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
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2021-2024 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
; RUN: %opt %use_old_pass_manager% -GenXPatternMatch -march=genx64 -mcpu=XeHPC -mtriple=spir64-unknown-unknown -S < %s | FileCheck %s
; optimize
define <32 x i32> @test_match_v32i32(<32 x i32> %op0, <32 x i32> %op1, <32 x i32> %op2) {
;CHECK: [[V1:%[^ ]+]] = call <32 x i32> @llvm.genx.add3.v32i32.v32i32(<32 x i32> %op0, <32 x i32> %op1, <32 x i32> %op2)
%1 = add <32 x i32> %op0, %op1
%2 = add <32 x i32> %1, %op2
ret <32 x i32> %2
}
; scalar value optimize
define i32 @test_match_i32(i32 %op0, i32 %op1, i32 %op2) {
;CHECK: [[SIMPLE_RES:%[^ ]+]] = call i32 @llvm.genx.add3.i32.i32(i32 %op0, i32 %op1, i32 %op2)
%1 = add i32 %op0, %op1
%2 = add i32 %1, %op2
ret i32 %2
}
; scalar value 2 const not opt
define i32 @test_match_i32cc(i32 %op0) {
;CHECK: [[TMP1:%[^ ]+]] = add i32 %op0, 7
;CHECK-NEXT: [[TMP2:%[^ ]+]] = add i32 [[TMP1]], 10
%1 = add i32 %op0, 7
%2 = add i32 %1, 10
ret i32 %2
}
; scalar value 1 const opt
define i32 @test_match_i32c(i32 %op0, i32 %op1) {
;CHECK: [[NEG_RES:%[^ ]+]] = call i32 @llvm.genx.add3.i32.i32(i32 %op0, i32 %op1, i32 10)
%1 = add i32 %op0, %op1
%2 = add i32 %1, 10
ret i32 %2
}
; float type - not opt
define <32 x float> @test_v32float(<32 x float> %op0, <32 x float> %op1, <32 x float> %op2) {
;CHECK: %1 = fadd <32 x float> %op0, %op1
%1 = fadd <32 x float> %op0, %op1
%2 = fadd <32 x float> %1, %op2
ret <32 x float> %2
}
; scalar value negative
; place negative values before, after places add3
define i32 @test_match_i32_neg(i32 %op0, i32 %op1, i32 %op2) {
;CHECK: [[NEG1:%[^ ]+]] = sub i32 0, %op1
;CHECK-NEXT: [[NEG2:%[^ ]+]] = sub i32 0, %op2
;CHECK-NEXT: [[NEG_RES:%[^ ]+]] = call i32 @llvm.genx.add3.i32.i32(i32 %op0, i32 [[NEG1]], i32 [[NEG2]])
%1 = sub i32 %op0, %op1
%2 = sub i32 %1, %op2
ret i32 %2
}
declare <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32>, i32, i32, i32, i16, i32)
; scalar value splat
; place rdregion before, after places add3 with vector type
define <4 x i32> @test_match_i32_splat(i32 %op0, i32 %op1, <4 x i32> %op2) {
;CHECK: [[BC1:%[^ ]+]] = bitcast i32 %op0 to <1 x i32>
;CHECK-NEXT: [[RD1:%[^ ]+]] = call <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32> [[BC1]], i32 0, i32 1, i32 0, i16 0, i32 undef)
;CHECK-NEXT: [[BC2:%[^ ]+]] = bitcast i32 %op1 to <1 x i32>
;CHECK-NEXT: [[RD2:%[^ ]+]] = call <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32> [[BC2]], i32 0, i32 1, i32 0, i16 0, i32 undef)
;CHECK-NEXT: [[RES:%[^ ]+]] = call <4 x i32> @llvm.genx.add3.v4i32.v4i32(<4 x i32> [[RD1]], <4 x i32> [[RD2]], <4 x i32> %op2)
;CHECK-NEXT: ret <4 x i32> [[RES]]
%1 = add i32 %op0, %op1
%bc = bitcast i32 %1 to <1 x i32>
%rd = call <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32> %bc, i32 0, i32 4, i32 0, i16 0, i32 undef)
%2 = add <4 x i32> %rd, %op2
ret <4 x i32> %2
}
; scalar value splat
; place rdregion before, after places add3 with vector type
define <4 x i32> @test_match_i32_splat2(i32 %op0, i32 %op1, <4 x i32> %op2) {
;CHECK: [[BC1:%[^ ]+]] = bitcast i32 %op0 to <1 x i32>
;CHECK-NEXT: [[RD1:%[^ ]+]] = call <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32> [[BC1]], i32 0, i32 1, i32 0, i16 0, i32 undef)
;CHECK-NEXT: [[BC2:%[^ ]+]] = bitcast i32 %op1 to <1 x i32>
;CHECK-NEXT: [[RD2:%[^ ]+]] = call <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32> [[BC2]], i32 0, i32 1, i32 0, i16 0, i32 undef)
;CHECK-NEXT: [[RES:%[^ ]+]] = call <4 x i32> @llvm.genx.add3.v4i32.v4i32(<4 x i32> %op2, <4 x i32> [[RD1]], <4 x i32> [[RD2]])
;CHECK-NEXT: ret <4 x i32> [[RES]]
%1 = add i32 %op0, %op1
%bc = bitcast i32 %1 to <1 x i32>
%rd = call <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32> %bc, i32 0, i32 4, i32 0, i16 0, i32 undef)
%2 = add <4 x i32> %op2, %rd
ret <4 x i32> %2
}
; scalar value splat
; place rdregion before, after places add3 with vector type
declare <4 x i16> @llvm.genx.rdregioni.v4i16.v1i16.i16(<1 x i16>, i32, i32, i32, i16, i32)
define <4 x i32> @test_match_zext_match(i16 %op0, i16 %op1, <4 x i32> %op2) {
;CHECK: [[BC1:%[^ ]+]] = bitcast i16 %op0 to <1 x i16>
;CHECK-NEXT: [[RD1:%[^ ]+]] = call <4 x i16> @llvm.genx.rdregioni.v4i16.v1i16.i16(<1 x i16> [[BC1]], i32 0, i32 1, i32 0, i16 0, i32 undef)
;CHECK-NEXT: [[EX1:%[^ ]+]] = zext <4 x i16> [[RD1]] to <4 x i32>
;CHECK-NEXT: [[BC2:%[^ ]+]] = bitcast i16 %op1 to <1 x i16>
;CHECK-NEXT: [[RD2:%[^ ]+]] = call <4 x i16> @llvm.genx.rdregioni.v4i16.v1i16.i16(<1 x i16> [[BC2]], i32 0, i32 1, i32 0, i16 0, i32 undef)
;CHECK-NEXT: [[EX2:%[^ ]+]] = zext <4 x i16> [[RD2]] to <4 x i32>
;CHECK-NEXT: [[RES:%[^ ]+]] = call <4 x i32> @llvm.genx.add3.v4i32.v4i32(<4 x i32> %op2, <4 x i32> [[EX1]], <4 x i32> [[EX2]])
;CHECK-NEXT: ret <4 x i32> [[RES]]
%op0_c = zext i16 %op0 to i32
%op1_c = zext i16 %op1 to i32
%1 = add i32 %op0_c, %op1_c
%bc = bitcast i32 %1 to <1 x i32>
%rd = call <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32> %bc, i32 0, i32 4, i32 0, i16 0, i32 undef)
%2 = add <4 x i32> %op2, %rd
ret <4 x i32> %2
}
; scalar value splat
; place rdregion before, after places add3 with vector type
define <4 x i32> @test_match_sext_match(i16 %op0, i16 %op1, <4 x i32> %op2) {
;CHECK: [[BC1:%[^ ]+]] = bitcast i16 %op0 to <1 x i16>
;CHECK-NEXT: [[RD1:%[^ ]+]] = call <4 x i16> @llvm.genx.rdregioni.v4i16.v1i16.i16(<1 x i16> [[BC1]], i32 0, i32 1, i32 0, i16 0, i32 undef)
;CHECK-NEXT: [[SX1:%[^ ]+]] = sext <4 x i16> [[RD1]] to <4 x i32>
;CHECK-NEXT: [[BC2:%[^ ]+]] = bitcast i16 %op1 to <1 x i16>
;CHECK-NEXT: [[RD2:%[^ ]+]] = call <4 x i16> @llvm.genx.rdregioni.v4i16.v1i16.i16(<1 x i16> [[BC2]], i32 0, i32 1, i32 0, i16 0, i32 undef)
;CHECK-NEXT: [[SX2:%[^ ]+]] = sext <4 x i16> [[RD2]] to <4 x i32>
;CHECK-NEXT: [[RES:%[^ ]+]] = call <4 x i32> @llvm.genx.add3.v4i32.v4i32(<4 x i32> [[SX1]], <4 x i32> [[SX2]], <4 x i32> %op2)
;CHECK-NEXT: ret <4 x i32> [[RES]]
%op0_c = sext i16 %op0 to i32
%op1_c = sext i16 %op1 to i32
%1 = add i32 %op0_c, %op1_c
%bc = bitcast i32 %1 to <1 x i32>
%rd = call <4 x i32> @llvm.genx.rdregioni.v4i32.v1i32.i16(<1 x i32> %bc, i32 0, i32 4, i32 0, i16 0, i32 undef)
%2 = add <4 x i32> %rd, %op2
ret <4 x i32> %2
}
declare i32 @llvm.genx.ssadd.sat.i32.i32(i32, i32)
; No optimization
define i32 @test_match_i32ssadd(i32 %op0, i32 %op1, i32 %op2) {
;CHECK: %1 = add i32 %op0, %op1
;CHECK-NEXT: %2 = call i32 @llvm.genx.ssadd.sat.i32.i32(i32 %1, i32 %op2)
%1 = add i32 %op0, %op1
%2 = call i32 @llvm.genx.ssadd.sat.i32.i32(i32 %1, i32 %op2)
ret i32 %2
}
define i32 @test_match_i32_imm(i32 %a, i32 %b) {
; CHECK: call i32 @llvm.genx.add3.i32.i32(i32 %a, i32 %b, i32 123)
%tmp = add i32 %a, 123
%res = add i32 %tmp, %b
ret i32 %res
}
declare <8 x i32> @llvm.genx.rdregioni.v8i32.v1i32.i16(<16 x i32>, i32, i32, i32, i16, i32)
define <8 x i32> @test_match_rdregion_ext_v8i8(<8 x i32> %acc, <16 x i8> %src) {
; CHECK: %tmp = add <8 x i32> %a, %b
; CHECK: %res = add <8 x i32> %acc, %tmp
%ext = zext <16 x i8> %src to <16 x i32>
%a = call <8 x i32> @llvm.genx.rdregioni.v8i32.v1i32.i16(<16 x i32> %ext, i32 8, i32 8, i32 1, i16 0, i32 undef)
%b = call <8 x i32> @llvm.genx.rdregioni.v8i32.v1i32.i16(<16 x i32> %ext, i32 8, i32 8, i32 1, i16 8, i32 undef)
%tmp = add <8 x i32> %a, %b
%res = add <8 x i32> %acc, %tmp
ret <8 x i32> %res
}
; =============== ADDITIONAL FAILED MATCH ===============
; optimize
define <32 x i8> @test_match_v32i8(<32 x i8> %op0, <32 x i8> %op1, <32 x i8> %op2) {
;CHECK: [[F1:%[^ ]+]] = add <32 x i8> %op0, %op1
;CHECK: add <32 x i8> [[F1]], %op2
%1 = add <32 x i8> %op0, %op1
%2 = add <32 x i8> %1, %op2
ret <32 x i8> %2
}
; scalar value optimize
define i8 @test_match_i8(i8 %op0, i8 %op1, i8 %op2) {
;CHECK: [[F2:%[^ ]+]] = add i8 %op0, %op1
;CHECK: add i8 [[F2:%[^ ]+]], %op2
%1 = add i8 %op0, %op1
%2 = add i8 %1, %op2
ret i8 %2
}
; scalar value 2 const not opt
define i8 @test_match_i8cc(i8 %op0) {
;CHECK: [[TMP1:%[^ ]+]] = add i8 %op0, 7
;CHECK-NEXT: [[TMP2:%[^ ]+]] = add i8 [[TMP1]], 10
%1 = add i8 %op0, 7
%2 = add i8 %1, 10
ret i8 %2
}
; scalar value 1 const opt
define i8 @test_match_i8c(i8 %op0, i8 %op1) {
;CHECK: [[F3:%[^ ]+]] = add i8 %op0, %op1
;CHECK: add i8 [[F3]], 10
%1 = add i8 %op0, %op1
%2 = add i8 %1, 10
ret i8 %2
}
; double type - not opt
define <32 x double> @test_v32double(<32 x double> %op0, <32 x double> %op1, <32 x double> %op2) {
;CHECK: %1 = fadd <32 x double> %op0, %op1
%1 = fadd <32 x double> %op0, %op1
%2 = fadd <32 x double> %1, %op2
ret <32 x double> %2
}
; optimize
define <32 x i64> @test_match_v32i64(<32 x i64> %op0, <32 x i64> %op1, <32 x i64> %op2) {
;CHECK: [[F4:%[^ ]+]] = add <32 x i64> %op0, %op1
;CHECK: add <32 x i64> [[F4]], %op2
%1 = add <32 x i64> %op0, %op1
%2 = add <32 x i64> %1, %op2
ret <32 x i64> %2
}
; scalar value optimize
define i64 @test_match_i64(i64 %op0, i64 %op1, i64 %op2) {
;CHECK: [[F5:%[^ ]+]] = add i64 %op0, %op1
;CHECK: add i64 [[F5]], %op2
%1 = add i64 %op0, %op1
%2 = add i64 %1, %op2
ret i64 %2
}
; scalar value 2 const not opt
define i64 @test_match_i64cc(i64 %op0) {
;CHECK: [[TMP1:%[^ ]+]] = add i64 %op0, 7
;CHECK-NEXT: [[TMP2:%[^ ]+]] = add i64 [[TMP1]], 10
%1 = add i64 %op0, 7
%2 = add i64 %1, 10
ret i64 %2
}
; scalar value 1 const opt
define i64 @test_match_i64c(i64 %op0, i64 %op1) {
;CHECK: [[F6:%[^ ]+]] = add i64 %op0, %op1
;CHECK: add i64 [[F6]], 10
%1 = add i64 %op0, %op1
%2 = add i64 %1, 10
ret i64 %2
}
|
