1;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck -check-prefixes=GCN,PREGFX10 %s 2;RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefixes=GCN,PREGFX10 %s 3;RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck -check-prefixes=GCN,GFX10 %s 4 5; GCN-LABEL: {{^}}tbuffer_load: 6; GCN: v_mov_b32_e32 [[ZEROREG:v[0-9]+]], 0 7; PREGFX10: 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 8; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_RESERVED_15,BUF_NUM_FORMAT_SSCALED] idxen glc 9; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_10_11_11,BUF_NUM_FORMAT_SNORM] idxen slc 10; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_10_11_11,BUF_NUM_FORMAT_SNORM] idxen glc 11; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:78 idxen 12; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_FMT_32_32_SINT] idxen glc 13; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_FMT_32_FLOAT] idxen slc 14; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_FMT_32_FLOAT] idxen glc dlc 15; GCN: s_waitcnt 16define amdgpu_vs {<4 x float>, <4 x float>, <4 x float>, <4 x float>} @tbuffer_load(<4 x i32> inreg) { 17main_body: 18 %vdata = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 78, i32 0) 19 %vdata_glc = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 63, i32 1) 20 %vdata_slc = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 22, i32 2) 21 %vdata_f32 = call <4 x float> @llvm.amdgcn.struct.tbuffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 22, i32 5) 22 %vdata.f = bitcast <4 x i32> %vdata to <4 x float> 23 %vdata_glc.f = bitcast <4 x i32> %vdata_glc to <4 x float> 24 %vdata_slc.f = bitcast <4 x i32> %vdata_slc to <4 x float> 25 %r0 = insertvalue {<4 x float>, <4 x float>, <4 x float>, <4 x float>} undef, <4 x float> %vdata.f, 0 26 %r1 = insertvalue {<4 x float>, <4 x float>, <4 x float>, <4 x float>} %r0, <4 x float> %vdata_glc.f, 1 27 %r2 = insertvalue {<4 x float>, <4 x float>, <4 x float>, <4 x float>} %r1, <4 x float> %vdata_slc.f, 2 28 %r3 = insertvalue {<4 x float>, <4 x float>, <4 x float>, <4 x float>} %r2, <4 x float> %vdata_f32, 3 29 ret {<4 x float>, <4 x float>, <4 x float>, <4 x float>} %r3 30} 31 32; GCN-LABEL: {{^}}tbuffer_load_immoffs: 33; GCN: v_mov_b32_e32 [[ZEROREG:v[0-9]+]], 0 34; PREGFX10: 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 35; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 0 format:78 idxen offset:42 36define amdgpu_vs <4 x float> @tbuffer_load_immoffs(<4 x i32> inreg) { 37main_body: 38 %vdata = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 42, i32 0, i32 78, i32 0) 39 %vdata.f = bitcast <4 x i32> %vdata to <4 x float> 40 ret <4 x float> %vdata.f 41} 42 43; GCN-LABEL: {{^}}tbuffer_load_immoffs_large 44; GCN: v_mov_b32_e32 [[ZEROREG:v[0-9]+]], 0 45; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 61 format:[BUF_DATA_FORMAT_RESERVED_15,BUF_NUM_FORMAT_USCALED] idxen offset:4095 46; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_SSCALED] idxen offset:73 47; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} format:[BUF_DATA_FORMAT_32_32_32,BUF_NUM_FORMAT_UINT] idxen offset:1 48; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, 61 format:[BUF_FMT_10_10_10_2_SSCALED] idxen offset:4095 49; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} format:[BUF_FMT_32_32_UINT] idxen offset:73 50; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, [[ZEROREG]], {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} format:[BUF_FMT_32_32_32_32_FLOAT] idxen offset:1 51; GCN: s_waitcnt 52define amdgpu_vs {<4 x float>, <4 x float>, <4 x float>} @tbuffer_load_immoffs_large(<4 x i32> inreg, i32 inreg %soffs) { 53 %vdata = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 4095, i32 61, i32 47, i32 0) 54 %vdata_glc = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 73, i32 %soffs, i32 62, i32 0) 55 %vdata_slc = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 1, i32 %soffs, i32 77, i32 0) 56 %vdata.f = bitcast <4 x i32> %vdata to <4 x float> 57 %vdata_glc.f = bitcast <4 x i32> %vdata_glc to <4 x float> 58 %vdata_slc.f = bitcast <4 x i32> %vdata_slc to <4 x float> 59 %r0 = insertvalue {<4 x float>, <4 x float>, <4 x float>} undef, <4 x float> %vdata.f, 0 60 %r1 = insertvalue {<4 x float>, <4 x float>, <4 x float>} %r0, <4 x float> %vdata_glc.f, 1 61 %r2 = insertvalue {<4 x float>, <4 x float>, <4 x float>} %r1, <4 x float> %vdata_slc.f, 2 62 ret {<4 x float>, <4 x float>, <4 x float>} %r2 63} 64 65; GCN-LABEL: {{^}}tbuffer_load_idx: 66; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] idxen 67; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:78 idxen 68define amdgpu_vs <4 x float> @tbuffer_load_idx(<4 x i32> inreg, i32 %vindex) { 69main_body: 70 %vdata = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 %vindex, i32 0, i32 0, i32 78, i32 0) 71 %vdata.f = bitcast <4 x i32> %vdata to <4 x float> 72 ret <4 x float> %vdata.f 73} 74 75; GCN-LABEL: {{^}}tbuffer_load_ofs: 76; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] idxen offen 77; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:78 idxen offen 78define amdgpu_vs <4 x float> @tbuffer_load_ofs(<4 x i32> inreg, i32 %voffs) { 79main_body: 80 %vdata = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 %voffs, i32 0, i32 78, i32 0) 81 %vdata.f = bitcast <4 x i32> %vdata to <4 x float> 82 ret <4 x float> %vdata.f 83} 84 85; GCN-LABEL: {{^}}tbuffer_load_ofs_imm: 86; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] idxen offen offset:52 87; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:78 idxen offen offset:52 88define amdgpu_vs <4 x float> @tbuffer_load_ofs_imm(<4 x i32> inreg, i32 %voffs) { 89main_body: 90 %ofs = add i32 %voffs, 52 91 %vdata = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 0, i32 %ofs, i32 0, i32 78, i32 0) 92 %vdata.f = bitcast <4 x i32> %vdata to <4 x float> 93 ret <4 x float> %vdata.f 94} 95 96; GCN-LABEL: {{^}}tbuffer_load_both: 97; PREGFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32_32,BUF_NUM_FORMAT_UINT] idxen offen 98; GFX10: tbuffer_load_format_xyzw {{v\[[0-9]+:[0-9]+\]}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:78 idxen offen 99define amdgpu_vs <4 x float> @tbuffer_load_both(<4 x i32> inreg, i32 %vindex, i32 %voffs) { 100main_body: 101 %vdata = call <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32> %0, i32 %vindex, i32 %voffs, i32 0, i32 78, i32 0) 102 %vdata.f = bitcast <4 x i32> %vdata to <4 x float> 103 ret <4 x float> %vdata.f 104} 105 106 107; GCN-LABEL: {{^}}buffer_load_xy: 108; PREGFX10: tbuffer_load_format_xy {{v\[[0-9]+:[0-9]+\]}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32,BUF_NUM_FORMAT_UINT] idxen 109; GFX10: tbuffer_load_format_xy {{v\[[0-9]+:[0-9]+\]}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_FMT_32_32_32_32_FLOAT] idxen 110define amdgpu_vs <2 x float> @buffer_load_xy(<4 x i32> inreg %rsrc) { 111 %vdata = call <2 x i32> @llvm.amdgcn.struct.tbuffer.load.v2i32(<4 x i32> %rsrc, i32 0, i32 0, i32 0, i32 77, i32 0) 112 %vdata.f = bitcast <2 x i32> %vdata to <2 x float> 113 ret <2 x float> %vdata.f 114} 115 116; GCN-LABEL: {{^}}buffer_load_x: 117; PREGFX10: tbuffer_load_format_x {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_DATA_FORMAT_32_32_32,BUF_NUM_FORMAT_UINT] idxen 118; GFX10: tbuffer_load_format_x {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 format:[BUF_FMT_32_32_32_32_FLOAT] idxen 119define amdgpu_vs float @buffer_load_x(<4 x i32> inreg %rsrc) { 120 %vdata = call i32 @llvm.amdgcn.struct.tbuffer.load.i32(<4 x i32> %rsrc, i32 0, i32 0, i32 0, i32 77, i32 0) 121 %vdata.f = bitcast i32 %vdata to float 122 ret float %vdata.f 123} 124 125declare i32 @llvm.amdgcn.struct.tbuffer.load.i32(<4 x i32>, i32, i32, i32, i32, i32) 126declare <2 x i32> @llvm.amdgcn.struct.tbuffer.load.v2i32(<4 x i32>, i32, i32, i32, i32, i32) 127declare <4 x i32> @llvm.amdgcn.struct.tbuffer.load.v4i32(<4 x i32>, i32, i32, i32, i32, i32) 128declare <4 x float> @llvm.amdgcn.struct.tbuffer.load.v4f32(<4 x i32>, i32, i32, i32, i32, i32) 129 130