41 lines · plain
1; RUN: opt -S -passes=loop-vectorize,instsimplify -force-vector-interleave=1 \2; RUN: -mcpu=neoverse-v1 -sve-tail-folding=disabled < %s | FileCheck %s --check-prefix=CHECK-EPILOG3; RUN: opt -S -passes=loop-vectorize,instsimplify -force-vector-interleave=1 \4; RUN: -mcpu=neoverse-v2 < %s | FileCheck %s --check-prefix=CHECK-EPILOG-V25; RUN: opt -S -passes=loop-vectorize,instsimplify -force-vector-interleave=1 \6; RUN: -mcpu=cortex-x2 < %s | FileCheck %s --check-prefix=CHECK-NO-EPILOG7 8target triple = "aarch64-unknown-linux-gnu"9 10define void @foo(ptr noalias nocapture readonly %p, ptr noalias nocapture %q, i64 %len) #0 {11; CHECK-EPILOG: vec.epilog.ph:12; CHECK-EPILOG: vec.epilog.vector.body:13; CHECK-EPILOG: load <8 x i16>14 15; The epilogue loop gets vectorised vscale x 2 x i16 wide.16; CHECK-EPILOG-V2: vec.epilog.ph:17; CHECK-EPILOG-V2: vec.epilog.vector.body:18; CHECK-EPILOG-V2: load <vscale x 2 x i16>19 20; CHECK-NO-EPILOG-NOT: vec.epilog.vector.ph:21; CHECK-NO-EPILOG-NOT: vec.epilog.vector.body:22entry:23 br label %for.body24 25for.body: ; preds = %entry, %for.body26 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]27 %arrayidx = getelementptr inbounds i16, ptr %p, i64 %indvars.iv28 %0 = load i16, ptr %arrayidx29 %add = add nuw nsw i16 %0, 230 %arrayidx3 = getelementptr inbounds i16, ptr %q, i64 %indvars.iv31 store i16 %add, ptr %arrayidx332 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 133 %exitcond = icmp eq i64 %indvars.iv.next, %len34 br i1 %exitcond, label %exit, label %for.body35 36exit: ; preds = %for.body37 ret void38}39 40attributes #0 = { "target-features"="+sve" vscale_range(1,16) }41