@@ -17,42 +17,16 @@ define void @loop_invariant_store(ptr %p, i64 %a, i8 %b) {
17
17
; CHECK-NEXT: [[TMP3:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT]] to <4 x i32>
18
18
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
19
19
; CHECK: [[VECTOR_BODY]]:
20
- ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE8:.* ]] ]
21
- ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE8 ]] ]
20
+ ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY ]] ]
21
+ ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY ]] ]
22
22
; CHECK-NEXT: [[TMP4:%.*]] = icmp ule <4 x i32> [[VEC_IND]], splat (i32 8)
23
23
; CHECK-NEXT: [[TMP5:%.*]] = icmp sge <4 x i32> [[VEC_IND]], splat (i32 2)
24
24
; CHECK-NEXT: [[TMP6:%.*]] = select <4 x i1> [[TMP4]], <4 x i1> [[TMP5]], <4 x i1> zeroinitializer
25
25
; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP6]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]]
26
26
; CHECK-NEXT: [[TMP7:%.*]] = shl <4 x i32> [[PREDPHI]], splat (i32 8)
27
27
; CHECK-NEXT: [[TMP8:%.*]] = trunc <4 x i32> [[TMP7]] to <4 x i8>
28
- ; CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i1> [[TMP4]], i32 0
29
- ; CHECK-NEXT: br i1 [[TMP16]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
30
- ; CHECK: [[PRED_STORE_IF]]:
31
- ; CHECK-NEXT: [[TMP17:%.*]] = extractelement <4 x i8> [[TMP8]], i32 0
32
- ; CHECK-NEXT: store i8 [[TMP17]], ptr [[P]], align 1
33
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE]]
34
- ; CHECK: [[PRED_STORE_CONTINUE]]:
35
- ; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x i1> [[TMP4]], i32 1
36
- ; CHECK-NEXT: br i1 [[TMP11]], label %[[PRED_STORE_IF3:.*]], label %[[PRED_STORE_CONTINUE4:.*]]
37
- ; CHECK: [[PRED_STORE_IF3]]:
38
- ; CHECK-NEXT: [[TMP12:%.*]] = extractelement <4 x i8> [[TMP8]], i32 1
39
- ; CHECK-NEXT: store i8 [[TMP12]], ptr [[P]], align 1
40
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE4]]
41
- ; CHECK: [[PRED_STORE_CONTINUE4]]:
42
- ; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x i1> [[TMP4]], i32 2
43
- ; CHECK-NEXT: br i1 [[TMP13]], label %[[PRED_STORE_IF5:.*]], label %[[PRED_STORE_CONTINUE6:.*]]
44
- ; CHECK: [[PRED_STORE_IF5]]:
45
- ; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i8> [[TMP8]], i32 2
46
- ; CHECK-NEXT: store i8 [[TMP14]], ptr [[P]], align 1
47
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE6]]
48
- ; CHECK: [[PRED_STORE_CONTINUE6]]:
49
- ; CHECK-NEXT: [[TMP15:%.*]] = extractelement <4 x i1> [[TMP4]], i32 3
50
- ; CHECK-NEXT: br i1 [[TMP15]], label %[[PRED_STORE_IF7:.*]], label %[[PRED_STORE_CONTINUE8]]
51
- ; CHECK: [[PRED_STORE_IF7]]:
52
28
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i8> [[TMP8]], i32 3
53
29
; CHECK-NEXT: store i8 [[TMP9]], ptr [[P]], align 1
54
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE8]]
55
- ; CHECK: [[PRED_STORE_CONTINUE8]]:
56
30
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
57
31
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
58
32
; CHECK-NEXT: [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], 12
@@ -127,43 +101,17 @@ define void @loop_invariant_udiv(ptr %p, i64 %a, i8 %b) {
127
101
; CHECK-NEXT: [[TMP3:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT]] to <4 x i32>
128
102
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
129
103
; CHECK: [[VECTOR_BODY]]:
130
- ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE8:.* ]] ]
131
- ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE8 ]] ]
104
+ ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY ]] ]
105
+ ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY ]] ]
132
106
; CHECK-NEXT: [[TMP4:%.*]] = icmp ule <4 x i32> [[VEC_IND]], splat (i32 8)
133
107
; CHECK-NEXT: [[TMP5:%.*]] = icmp sge <4 x i32> [[VEC_IND]], splat (i32 2)
134
108
; CHECK-NEXT: [[TMP6:%.*]] = select <4 x i1> [[TMP4]], <4 x i1> [[TMP5]], <4 x i1> zeroinitializer
135
109
; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP6]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]]
136
110
; CHECK-NEXT: [[TMP7:%.*]] = shl <4 x i32> [[PREDPHI]], splat (i32 8)
137
111
; CHECK-NEXT: [[TMP8:%.*]] = trunc <4 x i32> [[TMP7]] to <4 x i8>
138
112
; CHECK-NEXT: [[TMP9:%.*]] = udiv <4 x i8> [[TMP8]], [[BROADCAST_SPLAT]]
139
- ; CHECK-NEXT: [[TMP17:%.*]] = extractelement <4 x i1> [[TMP4]], i32 0
140
- ; CHECK-NEXT: br i1 [[TMP17]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
141
- ; CHECK: [[PRED_STORE_IF]]:
142
- ; CHECK-NEXT: [[TMP18:%.*]] = extractelement <4 x i8> [[TMP9]], i32 0
143
- ; CHECK-NEXT: store i8 [[TMP18]], ptr [[P]], align 1
144
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE]]
145
- ; CHECK: [[PRED_STORE_CONTINUE]]:
146
- ; CHECK-NEXT: [[TMP12:%.*]] = extractelement <4 x i1> [[TMP4]], i32 1
147
- ; CHECK-NEXT: br i1 [[TMP12]], label %[[PRED_STORE_IF3:.*]], label %[[PRED_STORE_CONTINUE4:.*]]
148
- ; CHECK: [[PRED_STORE_IF3]]:
149
- ; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x i8> [[TMP9]], i32 1
150
- ; CHECK-NEXT: store i8 [[TMP13]], ptr [[P]], align 1
151
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE4]]
152
- ; CHECK: [[PRED_STORE_CONTINUE4]]:
153
- ; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i1> [[TMP4]], i32 2
154
- ; CHECK-NEXT: br i1 [[TMP14]], label %[[PRED_STORE_IF5:.*]], label %[[PRED_STORE_CONTINUE6:.*]]
155
- ; CHECK: [[PRED_STORE_IF5]]:
156
- ; CHECK-NEXT: [[TMP15:%.*]] = extractelement <4 x i8> [[TMP9]], i32 2
157
- ; CHECK-NEXT: store i8 [[TMP15]], ptr [[P]], align 1
158
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE6]]
159
- ; CHECK: [[PRED_STORE_CONTINUE6]]:
160
- ; CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i1> [[TMP4]], i32 3
161
- ; CHECK-NEXT: br i1 [[TMP16]], label %[[PRED_STORE_IF7:.*]], label %[[PRED_STORE_CONTINUE8]]
162
- ; CHECK: [[PRED_STORE_IF7]]:
163
113
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <4 x i8> [[TMP9]], i32 3
164
114
; CHECK-NEXT: store i8 [[TMP10]], ptr [[P]], align 1
165
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE8]]
166
- ; CHECK: [[PRED_STORE_CONTINUE8]]:
167
115
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
168
116
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
169
117
; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i32 [[INDEX_NEXT]], 12
@@ -240,43 +188,17 @@ define void @loop_invariant_float_store(ptr noalias %p, ptr noalias %q, i64 %a,
240
188
; CHECK-NEXT: [[TMP3:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT]] to <4 x i32>
241
189
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
242
190
; CHECK: [[VECTOR_BODY]]:
243
- ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE8:.* ]] ]
244
- ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE8 ]] ]
191
+ ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY ]] ]
192
+ ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY ]] ]
245
193
; CHECK-NEXT: [[TMP4:%.*]] = icmp ule <4 x i32> [[VEC_IND]], splat (i32 8)
246
194
; CHECK-NEXT: store float 2.000000e+00, ptr [[Q]], align 4
247
195
; CHECK-NEXT: [[TMP5:%.*]] = icmp sge <4 x i32> [[VEC_IND]], splat (i32 2)
248
196
; CHECK-NEXT: [[TMP6:%.*]] = select <4 x i1> [[TMP4]], <4 x i1> [[TMP5]], <4 x i1> zeroinitializer
249
197
; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP6]], <4 x i32> [[TMP2]], <4 x i32> [[TMP3]]
250
198
; CHECK-NEXT: [[TMP7:%.*]] = shl <4 x i32> [[PREDPHI]], splat (i32 8)
251
199
; CHECK-NEXT: [[TMP8:%.*]] = trunc <4 x i32> [[TMP7]] to <4 x i8>
252
- ; CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i1> [[TMP4]], i32 0
253
- ; CHECK-NEXT: br i1 [[TMP16]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
254
- ; CHECK: [[PRED_STORE_IF]]:
255
- ; CHECK-NEXT: [[TMP17:%.*]] = extractelement <4 x i8> [[TMP8]], i32 0
256
- ; CHECK-NEXT: store i8 [[TMP17]], ptr [[P]], align 1
257
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE]]
258
- ; CHECK: [[PRED_STORE_CONTINUE]]:
259
- ; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x i1> [[TMP4]], i32 1
260
- ; CHECK-NEXT: br i1 [[TMP11]], label %[[PRED_STORE_IF3:.*]], label %[[PRED_STORE_CONTINUE4:.*]]
261
- ; CHECK: [[PRED_STORE_IF3]]:
262
- ; CHECK-NEXT: [[TMP12:%.*]] = extractelement <4 x i8> [[TMP8]], i32 1
263
- ; CHECK-NEXT: store i8 [[TMP12]], ptr [[P]], align 1
264
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE4]]
265
- ; CHECK: [[PRED_STORE_CONTINUE4]]:
266
- ; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x i1> [[TMP4]], i32 2
267
- ; CHECK-NEXT: br i1 [[TMP13]], label %[[PRED_STORE_IF5:.*]], label %[[PRED_STORE_CONTINUE6:.*]]
268
- ; CHECK: [[PRED_STORE_IF5]]:
269
- ; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i8> [[TMP8]], i32 2
270
- ; CHECK-NEXT: store i8 [[TMP14]], ptr [[P]], align 1
271
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE6]]
272
- ; CHECK: [[PRED_STORE_CONTINUE6]]:
273
- ; CHECK-NEXT: [[TMP15:%.*]] = extractelement <4 x i1> [[TMP4]], i32 3
274
- ; CHECK-NEXT: br i1 [[TMP15]], label %[[PRED_STORE_IF7:.*]], label %[[PRED_STORE_CONTINUE8]]
275
- ; CHECK: [[PRED_STORE_IF7]]:
276
200
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i8> [[TMP8]], i32 3
277
201
; CHECK-NEXT: store i8 [[TMP9]], ptr [[P]], align 1
278
- ; CHECK-NEXT: br label %[[PRED_STORE_CONTINUE8]]
279
- ; CHECK: [[PRED_STORE_CONTINUE8]]:
280
202
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
281
203
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
282
204
; CHECK-NEXT: [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], 12
0 commit comments