AMDGPU/load-global-i32: regenerate test using UTC (NFC)

[artagnon]

Fix the RUN lines so that UTC runs cleanly, and regenerate the test load-global-i32.ll using utils/update_llc_test_checks.py.

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: Ramkumar Ramachandra (artagnon)

Changes

Fix the RUN lines so that UTC runs cleanly, strip hand-written CHECK lines, and regenerate the test load-global-i32.ll using utils/update_llc_test_checks.py.

---

Patch is 226.21 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/73962.diff

1 Files Affected:

• (modified) llvm/test/CodeGen/AMDGPU/load-global-i32.ll (+4462-473)

diff --git a/llvm/test/CodeGen/AMDGPU/load-global-i32.ll b/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
index c4d9b4b2bb5ebbb..55f0773f7e05aea 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
@@ -1,113 +1,825 @@
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=SI-NOHSA -check-prefix=FUNC %s
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=GCNX3-HSA -check-prefix=FUNC %s
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=GCNX3-NOHSA -check-prefix=FUNC %s
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=r600 -mcpu=redwood < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=EG -check-prefix=FUNC %s
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn--amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn--amdhsa -mcpu=gfx908 -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCNX3-HSA %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCNX3-NOHSA %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-HSA -check-prefix=GCN-GFX900-HSA %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=gfx908 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-HSA -check-prefix=GCN-GFX908-HSA %s
 
-; FUNC-LABEL: {{^}}global_load_i32:
-; GCN-NOHSA: buffer_load_dword v{{[0-9]+}}
-; GCN-HSA: {{flat|global}}_load_dword
-
-; EG: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0
 define amdgpu_kernel void @global_load_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; SI-NOHSA-LABEL: global_load_i32:
+; SI-NOHSA:       ; %bb.0: ; %entry
+; SI-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NOHSA-NEXT:    s_mov_b32 s6, -1
+; SI-NOHSA-NEXT:    s_mov_b32 s10, s6
+; SI-NOHSA-NEXT:    s_mov_b32 s11, s7
+; SI-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NOHSA-NEXT:    s_mov_b32 s8, s2
+; SI-NOHSA-NEXT:    s_mov_b32 s9, s3
+; SI-NOHSA-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; SI-NOHSA-NEXT:    s_mov_b32 s4, s0
+; SI-NOHSA-NEXT:    s_mov_b32 s5, s1
+; SI-NOHSA-NEXT:    s_waitcnt vmcnt(0)
+; SI-NOHSA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SI-NOHSA-NEXT:    s_endpgm
+;
+; GCNX3-HSA-LABEL: global_load_i32:
+; GCNX3-HSA:       ; %bb.0: ; %entry
+; GCNX3-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCNX3-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GCNX3-HSA-NEXT:    flat_load_dword v2, v[0:1]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s1
+; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(0)
+; GCNX3-HSA-NEXT:    flat_store_dword v[0:1], v2
+; GCNX3-HSA-NEXT:    s_endpgm
+;
+; GCNX3-NOHSA-LABEL: global_load_i32:
+; GCNX3-NOHSA:       ; %bb.0: ; %entry
+; GCNX3-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s6, -1
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s10, s6
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s11, s7
+; GCNX3-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s8, s2
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s9, s3
+; GCNX3-NOHSA-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s4, s0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s5, s1
+; GCNX3-NOHSA-NEXT:    s_waitcnt vmcnt(0)
+; GCNX3-NOHSA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCNX3-NOHSA-NEXT:    s_endpgm
+;
+; EG-LABEL: global_load_i32:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], 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_32 T0.X, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; 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)
+;
+; GCN-HSA-LABEL: global_load_i32:
+; GCN-HSA:       ; %bb.0: ; %entry
+; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, 0
+; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-HSA-NEXT:    global_load_dword v1, v0, s[2:3]
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
+; GCN-HSA-NEXT:    global_store_dword v0, v1, s[0:1]
+; GCN-HSA-NEXT:    s_endpgm
 entry:
   %ld = load i32, ptr addrspace(1) %in
   store i32 %ld, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}global_load_v2i32:
-; GCN-NOHSA: buffer_load_dwordx2
-; GCN-HSA: {{flat|global}}_load_dwordx2
-
-; EG: VTX_READ_64
 define amdgpu_kernel void @global_load_v2i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; SI-NOHSA-LABEL: global_load_v2i32:
+; SI-NOHSA:       ; %bb.0: ; %entry
+; SI-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NOHSA-NEXT:    s_mov_b32 s6, -1
+; SI-NOHSA-NEXT:    s_mov_b32 s10, s6
+; SI-NOHSA-NEXT:    s_mov_b32 s11, s7
+; SI-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NOHSA-NEXT:    s_mov_b32 s8, s2
+; SI-NOHSA-NEXT:    s_mov_b32 s9, s3
+; SI-NOHSA-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; SI-NOHSA-NEXT:    s_mov_b32 s4, s0
+; SI-NOHSA-NEXT:    s_mov_b32 s5, s1
+; SI-NOHSA-NEXT:    s_waitcnt vmcnt(0)
+; SI-NOHSA-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NOHSA-NEXT:    s_endpgm
+;
+; GCNX3-HSA-LABEL: global_load_v2i32:
+; GCNX3-HSA:       ; %bb.0: ; %entry
+; GCNX3-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCNX3-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GCNX3-HSA-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v2, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v3, s1
+; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(0)
+; GCNX3-HSA-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GCNX3-HSA-NEXT:    s_endpgm
+;
+; GCNX3-NOHSA-LABEL: global_load_v2i32:
+; GCNX3-NOHSA:       ; %bb.0: ; %entry
+; GCNX3-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s6, -1
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s10, s6
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s11, s7
+; GCNX3-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s8, s2
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s9, s3
+; GCNX3-NOHSA-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s4, s0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s5, s1
+; GCNX3-NOHSA-NEXT:    s_waitcnt vmcnt(0)
+; GCNX3-NOHSA-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GCNX3-NOHSA-NEXT:    s_endpgm
+;
+; EG-LABEL: global_load_v2i32:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 1, @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_64 T0.XY, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; 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)
+;
+; GCN-HSA-LABEL: global_load_v2i32:
+; GCN-HSA:       ; %bb.0: ; %entry
+; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, 0
+; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-HSA-NEXT:    global_load_dwordx2 v[0:1], v2, s[2:3]
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
+; GCN-HSA-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GCN-HSA-NEXT:    s_endpgm
 entry:
   %ld = load <2 x i32>, ptr addrspace(1) %in
   store <2 x i32> %ld, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}global_load_v3i32:
-; SI-NOHSA: buffer_load_dwordx4
-; GCNX3-NOHSA: buffer_load_dwordx3
-; GCNX3-HSA: {{flat|global}}_load_dwordx3
-
-; EG: VTX_READ_128
 define amdgpu_kernel void @global_load_v3i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; SI-NOHSA-LABEL: global_load_v3i32:
+; SI-NOHSA:       ; %bb.0: ; %entry
+; SI-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NOHSA-NEXT:    s_mov_b32 s6, -1
+; SI-NOHSA-NEXT:    s_mov_b32 s10, s6
+; SI-NOHSA-NEXT:    s_mov_b32 s11, s7
+; SI-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NOHSA-NEXT:    s_mov_b32 s8, s2
+; SI-NOHSA-NEXT:    s_mov_b32 s9, s3
+; SI-NOHSA-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
+; SI-NOHSA-NEXT:    s_mov_b32 s4, s0
+; SI-NOHSA-NEXT:    s_mov_b32 s5, s1
+; SI-NOHSA-NEXT:    s_waitcnt vmcnt(0)
+; SI-NOHSA-NEXT:    buffer_store_dword v2, off, s[4:7], 0 offset:8
+; SI-NOHSA-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NOHSA-NEXT:    s_endpgm
+;
+; GCNX3-HSA-LABEL: global_load_v3i32:
+; GCNX3-HSA:       ; %bb.0: ; %entry
+; GCNX3-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCNX3-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GCNX3-HSA-NEXT:    flat_load_dwordx3 v[0:2], v[0:1]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v3, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(0)
+; GCNX3-HSA-NEXT:    flat_store_dwordx3 v[3:4], v[0:2]
+; GCNX3-HSA-NEXT:    s_endpgm
+;
+; GCNX3-NOHSA-LABEL: global_load_v3i32:
+; GCNX3-NOHSA:       ; %bb.0: ; %entry
+; GCNX3-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s6, -1
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s10, s6
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s11, s7
+; GCNX3-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s8, s2
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s9, s3
+; GCNX3-NOHSA-NEXT:    buffer_load_dwordx3 v[0:2], off, s[8:11], 0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s4, s0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s5, s1
+; GCNX3-NOHSA-NEXT:    s_waitcnt vmcnt(0)
+; GCNX3-NOHSA-NEXT:    buffer_store_dwordx3 v[0:2], off, s[4:7], 0
+; GCNX3-NOHSA-NEXT:    s_endpgm
+;
+; EG-LABEL: global_load_v3i32:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 6, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.X, T3.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     LSHR T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:     MOV * T2.X, T0.Z,
+; 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 * T3.X, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; GCN-HSA-LABEL: global_load_v3i32:
+; GCN-HSA:       ; %bb.0: ; %entry
+; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, 0
+; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-HSA-NEXT:    global_load_dwordx3 v[0:2], v3, s[2:3]
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
+; GCN-HSA-NEXT:    global_store_dwordx3 v3, v[0:2], s[0:1]
+; GCN-HSA-NEXT:    s_endpgm
 entry:
   %ld = load <3 x i32>, ptr addrspace(1) %in
   store <3 x i32> %ld, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}global_load_v4i32:
-; GCN-NOHSA: buffer_load_dwordx4
-; GCN-HSA: {{flat|global}}_load_dwordx4
-
-; EG: VTX_READ_128
 define amdgpu_kernel void @global_load_v4i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; SI-NOHSA-LABEL: global_load_v4i32:
+; SI-NOHSA:       ; %bb.0: ; %entry
+; SI-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NOHSA-NEXT:    s_mov_b32 s6, -1
+; SI-NOHSA-NEXT:    s_mov_b32 s10, s6
+; SI-NOHSA-NEXT:    s_mov_b32 s11, s7
+; SI-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NOHSA-NEXT:    s_mov_b32 s8, s2
+; SI-NOHSA-NEXT:    s_mov_b32 s9, s3
+; SI-NOHSA-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
+; SI-NOHSA-NEXT:    s_mov_b32 s4, s0
+; SI-NOHSA-NEXT:    s_mov_b32 s5, s1
+; SI-NOHSA-NEXT:    s_waitcnt vmcnt(0)
+; SI-NOHSA-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; SI-NOHSA-NEXT:    s_endpgm
+;
+; GCNX3-HSA-LABEL: global_load_v4i32:
+; GCNX3-HSA:       ; %bb.0: ; %entry
+; GCNX3-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCNX3-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s1
+; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(0)
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCNX3-HSA-NEXT:    s_endpgm
+;
+; GCNX3-NOHSA-LABEL: global_load_v4i32:
+; GCNX3-NOHSA:       ; %bb.0: ; %entry
+; GCNX3-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s6, -1
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s10, s6
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s11, s7
+; GCNX3-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s8, s2
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s9, s3
+; GCNX3-NOHSA-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s4, s0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s5, s1
+; GCNX3-NOHSA-NEXT:    s_waitcnt vmcnt(0)
+; GCNX3-NOHSA-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; GCNX3-NOHSA-NEXT:    s_endpgm
+;
+; EG-LABEL: global_load_v4i32:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 1, @9, 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:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; 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)
+;
+; GCN-HSA-LABEL: global_load_v4i32:
+; GCN-HSA:       ; %bb.0: ; %entry
+; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, 0
+; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-HSA-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3]
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
+; GCN-HSA-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GCN-HSA-NEXT:    s_endpgm
 entry:
   %ld = load <4 x i32>, ptr addrspace(1) %in
   store <4 x i32> %ld, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}global_load_v8i32:
-; GCN-NOHSA: buffer_load_dwordx4
-; GCN-NOHSA: buffer_load_dwordx4
-; GCN-HSA: {{flat|global}}_load_dwordx4
-; GCN-HSA: {{flat|global}}_load_dwordx4
-
-; EG: VTX_READ_128
-; EG: VTX_READ_128
 define amdgpu_kernel void @global_load_v8i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; SI-NOHSA-LABEL: global_load_v8i32:
+; SI-NOHSA:       ; %bb.0: ; %entry
+; SI-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NOHSA-NEXT:    s_mov_b32 s6, -1
+; SI-NOHSA-NEXT:    s_mov_b32 s10, s6
+; SI-NOHSA-NEXT:    s_mov_b32 s11, s7
+; SI-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NOHSA-NEXT:    s_mov_b32 s8, s2
+; SI-NOHSA-NEXT:    s_mov_b32 s9, s3
+; SI-NOHSA-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0 offset:16
+; SI-NOHSA-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0
+; SI-NOHSA-NEXT:    s_mov_b32 s4, s0
+; SI-NOHSA-NEXT:    s_mov_b32 s5, s1
+; SI-NOHSA-NEXT:    s_waitcnt vmcnt(1)
+; SI-NOHSA-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0 offset:16
+; SI-NOHSA-NEXT:    s_waitcnt vmcnt(1)
+; SI-NOHSA-NEXT:    buffer_store_dwordx4 v[4:7], off, s[4:7], 0
+; SI-NOHSA-NEXT:    s_endpgm
+;
+; GCNX3-HSA-LABEL: global_load_v8i32:
+; GCNX3-HSA:       ; %bb.0: ; %entry
+; GCNX3-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCNX3-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v5, s3
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v4, s2
+; GCNX3-HSA-NEXT:    s_add_u32 s2, s2, 16
+; GCNX3-HSA-NEXT:    s_addc_u32 s3, s3, 0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v0, s2
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v1, s3
+; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; GCNX3-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v9, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v8, s0
+; GCNX3-HSA-NEXT:    s_add_u32 s0, s0, 16
+; GCNX3-HSA-NEXT:    s_addc_u32 s1, s1, 0
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v11, s1
+; GCNX3-HSA-NEXT:    v_mov_b32_e32 v10, s0
+; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(1)
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
+; GCNX3-HSA-NEXT:    s_waitcnt vmcnt(1)
+; GCNX3-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[4:7]
+; GCNX3-HSA-NEXT:    s_endpgm
+;
+; GCNX3-NOHSA-LABEL: global_load_v8i32:
+; GCNX3-NOHSA:       ; %bb.0: ; %entry
+; GCNX3-NOHSA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s7, 0xf000
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s6, -1
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s10, s6
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s11, s7
+; GCNX3-NOHSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s8, s2
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s9, s3
+; GCNX3-NOHSA-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0 offset:16
+; GCNX3-NOHSA-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s4, s0
+; GCNX3-NOHSA-NEXT:    s_mov_b32 s5, s1
+; GCNX3-NOHSA-NEXT:    s_waitcnt vmcnt(1)
+; GCNX3-NOHSA-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0 offset:16
+; GCNX3-NOHSA-NEXT:    s_waitcnt vmcnt(1)
+; GCNX3-NOHSA-NEXT:    buffer_store_dwordx4 v[4:7], off, s[4:7], 0
+; GCNX3-NOHSA-NEXT:    s_endpgm
+;
+; EG-LABEL: global_load_v8i32:
+; EG:       ; %bb.0: ; %entry
+; EG-NEXT:    ALU 0, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 1 @6
+; EG-NEXT:    ALU 4, @11, 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:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_128 T1.XYZW, T0.X, 0, #1
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T0.X, 16, #1
+; EG-NEXT:    ALU clause starting at 10:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 11:
+; EG-NEXT:     ADD_INT * T2.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:     LSHR * T3.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+;
+; GCN-HSA-LABEL: global_load_v8i32:
+; GCN-HSA:       ; %bb.0: ; %entry
+; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, 0
+; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-HSA-NEXT:    global_load_dwordx4 v[0:3], v8, s[2:3] offset:16
+; GCN-HSA-NEXT:    global_load_dwordx4 v[4:7], v8, s[2:3]
+; ...
[truncated]

[jayfoad]

Seems reasonable to me.