1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -passes=loop-vectorize,instcombine -force-vector-width=4 -force-vector-interleave=1 -S | FileCheck %s 3 4target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128" 5 6; 7define void @vector_gep_stored(ptr %a, ptr %b, i64 %n) { 8; CHECK-LABEL: @vector_gep_stored( 9; CHECK-NEXT: entry: 10; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1) 11; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 4 12; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 13; CHECK: vector.ph: 14; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804 15; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 16; CHECK: vector.body: 17; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 18; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ] 19; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], <4 x i64> [[VEC_IND]] 20; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds ptr, ptr [[A:%.*]], i64 [[INDEX]] 21; CHECK-NEXT: store <4 x ptr> [[TMP0]], ptr [[TMP1]], align 8 22; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 23; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4) 24; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 25; CHECK-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 26; CHECK: middle.block: 27; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]] 28; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] 29; CHECK: scalar.ph: 30; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 31; CHECK-NEXT: br label [[FOR_BODY:%.*]] 32; CHECK: for.body: 33; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] 34; CHECK-NEXT: [[VAR0:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[I]] 35; CHECK-NEXT: [[VAR1:%.*]] = getelementptr inbounds nuw ptr, ptr [[A]], i64 [[I]] 36; CHECK-NEXT: store ptr [[VAR0]], ptr [[VAR1]], align 8 37; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1 38; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]] 39; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP3:![0-9]+]] 40; CHECK: for.end: 41; CHECK-NEXT: ret void 42; 43entry: 44 br label %for.body 45 46for.body: 47 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ] 48 %var0 = getelementptr inbounds i32, ptr %b, i64 %i 49 %var1 = getelementptr inbounds ptr, ptr %a, i64 %i 50 store ptr %var0, ptr %var1, align 8 51 %i.next = add nuw nsw i64 %i, 1 52 %cond = icmp slt i64 %i.next, %n 53 br i1 %cond, label %for.body, label %for.end 54 55for.end: 56 ret void 57} 58 59; 60define void @uniform_vector_gep_stored(ptr %a, ptr %b, i64 %n) { 61; CHECK-LABEL: @uniform_vector_gep_stored( 62; CHECK-NEXT: entry: 63; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1) 64; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 4 65; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 66; CHECK: vector.ph: 67; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804 68; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds nuw i8, ptr [[B:%.*]], i64 4 69; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x ptr> poison, ptr [[TMP0]], i64 0 70; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x ptr> [[DOTSPLATINSERT]], <4 x ptr> poison, <4 x i32> zeroinitializer 71; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 72; CHECK: vector.body: 73; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 74; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds ptr, ptr [[A:%.*]], i64 [[INDEX]] 75; CHECK-NEXT: store <4 x ptr> [[DOTSPLAT]], ptr [[TMP1]], align 8 76; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 77; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 78; CHECK-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] 79; CHECK: middle.block: 80; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]] 81; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] 82; CHECK: scalar.ph: 83; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 84; CHECK-NEXT: br label [[FOR_BODY:%.*]] 85; CHECK: for.body: 86; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] 87; CHECK-NEXT: [[VAR0:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 4 88; CHECK-NEXT: [[VAR1:%.*]] = getelementptr inbounds nuw ptr, ptr [[A]], i64 [[I]] 89; CHECK-NEXT: store ptr [[VAR0]], ptr [[VAR1]], align 8 90; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1 91; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]] 92; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP5:![0-9]+]] 93; CHECK: for.end: 94; CHECK-NEXT: ret void 95; 96entry: 97 br label %for.body 98 99for.body: 100 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ] 101 %var0 = getelementptr inbounds i32, ptr %b, i64 1 102 %var1 = getelementptr inbounds ptr, ptr %a, i64 %i 103 store ptr %var0, ptr %var1, align 8 104 %i.next = add nuw nsw i64 %i, 1 105 %cond = icmp slt i64 %i.next, %n 106 br i1 %cond, label %for.body, label %for.end 107 108for.end: 109 ret void 110} 111