1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s  -loop-vectorize -S | FileCheck %s
3
4; This is a bugpoint reduction of a test from PR43582:
5; https://bugs.llvm.org/show_bug.cgi?id=43582
6
7; ...but it's over-simplifying the underlying question:
8; TODO: Should this be vectorized rather than allowing the backend to load combine?
9;       The original code is a bswap pattern.
10
11target datalayout = "e-m:w-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
12target triple = "x86_64-w64-windows-gnu"
13
14define void @cff_index_load_offsets(i1 %cond, i8 %x, i8* %p) #0 {
15; CHECK-LABEL: @cff_index_load_offsets(
16; CHECK-NEXT:  entry:
17; CHECK-NEXT:    br i1 [[COND:%.*]], label [[IF_THEN:%.*]], label [[EXIT:%.*]]
18; CHECK:       if.then:
19; CHECK-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 undef, i64 4)
20; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[UMAX]], -1
21; CHECK-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 2
22; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
23; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 8
24; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
25; CHECK:       vector.ph:
26; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 8
27; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
28; CHECK-NEXT:    [[TMP3:%.*]] = mul i64 [[N_VEC]], 4
29; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, i8* null, i64 [[TMP3]]
30; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i8> poison, i8 [[X:%.*]], i32 0
31; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT]], <4 x i8> poison, <4 x i32> zeroinitializer
32; CHECK-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i8> poison, i8 [[X]], i32 0
33; CHECK-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT2]], <4 x i8> 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:    [[TMP8:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT]] to <4 x i32>
38; CHECK-NEXT:    [[TMP9:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT3]] to <4 x i32>
39; CHECK-NEXT:    [[TMP10:%.*]] = shl nuw <4 x i32> [[TMP8]], <i32 24, i32 24, i32 24, i32 24>
40; CHECK-NEXT:    [[TMP11:%.*]] = shl nuw <4 x i32> [[TMP9]], <i32 24, i32 24, i32 24, i32 24>
41; CHECK-NEXT:    [[TMP12:%.*]] = load i8, i8* [[P:%.*]], align 1, !tbaa [[TBAA1:![0-9]+]]
42; CHECK-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <4 x i8> poison, i8 [[TMP12]], i32 0
43; CHECK-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT4]], <4 x i8> poison, <4 x i32> zeroinitializer
44; CHECK-NEXT:    [[TMP13:%.*]] = load i8, i8* [[P]], align 1, !tbaa [[TBAA1]]
45; CHECK-NEXT:    [[BROADCAST_SPLATINSERT6:%.*]] = insertelement <4 x i8> poison, i8 [[TMP13]], i32 0
46; CHECK-NEXT:    [[BROADCAST_SPLAT7:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT6]], <4 x i8> poison, <4 x i32> zeroinitializer
47; CHECK-NEXT:    [[TMP14:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT5]] to <4 x i32>
48; CHECK-NEXT:    [[TMP15:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT7]] to <4 x i32>
49; CHECK-NEXT:    [[TMP16:%.*]] = shl nuw nsw <4 x i32> [[TMP14]], <i32 16, i32 16, i32 16, i32 16>
50; CHECK-NEXT:    [[TMP17:%.*]] = shl nuw nsw <4 x i32> [[TMP15]], <i32 16, i32 16, i32 16, i32 16>
51; CHECK-NEXT:    [[TMP18:%.*]] = or <4 x i32> [[TMP16]], [[TMP10]]
52; CHECK-NEXT:    [[TMP19:%.*]] = or <4 x i32> [[TMP17]], [[TMP11]]
53; CHECK-NEXT:    [[TMP20:%.*]] = load i8, i8* undef, align 1, !tbaa [[TBAA1]]
54; CHECK-NEXT:    [[BROADCAST_SPLATINSERT8:%.*]] = insertelement <4 x i8> poison, i8 [[TMP20]], i32 0
55; CHECK-NEXT:    [[BROADCAST_SPLAT9:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT8]], <4 x i8> poison, <4 x i32> zeroinitializer
56; CHECK-NEXT:    [[TMP21:%.*]] = load i8, i8* undef, align 1, !tbaa [[TBAA1]]
57; CHECK-NEXT:    [[BROADCAST_SPLATINSERT10:%.*]] = insertelement <4 x i8> poison, i8 [[TMP21]], i32 0
58; CHECK-NEXT:    [[BROADCAST_SPLAT11:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT10]], <4 x i8> poison, <4 x i32> zeroinitializer
59; CHECK-NEXT:    [[TMP22:%.*]] = or <4 x i32> [[TMP18]], zeroinitializer
60; CHECK-NEXT:    [[TMP23:%.*]] = or <4 x i32> [[TMP19]], zeroinitializer
61; CHECK-NEXT:    [[TMP24:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT9]] to <4 x i32>
62; CHECK-NEXT:    [[TMP25:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT11]] to <4 x i32>
63; CHECK-NEXT:    [[TMP26:%.*]] = or <4 x i32> [[TMP22]], [[TMP24]]
64; CHECK-NEXT:    [[TMP27:%.*]] = or <4 x i32> [[TMP23]], [[TMP25]]
65; CHECK-NEXT:    [[TMP28:%.*]] = extractelement <4 x i32> [[TMP26]], i32 0
66; CHECK-NEXT:    store i32 [[TMP28]], i32* undef, align 4, !tbaa [[TBAA4:![0-9]+]]
67; CHECK-NEXT:    [[TMP29:%.*]] = extractelement <4 x i32> [[TMP26]], i32 1
68; CHECK-NEXT:    store i32 [[TMP29]], i32* undef, align 4, !tbaa [[TBAA4]]
69; CHECK-NEXT:    [[TMP30:%.*]] = extractelement <4 x i32> [[TMP26]], i32 2
70; CHECK-NEXT:    store i32 [[TMP30]], i32* undef, align 4, !tbaa [[TBAA4]]
71; CHECK-NEXT:    [[TMP31:%.*]] = extractelement <4 x i32> [[TMP26]], i32 3
72; CHECK-NEXT:    store i32 [[TMP31]], i32* undef, align 4, !tbaa [[TBAA4]]
73; CHECK-NEXT:    [[TMP32:%.*]] = extractelement <4 x i32> [[TMP27]], i32 0
74; CHECK-NEXT:    store i32 [[TMP32]], i32* undef, align 4, !tbaa [[TBAA4]]
75; CHECK-NEXT:    [[TMP33:%.*]] = extractelement <4 x i32> [[TMP27]], i32 1
76; CHECK-NEXT:    store i32 [[TMP33]], i32* undef, align 4, !tbaa [[TBAA4]]
77; CHECK-NEXT:    [[TMP34:%.*]] = extractelement <4 x i32> [[TMP27]], i32 2
78; CHECK-NEXT:    store i32 [[TMP34]], i32* undef, align 4, !tbaa [[TBAA4]]
79; CHECK-NEXT:    [[TMP35:%.*]] = extractelement <4 x i32> [[TMP27]], i32 3
80; CHECK-NEXT:    store i32 [[TMP35]], i32* undef, align 4, !tbaa [[TBAA4]]
81; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
82; CHECK-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
83; CHECK-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
84; CHECK:       middle.block:
85; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
86; CHECK-NEXT:    br i1 [[CMP_N]], label [[SW_EPILOG:%.*]], label [[SCALAR_PH]]
87; CHECK:       scalar.ph:
88; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ null, [[IF_THEN]] ]
89; CHECK-NEXT:    br label [[FOR_BODY68:%.*]]
90; CHECK:       for.body68:
91; CHECK-NEXT:    [[P_359:%.*]] = phi i8* [ [[ADD_PTR86:%.*]], [[FOR_BODY68]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
92; CHECK-NEXT:    [[CONV70:%.*]] = zext i8 [[X]] to i32
93; CHECK-NEXT:    [[SHL71:%.*]] = shl nuw i32 [[CONV70]], 24
94; CHECK-NEXT:    [[TMP37:%.*]] = load i8, i8* [[P]], align 1, !tbaa [[TBAA1]]
95; CHECK-NEXT:    [[CONV73:%.*]] = zext i8 [[TMP37]] to i32
96; CHECK-NEXT:    [[SHL74:%.*]] = shl nuw nsw i32 [[CONV73]], 16
97; CHECK-NEXT:    [[OR75:%.*]] = or i32 [[SHL74]], [[SHL71]]
98; CHECK-NEXT:    [[TMP38:%.*]] = load i8, i8* undef, align 1, !tbaa [[TBAA1]]
99; CHECK-NEXT:    [[SHL78:%.*]] = shl nuw nsw i32 undef, 8
100; CHECK-NEXT:    [[OR79:%.*]] = or i32 [[OR75]], [[SHL78]]
101; CHECK-NEXT:    [[CONV81:%.*]] = zext i8 [[TMP38]] to i32
102; CHECK-NEXT:    [[OR83:%.*]] = or i32 [[OR79]], [[CONV81]]
103; CHECK-NEXT:    store i32 [[OR83]], i32* undef, align 4, !tbaa [[TBAA4]]
104; CHECK-NEXT:    [[ADD_PTR86]] = getelementptr inbounds i8, i8* [[P_359]], i64 4
105; CHECK-NEXT:    [[CMP66:%.*]] = icmp ult i8* [[ADD_PTR86]], undef
106; CHECK-NEXT:    br i1 [[CMP66]], label [[FOR_BODY68]], label [[SW_EPILOG]], !llvm.loop [[LOOP8:![0-9]+]]
107; CHECK:       sw.epilog:
108; CHECK-NEXT:    unreachable
109; CHECK:       Exit:
110; CHECK-NEXT:    ret void
111;
112entry:
113  br i1 %cond, label %if.then, label %Exit
114
115if.then:                                          ; preds = %entry
116  br label %for.body68
117
118for.body68:                                       ; preds = %for.body68, %if.then
119  %p.359 = phi i8* [ %add.ptr86, %for.body68 ], [ null, %if.then ]
120  %conv70 = zext i8 %x to i32
121  %shl71 = shl nuw i32 %conv70, 24
122  %0 = load i8, i8* %p, align 1, !tbaa !1
123  %conv73 = zext i8 %0 to i32
124  %shl74 = shl nuw nsw i32 %conv73, 16
125  %or75 = or i32 %shl74, %shl71
126  %1 = load i8, i8* undef, align 1, !tbaa !1
127  %shl78 = shl nuw nsw i32 undef, 8
128  %or79 = or i32 %or75, %shl78
129  %conv81 = zext i8 %1 to i32
130  %or83 = or i32 %or79, %conv81
131  store i32 %or83, i32* undef, align 4, !tbaa !4
132  %add.ptr86 = getelementptr inbounds i8, i8* %p.359, i64 4
133  %cmp66 = icmp ult i8* %add.ptr86, undef
134  br i1 %cmp66, label %for.body68, label %sw.epilog
135
136sw.epilog:                                        ; preds = %for.body68
137  unreachable
138
139Exit:                                             ; preds = %entry
140  ret void
141}
142
143attributes #0 = { "use-soft-float"="false" }
144
145!llvm.ident = !{!0}
146
147!0 = !{!"clang version 10.0.0 (https://github.com/llvm/llvm-project.git 0fedc26a0dc0066f3968b9fea6a4e1f746c8d5a4)"}
148!1 = !{!2, !2, i64 0}
149!2 = !{!"omnipotent char", !3, i64 0}
150!3 = !{!"Simple C/C++ TBAA"}
151!4 = !{!5, !5, i64 0}
152!5 = !{!"long", !2, i64 0}
153