1; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -instcombine -S | FileCheck %s 2 3target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128" 4 5; CHECK-LABEL: @vector_gep_stored( 6; CHECK: vector.body: 7; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ] 8; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ] 9; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* %b, <4 x i64> [[VEC_IND]] 10; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32*, i32** %a, i64 [[INDEX]] 11; CHECK-NEXT: [[TMP3:%.*]] = bitcast i32** [[TMP2]] to <4 x i32*>* 12; CHECK-NEXT: store <4 x i32*> [[TMP1]], <4 x i32*>* [[TMP3]], align 8 13; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4 14; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], <i64 4, i64 4, i64 4, i64 4> 15; CHECK: br i1 {{.*}}, label %middle.block, label %vector.body 16; 17define void @vector_gep_stored(i32** %a, i32 *%b, i64 %n) { 18entry: 19 br label %for.body 20 21for.body: 22 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ] 23 %tmp0 = getelementptr inbounds i32, i32* %b, i64 %i 24 %tmp1 = getelementptr inbounds i32*, i32** %a, i64 %i 25 store i32* %tmp0, i32** %tmp1, align 8 26 %i.next = add nuw nsw i64 %i, 1 27 %cond = icmp slt i64 %i.next, %n 28 br i1 %cond, label %for.body, label %for.end 29 30for.end: 31 ret void 32} 33 34; CHECK-LABEL: @uniform_vector_gep_stored( 35; CHECK: vector.body: 36; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ] 37; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* %b, i64 1 38; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32*> undef, i32* [[TMP1]], i32 0 39; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32*> [[DOTSPLATINSERT]], <4 x i32*> undef, <4 x i32> zeroinitializer 40; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32*, i32** %a, i64 [[INDEX]] 41; CHECK-NEXT: [[TMP3:%.*]] = bitcast i32** [[TMP2]] to <4 x i32*>* 42; CHECK-NEXT: store <4 x i32*> [[DOTSPLAT]], <4 x i32*>* [[TMP3]], align 8 43; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4 44; CHECK: br i1 {{.*}}, label %middle.block, label %vector.body 45; 46define void @uniform_vector_gep_stored(i32** %a, i32 *%b, i64 %n) { 47entry: 48 br label %for.body 49 50for.body: 51 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ] 52 %tmp0 = getelementptr inbounds i32, i32* %b, i64 1 53 %tmp1 = getelementptr inbounds i32*, i32** %a, i64 %i 54 store i32* %tmp0, i32** %tmp1, align 8 55 %i.next = add nuw nsw i64 %i, 1 56 %cond = icmp slt i64 %i.next, %n 57 br i1 %cond, label %for.body, label %for.end 58 59for.end: 60 ret void 61} 62