1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -S -loop-vectorize < %s -mcpu=skylake | FileCheck %s
3target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128-ni:10:11:12:13"
4target triple = "x86_64-unknown-linux-gnu"
5
6@jlplt_ijl_alloc_array_1d_10294_got = external dso_local local_unnamed_addr global void ()*
7
8define {} addrspace(10)* @japi1_vect_42283({} addrspace(10)** nocapture readonly %0, i32 %1) local_unnamed_addr #0 {
9; CHECK-LABEL: @japi1_vect_42283(
10; CHECK-NEXT:  top:
11; CHECK-NEXT:    [[TMP2:%.*]] = sext i32 [[TMP1:%.*]] to i64
12; CHECK-NEXT:    [[TMP3:%.*]] = load atomic {} addrspace(10)* ({} addrspace(10)*, i64)*, {} addrspace(10)* ({} addrspace(10)*, i64)** bitcast (void ()** @jlplt_ijl_alloc_array_1d_10294_got to {} addrspace(10)* ({} addrspace(10)*, i64)**) unordered, align 8
13; CHECK-NEXT:    [[TMP4:%.*]] = tail call {} addrspace(10)* [[TMP3]]({} addrspace(10)* null, i64 0)
14; CHECK-NEXT:    [[TMP5:%.*]] = bitcast {} addrspace(10)** [[TMP0:%.*]] to { {} addrspace(10)*, i64 } addrspace(10)**
15; CHECK-NEXT:    [[TMP6:%.*]] = load { {} addrspace(10)*, i64 } addrspace(10)*, { {} addrspace(10)*, i64 } addrspace(10)** [[TMP5]], align 8, !tbaa [[TBAA0:![0-9]+]]
16; CHECK-NEXT:    [[TMP7:%.*]] = bitcast {} addrspace(10)* [[TMP4]] to { {} addrspace(10)*, i64 } addrspace(13)* addrspace(10)*
17; CHECK-NEXT:    [[TMP8:%.*]] = addrspacecast { {} addrspace(10)*, i64 } addrspace(13)* addrspace(10)* [[TMP7]] to { {} addrspace(10)*, i64 } addrspace(13)* addrspace(11)*
18; CHECK-NEXT:    [[TMP9:%.*]] = load { {} addrspace(10)*, i64 } addrspace(13)*, { {} addrspace(10)*, i64 } addrspace(13)* addrspace(11)* [[TMP8]], align 8, !tbaa [[TBAA5:![0-9]+]]
19; CHECK-NEXT:    [[TMP10:%.*]] = bitcast { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]] to i8 addrspace(13)*
20; CHECK-NEXT:    [[DOTELT:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(10)* [[TMP6]], i64 0, i32 0
21; CHECK-NEXT:    [[DOTUNPACK:%.*]] = load {} addrspace(10)*, {} addrspace(10)* addrspace(10)* [[DOTELT]], align 8, !tbaa [[TBAA8:![0-9]+]]
22; CHECK-NEXT:    [[DOTELT1:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(10)* [[TMP6]], i64 0, i32 1
23; CHECK-NEXT:    [[DOTUNPACK2:%.*]] = load i64, i64 addrspace(10)* [[DOTELT1]], align 8, !tbaa [[TBAA8]]
24; CHECK-NEXT:    [[TMP11:%.*]] = add nsw i64 [[TMP2]], 1
25; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP11]], 28
26; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
27; CHECK:       vector.scevcheck:
28; CHECK-NEXT:    [[MUL:%.*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 16, i64 [[TMP2]])
29; CHECK-NEXT:    [[MUL_RESULT:%.*]] = extractvalue { i64, i1 } [[MUL]], 0
30; CHECK-NEXT:    [[MUL_OVERFLOW:%.*]] = extractvalue { i64, i1 } [[MUL]], 1
31; CHECK-NEXT:    [[TMP12:%.*]] = sub i64 0, [[MUL_RESULT]]
32; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i8, i8 addrspace(13)* [[TMP10]], i64 [[MUL_RESULT]]
33; CHECK-NEXT:    [[TMP14:%.*]] = icmp ult i8 addrspace(13)* [[TMP13]], [[TMP10]]
34; CHECK-NEXT:    [[TMP15:%.*]] = or i1 [[TMP14]], [[MUL_OVERFLOW]]
35; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], i64 0, i32 1
36; CHECK-NEXT:    [[SCEVGEP1:%.*]] = bitcast i64 addrspace(13)* [[SCEVGEP]] to { {} addrspace(10)*, i64 } addrspace(13)*
37; CHECK-NEXT:    [[MUL2:%.*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 16, i64 [[TMP2]])
38; CHECK-NEXT:    [[MUL_RESULT3:%.*]] = extractvalue { i64, i1 } [[MUL2]], 0
39; CHECK-NEXT:    [[MUL_OVERFLOW4:%.*]] = extractvalue { i64, i1 } [[MUL2]], 1
40; CHECK-NEXT:    [[SCEVGEP15:%.*]] = bitcast { {} addrspace(10)*, i64 } addrspace(13)* [[SCEVGEP1]] to i8 addrspace(13)*
41; CHECK-NEXT:    [[TMP16:%.*]] = sub i64 0, [[MUL_RESULT3]]
42; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i8, i8 addrspace(13)* [[SCEVGEP15]], i64 [[MUL_RESULT3]]
43; CHECK-NEXT:    [[TMP18:%.*]] = icmp ult i8 addrspace(13)* [[TMP17]], [[SCEVGEP15]]
44; CHECK-NEXT:    [[TMP19:%.*]] = or i1 [[TMP18]], [[MUL_OVERFLOW4]]
45; CHECK-NEXT:    [[TMP20:%.*]] = or i1 [[TMP15]], [[TMP19]]
46; CHECK-NEXT:    br i1 [[TMP20]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
47; CHECK:       vector.ph:
48; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP11]], 16
49; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP11]], [[N_MOD_VF]]
50; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x {} addrspace(10)*> poison, {} addrspace(10)* [[DOTUNPACK]], i32 0
51; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x {} addrspace(10)*> [[BROADCAST_SPLATINSERT]], <4 x {} addrspace(10)*> poison, <4 x i32> zeroinitializer
52; CHECK-NEXT:    [[BROADCAST_SPLATINSERT9:%.*]] = insertelement <4 x {} addrspace(10)*> poison, {} addrspace(10)* [[DOTUNPACK]], i32 0
53; CHECK-NEXT:    [[BROADCAST_SPLAT10:%.*]] = shufflevector <4 x {} addrspace(10)*> [[BROADCAST_SPLATINSERT9]], <4 x {} addrspace(10)*> poison, <4 x i32> zeroinitializer
54; CHECK-NEXT:    [[BROADCAST_SPLATINSERT11:%.*]] = insertelement <4 x {} addrspace(10)*> poison, {} addrspace(10)* [[DOTUNPACK]], i32 0
55; CHECK-NEXT:    [[BROADCAST_SPLAT12:%.*]] = shufflevector <4 x {} addrspace(10)*> [[BROADCAST_SPLATINSERT11]], <4 x {} addrspace(10)*> poison, <4 x i32> zeroinitializer
56; CHECK-NEXT:    [[BROADCAST_SPLATINSERT13:%.*]] = insertelement <4 x {} addrspace(10)*> poison, {} addrspace(10)* [[DOTUNPACK]], i32 0
57; CHECK-NEXT:    [[BROADCAST_SPLAT14:%.*]] = shufflevector <4 x {} addrspace(10)*> [[BROADCAST_SPLATINSERT13]], <4 x {} addrspace(10)*> poison, <4 x i32> zeroinitializer
58; CHECK-NEXT:    [[BROADCAST_SPLATINSERT15:%.*]] = insertelement <4 x i64> poison, i64 [[DOTUNPACK2]], i32 0
59; CHECK-NEXT:    [[BROADCAST_SPLAT16:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT15]], <4 x i64> poison, <4 x i32> zeroinitializer
60; CHECK-NEXT:    [[BROADCAST_SPLATINSERT17:%.*]] = insertelement <4 x i64> poison, i64 [[DOTUNPACK2]], i32 0
61; CHECK-NEXT:    [[BROADCAST_SPLAT18:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT17]], <4 x i64> poison, <4 x i32> zeroinitializer
62; CHECK-NEXT:    [[BROADCAST_SPLATINSERT19:%.*]] = insertelement <4 x i64> poison, i64 [[DOTUNPACK2]], i32 0
63; CHECK-NEXT:    [[BROADCAST_SPLAT20:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT19]], <4 x i64> poison, <4 x i32> zeroinitializer
64; CHECK-NEXT:    [[BROADCAST_SPLATINSERT21:%.*]] = insertelement <4 x i64> poison, i64 [[DOTUNPACK2]], i32 0
65; CHECK-NEXT:    [[BROADCAST_SPLAT22:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT21]], <4 x i64> poison, <4 x i32> zeroinitializer
66; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
67; CHECK:       vector.body:
68; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
69; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
70; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <4 x i64> [[VEC_IND]], <i64 4, i64 4, i64 4, i64 4>
71; CHECK-NEXT:    [[STEP_ADD6:%.*]] = add <4 x i64> [[STEP_ADD]], <i64 4, i64 4, i64 4, i64 4>
72; CHECK-NEXT:    [[STEP_ADD7:%.*]] = add <4 x i64> [[STEP_ADD6]], <i64 4, i64 4, i64 4, i64 4>
73; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], <4 x i64> [[VEC_IND]], i32 0
74; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], <4 x i64> [[STEP_ADD]], i32 0
75; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], <4 x i64> [[STEP_ADD6]], i32 0
76; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], <4 x i64> [[STEP_ADD7]], i32 0
77; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10sl_s.v4p13p10sl_s(<4 x {} addrspace(10)*> [[BROADCAST_SPLAT]], <4 x {} addrspace(10)* addrspace(13)*> [[TMP21]], i32 8, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa [[TBAA10:![0-9]+]]
78; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10sl_s.v4p13p10sl_s(<4 x {} addrspace(10)*> [[BROADCAST_SPLAT10]], <4 x {} addrspace(10)* addrspace(13)*> [[TMP22]], i32 8, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa [[TBAA10]]
79; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10sl_s.v4p13p10sl_s(<4 x {} addrspace(10)*> [[BROADCAST_SPLAT12]], <4 x {} addrspace(10)* addrspace(13)*> [[TMP23]], i32 8, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa [[TBAA10]]
80; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10sl_s.v4p13p10sl_s(<4 x {} addrspace(10)*> [[BROADCAST_SPLAT14]], <4 x {} addrspace(10)* addrspace(13)*> [[TMP24]], i32 8, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa [[TBAA10]]
81; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], <4 x i64> [[VEC_IND]], i32 1
82; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], <4 x i64> [[STEP_ADD]], i32 1
83; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], <4 x i64> [[STEP_ADD6]], i32 1
84; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], <4 x i64> [[STEP_ADD7]], i32 1
85; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13i64(<4 x i64> [[BROADCAST_SPLAT16]], <4 x i64 addrspace(13)*> [[TMP25]], i32 8, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa [[TBAA10]]
86; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13i64(<4 x i64> [[BROADCAST_SPLAT18]], <4 x i64 addrspace(13)*> [[TMP26]], i32 8, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa [[TBAA10]]
87; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13i64(<4 x i64> [[BROADCAST_SPLAT20]], <4 x i64 addrspace(13)*> [[TMP27]], i32 8, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa [[TBAA10]]
88; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13i64(<4 x i64> [[BROADCAST_SPLAT22]], <4 x i64 addrspace(13)*> [[TMP28]], i32 8, <4 x i1> <i1 true, i1 true, i1 true, i1 true>), !tbaa [[TBAA10]]
89; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
90; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[STEP_ADD7]], <i64 4, i64 4, i64 4, i64 4>
91; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
92; CHECK-NEXT:    br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
93; CHECK:       middle.block:
94; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP11]], [[N_VEC]]
95; CHECK-NEXT:    br i1 [[CMP_N]], label [[L44:%.*]], label [[SCALAR_PH]]
96; CHECK:       scalar.ph:
97; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[TOP:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
98; CHECK-NEXT:    br label [[L26:%.*]]
99; CHECK:       L26:
100; CHECK-NEXT:    [[VALUE_PHI5:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP30:%.*]], [[L26]] ]
101; CHECK-NEXT:    [[DOTREPACK:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], i64 [[VALUE_PHI5]], i32 0
102; CHECK-NEXT:    store {} addrspace(10)* [[DOTUNPACK]], {} addrspace(10)* addrspace(13)* [[DOTREPACK]], align 8, !tbaa [[TBAA10]]
103; CHECK-NEXT:    [[DOTREPACK4:%.*]] = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* [[TMP9]], i64 [[VALUE_PHI5]], i32 1
104; CHECK-NEXT:    store i64 [[DOTUNPACK2]], i64 addrspace(13)* [[DOTREPACK4]], align 8, !tbaa [[TBAA10]]
105; CHECK-NEXT:    [[TMP30]] = add i64 [[VALUE_PHI5]], 1
106; CHECK-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[VALUE_PHI5]], [[TMP2]]
107; CHECK-NEXT:    br i1 [[DOTNOT]], label [[L44]], label [[L26]], !llvm.loop [[LOOP14:![0-9]+]]
108; CHECK:       L44:
109; CHECK-NEXT:    ret {} addrspace(10)* null
110;
111top:
112  %2 = sext i32 %1 to i64
113  %3 = load atomic {} addrspace(10)* ({} addrspace(10)*, i64)*, {} addrspace(10)* ({} addrspace(10)*, i64)** bitcast (void ()** @jlplt_ijl_alloc_array_1d_10294_got to {} addrspace(10)* ({} addrspace(10)*, i64)**) unordered, align 8
114  %4 = tail call {} addrspace(10)* %3({} addrspace(10)* null, i64 0)
115  %5 = bitcast {} addrspace(10)** %0 to { {} addrspace(10)*, i64 } addrspace(10)**
116  %6 = load { {} addrspace(10)*, i64 } addrspace(10)*, { {} addrspace(10)*, i64 } addrspace(10)** %5, align 8, !tbaa !0
117  %7 = bitcast {} addrspace(10)* %4 to { {} addrspace(10)*, i64 } addrspace(13)* addrspace(10)*
118  %8 = addrspacecast { {} addrspace(10)*, i64 } addrspace(13)* addrspace(10)* %7 to { {} addrspace(10)*, i64 } addrspace(13)* addrspace(11)*
119  %9 = load { {} addrspace(10)*, i64 } addrspace(13)*, { {} addrspace(10)*, i64 } addrspace(13)* addrspace(11)* %8, align 8, !tbaa !5
120  %.elt = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(10)* %6, i64 0, i32 0
121  %.unpack = load {} addrspace(10)*, {} addrspace(10)* addrspace(10)* %.elt, align 8, !tbaa !8
122  %.elt1 = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(10)* %6, i64 0, i32 1
123  %.unpack2 = load i64, i64 addrspace(10)* %.elt1, align 8, !tbaa !8
124  br label %L26
125
126L26:                                              ; preds = %L26, %top
127  %value_phi5 = phi i64 [ 0, %top ], [ %10, %L26 ]
128  %.repack = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* %9, i64 %value_phi5, i32 0
129  store {} addrspace(10)* %.unpack, {} addrspace(10)* addrspace(13)* %.repack, align 8, !tbaa !10
130  %.repack4 = getelementptr inbounds { {} addrspace(10)*, i64 }, { {} addrspace(10)*, i64 } addrspace(13)* %9, i64 %value_phi5, i32 1
131  store i64 %.unpack2, i64 addrspace(13)* %.repack4, align 8, !tbaa !10
132  %10 = add i64 %value_phi5, 1
133  %.not = icmp eq i64 %value_phi5, %2
134  br i1 %.not, label %L44, label %L26
135
136L44:                                              ; preds = %L26
137  ret {} addrspace(10)* null
138}
139
140attributes #0 = { "target-cpu"="skylake-avx512" "target-features"="+xsaves,+xsavec,+prfchw,+lzcnt,+sahf,+pku,+avx512vl,+avx512bw,+avx512cd,+clwb,+clflushopt,+adx,+avx512dq,+avx512f,+bmi2,+avx2,+bmi,+fsgsbase,+f16c,+avx,+xsave,+aes,+popcnt,+movbe,+sse4.2,+sse4.1,+cx16,+fma,+ssse3,+pclmul,+sse3,-rdrnd,-rtm,-rdseed,-avx512ifma,-avx512pf,-avx512er,-sha,-prefetchwt1,-avx512vbmi,-waitpkg,-avx512vbmi2,-shstk,-gfni,-vaes,-vpclmulqdq,-avx512vnni,-avx512bitalg,-avx512vpopcntdq,-rdpid,-cldemote,-movdiri,-movdir64b,-enqcmd,-avx512vp2intersect,-serialize,-tsxldtrk,-pconfig,-amx-bf16,-amx-tile,-amx-int8,-sse4a,-xop,-lwp,-fma4,-tbm,-mwaitx,-xsaveopt,-clzero,-wbnoinvd,-avx512bf16,-ptwrite,+sse2,+mmx,+fxsr,+64bit,+cx8" }
141attributes #1 = { inaccessiblemem_or_argmemonly }
142attributes #2 = { allocsize(1) }
143
144!0 = !{!1, !1, i64 0}
145!1 = !{!"jtbaa_value", !2, i64 0}
146!2 = !{!"jtbaa_data", !3, i64 0}
147!3 = !{!"jtbaa", !4, i64 0}
148!4 = !{!"jtbaa"}
149!5 = !{!6, !6, i64 0}
150!6 = !{!"jtbaa_arrayptr", !7, i64 0}
151!7 = !{!"jtbaa_array", !3, i64 0}
152!8 = !{!9, !9, i64 0}
153!9 = !{!"jtbaa_immut", !1, i64 0}
154!10 = !{!11, !11, i64 0}
155!11 = !{!"jtbaa_arraybuf", !2, i64 0}
156