220 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc -mattr=+sve -force-streaming-compatible < %s | FileCheck %s3; RUN: llc -mattr=+sme -force-streaming < %s | FileCheck %s4; RUN: llc -force-streaming-compatible < %s | FileCheck %s --check-prefix=NONEON-NOSVE5 6target triple = "aarch64-unknown-linux-gnu"7 8declare void @def(ptr)9 10define void @alloc_v4i8(ptr %st_ptr) nounwind {11; CHECK-LABEL: alloc_v4i8:12; CHECK: // %bb.0:13; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill14; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill15; CHECK-NEXT: mov x19, x016; CHECK-NEXT: add x0, sp, #1217; CHECK-NEXT: add x20, sp, #1218; CHECK-NEXT: bl def19; CHECK-NEXT: ptrue p0.b, vl220; CHECK-NEXT: ld2b { z0.b, z1.b }, p0/z, [x20]21; CHECK-NEXT: ptrue p0.s, vl222; CHECK-NEXT: mov z2.b, z0.b[1]23; CHECK-NEXT: zip1 z0.s, z0.s, z2.s24; CHECK-NEXT: st1b { z0.s }, p0, [x19]25; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload26; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload27; CHECK-NEXT: ret28;29; NONEON-NOSVE-LABEL: alloc_v4i8:30; NONEON-NOSVE: // %bb.0:31; NONEON-NOSVE-NEXT: sub sp, sp, #3232; NONEON-NOSVE-NEXT: stp x30, x19, [sp, #16] // 16-byte Folded Spill33; NONEON-NOSVE-NEXT: mov x19, x034; NONEON-NOSVE-NEXT: add x0, sp, #1235; NONEON-NOSVE-NEXT: bl def36; NONEON-NOSVE-NEXT: ldrb w8, [sp, #14]37; NONEON-NOSVE-NEXT: ldrb w9, [sp, #12]38; NONEON-NOSVE-NEXT: strb w8, [x19, #1]39; NONEON-NOSVE-NEXT: strb w9, [x19]40; NONEON-NOSVE-NEXT: ldp x30, x19, [sp, #16] // 16-byte Folded Reload41; NONEON-NOSVE-NEXT: add sp, sp, #3242; NONEON-NOSVE-NEXT: ret43 %alloc = alloca [4 x i8]44 call void @def(ptr %alloc)45 %load = load <4 x i8>, ptr %alloc46 %strided.vec = shufflevector <4 x i8> %load, <4 x i8> poison, <2 x i32> <i32 0, i32 2>47 store <2 x i8> %strided.vec, ptr %st_ptr48 ret void49}50 51define void @alloc_v6i8(ptr %st_ptr) nounwind {52; CHECK-LABEL: alloc_v6i8:53; CHECK: // %bb.0:54; CHECK-NEXT: sub sp, sp, #3255; CHECK-NEXT: stp x30, x19, [sp, #16] // 16-byte Folded Spill56; CHECK-NEXT: mov x19, x057; CHECK-NEXT: add x0, sp, #858; CHECK-NEXT: bl def59; CHECK-NEXT: ldr d0, [sp, #8]60; CHECK-NEXT: ptrue p0.h, vl461; CHECK-NEXT: add x8, sp, #462; CHECK-NEXT: ptrue p1.s, vl263; CHECK-NEXT: mov z1.b, z0.b[3]64; CHECK-NEXT: mov z2.b, z0.b[1]65; CHECK-NEXT: mov z0.b, z0.b[5]66; CHECK-NEXT: zip1 z1.h, z2.h, z1.h67; CHECK-NEXT: zip1 z1.s, z1.s, z0.s68; CHECK-NEXT: st1b { z1.h }, p0, [x8]69; CHECK-NEXT: ld1h { z1.s }, p1/z, [x8]70; CHECK-NEXT: fmov w8, s071; CHECK-NEXT: strb w8, [x19, #2]72; CHECK-NEXT: str h1, [x19]73; CHECK-NEXT: ldp x30, x19, [sp, #16] // 16-byte Folded Reload74; CHECK-NEXT: add sp, sp, #3275; CHECK-NEXT: ret76;77; NONEON-NOSVE-LABEL: alloc_v6i8:78; NONEON-NOSVE: // %bb.0:79; NONEON-NOSVE-NEXT: sub sp, sp, #4880; NONEON-NOSVE-NEXT: stp x30, x19, [sp, #32] // 16-byte Folded Spill81; NONEON-NOSVE-NEXT: mov x19, x082; NONEON-NOSVE-NEXT: add x0, sp, #2483; NONEON-NOSVE-NEXT: bl def84; NONEON-NOSVE-NEXT: ldr x8, [sp, #24]85; NONEON-NOSVE-NEXT: str x8, [sp]86; NONEON-NOSVE-NEXT: ldr d0, [sp]87; NONEON-NOSVE-NEXT: str d0, [sp, #8]88; NONEON-NOSVE-NEXT: ldrb w8, [sp, #11]89; NONEON-NOSVE-NEXT: strb w8, [sp, #21]90; NONEON-NOSVE-NEXT: ldrb w8, [sp, #9]91; NONEON-NOSVE-NEXT: strb w8, [sp, #20]92; NONEON-NOSVE-NEXT: ldrb w8, [sp, #13]93; NONEON-NOSVE-NEXT: ldrh w9, [sp, #20]94; NONEON-NOSVE-NEXT: strb w8, [x19, #2]95; NONEON-NOSVE-NEXT: strh w9, [x19]96; NONEON-NOSVE-NEXT: ldp x30, x19, [sp, #32] // 16-byte Folded Reload97; NONEON-NOSVE-NEXT: add sp, sp, #4898; NONEON-NOSVE-NEXT: ret99 %alloc = alloca [6 x i8]100 call void @def(ptr %alloc)101 %load = load <6 x i8>, ptr %alloc102 %strided.vec = shufflevector <6 x i8> %load, <6 x i8> poison, <3 x i32> <i32 1, i32 3, i32 5>103 store <3 x i8> %strided.vec, ptr %st_ptr104 ret void105}106 107define void @alloc_v32i8(ptr %st_ptr) nounwind {108; CHECK-LABEL: alloc_v32i8:109; CHECK: // %bb.0:110; CHECK-NEXT: sub sp, sp, #48111; CHECK-NEXT: stp x30, x19, [sp, #32] // 16-byte Folded Spill112; CHECK-NEXT: mov x19, x0113; CHECK-NEXT: mov x0, sp114; CHECK-NEXT: bl def115; CHECK-NEXT: adrp x8, .LCPI2_0116; CHECK-NEXT: ldr q0, [sp]117; CHECK-NEXT: ldr q1, [x8, :lo12:.LCPI2_0]118; CHECK-NEXT: ldrb w8, [sp, #16]119; CHECK-NEXT: tbl z0.b, { z0.b }, z1.b120; CHECK-NEXT: strb w8, [x19, #8]121; CHECK-NEXT: str d0, [x19]122; CHECK-NEXT: ldp x30, x19, [sp, #32] // 16-byte Folded Reload123; CHECK-NEXT: add sp, sp, #48124; CHECK-NEXT: ret125;126; NONEON-NOSVE-LABEL: alloc_v32i8:127; NONEON-NOSVE: // %bb.0:128; NONEON-NOSVE-NEXT: sub sp, sp, #112129; NONEON-NOSVE-NEXT: stp x30, x19, [sp, #96] // 16-byte Folded Spill130; NONEON-NOSVE-NEXT: mov x19, x0131; NONEON-NOSVE-NEXT: add x0, sp, #64132; NONEON-NOSVE-NEXT: bl def133; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]134; NONEON-NOSVE-NEXT: str q0, [sp]135; NONEON-NOSVE-NEXT: ldrb w8, [sp, #14]136; NONEON-NOSVE-NEXT: str q1, [sp, #32]137; NONEON-NOSVE-NEXT: strb w8, [sp, #23]138; NONEON-NOSVE-NEXT: ldrb w8, [sp, #12]139; NONEON-NOSVE-NEXT: strb w8, [sp, #22]140; NONEON-NOSVE-NEXT: ldrb w8, [sp, #10]141; NONEON-NOSVE-NEXT: strb w8, [sp, #21]142; NONEON-NOSVE-NEXT: ldrb w8, [sp, #8]143; NONEON-NOSVE-NEXT: strb w8, [sp, #20]144; NONEON-NOSVE-NEXT: ldrb w8, [sp, #6]145; NONEON-NOSVE-NEXT: strb w8, [sp, #19]146; NONEON-NOSVE-NEXT: ldrb w8, [sp, #4]147; NONEON-NOSVE-NEXT: strb w8, [sp, #18]148; NONEON-NOSVE-NEXT: ldrb w8, [sp, #2]149; NONEON-NOSVE-NEXT: strb w8, [sp, #17]150; NONEON-NOSVE-NEXT: ldrb w8, [sp]151; NONEON-NOSVE-NEXT: strb w8, [sp, #16]152; NONEON-NOSVE-NEXT: ldrb w8, [sp, #32]153; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]154; NONEON-NOSVE-NEXT: strb w8, [x19, #8]155; NONEON-NOSVE-NEXT: str q0, [sp, #48]156; NONEON-NOSVE-NEXT: ldr x8, [sp, #48]157; NONEON-NOSVE-NEXT: str x8, [x19]158; NONEON-NOSVE-NEXT: ldp x30, x19, [sp, #96] // 16-byte Folded Reload159; NONEON-NOSVE-NEXT: add sp, sp, #112160; NONEON-NOSVE-NEXT: ret161 %alloc = alloca [32 x i8]162 call void @def(ptr %alloc)163 %load = load <32 x i8>, ptr %alloc164 %strided.vec = shufflevector <32 x i8> %load, <32 x i8> poison, <9 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16>165 store <9 x i8> %strided.vec, ptr %st_ptr166 ret void167}168 169 170define void @alloc_v8f64(ptr %st_ptr) nounwind {171; CHECK-LABEL: alloc_v8f64:172; CHECK: // %bb.0:173; CHECK-NEXT: sub sp, sp, #96174; CHECK-NEXT: stp x20, x19, [sp, #80] // 16-byte Folded Spill175; CHECK-NEXT: mov x19, x0176; CHECK-NEXT: mov x0, sp177; CHECK-NEXT: str x30, [sp, #64] // 8-byte Spill178; CHECK-NEXT: mov x20, sp179; CHECK-NEXT: bl def180; CHECK-NEXT: ptrue p0.d, vl2181; CHECK-NEXT: mov x8, #4 // =0x4182; CHECK-NEXT: ld2d { z0.d, z1.d }, p0/z, [x20]183; CHECK-NEXT: ld2d { z2.d, z3.d }, p0/z, [x20, x8, lsl #3]184; CHECK-NEXT: ldr x30, [sp, #64] // 8-byte Reload185; CHECK-NEXT: stp q0, q2, [x19]186; CHECK-NEXT: ldp x20, x19, [sp, #80] // 16-byte Folded Reload187; CHECK-NEXT: add sp, sp, #96188; CHECK-NEXT: ret189;190; NONEON-NOSVE-LABEL: alloc_v8f64:191; NONEON-NOSVE: // %bb.0:192; NONEON-NOSVE-NEXT: sub sp, sp, #176193; NONEON-NOSVE-NEXT: stp x30, x19, [sp, #160] // 16-byte Folded Spill194; NONEON-NOSVE-NEXT: mov x19, x0195; NONEON-NOSVE-NEXT: add x0, sp, #96196; NONEON-NOSVE-NEXT: bl def197; NONEON-NOSVE-NEXT: ldp q1, q0, [sp, #96]198; NONEON-NOSVE-NEXT: ldp q2, q3, [sp, #128]199; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #48]200; NONEON-NOSVE-NEXT: ldr d1, [sp, #64]201; NONEON-NOSVE-NEXT: ldr d0, [sp, #48]202; NONEON-NOSVE-NEXT: stp q2, q3, [sp]203; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #80]204; NONEON-NOSVE-NEXT: ldr d1, [sp, #16]205; NONEON-NOSVE-NEXT: ldr d0, [sp]206; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #32]207; NONEON-NOSVE-NEXT: ldr q1, [sp, #80]208; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]209; NONEON-NOSVE-NEXT: stp q1, q0, [x19]210; NONEON-NOSVE-NEXT: ldp x30, x19, [sp, #160] // 16-byte Folded Reload211; NONEON-NOSVE-NEXT: add sp, sp, #176212; NONEON-NOSVE-NEXT: ret213 %alloc = alloca [8 x double]214 call void @def(ptr %alloc)215 %load = load <8 x double>, ptr %alloc216 %strided.vec = shufflevector <8 x double> %load, <8 x double> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>217 store <4 x double> %strided.vec, ptr %st_ptr218 ret void219}220