brintos

brintos / llvm-project-archived public Read only

0
0
Text · 16.4 KiB · ee4b8da Raw
286 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc -mtriple=aarch64--linux-gnu -mattr=sve < %s | FileCheck %s3; RUN: llc -mtriple=aarch64--linux-gnu -mattr=sme -force-streaming < %s | FileCheck %s4 5; ld2b6define { <vscale x 16 x i8>, <vscale x 16 x i8> } @ld2.nxv32i8(<vscale x 16 x i1> %Pg, ptr %addr, i64 %a) {7; CHECK-LABEL: ld2.nxv32i8:8; CHECK:       // %bb.0:9; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0, x1]10; CHECK-NEXT:    ret11  %addr2 = getelementptr i8, ptr  %addr, i64 %a12  %res = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> %Pg, ptr %addr2)13  ret { <vscale x 16 x i8>, <vscale x 16 x i8> } %res14}15 16; ld2h17define { <vscale x 8 x i16>, <vscale x 8 x i16> } @ld2.nxv16i16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) {18; CHECK-LABEL: ld2.nxv16i16:19; CHECK:       // %bb.0:20; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, x1, lsl #1]21; CHECK-NEXT:    ret22  %addr2 = getelementptr i16, ptr  %addr, i64 %a23  %res = call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> %Pg, ptr %addr2)24  ret { <vscale x 8 x i16>, <vscale x 8 x i16> } %res25}26 27define { <vscale x 8 x half>, <vscale x 8 x half> } @ld2.nxv16f16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) {28; CHECK-LABEL: ld2.nxv16f16:29; CHECK:       // %bb.0:30; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, x1, lsl #1]31; CHECK-NEXT:    ret32  %addr2 = getelementptr half, ptr  %addr, i64 %a33  %res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> %Pg, ptr %addr2)34  ret { <vscale x 8 x half>, <vscale x 8 x half> } %res35}36 37define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ld2.nxv16bf16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) #0 {38; CHECK-LABEL: ld2.nxv16bf16:39; CHECK:       // %bb.0:40; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, x1, lsl #1]41; CHECK-NEXT:    ret42  %addr2 = getelementptr bfloat, ptr  %addr, i64 %a43  %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> %Pg, ptr %addr2)44  ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res45}46 47; ld2w48define { <vscale x 4 x i32>, <vscale x 4 x i32> } @ld2.nxv8i32(<vscale x 4 x i1> %Pg, ptr %addr, i64 %a) {49; CHECK-LABEL: ld2.nxv8i32:50; CHECK:       // %bb.0:51; CHECK-NEXT:    ld2w { z0.s, z1.s }, p0/z, [x0, x1, lsl #2]52; CHECK-NEXT:    ret53  %addr2 = getelementptr i32, ptr  %addr, i64 %a54  %res = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> %Pg, ptr %addr2)55  ret { <vscale x 4 x i32>, <vscale x 4 x i32> } %res56}57 58define { <vscale x 4 x float>, <vscale x 4 x float> } @ld2.nxv8f32(<vscale x 4 x i1> %Pg, ptr %addr, i64 %a) {59; CHECK-LABEL: ld2.nxv8f32:60; CHECK:       // %bb.0:61; CHECK-NEXT:    ld2w { z0.s, z1.s }, p0/z, [x0, x1, lsl #2]62; CHECK-NEXT:    ret63  %addr2 = getelementptr float, ptr  %addr, i64 %a64  %res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> %Pg, ptr %addr2)65  ret { <vscale x 4 x float>, <vscale x 4 x float> } %res66}67 68; ld2d69define { <vscale x 2 x i64>, <vscale x 2 x i64> } @ld2.nxv4i64(<vscale x 2 x i1> %Pg, ptr %addr, i64 %a) {70; CHECK-LABEL: ld2.nxv4i64:71; CHECK:       // %bb.0:72; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x0, x1, lsl #3]73; CHECK-NEXT:    ret74  %addr2 = getelementptr i64, ptr  %addr, i64 %a75  %res = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> %Pg, ptr %addr2)76  ret { <vscale x 2 x i64>, <vscale x 2 x i64> } %res77}78 79define { <vscale x 2 x double>, <vscale x 2 x double> } @ld2.nxv4f64(<vscale x 2 x i1> %Pg, ptr %addr, i64 %a) {80; CHECK-LABEL: ld2.nxv4f64:81; CHECK:       // %bb.0:82; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x0, x1, lsl #3]83; CHECK-NEXT:    ret84  %addr2 = getelementptr double, ptr  %addr, i64 %a85  %res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> %Pg, ptr %addr2)86  ret { <vscale x 2 x double>, <vscale x 2 x double> } %res87}88 89; ld3b90define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @ld3.nxv48i8(<vscale x 16 x i1> %Pg, ptr %addr, i64 %a) {91; CHECK-LABEL: ld3.nxv48i8:92; CHECK:       // %bb.0:93; CHECK-NEXT:    ld3b { z0.b - z2.b }, p0/z, [x0, x1]94; CHECK-NEXT:    ret95  %addr2 = getelementptr i8, ptr  %addr, i64 %a96  %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> %Pg, ptr %addr2)97  ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res98}99 100; ld3h101define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @ld3.nxv24i16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) {102; CHECK-LABEL: ld3.nxv24i16:103; CHECK:       // %bb.0:104; CHECK-NEXT:    ld3h { z0.h - z2.h }, p0/z, [x0, x1, lsl #1]105; CHECK-NEXT:    ret106  %addr2 = getelementptr i16, ptr  %addr, i64 %a107  %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> %Pg, ptr %addr2)108  ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res109}110 111define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @ld3.nxv24f16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) {112; CHECK-LABEL: ld3.nxv24f16:113; CHECK:       // %bb.0:114; CHECK-NEXT:    ld3h { z0.h - z2.h }, p0/z, [x0, x1, lsl #1]115; CHECK-NEXT:    ret116  %addr2 = getelementptr half, ptr  %addr, i64 %a117  %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> %Pg, ptr %addr2)118  ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res119}120 121define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ld3.nxv24bf16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) #0 {122; CHECK-LABEL: ld3.nxv24bf16:123; CHECK:       // %bb.0:124; CHECK-NEXT:    ld3h { z0.h - z2.h }, p0/z, [x0, x1, lsl #1]125; CHECK-NEXT:    ret126  %addr2 = getelementptr bfloat, ptr  %addr, i64 %a127  %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> %Pg, ptr %addr2)128  ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res129}130 131; ld3w132define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @ld3.nxv12i32(<vscale x 4 x i1> %Pg, ptr %addr, i64 %a) {133; CHECK-LABEL: ld3.nxv12i32:134; CHECK:       // %bb.0:135; CHECK-NEXT:    ld3w { z0.s - z2.s }, p0/z, [x0, x1, lsl #2]136; CHECK-NEXT:    ret137  %addr2 = getelementptr i32, ptr  %addr, i64 %a138  %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> %Pg, ptr %addr2)139  ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res140}141 142define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @ld3.nxv12f32(<vscale x 4 x i1> %Pg, ptr %addr, i64 %a) {143; CHECK-LABEL: ld3.nxv12f32:144; CHECK:       // %bb.0:145; CHECK-NEXT:    ld3w { z0.s - z2.s }, p0/z, [x0, x1, lsl #2]146; CHECK-NEXT:    ret147  %addr2 = getelementptr float, ptr  %addr, i64 %a148  %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> %Pg, ptr %addr2)149  ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res150}151 152; ld3d153define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @ld3.nxv6i64(<vscale x 2 x i1> %Pg, ptr %addr, i64 %a) {154; CHECK-LABEL: ld3.nxv6i64:155; CHECK:       // %bb.0:156; CHECK-NEXT:    ld3d { z0.d - z2.d }, p0/z, [x0, x1, lsl #3]157; CHECK-NEXT:    ret158  %addr2 = getelementptr i64, ptr  %addr, i64 %a159  %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> %Pg, ptr %addr2)160  ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } %res161}162 163define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @ld3.nxv6f64(<vscale x 2 x i1> %Pg, ptr %addr, i64 %a) {164; CHECK-LABEL: ld3.nxv6f64:165; CHECK:       // %bb.0:166; CHECK-NEXT:    ld3d { z0.d - z2.d }, p0/z, [x0, x1, lsl #3]167; CHECK-NEXT:    ret168  %addr2 = getelementptr double, ptr  %addr, i64 %a169  %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> %Pg, ptr %addr2)170  ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res171}172 173; ld4b174define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @ld4.nxv64i8(<vscale x 16 x i1> %Pg, ptr %addr, i64 %a) {175; CHECK-LABEL: ld4.nxv64i8:176; CHECK:       // %bb.0:177; CHECK-NEXT:    ld4b { z0.b - z3.b }, p0/z, [x0, x1]178; CHECK-NEXT:    ret179  %addr2 = getelementptr i8, ptr  %addr, i64 %a180  %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> %Pg, ptr %addr2)181  ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res182}183 184; ld4h185define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @ld4.nxv32i16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) {186; CHECK-LABEL: ld4.nxv32i16:187; CHECK:       // %bb.0:188; CHECK-NEXT:    ld4h { z0.h - z3.h }, p0/z, [x0, x1, lsl #1]189; CHECK-NEXT:    ret190  %addr2 = getelementptr i16, ptr  %addr, i64 %a191  %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> %Pg, ptr %addr2)192  ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res193}194 195define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @ld4.nxv32f16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) {196; CHECK-LABEL: ld4.nxv32f16:197; CHECK:       // %bb.0:198; CHECK-NEXT:    ld4h { z0.h - z3.h }, p0/z, [x0, x1, lsl #1]199; CHECK-NEXT:    ret200  %addr2 = getelementptr half, ptr  %addr, i64 %a201  %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> %Pg, ptr %addr2)202  ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res203}204 205define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ld4.nxv32bf16(<vscale x 8 x i1> %Pg, ptr %addr, i64 %a) #0 {206; CHECK-LABEL: ld4.nxv32bf16:207; CHECK:       // %bb.0:208; CHECK-NEXT:    ld4h { z0.h - z3.h }, p0/z, [x0, x1, lsl #1]209; CHECK-NEXT:    ret210  %addr2 = getelementptr bfloat, ptr  %addr, i64 %a211  %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> %Pg, ptr %addr2)212  ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res213}214 215; ld4w216define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @ld4.nxv16i32(<vscale x 4 x i1> %Pg, ptr %addr, i64 %a) {217; CHECK-LABEL: ld4.nxv16i32:218; CHECK:       // %bb.0:219; CHECK-NEXT:    ld4w { z0.s - z3.s }, p0/z, [x0, x1, lsl #2]220; CHECK-NEXT:    ret221  %addr2 = getelementptr i32, ptr  %addr, i64 %a222  %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> %Pg, ptr %addr2)223  ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res224}225 226define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @ld4.nxv16f32(<vscale x 4 x i1> %Pg, ptr %addr, i64 %a) {227; CHECK-LABEL: ld4.nxv16f32:228; CHECK:       // %bb.0:229; CHECK-NEXT:    ld4w { z0.s - z3.s }, p0/z, [x0, x1, lsl #2]230; CHECK-NEXT:    ret231  %addr2 = getelementptr float, ptr  %addr, i64 %a232  %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> %Pg, ptr %addr2)233  ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res234}235 236; ld4d237define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @ld4.nxv8i64(<vscale x 2 x i1> %Pg, ptr %addr, i64 %a) {238; CHECK-LABEL: ld4.nxv8i64:239; CHECK:       // %bb.0:240; CHECK-NEXT:    ld4d { z0.d - z3.d }, p0/z, [x0, x1, lsl #3]241; CHECK-NEXT:    ret242  %addr2 = getelementptr i64, ptr  %addr, i64 %a243  %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> %Pg, ptr %addr2)244  ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } %res245}246 247define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @ld4.nxv8f64(<vscale x 2 x i1> %Pg, ptr %addr, i64 %a) {248; CHECK-LABEL: ld4.nxv8f64:249; CHECK:       // %bb.0:250; CHECK-NEXT:    ld4d { z0.d - z3.d }, p0/z, [x0, x1, lsl #3]251; CHECK-NEXT:    ret252  %addr2 = getelementptr double, ptr  %addr, i64 %a253  %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> %Pg, ptr %addr2)254  ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res255}256 257declare { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1>, ptr)258declare { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1>, ptr)259declare { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1>, ptr)260declare { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1>, ptr)261declare { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1>, ptr)262declare { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1>, ptr)263declare { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1>, ptr)264declare { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1>, ptr)265 266declare { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1>, ptr)267declare { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1>, ptr)268declare { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1>, ptr)269declare { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1>, ptr)270declare { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1>, ptr)271declare { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1>, ptr)272declare { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1>, ptr)273declare { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1>, ptr)274 275declare { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1>, ptr)276declare { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1>, ptr)277declare { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1>, ptr)278declare { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1>, ptr)279declare { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1>, ptr)280declare { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1>, ptr)281declare { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1>, ptr)282declare { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1>, ptr)283 284; +bf16 is required for the bfloat version.285attributes #0 = { "target-features"="+bf16" }286