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
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
|
// RUN: mlir-opt %s -convert-vector-to-llvm -split-input-file | FileCheck %s
// RUN: mlir-opt %s -convert-vector-to-llvm='reassociate-fp-reductions' -split-input-file | FileCheck %s --check-prefix=REASSOC
// CHECK-LABEL: @reduce_add_f32(
// CHECK-SAME: %[[A:.*]]: vector<16xf32>)
// CHECK: %[[C:.*]] = llvm.mlir.constant(0.000000e+00 : f32) : f32
// CHECK: %[[V:.*]] = "llvm.intr.vector.reduce.fadd"(%[[C]], %[[A]])
// CHECK-SAME: <{fastmathFlags = #llvm.fastmath<none>}> : (f32, vector<16xf32>) -> f32
// CHECK: return %[[V]] : f32
//
// REASSOC-LABEL: @reduce_add_f32(
// REASSOC-SAME: %[[A:.*]]: vector<16xf32>)
// REASSOC: %[[C:.*]] = llvm.mlir.constant(0.000000e+00 : f32) : f32
// REASSOC: %[[V:.*]] = "llvm.intr.vector.reduce.fadd"(%[[C]], %[[A]])
// REASSOC-SAME: <{fastmathFlags = #llvm.fastmath<reassoc>}> : (f32, vector<16xf32>) -> f32
// REASSOC: return %[[V]] : f32
//
func.func @reduce_add_f32(%arg0: vector<16xf32>) -> f32 {
%0 = vector.reduction <add>, %arg0 : vector<16xf32> into f32
return %0 : f32
}
// -----
// CHECK-LABEL: @reduce_add_f32_always_reassoc(
// CHECK-SAME: %[[A:.*]]: vector<16xf32>)
// CHECK: %[[C:.*]] = llvm.mlir.constant(0.000000e+00 : f32) : f32
// CHECK: %[[V:.*]] = "llvm.intr.vector.reduce.fadd"(%[[C]], %[[A]])
/// Note: the reassoc flag remains even though the pass sets reassociate-fp-reduction to false.
/// Ponder whether this flag really is a property of the pass / pattern..
// CHECK-SAME: <{fastmathFlags = #llvm.fastmath<reassoc>}> : (f32, vector<16xf32>) -> f32
// CHECK: return %[[V]] : f32
//
// REASSOC-LABEL: @reduce_add_f32_always_reassoc(
// REASSOC-SAME: %[[A:.*]]: vector<16xf32>)
// REASSOC: %[[C:.*]] = llvm.mlir.constant(0.000000e+00 : f32) : f32
// REASSOC: %[[V:.*]] = "llvm.intr.vector.reduce.fadd"(%[[C]], %[[A]])
// REASSOC-SAME: <{fastmathFlags = #llvm.fastmath<reassoc>}> : (f32, vector<16xf32>) -> f32
// REASSOC: return %[[V]] : f32
//
func.func @reduce_add_f32_always_reassoc(%arg0: vector<16xf32>) -> f32 {
%0 = vector.reduction <add>, %arg0 fastmath<reassoc> : vector<16xf32> into f32
return %0 : f32
}
// -----
// CHECK-LABEL: @reduce_mul_f32(
// CHECK-SAME: %[[A:.*]]: vector<16xf32>)
// CHECK: %[[C:.*]] = llvm.mlir.constant(1.000000e+00 : f32) : f32
// CHECK: %[[V:.*]] = "llvm.intr.vector.reduce.fmul"(%[[C]], %[[A]])
// CHECK-SAME: <{fastmathFlags = #llvm.fastmath<nnan, ninf>}> : (f32, vector<16xf32>) -> f32
// CHECK: return %[[V]] : f32
//
// REASSOC-LABEL: @reduce_mul_f32(
// REASSOC-SAME: %[[A:.*]]: vector<16xf32>)
// REASSOC: %[[C:.*]] = llvm.mlir.constant(1.000000e+00 : f32) : f32
// REASSOC: %[[V:.*]] = "llvm.intr.vector.reduce.fmul"(%[[C]], %[[A]])
// REASSOC-SAME: <{fastmathFlags = #llvm.fastmath<nnan, ninf, reassoc>}> : (f32, vector<16xf32>) -> f32
// REASSOC: return %[[V]] : f32
//
func.func @reduce_mul_f32(%arg0: vector<16xf32>) -> f32 {
%0 = vector.reduction <mul>, %arg0 fastmath<nnan, ninf> : vector<16xf32> into f32
return %0 : f32
}
// -----
func.func @masked_reduce_add_f32(%arg0: vector<16xf32>, %mask : vector<16xi1>) -> f32 {
%0 = vector.mask %mask { vector.reduction <add>, %arg0 : vector<16xf32> into f32 } : vector<16xi1> -> f32
return %0 : f32
}
// CHECK-LABEL: func.func @masked_reduce_add_f32(
// CHECK-SAME: %[[INPUT:.*]]: vector<16xf32>,
// CHECK-SAME: %[[MASK:.*]]: vector<16xi1>) -> f32 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0.000000e+00 : f32) : f32
// CHECK: %[[VL:.*]] = llvm.mlir.constant(16 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.fadd"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (f32, vector<16xf32>, vector<16xi1>, i32) -> f32
// -----
func.func @masked_reduce_add_f32_scalable(%arg0: vector<[16]xf32>, %mask : vector<[16]xi1>) -> f32 {
%0 = vector.mask %mask { vector.reduction <add>, %arg0 : vector<[16]xf32> into f32 } : vector<[16]xi1> -> f32
return %0 : f32
}
// CHECK-LABEL: func.func @masked_reduce_add_f32_scalable(
// CHECK-SAME: %[[INPUT:.*]]: vector<[16]xf32>,
// CHECK-SAME: %[[MASK:.*]]: vector<[16]xi1>) -> f32 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0.000000e+00 : f32) : f32
// CHECK: %[[VL_BASE:.*]] = llvm.mlir.constant(16 : i32) : i32
// CHECK: %[[VSCALE:.*]] = "llvm.intr.vscale"() : () -> i64
// CHECK: %[[CAST_IDX:.*]] = builtin.unrealized_conversion_cast %[[VSCALE]] : i64 to index
// CHECK: %[[CAST_I32:.*]] = arith.index_cast %[[CAST_IDX]] : index to i32
// CHECK: %[[VL_MUL:.*]] = arith.muli %[[VL_BASE]], %[[CAST_I32]] : i32
// CHECK: "llvm.intr.vp.reduce.fadd"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL_MUL]]) : (f32, vector<[16]xf32>, vector<[16]xi1>, i32) -> f32
// -----
func.func @masked_reduce_mul_f32(%arg0: vector<16xf32>, %mask : vector<16xi1>) -> f32 {
%0 = vector.mask %mask { vector.reduction <mul>, %arg0 : vector<16xf32> into f32 } : vector<16xi1> -> f32
return %0 : f32
}
// CHECK-LABEL: func.func @masked_reduce_mul_f32(
// CHECK-SAME: %[[INPUT:.*]]: vector<16xf32>,
// CHECK-SAME: %[[MASK:.*]]: vector<16xi1>) -> f32 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(1.000000e+00 : f32) : f32
// CHECK: %[[VL:.*]] = llvm.mlir.constant(16 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.fmul"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (f32, vector<16xf32>, vector<16xi1>, i32) -> f32
// -----
func.func @masked_reduce_minf_f32(%arg0: vector<16xf32>, %mask : vector<16xi1>) -> f32 {
%0 = vector.mask %mask { vector.reduction <minnumf>, %arg0 : vector<16xf32> into f32 } : vector<16xi1> -> f32
return %0 : f32
}
// CHECK-LABEL: func.func @masked_reduce_minf_f32(
// CHECK-SAME: %[[INPUT:.*]]: vector<16xf32>,
// CHECK-SAME: %[[MASK:.*]]: vector<16xi1>) -> f32 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0xFFC00000 : f32) : f32
// CHECK: %[[VL:.*]] = llvm.mlir.constant(16 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.fmin"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (f32, vector<16xf32>, vector<16xi1>, i32) -> f32
// -----
func.func @masked_reduce_minf_f32_scalable(%arg0: vector<[16]xf32>, %mask : vector<[16]xi1>) -> f32 {
%0 = vector.mask %mask { vector.reduction <minnumf>, %arg0 : vector<[16]xf32> into f32 } : vector<[16]xi1> -> f32
return %0 : f32
}
// CHECK-LABEL: func.func @masked_reduce_minf_f32_scalable(
// CHECK-SAME: %[[INPUT:.*]]: vector<[16]xf32>,
// CHECK-SAME: %[[MASK:.*]]: vector<[16]xi1>) -> f32 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0xFFC00000 : f32) : f32
// CHECK: %[[VL_BASE:.*]] = llvm.mlir.constant(16 : i32) : i32
// CHECK: %[[VSCALE:.*]] = "llvm.intr.vscale"() : () -> i64
// CHECK: %[[CAST_IDX:.*]] = builtin.unrealized_conversion_cast %[[VSCALE]] : i64 to index
// CHECK: %[[CAST_I32:.*]] = arith.index_cast %[[CAST_IDX]] : index to i32
// CHECK: %[[VL_MUL:.*]] = arith.muli %[[VL_BASE]], %[[CAST_I32]] : i32
// CHECK: "llvm.intr.vp.reduce.fmin"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL_MUL]]) : (f32, vector<[16]xf32>, vector<[16]xi1>, i32) -> f32
// -----
func.func @masked_reduce_maxf_f32(%arg0: vector<16xf32>, %mask : vector<16xi1>) -> f32 {
%0 = vector.mask %mask { vector.reduction <maxnumf>, %arg0 : vector<16xf32> into f32 } : vector<16xi1> -> f32
return %0 : f32
}
// CHECK-LABEL: func.func @masked_reduce_maxf_f32(
// CHECK-SAME: %[[INPUT:.*]]: vector<16xf32>,
// CHECK-SAME: %[[MASK:.*]]: vector<16xi1>) -> f32 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0x7FC00000 : f32) : f32
// CHECK: %[[VL:.*]] = llvm.mlir.constant(16 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.fmax"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (f32, vector<16xf32>, vector<16xi1>, i32) -> f32
// -----
func.func @masked_reduce_maximumf_f32(%arg0: vector<16xf32>, %mask : vector<16xi1>) -> f32 {
%0 = vector.mask %mask { vector.reduction <maximumf>, %arg0 : vector<16xf32> into f32 } : vector<16xi1> -> f32
return %0 : f32
}
// CHECK-LABEL: func.func @masked_reduce_maximumf_f32(
// CHECK-SAME: %[[INPUT:.*]]: vector<16xf32>,
// CHECK-SAME: %[[MASK:.*]]: vector<16xi1>) -> f32 {
// CHECK: %[[MASK_NEUTRAL:.*]] = llvm.mlir.constant(dense<-1.401300e-45> : vector<16xf32>) : vector<16xf32>
// CHECK: %[[MASKED:.*]] = llvm.select %[[MASK]], %[[INPUT]], %[[MASK_NEUTRAL]] : vector<16xi1>, vector<16xf32>
// CHECK: %[[RESULT:.*]] = llvm.intr.vector.reduce.fmaximum(%[[MASKED]]) : (vector<16xf32>) -> f32
// CHECK: return %[[RESULT]]
// -----
func.func @masked_reduce_minimumf_f32(%arg0: vector<16xf32>, %mask : vector<16xi1>) -> f32 {
%0 = vector.mask %mask { vector.reduction <minimumf>, %arg0 : vector<16xf32> into f32 } : vector<16xi1> -> f32
return %0 : f32
}
// CHECK-LABEL: func.func @masked_reduce_minimumf_f32(
// CHECK-SAME: %[[INPUT:.*]]: vector<16xf32>,
// CHECK-SAME: %[[MASK:.*]]: vector<16xi1>) -> f32 {
// CHECK: %[[MASK_NEUTRAL:.*]] = llvm.mlir.constant(dense<3.40282347E+38> : vector<16xf32>) : vector<16xf32>
// CHECK: %[[MASKED:.*]] = llvm.select %[[MASK]], %[[INPUT]], %[[MASK_NEUTRAL]] : vector<16xi1>, vector<16xf32>
// CHECK: %[[RESULT:.*]] = llvm.intr.vector.reduce.fminimum(%[[MASKED]]) : (vector<16xf32>) -> f32
// CHECK: return %[[RESULT]]
// -----
func.func @masked_reduce_add_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <add>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_add_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.add"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_add_i8_scalable(%arg0: vector<[32]xi8>, %mask : vector<[32]xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <add>, %arg0 : vector<[32]xi8> into i8 } : vector<[32]xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_add_i8_scalable(
// CHECK-SAME: %[[INPUT:.*]]: vector<[32]xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<[32]xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0 : i8) : i8
// CHECK: %[[VL_BASE:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: %[[VSCALE:.*]] = "llvm.intr.vscale"() : () -> i64
// CHECK: %[[CAST_IDX:.*]] = builtin.unrealized_conversion_cast %[[VSCALE]] : i64 to index
// CHECK: %[[CAST_I32:.*]] = arith.index_cast %[[CAST_IDX]] : index to i32
// CHECK: %[[VL_MUL:.*]] = arith.muli %[[VL_BASE]], %[[CAST_I32]] : i32
// CHECK: "llvm.intr.vp.reduce.add"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL_MUL]]) : (i8, vector<[32]xi8>, vector<[32]xi1>, i32) -> i8
// -----
func.func @masked_reduce_mul_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <mul>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_mul_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(1 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: %[[VAL_4:.*]] = "llvm.intr.vp.reduce.mul"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_minui_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <minui>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_minui_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(-1 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.umin"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_minui_i8_scalable(%arg0: vector<[32]xi8>, %mask : vector<[32]xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <minui>, %arg0 : vector<[32]xi8> into i8 } : vector<[32]xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_minui_i8_scalable(
// CHECK-SAME: %[[INPUT:.*]]: vector<[32]xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<[32]xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(-1 : i8) : i8
// CHECK: %[[VL_BASE:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: %[[VSCALE:.*]] = "llvm.intr.vscale"() : () -> i64
// CHECK: %[[CAST_IDX:.*]] = builtin.unrealized_conversion_cast %[[VSCALE]] : i64 to index
// CHECK: %[[CAST_I32:.*]] = arith.index_cast %[[CAST_IDX]] : index to i32
// CHECK: %[[VL_MUL:.*]] = arith.muli %[[VL_BASE]], %[[CAST_I32]] : i32
// CHECK: "llvm.intr.vp.reduce.umin"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL_MUL]]) : (i8, vector<[32]xi8>, vector<[32]xi1>, i32) -> i8
// -----
func.func @masked_reduce_maxui_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <maxui>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_maxui_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.umax"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_minsi_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <minsi>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_minsi_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(127 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.smin"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_maxsi_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <maxsi>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_maxsi_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(-128 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.smax"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_maxsi_i8_scalable(%arg0: vector<[32]xi8>, %mask : vector<[32]xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <maxsi>, %arg0 : vector<[32]xi8> into i8 } : vector<[32]xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_maxsi_i8_scalable(
// CHECK-SAME: %[[INPUT:.*]]: vector<[32]xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<[32]xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(-128 : i8) : i8
// CHECK: %[[VL_BASE:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: %[[VSCALE:.*]] = "llvm.intr.vscale"() : () -> i64
// CHECK: %[[CAST_IDX:.*]] = builtin.unrealized_conversion_cast %[[VSCALE]] : i64 to index
// CHECK: %[[CAST_I32:.*]] = arith.index_cast %[[CAST_IDX]] : index to i32
// CHECK: %[[VL_MUL:.*]] = arith.muli %[[VL_BASE]], %[[CAST_I32]] : i32
// CHECK: "llvm.intr.vp.reduce.smax"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL_MUL]]) : (i8, vector<[32]xi8>, vector<[32]xi1>, i32) -> i8
// -----
func.func @masked_reduce_or_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <or>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_or_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.or"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_and_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <and>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_and_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(-1 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.and"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_xor_i8(%arg0: vector<32xi8>, %mask : vector<32xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <xor>, %arg0 : vector<32xi8> into i8 } : vector<32xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_xor_i8(
// CHECK-SAME: %[[INPUT:.*]]: vector<32xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<32xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0 : i8) : i8
// CHECK: %[[VL:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: "llvm.intr.vp.reduce.xor"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL]]) : (i8, vector<32xi8>, vector<32xi1>, i32) -> i8
// -----
func.func @masked_reduce_xor_i8_scalable(%arg0: vector<[32]xi8>, %mask : vector<[32]xi1>) -> i8 {
%0 = vector.mask %mask { vector.reduction <xor>, %arg0 : vector<[32]xi8> into i8 } : vector<[32]xi1> -> i8
return %0 : i8
}
// CHECK-LABEL: func.func @masked_reduce_xor_i8_scalable(
// CHECK-SAME: %[[INPUT:.*]]: vector<[32]xi8>,
// CHECK-SAME: %[[MASK:.*]]: vector<[32]xi1>) -> i8 {
// CHECK: %[[NEUTRAL:.*]] = llvm.mlir.constant(0 : i8) : i8
// CHECK: %[[VL_BASE:.*]] = llvm.mlir.constant(32 : i32) : i32
// CHECK: %[[VSCALE:.*]] = "llvm.intr.vscale"() : () -> i64
// CHECK: %[[CAST_IDX:.*]] = builtin.unrealized_conversion_cast %[[VSCALE]] : i64 to index
// CHECK: %[[CAST_I32:.*]] = arith.index_cast %[[CAST_IDX]] : index to i32
// CHECK: %[[VL_MUL:.*]] = arith.muli %[[VL_BASE]], %[[CAST_I32]] : i32
// CHECK: "llvm.intr.vp.reduce.xor"(%[[NEUTRAL]], %[[INPUT]], %[[MASK]], %[[VL_MUL]]) : (i8, vector<[32]xi8>, vector<[32]xi1>, i32) -> i8
|
