Lines Matching refs:AVX2

4 …-opt %s -test-vector-transpose-lowering=avx2=1 -split-input-file | FileCheck %s --check-prefix=AVX2
49 // AVX2-LABEL: func @transpose4x8
51   //      AVX2: vector.extract {{.*}}[0]
52   // AVX2-NEXT: vector.extract {{.*}}[1]
53   // AVX2-NEXT: vector.extract {{.*}}[2]
54   // AVX2-NEXT: vector.extract {{.*}}[3]
55   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
56   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
57   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
58   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
59   // AVX2-NEXT: vector.shuffle {{.*}} [0, 1, 8, 9, 4, 5, 12, 13] : vector<8xf32>, vector<8xf32>
60   // AVX2-NEXT: vector.shuffle {{.*}} [2, 3, 10, 11, 6, 7, 14, 15] : vector<8xf32>, vector<8xf32>
61   // AVX2-NEXT: vector.shuffle {{.*}} [0, 1, 8, 9, 4, 5, 12, 13] : vector<8xf32>, vector<8xf32>
62   // AVX2-NEXT: vector.shuffle {{.*}} [2, 3, 10, 11, 6, 7, 14, 15] : vector<8xf32>, vector<8xf32>
63   // AVX2-NEXT: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
64   // AVX2-NEXT: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
65   // AVX2-NEXT: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
66   // AVX2-NEXT: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
67   // AVX2-NEXT: vector.insert {{.*}}[0]
68   // AVX2-NEXT: vector.insert {{.*}}[1]
69   // AVX2-NEXT: vector.insert {{.*}}[2]
70   // AVX2-NEXT: vector.insert {{.*}}[3]
71   // AVX2-NEXT: vector.shape_cast {{.*}} vector<4x8xf32> to vector<32xf32>
72   // AVX2-NEXT: vector.shape_cast {{.*}} vector<32xf32> to vector<8x4xf32>
79 // AVX2-LABEL: func @transpose021_1x4x8
81   //      AVX2: vector.extract {{.*}}[0, 0]
82   // AVX2-NEXT: vector.extract {{.*}}[0, 1]
83   // AVX2-NEXT: vector.extract {{.*}}[0, 2]
84   // AVX2-NEXT: vector.extract {{.*}}[0, 3]
85   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
86   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
87   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
88   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
89   // AVX2-NEXT: vector.shuffle {{.*}} [0, 1, 8, 9, 4, 5, 12, 13] : vector<8xf32>, vector<8xf32>
90   // AVX2-NEXT: vector.shuffle {{.*}} [2, 3, 10, 11, 6, 7, 14, 15] : vector<8xf32>, vector<8xf32>
91   // AVX2-NEXT: vector.shuffle {{.*}} [0, 1, 8, 9, 4, 5, 12, 13] : vector<8xf32>, vector<8xf32>
92   // AVX2-NEXT: vector.shuffle {{.*}} [2, 3, 10, 11, 6, 7, 14, 15] : vector<8xf32>, vector<8xf32>
93   // AVX2-NEXT: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
94   // AVX2-NEXT: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
95   // AVX2-NEXT: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
96   // AVX2-NEXT: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
97   // AVX2-NEXT: vector.insert {{.*}}[0]
98   // AVX2-NEXT: vector.insert {{.*}}[1]
99   // AVX2-NEXT: vector.insert {{.*}}[2]
100   // AVX2-NEXT: vector.insert {{.*}}[3]
101   // AVX2-NEXT: vector.shape_cast {{.*}} vector<4x8xf32> to vector<32xf32>
102   // AVX2-NEXT: vector.shape_cast {{.*}} vector<32xf32> to vector<1x8x4xf32>
109 // AVX2-LABEL: func @transpose8x8
111   //      AVX2: vector.extract {{.*}}[0]
112   // AVX2-NEXT: vector.extract {{.*}}[1]
113   // AVX2-NEXT: vector.extract {{.*}}[2]
114   // AVX2-NEXT: vector.extract {{.*}}[3]
115   // AVX2-NEXT: vector.extract {{.*}}[4]
116   // AVX2-NEXT: vector.extract {{.*}}[5]
117   // AVX2-NEXT: vector.extract {{.*}}[6]
118   // AVX2-NEXT: vector.extract {{.*}}[7]
119   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
120   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
121   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
122   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
123   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
124   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
125   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
126   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
127   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
128 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
129 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
130 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
131 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
132 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
133 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
134 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
135 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
136   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
137   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
138   // AVX2-NEXT: vector.insert {{.*}}[0]
139   // AVX2-NEXT: vector.insert {{.*}}[1]
140   // AVX2-NEXT: vector.insert {{.*}}[2]
141   // AVX2-NEXT: vector.insert {{.*}}[3]
142   // AVX2-NEXT: vector.insert {{.*}}[4]
143   // AVX2-NEXT: vector.insert {{.*}}[5]
144   // AVX2-NEXT: vector.insert {{.*}}[6]
145   // AVX2-NEXT: vector.insert {{.*}}[7]
152 // AVX2-LABEL: func @transpose021_1x8x8
154   //      AVX2: vector.extract {{.*}}[0, 0]
155   // AVX2-NEXT: vector.extract {{.*}}[0, 1]
156   // AVX2-NEXT: vector.extract {{.*}}[0, 2]
157   // AVX2-NEXT: vector.extract {{.*}}[0, 3]
158   // AVX2-NEXT: vector.extract {{.*}}[0, 4]
159   // AVX2-NEXT: vector.extract {{.*}}[0, 5]
160   // AVX2-NEXT: vector.extract {{.*}}[0, 6]
161   // AVX2-NEXT: vector.extract {{.*}}[0, 7]
162   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
163   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
164   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
165   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
166   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
167   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
168   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
169   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
170   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
171 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
172 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
173 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
174 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
175 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
176 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
177 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
178 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
179   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
180   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
181   // AVX2-NEXT: vector.insert {{.*}}[0]
182   // AVX2-NEXT: vector.insert {{.*}}[1]
183   // AVX2-NEXT: vector.insert {{.*}}[2]
184   // AVX2-NEXT: vector.insert {{.*}}[3]
185   // AVX2-NEXT: vector.insert {{.*}}[4]
186   // AVX2-NEXT: vector.insert {{.*}}[5]
187   // AVX2-NEXT: vector.insert {{.*}}[6]
188   // AVX2-NEXT: vector.insert {{.*}}[7]
189   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<1x8x8xf32>
196 // AVX2-LABEL: func @transpose120_8x1x8
198   //      AVX2: vector.extract {{.*}}[0, 0]
199   // AVX2-NEXT: vector.extract {{.*}}[1, 0]
200   // AVX2-NEXT: vector.extract {{.*}}[2, 0]
201   // AVX2-NEXT: vector.extract {{.*}}[3, 0]
202   // AVX2-NEXT: vector.extract {{.*}}[4, 0]
203   // AVX2-NEXT: vector.extract {{.*}}[5, 0]
204   // AVX2-NEXT: vector.extract {{.*}}[6, 0]
205   // AVX2-NEXT: vector.extract {{.*}}[7, 0]
206   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
207   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
208   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
209   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
210   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
211   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
212   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
213   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
214   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
215 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
216 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
217 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
218 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
219 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
220 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
221 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
222 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
223   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
224   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
225   // AVX2-NEXT: vector.insert {{.*}}[0]
226   // AVX2-NEXT: vector.insert {{.*}}[1]
227   // AVX2-NEXT: vector.insert {{.*}}[2]
228   // AVX2-NEXT: vector.insert {{.*}}[3]
229   // AVX2-NEXT: vector.insert {{.*}}[4]
230   // AVX2-NEXT: vector.insert {{.*}}[5]
231   // AVX2-NEXT: vector.insert {{.*}}[6]
232   // AVX2-NEXT: vector.insert {{.*}}[7]
233   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<1x8x8xf32>
240 // AVX2-LABEL: func @transpose120_8x8x1
242   //      AVX2: vector.shape_cast %{{.*}} : vector<8x8x1xf32> to vector<8x8xf32>
243   // AVX2-NEXT: vector.extract {{.*}}[0]
244   // AVX2-NEXT: vector.extract {{.*}}[1]
245   // AVX2-NEXT: vector.extract {{.*}}[2]
246   // AVX2-NEXT: vector.extract {{.*}}[3]
247   // AVX2-NEXT: vector.extract {{.*}}[4]
248   // AVX2-NEXT: vector.extract {{.*}}[5]
249   // AVX2-NEXT: vector.extract {{.*}}[6]
250   // AVX2-NEXT: vector.extract {{.*}}[7]
251   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
252   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
253   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
254   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
255   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
256   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
257   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
258   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
259   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
260 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
261 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
262 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
263 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
264 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
265 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
266 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
267 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
268   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
269   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
270   // AVX2-NEXT: vector.insert {{.*}}[0]
271   // AVX2-NEXT: vector.insert {{.*}}[1]
272   // AVX2-NEXT: vector.insert {{.*}}[2]
273   // AVX2-NEXT: vector.insert {{.*}}[3]
274   // AVX2-NEXT: vector.insert {{.*}}[4]
275   // AVX2-NEXT: vector.insert {{.*}}[5]
276   // AVX2-NEXT: vector.insert {{.*}}[6]
277   // AVX2-NEXT: vector.insert {{.*}}[7]
278   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<8x1x8xf32>
285 // AVX2-LABEL: func @transpose102_8x8x1
287   //      AVX2: vector.shape_cast %{{.*}} : vector<8x8x1xf32> to vector<8x8xf32>
288   // AVX2-NEXT: vector.extract {{.*}}[0]
289   // AVX2-NEXT: vector.extract {{.*}}[1]
290   // AVX2-NEXT: vector.extract {{.*}}[2]
291   // AVX2-NEXT: vector.extract {{.*}}[3]
292   // AVX2-NEXT: vector.extract {{.*}}[4]
293   // AVX2-NEXT: vector.extract {{.*}}[5]
294   // AVX2-NEXT: vector.extract {{.*}}[6]
295   // AVX2-NEXT: vector.extract {{.*}}[7]
296   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
297   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
298   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
299   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
300   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
301   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
302   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
303   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
304   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
305 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
306 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
307 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
308 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
309 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
310 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
311 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
312 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
313   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
314   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
315   // AVX2-NEXT: vector.insert {{.*}}[0]
316   // AVX2-NEXT: vector.insert {{.*}}[1]
317   // AVX2-NEXT: vector.insert {{.*}}[2]
318   // AVX2-NEXT: vector.insert {{.*}}[3]
319   // AVX2-NEXT: vector.insert {{.*}}[4]
320   // AVX2-NEXT: vector.insert {{.*}}[5]
321   // AVX2-NEXT: vector.insert {{.*}}[6]
322   // AVX2-NEXT: vector.insert {{.*}}[7]
323   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<8x8x1xf32>
330 // AVX2-LABEL: func @transpose201_8x1x8
332   //      AVX2: vector.extract {{.*}}[0, 0]
333   // AVX2-NEXT: vector.extract {{.*}}[1, 0]
334   // AVX2-NEXT: vector.extract {{.*}}[2, 0]
335   // AVX2-NEXT: vector.extract {{.*}}[3, 0]
336   // AVX2-NEXT: vector.extract {{.*}}[4, 0]
337   // AVX2-NEXT: vector.extract {{.*}}[5, 0]
338   // AVX2-NEXT: vector.extract {{.*}}[6, 0]
339   // AVX2-NEXT: vector.extract {{.*}}[7, 0]
340   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
341   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
342   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
343   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
344   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
345   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
346   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
347   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
348   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
349 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
350 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
351 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
352 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
353 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
354 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
355 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
356 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
357   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
358   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
359   // AVX2-NEXT: vector.insert {{.*}}[0]
360   // AVX2-NEXT: vector.insert {{.*}}[1]
361   // AVX2-NEXT: vector.insert {{.*}}[2]
362   // AVX2-NEXT: vector.insert {{.*}}[3]
363   // AVX2-NEXT: vector.insert {{.*}}[4]
364   // AVX2-NEXT: vector.insert {{.*}}[5]
365   // AVX2-NEXT: vector.insert {{.*}}[6]
366   // AVX2-NEXT: vector.insert {{.*}}[7]
367   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<8x8x1xf32>
374 // AVX2-LABEL: func @transpose201_1x8x8
376   //      AVX2: vector.extract {{.*}}[0, 0]
377   // AVX2-NEXT: vector.extract {{.*}}[0, 1]
378   // AVX2-NEXT: vector.extract {{.*}}[0, 2]
379   // AVX2-NEXT: vector.extract {{.*}}[0, 3]
380   // AVX2-NEXT: vector.extract {{.*}}[0, 4]
381   // AVX2-NEXT: vector.extract {{.*}}[0, 5]
382   // AVX2-NEXT: vector.extract {{.*}}[0, 6]
383   // AVX2-NEXT: vector.extract {{.*}}[0, 7]
384   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
385   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
386   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
387   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
388   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
389   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
390   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
391   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
392   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
393 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
394 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
395 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
396 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
397 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
398 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
399 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
400 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
401   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
402   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
403   // AVX2-NEXT: vector.insert {{.*}}[0]
404   // AVX2-NEXT: vector.insert {{.*}}[1]
405   // AVX2-NEXT: vector.insert {{.*}}[2]
406   // AVX2-NEXT: vector.insert {{.*}}[3]
407   // AVX2-NEXT: vector.insert {{.*}}[4]
408   // AVX2-NEXT: vector.insert {{.*}}[5]
409   // AVX2-NEXT: vector.insert {{.*}}[6]
410   // AVX2-NEXT: vector.insert {{.*}}[7]
411   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<8x1x8xf32>
418 // AVX2-LABEL: func @transpose210_8x1x8
420   //      AVX2: vector.extract {{.*}}[0, 0]
421   // AVX2-NEXT: vector.extract {{.*}}[1, 0]
422   // AVX2-NEXT: vector.extract {{.*}}[2, 0]
423   // AVX2-NEXT: vector.extract {{.*}}[3, 0]
424   // AVX2-NEXT: vector.extract {{.*}}[4, 0]
425   // AVX2-NEXT: vector.extract {{.*}}[5, 0]
426   // AVX2-NEXT: vector.extract {{.*}}[6, 0]
427   // AVX2-NEXT: vector.extract {{.*}}[7, 0]
428   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
429   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
430   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
431   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
432   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
433   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
434   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
435   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
436   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
437 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
438 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
439 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
440 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
441 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
442 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
443 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
444 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
445   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
446   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
447   // AVX2-NEXT: vector.insert {{.*}}[0]
448   // AVX2-NEXT: vector.insert {{.*}}[1]
449   // AVX2-NEXT: vector.insert {{.*}}[2]
450   // AVX2-NEXT: vector.insert {{.*}}[3]
451   // AVX2-NEXT: vector.insert {{.*}}[4]
452   // AVX2-NEXT: vector.insert {{.*}}[5]
453   // AVX2-NEXT: vector.insert {{.*}}[6]
454   // AVX2-NEXT: vector.insert {{.*}}[7]
455   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<8x1x8xf32>
462 // AVX2-LABEL: func @transpose210_8x8x1
464   //      AVX2: vector.shape_cast %{{.*}} : vector<8x8x1xf32> to vector<8x8xf32>
465   // AVX2-NEXT: vector.extract {{.*}}[0]
466   // AVX2-NEXT: vector.extract {{.*}}[1]
467   // AVX2-NEXT: vector.extract {{.*}}[2]
468   // AVX2-NEXT: vector.extract {{.*}}[3]
469   // AVX2-NEXT: vector.extract {{.*}}[4]
470   // AVX2-NEXT: vector.extract {{.*}}[5]
471   // AVX2-NEXT: vector.extract {{.*}}[6]
472   // AVX2-NEXT: vector.extract {{.*}}[7]
473   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
474   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
475   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
476   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
477   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
478   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
479   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
480   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
481   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
482 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
483 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
484 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
485 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
486 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
487 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
488 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
489 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
490   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
491   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
492   // AVX2-NEXT: vector.insert {{.*}}[0]
493   // AVX2-NEXT: vector.insert {{.*}}[1]
494   // AVX2-NEXT: vector.insert {{.*}}[2]
495   // AVX2-NEXT: vector.insert {{.*}}[3]
496   // AVX2-NEXT: vector.insert {{.*}}[4]
497   // AVX2-NEXT: vector.insert {{.*}}[5]
498   // AVX2-NEXT: vector.insert {{.*}}[6]
499   // AVX2-NEXT: vector.insert {{.*}}[7]
500   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<1x8x8xf32>
507 // AVX2-LABEL: func @transpose210_1x8x8
509   //      AVX2: vector.extract {{.*}}[0, 0]
510   // AVX2-NEXT: vector.extract {{.*}}[0, 1]
511   // AVX2-NEXT: vector.extract {{.*}}[0, 2]
512   // AVX2-NEXT: vector.extract {{.*}}[0, 3]
513   // AVX2-NEXT: vector.extract {{.*}}[0, 4]
514   // AVX2-NEXT: vector.extract {{.*}}[0, 5]
515   // AVX2-NEXT: vector.extract {{.*}}[0, 6]
516   // AVX2-NEXT: vector.extract {{.*}}[0, 7]
517   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
518   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
519   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
520   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
521   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
522   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
523   // AVX2-NEXT: vector.shuffle {{.*}} [0, 8, 1, 9, 4, 12, 5, 13] : vector<8xf32>, vector<8xf32>
524   // AVX2-NEXT: vector.shuffle {{.*}} [2, 10, 3, 11, 6, 14, 7, 15] : vector<8xf32>, vector<8xf32>
525   // AVX2-COUNT-4: vector.shuffle {{.*}} [2, 3, 8, 9, 6, 7, 12, 13] : vector<8xf32>, vector<8xf32>
526 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
527 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
528 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
529 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
530 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
531 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
532 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0xcc", "=x,x,x" {{.*}} : (…
533 …// AVX2-NEXT: llvm.inline_asm asm_dialect = intel "vblendps $0, $1, $2, 0x33", "=x,x,x" {{.*}} : (…
534   // AVX2-COUNT-4: vector.shuffle {{.*}} [0, 1, 2, 3, 8, 9, 10, 11] : vector<8xf32>, vector<8xf32>
535   // AVX2-COUNT-4: vector.shuffle {{.*}} [4, 5, 6, 7, 12, 13, 14, 15] : vector<8xf32>, vector<8xf32>
536   // AVX2-NEXT: vector.insert {{.*}}[0]
537   // AVX2-NEXT: vector.insert {{.*}}[1]
538   // AVX2-NEXT: vector.insert {{.*}}[2]
539   // AVX2-NEXT: vector.insert {{.*}}[3]
540   // AVX2-NEXT: vector.insert {{.*}}[4]
541   // AVX2-NEXT: vector.insert {{.*}}[5]
542   // AVX2-NEXT: vector.insert {{.*}}[6]
543   // AVX2-NEXT: vector.insert {{.*}}[7]
544   // AVX2-NEXT: vector.shape_cast %{{.*}} : vector<8x8xf32> to vector<8x8x1xf32>
556 // AVX2-NOT: vector.shuffle
560 // AVX2-LABEL: func @transpose021_8x1x8
566 // AVX2-NOT: vector.shuffle
570 // AVX2-LABEL: func @transpose021_8x8x1
576 // AVX2-NOT: vector.shuffle
581 // AVX2-LABEL: func @transpose102_1x8x8
603 // AVX2-NOT: vector.shuffle
608 // AVX2-LABEL: func @transpose102_8x1x8
630 // AVX2-NOT: vector.shuffle
635 // AVX2-LABEL: func @transpose1023_1x1x8x8
644 // AVX2-NOT: vector.shuffle
648 // AVX2-LABEL: func @transpose120_1x8x8
655 // AVX2-NOT: vector.shuffle
659 // AVX2-LABEL: func @transpose201_8x8x1
665 // AVX2-NOT: vector.shuffle