1 ; RUN: opt -loop-vectorize -force-vector-width=4 -S %s | FileCheck %s
3 @dst = external global [32 x i16], align 1
5 define void @blend_uniform_iv_trunc(i1 %c) {
6 ; CHECK-LABEL: @blend_uniform_iv_trunc(
8 ; CHECK-NEXT: [[MASK0:%.*]] = insertelement <4 x i1> poison, i1 %c, i32 0
9 ; CHECK-NEXT: [[MASK1:%.*]] = shufflevector <4 x i1> [[MASK0]], <4 x i1> poison, <4 x i32> zeroinitializer
12 ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
13 ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
14 ; CHECK-NEXT: [[TMP1:%.*]] = trunc i64 [[INDEX]] to i16
15 ; CHECK-NEXT: [[TMP2:%.*]] = add i16 [[TMP1]], 0
16 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i16> poison, i16 [[TMP2]], i32 0
17 ; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT1]], <4 x i16> poison, <4 x i32> zeroinitializer
18 ; CHECK-NEXT: [[TMP3:%.*]] = xor <4 x i1> [[MASK1]], <i1 true, i1 true, i1 true, i1 true>
19 ; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[MASK1]], <4 x i16> [[BROADCAST_SPLAT2]], <4 x i16> undef
20 ; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x i16> [[PREDPHI]], i32 0
21 ; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i16 [[TMP4]]
22 ; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP5]], i32 0
23 ; CHECK-NEXT: [[TMP7:%.*]] = bitcast i16* [[TMP6]] to <4 x i16>*
24 ; CHECK-NEXT: store <4 x i16> zeroinitializer, <4 x i16>* [[TMP7]], align 2
25 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
26 ; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
27 ; CHECK-NEXT: br i1 [[TMP8]], label %middle.block, label %vector.body
32 loop.header: ; preds = %loop.latch, %entry
33 %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
34 %iv.trunc.2 = trunc i64 %iv to i16
35 br i1 %c, label %loop.next, label %loop.latch
37 loop.next: ; preds = %loop.header
40 loop.latch: ; preds = %loop.next, %loop.header
41 %blend = phi i16 [ undef, %loop.header ], [ %iv.trunc.2, %loop.next ]
42 %dst.ptr = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i16 %blend
43 store i16 0, i16* %dst.ptr
44 %iv.next = add nuw nsw i64 %iv, 1
45 %cmp439 = icmp ult i64 %iv, 31
46 br i1 %cmp439, label %loop.header, label %exit
48 exit: ; preds = %loop.latch
52 define void @blend_uniform_iv(i1 %c) {
53 ; CHECK-LABEL: @blend_uniform_iv(
55 ; CHECK-NEXT: [[MASK0:%.*]] = insertelement <4 x i1> poison, i1 %c, i32 0
56 ; CHECK-NEXT: [[MASK1:%.*]] = shufflevector <4 x i1> [[MASK0]], <4 x i1> poison, <4 x i32> zeroinitializer
59 ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
60 ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
61 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[TMP0]], i32 0
62 ; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer
63 ; CHECK-NEXT: [[TMP1:%.*]] = xor <4 x i1> [[MASK1]], <i1 true, i1 true, i1 true, i1 true>
64 ; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[MASK1]], <4 x i64> [[BROADCAST_SPLAT2]], <4 x i64> undef
65 ; CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x i64> [[PREDPHI]], i32 0
66 ; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i64 [[TMP2]]
67 ; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 0
68 ; CHECK-NEXT: [[TMP5:%.*]] = bitcast i16* [[TMP4]] to <4 x i16>*
69 ; CHECK-NEXT: store <4 x i16> zeroinitializer, <4 x i16>* [[TMP5]], align 2
70 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
71 ; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
72 ; CHECK-NEXT: br i1 [[TMP6]], label %middle.block, label %vector.body
77 loop.header: ; preds = %loop.latch, %entry
78 %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
79 br i1 %c, label %loop.next, label %loop.latch
81 loop.next: ; preds = %loop.header
84 loop.latch: ; preds = %loop.next, %loop.header
85 %blend = phi i64 [ undef, %loop.header ], [ %iv, %loop.next ]
86 %dst.ptr = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i64 %blend
87 store i16 0, i16* %dst.ptr
88 %iv.next = add nuw nsw i64 %iv, 1
89 %cmp439 = icmp ult i64 %iv, 31
90 br i1 %cmp439, label %loop.header, label %exit
92 exit: ; preds = %loop.latch
96 define void @blend_chain_iv(i1 %c) {
97 ; CHECK-LABEL: @blend_chain_iv(
99 ; CHECK-NEXT: [[MASK0:%.*]] = insertelement <4 x i1> poison, i1 %c, i32 0
100 ; CHECK-NEXT: [[MASK1:%.*]] = shufflevector <4 x i1> [[MASK0]], <4 x i1> poison, <4 x i32> zeroinitializer
102 ; CHECK: vector.body:
103 ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
104 ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
105 ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
106 ; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1
107 ; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 2
108 ; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 3
109 ; CHECK-NEXT: [[TMP4:%.*]] = xor <4 x i1> [[MASK1]], <i1 true, i1 true, i1 true, i1 true>
110 ; CHECK-NEXT: [[TMP5:%.*]] = select <4 x i1> [[MASK1]], <4 x i1> [[TMP4]], <4 x i1> zeroinitializer
111 ; CHECK-NEXT: [[TMP6:%.*]] = select <4 x i1> [[MASK1]], <4 x i1> [[MASK1]], <4 x i1> zeroinitializer
112 ; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP6]], <4 x i64> [[VEC_IND]], <4 x i64> undef
113 ; CHECK-NEXT: [[TMP7:%.*]] = xor <4 x i1> [[MASK1]], <i1 true, i1 true, i1 true, i1 true>
114 ; CHECK-NEXT: [[TMP8:%.*]] = or <4 x i1> [[TMP6]], [[TMP5]]
115 ; CHECK-NEXT: [[PREDPHI1:%.*]] = select <4 x i1> [[TMP8]], <4 x i64> [[PREDPHI]], <4 x i64> undef
116 ; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i64> [[PREDPHI1]], i32 0
117 ; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i64 [[TMP9]]
118 ; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x i64> [[PREDPHI1]], i32 1
119 ; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i64 [[TMP11]]
120 ; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x i64> [[PREDPHI1]], i32 2
121 ; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i64 [[TMP13]]
122 ; CHECK-NEXT: [[TMP15:%.*]] = extractelement <4 x i64> [[PREDPHI1]], i32 3
123 ; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i64 [[TMP15]]
124 ; CHECK-NEXT: store i16 0, i16* [[TMP10]], align 2
125 ; CHECK-NEXT: store i16 0, i16* [[TMP12]], align 2
126 ; CHECK-NEXT: store i16 0, i16* [[TMP14]], align 2
127 ; CHECK-NEXT: store i16 0, i16* [[TMP16]], align 2
128 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
129 ; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], <i64 4, i64 4, i64 4, i64 4>
130 ; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
131 ; CHECK-NEXT: br i1 [[TMP17]], label %middle.block, label %vector.body
134 br label %loop.header
136 loop.header: ; preds = %loop.latch, %entry
137 %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
138 br i1 %c, label %loop.next, label %loop.latch
140 loop.next: ; preds = %loop.header
141 br i1 %c, label %loop.next.2, label %loop.next.3
144 br label %loop.next.3
147 %blend.1 = phi i64 [ undef, %loop.next ], [ %iv, %loop.next.2 ]
150 loop.latch: ; preds = %loop.next, %loop.header
151 %blend = phi i64 [ undef, %loop.header ], [ %blend.1, %loop.next.3 ]
152 %dst.ptr = getelementptr inbounds [32 x i16], [32 x i16]* @dst, i16 0, i64 %blend
153 store i16 0, i16* %dst.ptr
154 %iv.next = add nuw nsw i64 %iv, 1
155 %cmp439 = icmp ult i64 %iv, 31
156 br i1 %cmp439, label %loop.header, label %exit
158 exit: ; preds = %loop.latch