1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -instcombine -S | FileCheck %s
; This turns into a&1 != 0
define <2 x i1> @test1(<2 x i64> %a) {
; CHECK-LABEL: @test1(
; CHECK-NEXT: [[TMP1:%.*]] = and <2 x i64> %a, <i64 1, i64 1>
; CHECK-NEXT: [[T:%.*]] = icmp ne <2 x i64> [[TMP1]], zeroinitializer
; CHECK-NEXT: ret <2 x i1> [[T]]
;
%t = trunc <2 x i64> %a to <2 x i1>
ret <2 x i1> %t
}
; The ashr turns into an lshr.
define <2 x i64> @test2(<2 x i64> %a) {
; CHECK-LABEL: @test2(
; CHECK-NEXT: [[B:%.*]] = and <2 x i64> %a, <i64 65534, i64 65534>
; CHECK-NEXT: [[TMP1:%.*]] = lshr exact <2 x i64> [[B]], <i64 1, i64 1>
; CHECK-NEXT: ret <2 x i64> [[TMP1]]
;
%b = and <2 x i64> %a, <i64 65535, i64 65535>
%t = ashr <2 x i64> %b, <i64 1, i64 1>
ret <2 x i64> %t
}
define <2 x i64> @test3(<4 x float> %a, <4 x float> %b) {
; CHECK-LABEL: @test3(
; CHECK-NEXT: [[TMP1:%.*]] = fcmp ord <4 x float> %a, %b
; CHECK-NEXT: [[AND:%.*]] = sext <4 x i1> [[TMP1]] to <4 x i32>
; CHECK-NEXT: [[CONV:%.*]] = bitcast <4 x i32> [[AND]] to <2 x i64>
; CHECK-NEXT: ret <2 x i64> [[CONV]]
;
%cmp = fcmp ord <4 x float> %a, zeroinitializer
%sext = sext <4 x i1> %cmp to <4 x i32>
%cmp4 = fcmp ord <4 x float> %b, zeroinitializer
%sext5 = sext <4 x i1> %cmp4 to <4 x i32>
%and = and <4 x i32> %sext, %sext5
%conv = bitcast <4 x i32> %and to <2 x i64>
ret <2 x i64> %conv
}
define <2 x i64> @test4(<4 x float> %a, <4 x float> %b) {
; CHECK-LABEL: @test4(
; CHECK-NEXT: [[TMP1:%.*]] = fcmp uno <4 x float> %a, %b
; CHECK-NEXT: [[OR:%.*]] = sext <4 x i1> [[TMP1]] to <4 x i32>
; CHECK-NEXT: [[CONV:%.*]] = bitcast <4 x i32> [[OR]] to <2 x i64>
; CHECK-NEXT: ret <2 x i64> [[CONV]]
;
%cmp = fcmp uno <4 x float> %a, zeroinitializer
%sext = sext <4 x i1> %cmp to <4 x i32>
%cmp4 = fcmp uno <4 x float> %b, zeroinitializer
%sext5 = sext <4 x i1> %cmp4 to <4 x i32>
%or = or <4 x i32> %sext, %sext5
%conv = bitcast <4 x i32> %or to <2 x i64>
ret <2 x i64> %conv
}
; rdar://7434900
define <2 x i64> @test5(<4 x float> %a, <4 x float> %b) {
; CHECK-LABEL: @test5(
; CHECK-NEXT: [[CMP:%.*]] = fcmp ult <4 x float> %a, zeroinitializer
; CHECK-NEXT: [[CMP4:%.*]] = fcmp ult <4 x float> %b, zeroinitializer
; CHECK-NEXT: [[NARROW:%.*]] = and <4 x i1> [[CMP4]], [[CMP]]
; CHECK-NEXT: [[AND:%.*]] = sext <4 x i1> [[NARROW]] to <4 x i32>
; CHECK-NEXT: [[CONV:%.*]] = bitcast <4 x i32> [[AND]] to <2 x i64>
; CHECK-NEXT: ret <2 x i64> [[CONV]]
;
%cmp = fcmp ult <4 x float> %a, zeroinitializer
%sext = sext <4 x i1> %cmp to <4 x i32>
%cmp4 = fcmp ult <4 x float> %b, zeroinitializer
%sext5 = sext <4 x i1> %cmp4 to <4 x i32>
%and = and <4 x i32> %sext, %sext5
%conv = bitcast <4 x i32> %and to <2 x i64>
ret <2 x i64> %conv
}
define void @convert(<2 x i32>* %dst.addr, <2 x i64> %src) {
; CHECK-LABEL: @convert(
; CHECK-NEXT: [[VAL:%.*]] = trunc <2 x i64> %src to <2 x i32>
; CHECK-NEXT: [[ADD:%.*]] = add <2 x i32> [[VAL]], <i32 1, i32 1>
; CHECK-NEXT: store <2 x i32> [[ADD]], <2 x i32>* %dst.addr, align 8
; CHECK-NEXT: ret void
;
%val = trunc <2 x i64> %src to <2 x i32>
%add = add <2 x i32> %val, <i32 1, i32 1>
store <2 x i32> %add, <2 x i32>* %dst.addr
ret void
}
define <2 x i65> @foo(<2 x i64> %t) {
; CHECK-LABEL: @foo(
; CHECK-NEXT: [[TMP1:%.*]] = and <2 x i64> %t, <i64 4294967295, i64 4294967295>
; CHECK-NEXT: [[B:%.*]] = zext <2 x i64> [[TMP1]] to <2 x i65>
; CHECK-NEXT: ret <2 x i65> [[B]]
;
%a = trunc <2 x i64> %t to <2 x i32>
%b = zext <2 x i32> %a to <2 x i65>
ret <2 x i65> %b
}
define <2 x i64> @bar(<2 x i65> %t) {
; CHECK-LABEL: @bar(
; CHECK-NEXT: [[A:%.*]] = trunc <2 x i65> %t to <2 x i64>
; CHECK-NEXT: [[B:%.*]] = and <2 x i64> [[A]], <i64 4294967295, i64 4294967295>
; CHECK-NEXT: ret <2 x i64> [[B]]
;
%a = trunc <2 x i65> %t to <2 x i32>
%b = zext <2 x i32> %a to <2 x i64>
ret <2 x i64> %b
}
define <2 x i64> @bars(<2 x i65> %t) {
; CHECK-LABEL: @bars(
; CHECK-NEXT: [[A:%.*]] = trunc <2 x i65> %t to <2 x i64>
; CHECK-NEXT: [[SEXT:%.*]] = shl <2 x i64> [[A]], <i64 32, i64 32>
; CHECK-NEXT: [[B:%.*]] = ashr exact <2 x i64> [[SEXT]], <i64 32, i64 32>
; CHECK-NEXT: ret <2 x i64> [[B]]
;
%a = trunc <2 x i65> %t to <2 x i32>
%b = sext <2 x i32> %a to <2 x i64>
ret <2 x i64> %b
}
define <2 x i64> @quxs(<2 x i64> %t) {
; CHECK-LABEL: @quxs(
; CHECK-NEXT: [[SEXT:%.*]] = shl <2 x i64> %t, <i64 32, i64 32>
; CHECK-NEXT: [[B:%.*]] = ashr exact <2 x i64> [[SEXT]], <i64 32, i64 32>
; CHECK-NEXT: ret <2 x i64> [[B]]
;
%a = trunc <2 x i64> %t to <2 x i32>
%b = sext <2 x i32> %a to <2 x i64>
ret <2 x i64> %b
}
define <2 x i64> @quxt(<2 x i64> %t) {
; CHECK-LABEL: @quxt(
; CHECK-NEXT: [[A:%.*]] = shl <2 x i64> %t, <i64 32, i64 32>
; CHECK-NEXT: [[B:%.*]] = ashr exact <2 x i64> [[A]], <i64 32, i64 32>
; CHECK-NEXT: ret <2 x i64> [[B]]
;
%a = shl <2 x i64> %t, <i64 32, i64 32>
%b = ashr <2 x i64> %a, <i64 32, i64 32>
ret <2 x i64> %b
}
define <2 x double> @fa(<2 x double> %t) {
; CHECK-LABEL: @fa(
; CHECK-NEXT: [[A:%.*]] = fptrunc <2 x double> %t to <2 x float>
; CHECK-NEXT: [[B:%.*]] = fpext <2 x float> [[A]] to <2 x double>
; CHECK-NEXT: ret <2 x double> [[B]]
;
%a = fptrunc <2 x double> %t to <2 x float>
%b = fpext <2 x float> %a to <2 x double>
ret <2 x double> %b
}
define <2 x double> @fb(<2 x double> %t) {
; CHECK-LABEL: @fb(
; CHECK-NEXT: [[A:%.*]] = fptoui <2 x double> %t to <2 x i64>
; CHECK-NEXT: [[B:%.*]] = uitofp <2 x i64> [[A]] to <2 x double>
; CHECK-NEXT: ret <2 x double> [[B]]
;
%a = fptoui <2 x double> %t to <2 x i64>
%b = uitofp <2 x i64> %a to <2 x double>
ret <2 x double> %b
}
define <2 x double> @fc(<2 x double> %t) {
; CHECK-LABEL: @fc(
; CHECK-NEXT: [[A:%.*]] = fptosi <2 x double> %t to <2 x i64>
; CHECK-NEXT: [[B:%.*]] = sitofp <2 x i64> [[A]] to <2 x double>
; CHECK-NEXT: ret <2 x double> [[B]]
;
%a = fptosi <2 x double> %t to <2 x i64>
%b = sitofp <2 x i64> %a to <2 x double>
ret <2 x double> %b
}
; PR9228
define <4 x float> @f(i32 %a) {
; CHECK-LABEL: @f(
; CHECK-NEXT: ret <4 x float> undef
;
%dim = insertelement <4 x i32> undef, i32 %a, i32 0
%dim30 = insertelement <4 x i32> %dim, i32 %a, i32 1
%dim31 = insertelement <4 x i32> %dim30, i32 %a, i32 2
%dim32 = insertelement <4 x i32> %dim31, i32 %a, i32 3
%offset_ptr = getelementptr <4 x float>, <4 x float>* null, i32 1
%offset_int = ptrtoint <4 x float>* %offset_ptr to i64
%sizeof32 = trunc i64 %offset_int to i32
%smearinsert33 = insertelement <4 x i32> undef, i32 %sizeof32, i32 0
%smearinsert34 = insertelement <4 x i32> %smearinsert33, i32 %sizeof32, i32 1
%smearinsert35 = insertelement <4 x i32> %smearinsert34, i32 %sizeof32, i32 2
%smearinsert36 = insertelement <4 x i32> %smearinsert35, i32 %sizeof32, i32 3
%delta_scale = mul <4 x i32> %dim32, %smearinsert36
%offset_delta = add <4 x i32> zeroinitializer, %delta_scale
%offset_varying_delta = add <4 x i32> %offset_delta, undef
ret <4 x float> undef
}
define <8 x i32> @pr24458(<8 x float> %n) {
; CHECK-LABEL: @pr24458(
; CHECK-NEXT: ret <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>
;
%notequal_b_load_.i = fcmp une <8 x float> %n, zeroinitializer
%equal_a_load72_.i = fcmp ueq <8 x float> %n, zeroinitializer
%notequal_b_load__to_boolvec.i = sext <8 x i1> %notequal_b_load_.i to <8 x i32>
%equal_a_load72__to_boolvec.i = sext <8 x i1> %equal_a_load72_.i to <8 x i32>
%wrong = or <8 x i32> %notequal_b_load__to_boolvec.i, %equal_a_load72__to_boolvec.i
ret <8 x i32> %wrong
}
; Hoist a trunc to a scalar if we're inserting into an undef vector.
; trunc (inselt undef, X, Index) --> inselt undef, (trunc X), Index
define <3 x i16> @trunc_inselt_undef(i32 %x) {
; CHECK-LABEL: @trunc_inselt_undef(
; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 %x to i16
; CHECK-NEXT: [[TRUNC:%.*]] = insertelement <3 x i16> undef, i16 [[TMP1]], i32 1
; CHECK-NEXT: ret <3 x i16> [[TRUNC]]
;
%vec = insertelement <3 x i32> undef, i32 %x, i32 1
%trunc = trunc <3 x i32> %vec to <3 x i16>
ret <3 x i16> %trunc
}
; Hoist a trunc to a scalar if we're inserting into an undef vector.
; trunc (inselt undef, X, Index) --> inselt undef, (trunc X), Index
define <2 x float> @fptrunc_inselt_undef(double %x, i32 %index) {
; CHECK-LABEL: @fptrunc_inselt_undef(
; CHECK-NEXT: [[TMP1:%.*]] = fptrunc double %x to float
; CHECK-NEXT: [[TRUNC:%.*]] = insertelement <2 x float> undef, float [[TMP1]], i32 %index
; CHECK-NEXT: ret <2 x float> [[TRUNC]]
;
%vec = insertelement <2 x double> <double undef, double undef>, double %x, i32 %index
%trunc = fptrunc <2 x double> %vec to <2 x float>
ret <2 x float> %trunc
}
; TODO: Strengthen the backend, so we can have this canonicalization.
; Insert a scalar int into a constant vector and truncate:
; trunc (inselt C, X, Index) --> inselt C, (trunc X), Index
define <3 x i16> @trunc_inselt1(i32 %x) {
; CHECK-LABEL: @trunc_inselt1(
; CHECK-NEXT: [[VEC:%.*]] = insertelement <3 x i32> <i32 3, i32 undef, i32 65536>, i32 %x, i32 1
; CHECK-NEXT: [[TRUNC:%.*]] = trunc <3 x i32> [[VEC]] to <3 x i16>
; CHECK-NEXT: ret <3 x i16> [[TRUNC]]
;
%vec = insertelement <3 x i32> <i32 3, i32 -2, i32 65536>, i32 %x, i32 1
%trunc = trunc <3 x i32> %vec to <3 x i16>
ret <3 x i16> %trunc
}
; TODO: Strengthen the backend, so we can have this canonicalization.
; Insert a scalar FP into a constant vector and FP truncate:
; fptrunc (inselt C, X, Index) --> inselt C, (fptrunc X), Index
define <2 x float> @fptrunc_inselt1(double %x, i32 %index) {
; CHECK-LABEL: @fptrunc_inselt1(
; CHECK-NEXT: [[VEC:%.*]] = insertelement <2 x double> <double undef, double 3.000000e+00>, double %x, i32 %index
; CHECK-NEXT: [[TRUNC:%.*]] = fptrunc <2 x double> [[VEC]] to <2 x float>
; CHECK-NEXT: ret <2 x float> [[TRUNC]]
;
%vec = insertelement <2 x double> <double undef, double 3.0>, double %x, i32 %index
%trunc = fptrunc <2 x double> %vec to <2 x float>
ret <2 x float> %trunc
}
; TODO: Strengthen the backend, so we can have this canonicalization.
; Insert a scalar int constant into a vector and truncate:
; trunc (inselt X, C, Index) --> inselt (trunc X), C', Index
define <8 x i16> @trunc_inselt2(<8 x i32> %x, i32 %index) {
; CHECK-LABEL: @trunc_inselt2(
; CHECK-NEXT: [[VEC:%.*]] = insertelement <8 x i32> %x, i32 1048576, i32 %index
; CHECK-NEXT: [[TRUNC:%.*]] = trunc <8 x i32> [[VEC]] to <8 x i16>
; CHECK-NEXT: ret <8 x i16> [[TRUNC]]
;
%vec = insertelement <8 x i32> %x, i32 1048576, i32 %index
%trunc = trunc <8 x i32> %vec to <8 x i16>
ret <8 x i16> %trunc
}
; TODO: Strengthen the backend, so we can have this canonicalization.
; Insert a scalar FP constant into a vector and FP truncate:
; fptrunc (inselt X, C, Index) --> inselt (fptrunc X), C', Index
define <3 x float> @fptrunc_inselt2(<3 x double> %x) {
; CHECK-LABEL: @fptrunc_inselt2(
; CHECK-NEXT: [[VEC:%.*]] = insertelement <3 x double> %x, double 4.000000e+00, i32 2
; CHECK-NEXT: [[TRUNC:%.*]] = fptrunc <3 x double> [[VEC]] to <3 x float>
; CHECK-NEXT: ret <3 x float> [[TRUNC]]
;
%vec = insertelement <3 x double> %x, double 4.0, i32 2
%trunc = fptrunc <3 x double> %vec to <3 x float>
ret <3 x float> %trunc
}
|
