1;RUN: llc < %s -march=amdgcn -mcpu=gfx600 -verify-machineinstrs | FileCheck %s -check-prefixes=GCN,SI
2;RUN: llc < %s -march=amdgcn -mcpu=gfx700 -verify-machineinstrs | FileCheck %s -check-prefixes=GCN,GCNX3
3
4; GCN-LABEL: {{^}}tbuffer_raw_load_immoffs_x3:
5; SI: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] offset:42
6; GCNX3: tbuffer_load_format_xyz {{v\[[0-9]+:[0-9]+\]}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] offset:42
7define amdgpu_vs <3 x float> @tbuffer_raw_load_immoffs_x3(<4 x i32> inreg) {
8main_body:
9    %vdata   = call <3 x i32> @llvm.amdgcn.raw.tbuffer.load.v3i32(<4 x i32> %0, i32 42, i32 0, i32 78, i32 0)
10    %vdata.f = bitcast <3 x i32> %vdata to <3 x float>
11    ret <3 x float> %vdata.f
12}
13
14
15; GCN-LABEL: {{^}}tbuffer_struct_load_immoffs_x3:
16; GCN: v_mov_b32_e32 [[ZEROREG:v[0-9]+]], 0
17; SI: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] idxen offset:42
18; GCNX3: tbuffer_load_format_xyz {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] idxen offset:42
19define amdgpu_vs <3 x float> @tbuffer_struct_load_immoffs_x3(<4 x i32> inreg) {
20main_body:
21    %vdata   = call <3 x i32> @llvm.amdgcn.struct.tbuffer.load.v3i32(<4 x i32> %0, i32 0, i32 42, i32 0, i32 78, i32 0)
22    %vdata.f = bitcast <3 x i32> %vdata to <3 x float>
23    ret <3 x float> %vdata.f
24}
25
26
27; GCN-LABEL: {{^}}tbuffer_load_format_immoffs_x3:
28; SI: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] offset:42
29; GCNX3: tbuffer_load_format_xyz {{v\[[0-9]+:[0-9]+\]}}, off, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] offset:42
30define amdgpu_vs <3 x float> @tbuffer_load_format_immoffs_x3(<4 x i32> inreg) {
31main_body:
32    %vdata   = call <3 x i32> @llvm.amdgcn.tbuffer.load.v3i32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 42, i32 14, i32 4, i1 0, i1 0)
33    %vdata.f = bitcast <3 x i32> %vdata to <3 x float>
34    ret <3 x float> %vdata.f
35}
36
37declare <3 x i32> @llvm.amdgcn.raw.tbuffer.load.v3i32(<4 x i32>, i32, i32, i32, i32)
38declare <3 x i32> @llvm.amdgcn.struct.tbuffer.load.v3i32(<4 x i32>, i32, i32, i32, i32, i32)
39declare <3 x i32> @llvm.amdgcn.tbuffer.load.v3i32(<4 x i32>, i32, i32, i32, i32, i32, i32, i1, i1)
40
41