52 lines · plain
1; RUN: opt -mcpu=kryo -mtriple=aarch64-gnu-linux -passes=loop-data-prefetch -max-prefetch-iters-ahead=1000 -S < %s | FileCheck %s --check-prefix=LARGE_PREFETCH --check-prefix=ALL2; RUN: opt -mcpu=kryo -mtriple=aarch64-gnu-linux -passes=loop-data-prefetch -S < %s | FileCheck %s --check-prefix=NO_LARGE_PREFETCH --check-prefix=ALL3 4target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-n32:64-S128"5 6; ALL-LABEL: @small_stride(7define void @small_stride(ptr nocapture %a, ptr nocapture readonly %b) {8entry:9 br label %for.body10 11; ALL: for.body:12for.body: ; preds = %for.body, %entry13 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]14 %arrayidx = getelementptr inbounds double, ptr %b, i64 %indvars.iv15; ALL-NOT: call void @llvm.prefetch16 %0 = load double, ptr %arrayidx, align 817 %add = fadd double %0, 1.000000e+0018 %arrayidx2 = getelementptr inbounds double, ptr %a, i64 %indvars.iv19 store double %add, ptr %arrayidx2, align 820 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 121 %exitcond = icmp eq i64 %indvars.iv.next, 160022 br i1 %exitcond, label %for.end, label %for.body23 24; ALL: for.end:25for.end: ; preds = %for.body26 ret void27}28 29; ALL-LABEL: @large_stride(30define void @large_stride(ptr nocapture %a, ptr nocapture readonly %b) {31entry:32 br label %for.body33 34; ALL: for.body:35for.body: ; preds = %for.body, %entry36 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]37 %arrayidx = getelementptr inbounds double, ptr %b, i64 %indvars.iv38; LARGE_PREFETCH: call void @llvm.prefetch39; NO_LARGE_PREFETCH-NOT: call void @llvm.prefetch40 %0 = load double, ptr %arrayidx, align 841 %add = fadd double %0, 1.000000e+0042 %arrayidx2 = getelementptr inbounds double, ptr %a, i64 %indvars.iv43 store double %add, ptr %arrayidx2, align 844 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 150 45 %exitcond = icmp eq i64 %indvars.iv.next, 16000046 br i1 %exitcond, label %for.end, label %for.body47 48; ALL: for.end:49for.end: ; preds = %for.body50 ret void51}52