87 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py2; RUN: opt < %s -passes=loop-vectorize -force-vector-width=2 -enable-masked-interleaved-mem-accesses -enable-interleaved-mem-accesses -S | FileCheck %s3 4target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"5 6@A = external dso_local local_unnamed_addr global [40 x [4 x i16]], align 17 8; Make sure interleave group of loads with gap is considered masked with fold-tail,9; and forbidden with reverse access.10 11define dso_local i16 @reverse_interleave_load_fold_mask() optsize {12; CHECK-LABEL: @reverse_interleave_load_fold_mask(13; CHECK-NEXT: entry:14; CHECK-NEXT: br label [[VECTOR_PH:%.*]]15; CHECK: vector.ph:16; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]17; CHECK: vector.body:18; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE4:%.*]] ]19; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i16> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP25:%.*]], [[PRED_LOAD_CONTINUE4]] ]20; CHECK-NEXT: [[TMP0:%.*]] = trunc i32 [[INDEX]] to i1621; CHECK-NEXT: [[OFFSET_IDX:%.*]] = sub i16 41, [[TMP0]]22; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <2 x i32> poison, i32 [[INDEX]], i64 023; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT1]], <2 x i32> poison, <2 x i32> zeroinitializer24; CHECK-NEXT: [[VEC_IV:%.*]] = add <2 x i32> [[BROADCAST_SPLAT2]], <i32 0, i32 1>25; CHECK-NEXT: [[TMP1:%.*]] = icmp ule <2 x i32> [[VEC_IV]], splat (i32 40)26; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x i1> [[TMP1]], i32 027; CHECK-NEXT: br i1 [[TMP2]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]28; CHECK: pred.load.if:29; CHECK-NEXT: [[TMP3:%.*]] = add i16 [[OFFSET_IDX]], 030; CHECK-NEXT: [[TMP4:%.*]] = add nsw i16 [[TMP3]], -131; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [40 x [4 x i16]], ptr @A, i16 0, i16 [[TMP4]], i16 032; CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[TMP5]], align 133; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i16> poison, i16 [[TMP6]], i32 034; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [40 x [4 x i16]], ptr @A, i16 0, i16 [[TMP4]], i16 335; CHECK-NEXT: [[TMP9:%.*]] = load i16, ptr [[TMP8]], align 136; CHECK-NEXT: [[TMP10:%.*]] = insertelement <2 x i16> poison, i16 [[TMP9]], i32 037; CHECK-NEXT: br label [[PRED_LOAD_CONTINUE]]38; CHECK: pred.load.continue:39; CHECK-NEXT: [[TMP11:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_LOAD_IF]] ]40; CHECK-NEXT: [[TMP12:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP10]], [[PRED_LOAD_IF]] ]41; CHECK-NEXT: [[TMP13:%.*]] = extractelement <2 x i1> [[TMP1]], i32 142; CHECK-NEXT: br i1 [[TMP13]], label [[PRED_LOAD_IF3:%.*]], label [[PRED_LOAD_CONTINUE4]]43; CHECK: pred.load.if1:44; CHECK-NEXT: [[TMP14:%.*]] = add i16 [[OFFSET_IDX]], -145; CHECK-NEXT: [[TMP15:%.*]] = add nsw i16 [[TMP14]], -146; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds [40 x [4 x i16]], ptr @A, i16 0, i16 [[TMP15]], i16 047; CHECK-NEXT: [[TMP17:%.*]] = load i16, ptr [[TMP16]], align 148; CHECK-NEXT: [[TMP18:%.*]] = insertelement <2 x i16> [[TMP11]], i16 [[TMP17]], i32 149; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds [40 x [4 x i16]], ptr @A, i16 0, i16 [[TMP15]], i16 350; CHECK-NEXT: [[TMP20:%.*]] = load i16, ptr [[TMP19]], align 151; CHECK-NEXT: [[TMP21:%.*]] = insertelement <2 x i16> [[TMP12]], i16 [[TMP20]], i32 152; CHECK-NEXT: br label [[PRED_LOAD_CONTINUE4]]53; CHECK: pred.load.continue2:54; CHECK-NEXT: [[TMP22:%.*]] = phi <2 x i16> [ [[TMP11]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP18]], [[PRED_LOAD_IF3]] ]55; CHECK-NEXT: [[TMP23:%.*]] = phi <2 x i16> [ [[TMP12]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP21]], [[PRED_LOAD_IF3]] ]56; CHECK-NEXT: [[TMP24:%.*]] = add nsw <2 x i16> [[TMP22]], [[TMP23]]57; CHECK-NEXT: [[TMP25]] = add <2 x i16> [[VEC_PHI]], [[TMP24]]58; CHECK-NEXT: [[TMP26:%.*]] = select <2 x i1> [[TMP1]], <2 x i16> [[TMP25]], <2 x i16> [[VEC_PHI]]59; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 260; CHECK-NEXT: [[TMP27:%.*]] = icmp eq i32 [[INDEX_NEXT]], 4261; CHECK-NEXT: br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]62; CHECK: middle.block:63; CHECK-NEXT: [[TMP28:%.*]] = call i16 @llvm.vector.reduce.add.v2i16(<2 x i16> [[TMP26]])64; CHECK-NEXT: br label [[LOOP:%.*]]65; CHECK: exit:66; CHECK-NEXT: ret i16 [[TMP28]]67;68entry:69 br label %loop70 71loop:72 %iv = phi i16 [ 41, %entry ], [ %ivMinus1, %loop ]73 %sum = phi i16 [ 0, %entry ], [ %prevSum, %loop ]74 %ivMinus1 = add nsw i16 %iv, -175 %gepA0 = getelementptr inbounds [40 x [4 x i16]], ptr @A, i16 0, i16 %ivMinus1, i16 076 %0 = load i16, ptr %gepA0, align 177 %gepA3 = getelementptr inbounds [40 x [4 x i16]], ptr @A, i16 0, i16 %ivMinus1, i16 378 %1 = load i16, ptr %gepA3, align 179 %add = add nsw i16 %0, %180 %prevSum = add nsw i16 %sum, %add81 %cmp = icmp ugt i16 %iv, 182 br i1 %cmp, label %loop, label %exit83 84exit:85 ret i16 %prevSum86}87