1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -passes=loop-vectorize -force-vector-interleave=4 -pass-remarks='loop-vectorize' -disable-output -S 2>&1 | FileCheck %s --check-prefix=CHECK-REMARKS 3; RUN: opt < %s -passes=loop-vectorize -force-vector-interleave=4 -S | FileCheck %s 4 5; These tests are to check that fold-tail procedure produces correct scalar code when 6; loop-vectorization is only unrolling but not vectorizing. 7 8; CHECK-REMARKS: remark: {{.*}} interleaved loop (interleaved count: 4) 9; CHECK-REMARKS-NEXT: remark: {{.*}} interleaved loop (interleaved count: 4) 10; CHECK-REMARKS-NOT: remark: {{.*}} vectorized loop 11 12define void @VF1-VPlanExe(ptr %dst) { 13; CHECK-LABEL: @VF1-VPlanExe( 14; CHECK-NEXT: entry: 15; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 16; CHECK: vector.ph: 17; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 18; CHECK: vector.body: 19; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ] 20; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 21; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1 22; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 2 23; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 3 24; CHECK-NEXT: [[TMP4:%.*]] = icmp ule i64 [[TMP0]], 14 25; CHECK-NEXT: [[TMP5:%.*]] = icmp ule i64 [[TMP1]], 14 26; CHECK-NEXT: [[TMP6:%.*]] = icmp ule i64 [[TMP2]], 14 27; CHECK-NEXT: [[TMP7:%.*]] = icmp ule i64 [[TMP3]], 14 28; CHECK-NEXT: br i1 [[TMP4]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]] 29; CHECK: pred.store.if: 30; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 [[TMP0]] 31; CHECK-NEXT: store i32 0, ptr [[TMP8]], align 4 32; CHECK-NEXT: br label [[PRED_STORE_CONTINUE]] 33; CHECK: pred.store.continue: 34; CHECK-NEXT: br i1 [[TMP5]], label [[PRED_STORE_IF1:%.*]], label [[PRED_STORE_CONTINUE2:%.*]] 35; CHECK: pred.store.if1: 36; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP1]] 37; CHECK-NEXT: store i32 0, ptr [[TMP9]], align 4 38; CHECK-NEXT: br label [[PRED_STORE_CONTINUE2]] 39; CHECK: pred.store.continue2: 40; CHECK-NEXT: br i1 [[TMP6]], label [[PRED_STORE_IF3:%.*]], label [[PRED_STORE_CONTINUE4:%.*]] 41; CHECK: pred.store.if3: 42; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP2]] 43; CHECK-NEXT: store i32 0, ptr [[TMP10]], align 4 44; CHECK-NEXT: br label [[PRED_STORE_CONTINUE4]] 45; CHECK: pred.store.continue4: 46; CHECK-NEXT: br i1 [[TMP7]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6]] 47; CHECK: pred.store.if5: 48; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP3]] 49; CHECK-NEXT: store i32 0, ptr [[TMP11]], align 4 50; CHECK-NEXT: br label [[PRED_STORE_CONTINUE6]] 51; CHECK: pred.store.continue6: 52; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 53; CHECK-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16 54; CHECK-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 55; CHECK: middle.block: 56; CHECK-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]] 57; CHECK: scalar.ph: 58; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 59; CHECK-NEXT: br label [[FOR_BODY:%.*]] 60; CHECK: for.cond.cleanup: 61; CHECK-NEXT: ret void 62; CHECK: for.body: 63; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 64; CHECK-NEXT: [[DST_PTR:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDVARS_IV]] 65; CHECK-NEXT: store i32 0, ptr [[DST_PTR]], align 4 66; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 67; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 15 68; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]] 69; 70entry: 71 br label %for.body 72 73for.cond.cleanup: 74 ret void 75 76for.body: 77 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] 78 %dst.ptr = getelementptr inbounds i32, ptr %dst, i64 %indvars.iv 79 store i32 0, ptr %dst.ptr 80 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 81 %exitcond = icmp eq i64 %indvars.iv.next, 15 82 br i1 %exitcond, label %for.cond.cleanup, label %for.body 83} 84 85define void @VF1-VPWidenCanonicalIVRecipeExe(ptr %ptr1) { 86; CHECK-LABEL: @VF1-VPWidenCanonicalIVRecipeExe( 87; CHECK-NEXT: entry: 88; CHECK-NEXT: [[PTR2:%.*]] = getelementptr inbounds double, ptr [[PTR1:%.*]], i64 15 89; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 90; CHECK: vector.ph: 91; CHECK-NEXT: [[IND_END:%.*]] = getelementptr i8, ptr [[PTR1]], i64 128 92; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 93; CHECK: vector.body: 94; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE12:%.*]] ] 95; CHECK-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8 96; CHECK-NEXT: [[VEC_IV:%.*]] = add i64 [[INDEX]], 0 97; CHECK-NEXT: [[VEC_IV4:%.*]] = add i64 [[INDEX]], 1 98; CHECK-NEXT: [[VEC_IV5:%.*]] = add i64 [[INDEX]], 2 99; CHECK-NEXT: [[VEC_IV6:%.*]] = add i64 [[INDEX]], 3 100; CHECK-NEXT: [[TMP0:%.*]] = icmp ule i64 [[VEC_IV]], 14 101; CHECK-NEXT: [[TMP1:%.*]] = icmp ule i64 [[VEC_IV4]], 14 102; CHECK-NEXT: [[TMP2:%.*]] = icmp ule i64 [[VEC_IV5]], 14 103; CHECK-NEXT: [[TMP3:%.*]] = icmp ule i64 [[VEC_IV6]], 14 104; CHECK-NEXT: br i1 [[TMP0]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]] 105; CHECK: pred.store.if: 106; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 0 107; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR1]], i64 [[TMP4]] 108; CHECK-NEXT: store double 0.000000e+00, ptr [[NEXT_GEP]], align 8 109; CHECK-NEXT: br label [[PRED_STORE_CONTINUE]] 110; CHECK: pred.store.continue: 111; CHECK-NEXT: br i1 [[TMP1]], label [[PRED_STORE_IF7:%.*]], label [[PRED_STORE_CONTINUE8:%.*]] 112; CHECK: pred.store.if7: 113; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 8 114; CHECK-NEXT: [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[PTR1]], i64 [[TMP5]] 115; CHECK-NEXT: store double 0.000000e+00, ptr [[NEXT_GEP1]], align 8 116; CHECK-NEXT: br label [[PRED_STORE_CONTINUE8]] 117; CHECK: pred.store.continue8: 118; CHECK-NEXT: br i1 [[TMP2]], label [[PRED_STORE_IF9:%.*]], label [[PRED_STORE_CONTINUE10:%.*]] 119; CHECK: pred.store.if9: 120; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 16 121; CHECK-NEXT: [[NEXT_GEP2:%.*]] = getelementptr i8, ptr [[PTR1]], i64 [[TMP6]] 122; CHECK-NEXT: store double 0.000000e+00, ptr [[NEXT_GEP2]], align 8 123; CHECK-NEXT: br label [[PRED_STORE_CONTINUE10]] 124; CHECK: pred.store.continue10: 125; CHECK-NEXT: br i1 [[TMP3]], label [[PRED_STORE_IF11:%.*]], label [[PRED_STORE_CONTINUE12]] 126; CHECK: pred.store.if11: 127; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], 24 128; CHECK-NEXT: [[NEXT_GEP3:%.*]] = getelementptr i8, ptr [[PTR1]], i64 [[TMP7]] 129; CHECK-NEXT: store double 0.000000e+00, ptr [[NEXT_GEP3]], align 8 130; CHECK-NEXT: br label [[PRED_STORE_CONTINUE12]] 131; CHECK: pred.store.continue12: 132; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 133; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16 134; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] 135; CHECK: middle.block: 136; CHECK-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]] 137; CHECK: scalar.ph: 138; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[PTR1]], [[ENTRY:%.*]] ] 139; CHECK-NEXT: br label [[FOR_BODY:%.*]] 140; CHECK: for.cond.cleanup: 141; CHECK-NEXT: ret void 142; CHECK: for.body: 143; CHECK-NEXT: [[ADDR:%.*]] = phi ptr [ [[PTR:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] 144; CHECK-NEXT: store double 0.000000e+00, ptr [[ADDR]], align 8 145; CHECK-NEXT: [[PTR]] = getelementptr inbounds double, ptr [[ADDR]], i64 1 146; CHECK-NEXT: [[COND:%.*]] = icmp eq ptr [[PTR]], [[PTR2]] 147; CHECK-NEXT: br i1 [[COND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] 148; 149entry: 150 %ptr2 = getelementptr inbounds double, ptr %ptr1, i64 15 151 br label %for.body 152 153for.cond.cleanup: 154 ret void 155 156for.body: 157 %addr = phi ptr [ %ptr, %for.body ], [ %ptr1, %entry ] 158 store double 0.0, ptr %addr 159 %ptr = getelementptr inbounds double, ptr %addr, i64 1 160 %cond = icmp eq ptr %ptr, %ptr2 161 br i1 %cond, label %for.cond.cleanup, label %for.body 162} 163;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: 164; CHECK-REMARKS: {{.*}} 165