1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt -passes=loop-vectorize -force-vector-width=4 -S %s | FileCheck %s 3 4define void @test(ptr %A, i32 %x) { 5; CHECK-LABEL: @test( 6; CHECK-NEXT: entry: 7; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]] 8; CHECK: vector.scevcheck: 9; CHECK-NEXT: [[IDENT_CHECK:%.*]] = icmp ne i32 [[X:%.*]], 1 10; CHECK-NEXT: br i1 [[IDENT_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] 11; CHECK: vector.ph: 12; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 13; CHECK: vector.body: 14; CHECK-NEXT: [[OFFSET_IDX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 15; CHECK-NEXT: [[TMP0:%.*]] = trunc i64 [[OFFSET_IDX]] to i32 16; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], 0 17; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[OFFSET_IDX]], 0 18; CHECK-NEXT: [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1 19; CHECK-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32 20; CHECK-NEXT: [[TMP5:%.*]] = mul i32 [[TMP4]], [[X]] 21; CHECK-NEXT: [[TMP6:%.*]] = zext i32 [[TMP5]] to i64 22; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[TMP6]] 23; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0 24; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP8]], align 4 25; CHECK-NEXT: [[TMP9:%.*]] = mul i32 [[TMP1]], [[X]] 26; CHECK-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64 27; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP10]] 28; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0 29; CHECK-NEXT: store <4 x float> [[WIDE_LOAD]], ptr [[TMP12]], align 4 30; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[OFFSET_IDX]], 4 31; CHECK-NEXT: [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef 32; CHECK-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 33; CHECK: middle.block: 34; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 undef, undef 35; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 36; CHECK: scalar.ph: 37; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ] 38; CHECK-NEXT: br label [[LOOP:%.*]] 39; CHECK: loop: 40; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] 41; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 42; CHECK-NEXT: [[T_IV_NEXT:%.*]] = trunc i64 [[IV_NEXT]] to i32 43; CHECK-NEXT: [[MUL_IV_NEXT:%.*]] = mul i32 [[T_IV_NEXT]], [[X]] 44; CHECK-NEXT: [[IDX_1:%.*]] = zext i32 [[MUL_IV_NEXT]] to i64 45; CHECK-NEXT: [[ARRAYIDX1215:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IDX_1]] 46; CHECK-NEXT: [[LV:%.*]] = load float, ptr [[ARRAYIDX1215]], align 4 47; CHECK-NEXT: [[T_IV:%.*]] = trunc i64 [[IV]] to i32 48; CHECK-NEXT: [[MUL_IV:%.*]] = mul i32 [[T_IV]], [[X]] 49; CHECK-NEXT: [[IDX_2:%.*]] = zext i32 [[MUL_IV]] to i64 50; CHECK-NEXT: [[ARRAYIDX1209:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IDX_2]] 51; CHECK-NEXT: store float [[LV]], ptr [[ARRAYIDX1209]], align 4 52; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], undef 53; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP2:![0-9]+]] 54; CHECK: exit: 55; CHECK-NEXT: ret void 56; 57entry: 58 br label %loop 59 60loop: ; preds = %loop, %entry 61 %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ] 62 %iv.next = add nuw nsw i64 %iv, 1 63 %t.iv.next = trunc i64 %iv.next to i32 64 %mul.iv.next = mul i32 %t.iv.next, %x 65 %idx.1 = zext i32 %mul.iv.next to i64 66 %arrayidx1215 = getelementptr inbounds float, ptr %A, i64 %idx.1 67 %lv = load float, ptr %arrayidx1215, align 4 68 69 %t.iv = trunc i64 %iv to i32 70 %mul.iv = mul i32 %t.iv, %x 71 %idx.2 = zext i32 %mul.iv to i64 72 %arrayidx1209 = getelementptr inbounds float, ptr %A, i64 %idx.2 73 store float %lv, ptr %arrayidx1209, align 4 74 %ec = icmp eq i64 %iv.next, undef 75 br i1 %ec, label %exit, label %loop 76 77exit: ; preds = %loop 78 ret void 79} 80