374 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 22; RUN: opt < %s -interleaved-access -S | FileCheck %s --check-prefix=NEON3; RUN: opt < %s -interleaved-access -mtriple=aarch64-linux-gnu -mattr=+sve -force-streaming-compatible -S | FileCheck %s --check-prefix=SVE-FIXED4; RUN: opt < %s -passes=interleaved-access -S | FileCheck %s --check-prefix=NEON5; RUN: opt < %s -passes=interleaved-access -mtriple=aarch64-linux-gnu -mattr=+sve -force-streaming-compatible -S | FileCheck %s --check-prefix=SVE-FIXED6 7target triple = "aarch64-linux-gnu"8 9define void @deinterleave_i8_factor2(ptr %ptr) {10; NEON-LABEL: define void @deinterleave_i8_factor211; NEON-SAME: (ptr [[PTR:%.*]]) {12; NEON-NEXT: [[LDN:%.*]] = call { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2.v16i8.p0(ptr [[PTR]])13; NEON-NEXT: [[EXTRACT1:%.*]] = extractvalue { <16 x i8>, <16 x i8> } [[LDN]], 014; NEON-NEXT: [[EXTRACT2:%.*]] = extractvalue { <16 x i8>, <16 x i8> } [[LDN]], 115; NEON-NEXT: ret void16;17; SVE-FIXED-LABEL: define void @deinterleave_i8_factor218; SVE-FIXED-SAME: (ptr [[PTR:%.*]]) #[[ATTR0:[0-9]+]] {19; SVE-FIXED-NEXT: [[LOAD:%.*]] = load <32 x i8>, ptr [[PTR]], align 120; SVE-FIXED-NEXT: [[DEINTERLEAVE:%.*]] = tail call { <16 x i8>, <16 x i8> } @llvm.vector.deinterleave2.v32i8(<32 x i8> [[LOAD]])21; SVE-FIXED-NEXT: [[EXTRACT1:%.*]] = extractvalue { <16 x i8>, <16 x i8> } [[DEINTERLEAVE]], 022; SVE-FIXED-NEXT: [[EXTRACT2:%.*]] = extractvalue { <16 x i8>, <16 x i8> } [[DEINTERLEAVE]], 123; SVE-FIXED-NEXT: ret void24;25 %load = load <32 x i8>, ptr %ptr, align 126 %deinterleave = tail call { <16 x i8>, <16 x i8> } @llvm.vector.deinterleave2.v32i8(<32 x i8> %load)27 %extract1 = extractvalue { <16 x i8>, <16 x i8> } %deinterleave, 028 %extract2 = extractvalue { <16 x i8>, <16 x i8> } %deinterleave, 129 ret void30}31 32define void @deinterleave_i16_factor2(ptr %ptr) {33; NEON-LABEL: define void @deinterleave_i16_factor234; NEON-SAME: (ptr [[PTR:%.*]]) {35; NEON-NEXT: [[LDN:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2.v8i16.p0(ptr [[PTR]])36; NEON-NEXT: [[EXTRACT1:%.*]] = extractvalue { <8 x i16>, <8 x i16> } [[LDN]], 037; NEON-NEXT: [[EXTRACT2:%.*]] = extractvalue { <8 x i16>, <8 x i16> } [[LDN]], 138; NEON-NEXT: ret void39;40; SVE-FIXED-LABEL: define void @deinterleave_i16_factor241; SVE-FIXED-SAME: (ptr [[PTR:%.*]]) #[[ATTR0]] {42; SVE-FIXED-NEXT: [[LOAD:%.*]] = load <16 x i16>, ptr [[PTR]], align 243; SVE-FIXED-NEXT: [[DEINTERLEAVE:%.*]] = tail call { <8 x i16>, <8 x i16> } @llvm.vector.deinterleave2.v16i16(<16 x i16> [[LOAD]])44; SVE-FIXED-NEXT: [[EXTRACT1:%.*]] = extractvalue { <8 x i16>, <8 x i16> } [[DEINTERLEAVE]], 045; SVE-FIXED-NEXT: [[EXTRACT2:%.*]] = extractvalue { <8 x i16>, <8 x i16> } [[DEINTERLEAVE]], 146; SVE-FIXED-NEXT: ret void47;48 %load = load <16 x i16>, ptr %ptr, align 249 %deinterleave = tail call { <8 x i16>, <8 x i16> } @llvm.vector.deinterleave2.v16i16(<16 x i16> %load)50 %extract1 = extractvalue { <8 x i16>, <8 x i16> } %deinterleave, 051 %extract2 = extractvalue { <8 x i16>, <8 x i16> } %deinterleave, 152 ret void53}54 55define void @deinterleave_8xi32_factor2(ptr %ptr) {56; NEON-LABEL: define void @deinterleave_8xi32_factor257; NEON-SAME: (ptr [[PTR:%.*]]) {58; NEON-NEXT: [[LDN:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0(ptr [[PTR]])59; NEON-NEXT: [[EXTRACT1:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[LDN]], 060; NEON-NEXT: [[EXTRACT2:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[LDN]], 161; NEON-NEXT: ret void62;63; SVE-FIXED-LABEL: define void @deinterleave_8xi32_factor264; SVE-FIXED-SAME: (ptr [[PTR:%.*]]) #[[ATTR0]] {65; SVE-FIXED-NEXT: [[LOAD:%.*]] = load <8 x i32>, ptr [[PTR]], align 466; SVE-FIXED-NEXT: [[DEINTERLEAVE:%.*]] = tail call { <4 x i32>, <4 x i32> } @llvm.vector.deinterleave2.v8i32(<8 x i32> [[LOAD]])67; SVE-FIXED-NEXT: [[EXTRACT1:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[DEINTERLEAVE]], 068; SVE-FIXED-NEXT: [[EXTRACT2:%.*]] = extractvalue { <4 x i32>, <4 x i32> } [[DEINTERLEAVE]], 169; SVE-FIXED-NEXT: ret void70;71 %load = load <8 x i32>, ptr %ptr, align 472 %deinterleave = tail call { <4 x i32>, <4 x i32> } @llvm.vector.deinterleave2.v8i32(<8 x i32> %load)73 %extract1 = extractvalue { <4 x i32>, <4 x i32> } %deinterleave, 074 %extract2 = extractvalue { <4 x i32>, <4 x i32> } %deinterleave, 175 ret void76}77 78define void @deinterleave_i64_factor2(ptr %ptr) {79; NEON-LABEL: define void @deinterleave_i64_factor280; NEON-SAME: (ptr [[PTR:%.*]]) {81; NEON-NEXT: [[LDN:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2.v2i64.p0(ptr [[PTR]])82; NEON-NEXT: [[EXTRACT1:%.*]] = extractvalue { <2 x i64>, <2 x i64> } [[LDN]], 083; NEON-NEXT: [[EXTRACT2:%.*]] = extractvalue { <2 x i64>, <2 x i64> } [[LDN]], 184; NEON-NEXT: ret void85;86; SVE-FIXED-LABEL: define void @deinterleave_i64_factor287; SVE-FIXED-SAME: (ptr [[PTR:%.*]]) #[[ATTR0]] {88; SVE-FIXED-NEXT: [[LOAD:%.*]] = load <4 x i64>, ptr [[PTR]], align 889; SVE-FIXED-NEXT: [[DEINTERLEAVE:%.*]] = tail call { <2 x i64>, <2 x i64> } @llvm.vector.deinterleave2.v4i64(<4 x i64> [[LOAD]])90; SVE-FIXED-NEXT: [[EXTRACT1:%.*]] = extractvalue { <2 x i64>, <2 x i64> } [[DEINTERLEAVE]], 091; SVE-FIXED-NEXT: [[EXTRACT2:%.*]] = extractvalue { <2 x i64>, <2 x i64> } [[DEINTERLEAVE]], 192; SVE-FIXED-NEXT: ret void93;94 %load = load <4 x i64>, ptr %ptr, align 895 %deinterleave = tail call { <2 x i64>, <2 x i64> } @llvm.vector.deinterleave2.v4i64(<4 x i64> %load)96 %extract1 = extractvalue { <2 x i64>, <2 x i64> } %deinterleave, 097 %extract2 = extractvalue { <2 x i64>, <2 x i64> } %deinterleave, 198 ret void99}100 101define void @deinterleave_float_factor2(ptr %ptr) {102; NEON-LABEL: define void @deinterleave_float_factor2103; NEON-SAME: (ptr [[PTR:%.*]]) {104; NEON-NEXT: [[LDN:%.*]] = call { <4 x float>, <4 x float> } @llvm.aarch64.neon.ld2.v4f32.p0(ptr [[PTR]])105; NEON-NEXT: [[EXTRACT1:%.*]] = extractvalue { <4 x float>, <4 x float> } [[LDN]], 0106; NEON-NEXT: [[EXTRACT2:%.*]] = extractvalue { <4 x float>, <4 x float> } [[LDN]], 1107; NEON-NEXT: ret void108;109; SVE-FIXED-LABEL: define void @deinterleave_float_factor2110; SVE-FIXED-SAME: (ptr [[PTR:%.*]]) #[[ATTR0]] {111; SVE-FIXED-NEXT: [[LOAD:%.*]] = load <8 x float>, ptr [[PTR]], align 4112; SVE-FIXED-NEXT: [[DEINTERLEAVE:%.*]] = tail call { <4 x float>, <4 x float> } @llvm.vector.deinterleave2.v8f32(<8 x float> [[LOAD]])113; SVE-FIXED-NEXT: [[EXTRACT1:%.*]] = extractvalue { <4 x float>, <4 x float> } [[DEINTERLEAVE]], 0114; SVE-FIXED-NEXT: [[EXTRACT2:%.*]] = extractvalue { <4 x float>, <4 x float> } [[DEINTERLEAVE]], 1115; SVE-FIXED-NEXT: ret void116;117 %load = load <8 x float>, ptr %ptr, align 4118 %deinterleave = tail call { <4 x float>, <4 x float> } @llvm.vector.deinterleave2.v8f32(<8 x float> %load)119 %extract1 = extractvalue { <4 x float>, <4 x float> } %deinterleave, 0120 %extract2 = extractvalue { <4 x float>, <4 x float> } %deinterleave, 1121 ret void122}123 124define void @deinterleave_double_factor2(ptr %ptr) {125; NEON-LABEL: define void @deinterleave_double_factor2126; NEON-SAME: (ptr [[PTR:%.*]]) {127; NEON-NEXT: [[LDN:%.*]] = call { <2 x double>, <2 x double> } @llvm.aarch64.neon.ld2.v2f64.p0(ptr [[PTR]])128; NEON-NEXT: [[EXTRACT1:%.*]] = extractvalue { <2 x double>, <2 x double> } [[LDN]], 0129; NEON-NEXT: [[EXTRACT2:%.*]] = extractvalue { <2 x double>, <2 x double> } [[LDN]], 1130; NEON-NEXT: ret void131;132; SVE-FIXED-LABEL: define void @deinterleave_double_factor2133; SVE-FIXED-SAME: (ptr [[PTR:%.*]]) #[[ATTR0]] {134; SVE-FIXED-NEXT: [[LOAD:%.*]] = load <4 x double>, ptr [[PTR]], align 8135; SVE-FIXED-NEXT: [[DEINTERLEAVE:%.*]] = tail call { <2 x double>, <2 x double> } @llvm.vector.deinterleave2.v4f64(<4 x double> [[LOAD]])136; SVE-FIXED-NEXT: [[EXTRACT1:%.*]] = extractvalue { <2 x double>, <2 x double> } [[DEINTERLEAVE]], 0137; SVE-FIXED-NEXT: [[EXTRACT2:%.*]] = extractvalue { <2 x double>, <2 x double> } [[DEINTERLEAVE]], 1138; SVE-FIXED-NEXT: ret void139;140 %load = load <4 x double>, ptr %ptr, align 8141 %deinterleave = tail call { <2 x double>, <2 x double> } @llvm.vector.deinterleave2.v4f64(<4 x double> %load)142 %extract1 = extractvalue { <2 x double>, <2 x double> } %deinterleave, 0143 %extract2 = extractvalue { <2 x double>, <2 x double> } %deinterleave, 1144 ret void145}146 147define void @deinterleave_ptr_factor2(ptr %ptr) {148; NEON-LABEL: define void @deinterleave_ptr_factor2149; NEON-SAME: (ptr [[PTR:%.*]]) {150; NEON-NEXT: [[LDN:%.*]] = call { <2 x ptr>, <2 x ptr> } @llvm.aarch64.neon.ld2.v2p0.p0(ptr [[PTR]])151; NEON-NEXT: [[EXTRACT1:%.*]] = extractvalue { <2 x ptr>, <2 x ptr> } [[LDN]], 0152; NEON-NEXT: [[EXTRACT2:%.*]] = extractvalue { <2 x ptr>, <2 x ptr> } [[LDN]], 1153; NEON-NEXT: ret void154;155; SVE-FIXED-LABEL: define void @deinterleave_ptr_factor2156; SVE-FIXED-SAME: (ptr [[PTR:%.*]]) #[[ATTR0]] {157; SVE-FIXED-NEXT: [[LOAD:%.*]] = load <4 x ptr>, ptr [[PTR]], align 8158; SVE-FIXED-NEXT: [[DEINTERLEAVE:%.*]] = tail call { <2 x ptr>, <2 x ptr> } @llvm.vector.deinterleave2.v4p0(<4 x ptr> [[LOAD]])159; SVE-FIXED-NEXT: [[EXTRACT1:%.*]] = extractvalue { <2 x ptr>, <2 x ptr> } [[DEINTERLEAVE]], 0160; SVE-FIXED-NEXT: [[EXTRACT2:%.*]] = extractvalue { <2 x ptr>, <2 x ptr> } [[DEINTERLEAVE]], 1161; SVE-FIXED-NEXT: ret void162;163 %load = load <4 x ptr>, ptr %ptr, align 8164 %deinterleave = tail call { <2 x ptr>, <2 x ptr> } @llvm.vector.deinterleave2.v4p0(<4 x ptr> %load)165 %extract1 = extractvalue { <2 x ptr>, <2 x ptr> } %deinterleave, 0166 %extract2 = extractvalue { <2 x ptr>, <2 x ptr> } %deinterleave, 1167 ret void168}169 170define void @interleave_i8_factor2(ptr %ptr, <16 x i8> %l, <16 x i8> %r) {171; NEON-LABEL: define void @interleave_i8_factor2172; NEON-SAME: (ptr [[PTR:%.*]], <16 x i8> [[L:%.*]], <16 x i8> [[R:%.*]]) {173; NEON-NEXT: call void @llvm.aarch64.neon.st2.v16i8.p0(<16 x i8> [[L]], <16 x i8> [[R]], ptr [[PTR]])174; NEON-NEXT: ret void175;176; SVE-FIXED-LABEL: define void @interleave_i8_factor2177; SVE-FIXED-SAME: (ptr [[PTR:%.*]], <16 x i8> [[L:%.*]], <16 x i8> [[R:%.*]]) #[[ATTR0]] {178; SVE-FIXED-NEXT: [[INTERLEAVE:%.*]] = tail call <32 x i8> @llvm.vector.interleave2.v32i8(<16 x i8> [[L]], <16 x i8> [[R]])179; SVE-FIXED-NEXT: store <32 x i8> [[INTERLEAVE]], ptr [[PTR]], align 1180; SVE-FIXED-NEXT: ret void181;182 %interleave = tail call <32 x i8> @llvm.vector.interleave2.v32i8(<16 x i8> %l, <16 x i8> %r)183 store <32 x i8> %interleave, ptr %ptr, align 1184 ret void185}186 187define void @interleave_i16_factor2(ptr %ptr, <8 x i16> %l, <8 x i16> %r) {188; NEON-LABEL: define void @interleave_i16_factor2189; NEON-SAME: (ptr [[PTR:%.*]], <8 x i16> [[L:%.*]], <8 x i16> [[R:%.*]]) {190; NEON-NEXT: call void @llvm.aarch64.neon.st2.v8i16.p0(<8 x i16> [[L]], <8 x i16> [[R]], ptr [[PTR]])191; NEON-NEXT: ret void192;193; SVE-FIXED-LABEL: define void @interleave_i16_factor2194; SVE-FIXED-SAME: (ptr [[PTR:%.*]], <8 x i16> [[L:%.*]], <8 x i16> [[R:%.*]]) #[[ATTR0]] {195; SVE-FIXED-NEXT: [[INTERLEAVE:%.*]] = tail call <16 x i16> @llvm.vector.interleave2.v16i16(<8 x i16> [[L]], <8 x i16> [[R]])196; SVE-FIXED-NEXT: store <16 x i16> [[INTERLEAVE]], ptr [[PTR]], align 2197; SVE-FIXED-NEXT: ret void198;199 %interleave = tail call <16 x i16> @llvm.vector.interleave2.v16i16(<8 x i16> %l, <8 x i16> %r)200 store <16 x i16> %interleave, ptr %ptr, align 2201 ret void202}203 204define void @interleave_i32_factor2(ptr %ptr, <4 x i32> %l, <4 x i32> %r) {205; NEON-LABEL: define void @interleave_i32_factor2206; NEON-SAME: (ptr [[PTR:%.*]], <4 x i32> [[L:%.*]], <4 x i32> [[R:%.*]]) {207; NEON-NEXT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> [[L]], <4 x i32> [[R]], ptr [[PTR]])208; NEON-NEXT: ret void209;210; SVE-FIXED-LABEL: define void @interleave_i32_factor2211; SVE-FIXED-SAME: (ptr [[PTR:%.*]], <4 x i32> [[L:%.*]], <4 x i32> [[R:%.*]]) #[[ATTR0]] {212; SVE-FIXED-NEXT: [[INTERLEAVE:%.*]] = tail call <8 x i32> @llvm.vector.interleave2.v8i32(<4 x i32> [[L]], <4 x i32> [[R]])213; SVE-FIXED-NEXT: store <8 x i32> [[INTERLEAVE]], ptr [[PTR]], align 4214; SVE-FIXED-NEXT: ret void215;216 %interleave = tail call <8 x i32> @llvm.vector.interleave2.v8i32(<4 x i32> %l, <4 x i32> %r)217 store <8 x i32> %interleave, ptr %ptr, align 4218 ret void219}220 221define void @interleave_i64_factor2(ptr %ptr, <2 x i64> %l, <2 x i64> %r) {222; NEON-LABEL: define void @interleave_i64_factor2223; NEON-SAME: (ptr [[PTR:%.*]], <2 x i64> [[L:%.*]], <2 x i64> [[R:%.*]]) {224; NEON-NEXT: call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> [[L]], <2 x i64> [[R]], ptr [[PTR]])225; NEON-NEXT: ret void226;227; SVE-FIXED-LABEL: define void @interleave_i64_factor2228; SVE-FIXED-SAME: (ptr [[PTR:%.*]], <2 x i64> [[L:%.*]], <2 x i64> [[R:%.*]]) #[[ATTR0]] {229; SVE-FIXED-NEXT: [[INTERLEAVE:%.*]] = tail call <4 x i64> @llvm.vector.interleave2.v4i64(<2 x i64> [[L]], <2 x i64> [[R]])230; SVE-FIXED-NEXT: store <4 x i64> [[INTERLEAVE]], ptr [[PTR]], align 8231; SVE-FIXED-NEXT: ret void232;233 %interleave = tail call <4 x i64> @llvm.vector.interleave2.v4i64(<2 x i64> %l, <2 x i64> %r)234 store <4 x i64> %interleave, ptr %ptr, align 8235 ret void236}237 238define void @interleave_float_factor2(ptr %ptr, <4 x float> %l, <4 x float> %r) {239; NEON-LABEL: define void @interleave_float_factor2240; NEON-SAME: (ptr [[PTR:%.*]], <4 x float> [[L:%.*]], <4 x float> [[R:%.*]]) {241; NEON-NEXT: call void @llvm.aarch64.neon.st2.v4f32.p0(<4 x float> [[L]], <4 x float> [[R]], ptr [[PTR]])242; NEON-NEXT: ret void243;244; SVE-FIXED-LABEL: define void @interleave_float_factor2245; SVE-FIXED-SAME: (ptr [[PTR:%.*]], <4 x float> [[L:%.*]], <4 x float> [[R:%.*]]) #[[ATTR0]] {246; SVE-FIXED-NEXT: [[INTERLEAVE:%.*]] = tail call <8 x float> @llvm.vector.interleave2.v8f32(<4 x float> [[L]], <4 x float> [[R]])247; SVE-FIXED-NEXT: store <8 x float> [[INTERLEAVE]], ptr [[PTR]], align 4248; SVE-FIXED-NEXT: ret void249;250 %interleave = tail call <8 x float> @llvm.vector.interleave2.v8f32(<4 x float> %l, <4 x float> %r)251 store <8 x float> %interleave, ptr %ptr, align 4252 ret void253}254 255define void @interleave_double_factor2(ptr %ptr, <2 x double> %l, <2 x double> %r) {256; NEON-LABEL: define void @interleave_double_factor2257; NEON-SAME: (ptr [[PTR:%.*]], <2 x double> [[L:%.*]], <2 x double> [[R:%.*]]) {258; NEON-NEXT: call void @llvm.aarch64.neon.st2.v2f64.p0(<2 x double> [[L]], <2 x double> [[R]], ptr [[PTR]])259; NEON-NEXT: ret void260;261; SVE-FIXED-LABEL: define void @interleave_double_factor2262; SVE-FIXED-SAME: (ptr [[PTR:%.*]], <2 x double> [[L:%.*]], <2 x double> [[R:%.*]]) #[[ATTR0]] {263; SVE-FIXED-NEXT: [[INTERLEAVE:%.*]] = tail call <4 x double> @llvm.vector.interleave2.v4f64(<2 x double> [[L]], <2 x double> [[R]])264; SVE-FIXED-NEXT: store <4 x double> [[INTERLEAVE]], ptr [[PTR]], align 4265; SVE-FIXED-NEXT: ret void266;267 %interleave = tail call <4 x double> @llvm.vector.interleave2.v4f64(<2 x double> %l, <2 x double> %r)268 store <4 x double> %interleave, ptr %ptr, align 4269 ret void270}271 272define void @interleave_ptr_factor2(ptr %ptr, <2 x ptr> %l, <2 x ptr> %r) {273; NEON-LABEL: define void @interleave_ptr_factor2274; NEON-SAME: (ptr [[PTR:%.*]], <2 x ptr> [[L:%.*]], <2 x ptr> [[R:%.*]]) {275; NEON-NEXT: call void @llvm.aarch64.neon.st2.v2p0.p0(<2 x ptr> [[L]], <2 x ptr> [[R]], ptr [[PTR]])276; NEON-NEXT: ret void277;278; SVE-FIXED-LABEL: define void @interleave_ptr_factor2279; SVE-FIXED-SAME: (ptr [[PTR:%.*]], <2 x ptr> [[L:%.*]], <2 x ptr> [[R:%.*]]) #[[ATTR0]] {280; SVE-FIXED-NEXT: [[INTERLEAVE:%.*]] = tail call <4 x ptr> @llvm.vector.interleave2.v4p0(<2 x ptr> [[L]], <2 x ptr> [[R]])281; SVE-FIXED-NEXT: store <4 x ptr> [[INTERLEAVE]], ptr [[PTR]], align 4282; SVE-FIXED-NEXT: ret void283;284 %interleave = tail call <4 x ptr> @llvm.vector.interleave2.v4p0(<2 x ptr> %l, <2 x ptr> %r)285 store <4 x ptr> %interleave, ptr %ptr, align 4286 ret void287}288 289define void @deinterleave_wide_i16_factor2(ptr %ptr) #0 {290; NEON-LABEL: define void @deinterleave_wide_i16_factor2291; NEON-SAME: (ptr [[PTR:%.*]]) {292; NEON-NEXT: [[TMP1:%.*]] = getelementptr <8 x i16>, ptr [[PTR]], i64 0293; NEON-NEXT: [[LDN:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2.v8i16.p0(ptr [[TMP1]])294; NEON-NEXT: [[TMP2:%.*]] = extractvalue { <8 x i16>, <8 x i16> } [[LDN]], 0295; NEON-NEXT: [[TMP3:%.*]] = call <16 x i16> @llvm.vector.insert.v16i16.v8i16(<16 x i16> poison, <8 x i16> [[TMP2]], i64 0)296; NEON-NEXT: [[TMP4:%.*]] = extractvalue { <8 x i16>, <8 x i16> } [[LDN]], 1297; NEON-NEXT: [[TMP5:%.*]] = call <16 x i16> @llvm.vector.insert.v16i16.v8i16(<16 x i16> poison, <8 x i16> [[TMP4]], i64 0)298; NEON-NEXT: [[TMP6:%.*]] = getelementptr <8 x i16>, ptr [[PTR]], i64 2299; NEON-NEXT: [[LDN1:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2.v8i16.p0(ptr [[TMP6]])300; NEON-NEXT: [[TMP7:%.*]] = extractvalue { <8 x i16>, <8 x i16> } [[LDN1]], 0301; NEON-NEXT: [[TMP8:%.*]] = call <16 x i16> @llvm.vector.insert.v16i16.v8i16(<16 x i16> [[TMP3]], <8 x i16> [[TMP7]], i64 8)302; NEON-NEXT: [[TMP9:%.*]] = extractvalue { <8 x i16>, <8 x i16> } [[LDN1]], 1303; NEON-NEXT: [[TMP10:%.*]] = call <16 x i16> @llvm.vector.insert.v16i16.v8i16(<16 x i16> [[TMP5]], <8 x i16> [[TMP9]], i64 8)304; NEON-NEXT: [[TMP11:%.*]] = insertvalue { <16 x i16>, <16 x i16> } poison, <16 x i16> [[TMP8]], 0305; NEON-NEXT: [[TMP12:%.*]] = insertvalue { <16 x i16>, <16 x i16> } [[TMP11]], <16 x i16> [[TMP10]], 1306; NEON-NEXT: [[EXTRACT1:%.*]] = extractvalue { <16 x i16>, <16 x i16> } [[TMP12]], 0307; NEON-NEXT: [[EXTRACT2:%.*]] = extractvalue { <16 x i16>, <16 x i16> } [[TMP12]], 1308; NEON-NEXT: ret void309;310; SVE-FIXED-LABEL: define void @deinterleave_wide_i16_factor2311; SVE-FIXED-SAME: (ptr [[PTR:%.*]]) #[[ATTR0]] {312; SVE-FIXED-NEXT: [[LOAD:%.*]] = load <32 x i16>, ptr [[PTR]], align 2313; SVE-FIXED-NEXT: [[DEINTERLEAVE:%.*]] = tail call { <16 x i16>, <16 x i16> } @llvm.vector.deinterleave2.v32i16(<32 x i16> [[LOAD]])314; SVE-FIXED-NEXT: [[EXTRACT1:%.*]] = extractvalue { <16 x i16>, <16 x i16> } [[DEINTERLEAVE]], 0315; SVE-FIXED-NEXT: [[EXTRACT2:%.*]] = extractvalue { <16 x i16>, <16 x i16> } [[DEINTERLEAVE]], 1316; SVE-FIXED-NEXT: ret void317;318 %load = load <32 x i16>, ptr %ptr, align 2319 %deinterleave = tail call { <16 x i16>, <16 x i16> } @llvm.vector.deinterleave2.v32i16(<32 x i16> %load)320 %extract1 = extractvalue { <16 x i16>, <16 x i16> } %deinterleave, 0321 %extract2 = extractvalue { <16 x i16>, <16 x i16> } %deinterleave, 1322 ret void323}324 325define void @interleave_wide_ptr_factor2(ptr %ptr, <8 x ptr> %l, <8 x ptr> %r) {326; NEON-LABEL: define void @interleave_wide_ptr_factor2327; NEON-SAME: (ptr [[PTR:%.*]], <8 x ptr> [[L:%.*]], <8 x ptr> [[R:%.*]]) {328; NEON-NEXT: [[TMP1:%.*]] = getelementptr <2 x ptr>, ptr [[PTR]], i64 0329; NEON-NEXT: [[TMP2:%.*]] = call <2 x ptr> @llvm.vector.extract.v2p0.v8p0(<8 x ptr> [[L]], i64 0)330; NEON-NEXT: [[TMP3:%.*]] = call <2 x ptr> @llvm.vector.extract.v2p0.v8p0(<8 x ptr> [[R]], i64 0)331; NEON-NEXT: call void @llvm.aarch64.neon.st2.v2p0.p0(<2 x ptr> [[TMP2]], <2 x ptr> [[TMP3]], ptr [[TMP1]])332; NEON-NEXT: [[TMP4:%.*]] = getelementptr <2 x ptr>, ptr [[PTR]], i64 2333; NEON-NEXT: [[TMP5:%.*]] = call <2 x ptr> @llvm.vector.extract.v2p0.v8p0(<8 x ptr> [[L]], i64 2)334; NEON-NEXT: [[TMP6:%.*]] = call <2 x ptr> @llvm.vector.extract.v2p0.v8p0(<8 x ptr> [[R]], i64 2)335; NEON-NEXT: call void @llvm.aarch64.neon.st2.v2p0.p0(<2 x ptr> [[TMP5]], <2 x ptr> [[TMP6]], ptr [[TMP4]])336; NEON-NEXT: [[TMP7:%.*]] = getelementptr <2 x ptr>, ptr [[PTR]], i64 4337; NEON-NEXT: [[TMP8:%.*]] = call <2 x ptr> @llvm.vector.extract.v2p0.v8p0(<8 x ptr> [[L]], i64 4)338; NEON-NEXT: [[TMP9:%.*]] = call <2 x ptr> @llvm.vector.extract.v2p0.v8p0(<8 x ptr> [[R]], i64 4)339; NEON-NEXT: call void @llvm.aarch64.neon.st2.v2p0.p0(<2 x ptr> [[TMP8]], <2 x ptr> [[TMP9]], ptr [[TMP7]])340; NEON-NEXT: [[TMP10:%.*]] = getelementptr <2 x ptr>, ptr [[PTR]], i64 6341; NEON-NEXT: [[TMP11:%.*]] = call <2 x ptr> @llvm.vector.extract.v2p0.v8p0(<8 x ptr> [[L]], i64 6)342; NEON-NEXT: [[TMP12:%.*]] = call <2 x ptr> @llvm.vector.extract.v2p0.v8p0(<8 x ptr> [[R]], i64 6)343; NEON-NEXT: call void @llvm.aarch64.neon.st2.v2p0.p0(<2 x ptr> [[TMP11]], <2 x ptr> [[TMP12]], ptr [[TMP10]])344; NEON-NEXT: ret void345;346; SVE-FIXED-LABEL: define void @interleave_wide_ptr_factor2347; SVE-FIXED-SAME: (ptr [[PTR:%.*]], <8 x ptr> [[L:%.*]], <8 x ptr> [[R:%.*]]) #[[ATTR0]] {348; SVE-FIXED-NEXT: [[INTERLEAVE:%.*]] = tail call <16 x ptr> @llvm.vector.interleave2.v16p0(<8 x ptr> [[L]], <8 x ptr> [[R]])349; SVE-FIXED-NEXT: store <16 x ptr> [[INTERLEAVE]], ptr [[PTR]], align 4350; SVE-FIXED-NEXT: ret void351;352 %interleave = tail call <16 x ptr> @llvm.vector.interleave2.v16p0(<8 x ptr> %l, <8 x ptr> %r)353 store <16 x ptr> %interleave, ptr %ptr, align 4354 ret void355}356 357declare { <16 x i8>, <16 x i8> } @llvm.vector.deinterleave2.v32i8(<32 x i8>)358declare { <8 x i16>, <8 x i16> } @llvm.vector.deinterleave2.v16i16(<16 x i16>)359declare { <4 x i32>, <4 x i32> } @llvm.vector.deinterleave2.v8i32(<8 x i32>)360declare { <2 x i64>, <2 x i64> } @llvm.vector.deinterleave2.v4i64(<4 x i64>)361declare { <4 x float>, <4 x float> } @llvm.vector.deinterleave2.v8f32(<8 x float>)362declare { <2 x double>, <2 x double> } @llvm.vector.deinterleave2.v4f64(<4 x double>)363declare { <2 x ptr>, <2 x ptr> } @llvm.vector.deinterleave2.v4p0(<4 x ptr>)364declare { <16 x i16>, <16 x i16> } @llvm.vector.deinterleave2.v32i16(<32 x i16>)365 366declare <32 x i8> @llvm.vector.interleave2.v32i8(<16 x i8>, <16 x i8>)367declare <16 x i16> @llvm.vector.interleave2.v16i16(<8 x i16>, <8 x i16>)368declare <8 x i32> @llvm.vector.interleave2.v8i32(<4 x i32>, <4 x i32>)369declare <4 x i64> @llvm.vector.interleave2.v4i64(<2 x i64>, <2 x i64>)370declare <8 x float> @llvm.vector.interleave2.v8f32(<4 x float>, <4 x float>)371declare <4 x double> @llvm.vector.interleave2.v4f64(<2 x double>, <2 x double>)372declare <4 x ptr> @llvm.vector.interleave2.v4p0(<2 x ptr>, <2 x ptr>)373declare <16 x ptr> @llvm.vector.interleave2.v16p0(<8 x ptr>, <8 x ptr>)374