1; This is the loop in c++ being vectorize in this file with 2; experimental.vector.reverse 3 4;#pragma clang loop vectorize_width(4, scalable) 5; for (long int i = N - 1; i >= 0; i--) 6; { 7; if (cond[i]) 8; a[i] += 1; 9; } 10 11; The test checks if the mask is being correctly created, reverted and used 12 13; RUN: opt -loop-vectorize -dce -instcombine -mtriple aarch64-linux-gnu -S < %s 2>%t | FileCheck %s 14 15; RUN: FileCheck --check-prefix=WARN --allow-empty %s <%t 16 17; If this check fails please read test/CodeGen/AArch64/README for instructions on how to resolve it. 18; WARN-NOT: warning 19 20 21target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128" 22target triple = "aarch64-unknown-linux-gnu" 23 24define void @vector_reverse_mask_nxv4i1(double* %a, double* %cond, i64 %N) #0 { 25; CHECK-LABEL: vector.body: 26; CHECK: %[[REVERSE6:.*]] = call <vscale x 4 x i1> @llvm.experimental.vector.reverse.nxv4i1(<vscale x 4 x i1> %{{.*}}) 27; CHECK: %[[WIDEMSKLOAD:.*]] = call <vscale x 4 x double> @llvm.masked.load.nxv4f64.p0nxv4f64(<vscale x 4 x double>* nonnull %{{.*}}, i32 8, <vscale x 4 x i1> %[[REVERSE6]], <vscale x 4 x double> poison) 28; CHECK-NEXT: %[[REVERSE7:.*]] = call <vscale x 4 x double> @llvm.experimental.vector.reverse.nxv4f64(<vscale x 4 x double> %[[WIDEMSKLOAD]]) 29; CHECK-NEXT: %[[FADD:.*]] = fadd <vscale x 4 x double> %[[REVERSE7]] 30; CHECK-NEXT: %[[REVERSE8:.*]] = call <vscale x 4 x double> @llvm.experimental.vector.reverse.nxv4f64(<vscale x 4 x double> %[[FADD]]) 31; CHECK: %[[REVERSE9:.*]] = call <vscale x 4 x i1> @llvm.experimental.vector.reverse.nxv4i1(<vscale x 4 x i1> %{{.*}}) 32; CHECK: call void @llvm.masked.store.nxv4f64.p0nxv4f64(<vscale x 4 x double> %[[REVERSE8]], <vscale x 4 x double>* %{{.*}}, i32 8, <vscale x 4 x i1> %[[REVERSE9]] 33 34entry: 35 %cmp7 = icmp sgt i64 %N, 0 36 br i1 %cmp7, label %for.body, label %for.cond.cleanup 37 38for.cond.cleanup: ; preds = %for.cond.cleanup, %entry 39 ret void 40 41for.body: ; preds = %for.body, %entry 42 %i.08.in = phi i64 [ %i.08, %for.inc ], [ %N, %entry ] 43 %i.08 = add nsw i64 %i.08.in, -1 44 %arrayidx = getelementptr inbounds double, double* %cond, i64 %i.08 45 %0 = load double, double* %arrayidx, align 8 46 %tobool = fcmp une double %0, 0.000000e+00 47 br i1 %tobool, label %if.then, label %for.inc 48 49if.then: ; preds = %for.body 50 %arrayidx1 = getelementptr inbounds double, double* %a, i64 %i.08 51 %1 = load double, double* %arrayidx1, align 8 52 %add = fadd double %1, 1.000000e+00 53 store double %add, double* %arrayidx1, align 8 54 br label %for.inc 55 56for.inc: ; preds = %for.body, %if.then 57 %cmp = icmp sgt i64 %i.08.in, 1 58 br i1 %cmp, label %for.body, label %for.cond.cleanup, !llvm.loop !0 59} 60 61attributes #0 = {"target-cpu"="generic" "target-features"="+neon,+sve"} 62 63 64!0 = distinct !{!0, !1, !2, !3, !4} 65!1 = !{!"llvm.loop.mustprogress"} 66!2 = !{!"llvm.loop.vectorize.width", i32 4} 67!3 = !{!"llvm.loop.vectorize.scalable.enable", i1 true} 68!4 = !{!"llvm.loop.vectorize.enable", i1 true} 69