91 lines · plain
1; RUN: llc -mtriple=aarch64-none-linux-gnu -align-loops=32 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-DEFAULT2; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=neoverse-n1 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-163; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=neoverse-n2 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-164; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=neoverse-v1 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-165; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-x1 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-166; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-x2 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-167; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-a35 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-88; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-a53 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-89; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-a55 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-810; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-a57 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-811; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-a510 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-812; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-a75 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-813; RUN: llc -mtriple=aarch64-none-linux-gnu -mcpu=cortex-a710 < %s -o -| FileCheck %s --check-prefixes=CHECK,CHECK-1614 15define i32 @a(i32 %x, ptr nocapture readonly %y, ptr nocapture readonly %z) {16; CHECK-DEFAULT: .p2align 517; CHECK-8: .p2align 4, , 818; CHECK-16: .p2align 5, , 1619; CHECK-NEXT: .LBB0_5: // %vector.body20; CHECK-DEFAULT: .p2align 521; CHECK-8: .p2align 4, , 822; CHECK-16: .p2align 5, , 1623; CHECK-NEXT: .LBB0_8: // %for.body24entry:25 %cmp10 = icmp sgt i32 %x, 026 br i1 %cmp10, label %for.body.preheader, label %for.cond.cleanup27 28for.body.preheader: ; preds = %entry29 %wide.trip.count = zext i32 %x to i6430 %min.iters.check = icmp ult i32 %x, 831 br i1 %min.iters.check, label %for.body.preheader17, label %vector.ph32 33vector.ph: ; preds = %for.body.preheader34 %n.vec = and i64 %wide.trip.count, 429496728835 br label %vector.body36 37vector.body: ; preds = %vector.body, %vector.ph38 %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]39 %vec.phi = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ %10, %vector.body ]40 %vec.phi13 = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ %11, %vector.body ]41 %0 = getelementptr inbounds i32, ptr %y, i64 %index42 %1 = bitcast ptr %0 to ptr43 %wide.load = load <4 x i32>, ptr %1, align 444 %2 = getelementptr inbounds i32, ptr %0, i64 445 %3 = bitcast ptr %2 to ptr46 %wide.load14 = load <4 x i32>, ptr %3, align 447 %4 = getelementptr inbounds i32, ptr %z, i64 %index48 %5 = bitcast ptr %4 to ptr49 %wide.load15 = load <4 x i32>, ptr %5, align 450 %6 = getelementptr inbounds i32, ptr %4, i64 451 %7 = bitcast ptr %6 to ptr52 %wide.load16 = load <4 x i32>, ptr %7, align 453 %8 = add <4 x i32> %wide.load, %vec.phi54 %9 = add <4 x i32> %wide.load14, %vec.phi1355 %10 = add <4 x i32> %8, %wide.load1556 %11 = add <4 x i32> %9, %wide.load1657 %index.next = add nuw i64 %index, 858 %12 = icmp eq i64 %index.next, %n.vec59 br i1 %12, label %middle.block, label %vector.body60 61middle.block: ; preds = %vector.body62 %bin.rdx = add <4 x i32> %11, %1063 %13 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %bin.rdx)64 %cmp.n = icmp eq i64 %n.vec, %wide.trip.count65 br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader1766 67for.body.preheader17: ; preds = %for.body.preheader, %middle.block68 %indvars.iv.ph = phi i64 [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]69 %b.011.ph = phi i32 [ 0, %for.body.preheader ], [ %13, %middle.block ]70 br label %for.body71 72for.cond.cleanup: ; preds = %for.body, %middle.block, %entry73 %b.0.lcssa = phi i32 [ 0, %entry ], [ %13, %middle.block ], [ %add3, %for.body ]74 ret i32 %b.0.lcssa75 76for.body: ; preds = %for.body.preheader17, %for.body77 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ %indvars.iv.ph, %for.body.preheader17 ]78 %b.011 = phi i32 [ %add3, %for.body ], [ %b.011.ph, %for.body.preheader17 ]79 %arrayidx = getelementptr inbounds i32, ptr %y, i64 %indvars.iv80 %14 = load i32, ptr %arrayidx, align 481 %arrayidx2 = getelementptr inbounds i32, ptr %z, i64 %indvars.iv82 %15 = load i32, ptr %arrayidx2, align 483 %add = add i32 %14, %b.01184 %add3 = add i32 %add, %1585 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 186 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count87 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body88}89 90declare i32 @llvm.vector.reduce.add.v4i32(<4 x i32>)91