xref: /llvm-project/llvm/test/Transforms/LoopVectorize/X86/pr54634.ll (revision 4ad0fdd1631eeae432714c03ede01a10dc00025d) 
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -S -passes=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 ptr
7
8define ptr addrspace(10) @japi1_vect_42283(ptr nocapture readonly %0, i32 %1) local_unnamed_addr #0 {
9; CHECK-LABEL: @japi1_vect_42283(
10; CHECK-NEXT:  iter.check:
11; CHECK-NEXT:    [[TMP2:%.*]] = sext i32 [[TMP1:%.*]] to i64
12; CHECK-NEXT:    [[TMP3:%.*]] = load atomic ptr, ptr @jlplt_ijl_alloc_array_1d_10294_got unordered, align 8
13; CHECK-NEXT:    [[TMP4:%.*]] = tail call ptr addrspace(10) [[TMP3]](ptr addrspace(10) null, i64 0)
14; CHECK-NEXT:    [[TMP5:%.*]] = load ptr addrspace(10), ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA0:![0-9]+]]
15; CHECK-NEXT:    [[TMP6:%.*]] = addrspacecast ptr addrspace(10) [[TMP4]] to ptr addrspace(11)
16; CHECK-NEXT:    [[TMP7:%.*]] = load ptr addrspace(13), ptr addrspace(11) [[TMP6]], align 8, !tbaa [[TBAA5:![0-9]+]]
17; CHECK-NEXT:    [[DOTELT:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(10) [[TMP5]], i64 0, i32 0
18; CHECK-NEXT:    [[DOTUNPACK:%.*]] = load ptr addrspace(10), ptr addrspace(10) [[DOTELT]], align 8, !tbaa [[TBAA8:![0-9]+]]
19; CHECK-NEXT:    [[DOTELT1:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(10) [[TMP5]], i64 0, i32 1
20; CHECK-NEXT:    [[DOTUNPACK2:%.*]] = load i64, ptr addrspace(10) [[DOTELT1]], align 8, !tbaa [[TBAA8]]
21; CHECK-NEXT:    [[TMP8:%.*]] = add nsw i64 [[TMP2]], 1
22; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP8]], 4
23; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[TOP:%.*]]
24; CHECK:       vector.main.loop.iter.check:
25; CHECK-NEXT:    [[TMP17:%.*]] = icmp ult i64 [[TMP8]], 16
26; CHECK-NEXT:    br i1 [[TMP17]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
27; CHECK:       vector.ph:
28; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP8]], 16
29; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP8]], [[N_MOD_VF]]
30; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x ptr addrspace(10)> poison, ptr addrspace(10) [[DOTUNPACK]], i64 0
31; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x ptr addrspace(10)> [[BROADCAST_SPLATINSERT]], <4 x ptr addrspace(10)> poison, <4 x i32> zeroinitializer
32; CHECK-NEXT:    [[BROADCAST_SPLATINSERT7:%.*]] = insertelement <4 x i64> poison, i64 [[DOTUNPACK2]], i64 0
33; CHECK-NEXT:    [[BROADCAST_SPLAT8:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT7]], <4 x i64> poison, <4 x i32> zeroinitializer
34; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
35; CHECK:       vector.body:
36; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
37; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
38; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
39; CHECK-NEXT:    [[STEP_ADD4:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
40; CHECK-NEXT:    [[STEP_ADD5:%.*]] = add <4 x i64> [[STEP_ADD4]], splat (i64 4)
41; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[VEC_IND]], i32 0
42; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[STEP_ADD]], i32 0
43; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[STEP_ADD4]], i32 0
44; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[STEP_ADD5]], i32 0
45; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10.v4p13(<4 x ptr addrspace(10)> [[BROADCAST_SPLAT]], <4 x ptr addrspace(13)> [[TMP18]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10:![0-9]+]]
46; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10.v4p13(<4 x ptr addrspace(10)> [[BROADCAST_SPLAT]], <4 x ptr addrspace(13)> [[TMP19]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
47; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10.v4p13(<4 x ptr addrspace(10)> [[BROADCAST_SPLAT]], <4 x ptr addrspace(13)> [[TMP20]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
48; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10.v4p13(<4 x ptr addrspace(10)> [[BROADCAST_SPLAT]], <4 x ptr addrspace(13)> [[TMP21]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
49; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[VEC_IND]], i32 1
50; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[STEP_ADD]], i32 1
51; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[STEP_ADD4]], i32 1
52; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[STEP_ADD5]], i32 1
53; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13(<4 x i64> [[BROADCAST_SPLAT8]], <4 x ptr addrspace(13)> [[TMP22]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
54; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13(<4 x i64> [[BROADCAST_SPLAT8]], <4 x ptr addrspace(13)> [[TMP23]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
55; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13(<4 x i64> [[BROADCAST_SPLAT8]], <4 x ptr addrspace(13)> [[TMP24]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
56; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13(<4 x i64> [[BROADCAST_SPLAT8]], <4 x ptr addrspace(13)> [[TMP25]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
57; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
58; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[STEP_ADD5]], splat (i64 4)
59; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
60; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK1:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
61; CHECK:       middle.block:
62; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP8]], [[N_VEC]]
63; CHECK-NEXT:    br i1 [[CMP_N]], label [[L44:%.*]], label [[MIDDLE_BLOCK:%.*]]
64; CHECK:       vec.epilog.iter.check:
65; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[TMP8]], [[N_VEC]]
66; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
67; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[SCALAR_PH]]
68; CHECK:       vec.epilog.ph:
69; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[TOP]] ]
70; CHECK-NEXT:    [[N_MOD_VF4:%.*]] = urem i64 [[TMP8]], 4
71; CHECK-NEXT:    [[N_VEC5:%.*]] = sub i64 [[TMP8]], [[N_MOD_VF4]]
72; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[VEC_EPILOG_RESUME_VAL]], i64 0
73; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i64> [[DOTSPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
74; CHECK-NEXT:    [[INDUCTION:%.*]] = add <4 x i64> [[DOTSPLAT]], <i64 0, i64 1, i64 2, i64 3>
75; CHECK-NEXT:    [[BROADCAST_SPLATINSERT10:%.*]] = insertelement <4 x ptr addrspace(10)> poison, ptr addrspace(10) [[DOTUNPACK]], i64 0
76; CHECK-NEXT:    [[BROADCAST_SPLAT11:%.*]] = shufflevector <4 x ptr addrspace(10)> [[BROADCAST_SPLATINSERT10]], <4 x ptr addrspace(10)> poison, <4 x i32> zeroinitializer
77; CHECK-NEXT:    [[BROADCAST_SPLATINSERT12:%.*]] = insertelement <4 x i64> poison, i64 [[DOTUNPACK2]], i64 0
78; CHECK-NEXT:    [[BROADCAST_SPLAT13:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT12]], <4 x i64> poison, <4 x i32> zeroinitializer
79; CHECK-NEXT:    br label [[L26:%.*]]
80; CHECK:       vec.epilog.vector.body:
81; CHECK-NEXT:    [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDEX_NEXT14:%.*]], [[L26]] ]
82; CHECK-NEXT:    [[VEC_IND8:%.*]] = phi <4 x i64> [ [[INDUCTION]], [[SCALAR_PH]] ], [ [[VEC_IND_NEXT9:%.*]], [[L26]] ]
83; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[VEC_IND8]], i32 0
84; CHECK-NEXT:    call void @llvm.masked.scatter.v4p10.v4p13(<4 x ptr addrspace(10)> [[BROADCAST_SPLAT11]], <4 x ptr addrspace(13)> [[TMP28]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
85; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], <4 x i64> [[VEC_IND8]], i32 1
86; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p13(<4 x i64> [[BROADCAST_SPLAT13]], <4 x ptr addrspace(13)> [[TMP29]], i32 8, <4 x i1> splat (i1 true)), !tbaa [[TBAA10]]
87; CHECK-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX7]], 4
88; CHECK-NEXT:    [[VEC_IND_NEXT9]] = add <4 x i64> [[VEC_IND8]], splat (i64 4)
89; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT14]], [[N_VEC5]]
90; CHECK-NEXT:    br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[L26]], !llvm.loop [[LOOP15:![0-9]+]]
91; CHECK:       vec.epilog.middle.block:
92; CHECK-NEXT:    [[CMP_N15:%.*]] = icmp eq i64 [[TMP8]], [[N_VEC5]]
93; CHECK-NEXT:    br i1 [[CMP_N15]], label [[L44]], label [[VEC_EPILOG_SCALAR_PH]]
94; CHECK:       vec.epilog.scalar.ph:
95; CHECK-NEXT:    [[BC_RESUME_VAL6:%.*]] = phi i64 [ [[N_VEC5]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ 0, [[ITER_CHECK:%.*]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
96; CHECK-NEXT:    br label [[L27:%.*]]
97; CHECK:       L26:
98; CHECK-NEXT:    [[VALUE_PHI5:%.*]] = phi i64 [ [[BC_RESUME_VAL6]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[TMP27:%.*]], [[L27]] ]
99; CHECK-NEXT:    [[DOTREPACK:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], i64 [[VALUE_PHI5]], i32 0
100; CHECK-NEXT:    store ptr addrspace(10) [[DOTUNPACK]], ptr addrspace(13) [[DOTREPACK]], align 8, !tbaa [[TBAA10]]
101; CHECK-NEXT:    [[DOTREPACK4:%.*]] = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) [[TMP7]], i64 [[VALUE_PHI5]], i32 1
102; CHECK-NEXT:    store i64 [[DOTUNPACK2]], ptr addrspace(13) [[DOTREPACK4]], align 8, !tbaa [[TBAA10]]
103; CHECK-NEXT:    [[TMP27]] = add i64 [[VALUE_PHI5]], 1
104; CHECK-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[VALUE_PHI5]], [[TMP2]]
105; CHECK-NEXT:    br i1 [[DOTNOT]], label [[L44]], label [[L27]], !llvm.loop [[LOOP16:![0-9]+]]
106; CHECK:       L44:
107; CHECK-NEXT:    ret ptr addrspace(10) null
108;
109top:
110  %2 = sext i32 %1 to i64
111  %3 = load atomic ptr addrspace(10) (ptr addrspace(10), i64)*, ptr addrspace(10) (ptr addrspace(10), i64)** bitcast (ptr @jlplt_ijl_alloc_array_1d_10294_got to ptr addrspace(10) (ptr addrspace(10), i64)**) unordered, align 8
112  %4 = tail call ptr addrspace(10) %3(ptr addrspace(10) null, i64 0)
113  %5 = load ptr addrspace(10), ptr %0, align 8, !tbaa !0
114  %6 = addrspacecast ptr addrspace(10) %4 to ptr addrspace(11)
115  %7 = load ptr addrspace(13), ptr addrspace(11) %6, align 8, !tbaa !5
116  %.elt = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(10) %5, i64 0, i32 0
117  %.unpack = load ptr addrspace(10), ptr addrspace(10) %.elt, align 8, !tbaa !8
118  %.elt1 = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(10) %5, i64 0, i32 1
119  %.unpack2 = load i64, ptr addrspace(10) %.elt1, align 8, !tbaa !8
120  br label %L26
121
122L26:                                              ; preds = %L26, %top
123  %value_phi5 = phi i64 [ 0, %top ], [ %8, %L26 ]
124  %.repack = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) %7, i64 %value_phi5, i32 0
125  store ptr addrspace(10) %.unpack, ptr addrspace(13) %.repack, align 8, !tbaa !10
126  %.repack4 = getelementptr inbounds { ptr addrspace(10), i64 }, ptr addrspace(13) %7, i64 %value_phi5, i32 1
127  store i64 %.unpack2, ptr addrspace(13) %.repack4, align 8, !tbaa !10
128  %8 = add i64 %value_phi5, 1
129  %.not = icmp eq i64 %value_phi5, %2
130  br i1 %.not, label %L44, label %L26
131
132L44:                                              ; preds = %L26
133  ret ptr addrspace(10) null
134}
135
136attributes #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,-sha,-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" }
137attributes #1 = { inaccessiblemem_or_argmemonly }
138attributes #2 = { allocsize(1) }
139
140!0 = !{!1, !1, i64 0}
141!1 = !{!"jtbaa_value", !2, i64 0}
142!2 = !{!"jtbaa_data", !3, i64 0}
143!3 = !{!"jtbaa", !4, i64 0}
144!4 = !{!"jtbaa"}
145!5 = !{!6, !6, i64 0}
146!6 = !{!"jtbaa_arrayptr", !7, i64 0}
147!7 = !{!"jtbaa_array", !3, i64 0}
148!8 = !{!9, !9, i64 0}
149!9 = !{!"jtbaa_immut", !1, i64 0}
150!10 = !{!11, !11, i64 0}
151!11 = !{!"jtbaa_arraybuf", !2, i64 0}
152