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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
; fold (urem x, 1) -> 0
define i32 @combine_urem_by_one(i32 %x) {
; CHECK-LABEL: combine_urem_by_one:
; CHECK: # %bb.0:
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: retq
%1 = urem i32 %x, 1
ret i32 %1
}
define <4 x i32> @combine_vec_urem_by_one(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_one:
; SSE: # %bb.0:
; SSE-NEXT: xorps %xmm0, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: combine_vec_urem_by_one:
; AVX: # %bb.0:
; AVX-NEXT: vxorps %xmm0, %xmm0, %xmm0
; AVX-NEXT: retq
%1 = urem <4 x i32> %x, <i32 1, i32 1, i32 1, i32 1>
ret <4 x i32> %1
}
; fold (urem x, -1) -> select((icmp eq x, -1), 0, x)
define i32 @combine_urem_by_negone(i32 %x) {
; CHECK-LABEL: combine_urem_by_negone:
; CHECK: # %bb.0:
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: cmpl $-1, %edi
; CHECK-NEXT: cmovnel %edi, %eax
; CHECK-NEXT: retq
%1 = urem i32 %x, -1
ret i32 %1
}
define <4 x i32> @combine_vec_urem_by_negone(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_negone:
; SSE: # %bb.0:
; SSE-NEXT: pcmpeqd %xmm1, %xmm1
; SSE-NEXT: pcmpeqd %xmm0, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: combine_vec_urem_by_negone:
; AVX: # %bb.0:
; AVX-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vpandn %xmm0, %xmm1, %xmm0
; AVX-NEXT: retq
%1 = urem <4 x i32> %x, <i32 -1, i32 -1, i32 -1, i32 -1>
ret <4 x i32> %1
}
; fold (urem x, INT_MIN) -> (and x, ~INT_MIN)
define i32 @combine_urem_by_minsigned(i32 %x) {
; CHECK-LABEL: combine_urem_by_minsigned:
; CHECK: # %bb.0:
; CHECK-NEXT: andl $2147483647, %edi # imm = 0x7FFFFFFF
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: retq
%1 = urem i32 %x, -2147483648
ret i32 %1
}
define <4 x i32> @combine_vec_urem_by_minsigned(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_minsigned:
; SSE: # %bb.0:
; SSE-NEXT: andps {{.*}}(%rip), %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_minsigned:
; AVX1: # %bb.0:
; AVX1-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_minsigned:
; AVX2: # %bb.0:
; AVX2-NEXT: vbroadcastss {{.*#+}} xmm1 = [2147483647,2147483647,2147483647,2147483647]
; AVX2-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = urem <4 x i32> %x, <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>
ret <4 x i32> %1
}
; TODO fold (urem x, x) -> 0
define i32 @combine_urem_dupe(i32 %x) {
; CHECK-LABEL: combine_urem_dupe:
; CHECK: # %bb.0:
; CHECK-NEXT: xorl %edx, %edx
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: divl %edi
; CHECK-NEXT: movl %edx, %eax
; CHECK-NEXT: retq
%1 = urem i32 %x, %x
ret i32 %1
}
define <4 x i32> @combine_vec_urem_dupe(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_dupe:
; SSE: # %bb.0:
; SSE-NEXT: pextrd $1, %xmm0, %eax
; SSE-NEXT: xorl %edx, %edx
; SSE-NEXT: divl %eax
; SSE-NEXT: movl %edx, %ecx
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: xorl %edx, %edx
; SSE-NEXT: divl %eax
; SSE-NEXT: movd %edx, %xmm1
; SSE-NEXT: pinsrd $1, %ecx, %xmm1
; SSE-NEXT: pextrd $2, %xmm0, %eax
; SSE-NEXT: xorl %edx, %edx
; SSE-NEXT: divl %eax
; SSE-NEXT: pinsrd $2, %edx, %xmm1
; SSE-NEXT: pextrd $3, %xmm0, %eax
; SSE-NEXT: xorl %edx, %edx
; SSE-NEXT: divl %eax
; SSE-NEXT: pinsrd $3, %edx, %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: combine_vec_urem_dupe:
; AVX: # %bb.0:
; AVX-NEXT: vpextrd $1, %xmm0, %eax
; AVX-NEXT: xorl %edx, %edx
; AVX-NEXT: divl %eax
; AVX-NEXT: movl %edx, %ecx
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: xorl %edx, %edx
; AVX-NEXT: divl %eax
; AVX-NEXT: vmovd %edx, %xmm1
; AVX-NEXT: vpinsrd $1, %ecx, %xmm1, %xmm1
; AVX-NEXT: vpextrd $2, %xmm0, %eax
; AVX-NEXT: xorl %edx, %edx
; AVX-NEXT: divl %eax
; AVX-NEXT: vpinsrd $2, %edx, %xmm1, %xmm1
; AVX-NEXT: vpextrd $3, %xmm0, %eax
; AVX-NEXT: xorl %edx, %edx
; AVX-NEXT: divl %eax
; AVX-NEXT: vpinsrd $3, %edx, %xmm1, %xmm0
; AVX-NEXT: retq
%1 = urem <4 x i32> %x, %x
ret <4 x i32> %1
}
; fold (urem x, pow2) -> (and x, (pow2-1))
define <4 x i32> @combine_vec_urem_by_pow2a(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_pow2a:
; SSE: # %bb.0:
; SSE-NEXT: andps {{.*}}(%rip), %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_pow2a:
; AVX1: # %bb.0:
; AVX1-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_pow2a:
; AVX2: # %bb.0:
; AVX2-NEXT: vbroadcastss {{.*#+}} xmm1 = [3,3,3,3]
; AVX2-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = urem <4 x i32> %x, <i32 4, i32 4, i32 4, i32 4>
ret <4 x i32> %1
}
define <4 x i32> @combine_vec_urem_by_pow2b(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_pow2b:
; SSE: # %bb.0:
; SSE-NEXT: andps {{.*}}(%rip), %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: combine_vec_urem_by_pow2b:
; AVX: # %bb.0:
; AVX-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
; AVX-NEXT: retq
%1 = urem <4 x i32> %x, <i32 1, i32 4, i32 8, i32 16>
ret <4 x i32> %1
}
define <4 x i32> @combine_vec_urem_by_pow2c(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: combine_vec_urem_by_pow2c:
; SSE: # %bb.0:
; SSE-NEXT: pslld $23, %xmm1
; SSE-NEXT: paddd {{.*}}(%rip), %xmm1
; SSE-NEXT: cvttps2dq %xmm1, %xmm1
; SSE-NEXT: pcmpeqd %xmm2, %xmm2
; SSE-NEXT: paddd %xmm1, %xmm2
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_pow2c:
; AVX1: # %bb.0:
; AVX1-NEXT: vpslld $23, %xmm1, %xmm1
; AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vcvttps2dq %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_pow2c:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
; AVX2-NEXT: vpsllvd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y
%2 = urem <4 x i32> %x, %1
ret <4 x i32> %2
}
define <4 x i32> @combine_vec_urem_by_pow2d(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: combine_vec_urem_by_pow2d:
; SSE: # %bb.0:
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
; SSE-NEXT: movdqa {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
; SSE-NEXT: movdqa %xmm3, %xmm4
; SSE-NEXT: psrld %xmm2, %xmm4
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,3,0,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm2[2,3,3,3,4,5,6,7]
; SSE-NEXT: movdqa %xmm3, %xmm6
; SSE-NEXT: psrld %xmm5, %xmm6
; SSE-NEXT: pblendw {{.*#+}} xmm6 = xmm4[0,1,2,3],xmm6[4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
; SSE-NEXT: movdqa %xmm3, %xmm4
; SSE-NEXT: psrld %xmm1, %xmm4
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[0,1,1,1,4,5,6,7]
; SSE-NEXT: psrld %xmm1, %xmm3
; SSE-NEXT: pblendw {{.*#+}} xmm3 = xmm4[0,1,2,3],xmm3[4,5,6,7]
; SSE-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,3],xmm3[4,5],xmm6[6,7]
; SSE-NEXT: pcmpeqd %xmm1, %xmm1
; SSE-NEXT: paddd %xmm3, %xmm1
; SSE-NEXT: pand %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_pow2d:
; AVX1: # %bb.0:
; AVX1-NEXT: vpsrldq {{.*#+}} xmm2 = xmm1[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
; AVX1-NEXT: vpsrld %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpsrlq $32, %xmm1, %xmm4
; AVX1-NEXT: vpsrld %xmm4, %xmm3, %xmm4
; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3],xmm2[4,5,6,7]
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm4 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; AVX1-NEXT: vpsrld %xmm4, %xmm3, %xmm4
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX1-NEXT: vpsrld %xmm1, %xmm3, %xmm1
; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm4[4,5,6,7]
; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_pow2d:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [2147483648,2147483648,2147483648,2147483648]
; AVX2-NEXT: vpsrlvd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = lshr <4 x i32> <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>, %y
%2 = urem <4 x i32> %x, %1
ret <4 x i32> %2
}
; fold (urem x, (shl pow2, y)) -> (and x, (add (shl pow2, y), -1))
define <4 x i32> @combine_vec_urem_by_shl_pow2a(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: combine_vec_urem_by_shl_pow2a:
; SSE: # %bb.0:
; SSE-NEXT: pslld $23, %xmm1
; SSE-NEXT: paddd {{.*}}(%rip), %xmm1
; SSE-NEXT: cvttps2dq %xmm1, %xmm1
; SSE-NEXT: pslld $2, %xmm1
; SSE-NEXT: pcmpeqd %xmm2, %xmm2
; SSE-NEXT: paddd %xmm1, %xmm2
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_shl_pow2a:
; AVX1: # %bb.0:
; AVX1-NEXT: vpslld $23, %xmm1, %xmm1
; AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vcvttps2dq %xmm1, %xmm1
; AVX1-NEXT: vpslld $2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_shl_pow2a:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [4,4,4,4]
; AVX2-NEXT: vpsllvd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = shl <4 x i32> <i32 4, i32 4, i32 4, i32 4>, %y
%2 = urem <4 x i32> %x, %1
ret <4 x i32> %2
}
define <4 x i32> @combine_vec_urem_by_shl_pow2b(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: combine_vec_urem_by_shl_pow2b:
; SSE: # %bb.0:
; SSE-NEXT: pslld $23, %xmm1
; SSE-NEXT: paddd {{.*}}(%rip), %xmm1
; SSE-NEXT: cvttps2dq %xmm1, %xmm1
; SSE-NEXT: pmulld {{.*}}(%rip), %xmm1
; SSE-NEXT: pcmpeqd %xmm2, %xmm2
; SSE-NEXT: paddd %xmm1, %xmm2
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_shl_pow2b:
; AVX1: # %bb.0:
; AVX1-NEXT: vpslld $23, %xmm1, %xmm1
; AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vcvttps2dq %xmm1, %xmm1
; AVX1-NEXT: vpmulld {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_shl_pow2b:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [1,4,8,16]
; AVX2-NEXT: vpsllvd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = shl <4 x i32> <i32 1, i32 4, i32 8, i32 16>, %y
%2 = urem <4 x i32> %x, %1
ret <4 x i32> %2
}
|
