T8.X
+; CM-NEXT:    MEM_RAT MSKOR T5.XW, T7.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T6.X, T5.X, 44, #3
+; CM-NEXT:     VTX_READ_8 T7.X, T5.X, 47, #3
+; CM-NEXT:     VTX_READ_8 T8.X, T5.X, 45, #3
+; CM-NEXT:     VTX_READ_8 T9.X, T5.X, 46, #3
+; CM-NEXT:     VTX_READ_8 T5.X, T5.X, 48, #3
+; CM-NEXT:    ALU clause starting at 16:
+; CM-NEXT:     MOV * T5.X, 0.0,
+; CM-NEXT:    ALU clause starting at 17:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT * T1.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T5.X, literal.x,
+; CM-NEXT:     LSHL * T1.W, PV.W, literal.y,
+; CM-NEXT:    255(3.573311e-43), 3(4.203895e-45)
+; CM-NEXT:     LSHL T5.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T5.W, literal.x, PV.W,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     MOV T5.Y, 0.0,
+; CM-NEXT:     MOV T5.Z, 0.0,
+; CM-NEXT:     AND_INT * T1.W, T9.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Y, T8.X, literal.x,
+; CM-NEXT:     LSHL T0.Z, PV.W, literal.y,
+; CM-NEXT:     LSHL * T1.W, T7.X, literal.z, BS:VEC_120/SCL_212
+; CM-NEXT:    255(3.573311e-43), 16(2.242078e-44)
+; CM-NEXT:    24(3.363116e-44), 0(0.000000e+00)
+; CM-NEXT:     OR_INT T0.Z, PV.W, PV.Z,
+; CM-NEXT:     LSHL * T1.W, PV.Y, literal.x,
+; CM-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T7.X, T0.W, literal.x,
+; CM-NEXT:     OR_INT T0.Z, PV.Z, PV.W,
+; CM-NEXT:     AND_INT * T0.W, T6.X, literal.y,
+; CM-NEXT:    2(2.802597e-45), 255(3.573311e-43)
+; CM-NEXT:     OR_INT * T6.X, PV.Z, PV.W,
+; CM-NEXT:     LSHR * T8.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <5 x i8> %in, <5 x i8> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v5i16_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 32
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; EGCM-DAG: VTX_READ_16 T{{[0-9]}}.X, T{{[0-9]}}.X, 58
-; EGCM-DAG: VTX_READ_16 T{{[0-9]}}.X, T{{[0-9]}}.X, 58
-; EGCM-DAG: VTX_READ_16 T{{[0-9]}}.X, T{{[0-9]}}.X, 58
-
-; SI: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0xd
-
-; VI-HSA: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x8
-; VI-MESA: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x2c
 define amdgpu_kernel void @v5i16_arg(<5 x i16> addrspace(1)* nocapture %out, <5 x i16> %in) nounwind {
+; SI-LABEL: v5i16_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xf
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s2
+; SI-NEXT:    buffer_store_short v0, off, s[4:7], 0 offset:8
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    v_mov_b32_e32 v1, s1
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v5i16_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x34
+; VI-NEXT:    s_load_dword s1, s[0:1], 0x3c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_add_u32 s0, s2, 8
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v4, s1
+; VI-NEXT:    s_addc_u32 s1, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_short v[2:3], v4
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v3, s5
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v5i16_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x10
+; GFX9-NEXT:    s_load_dword s6, s[4:5], 0x18
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s6
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_short v2, v3, s[0:1] offset:8
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v5i16_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @20, KC0[], KC1[]
+; EG-NEXT:    TEX 4 @10
+; EG-NEXT:    ALU 65, @21, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T5.XW, T9.X
+; EG-NEXT:    MEM_RAT MSKOR T4.XW, T7.X
+; EG-NEXT:    MEM_RAT MSKOR T3.XW, T2.X
+; EG-NEXT:    MEM_RAT MSKOR T6.XW, T1.X
+; EG-NEXT:    MEM_RAT MSKOR T8.XW, T0.X
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 10:
+; EG-NEXT:     VTX_READ_16 T1.X, T0.X, 58, #3
+; EG-NEXT:     VTX_READ_16 T2.X, T0.X, 56, #3
+; EG-NEXT:     VTX_READ_16 T3.X, T0.X, 54, #3
+; EG-NEXT:     VTX_READ_16 T4.X, T0.X, 52, #3
+; EG-NEXT:     VTX_READ_16 T0.X, T0.X, 60, #3
+; EG-NEXT:    ALU clause starting at 20:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 21:
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T1.W, PV.W, literal.x,
+; EG-NEXT:     AND_INT * T2.W, T0.X, literal.y,
+; EG-NEXT:    3(4.203895e-45), 65535(9.183409e-41)
+; EG-NEXT:     LSHL * T1.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T5.X, T2.W, PV.W,
+; EG-NEXT:     LSHL * T5.W, literal.x, PV.W,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T5.Y, 0.0,
+; EG-NEXT:     AND_INT T1.W, KC0[2].Y, literal.x,
+; EG-NEXT:     AND_INT * T2.W, T4.X, literal.y,
+; EG-NEXT:    3(4.203895e-45), 65535(9.183409e-41)
+; EG-NEXT:     LSHL * T1.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T4.X, T2.W, PV.W,
+; EG-NEXT:     LSHL * T4.W, literal.x, PV.W,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T4.Y, 0.0,
+; EG-NEXT:     ADD_INT * T1.W, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T2.W, PV.W, literal.x,
+; EG-NEXT:     AND_INT * T3.W, T3.X, literal.y,
+; EG-NEXT:    3(4.203895e-45), 65535(9.183409e-41)
+; EG-NEXT:     LSHL * T2.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.X, T3.W, PV.W,
+; EG-NEXT:     LSHL * T3.W, literal.x, PV.W,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T3.Y, 0.0,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T6.W, PV.W, literal.x,
+; EG-NEXT:     AND_INT * T7.W, T2.X, literal.y,
+; EG-NEXT:    3(4.203895e-45), 65535(9.183409e-41)
+; EG-NEXT:     LSHL * T6.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T6.X, T7.W, PV.W,
+; EG-NEXT:     LSHL * T6.W, literal.x, PV.W,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T6.Y, 0.0,
+; EG-NEXT:     ADD_INT * T7.W, KC0[2].Y, literal.x,
+; EG-NEXT:    6(8.407791e-45), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T8.W, PV.W, literal.x,
+; EG-NEXT:     AND_INT * T9.W, T1.X, literal.y,
+; EG-NEXT:    3(4.203895e-45), 65535(9.183409e-41)
+; EG-NEXT:     LSHL * T8.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T8.X, T9.W, PV.W,
+; EG-NEXT:     LSHL * T8.W, literal.x, PV.W,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T8.Y, 0.0,
+; EG-NEXT:     MOV T5.Z, 0.0,
+; EG-NEXT:     MOV * T4.Z, 0.0,
+; EG-NEXT:     MOV T3.Z, 0.0,
+; EG-NEXT:     MOV * T6.Z, 0.0,
+; EG-NEXT:     MOV * T8.Z, 0.0,
+; EG-NEXT:     LSHR T0.X, T7.W, literal.x,
+; EG-NEXT:     LSHR * T1.X, T2.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHR T2.X, T1.W, literal.x,
+; EG-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T9.X, T0.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v5i16_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 0, @20, KC0[], KC1[]
+; CM-NEXT:    TEX 4 @10
+; CM-NEXT:    ALU 67, @21, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T5.XW, T9.X
+; CM-NEXT:    MEM_RAT MSKOR T4.XW, T7.X
+; CM-NEXT:    MEM_RAT MSKOR T3.XW, T2.X
+; CM-NEXT:    MEM_RAT MSKOR T6.XW, T1.X
+; CM-NEXT:    MEM_RAT MSKOR T8.XW, T0.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 10:
+; CM-NEXT:     VTX_READ_16 T1.X, T0.X, 58, #3
+; CM-NEXT:     VTX_READ_16 T2.X, T0.X, 56, #3
+; CM-NEXT:     VTX_READ_16 T3.X, T0.X, 54, #3
+; CM-NEXT:     VTX_READ_16 T4.X, T0.X, 52, #3
+; CM-NEXT:     VTX_READ_16 T0.X, T0.X, 60, #3
+; CM-NEXT:    ALU clause starting at 20:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 21:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; CM-NEXT:     AND_INT * T1.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.X, literal.x,
+; CM-NEXT:     LSHL * T1.W, PV.W, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; CM-NEXT:     LSHL T5.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T5.W, literal.x, PV.W,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T5.Y, 0.0,
+; CM-NEXT:     AND_INT * T1.W, KC0[2].Y, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T4.X, literal.x,
+; CM-NEXT:     LSHL * T1.W, PV.W, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; CM-NEXT:     LSHL T4.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T4.W, literal.x, PV.W,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T4.Y, 0.0,
+; CM-NEXT:     ADD_INT * T1.W, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT * T2.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T3.X, literal.x,
+; CM-NEXT:     LSHL * T2.W, PV.W, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; CM-NEXT:     LSHL T3.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T3.W, literal.x, PV.W,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T3.Y, 0.0,
+; CM-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; CM-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT * T6.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T2.X, literal.x,
+; CM-NEXT:     LSHL * T6.W, PV.W, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; CM-NEXT:     LSHL T6.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T6.W, literal.x, PV.W,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T6.Y, 0.0,
+; CM-NEXT:     ADD_INT * T7.W, KC0[2].Y, literal.x,
+; CM-NEXT:    6(8.407791e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT * T8.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T1.X, literal.x,
+; CM-NEXT:     LSHL * T8.W, PV.W, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 3(4.203895e-45)
+; CM-NEXT:     LSHL T8.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T8.W, literal.x, PV.W,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T8.Y, 0.0,
+; CM-NEXT:     MOV * T5.Z, 0.0,
+; CM-NEXT:     MOV * T4.Z, 0.0,
+; CM-NEXT:     MOV * T3.Z, 0.0,
+; CM-NEXT:     MOV * T6.Z, 0.0,
+; CM-NEXT:     MOV * T8.Z, 0.0,
+; CM-NEXT:     LSHR * T0.X, T7.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T1.X, T2.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T2.X, T1.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T9.X, T0.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <5 x i16> %in, <5 x i16> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v5i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 64
-; HSA-GFX9: kernarg_segment_alignment = 5
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].W
-; SI: s_load_dwordx4 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x11
-; MESA-VI: s_load_dwordx4 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x44
-; HSA-GFX9: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x20
 define amdgpu_kernel void @v5i32_arg(<5 x i32> addrspace(1)* nocapture %out, <5 x i32> %in) nounwind {
+; SI-LABEL: v5i32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s8, s[0:1], 0x15
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x11
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    v_mov_b32_e32 v1, s1
+; SI-NEXT:    v_mov_b32_e32 v2, s2
+; SI-NEXT:    v_mov_b32_e32 v3, s3
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v5i32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x44
+; VI-NEXT:    s_load_dword s1, s[0:1], 0x54
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_add_u32 s0, s2, 16
+; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s1
+; VI-NEXT:    s_addc_u32 s1, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v5i32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x20
+; GFX9-NEXT:    s_load_dword s8, s[4:5], 0x30
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v5, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dword v4, v5, s[6:7] offset:16
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v5i32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 10, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.X, T2.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.W, KC0[5].X,
+; EG-NEXT:     MOV * T0.Z, KC0[4].W,
+; EG-NEXT:     MOV * T0.Y, KC0[4].Z,
+; EG-NEXT:     MOV T0.X, KC0[4].Y,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     ADD_INT * T1.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T2.X, PV.W, literal.x,
+; EG-NEXT:     MOV * T3.X, KC0[5].Y,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v5i32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 10, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T3.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     ADD_INT T0.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T0.W, KC0[5].X,
+; CM-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T1.X, PV.Z, literal.x,
+; CM-NEXT:     MOV * T0.Z, KC0[4].W,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV T2.X, KC0[5].Y,
+; CM-NEXT:     MOV * T0.Y, KC0[4].Z,
+; CM-NEXT:     MOV * T0.X, KC0[4].Y,
+; CM-NEXT:     LSHR * T3.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <5 x i32> %in, <5 x i32> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v5f32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 64
-; HSA-GFX9: kernarg_segment_alignment = 5
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].W
-; SI: s_load_dwordx4 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x11
-; MESA-VI: s_load_dwordx4 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x44
-; HSA-GFX9: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x20
 define amdgpu_kernel void @v5f32_arg(<5 x float> addrspace(1)* nocapture %out, <5 x float> %in) nounwind {
+; SI-LABEL: v5f32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s8, s[0:1], 0x15
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x11
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    v_mov_b32_e32 v1, s1
+; SI-NEXT:    v_mov_b32_e32 v2, s2
+; SI-NEXT:    v_mov_b32_e32 v3, s3
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v5f32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x44
+; VI-NEXT:    s_load_dword s1, s[0:1], 0x54
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_add_u32 s0, s2, 16
+; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    s_addc_u32 s1, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s1
+; VI-NEXT:    v_mov_b32_e32 v1, s0
+; VI-NEXT:    flat_store_dword v[1:2], v3
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v5f32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x20
+; GFX9-NEXT:    s_load_dword s8, s[4:5], 0x30
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    global_store_dword v4, v0, s[6:7] offset:16
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v5f32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 10, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.X, T2.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.W, KC0[5].X,
+; EG-NEXT:     MOV * T0.Z, KC0[4].W,
+; EG-NEXT:     MOV * T0.Y, KC0[4].Z,
+; EG-NEXT:     MOV T0.X, KC0[4].Y,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     ADD_INT * T1.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T2.X, PV.W, literal.x,
+; EG-NEXT:     MOV * T3.X, KC0[5].Y,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v5f32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 10, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T3.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     ADD_INT T0.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T0.W, KC0[5].X,
+; CM-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T1.X, PV.Z, literal.x,
+; CM-NEXT:     MOV * T0.Z, KC0[4].W,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV T2.X, KC0[5].Y,
+; CM-NEXT:     MOV * T0.Y, KC0[4].Z,
+; CM-NEXT:     MOV * T0.X, KC0[4].Y,
+; CM-NEXT:     LSHR * T3.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <5 x float> %in, <5 x float> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v5i64_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 128
-; HSA-GFX9: kernarg_segment_alignment = 6
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].Z
-; SI-DAG: s_load_dwordx8 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x19
-; SI-DAG: s_load_dwordx2 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x21
-; MESA-VI-DAG: s_load_dwordx8 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x64
-; MESA-VI-DAG: s_load_dwordx2 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x84
-; HSA-GFX9-DAG: s_load_dwordx8 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x40
-; HSA-GFX9-DAG: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x60
 define amdgpu_kernel void @v5i64_arg(<5 x i64> addrspace(1)* nocapture %out, <5 x i64> %in) nounwind {
+; SI-LABEL: v5i64_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x19
+; SI-NEXT:    s_load_dwordx2 s[12:13], s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x21
+; SI-NEXT:    s_mov_b32 s15, 0xf000
+; SI-NEXT:    s_mov_b32 s14, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    v_mov_b32_e32 v2, s10
+; SI-NEXT:    v_mov_b32_e32 v3, s11
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[12:15], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[12:15], 0
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    v_mov_b32_e32 v1, s1
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[12:15], 0 offset:32
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v5i64_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x64
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x84
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s8
+; VI-NEXT:    s_add_u32 s8, s2, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s9
+; VI-NEXT:    s_addc_u32 s9, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s8
+; VI-NEXT:    v_mov_b32_e32 v2, s10
+; VI-NEXT:    v_mov_b32_e32 v3, s11
+; VI-NEXT:    v_mov_b32_e32 v5, s9
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    s_add_u32 s2, s2, 32
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_addc_u32 s3, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v5i64_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x40
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x60
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-NEXT:    v_mov_b32_e32 v2, s14
+; GFX9-NEXT:    v_mov_b32_e32 v3, s15
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7] offset:16
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    v_mov_b32_e32 v2, s10
+; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_store_dwordx2 v4, v[0:1], s[6:7] offset:32
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v5i64_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 18, @6, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.XY, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.XYZW, T3.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T2.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 6:
+; EG-NEXT:     MOV * T0.W, KC0[7].X,
+; EG-NEXT:     MOV * T0.Z, KC0[6].W,
+; EG-NEXT:     MOV T0.Y, KC0[6].Z,
+; EG-NEXT:     MOV * T1.W, KC0[8].X,
+; EG-NEXT:     MOV T0.X, KC0[6].Y,
+; EG-NEXT:     MOV * T1.Z, KC0[7].W,
+; EG-NEXT:     LSHR T2.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T1.Y, KC0[7].Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T1.X, KC0[7].Y,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T3.X, PV.W, literal.x,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.y,
+; EG-NEXT:    2(2.802597e-45), 32(4.484155e-44)
+; EG-NEXT:     LSHR T4.X, PV.W, literal.x,
+; EG-NEXT:     MOV T5.Y, KC0[8].Z,
+; EG-NEXT:     MOV * T5.X, KC0[8].Y,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v5i64_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 18, @6, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2, T5.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T4.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1, T3.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 6:
+; CM-NEXT:     MOV * T0.W, KC0[8].X,
+; CM-NEXT:     MOV T1.Y, KC0[8].Z,
+; CM-NEXT:     MOV * T0.Z, KC0[7].W,
+; CM-NEXT:     MOV T1.X, KC0[8].Y,
+; CM-NEXT:     MOV * T0.Y, KC0[7].Z,
+; CM-NEXT:     MOV T0.X, KC0[7].Y,
+; CM-NEXT:     ADD_INT T1.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T2.W, KC0[7].X,
+; CM-NEXT:    32(4.484155e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T3.X, PV.Z, literal.x,
+; CM-NEXT:     MOV T2.Z, KC0[6].W,
+; CM-NEXT:     ADD_INT * T1.W, KC0[2].Y, literal.y,
+; CM-NEXT:    2(2.802597e-45), 16(2.242078e-44)
+; CM-NEXT:     LSHR T4.X, PV.W, literal.x,
+; CM-NEXT:     MOV * T2.Y, KC0[6].Z,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T2.X, KC0[6].Y,
+; CM-NEXT:     LSHR * T5.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <5 x i64> %in, <5 x i64> addrspace(1)* %out, align 8
   ret void
 }
 
-; FUNC-LABEL: {{^}}v5f64_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 128
-; HSA-GFX9: kernarg_segment_alignment = 6
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].Z
-; SI-DAG: s_load_dwordx8 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x19
-; SI-DAG: s_load_dwordx2 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x21
-; MESA-VI-DAG: s_load_dwordx8 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x64
-; MESA-VI-DAG: s_load_dwordx2 s{{\[[0-9]:[0-9]+\]}}, s[0:1], 0x84
-; HSA-GFX9-DAG: s_load_dwordx8 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x40
-; HSA-GFX9-DAG: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x60
 define amdgpu_kernel void @v5f64_arg(<5 x double> addrspace(1)* nocapture %out, <5 x double> %in) nounwind {
+; SI-LABEL: v5f64_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x19
+; SI-NEXT:    s_load_dwordx2 s[12:13], s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x21
+; SI-NEXT:    s_mov_b32 s15, 0xf000
+; SI-NEXT:    s_mov_b32 s14, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    v_mov_b32_e32 v2, s10
+; SI-NEXT:    v_mov_b32_e32 v3, s11
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[12:15], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[12:15], 0
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    v_mov_b32_e32 v1, s1
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[12:15], 0 offset:32
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v5f64_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x64
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x84
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s8
+; VI-NEXT:    s_add_u32 s8, s2, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s9
+; VI-NEXT:    s_addc_u32 s9, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s8
+; VI-NEXT:    v_mov_b32_e32 v2, s10
+; VI-NEXT:    v_mov_b32_e32 v3, s11
+; VI-NEXT:    v_mov_b32_e32 v5, s9
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    s_add_u32 s2, s2, 32
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_addc_u32 s3, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v5f64_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x40
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x60
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-NEXT:    v_mov_b32_e32 v2, s14
+; GFX9-NEXT:    v_mov_b32_e32 v3, s15
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7] offset:16
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    v_mov_b32_e32 v2, s10
+; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_store_dwordx2 v4, v[0:1], s[6:7] offset:32
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v5f64_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 18, @6, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.XY, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.XYZW, T3.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T2.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 6:
+; EG-NEXT:     MOV * T0.W, KC0[7].X,
+; EG-NEXT:     MOV * T0.Z, KC0[6].W,
+; EG-NEXT:     MOV T0.Y, KC0[6].Z,
+; EG-NEXT:     MOV * T1.W, KC0[8].X,
+; EG-NEXT:     MOV T0.X, KC0[6].Y,
+; EG-NEXT:     MOV * T1.Z, KC0[7].W,
+; EG-NEXT:     LSHR T2.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T1.Y, KC0[7].Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T1.X, KC0[7].Y,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T3.X, PV.W, literal.x,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.y,
+; EG-NEXT:    2(2.802597e-45), 32(4.484155e-44)
+; EG-NEXT:     LSHR T4.X, PV.W, literal.x,
+; EG-NEXT:     MOV T5.Y, KC0[8].Z,
+; EG-NEXT:     MOV * T5.X, KC0[8].Y,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v5f64_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 18, @6, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2, T5.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T4.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1, T3.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 6:
+; CM-NEXT:     MOV * T0.W, KC0[8].X,
+; CM-NEXT:     MOV T1.Y, KC0[8].Z,
+; CM-NEXT:     MOV * T0.Z, KC0[7].W,
+; CM-NEXT:     MOV T1.X, KC0[8].Y,
+; CM-NEXT:     MOV * T0.Y, KC0[7].Z,
+; CM-NEXT:     MOV T0.X, KC0[7].Y,
+; CM-NEXT:     ADD_INT T1.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T2.W, KC0[7].X,
+; CM-NEXT:    32(4.484155e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T3.X, PV.Z, literal.x,
+; CM-NEXT:     MOV T2.Z, KC0[6].W,
+; CM-NEXT:     ADD_INT * T1.W, KC0[2].Y, literal.y,
+; CM-NEXT:    2(2.802597e-45), 16(2.242078e-44)
+; CM-NEXT:     LSHR T4.X, PV.W, literal.x,
+; CM-NEXT:     MOV * T2.Y, KC0[6].Z,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T2.X, KC0[6].Y,
+; CM-NEXT:     LSHR * T5.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <5 x double> %in, <5 x double> addrspace(1)* %out, align 8
   ret void
 }
 
 ; FIXME: Lots of unpack and re-pack junk on VI
-; FUNC-LABEL: {{^}}v8i8_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 16
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-
-; SI-NOT: {{buffer|flat|global}}_load
-; SI: s_load_dwordx2 s
-; SI-NEXT: s_load_dwordx2 s
-; SI-NOT: {{buffer|flat|global}}_load
-
-; VI: s_load_dwordx2 s
-; VI-NEXT: s_load_dwordx2 s
-; VI-NOT: lshl
-; VI-NOT: _or
-; VI-NOT: _sdwa
 define amdgpu_kernel void @v8i8_arg(<8 x i8> addrspace(1)* %out, <8 x i8> %in) {
+; SI-LABEL: v8i8_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v8i8_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v8i8_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v8i8_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 1, @36, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @20
+; EG-NEXT:    ALU 5, @38, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @22
+; EG-NEXT:    ALU 5, @44, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @24
+; EG-NEXT:    ALU 7, @50, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @26
+; EG-NEXT:    ALU 7, @58, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @28
+; EG-NEXT:    ALU 7, @66, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @30
+; EG-NEXT:    ALU 7, @74, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @32
+; EG-NEXT:    ALU 5, @82, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @34
+; EG-NEXT:    ALU 5, @88, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.XY, T6.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 20:
+; EG-NEXT:     VTX_READ_8 T6.X, T5.X, 51, #3
+; EG-NEXT:    Fetch clause starting at 22:
+; EG-NEXT:     VTX_READ_8 T6.X, T5.X, 47, #3
+; EG-NEXT:    Fetch clause starting at 24:
+; EG-NEXT:     VTX_READ_8 T6.X, T5.X, 50, #3
+; EG-NEXT:    Fetch clause starting at 26:
+; EG-NEXT:     VTX_READ_8 T6.X, T5.X, 46, #3
+; EG-NEXT:    Fetch clause starting at 28:
+; EG-NEXT:     VTX_READ_8 T6.X, T5.X, 49, #3
+; EG-NEXT:    Fetch clause starting at 30:
+; EG-NEXT:     VTX_READ_8 T6.X, T5.X, 45, #3
+; EG-NEXT:    Fetch clause starting at 32:
+; EG-NEXT:     VTX_READ_8 T6.X, T5.X, 48, #3
+; EG-NEXT:    Fetch clause starting at 34:
+; EG-NEXT:     VTX_READ_8 T5.X, T5.X, 44, #3
+; EG-NEXT:    ALU clause starting at 36:
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:     MOV * T5.X, 0.0,
+; EG-NEXT:    ALU clause starting at 38:
+; EG-NEXT:     LSHL T0.W, T6.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    24(3.363116e-44), 16777215(2.350989e-38)
+; EG-NEXT:     OR_INT * T0.W, PS, PV.W,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 44:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T6.X, literal.y,
+; EG-NEXT:    16777215(2.350989e-38), 24(3.363116e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 50:
+; EG-NEXT:     AND_INT T0.W, T6.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -16711681(-1.714704e+38)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 58:
+; EG-NEXT:     AND_INT T0.W, T6.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -16711681(-1.714704e+38)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 66:
+; EG-NEXT:     AND_INT T0.W, T6.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -65281(nan)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 74:
+; EG-NEXT:     AND_INT T0.W, T6.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -65281(nan)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 82:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T6.X, literal.y,
+; EG-NEXT:    -256(nan), 255(3.573311e-43)
+; EG-NEXT:     OR_INT * T5.Y, PV.W, PS,
+; EG-NEXT:     MOV T2.X, PV.Y,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 88:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T5.X, literal.y,
+; EG-NEXT:    -256(nan), 255(3.573311e-43)
+; EG-NEXT:     OR_INT T5.X, PV.W, PS,
+; EG-NEXT:     LSHR * T6.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v8i8_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 1, @36, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @20
+; CM-NEXT:    ALU 5, @38, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @22
+; CM-NEXT:    ALU 5, @44, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @24
+; CM-NEXT:    ALU 7, @50, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @26
+; CM-NEXT:    ALU 7, @58, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @28
+; CM-NEXT:    ALU 7, @66, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @30
+; CM-NEXT:    ALU 7, @74, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @32
+; CM-NEXT:    ALU 5, @82, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @34
+; CM-NEXT:    ALU 5, @88, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T5, T6.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 20:
+; CM-NEXT:     VTX_READ_8 T6.X, T5.X, 51, #3
+; CM-NEXT:    Fetch clause starting at 22:
+; CM-NEXT:     VTX_READ_8 T6.X, T5.X, 47, #3
+; CM-NEXT:    Fetch clause starting at 24:
+; CM-NEXT:     VTX_READ_8 T6.X, T5.X, 50, #3
+; CM-NEXT:    Fetch clause starting at 26:
+; CM-NEXT:     VTX_READ_8 T6.X, T5.X, 46, #3
+; CM-NEXT:    Fetch clause starting at 28:
+; CM-NEXT:     VTX_READ_8 T6.X, T5.X, 49, #3
+; CM-NEXT:    Fetch clause starting at 30:
+; CM-NEXT:     VTX_READ_8 T6.X, T5.X, 45, #3
+; CM-NEXT:    Fetch clause starting at 32:
+; CM-NEXT:     VTX_READ_8 T6.X, T5.X, 48, #3
+; CM-NEXT:    Fetch clause starting at 34:
+; CM-NEXT:     VTX_READ_8 T5.X, T5.X, 44, #3
+; CM-NEXT:    ALU clause starting at 36:
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:     MOV * T5.X, 0.0,
+; CM-NEXT:    ALU clause starting at 38:
+; CM-NEXT:     LSHL T0.Z, T6.X, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T0.Y, literal.y,
+; CM-NEXT:    24(3.363116e-44), 16777215(2.350989e-38)
+; CM-NEXT:     OR_INT * T0.W, PV.W, PV.Z,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 44:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T6.X, literal.y,
+; CM-NEXT:    16777215(2.350989e-38), 24(3.363116e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 50:
+; CM-NEXT:     AND_INT * T0.W, T6.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -16711681(-1.714704e+38), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 58:
+; CM-NEXT:     AND_INT * T0.W, T6.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -16711681(-1.714704e+38), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 66:
+; CM-NEXT:     AND_INT * T0.W, T6.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -65281(nan), 8(1.121039e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 74:
+; CM-NEXT:     AND_INT * T0.W, T6.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -65281(nan), 8(1.121039e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 82:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T6.X, literal.y,
+; CM-NEXT:    -256(nan), 255(3.573311e-43)
+; CM-NEXT:     OR_INT * T5.Y, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.Y,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 88:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T5.X, literal.y,
+; CM-NEXT:    -256(nan), 255(3.573311e-43)
+; CM-NEXT:     OR_INT * T5.X, PV.Z, PV.W,
+; CM-NEXT:     LSHR * T6.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <8 x i8> %in, <8 x i8> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}v8i16_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 32
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-
-; SI: s_load_dwordx4
-; SI-NEXT: s_load_dwordx2
-; SI-NOT: {{buffer|flat|global}}_load
-
-
-; MESA-VI: s_load_dwordx4 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x34
-
-; HSA-GFX9: s_load_dwordx4 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x10
 define amdgpu_kernel void @v8i16_arg(<8 x i16> addrspace(1)* %out, <8 x i16> %in) {
+; SI-LABEL: v8i16_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v8i16_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v8i16_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v8i16_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 1, @36, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @20
+; EG-NEXT:    ALU 5, @38, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @22
+; EG-NEXT:    ALU 5, @44, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @24
+; EG-NEXT:    ALU 5, @50, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @26
+; EG-NEXT:    ALU 5, @56, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @28
+; EG-NEXT:    ALU 5, @62, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @30
+; EG-NEXT:    ALU 5, @68, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @32
+; EG-NEXT:    ALU 5, @74, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @34
+; EG-NEXT:    ALU 8, @80, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T7.XYZW, T8.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 20:
+; EG-NEXT:     VTX_READ_16 T8.X, T7.X, 66, #3
+; EG-NEXT:    Fetch clause starting at 22:
+; EG-NEXT:     VTX_READ_16 T8.X, T7.X, 58, #3
+; EG-NEXT:    Fetch clause starting at 24:
+; EG-NEXT:     VTX_READ_16 T8.X, T7.X, 64, #3
+; EG-NEXT:    Fetch clause starting at 26:
+; EG-NEXT:     VTX_READ_16 T8.X, T7.X, 56, #3
+; EG-NEXT:    Fetch clause starting at 28:
+; EG-NEXT:     VTX_READ_16 T8.X, T7.X, 62, #3
+; EG-NEXT:    Fetch clause starting at 30:
+; EG-NEXT:     VTX_READ_16 T8.X, T7.X, 54, #3
+; EG-NEXT:    Fetch clause starting at 32:
+; EG-NEXT:     VTX_READ_16 T8.X, T7.X, 60, #3
+; EG-NEXT:    Fetch clause starting at 34:
+; EG-NEXT:     VTX_READ_16 T7.X, T7.X, 52, #3
+; EG-NEXT:    ALU clause starting at 36:
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:     MOV * T7.X, 0.0,
+; EG-NEXT:    ALU clause starting at 38:
+; EG-NEXT:     LSHL T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PS, PV.W,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T5.X,
+; EG-NEXT:    ALU clause starting at 44:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T8.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T5.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 50:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T8.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T5.X,
+; EG-NEXT:    ALU clause starting at 56:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T8.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T5.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 62:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T8.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T4.X,
+; EG-NEXT:    ALU clause starting at 68:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T8.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T4.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 74:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T8.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T7.Z, PV.W, PS,
+; EG-NEXT:     MOV T2.X, PV.Z,
+; EG-NEXT:     MOV * T0.Y, T4.X,
+; EG-NEXT:    ALU clause starting at 80:
+; EG-NEXT:     LSHR T8.X, KC0[2].Y, literal.x,
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.y,
+; EG-NEXT:     AND_INT * T1.W, T7.X, literal.z,
+; EG-NEXT:    2(2.802597e-45), -65536(nan)
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T7.X, PV.W, PS,
+; EG-NEXT:     MOV T4.X, PV.X,
+; EG-NEXT:     MOV * T7.W, T3.X,
+; EG-NEXT:     MOV * T7.Y, T5.X,
+;
+; CM-LABEL: v8i16_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 1, @36, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @20
+; CM-NEXT:    ALU 5, @38, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @22
+; CM-NEXT:    ALU 5, @44, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @24
+; CM-NEXT:    ALU 5, @50, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @26
+; CM-NEXT:    ALU 5, @56, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @28
+; CM-NEXT:    ALU 5, @62, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @30
+; CM-NEXT:    ALU 5, @68, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @32
+; CM-NEXT:    ALU 5, @74, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @34
+; CM-NEXT:    ALU 8, @80, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T7, T8.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 20:
+; CM-NEXT:     VTX_READ_16 T8.X, T7.X, 66, #3
+; CM-NEXT:    Fetch clause starting at 22:
+; CM-NEXT:     VTX_READ_16 T8.X, T7.X, 58, #3
+; CM-NEXT:    Fetch clause starting at 24:
+; CM-NEXT:     VTX_READ_16 T8.X, T7.X, 64, #3
+; CM-NEXT:    Fetch clause starting at 26:
+; CM-NEXT:     VTX_READ_16 T8.X, T7.X, 56, #3
+; CM-NEXT:    Fetch clause starting at 28:
+; CM-NEXT:     VTX_READ_16 T8.X, T7.X, 62, #3
+; CM-NEXT:    Fetch clause starting at 30:
+; CM-NEXT:     VTX_READ_16 T8.X, T7.X, 54, #3
+; CM-NEXT:    Fetch clause starting at 32:
+; CM-NEXT:     VTX_READ_16 T8.X, T7.X, 60, #3
+; CM-NEXT:    Fetch clause starting at 34:
+; CM-NEXT:     VTX_READ_16 T7.X, T7.X, 52, #3
+; CM-NEXT:    ALU clause starting at 36:
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:     MOV * T7.X, 0.0,
+; CM-NEXT:    ALU clause starting at 38:
+; CM-NEXT:     LSHL T0.Z, T8.X, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T0.Y, literal.y,
+; CM-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.W, PV.Z,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T5.X,
+; CM-NEXT:    ALU clause starting at 44:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T8.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T5.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 50:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T5.X,
+; CM-NEXT:    ALU clause starting at 56:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T5.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 62:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T8.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T4.X,
+; CM-NEXT:    ALU clause starting at 68:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T8.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T4.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 74:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T7.Z, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.Z,
+; CM-NEXT:     MOV * T0.Y, T4.X,
+; CM-NEXT:    ALU clause starting at 80:
+; CM-NEXT:     LSHR T8.X, KC0[2].Y, literal.x,
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.y,
+; CM-NEXT:     AND_INT * T0.W, T7.X, literal.z,
+; CM-NEXT:    2(2.802597e-45), -65536(nan)
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     OR_INT * T7.X, PV.Z, PV.W,
+; CM-NEXT:     MOV T4.X, PV.X,
+; CM-NEXT:     MOV * T7.W, T3.X,
+; CM-NEXT:     MOV * T7.Y, T5.X,
 entry:
   store <8 x i16> %in, <8 x i16> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}v8i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 64
-; HSA-GFX9: kernarg_segment_alignment = 5
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[5].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[5].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[5].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[5].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].X
-
-; SI: s_load_dwordx8 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x11
-; MESA-VI: s_load_dwordx8 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x44
-; HSA-GFX9: s_load_dwordx8 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x20
 define amdgpu_kernel void @v8i32_arg(<8 x i32> addrspace(1)* nocapture %out, <8 x i32> %in) nounwind {
+; SI-LABEL: v8i32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x11
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    v_mov_b32_e32 v2, s10
+; SI-NEXT:    v_mov_b32_e32 v3, s11
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v8i32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx8 s[0:7], s[0:1], 0x44
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_add_u32 s4, s8, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_addc_u32 s5, s9, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    v_mov_b32_e32 v4, s8
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    v_mov_b32_e32 v5, s9
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v8i32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x20
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-NEXT:    v_mov_b32_e32 v2, s14
+; GFX9-NEXT:    v_mov_b32_e32 v3, s15
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    v_mov_b32_e32 v2, s10
+; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v8i32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 13, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.XYZW, T3.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T2.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.W, KC0[5].X,
+; EG-NEXT:     MOV * T0.Z, KC0[4].W,
+; EG-NEXT:     MOV T0.Y, KC0[4].Z,
+; EG-NEXT:     MOV * T1.W, KC0[6].X,
+; EG-NEXT:     MOV T0.X, KC0[4].Y,
+; EG-NEXT:     MOV * T1.Z, KC0[5].W,
+; EG-NEXT:     LSHR T2.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T1.Y, KC0[5].Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T1.X, KC0[5].Y,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T3.X, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v8i32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 13, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1, T3.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T2.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     MOV * T0.W, KC0[6].X,
+; CM-NEXT:     MOV * T0.Z, KC0[5].W,
+; CM-NEXT:     MOV * T0.Y, KC0[5].Z,
+; CM-NEXT:     MOV T0.X, KC0[5].Y,
+; CM-NEXT:     MOV * T1.W, KC0[5].X,
+; CM-NEXT:     MOV T1.Z, KC0[4].W,
+; CM-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; CM-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T2.X, PV.W, literal.x,
+; CM-NEXT:     MOV * T1.Y, KC0[4].Z,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T1.X, KC0[4].Y,
+; CM-NEXT:     LSHR * T3.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <8 x i32> %in, <8 x i32> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v8f32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 64
-; HSA-GFX9: kernarg_segment_alignment = 5
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[4].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[5].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[5].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[5].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[5].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].X
-; SI: s_load_dwordx8 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x11
 define amdgpu_kernel void @v8f32_arg(<8 x float> addrspace(1)* nocapture %out, <8 x float> %in) nounwind {
+; SI-LABEL: v8f32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x11
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    v_mov_b32_e32 v2, s10
+; SI-NEXT:    v_mov_b32_e32 v3, s11
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v8f32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx8 s[0:7], s[0:1], 0x44
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_add_u32 s4, s8, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_addc_u32 s5, s9, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    v_mov_b32_e32 v4, s8
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    v_mov_b32_e32 v5, s9
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v8f32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x20
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-NEXT:    v_mov_b32_e32 v2, s14
+; GFX9-NEXT:    v_mov_b32_e32 v3, s15
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    v_mov_b32_e32 v2, s10
+; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v8f32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 13, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.XYZW, T3.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T2.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.W, KC0[5].X,
+; EG-NEXT:     MOV * T0.Z, KC0[4].W,
+; EG-NEXT:     MOV T0.Y, KC0[4].Z,
+; EG-NEXT:     MOV * T1.W, KC0[6].X,
+; EG-NEXT:     MOV T0.X, KC0[4].Y,
+; EG-NEXT:     MOV * T1.Z, KC0[5].W,
+; EG-NEXT:     LSHR T2.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T1.Y, KC0[5].Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T1.X, KC0[5].Y,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T3.X, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v8f32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 13, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1, T3.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T2.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     MOV * T0.W, KC0[6].X,
+; CM-NEXT:     MOV * T0.Z, KC0[5].W,
+; CM-NEXT:     MOV * T0.Y, KC0[5].Z,
+; CM-NEXT:     MOV T0.X, KC0[5].Y,
+; CM-NEXT:     MOV * T1.W, KC0[5].X,
+; CM-NEXT:     MOV T1.Z, KC0[4].W,
+; CM-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; CM-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T2.X, PV.W, literal.x,
+; CM-NEXT:     MOV * T1.Y, KC0[4].Z,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T1.X, KC0[4].Y,
+; CM-NEXT:     LSHR * T3.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <8 x float> %in, <8 x float> addrspace(1)* %out, align 4
   ret void
 }
 
 ; FIXME: Pack/repack on VI
-
-; FUNC-LABEL: {{^}}v16i8_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 32
-; HSA-GFX9: kernarg_segment_alignment = 4
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-; EGCM: VTX_READ_8
-
-; SI: s_load_dwordx4 s
-; SI-NEXT: s_load_dwordx2 s
-; SI-NOT: {{buffer|flat|global}}_load
-
-
-; VI: s_load_dwordx4 s
-; VI-NOT: shr
-; VI-NOT: shl
-; VI-NOT: _sdwa
-; VI-NOT: _or_
 define amdgpu_kernel void @v16i8_arg(<16 x i8> addrspace(1)* %out, <16 x i8> %in) {
+; SI-LABEL: v16i8_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xd
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v16i8_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v16i8_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v16i8_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 1, @68, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @36
+; EG-NEXT:    ALU 5, @70, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @38
+; EG-NEXT:    ALU 5, @76, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @40
+; EG-NEXT:    ALU 5, @82, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @42
+; EG-NEXT:    ALU 5, @88, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @44
+; EG-NEXT:    ALU 7, @94, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @46
+; EG-NEXT:    ALU 7, @102, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @48
+; EG-NEXT:    ALU 7, @110, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @50
+; EG-NEXT:    ALU 7, @118, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @52
+; EG-NEXT:    ALU 7, @126, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @54
+; EG-NEXT:    ALU 7, @134, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @56
+; EG-NEXT:    ALU 7, @142, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @58
+; EG-NEXT:    ALU 7, @150, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @60
+; EG-NEXT:    ALU 5, @158, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @62
+; EG-NEXT:    ALU 5, @164, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @64
+; EG-NEXT:    ALU 5, @170, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @66
+; EG-NEXT:    ALU 5, @176, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T7.XYZW, T8.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 36:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 67, #3
+; EG-NEXT:    Fetch clause starting at 38:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 63, #3
+; EG-NEXT:    Fetch clause starting at 40:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 59, #3
+; EG-NEXT:    Fetch clause starting at 42:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 55, #3
+; EG-NEXT:    Fetch clause starting at 44:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 66, #3
+; EG-NEXT:    Fetch clause starting at 46:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 62, #3
+; EG-NEXT:    Fetch clause starting at 48:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 58, #3
+; EG-NEXT:    Fetch clause starting at 50:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 54, #3
+; EG-NEXT:    Fetch clause starting at 52:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 65, #3
+; EG-NEXT:    Fetch clause starting at 54:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 61, #3
+; EG-NEXT:    Fetch clause starting at 56:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 57, #3
+; EG-NEXT:    Fetch clause starting at 58:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 53, #3
+; EG-NEXT:    Fetch clause starting at 60:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 64, #3
+; EG-NEXT:    Fetch clause starting at 62:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 60, #3
+; EG-NEXT:    Fetch clause starting at 64:
+; EG-NEXT:     VTX_READ_8 T8.X, T7.X, 56, #3
+; EG-NEXT:    Fetch clause starting at 66:
+; EG-NEXT:     VTX_READ_8 T7.X, T7.X, 52, #3
+; EG-NEXT:    ALU clause starting at 68:
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:     MOV * T7.X, 0.0,
+; EG-NEXT:    ALU clause starting at 70:
+; EG-NEXT:     LSHL T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    24(3.363116e-44), 16777215(2.350989e-38)
+; EG-NEXT:     OR_INT * T0.W, PS, PV.W,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 76:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T8.X, literal.y,
+; EG-NEXT:    16777215(2.350989e-38), 24(3.363116e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T4.X,
+; EG-NEXT:    ALU clause starting at 82:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T8.X, literal.y,
+; EG-NEXT:    16777215(2.350989e-38), 24(3.363116e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T4.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T5.X,
+; EG-NEXT:    ALU clause starting at 88:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T8.X, literal.y,
+; EG-NEXT:    16777215(2.350989e-38), 24(3.363116e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T5.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 94:
+; EG-NEXT:     AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -16711681(-1.714704e+38)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 102:
+; EG-NEXT:     AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -16711681(-1.714704e+38)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T4.X,
+; EG-NEXT:    ALU clause starting at 110:
+; EG-NEXT:     AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -16711681(-1.714704e+38)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T4.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T5.X,
+; EG-NEXT:    ALU clause starting at 118:
+; EG-NEXT:     AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -16711681(-1.714704e+38)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T5.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 126:
+; EG-NEXT:     AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -65281(nan)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 134:
+; EG-NEXT:     AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -65281(nan)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T4.X,
+; EG-NEXT:    ALU clause starting at 142:
+; EG-NEXT:     AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -65281(nan)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T4.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T5.X,
+; EG-NEXT:    ALU clause starting at 150:
+; EG-NEXT:     AND_INT T0.W, T8.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    255(3.573311e-43), -65281(nan)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T0.W, T1.W, PV.W,
+; EG-NEXT:     MOV T5.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 158:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T8.X, literal.y,
+; EG-NEXT:    -256(nan), 255(3.573311e-43)
+; EG-NEXT:     OR_INT * T7.W, PV.W, PS,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 164:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T8.X, literal.y,
+; EG-NEXT:    -256(nan), 255(3.573311e-43)
+; EG-NEXT:     OR_INT * T7.Z, PV.W, PS,
+; EG-NEXT:     MOV T3.X, PV.Z,
+; EG-NEXT:     MOV * T0.Y, T4.X,
+; EG-NEXT:    ALU clause starting at 170:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T8.X, literal.y,
+; EG-NEXT:    -256(nan), 255(3.573311e-43)
+; EG-NEXT:     OR_INT * T7.Y, PV.W, PS,
+; EG-NEXT:     MOV T4.X, PV.Y,
+; EG-NEXT:     MOV * T0.Y, T5.X,
+; EG-NEXT:    ALU clause starting at 176:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T7.X, literal.y,
+; EG-NEXT:    -256(nan), 255(3.573311e-43)
+; EG-NEXT:     OR_INT T7.X, PV.W, PS,
+; EG-NEXT:     LSHR * T8.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v16i8_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 1, @68, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @36
+; CM-NEXT:    ALU 5, @70, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @38
+; CM-NEXT:    ALU 5, @76, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @40
+; CM-NEXT:    ALU 5, @82, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @42
+; CM-NEXT:    ALU 5, @88, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @44
+; CM-NEXT:    ALU 7, @94, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @46
+; CM-NEXT:    ALU 7, @102, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @48
+; CM-NEXT:    ALU 7, @110, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @50
+; CM-NEXT:    ALU 7, @118, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @52
+; CM-NEXT:    ALU 7, @126, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @54
+; CM-NEXT:    ALU 7, @134, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @56
+; CM-NEXT:    ALU 7, @142, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @58
+; CM-NEXT:    ALU 7, @150, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @60
+; CM-NEXT:    ALU 5, @158, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @62
+; CM-NEXT:    ALU 5, @164, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @64
+; CM-NEXT:    ALU 5, @170, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @66
+; CM-NEXT:    ALU 5, @176, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T7, T8.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 36:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 67, #3
+; CM-NEXT:    Fetch clause starting at 38:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 63, #3
+; CM-NEXT:    Fetch clause starting at 40:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 59, #3
+; CM-NEXT:    Fetch clause starting at 42:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 55, #3
+; CM-NEXT:    Fetch clause starting at 44:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 66, #3
+; CM-NEXT:    Fetch clause starting at 46:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 62, #3
+; CM-NEXT:    Fetch clause starting at 48:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 58, #3
+; CM-NEXT:    Fetch clause starting at 50:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 54, #3
+; CM-NEXT:    Fetch clause starting at 52:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 65, #3
+; CM-NEXT:    Fetch clause starting at 54:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 61, #3
+; CM-NEXT:    Fetch clause starting at 56:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 57, #3
+; CM-NEXT:    Fetch clause starting at 58:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 53, #3
+; CM-NEXT:    Fetch clause starting at 60:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 64, #3
+; CM-NEXT:    Fetch clause starting at 62:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 60, #3
+; CM-NEXT:    Fetch clause starting at 64:
+; CM-NEXT:     VTX_READ_8 T8.X, T7.X, 56, #3
+; CM-NEXT:    Fetch clause starting at 66:
+; CM-NEXT:     VTX_READ_8 T7.X, T7.X, 52, #3
+; CM-NEXT:    ALU clause starting at 68:
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:     MOV * T7.X, 0.0,
+; CM-NEXT:    ALU clause starting at 70:
+; CM-NEXT:     LSHL T0.Z, T8.X, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T0.Y, literal.y,
+; CM-NEXT:    24(3.363116e-44), 16777215(2.350989e-38)
+; CM-NEXT:     OR_INT * T0.W, PV.W, PV.Z,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 76:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T8.X, literal.y,
+; CM-NEXT:    16777215(2.350989e-38), 24(3.363116e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T4.X,
+; CM-NEXT:    ALU clause starting at 82:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T8.X, literal.y,
+; CM-NEXT:    16777215(2.350989e-38), 24(3.363116e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T4.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T5.X,
+; CM-NEXT:    ALU clause starting at 88:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T8.X, literal.y,
+; CM-NEXT:    16777215(2.350989e-38), 24(3.363116e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T5.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 94:
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -16711681(-1.714704e+38), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 102:
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -16711681(-1.714704e+38), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T4.X,
+; CM-NEXT:    ALU clause starting at 110:
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -16711681(-1.714704e+38), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T4.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T5.X,
+; CM-NEXT:    ALU clause starting at 118:
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -16711681(-1.714704e+38), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T5.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 126:
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -65281(nan), 8(1.121039e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 134:
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -65281(nan), 8(1.121039e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T4.X,
+; CM-NEXT:    ALU clause starting at 142:
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -65281(nan), 8(1.121039e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T4.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T5.X,
+; CM-NEXT:    ALU clause starting at 150:
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.y,
+; CM-NEXT:    -65281(nan), 8(1.121039e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T5.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 158:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.y,
+; CM-NEXT:    -256(nan), 255(3.573311e-43)
+; CM-NEXT:     OR_INT * T7.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 164:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.y,
+; CM-NEXT:    -256(nan), 255(3.573311e-43)
+; CM-NEXT:     OR_INT * T7.Z, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.Z,
+; CM-NEXT:     MOV * T0.Y, T4.X,
+; CM-NEXT:    ALU clause starting at 170:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T8.X, literal.y,
+; CM-NEXT:    -256(nan), 255(3.573311e-43)
+; CM-NEXT:     OR_INT * T7.Y, PV.Z, PV.W,
+; CM-NEXT:     MOV T4.X, PV.Y,
+; CM-NEXT:     MOV * T0.Y, T5.X,
+; CM-NEXT:    ALU clause starting at 176:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T7.X, literal.y,
+; CM-NEXT:    -256(nan), 255(3.573311e-43)
+; CM-NEXT:     OR_INT * T7.X, PV.Z, PV.W,
+; CM-NEXT:     LSHR * T8.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <16 x i8> %in, <16 x i8> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}v16i16_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 64
-; HSA-GFX9: kernarg_segment_alignment = 5
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-; EGCM: VTX_READ_16
-
-; SI: s_load_dwordx8 s
-; SI-NEXT: s_load_dwordx2 s
-; SI-NOT: {{buffer|flat|global}}_load
-
-
-; MESA-VI: s_load_dwordx8 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x44
-
-; HSA-GFX9: s_load_dwordx8 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x20
 define amdgpu_kernel void @v16i16_arg(<16 x i16> addrspace(1)* %out, <16 x i16> %in) {
+; SI-LABEL: v16i16_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x11
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    v_mov_b32_e32 v2, s10
+; SI-NEXT:    v_mov_b32_e32 v3, s11
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v16i16_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx8 s[0:7], s[0:1], 0x44
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_add_u32 s4, s8, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_addc_u32 s5, s9, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    v_mov_b32_e32 v4, s8
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    v_mov_b32_e32 v5, s9
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v16i16_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x20
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-NEXT:    v_mov_b32_e32 v2, s14
+; GFX9-NEXT:    v_mov_b32_e32 v3, s15
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    v_mov_b32_e32 v2, s10
+; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v16i16_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 1, @68, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @36
+; EG-NEXT:    ALU 5, @70, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @38
+; EG-NEXT:    ALU 5, @76, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @40
+; EG-NEXT:    ALU 5, @82, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @42
+; EG-NEXT:    ALU 5, @88, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @44
+; EG-NEXT:    ALU 5, @94, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @46
+; EG-NEXT:    ALU 5, @100, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @48
+; EG-NEXT:    ALU 5, @106, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @50
+; EG-NEXT:    ALU 5, @112, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @52
+; EG-NEXT:    ALU 5, @118, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @54
+; EG-NEXT:    ALU 5, @124, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @56
+; EG-NEXT:    ALU 5, @130, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @58
+; EG-NEXT:    ALU 5, @136, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @60
+; EG-NEXT:    ALU 5, @142, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @62
+; EG-NEXT:    ALU 5, @148, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @64
+; EG-NEXT:    ALU 5, @154, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @66
+; EG-NEXT:    ALU 13, @160, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T12.XYZW, T14.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T11.XYZW, T13.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 36:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 98, #3
+; EG-NEXT:    Fetch clause starting at 38:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 90, #3
+; EG-NEXT:    Fetch clause starting at 40:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 82, #3
+; EG-NEXT:    Fetch clause starting at 42:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 74, #3
+; EG-NEXT:    Fetch clause starting at 44:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 96, #3
+; EG-NEXT:    Fetch clause starting at 46:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 88, #3
+; EG-NEXT:    Fetch clause starting at 48:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 80, #3
+; EG-NEXT:    Fetch clause starting at 50:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 72, #3
+; EG-NEXT:    Fetch clause starting at 52:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 94, #3
+; EG-NEXT:    Fetch clause starting at 54:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 86, #3
+; EG-NEXT:    Fetch clause starting at 56:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 78, #3
+; EG-NEXT:    Fetch clause starting at 58:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 70, #3
+; EG-NEXT:    Fetch clause starting at 60:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 92, #3
+; EG-NEXT:    Fetch clause starting at 62:
+; EG-NEXT:     VTX_READ_16 T12.X, T11.X, 84, #3
+; EG-NEXT:    Fetch clause starting at 64:
+; EG-NEXT:     VTX_READ_16 T13.X, T11.X, 76, #3
+; EG-NEXT:    Fetch clause starting at 66:
+; EG-NEXT:     VTX_READ_16 T11.X, T11.X, 68, #3
+; EG-NEXT:    ALU clause starting at 68:
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:     MOV * T11.X, 0.0,
+; EG-NEXT:    ALU clause starting at 70:
+; EG-NEXT:     LSHL T0.W, T12.X, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.Y, literal.y,
+; EG-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PS, PV.W,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T5.X,
+; EG-NEXT:    ALU clause starting at 76:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T12.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T5.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T7.X,
+; EG-NEXT:    ALU clause starting at 82:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T12.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T7.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T9.X,
+; EG-NEXT:    ALU clause starting at 88:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T12.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T9.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T3.X,
+; EG-NEXT:    ALU clause starting at 94:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T12.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T3.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T5.X,
+; EG-NEXT:    ALU clause starting at 100:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T12.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T5.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T7.X,
+; EG-NEXT:    ALU clause starting at 106:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T12.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T7.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T9.X,
+; EG-NEXT:    ALU clause starting at 112:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T12.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T9.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 118:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T12.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T2.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T4.X,
+; EG-NEXT:    ALU clause starting at 124:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T12.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T4.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T6.X,
+; EG-NEXT:    ALU clause starting at 130:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T12.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T6.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T8.X,
+; EG-NEXT:    ALU clause starting at 136:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     LSHL * T1.W, T12.X, literal.y,
+; EG-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; EG-NEXT:     OR_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MOV T8.X, PV.W,
+; EG-NEXT:     MOV * T0.Y, T2.X,
+; EG-NEXT:    ALU clause starting at 142:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T12.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T12.Z, PV.W, PS,
+; EG-NEXT:     MOV T2.X, PV.Z,
+; EG-NEXT:     MOV * T0.Y, T4.X,
+; EG-NEXT:    ALU clause starting at 148:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T12.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T12.X, PV.W, PS,
+; EG-NEXT:     MOV T4.X, PV.X,
+; EG-NEXT:     MOV * T0.Y, T6.X,
+; EG-NEXT:    ALU clause starting at 154:
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T13.X, literal.y,
+; EG-NEXT:    -65536(nan), 65535(9.183409e-41)
+; EG-NEXT:     OR_INT * T11.Z, PV.W, PS,
+; EG-NEXT:     MOV T6.X, PV.Z,
+; EG-NEXT:     MOV * T0.Y, T8.X,
+; EG-NEXT:    ALU clause starting at 160:
+; EG-NEXT:     LSHR T13.X, KC0[2].Y, literal.x,
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.y,
+; EG-NEXT:    2(2.802597e-45), 16(2.242078e-44)
+; EG-NEXT:     LSHR T14.X, PV.W, literal.x,
+; EG-NEXT:     AND_INT T0.W, T0.Y, literal.y,
+; EG-NEXT:     AND_INT * T1.W, T11.X, literal.z,
+; EG-NEXT:    2(2.802597e-45), -65536(nan)
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T11.X, PV.W, PS,
+; EG-NEXT:     MOV T8.X, PV.X,
+; EG-NEXT:     MOV * T12.W, T3.X,
+; EG-NEXT:     MOV T12.Y, T5.X,
+; EG-NEXT:     MOV T11.W, T7.X, BS:VEC_120/SCL_212
+; EG-NEXT:     MOV * T11.Y, T9.X,
+;
+; CM-LABEL: v16i16_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 1, @68, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @36
+; CM-NEXT:    ALU 5, @70, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @38
+; CM-NEXT:    ALU 5, @76, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @40
+; CM-NEXT:    ALU 5, @82, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @42
+; CM-NEXT:    ALU 5, @88, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @44
+; CM-NEXT:    ALU 5, @94, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @46
+; CM-NEXT:    ALU 5, @100, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @48
+; CM-NEXT:    ALU 5, @106, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @50
+; CM-NEXT:    ALU 5, @112, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @52
+; CM-NEXT:    ALU 5, @118, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @54
+; CM-NEXT:    ALU 5, @124, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @56
+; CM-NEXT:    ALU 5, @130, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @58
+; CM-NEXT:    ALU 5, @136, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @60
+; CM-NEXT:    ALU 5, @142, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @62
+; CM-NEXT:    ALU 5, @148, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @64
+; CM-NEXT:    ALU 5, @154, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @66
+; CM-NEXT:    ALU 14, @160, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T11, T14.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T12, T13.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    Fetch clause starting at 36:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 98, #3
+; CM-NEXT:    Fetch clause starting at 38:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 90, #3
+; CM-NEXT:    Fetch clause starting at 40:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 82, #3
+; CM-NEXT:    Fetch clause starting at 42:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 74, #3
+; CM-NEXT:    Fetch clause starting at 44:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 96, #3
+; CM-NEXT:    Fetch clause starting at 46:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 88, #3
+; CM-NEXT:    Fetch clause starting at 48:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 80, #3
+; CM-NEXT:    Fetch clause starting at 50:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 72, #3
+; CM-NEXT:    Fetch clause starting at 52:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 94, #3
+; CM-NEXT:    Fetch clause starting at 54:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 86, #3
+; CM-NEXT:    Fetch clause starting at 56:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 78, #3
+; CM-NEXT:    Fetch clause starting at 58:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 70, #3
+; CM-NEXT:    Fetch clause starting at 60:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 92, #3
+; CM-NEXT:    Fetch clause starting at 62:
+; CM-NEXT:     VTX_READ_16 T12.X, T11.X, 84, #3
+; CM-NEXT:    Fetch clause starting at 64:
+; CM-NEXT:     VTX_READ_16 T13.X, T11.X, 76, #3
+; CM-NEXT:    Fetch clause starting at 66:
+; CM-NEXT:     VTX_READ_16 T11.X, T11.X, 68, #3
+; CM-NEXT:    ALU clause starting at 68:
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:     MOV * T11.X, 0.0,
+; CM-NEXT:    ALU clause starting at 70:
+; CM-NEXT:     LSHL T0.Z, T12.X, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T0.Y, literal.y,
+; CM-NEXT:    16(2.242078e-44), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.W, PV.Z,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T5.X,
+; CM-NEXT:    ALU clause starting at 76:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T12.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T5.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T7.X,
+; CM-NEXT:    ALU clause starting at 82:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T12.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T7.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T9.X,
+; CM-NEXT:    ALU clause starting at 88:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T12.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T9.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T3.X,
+; CM-NEXT:    ALU clause starting at 94:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T12.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T3.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T5.X,
+; CM-NEXT:    ALU clause starting at 100:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T12.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T5.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T7.X,
+; CM-NEXT:    ALU clause starting at 106:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T12.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T7.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T9.X,
+; CM-NEXT:    ALU clause starting at 112:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T12.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T9.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 118:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T12.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T4.X,
+; CM-NEXT:    ALU clause starting at 124:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T12.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T4.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T6.X,
+; CM-NEXT:    ALU clause starting at 130:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T12.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T6.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T8.X,
+; CM-NEXT:    ALU clause starting at 136:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     LSHL * T0.W, T12.X, literal.y,
+; CM-NEXT:    65535(9.183409e-41), 16(2.242078e-44)
+; CM-NEXT:     OR_INT * T0.W, PV.Z, PV.W,
+; CM-NEXT:     MOV T8.X, PV.W,
+; CM-NEXT:     MOV * T0.Y, T2.X,
+; CM-NEXT:    ALU clause starting at 142:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T12.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T12.Z, PV.Z, PV.W,
+; CM-NEXT:     MOV T2.X, PV.Z,
+; CM-NEXT:     MOV * T0.Y, T4.X,
+; CM-NEXT:    ALU clause starting at 148:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T12.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T12.X, PV.Z, PV.W,
+; CM-NEXT:     MOV T4.X, PV.X,
+; CM-NEXT:     MOV * T0.Y, T6.X,
+; CM-NEXT:    ALU clause starting at 154:
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.x,
+; CM-NEXT:     AND_INT * T0.W, T13.X, literal.y,
+; CM-NEXT:    -65536(nan), 65535(9.183409e-41)
+; CM-NEXT:     OR_INT * T11.Z, PV.Z, PV.W,
+; CM-NEXT:     MOV T6.X, PV.Z,
+; CM-NEXT:     MOV * T0.Y, T8.X,
+; CM-NEXT:    ALU clause starting at 160:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T13.X, PV.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHR T14.X, KC0[2].Y, literal.x,
+; CM-NEXT:     AND_INT T0.Z, T0.Y, literal.y,
+; CM-NEXT:     AND_INT * T0.W, T11.X, literal.z,
+; CM-NEXT:    2(2.802597e-45), -65536(nan)
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     OR_INT * T11.X, PV.Z, PV.W,
+; CM-NEXT:     MOV T8.X, PV.X,
+; CM-NEXT:     MOV * T12.W, T3.X,
+; CM-NEXT:     MOV T12.Y, T5.X,
+; CM-NEXT:     MOV * T11.W, T7.X, BS:VEC_120/SCL_212
+; CM-NEXT:     MOV * T11.Y, T9.X,
 entry:
   store <16 x i16> %in, <16 x i16> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}v16i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 128
-; HSA-GFX9: kernarg_segment_alignment = 6
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[9].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[9].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[9].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[9].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[10].X
-; SI: s_load_dwordx16 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x19
-; MESA-VI: s_load_dwordx16 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x64
-; HSA-GFX9: s_load_dwordx16 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x40
 define amdgpu_kernel void @v16i32_arg(<16 x i32> addrspace(1)* nocapture %out, <16 x i32> %in) nounwind {
+; SI-LABEL: v16i32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx16 s[4:19], s[0:1], 0x19
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s16
+; SI-NEXT:    v_mov_b32_e32 v1, s17
+; SI-NEXT:    v_mov_b32_e32 v2, s18
+; SI-NEXT:    v_mov_b32_e32 v3, s19
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s12
+; SI-NEXT:    v_mov_b32_e32 v1, s13
+; SI-NEXT:    v_mov_b32_e32 v2, s14
+; SI-NEXT:    v_mov_b32_e32 v3, s15
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:32
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    v_mov_b32_e32 v2, s10
+; SI-NEXT:    v_mov_b32_e32 v3, s11
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v16i32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[16:17], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx16 s[0:15], s[0:1], 0x64
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s12
+; VI-NEXT:    s_add_u32 s12, s16, 48
+; VI-NEXT:    v_mov_b32_e32 v1, s13
+; VI-NEXT:    s_addc_u32 s13, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s12
+; VI-NEXT:    v_mov_b32_e32 v2, s14
+; VI-NEXT:    v_mov_b32_e32 v3, s15
+; VI-NEXT:    v_mov_b32_e32 v5, s13
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_nop 0
+; VI-NEXT:    v_mov_b32_e32 v0, s8
+; VI-NEXT:    s_add_u32 s8, s16, 32
+; VI-NEXT:    v_mov_b32_e32 v1, s9
+; VI-NEXT:    s_addc_u32 s9, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s8
+; VI-NEXT:    v_mov_b32_e32 v2, s10
+; VI-NEXT:    v_mov_b32_e32 v3, s11
+; VI-NEXT:    v_mov_b32_e32 v5, s9
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_nop 0
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_add_u32 s4, s16, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_addc_u32 s5, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    v_mov_b32_e32 v4, s16
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    v_mov_b32_e32 v5, s17
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v16i32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_load_dwordx16 s[0:15], s[4:5], 0x40
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-NEXT:    v_mov_b32_e32 v2, s14
+; GFX9-NEXT:    v_mov_b32_e32 v3, s15
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[16:17] offset:48
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    v_mov_b32_e32 v2, s10
+; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[16:17] offset:32
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-NEXT:    v_mov_b32_e32 v2, s6
+; GFX9-NEXT:    v_mov_b32_e32 v3, s7
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[16:17] offset:16
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[16:17]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v16i32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 29, @6, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.XYZW, T7.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.XYZW, T6.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.XYZW, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T2.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 6:
+; EG-NEXT:     MOV * T0.W, KC0[7].X,
+; EG-NEXT:     MOV * T0.Z, KC0[6].W,
+; EG-NEXT:     MOV T0.Y, KC0[6].Z,
+; EG-NEXT:     MOV * T1.W, KC0[8].X,
+; EG-NEXT:     MOV T0.X, KC0[6].Y,
+; EG-NEXT:     MOV * T1.Z, KC0[7].W,
+; EG-NEXT:     LSHR T2.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T1.Y, KC0[7].Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV * T3.W, KC0[9].X,
+; EG-NEXT:     MOV T1.X, KC0[7].Y,
+; EG-NEXT:     MOV * T3.Z, KC0[8].W,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T4.X, PV.W, literal.x,
+; EG-NEXT:     MOV T3.Y, KC0[8].Z,
+; EG-NEXT:     MOV * T5.W, KC0[10].X,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T3.X, KC0[8].Y,
+; EG-NEXT:     MOV * T5.Z, KC0[9].W,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    32(4.484155e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T6.X, PV.W, literal.x,
+; EG-NEXT:     MOV T5.Y, KC0[9].Z,
+; EG-NEXT:     MOV * T5.X, KC0[9].Y,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    48(6.726233e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T7.X, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v16i32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 28, @6, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T4, T7.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1, T6.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2, T5.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T3.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 6:
+; CM-NEXT:     MOV * T0.W, KC0[10].X,
+; CM-NEXT:     MOV * T0.Z, KC0[9].W,
+; CM-NEXT:     MOV * T0.Y, KC0[9].Z,
+; CM-NEXT:     MOV T0.X, KC0[9].Y,
+; CM-NEXT:     ADD_INT T1.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T2.W, KC0[9].X,
+; CM-NEXT:    48(6.726233e-44), 0(0.000000e+00)
+; CM-NEXT:     MOV T2.Z, KC0[8].W,
+; CM-NEXT:     MOV * T1.W, KC0[8].X,
+; CM-NEXT:     LSHR T3.X, T1.Z, literal.x,
+; CM-NEXT:     MOV T2.Y, KC0[8].Z,
+; CM-NEXT:     MOV * T1.Z, KC0[7].W,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV T2.X, KC0[8].Y,
+; CM-NEXT:     MOV * T1.Y, KC0[7].Z,
+; CM-NEXT:     MOV T1.X, KC0[7].Y,
+; CM-NEXT:     ADD_INT T3.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T4.W, KC0[7].X,
+; CM-NEXT:    32(4.484155e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T5.X, PV.Z, literal.x,
+; CM-NEXT:     MOV T4.Z, KC0[6].W,
+; CM-NEXT:     ADD_INT * T3.W, KC0[2].Y, literal.y,
+; CM-NEXT:    2(2.802597e-45), 16(2.242078e-44)
+; CM-NEXT:     LSHR T6.X, PV.W, literal.x,
+; CM-NEXT:     MOV * T4.Y, KC0[6].Z,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T4.X, KC0[6].Y,
+; CM-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <16 x i32> %in, <16 x i32> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}v16f32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 128
-; HSA-GFX9: kernarg_segment_alignment = 6
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[6].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[7].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[8].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[9].X
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[9].Y
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[9].Z
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[9].W
-; EGCM-DAG: T{{[0-9]\.[XYZW]}}, KC0[10].X
-; SI: s_load_dwordx16 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x19
-; MESA-VI: s_load_dwordx16 s{{\[[0-9]+:[0-9]+\]}}, s[0:1], 0x64
-; HSA-GFX9: s_load_dwordx16 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x40
 define amdgpu_kernel void @v16f32_arg(<16 x float> addrspace(1)* nocapture %out, <16 x float> %in) nounwind {
+; SI-LABEL: v16f32_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx16 s[4:19], s[0:1], 0x19
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s16
+; SI-NEXT:    v_mov_b32_e32 v1, s17
+; SI-NEXT:    v_mov_b32_e32 v2, s18
+; SI-NEXT:    v_mov_b32_e32 v3, s19
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s12
+; SI-NEXT:    v_mov_b32_e32 v1, s13
+; SI-NEXT:    v_mov_b32_e32 v2, s14
+; SI-NEXT:    v_mov_b32_e32 v3, s15
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:32
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    v_mov_b32_e32 v2, s10
+; SI-NEXT:    v_mov_b32_e32 v3, s11
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: v16f32_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[16:17], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx16 s[0:15], s[0:1], 0x64
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s12
+; VI-NEXT:    s_add_u32 s12, s16, 48
+; VI-NEXT:    v_mov_b32_e32 v1, s13
+; VI-NEXT:    s_addc_u32 s13, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s12
+; VI-NEXT:    v_mov_b32_e32 v2, s14
+; VI-NEXT:    v_mov_b32_e32 v3, s15
+; VI-NEXT:    v_mov_b32_e32 v5, s13
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_nop 0
+; VI-NEXT:    v_mov_b32_e32 v0, s8
+; VI-NEXT:    s_add_u32 s8, s16, 32
+; VI-NEXT:    v_mov_b32_e32 v1, s9
+; VI-NEXT:    s_addc_u32 s9, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s8
+; VI-NEXT:    v_mov_b32_e32 v2, s10
+; VI-NEXT:    v_mov_b32_e32 v3, s11
+; VI-NEXT:    v_mov_b32_e32 v5, s9
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_nop 0
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_add_u32 s4, s16, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_addc_u32 s5, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    v_mov_b32_e32 v4, s16
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    v_mov_b32_e32 v5, s17
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: v16f32_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_load_dwordx16 s[0:15], s[4:5], 0x40
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-NEXT:    v_mov_b32_e32 v2, s14
+; GFX9-NEXT:    v_mov_b32_e32 v3, s15
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[16:17] offset:48
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    v_mov_b32_e32 v2, s10
+; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[16:17] offset:32
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s5
+; GFX9-NEXT:    v_mov_b32_e32 v2, s6
+; GFX9-NEXT:    v_mov_b32_e32 v3, s7
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[16:17] offset:16
+; GFX9-NEXT:    s_nop 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[16:17]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: v16f32_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 29, @6, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.XYZW, T7.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.XYZW, T6.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.XYZW, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T2.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 6:
+; EG-NEXT:     MOV * T0.W, KC0[7].X,
+; EG-NEXT:     MOV * T0.Z, KC0[6].W,
+; EG-NEXT:     MOV T0.Y, KC0[6].Z,
+; EG-NEXT:     MOV * T1.W, KC0[8].X,
+; EG-NEXT:     MOV T0.X, KC0[6].Y,
+; EG-NEXT:     MOV * T1.Z, KC0[7].W,
+; EG-NEXT:     LSHR T2.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T1.Y, KC0[7].Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV * T3.W, KC0[9].X,
+; EG-NEXT:     MOV T1.X, KC0[7].Y,
+; EG-NEXT:     MOV * T3.Z, KC0[8].W,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T4.X, PV.W, literal.x,
+; EG-NEXT:     MOV T3.Y, KC0[8].Z,
+; EG-NEXT:     MOV * T5.W, KC0[10].X,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T3.X, KC0[8].Y,
+; EG-NEXT:     MOV * T5.Z, KC0[9].W,
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    32(4.484155e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T6.X, PV.W, literal.x,
+; EG-NEXT:     MOV T5.Y, KC0[9].Z,
+; EG-NEXT:     MOV * T5.X, KC0[9].Y,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; EG-NEXT:    48(6.726233e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T7.X, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: v16f32_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 28, @6, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T4, T7.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1, T6.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2, T5.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T3.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 6:
+; CM-NEXT:     MOV * T0.W, KC0[10].X,
+; CM-NEXT:     MOV * T0.Z, KC0[9].W,
+; CM-NEXT:     MOV * T0.Y, KC0[9].Z,
+; CM-NEXT:     MOV T0.X, KC0[9].Y,
+; CM-NEXT:     ADD_INT T1.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T2.W, KC0[9].X,
+; CM-NEXT:    48(6.726233e-44), 0(0.000000e+00)
+; CM-NEXT:     MOV T2.Z, KC0[8].W,
+; CM-NEXT:     MOV * T1.W, KC0[8].X,
+; CM-NEXT:     LSHR T3.X, T1.Z, literal.x,
+; CM-NEXT:     MOV T2.Y, KC0[8].Z,
+; CM-NEXT:     MOV * T1.Z, KC0[7].W,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV T2.X, KC0[8].Y,
+; CM-NEXT:     MOV * T1.Y, KC0[7].Z,
+; CM-NEXT:     MOV T1.X, KC0[7].Y,
+; CM-NEXT:     ADD_INT T3.Z, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T4.W, KC0[7].X,
+; CM-NEXT:    32(4.484155e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR T5.X, PV.Z, literal.x,
+; CM-NEXT:     MOV T4.Z, KC0[6].W,
+; CM-NEXT:     ADD_INT * T3.W, KC0[2].Y, literal.y,
+; CM-NEXT:    2(2.802597e-45), 16(2.242078e-44)
+; CM-NEXT:     LSHR T6.X, PV.W, literal.x,
+; CM-NEXT:     MOV * T4.Y, KC0[6].Z,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T4.X, KC0[6].Y,
+; CM-NEXT:     LSHR * T7.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store <16 x float> %in, <16 x float> addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}kernel_arg_i64:
-; MESA-VI: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[0:1], 0x24
-; HSA-GFX9: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x0
-
-; MESA-GCN: buffer_store_dwordx2
 define amdgpu_kernel void @kernel_arg_i64(i64 addrspace(1)* %out, i64 %a) nounwind {
+; SI-LABEL: kernel_arg_i64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    v_mov_b32_e32 v0, s2
+; SI-NEXT:    v_mov_b32_e32 v1, s3
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: kernel_arg_i64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: kernel_arg_i64:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: kernel_arg_i64:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.Y, KC0[3].X,
+; EG-NEXT:     MOV T0.X, KC0[2].W,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: kernel_arg_i64:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     MOV * T0.Y, KC0[3].X,
+; CM-NEXT:     MOV * T0.X, KC0[2].W,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   store i64 %a, i64 addrspace(1)* %out, align 8
   ret void
 }
 
-; FUNC-LABEL: {{^}}f64_kernel_arg:
-; SI-DAG: s_load_dwordx4 s[{{[0-9]:[0-9]}}], s[0:1], 0x9
-; MESA-VI-DAG: s_load_dwordx4 s[{{[0-9]:[0-9]}}], s[0:1], 0x24
-; MESA-GCN: buffer_store_dwordx2
-
-; HSA-GFX9: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x0
 define amdgpu_kernel void @f64_kernel_arg(double addrspace(1)* %out, double  %in) {
+; SI-LABEL: f64_kernel_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    v_mov_b32_e32 v0, s2
+; SI-NEXT:    v_mov_b32_e32 v1, s3
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: f64_kernel_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: f64_kernel_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: f64_kernel_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 4:
+; EG-NEXT:     MOV * T0.Y, KC0[3].X,
+; EG-NEXT:     MOV T0.X, KC0[2].W,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: f64_kernel_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 3, @4, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 4:
+; CM-NEXT:     MOV * T0.Y, KC0[3].X,
+; CM-NEXT:     MOV * T0.X, KC0[2].W,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store double %in, double addrspace(1)* %out
   ret void
 }
 
 ; XFUNC-LABEL: {{^}}kernel_arg_v1i64:
 ; XGCN: s_load_dwordx2
 ; XGCN: s_load_dwordx2
 ; XGCN: buffer_store_dwordx2
 ; define amdgpu_kernel void @kernel_arg_v1i64(<1 x i64> addrspace(1)* %out, <1 x i64> %a) nounwind {
 ;   store <1 x i64> %a, <1 x i64> addrspace(1)* %out, align 8
 ;   ret void
 ; }
 
-; FUNC-LABEL: {{^}}i65_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 24
-; HSA-GFX9: kernarg_segment_alignment = 4
-; HSA-GFX9: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x0
-; HSA-GFX9: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x8
 define amdgpu_kernel void @i65_arg(i65 addrspace(1)* nocapture %out, i65 %in) nounwind {
+; SI-LABEL: i65_arg:
+; SI:       ; %bb.0: ; %entry
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xd
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_and_b32 s6, s2, 1
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    buffer_store_byte v2, off, s[0:3], 0 offset:8
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i65_arg:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x2c
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    s_and_b32 s1, s0, 1
+; VI-NEXT:    s_add_u32 s0, s2, 8
+; VI-NEXT:    v_mov_b32_e32 v4, s1
+; VI-NEXT:    s_addc_u32 s1, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_byte v[2:3], v4
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    v_mov_b32_e32 v3, s5
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i65_arg:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x8
+; GFX9-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    s_and_b32 s4, s6, 1
+; GFX9-NEXT:    v_mov_b32_e32 v3, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_byte v2, v3, s[0:1] offset:8
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i65_arg:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 20, @6, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.X, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.X, T2.X, 0
+; EG-NEXT:    MEM_RAT MSKOR T1.XW, T0.X
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 6:
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     AND_INT * T1.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T1.W, PV.W, literal.x,
+; EG-NEXT:     AND_INT * T2.W, KC0[3].Y, 1,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T1.X, PS, PV.W,
+; EG-NEXT:     LSHL * T1.W, literal.x, PV.W,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     MOV T1.Y, 0.0,
+; EG-NEXT:     MOV * T1.Z, 0.0,
+; EG-NEXT:     LSHR T0.X, T0.W, literal.x,
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.y,
+; EG-NEXT:    2(2.802597e-45), 4(5.605194e-45)
+; EG-NEXT:     LSHR T2.X, PV.W, literal.x,
+; EG-NEXT:     MOV * T3.X, KC0[3].X,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHR T4.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T5.X, KC0[2].W,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: i65_arg:
+; CM:       ; %bb.0: ; %entry
+; CM-NEXT:    ALU 21, @6, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T1.XW, T5.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T4.X, T3.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2.X, T0.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 6:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; CM-NEXT:     AND_INT * T1.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHL T0.Z, PV.W, literal.x,
+; CM-NEXT:     AND_INT * T1.W, KC0[3].Y, 1,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHL T1.X, PV.W, PV.Z,
+; CM-NEXT:     LSHL * T1.W, literal.x, PV.Z,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     MOV T1.Y, 0.0,
+; CM-NEXT:     MOV * T1.Z, 0.0,
+; CM-NEXT:     LSHR * T0.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV T2.X, KC0[2].W,
+; CM-NEXT:     ADD_INT * T2.W, KC0[2].Y, literal.x,
+; CM-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T3.X, PV.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T4.X, KC0[3].X,
+; CM-NEXT:     LSHR * T5.X, T0.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
 entry:
   store i65 %in, i65 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i1_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; GCN: s_load_dword s
-; GCN: s_and_b32
-; GCN: {{buffer|flat|global}}_store_byte
 define amdgpu_kernel void @i1_arg(i1 addrspace(1)* %out, i1 %x) nounwind {
+; SI-LABEL: i1_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_and_b32 s4, s2, 1
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_byte v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i1_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_and_b32 s0, s0, 1
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_byte v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i1_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_and_b32 s2, s2, 1
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_byte v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i1_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 11, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T0.XW, T1.X
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     AND_INT T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:     AND_INT * T1.W, T0.X, 1,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T0.X, T1.W, PV.W,
+; EG-NEXT:     LSHL * T0.W, literal.x, PV.W,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     MOV T0.Y, 0.0,
+; EG-NEXT:     MOV * T0.Z, 0.0,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: i1_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 11, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T0.XW, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     AND_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     AND_INT T0.Z, T0.X, 1,
+; CM-NEXT:     LSHL * T0.W, PV.W, literal.x,
+; CM-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; CM-NEXT:     LSHL T0.X, PV.Z, PV.W,
+; CM-NEXT:     LSHL * T0.W, literal.x, PV.W,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     MOV T0.Y, 0.0,
+; CM-NEXT:     MOV * T0.Z, 0.0,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   store i1 %x, i1 addrspace(1)* %out, align 1
   ret void
 }
 
-; FUNC-LABEL: {{^}}i1_arg_zext_i32:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; GCN: s_load_dword
-; GCN: {{buffer|flat|global}}_store_dword
 define amdgpu_kernel void @i1_arg_zext_i32(i32 addrspace(1)* %out, i1 %x) nounwind {
+; SI-LABEL: i1_arg_zext_i32:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_and_b32 s4, s2, 1
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i1_arg_zext_i32:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_and_b32 s0, s0, 1
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i1_arg_zext_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_and_b32 s2, s2, 1
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i1_arg_zext_i32:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 1, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: i1_arg_zext_i32:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 1, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = zext i1 %x to i32
   store i32 %ext, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i1_arg_zext_i64:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; GCN: s_load_dword s
-; GCN: {{buffer|flat|global}}_store_dwordx2
 define amdgpu_kernel void @i1_arg_zext_i64(i64 addrspace(1)* %out, i1 %x) nounwind {
+; SI-LABEL: i1_arg_zext_i64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s4, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_and_b32 s4, s4, 1
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i1_arg_zext_i64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    v_mov_b32_e32 v1, 0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    s_and_b32 s0, s0, 1
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i1_arg_zext_i64:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_and_b32 s2, s2, 1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    global_store_dwordx2 v1, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i1_arg_zext_i64:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     MOV * T0.Y, 0.0,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: i1_arg_zext_i64:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     MOV * T0.Y, 0.0,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = zext i1 %x to i64
   store i64 %ext, i64 addrspace(1)* %out, align 8
   ret void
 }
 
-; FUNC-LABEL: {{^}}i1_arg_sext_i32:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; GCN: s_load_dword
-; GCN: {{buffer|flat|global}}_store_dword
 define amdgpu_kernel void @i1_arg_sext_i32(i32 addrspace(1)* %out, i1 %x) nounwind {
+; SI-LABEL: i1_arg_sext_i32:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_bfe_i32 s4, s2, 0x10000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i1_arg_sext_i32:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_bfe_i32 s0, s0, 0x10000
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i1_arg_sext_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_bfe_i32 s2, s2, 0x10000
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i1_arg_sext_i32:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     BFE_INT T0.X, T0.X, 0.0, 1,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: i1_arg_sext_i32:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     BFE_INT * T0.X, T0.X, 0.0, 1,
+; CM-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = sext i1 %x to i32
   store i32 %ext, i32addrspace(1)* %out, align 4
   ret void
 }
 
-; FUNC-LABEL: {{^}}i1_arg_sext_i64:
-; HSA-GFX9: kernarg_segment_byte_size = 12
-; HSA-GFX9: kernarg_segment_alignment = 4
-
-; GCN: s_load_dword
-; GCN: s_bfe_i64
-; GCN: {{buffer|flat|global}}_store_dwordx2
 define amdgpu_kernel void @i1_arg_sext_i64(i64 addrspace(1)* %out, i1 %x) nounwind {
+; SI-LABEL: i1_arg_sext_i64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s2, s[0:1], 0xb
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_bfe_i64 s[4:5], s[2:3], 0x10000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: i1_arg_sext_i64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x2c
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    s_bfe_i64 s[0:1], s[0:1], 0x10000
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: i1_arg_sext_i64:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_bfe_i64 s[2:3], s[2:3], 0x10000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: i1_arg_sext_i64:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 3, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     BFE_INT T0.X, T0.X, 0.0, 1,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV * T0.Y, PV.X,
+;
+; CM-LABEL: i1_arg_sext_i64:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 3, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T1.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_8 T0.X, T0.X, 40, #3
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     BFE_INT * T0.X, T0.X, 0.0, 1,
+; CM-NEXT:     LSHR T1.X, KC0[2].Y, literal.x,
+; CM-NEXT:     MOV * T0.Y, PV.X,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %ext = sext i1 %x to i64
   store i64 %ext, i64 addrspace(1)* %out, align 8
   ret void
 }
 
-; FUNC-LABEL: {{^}}empty_struct_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 0
 define amdgpu_kernel void @empty_struct_arg({} %in) nounwind {
+; SI-LABEL: empty_struct_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: empty_struct_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: empty_struct_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_endpgm
+;
+; EGCM-LABEL: empty_struct_arg:
+; EGCM:       ; %bb.0:
+; EGCM-NEXT:    CF_END
+; EGCM-NEXT:    PAD
   ret void
 }
 
 ; The correct load offsets for these:
 ; load 4 from 0,
 ; load 8 from 8
 ; load 4 from 24
 ; load 8 from 32
 
 ; With the SelectionDAG argument lowering, the alignments for the
 ; struct members is not properly considered, making these wrong.
 
 ; FIXME: Total argument size is computed wrong
-; FUNC-LABEL: {{^}}struct_argument_alignment:
-; HSA-GFX9: kernarg_segment_byte_size = 40
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0x0
-; HSA-GFX9: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x8
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0x18
-; HSA-GFX9: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x20
 define amdgpu_kernel void @struct_argument_alignment({i32, i64} %arg0, i8, {i32, i64} %arg1) {
+; SI-LABEL: struct_argument_alignment:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s8, s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xb
+; SI-NEXT:    s_load_dword s9, s[0:1], 0xf
+; SI-NEXT:    s_load_dwordx2 s[6:7], s[0:1], 0x11
+; SI-NEXT:    s_mov_b32 s0, 0
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_mov_b32 s1, s0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s9
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s6
+; SI-NEXT:    v_mov_b32_e32 v1, s7
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: struct_argument_alignment:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s4, s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x2c
+; VI-NEXT:    s_load_dword s5, s[0:1], 0x3c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x44
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v2, s4
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v2, s5
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: struct_argument_alignment:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dword s6, s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x8
+; GFX9-NEXT:    s_load_dword s7, s[4:5], 0x18
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x20
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v2, s6
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v3, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s0
+; GFX9-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v2, s7
+; GFX9-NEXT:    global_store_dword v[0:1], v2, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: struct_argument_alignment:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 9, @8, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T7.X, T6.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.X, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.X, T6.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.X, T6.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T6.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV T0.X, KC0[4].Y,
+; EG-NEXT:     MOV * T1.X, KC0[4].Z,
+; EG-NEXT:     MOV T2.X, KC0[3].W,
+; EG-NEXT:     MOV * T3.X, KC0[2].W,
+; EG-NEXT:     MOV T4.X, literal.x,
+; EG-NEXT:     MOV * T5.X, KC0[3].X,
+; EG-NEXT:    1(1.401298e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T6.X, literal.x,
+; EG-NEXT:     MOV * T7.X, KC0[2].Y,
+; EG-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+;
+; CM-LABEL: struct_argument_alignment:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 9, @8, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T7.X, T6.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T5.X, T4.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T3.X, T6.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2.X, T6.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T4.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T6.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, KC0[4].Y,
+; CM-NEXT:     MOV * T1.X, KC0[4].Z,
+; CM-NEXT:     MOV * T2.X, KC0[3].W,
+; CM-NEXT:     MOV * T3.X, KC0[2].W,
+; CM-NEXT:     MOV * T4.X, literal.x,
+; CM-NEXT:    1(1.401298e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T5.X, KC0[3].X,
+; CM-NEXT:     MOV * T6.X, literal.x,
+; CM-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; CM-NEXT:     MOV * T7.X, KC0[2].Y,
   %val0 = extractvalue {i32, i64} %arg0, 0
   %val1 = extractvalue {i32, i64} %arg0, 1
   %val2 = extractvalue {i32, i64} %arg1, 0
   %val3 = extractvalue {i32, i64} %arg1, 1
   store volatile i32 %val0, i32 addrspace(1)* null
   store volatile i64 %val1, i64 addrspace(1)* null
   store volatile i32 %val2, i32 addrspace(1)* null
   store volatile i64 %val3, i64 addrspace(1)* null
   ret void
 }
 
 ; No padding between i8 and next struct, but round up at end to 4 byte
 ; multiple.
-; FUNC-LABEL: {{^}}packed_struct_argument_alignment:
-; HSA-GFX9: kernarg_segment_byte_size = 28
-; HSA-GFX9-DAG: s_load_dword s{{[0-9]+}}, s[4:5], 0x0
-; HSA-GFX9-DAG: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x4
-; HSA-GFX9-DAG: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0{{$}}
-; HSA-GFX9: global_load_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, [[ZERO]], s{{\[[0-9]+:[0-9]+\]}} offset:17
-; HSA-GFX9: global_load_dword v{{[0-9]+}}, [[ZERO]], s{{\[[0-9]+:[0-9]+\]}} offset:13
 define amdgpu_kernel void @packed_struct_argument_alignment(<{i32, i64}> %arg0, i8, <{i32, i64}> %arg1) {
+; SI-LABEL: packed_struct_argument_alignment:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_load_dword s6, s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xa
+; SI-NEXT:    buffer_load_ubyte v4, off, s[0:3], 0 offset:49
+; SI-NEXT:    buffer_load_ubyte v5, off, s[0:3], 0 offset:50
+; SI-NEXT:    buffer_load_ubyte v6, off, s[0:3], 0 offset:51
+; SI-NEXT:    buffer_load_ubyte v7, off, s[0:3], 0 offset:52
+; SI-NEXT:    buffer_load_dwordx2 v[0:1], off, s[0:3], 0 offset:53
+; SI-NEXT:    s_mov_b32 s0, 0
+; SI-NEXT:    s_mov_b32 s1, s0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    buffer_store_dword v2, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v2, s4
+; SI-NEXT:    v_mov_b32_e32 v3, s5
+; SI-NEXT:    buffer_store_dwordx2 v[2:3], off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_lshlrev_b32_e32 v2, 8, v5
+; SI-NEXT:    v_lshlrev_b32_e32 v3, 8, v7
+; SI-NEXT:    v_or_b32_e32 v2, v2, v4
+; SI-NEXT:    v_or_b32_e32 v3, v3, v6
+; SI-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT:    v_or_b32_e32 v2, v3, v2
+; SI-NEXT:    buffer_store_dword v2, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: packed_struct_argument_alignment:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_add_u32 s2, s0, 49
+; VI-NEXT:    s_addc_u32 s3, s1, 0
+; VI-NEXT:    s_add_u32 s4, s0, 50
+; VI-NEXT:    s_addc_u32 s5, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    s_add_u32 s2, s2, 3
+; VI-NEXT:    s_addc_u32 s3, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    s_add_u32 s2, s0, 51
+; VI-NEXT:    s_addc_u32 s3, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v7, s3
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_mov_b32_e32 v6, s2
+; VI-NEXT:    s_load_dword s4, s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x28
+; VI-NEXT:    flat_load_ubyte v8, v[0:1]
+; VI-NEXT:    flat_load_ubyte v9, v[2:3]
+; VI-NEXT:    flat_load_ubyte v10, v[4:5]
+; VI-NEXT:    flat_load_ubyte v6, v[6:7]
+; VI-NEXT:    s_add_u32 s0, s0, 53
+; VI-NEXT:    s_addc_u32 s1, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v2, 0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v3, 0
+; VI-NEXT:    v_mov_b32_e32 v7, s4
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    flat_store_dword v[2:3], v7
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[4:5]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_lshlrev_b32_e32 v4, 8, v8
+; VI-NEXT:    v_or_b32_e32 v4, v4, v9
+; VI-NEXT:    v_lshlrev_b32_e32 v5, 8, v10
+; VI-NEXT:    v_or_b32_sdwa v5, v5, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; VI-NEXT:    v_or_b32_e32 v4, v5, v4
+; VI-NEXT:    flat_store_dword v[2:3], v4
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: packed_struct_argument_alignment:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x4
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v2, s[4:5] offset:17
+; GFX9-NEXT:    global_load_dword v6, v2, s[4:5] offset:13
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mov_b32_e32 v3, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v7, s2
+; GFX9-NEXT:    v_mov_b32_e32 v5, s1
+; GFX9-NEXT:    v_mov_b32_e32 v4, s0
+; GFX9-NEXT:    global_store_dword v[2:3], v7, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[4:5], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_dword v[2:3], v6, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_dwordx2 v[2:3], v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: packed_struct_argument_alignment:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 6, @18, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T4.X, T3.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.X, T1.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T3.X, 0
+; EG-NEXT:    ALU 2, @25, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @12
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T3.X, 0
+; EG-NEXT:    TEX 0 @14
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.X, T1.X, 0
+; EG-NEXT:    TEX 0 @16
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T4.X, T3.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 12:
+; EG-NEXT:     VTX_READ_32 T0.X, T0.X, 49, #3
+; EG-NEXT:    Fetch clause starting at 14:
+; EG-NEXT:     VTX_READ_32 T2.X, T2.X, 57, #3
+; EG-NEXT:    Fetch clause starting at 16:
+; EG-NEXT:     VTX_READ_32 T4.X, T4.X, 53, #3
+; EG-NEXT:    ALU clause starting at 18:
+; EG-NEXT:     MOV T0.X, KC0[2].Z,
+; EG-NEXT:     MOV * T1.X, literal.x,
+; EG-NEXT:    1(1.401298e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T2.X, KC0[2].W,
+; EG-NEXT:     MOV * T3.X, literal.x,
+; EG-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; EG-NEXT:     MOV * T4.X, KC0[2].Y,
+; EG-NEXT:    ALU clause starting at 25:
+; EG-NEXT:     MOV T0.X, 0.0,
+; EG-NEXT:     MOV * T2.X, 0.0,
+; EG-NEXT:     MOV * T4.X, 0.0,
+;
+; CM-LABEL: packed_struct_argument_alignment:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 6, @18, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T4.X, T3.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2.X, T1.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T3.X
+; CM-NEXT:    ALU 2, @25, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @12
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T3.X
+; CM-NEXT:    TEX 0 @14
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2.X, T1.X
+; CM-NEXT:    TEX 0 @16
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T4.X, T3.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    Fetch clause starting at 12:
+; CM-NEXT:     VTX_READ_32 T0.X, T0.X, 49, #3
+; CM-NEXT:    Fetch clause starting at 14:
+; CM-NEXT:     VTX_READ_32 T2.X, T2.X, 57, #3
+; CM-NEXT:    Fetch clause starting at 16:
+; CM-NEXT:     VTX_READ_32 T4.X, T4.X, 53, #3
+; CM-NEXT:    ALU clause starting at 18:
+; CM-NEXT:     MOV * T0.X, KC0[2].Z,
+; CM-NEXT:     MOV * T1.X, literal.x,
+; CM-NEXT:    1(1.401298e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T2.X, KC0[2].W,
+; CM-NEXT:     MOV * T3.X, literal.x,
+; CM-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; CM-NEXT:     MOV * T4.X, KC0[2].Y,
+; CM-NEXT:    ALU clause starting at 25:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:     MOV * T2.X, 0.0,
+; CM-NEXT:     MOV * T4.X, 0.0,
   %val0 = extractvalue <{i32, i64}> %arg0, 0
   %val1 = extractvalue <{i32, i64}> %arg0, 1
   %val2 = extractvalue <{i32, i64}> %arg1, 0
   %val3 = extractvalue <{i32, i64}> %arg1, 1
   store volatile i32 %val0, i32 addrspace(1)* null
   store volatile i64 %val1, i64 addrspace(1)* null
   store volatile i32 %val2, i32 addrspace(1)* null
   store volatile i64 %val3, i64 addrspace(1)* null
   ret void
 }
 
-; GCN-LABEL: {{^}}struct_argument_alignment_after:
-; HSA-GFX9: kernarg_segment_byte_size = 64
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0x0
-; HSA-GFX9: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x8
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0x18
-; HSA-GFX9: s_load_dwordx2 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x20
-; HSA-GFX9: s_load_dwordx4 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x30
 define amdgpu_kernel void @struct_argument_alignment_after({i32, i64} %arg0, i8, {i32, i64} %arg2, i8, <4 x i32> %arg4) {
+; SI-LABEL: struct_argument_alignment_after:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s12, s[0:1], 0x9
+; SI-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0xb
+; SI-NEXT:    s_load_dword s13, s[0:1], 0xf
+; SI-NEXT:    s_load_dwordx2 s[10:11], s[0:1], 0x11
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x15
+; SI-NEXT:    s_mov_b32 s4, 0
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s5, s4
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s12
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s13
+; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s10
+; SI-NEXT:    v_mov_b32_e32 v1, s11
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    v_mov_b32_e32 v1, s1
+; SI-NEXT:    v_mov_b32_e32 v2, s2
+; SI-NEXT:    v_mov_b32_e32 v3, s3
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: struct_argument_alignment_after:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s8, s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x2c
+; VI-NEXT:    s_load_dword s9, s[0:1], 0x3c
+; VI-NEXT:    s_load_dwordx2 s[6:7], s[0:1], 0x44
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x54
+; VI-NEXT:    v_mov_b32_e32 v4, 0
+; VI-NEXT:    v_mov_b32_e32 v5, 0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s8
+; VI-NEXT:    flat_store_dword v[4:5], v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s9
+; VI-NEXT:    flat_store_dword v[4:5], v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: struct_argument_alignment_after:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dword s10, s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x8
+; GFX9-NEXT:    s_load_dword s11, s[4:5], 0x18
+; GFX9-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x20
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x30
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    v_mov_b32_e32 v5, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s10
+; GFX9-NEXT:    global_store_dword v[4:5], v0, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s6
+; GFX9-NEXT:    v_mov_b32_e32 v1, s7
+; GFX9-NEXT:    global_store_dwordx2 v[4:5], v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s11
+; GFX9-NEXT:    global_store_dword v[4:5], v0, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    global_store_dwordx2 v[4:5], v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v2, s2
+; GFX9-NEXT:    v_mov_b32_e32 v3, s3
+; GFX9-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: struct_argument_alignment_after:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 13, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T8.X, T7.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T6.X, T5.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T4.X, T7.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.X, T7.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.X, T5.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T7.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T7.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    ALU clause starting at 10:
+; EG-NEXT:     MOV * T0.W, KC0[6].X,
+; EG-NEXT:     MOV * T0.Z, KC0[5].W,
+; EG-NEXT:     MOV * T0.Y, KC0[5].Z,
+; EG-NEXT:     MOV T0.X, KC0[5].Y,
+; EG-NEXT:     MOV * T1.X, KC0[4].Y,
+; EG-NEXT:     MOV T2.X, KC0[4].Z,
+; EG-NEXT:     MOV * T3.X, KC0[3].W,
+; EG-NEXT:     MOV T4.X, KC0[2].W,
+; EG-NEXT:     MOV * T5.X, literal.x,
+; EG-NEXT:    1(1.401298e-45), 0(0.000000e+00)
+; EG-NEXT:     MOV T6.X, KC0[3].X,
+; EG-NEXT:     MOV * T7.X, literal.x,
+; EG-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; EG-NEXT:     MOV * T8.X, KC0[2].Y,
+;
+; CM-LABEL: struct_argument_alignment_after:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 13, @10, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T8.X, T7.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T6.X, T5.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T4.X, T7.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T3.X, T7.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2.X, T5.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T7.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T7.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    ALU clause starting at 10:
+; CM-NEXT:     MOV * T0.W, KC0[6].X,
+; CM-NEXT:     MOV * T0.Z, KC0[5].W,
+; CM-NEXT:     MOV * T0.Y, KC0[5].Z,
+; CM-NEXT:     MOV * T0.X, KC0[5].Y,
+; CM-NEXT:     MOV * T1.X, KC0[4].Y,
+; CM-NEXT:     MOV * T2.X, KC0[4].Z,
+; CM-NEXT:     MOV * T3.X, KC0[3].W,
+; CM-NEXT:     MOV * T4.X, KC0[2].W,
+; CM-NEXT:     MOV * T5.X, literal.x,
+; CM-NEXT:    1(1.401298e-45), 0(0.000000e+00)
+; CM-NEXT:     MOV * T6.X, KC0[3].X,
+; CM-NEXT:     MOV * T7.X, literal.x,
+; CM-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; CM-NEXT:     MOV * T8.X, KC0[2].Y,
   %val0 = extractvalue {i32, i64} %arg0, 0
   %val1 = extractvalue {i32, i64} %arg0, 1
   %val2 = extractvalue {i32, i64} %arg2, 0
   %val3 = extractvalue {i32, i64} %arg2, 1
   store volatile i32 %val0, i32 addrspace(1)* null
   store volatile i64 %val1, i64 addrspace(1)* null
   store volatile i32 %val2, i32 addrspace(1)* null
   store volatile i64 %val3, i64 addrspace(1)* null
   store volatile <4 x i32> %arg4, <4 x i32> addrspace(1)* null
   ret void
 }
 
-; GCN-LABEL: {{^}}array_3xi32:
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0x0
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0x4
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0x8
-; HSA-GFX9: s_load_dword s{{[0-9]+}}, s[4:5], 0xc
 define amdgpu_kernel void @array_3xi32(i16 %arg0, [3 x i32] %arg1) {
+; SI-LABEL: array_3xi32:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s4, s[0:1], 0xc
+; SI-NEXT:    s_load_dword s5, s[0:1], 0x9
+; SI-NEXT:    s_load_dword s6, s[0:1], 0xa
+; SI-NEXT:    s_load_dword s0, s[0:1], 0xb
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s5
+; SI-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s6
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: array_3xi32:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x24
+; VI-NEXT:    s_load_dword s3, s[0:1], 0x28
+; VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x30
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s0
+; VI-NEXT:    flat_store_short v[0:1], v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    flat_store_dword v[0:1], v1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    flat_store_dword v[0:1], v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s3
+; VI-NEXT:    flat_store_dword v[0:1], v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: array_3xi32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dword s0, s[4:5], 0x0
+; GFX9-NEXT:    s_load_dword s1, s[4:5], 0x4
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    s_load_dword s3, s[4:5], 0xc
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s3
+; GFX9-NEXT:    global_store_short v[0:1], v0, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_dword v[0:1], v1, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    global_store_dword v[0:1], v0, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s1
+; GFX9-NEXT:    global_store_dword v[0:1], v0, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: array_3xi32:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @10, KC0[], KC1[]
+; EG-NEXT:    TEX 0 @8
+; EG-NEXT:    ALU 9, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T0.XW, T4.X
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.X, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.X, T4.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T4.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 8:
+; EG-NEXT:     VTX_READ_16 T0.X, T0.X, 36, #3
+; EG-NEXT:    ALU clause starting at 10:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 11:
+; EG-NEXT:     AND_INT T0.X, T0.X, literal.x,
+; EG-NEXT:     MOV * T0.W, literal.x,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T0.Y, 0.0,
+; EG-NEXT:     MOV * T0.Z, 0.0,
+; EG-NEXT:     MOV T1.X, KC0[2].Z,
+; EG-NEXT:     MOV * T2.X, KC0[2].W,
+; EG-NEXT:     MOV T3.X, KC0[3].X,
+; EG-NEXT:     MOV * T4.X, literal.x,
+; EG-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+;
+; CM-LABEL: array_3xi32:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @10, KC0[], KC1[]
+; CM-NEXT:    TEX 0 @8
+; CM-NEXT:    ALU 9, @11, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T0.XW, T4.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T3.X, T4.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2.X, T4.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T4.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    Fetch clause starting at 8:
+; CM-NEXT:     VTX_READ_16 T0.X, T0.X, 36, #3
+; CM-NEXT:    ALU clause starting at 10:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 11:
+; CM-NEXT:     AND_INT T0.X, T0.X, literal.x,
+; CM-NEXT:     MOV * T0.W, literal.x,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T0.Y, 0.0,
+; CM-NEXT:     MOV * T0.Z, 0.0,
+; CM-NEXT:     MOV * T1.X, KC0[2].Z,
+; CM-NEXT:     MOV * T2.X, KC0[2].W,
+; CM-NEXT:     MOV * T3.X, KC0[3].X,
+; CM-NEXT:     MOV * T4.X, literal.x,
+; CM-NEXT:    0(0.000000e+00), 0(0.000000e+00)
   store volatile i16 %arg0, i16 addrspace(1)* undef
   store volatile [3 x i32] %arg1, [3 x i32] addrspace(1)* undef
   ret void
 }
 
 ; FIXME: Why not all scalar loads?
-; GCN-LABEL: {{^}}array_3xi16:
-; HSA-GFX9-DAG: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0{{$}}
-; HSA-GFX9: global_load_ushort v{{[0-9]+}}, [[ZERO]], s{{\[[0-9]+:[0-9]+\]}} offset:2
-; HSA-GFX9: global_load_ushort v{{[0-9]+}}, [[ZERO]], s{{\[[0-9]+:[0-9]+\]}} offset:4
-; HSA-GFX9: global_load_ushort v{{[0-9]+}}, [[ZERO]], s{{\[[0-9]+:[0-9]+\]}} offset:6
 define amdgpu_kernel void @array_3xi16(i8 %arg0, [3 x i16] %arg1) {
+; SI-LABEL: array_3xi16:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dword s4, s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    buffer_load_ushort v0, off, s[0:3], 0 offset:42
+; SI-NEXT:    buffer_load_ushort v1, off, s[0:3], 0 offset:40
+; SI-NEXT:    buffer_load_ushort v2, off, s[0:3], 0 offset:38
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v3, s4
+; SI-NEXT:    buffer_store_byte v3, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    buffer_store_short v1, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    buffer_store_short v2, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: array_3xi16:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_add_u32 s2, s0, 38
+; VI-NEXT:    s_addc_u32 s3, s1, 0
+; VI-NEXT:    s_add_u32 s4, s2, 2
+; VI-NEXT:    s_addc_u32 s5, s3, 0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    s_add_u32 s2, s0, 42
+; VI-NEXT:    s_addc_u32 s3, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v5, s3
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    v_mov_b32_e32 v4, s2
+; VI-NEXT:    flat_load_ushort v0, v[0:1]
+; VI-NEXT:    flat_load_ushort v1, v[2:3]
+; VI-NEXT:    flat_load_ushort v2, v[4:5]
+; VI-NEXT:    s_load_dword s0, s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v3, s0
+; VI-NEXT:    s_waitcnt vmcnt(1)
+; VI-NEXT:    flat_store_byte v[0:1], v3
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    flat_store_short v[0:1], v2
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    flat_store_short v[0:1], v1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    flat_store_short v[0:1], v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: array_3xi16:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    global_load_ushort v1, v0, s[4:5] offset:2
+; GFX9-NEXT:    global_load_ushort v2, v0, s[4:5] offset:4
+; GFX9-NEXT:    global_load_ushort v3, v0, s[4:5] offset:6
+; GFX9-NEXT:    s_load_dword s0, s[4:5], 0x0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    s_waitcnt vmcnt(2)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_short v[0:1], v3, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_short v[0:1], v2, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_short v[0:1], v1, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: array_3xi16:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @20, KC0[], KC1[]
+; EG-NEXT:    TEX 1 @12
+; EG-NEXT:    ALU 11, @21, KC0[], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T1.XW, T3.X
+; EG-NEXT:    MEM_RAT MSKOR T2.XW, T3.X
+; EG-NEXT:    TEX 0 @16
+; EG-NEXT:    ALU 3, @33, KC0[], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T2.XW, T3.X
+; EG-NEXT:    TEX 0 @18
+; EG-NEXT:    ALU 3, @37, KC0[], KC1[]
+; EG-NEXT:    MEM_RAT MSKOR T2.XW, T3.X
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 12:
+; EG-NEXT:     VTX_READ_8 T1.X, T0.X, 36, #3
+; EG-NEXT:     VTX_READ_16 T2.X, T0.X, 42, #3
+; EG-NEXT:    Fetch clause starting at 16:
+; EG-NEXT:     VTX_READ_16 T1.X, T0.X, 40, #3
+; EG-NEXT:    Fetch clause starting at 18:
+; EG-NEXT:     VTX_READ_16 T0.X, T0.X, 38, #3
+; EG-NEXT:    ALU clause starting at 20:
+; EG-NEXT:     MOV * T0.X, 0.0,
+; EG-NEXT:    ALU clause starting at 21:
+; EG-NEXT:     AND_INT T1.X, T1.X, literal.x,
+; EG-NEXT:     MOV * T1.W, literal.x,
+; EG-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EG-NEXT:     MOV * T1.Y, 0.0,
+; EG-NEXT:     AND_INT T2.X, T2.X, literal.x,
+; EG-NEXT:     MOV * T2.W, literal.x,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:     MOV T2.Y, 0.0,
+; EG-NEXT:     MOV T1.Z, 0.0,
+; EG-NEXT:     MOV * T2.Z, 0.0,
+; EG-NEXT:     MOV * T3.X, literal.x,
+; EG-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 33:
+; EG-NEXT:     AND_INT T2.X, T1.X, literal.x,
+; EG-NEXT:     MOV T2.Y, 0.0,
+; EG-NEXT:     MOV * T2.Z, 0.0,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 37:
+; EG-NEXT:     AND_INT T2.X, T0.X, literal.x,
+; EG-NEXT:     MOV T2.Y, 0.0,
+; EG-NEXT:     MOV * T2.Z, 0.0,
+; EG-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+;
+; CM-LABEL: array_3xi16:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @20, KC0[], KC1[]
+; CM-NEXT:    TEX 1 @12
+; CM-NEXT:    ALU 11, @21, KC0[], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T1.XW, T3.X
+; CM-NEXT:    MEM_RAT MSKOR T2.XW, T3.X
+; CM-NEXT:    TEX 0 @16
+; CM-NEXT:    ALU 3, @33, KC0[], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T2.XW, T3.X
+; CM-NEXT:    TEX 0 @18
+; CM-NEXT:    ALU 3, @37, KC0[], KC1[]
+; CM-NEXT:    MEM_RAT MSKOR T2.XW, T3.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    Fetch clause starting at 12:
+; CM-NEXT:     VTX_READ_8 T1.X, T0.X, 36, #3
+; CM-NEXT:     VTX_READ_16 T2.X, T0.X, 42, #3
+; CM-NEXT:    Fetch clause starting at 16:
+; CM-NEXT:     VTX_READ_16 T1.X, T0.X, 40, #3
+; CM-NEXT:    Fetch clause starting at 18:
+; CM-NEXT:     VTX_READ_16 T0.X, T0.X, 38, #3
+; CM-NEXT:    ALU clause starting at 20:
+; CM-NEXT:     MOV * T0.X, 0.0,
+; CM-NEXT:    ALU clause starting at 21:
+; CM-NEXT:     AND_INT T1.X, T1.X, literal.x,
+; CM-NEXT:     MOV * T1.W, literal.x,
+; CM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; CM-NEXT:     MOV * T1.Y, 0.0,
+; CM-NEXT:     AND_INT T2.X, T2.X, literal.x,
+; CM-NEXT:     MOV * T2.W, literal.x,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:     MOV T2.Y, 0.0,
+; CM-NEXT:     MOV * T1.Z, 0.0,
+; CM-NEXT:     MOV * T2.Z, 0.0,
+; CM-NEXT:     MOV * T3.X, literal.x,
+; CM-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; CM-NEXT:    ALU clause starting at 33:
+; CM-NEXT:     AND_INT T2.X, T1.X, literal.x,
+; CM-NEXT:     MOV T2.Y, 0.0,
+; CM-NEXT:     MOV * T2.Z, 0.0,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
+; CM-NEXT:    ALU clause starting at 37:
+; CM-NEXT:     AND_INT T2.X, T0.X, literal.x,
+; CM-NEXT:     MOV T2.Y, 0.0,
+; CM-NEXT:     MOV * T2.Z, 0.0,
+; CM-NEXT:    65535(9.183409e-41), 0(0.000000e+00)
   store volatile i8 %arg0, i8 addrspace(1)* undef
   store volatile [3 x i16] %arg1, [3 x i16] addrspace(1)* undef
   ret void
 }
 
-; GCN-LABEL: {{^}}small_array_round_down_offset:
-; HSA-GFX9-DAG: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0{{$}}
-; HSA-GFX9: global_load_ubyte v{{[0-9]+}}, [[ZERO]], s{{\[[0-9]+:[0-9]+\]}} offset:1
 define amdgpu_kernel void @small_array_round_down_offset(i8, [1 x i8] %arg) {
+; SI-LABEL: small_array_round_down_offset:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    buffer_load_ubyte v0, off, s[0:3], 0 offset:37
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    buffer_store_byte v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: small_array_round_down_offset:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_add_u32 s0, s0, 37
+; VI-NEXT:    s_addc_u32 s1, s1, 0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    flat_load_ubyte v0, v[0:1]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    flat_store_byte v[0:1], v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: small_array_round_down_offset:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    global_load_ubyte v0, v0, s[4:5] offset:1
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; EGCM-LABEL: small_array_round_down_offset:
+; EGCM:       ; %bb.0:
+; EGCM-NEXT:    ALU 0, @8, KC0[], KC1[]
+; EGCM-NEXT:    TEX 0 @6
+; EGCM-NEXT:    ALU 6, @9, KC0[], KC1[]
+; EGCM-NEXT:    MEM_RAT MSKOR T0.XW, T1.X
+; EGCM-NEXT:    CF_END
+; EGCM-NEXT:    PAD
+; EGCM-NEXT:    Fetch clause starting at 6:
+; EGCM-NEXT:     VTX_READ_8 T0.X, T0.X, 37, #3
+; EGCM-NEXT:    ALU clause starting at 8:
+; EGCM-NEXT:     MOV * T0.X, 0.0,
+; EGCM-NEXT:    ALU clause starting at 9:
+; EGCM-NEXT:     AND_INT T0.X, T0.X, literal.x,
+; EGCM-NEXT:     MOV * T0.W, literal.x,
+; EGCM-NEXT:    255(3.573311e-43), 0(0.000000e+00)
+; EGCM-NEXT:     MOV T0.Y, 0.0,
+; EGCM-NEXT:     MOV * T0.Z, 0.0,
+; EGCM-NEXT:     MOV * T1.X, literal.x,
+; EGCM-NEXT:    0(0.000000e+00), 0(0.000000e+00)
   %val = extractvalue [1 x i8] %arg, 0
   store volatile i8 %val, i8 addrspace(1)* undef
   ret void
 }
 
-; GCN-LABEL: {{^}}byref_align_constant_i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 264
-; HSA-GFX9-DAG: s_load_dwordx2 {{s\[[0-9]+:[0-9]+\]}}, s[4:5], 0x100{{$}}
 define amdgpu_kernel void @byref_align_constant_i32_arg(i32 addrspace(1)* nocapture %out, i32 addrspace(4)* byref(i32) align(256) %in.byref, i32 %after.offset) {
+; SI-LABEL: byref_align_constant_i32_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x49
+; SI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s3, 0xf000
+; SI-NEXT:    s_mov_b32 s2, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s5
+; SI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: byref_align_constant_i32_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x124
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    flat_store_dword v[0:1], v3
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: byref_align_constant_i32_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[4:5], 0x100
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, s2
+; GFX9-NEXT:    v_mov_b32_e32 v2, s3
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_dword v0, v2, s[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: byref_align_constant_i32_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T2.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T2.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, KC0[18].Y,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     MOV T1.X, KC0[18].Z,
+; EG-NEXT:     LSHR * T2.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: byref_align_constant_i32_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    TEX 0 @6
+; CM-NEXT:    ALU 2, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0.X, T2.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T2.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    Fetch clause starting at 6:
+; CM-NEXT:     VTX_READ_32 T0.X, T0.X, 0, #1
+; CM-NEXT:    ALU clause starting at 8:
+; CM-NEXT:     MOV * T0.X, KC0[18].Y,
+; CM-NEXT:    ALU clause starting at 9:
+; CM-NEXT:     MOV * T1.X, KC0[18].Z,
+; CM-NEXT:     LSHR * T2.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %in = load i32, i32 addrspace(4)* %in.byref
   store volatile i32 %in, i32 addrspace(1)* %out, align 4
   store volatile i32 %after.offset, i32 addrspace(1)* %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}byref_natural_align_constant_v16i32_arg:
-; HSA-GFX9: kernarg_segment_byte_size = 132
-; HSA-GFX9-DAG: s_load_dword s{{[0-9]+}}, s[4:5], 0x80
-; HSA-GFX9-DAG: s_load_dwordx16 s{{\[[0-9]+:[0-9]+\]}}, s[4:5], 0x40{{$}}
 define amdgpu_kernel void @byref_natural_align_constant_v16i32_arg(i32 addrspace(1)* nocapture %out, i8, <16 x i32> addrspace(4)* byref(<16 x i32>) %in.byref, i32 %after.offset) {
+; SI-LABEL: byref_natural_align_constant_v16i32_arg:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx16 s[4:19], s[0:1], 0x19
+; SI-NEXT:    s_load_dwordx2 s[20:21], s[0:1], 0x9
+; SI-NEXT:    s_load_dword s0, s[0:1], 0x29
+; SI-NEXT:    s_mov_b32 s23, 0xf000
+; SI-NEXT:    s_mov_b32 s22, -1
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s16
+; SI-NEXT:    v_mov_b32_e32 v1, s17
+; SI-NEXT:    v_mov_b32_e32 v2, s18
+; SI-NEXT:    v_mov_b32_e32 v3, s19
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[20:23], 0 offset:48
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s12
+; SI-NEXT:    v_mov_b32_e32 v1, s13
+; SI-NEXT:    v_mov_b32_e32 v2, s14
+; SI-NEXT:    v_mov_b32_e32 v3, s15
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[20:23], 0 offset:32
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s8
+; SI-NEXT:    v_mov_b32_e32 v1, s9
+; SI-NEXT:    v_mov_b32_e32 v2, s10
+; SI-NEXT:    v_mov_b32_e32 v3, s11
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[20:23], 0 offset:16
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s4
+; SI-NEXT:    v_mov_b32_e32 v1, s5
+; SI-NEXT:    v_mov_b32_e32 v2, s6
+; SI-NEXT:    v_mov_b32_e32 v3, s7
+; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[20:23], 0
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v0, s0
+; SI-NEXT:    buffer_store_dword v0, off, s[20:23], 0
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: byref_natural_align_constant_v16i32_arg:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx2 s[16:17], s[0:1], 0x24
+; VI-NEXT:    s_load_dword s18, s[0:1], 0xa4
+; VI-NEXT:    s_load_dwordx16 s[0:15], s[0:1], 0x64
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s12
+; VI-NEXT:    s_add_u32 s12, s16, 48
+; VI-NEXT:    v_mov_b32_e32 v1, s13
+; VI-NEXT:    s_addc_u32 s13, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s12
+; VI-NEXT:    v_mov_b32_e32 v2, s14
+; VI-NEXT:    v_mov_b32_e32 v3, s15
+; VI-NEXT:    v_mov_b32_e32 v5, s13
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s8
+; VI-NEXT:    s_add_u32 s8, s16, 32
+; VI-NEXT:    v_mov_b32_e32 v1, s9
+; VI-NEXT:    s_addc_u32 s9, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s8
+; VI-NEXT:    v_mov_b32_e32 v2, s10
+; VI-NEXT:    v_mov_b32_e32 v3, s11
+; VI-NEXT:    v_mov_b32_e32 v5, s9
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_add_u32 s4, s16, 16
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_addc_u32 s5, s17, 0
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v2, s6
+; VI-NEXT:    v_mov_b32_e32 v3, s7
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v4, s16
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    v_mov_b32_e32 v5, s17
+; VI-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s18
+; VI-NEXT:    flat_store_dword v[4:5], v0
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: byref_natural_align_constant_v16i32_arg:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x80
+; GFX9-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x40
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s20
+; GFX9-NEXT:    v_mov_b32_e32 v1, s21
+; GFX9-NEXT:    v_mov_b32_e32 v2, s22
+; GFX9-NEXT:    v_mov_b32_e32 v3, s23
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:48
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s16
+; GFX9-NEXT:    v_mov_b32_e32 v1, s17
+; GFX9-NEXT:    v_mov_b32_e32 v2, s18
+; GFX9-NEXT:    v_mov_b32_e32 v3, s19
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:32
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s12
+; GFX9-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-NEXT:    v_mov_b32_e32 v2, s14
+; GFX9-NEXT:    v_mov_b32_e32 v3, s15
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s8
+; GFX9-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-NEXT:    v_mov_b32_e32 v2, s10
+; GFX9-NEXT:    v_mov_b32_e32 v3, s11
+; GFX9-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    global_store_dword v4, v0, s[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; EG-LABEL: byref_natural_align_constant_v16i32_arg:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @24, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @16
+; EG-NEXT:    ALU 3, @25, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.XYZW, T2.X, 0
+; EG-NEXT:    ALU 3, @29, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @18
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.XYZW, T1.X, 0
+; EG-NEXT:    ALU 3, @33, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @20
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.XYZW, T1.X, 0
+; EG-NEXT:    ALU 2, @37, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @22
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XYZW, T2.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T1.X, T2.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 16:
+; EG-NEXT:     VTX_READ_128 T1.XYZW, T0.X, 48, #1
+; EG-NEXT:    Fetch clause starting at 18:
+; EG-NEXT:     VTX_READ_128 T2.XYZW, T0.X, 32, #1
+; EG-NEXT:    Fetch clause starting at 20:
+; EG-NEXT:     VTX_READ_128 T2.XYZW, T0.X, 16, #1
+; EG-NEXT:    Fetch clause starting at 22:
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 24:
+; EG-NEXT:     MOV * T0.X, KC0[6].Y,
+; EG-NEXT:    ALU clause starting at 25:
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    48(6.726233e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T2.X, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 29:
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    32(4.484155e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T1.X, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 33:
+; EG-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T1.X, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 37:
+; EG-NEXT:     MOV T1.X, KC0[10].Y,
+; EG-NEXT:     LSHR * T2.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; CM-LABEL: byref_natural_align_constant_v16i32_arg:
+; CM:       ; %bb.0:
+; CM-NEXT:    ALU 0, @24, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    TEX 0 @16
+; CM-NEXT:    ALU 3, @25, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1, T2.X
+; CM-NEXT:    ALU 3, @29, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    TEX 0 @18
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2, T1.X
+; CM-NEXT:    ALU 3, @33, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    TEX 0 @20
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T2, T1.X
+; CM-NEXT:    ALU 2, @37, KC0[CB0:0-32], KC1[]
+; CM-NEXT:    TEX 0 @22
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T0, T2.X
+; CM-NEXT:    MEM_RAT_CACHELESS STORE_DWORD T1.X, T2.X
+; CM-NEXT:    CF_END
+; CM-NEXT:    PAD
+; CM-NEXT:    Fetch clause starting at 16:
+; CM-NEXT:     VTX_READ_128 T1.XYZW, T0.X, 48, #1
+; CM-NEXT:    Fetch clause starting at 18:
+; CM-NEXT:     VTX_READ_128 T2.XYZW, T0.X, 32, #1
+; CM-NEXT:    Fetch clause starting at 20:
+; CM-NEXT:     VTX_READ_128 T2.XYZW, T0.X, 16, #1
+; CM-NEXT:    Fetch clause starting at 22:
+; CM-NEXT:     VTX_READ_128 T0.XYZW, T0.X, 0, #1
+; CM-NEXT:    ALU clause starting at 24:
+; CM-NEXT:     MOV * T0.X, KC0[6].Y,
+; CM-NEXT:    ALU clause starting at 25:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    48(6.726233e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T2.X, PV.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:    ALU clause starting at 29:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    32(4.484155e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T1.X, PV.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:    ALU clause starting at 33:
+; CM-NEXT:     ADD_INT * T0.W, KC0[2].Y, literal.x,
+; CM-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; CM-NEXT:     LSHR * T1.X, PV.W, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; CM-NEXT:    ALU clause starting at 37:
+; CM-NEXT:     MOV * T1.X, KC0[10].Y,
+; CM-NEXT:     LSHR * T2.X, KC0[2].Y, literal.x,
+; CM-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %in = load <16 x i32>, <16 x i32> addrspace(4)* %in.byref
   %cast.out = bitcast i32 addrspace(1)* %out to <16 x i32> addrspace(1)*
   store volatile <16 x i32> %in, <16 x i32> addrspace(1)* %cast.out, align 4
   store volatile i32 %after.offset, i32 addrspace(1)* %out, align 4
   ret void
 }