1; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2 < %s 2>%t | FileCheck %s 2; RUN: FileCheck --check-prefix=WARN --allow-empty %s <%t 3 4; If this check fails please read test/CodeGen/AArch64/README for instructions on how to resolve it. 5; WARN-NOT: warning 6 7; 8; TBL2 9; 10 11define <vscale x 16 x i8> @tbl2_b(<vscale x 16 x i8> %a, <vscale x 16 x i8> %unused, 12 <vscale x 16 x i8> %b, <vscale x 16 x i8> %c) { 13; CHECK-LABEL: tbl2_b: 14; CHECK: mov z1.d, z0.d 15; CHECK-NEXT: tbl z0.b, { z1.b, z2.b }, z3.b 16; CHECK-NEXT: ret 17 %out = call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> %a, 18 <vscale x 16 x i8> %b, 19 <vscale x 16 x i8> %c) 20 ret <vscale x 16 x i8> %out 21} 22 23define <vscale x 8 x i16> @tbl2_h(<vscale x 8 x i16> %a, <vscale x 16 x i8> %unused, 24 <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) { 25; CHECK-LABEL: tbl2_h: 26; CHECK: mov z1.d, z0.d 27; CHECK-NEXT: tbl z0.h, { z1.h, z2.h }, z3.h 28; CHECK-NEXT: ret 29 %out = call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> %a, 30 <vscale x 8 x i16> %b, 31 <vscale x 8 x i16> %c) 32 ret <vscale x 8 x i16> %out 33} 34 35define <vscale x 4 x i32> @tbl2_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %unused, 36 <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) { 37; CHECK-LABEL: tbl2_s: 38; CHECK: mov z1.d, z0.d 39; CHECK-NEXT: tbl z0.s, { z1.s, z2.s }, z3.s 40; CHECK-NEXT: ret 41 %out = call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> %a, 42 <vscale x 4 x i32> %b, 43 <vscale x 4 x i32> %c) 44 ret <vscale x 4 x i32> %out 45} 46 47define <vscale x 2 x i64> @tbl2_d(<vscale x 2 x i64> %a, <vscale x 2 x i64> %unused, 48 <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) { 49; CHECK-LABEL: tbl2_d: 50; CHECK: mov z1.d, z0.d 51; CHECK-NEXT: tbl z0.d, { z1.d, z2.d }, z3.d 52; CHECK-NEXT: ret 53 %out = call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> %a, 54 <vscale x 2 x i64> %b, 55 <vscale x 2 x i64> %c) 56 ret <vscale x 2 x i64> %out 57} 58 59define <vscale x 8 x half> @tbl2_fh(<vscale x 8 x half> %a, <vscale x 8 x half> %unused, 60 <vscale x 8 x half> %b, <vscale x 8 x i16> %c) { 61; CHECK-LABEL: tbl2_fh: 62; CHECK: mov z1.d, z0.d 63; CHECK-NEXT: tbl z0.h, { z1.h, z2.h }, z3.h 64; CHECK-NEXT: ret 65 %out = call <vscale x 8 x half> @llvm.aarch64.sve.tbl2.nxv8f16(<vscale x 8 x half> %a, 66 <vscale x 8 x half> %b, 67 <vscale x 8 x i16> %c) 68 ret <vscale x 8 x half> %out 69} 70 71define <vscale x 4 x float> @tbl2_fs(<vscale x 4 x float> %a, <vscale x 4 x float> %unused, 72 <vscale x 4 x float> %b, <vscale x 4 x i32> %c) { 73; CHECK-LABEL: tbl2_fs: 74; CHECK: mov z1.d, z0.d 75; CHECK-NEXT: tbl z0.s, { z1.s, z2.s }, z3.s 76; CHECK-NEXT: ret 77 %out = call <vscale x 4 x float> @llvm.aarch64.sve.tbl2.nxv4f32(<vscale x 4 x float> %a, 78 <vscale x 4 x float> %b, 79 <vscale x 4 x i32> %c) 80 ret <vscale x 4 x float> %out 81} 82 83define <vscale x 2 x double> @tbl2_fd(<vscale x 2 x double> %a, <vscale x 2 x double> %unused, 84 <vscale x 2 x double> %b, <vscale x 2 x i64> %c) { 85; CHECK-LABEL: tbl2_fd: 86; CHECK: mov z1.d, z0.d 87; CHECK-NEXT: tbl z0.d, { z1.d, z2.d }, z3.d 88; CHECK-NEXT: ret 89 %out = call <vscale x 2 x double> @llvm.aarch64.sve.tbl2.nxv2f64(<vscale x 2 x double> %a, 90 <vscale x 2 x double> %b, 91 <vscale x 2 x i64> %c) 92 ret <vscale x 2 x double> %out 93} 94 95; 96; TBX 97; 98 99define <vscale x 16 x i8> @tbx_b(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b, <vscale x 16 x i8> %c) { 100; CHECK-LABEL: tbx_b: 101; CHECK: tbx z0.b, z1.b, z2.b 102; CHECK-NEXT: ret 103 %out = call <vscale x 16 x i8> @llvm.aarch64.sve.tbx.nxv16i8(<vscale x 16 x i8> %a, 104 <vscale x 16 x i8> %b, 105 <vscale x 16 x i8> %c) 106 ret <vscale x 16 x i8> %out 107} 108 109define <vscale x 8 x i16> @tbx_h(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) { 110; CHECK-LABEL: tbx_h: 111; CHECK: tbx z0.h, z1.h, z2.h 112; CHECK-NEXT: ret 113 %out = call <vscale x 8 x i16> @llvm.aarch64.sve.tbx.nxv8i16(<vscale x 8 x i16> %a, 114 <vscale x 8 x i16> %b, 115 <vscale x 8 x i16> %c) 116 ret <vscale x 8 x i16> %out 117} 118 119define <vscale x 8 x half> @ftbx_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x i16> %c) { 120; CHECK-LABEL: ftbx_h: 121; CHECK: tbx z0.h, z1.h, z2.h 122; CHECK-NEXT: ret 123 %out = call <vscale x 8 x half> @llvm.aarch64.sve.tbx.nxv8f16(<vscale x 8 x half> %a, 124 <vscale x 8 x half> %b, 125 <vscale x 8 x i16> %c) 126 ret <vscale x 8 x half> %out 127} 128 129define <vscale x 8 x bfloat> @ftbx_h_bf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, <vscale x 8 x i16> %c) #0 { 130; CHECK-LABEL: ftbx_h_bf16: 131; CHECK: tbx z0.h, z1.h, z2.h 132; CHECK-NEXT: ret 133 %out = call <vscale x 8 x bfloat> @llvm.aarch64.sve.tbx.nxv8bf16(<vscale x 8 x bfloat> %a, 134 <vscale x 8 x bfloat> %b, 135 <vscale x 8 x i16> %c) 136 ret <vscale x 8 x bfloat> %out 137} 138 139define <vscale x 4 x i32> @tbx_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) { 140; CHECK-LABEL: tbx_s: 141; CHECK: tbx z0.s, z1.s, z2.s 142; CHECK-NEXT: ret 143 %out = call <vscale x 4 x i32> @llvm.aarch64.sve.tbx.nxv4i32(<vscale x 4 x i32> %a, 144 <vscale x 4 x i32> %b, 145 <vscale x 4 x i32> %c) 146 ret <vscale x 4 x i32> %out 147} 148 149define <vscale x 4 x float> @ftbx_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x i32> %c) { 150; CHECK-LABEL: ftbx_s: 151; CHECK: tbx z0.s, z1.s, z2.s 152; CHECK-NEXT: ret 153 %out = call <vscale x 4 x float> @llvm.aarch64.sve.tbx.nxv4f32(<vscale x 4 x float> %a, 154 <vscale x 4 x float> %b, 155 <vscale x 4 x i32> %c) 156 ret <vscale x 4 x float> %out 157} 158 159define <vscale x 2 x i64> @tbx_d(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) { 160; CHECK-LABEL: tbx_d: 161; CHECK: tbx z0.d, z1.d, z2.d 162; CHECK-NEXT: ret 163 %out = call <vscale x 2 x i64> @llvm.aarch64.sve.tbx.nxv2i64(<vscale x 2 x i64> %a, 164 <vscale x 2 x i64> %b, 165 <vscale x 2 x i64> %c) 166 ret <vscale x 2 x i64> %out 167} 168 169define <vscale x 2 x double> @ftbx_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x i64> %c) { 170; CHECK-LABEL: ftbx_d: 171; CHECK: tbx z0.d, z1.d, z2.d 172; CHECK-NEXT: ret 173 %out = call <vscale x 2 x double> @llvm.aarch64.sve.tbx.nxv2f64(<vscale x 2 x double> %a, 174 <vscale x 2 x double> %b, 175 <vscale x 2 x i64> %c) 176 ret <vscale x 2 x double> %out 177} 178 179declare <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>) 180declare <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>) 181declare <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>) 182declare <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>) 183 184declare <vscale x 8 x half> @llvm.aarch64.sve.tbl2.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i16>) 185declare <vscale x 4 x float> @llvm.aarch64.sve.tbl2.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i32>) 186declare <vscale x 2 x double> @llvm.aarch64.sve.tbl2.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i64>) 187 188declare <vscale x 16 x i8> @llvm.aarch64.sve.tbx.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>) 189declare <vscale x 8 x i16> @llvm.aarch64.sve.tbx.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>) 190declare <vscale x 4 x i32> @llvm.aarch64.sve.tbx.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>) 191declare <vscale x 2 x i64> @llvm.aarch64.sve.tbx.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>) 192 193declare <vscale x 8 x half> @llvm.aarch64.sve.tbx.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x i16>) 194declare <vscale x 4 x float> @llvm.aarch64.sve.tbx.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x i32>) 195declare <vscale x 2 x double> @llvm.aarch64.sve.tbx.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x i64>) 196 197declare <vscale x 8 x bfloat> @llvm.aarch64.sve.tbx.nxv8bf16(<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x i16>) 198 199; +bf16 is required for the bfloat version. 200attributes #0 = { "target-features"="+sve,+bf16" } 201