46 lines · plain
1; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -mtriple=arm64-apple-darinw -S %s | FileCheck %s2 3; In the loop below, both the current and previous values of a first-order4; recurrence are stored in an interleave group.5define void @interleaved_store_first_order_recurrence(ptr noalias %src, ptr %dst) {6; CHECK-LABEL: @interleaved_store_first_order_recurrence(7; CHECK: vector.body:8; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]9; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 99>, %vector.ph ], [ [[BROADCAST_SPLAT:%.*]], %vector.body ]10; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[SRC:%.*]], align 411; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i64 012; CHECK-NEXT: [[BROADCAST_SPLAT]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer13; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[BROADCAST_SPLAT]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>14; CHECK-NEXT: [[TMP3:%.*]] = mul nuw nsw i64 [[INDEX]], 315; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 [[TMP3]]16; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i32> zeroinitializer, <4 x i32> [[TMP2]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>17; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLAT]], <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>18; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <8 x i32> [[TMP9]], <8 x i32> [[TMP10]], <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>19; CHECK-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <12 x i32> [[TMP11]], <12 x i32> poison, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11>20; CHECK-NEXT: store <12 x i32> [[INTERLEAVED_VEC]], ptr [[TMP4]], align 421; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 422; CHECK-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100023; CHECK-NEXT: br i1 [[TMP12]], label %middle.block, label %vector.body24;25entry:26 br label %loop27 28loop:29 %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]30 %for = phi i32 [ 99, %entry ],[ %for.next, %loop ]31 %for.next = load i32, ptr %src, align 432 %off = mul nuw nsw i64 %iv, 333 %gep.1 = getelementptr inbounds i32, ptr %dst, i64 %off34 store i32 0, ptr %gep.1, align 435 %gep.2 = getelementptr inbounds i32, ptr %gep.1, i64 136 store i32 %for, ptr %gep.2, align 437 %gep.3 = getelementptr inbounds i32, ptr %gep.1, i64 238 store i32 %for.next, ptr %gep.3, align 439 %iv.next = add nuw nsw i64 %iv, 140 %ec = icmp eq i64 %iv.next, 100041 br i1 %ec, label %exit, label %loop42 43exit:44 ret void45}46