33 lines · plain
1; REQUIRES: asserts2; RUN: opt -mtriple=s390x-unknown-linux -mcpu=z13 -passes=loop-vectorize \3; RUN: -force-vector-width=4 -debug-only=loop-vectorize \4; RUN: -disable-output -enable-interleaved-mem-accesses=false < %s 2>&1 | \5; RUN: FileCheck %s6;7; Check that a scalarized load/store does not get a cost for insterts/8; extracts, since z13 supports element load/store.9 10define void @fun(ptr %data, i64 %n) {11entry:12 br label %for.body13 14for.body:15 %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]16 %tmp0 = getelementptr inbounds i32, ptr %data, i64 %i17 %tmp1 = load i32, ptr %tmp0, align 418 %tmp2 = add i32 %tmp1, 119 store i32 %tmp2, ptr %tmp0, align 420 %i.next = add nuw nsw i64 %i, 221 %cond = icmp slt i64 %i.next, %n22 br i1 %cond, label %for.body, label %for.end23 24for.end:25 ret void26 27; CHECK: LV: Scalarizing: %tmp1 = load i32, ptr %tmp0, align 428; CHECK: LV: Scalarizing: store i32 %tmp2, ptr %tmp0, align 429 30; CHECK: Cost of 4 for VF 4: REPLICATE ir<%tmp1> = load ir<%tmp0>31; CHECK: Cost of 4 for VF 4: REPLICATE store ir<%tmp2>, ir<%tmp0>32}33