Lines Matching full:tensor
4 func.func @generalize_matmul_tensor_f16f64f32(%A : tensor<16x8xf16>, %B: tensor<8x32xf64>, %C: tensor<16x32xf32>) -> tensor<16x32xf32> {
5 %0 = linalg.matmul ins(%A, %B: tensor<16x8xf16>, tensor<8x32xf64>)
6 outs(%C: tensor<16x32xf32>) -> tensor<16x32xf32>
7 return %0: tensor<16x32xf32>
18 // CHECK-NEXT: -> tensor<16x32xf32>
23 func.func @generalize_matmul_tensor_i16i64i32(%A : tensor<16x8xi16>, %B: tensor<8x32xi64>, %C: tensor<16x32xi32>) -> tensor<16x32xi32> {
24 %0 = linalg.matmul ins(%A, %B: tensor<16x8xi16>, tensor<8x32xi64>)
25 outs(%C: tensor<16x32xi32>) -> tensor<16x32xi32>
26 return %0: tensor<16x32xi32>
37 // CHECK-NEXT: -> tensor<16x32xi32>
43 func.func @generalize_matmul_tensor_i16i64i32_unsigned(%A : tensor<16x8xi16>, %B: tensor<8x32xi64>, %C: tensor<16x32xi32>) -> tensor<16x32xi32> {
45 ins(%A, %B: tensor<16x8xi16>, tensor<8x32xi64>)
46 outs(%C: tensor<16x32xi32>) -> tensor<16x32xi32>
47 return %0: tensor<16x32xi32>
55 func.func @generalize_matmul_tensor_i16i64f32(%A : tensor<16x8xi16>, %B: tensor<8x32xi64>, %C: tensor<16x32xf32>) -> tensor<16x32xf32> {
56 %0 = linalg.matmul ins(%A, %B: tensor<16x8xi16>, tensor<8x32xi64>)
57 outs(%C: tensor<16x32xf32>) -> tensor<16x32xf32>
58 return %0: tensor<16x32xf32>
68 func.func @generalize_matmul_tensor_f16f64i32(%A : tensor<16x8xf16>, %B: tensor<8x32xf64>, %C: tensor<16x32xi32>) -> tensor<16x32xi32> {
69 %0 = linalg.matmul ins(%A, %B: tensor<16x8xf16>, tensor<8x32xf64>)
70 outs(%C: tensor<16x32xi32>) -> tensor<16x32xi32>
71 return %0: tensor<16x32xi32>
81 func.func @generalize_matmul_unsigned_tensor_i16i64i32(%A : tensor<16x8xi16>, %B: tensor<8x32xi64>, %C: tensor<16x32xi32>) -> tensor<16x32xi32> {
83 ins(%A, %B: tensor<16x8xi16>, tensor<8x32xi64>)
84 outs(%C: tensor<16x32xi32>) -> tensor<16x32xi32>
85 return %0: tensor<16x32xi32>
95 func.func @generalize_matmul_unsigned_tensor_i16i64f32(%A : tensor<16x8xi16>, %B: tensor<8x32xi64>, %C: tensor<16x32xf32>) -> tensor<16x32xf32> {
97 ins(%A, %B: tensor<16x8xi16>, tensor<8x32xi64>)
98 outs(%C: tensor<16x32xf32>) -> tensor<16x32xf32>
99 return %0: tensor<16x32xf32>
109 func.func @generalize_matmul_unsigned_tensor_f16f64i32(%A : tensor<16x8xf16>, %B: tensor<8x32xf64>, %C: tensor<16x32xi32>) -> tensor<16x32xi32> {
111 ins(%A, %B: tensor<16x8xf16>, tensor<8x32xf64>)
112 outs(%C: tensor<16x32xi32>) -> tensor<16x32xi32>
113 return %0: tensor<16x32xi32>
124 %A: tensor<16x8xf16>,
125 %B: tensor<8x32xf64>,
126 %C: tensor<16x32xf32>) -> tensor<16x32xf32> {
131 ins(%A, %B: tensor<16x8xf16>, tensor<8x32xf64>)
132 outs(%C: tensor<16x32xf32>) -> tensor<16x32xf32>
133 return %0: tensor<16x32xf32>
144 // CHECK-NEXT: -> tensor<16x32xf32>
149 %A: tensor<24x12xf16>,
150 %B: tensor<12x25xf16>,
151 %C: tensor<24x25xf32>) -> tensor<24x25xf16> {
156 ins(%A, %B : tensor<24x12xf16>, tensor<12x25xf16>)
157 outs(%C : tensor<24x25xf32>) -> tensor<24x25xf32>
158 %1 = arith.truncf %0 : tensor<24x25xf32> to tensor<24x25xf16>
159 func.return %1 : tensor<24x25xf16>
170 // CHECK-NEXT: -> tensor<24x25xf32>
171 // CHECK-NEXT: %[[RES:.+]] = arith.truncf {{.*}} : tensor<24x25xf32> to tensor<24x25xf16>
175 func.func @generalize_pooling_nhwc_max_f32(%input : tensor<1x4x16x1xf32>, %shape: tensor<2x2xf32>, %output: tensor<1x2x4x1xf32>) -> tensor<1x2x4x1xf32> {
176 %0 = linalg.pooling_nhwc_max {dilations = dense<[1, 2]> : tensor<2xi64>, strides = dense<[2, 4]> : tensor<2xi64>}
177 ins(%input, %shape : tensor<1x4x16x1xf32>, tensor<2x2xf32>) outs(%output : tensor<1x2x4x1xf32>) -> tensor<1x2x4x1xf32>
178 return %0: tensor<1x2x4x1xf32>
185 // CHECK-NEXT: -> tensor<1x2x4x1xf32>
189 func.func @generalize_pooling_nwc_max_f32(%input : tensor<1x16x1xf32>, %shape: tensor<2xf32>, %output: tensor<1x4x1xf32>) -> tensor<1x4x1xf32> {
190 %0 = linalg.pooling_nwc_max {dilations = dense<[2]> : tensor<1xi64>, strides = dense<[4]> : tensor<1xi64>}
191 ins(%input, %shape : tensor<1x16x1xf32>, tensor<2xf32>) outs(%output : tensor<1x4x1xf32>) -> tensor<1x4x1xf32>
192 return %0: tensor<1x4x1xf32>
199 // CHECK-NEXT: -> tensor<1x4x1xf32>
203 func.func @generalize_pooling_nhwc_max_i32(%input : tensor<1x4x16x1xi32>, %shape: tensor<2x2xi32>, %output: tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32> {
204 %0 = linalg.pooling_nhwc_max {dilations = dense<[1, 2]> : tensor<2xi64>, strides = dense<[2, 4]> : tensor<2xi64>}
205 ins(%input, %shape : tensor<1x4x16x1xi32>, tensor<2x2xi32>) outs(%output : tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32>
206 return %0: tensor<1x2x4x1xi32>
215 func.func @generalize_pooling_nwc_max_i32(%input : tensor<1x16x1xi32>, %shape: tensor<2xi32>, %output: tensor<1x4x1xi32>) -> tensor<1x4x1xi32> {
216 %0 = linalg.pooling_nwc_max {dilations = dense<[2]> : tensor<1xi64>, strides = dense<[4]> : tensor<1xi64>}
217 ins(%input, %shape : tensor<1x16x1xi32>, tensor<2xi32>) outs(%output : tensor<1x4x1xi32>) -> tensor<1x4x1xi32>
218 return %0: tensor<1x4x1xi32>
227 func.func @generalize_pooling_nhwc_max_unsigned_i32(%input : tensor<1x4x16x1xi32>, %shape: tensor<2x2xi32>, %output: tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32> {
228 %0 = linalg.pooling_nhwc_max_unsigned {dilations = dense<[1, 2]> : tensor<2xi64>, strides = dense<[2, 4]> : tensor<2xi64>}
229 ins(%input, %shape : tensor<1x4x16x1xi32>, tensor<2x2xi32>) outs(%output : tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32>
230 return %0: tensor<1x2x4x1xi32>
239 func.func @generalize_pooling_nwc_max_unsigned_i32(%input : tensor<1x16x1xi32>, %shape: tensor<2xi32>, %output: tensor<1x4x1xi32>) -> tensor<1x4x1xi32> {
240 %0 = linalg.pooling_nwc_max_unsigned {dilations = dense<[2]> : tensor<1xi64>, strides = dense<[4]> : tensor<1xi64>}
241 ins(%input, %shape : tensor<1x16x1xi32>, tensor<2xi32>) outs(%output : tensor<1x4x1xi32>) -> tensor<1x4x1xi32>
242 return %0: tensor<1x4x1xi32>
251 func.func @generalize_pooling_nhwc_min_f32(%input : tensor<1x4x16x1xf32>, %shape: tensor<2x2xf32>, %output: tensor<1x2x4x1xf32>) -> tensor<1x2x4x1xf32> {
252 %0 = linalg.pooling_nhwc_min {dilations = dense<[1, 2]> : tensor<2xi64>, strides = dense<[2, 4]> : tensor<2xi64>}
253 ins(%input, %shape : tensor<1x4x16x1xf32>, tensor<2x2xf32>) outs(%output : tensor<1x2x4x1xf32>) -> tensor<1x2x4x1xf32>
254 return %0: tensor<1x2x4x1xf32>
261 // CHECK-NEXT: -> tensor<1x2x4x1xf32>
265 func.func @generalize_pooling_nwc_min_f32(%input : tensor<1x16x1xf32>, %shape: tensor<2xf32>, %output: tensor<1x4x1xf32>) -> tensor<1x4x1xf32> {
266 %0 = linalg.pooling_nwc_min {dilations = dense<[2]> : tensor<1xi64>, strides = dense<[4]> : tensor<1xi64>}
267 ins(%input, %shape : tensor<1x16x1xf32>, tensor<2xf32>) outs(%output : tensor<1x4x1xf32>) -> tensor<1x4x1xf32>
268 return %0: tensor<1x4x1xf32>
275 // CHECK-NEXT: -> tensor<1x4x1xf32>
279 func.func @generalize_pooling_nhwc_min_i32(%input : tensor<1x4x16x1xi32>, %shape: tensor<2x2xi32>, %output: tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32> {
280 %0 = linalg.pooling_nhwc_min {dilations = dense<[1, 2]> : tensor<2xi64>, strides = dense<[2, 4]> : tensor<2xi64>}
281 ins(%input, %shape : tensor<1x4x16x1xi32>, tensor<2x2xi32>) outs(%output : tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32>
282 return %0: tensor<1x2x4x1xi32>
291 func.func @generalize_pooling_nwc_min_i32(%input : tensor<1x16x1xi32>, %shape: tensor<2xi32>, %output: tensor<1x4x1xi32>) -> tensor<1x4x1xi32> {
292 %0 = linalg.pooling_nwc_min {dilations = dense<[2]> : tensor<1xi64>, strides = dense<[4]> : tensor<1xi64>}
293 ins(%input, %shape : tensor<1x16x1xi32>, tensor<2xi32>) outs(%output : tensor<1x4x1xi32>) -> tensor<1x4x1xi32>
294 return %0: tensor<1x4x1xi32>
303 func.func @generalize_pooling_nhwc_min_unsigned_i32(%input : tensor<1x4x16x1xi32>, %shape: tensor<2x2xi32>, %output: tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32> {
304 %0 = linalg.pooling_nhwc_min_unsigned {dilations = dense<[1, 2]> : tensor<2xi64>, strides = dense<[2, 4]> : tensor<2xi64>}
305 ins(%input, %shape : tensor<1x4x16x1xi32>, tensor<2x2xi32>) outs(%output : tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32>
306 return %0: tensor<1x2x4x1xi32>
315 func.func @generalize_pooling_nwc_min_unsigned_i32(%input : tensor<1x16x1xi32>, %shape: tensor<2xi32>, %output: tensor<1x4x1xi32>) -> tensor<1x4x1xi32> {
316 %0 = linalg.pooling_nwc_min_unsigned {dilations = dense<[2]> : tensor<1xi64>, strides = dense<[4]> : tensor<1xi64>}
317 ins(%input, %shape : tensor<1x16x1xi32>, tensor<2xi32>) outs(%output : tensor<1x4x1xi32>) -> tensor<1x4x1xi32>
318 return %0: tensor<1x4x1xi32>
327 func.func @generalize_pooling_nhwc_sum_f32(%input : tensor<1x4x16x1xf32>, %shape: tensor<2x2xf32>, %output: tensor<1x2x4x1xf32>) -> tensor<1x2x4x1xf32> {
328 %0 = linalg.pooling_nhwc_sum {dilations = dense<[1, 2]> : tensor<2xi64>, strides = dense<[2, 4]> : tensor<2xi64>}
329 ins(%input, %shape : tensor<1x4x16x1xf32>, tensor<2x2xf32>) outs(%output : tensor<1x2x4x1xf32>) -> tensor<1x2x4x1xf32>
330 return %0: tensor<1x2x4x1xf32>
337 // CHECK-NEXT: -> tensor<1x2x4x1xf32>
341 func.func @generalize_pooling_nwc_sum_f32(%input : tensor<1x16x1xf32>, %shape: tensor<2xf32>, %output: tensor<1x4x1xf32>) -> tensor<1x4x1xf32> {
342 %0 = linalg.pooling_nwc_sum {dilations = dense<[2]> : tensor<1xi64>, strides = dense<[4]> : tensor<1xi64>}
343 ins(%input, %shape : tensor<1x16x1xf32>, tensor<2xf32>) outs(%output : tensor<1x4x1xf32>) -> tensor<1x4x1xf32>
344 return %0: tensor<1x4x1xf32>
351 // CHECK-NEXT: -> tensor<1x4x1xf32>
355 func.func @generalize_pooling_nhwc_sum_i32(%input : tensor<1x4x16x1xi32>, %shape: tensor<2x2xi32>, %output: tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32> {
356 %0 = linalg.pooling_nhwc_sum {dilations = dense<[1, 2]> : tensor<2xi64>, strides = dense<[2, 4]> : tensor<2xi64>}
357 ins(%input, %shape : tensor<1x4x16x1xi32>, tensor<2x2xi32>) outs(%output : tensor<1x2x4x1xi32>) -> tensor<1x2x4x1xi32>
358 return %0: tensor<1x2x4x1xi32>
365 // CHECK-NEXT: -> tensor<1x2x4x1xi32>
369 func.func @generalize_pooling_nwc_sum_i32(%input : tensor<1x16x1xi32>, %shape: tensor<2xi32>, %output: tensor<1x4x1xi32>) -> tensor<1x4x1xi32> {
370 %0 = linalg.pooling_nwc_sum {dilations = dense<[2]> : tensor<1xi64>, strides = dense<[4]> : tensor<1xi64>}
371 ins(%input, %shape : tensor<1x16x1xi32>, tensor<2xi32>) outs(%output : tensor<1x4x1xi32>) -> tensor<1x4x1xi32>
372 return %0: tensor<1x4x1xi32>
379 // CHECK-NEXT: -> tensor<1x4x1xi32>
383 func.func @generalize_fill_0d(%value: f64, %O: tensor<f32>) -> tensor<f32> {
384 %0 = linalg.fill ins(%value: f64) outs(%O : tensor<f32>) -> tensor<f32>
385 return %0: tensor<f32>
412 func.func @generalize_index(%min: f64, %max: f64, %seed: i32, %O: tensor<16x32xf32>) -> tensor<16x32xf32> {
413 %0 = linalg.fill_rng_2d ins(%min, %max, %seed: f64, f64, i32) outs(%O : tensor<16x32xf32>) -> tensor<16x32xf32>
414 return %0: tensor<16x32xf32>
425 func.func @generalize_const(%min: f64, %max: f64, %seed: i32, %O: tensor<16x32xf32>) -> tensor<16x32xf32> {
426 %0 = linalg.fill_rng_2d ins(%min, %max, %seed: f64, f64, i32) outs(%O : tensor<16x32xf32>) -> tensor<16x32xf32>
427 return %0: tensor<16x32xf32>
438 func.func @generalize_elemwise_exp(%lhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
439 %0 = linalg.elemwise_unary ins(%lhs: tensor<4x8xf32>) outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
440 return %0: tensor<4x8xf32>
449 func.func @generalize_elemwise_log(%lhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
451 ins(%lhs: tensor<4x8xf32>) outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
452 return %0: tensor<4x8xf32>
461 func.func @generalize_elemwise_abs(%lhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
463 ins(%lhs: tensor<4x8xf32>) outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
464 return %0: tensor<4x8xf32>
473 func.func @generalize_elemwise_ceil(%lhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
475 ins(%lhs: tensor<4x8xf32>) outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
476 return %0: tensor<4x8xf32>
485 func.func @generalize_elemwise_floor(%lhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
487 ins(%lhs: tensor<4x8xf32>) outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
488 return %0: tensor<4x8xf32>
497 func.func @generalize_elemwise_negf(%lhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
499 ins(%lhs: tensor<4x8xf32>) outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
500 return %0: tensor<4x8xf32>
509 func.func @generalize_elemwise_add(%lhs : tensor<4x8xf32>, %rhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
510 %0 = linalg.elemwise_binary ins(%lhs, %rhs: tensor<4x8xf32>, tensor<4x8xf32>)
511 outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
512 return %0: tensor<4x8xf32>
521 func.func @generalize_elemwise_mul(%lhs : tensor<4x8xf32>, %rhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
523 ins(%lhs, %rhs: tensor<4x8xf32>, tensor<4x8xf32>)
524 outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
525 return %0: tensor<4x8xf32>
534 func.func @generalize_elemwise_rank_zero(%lhs : tensor<f32>, %rhs : tensor<f32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
536 ins(%lhs, %rhs: tensor<f32>, tensor<f32>)
537 outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
538 return %0: tensor<4x8xf32>
549 func.func @generalize_copy(%lhs : tensor<4x8xf32>, %output : tensor<4x8xf32>) -> tensor<4x8xf32> {
550 %0 = linalg.copy ins(%lhs: tensor<4x8xf32>) outs(%output: tensor<4x8xf32>) -> tensor<4x8xf32>
551 return %0: tensor<4x8xf32>