cxx11_tensor_convolution_sycl.cpp
Go to the documentation of this file.
1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2016
5 // Mehdi Goli Codeplay Software Ltd.
6 // Ralph Potter Codeplay Software Ltd.
7 // Luke Iwanski Codeplay Software Ltd.
8 // Contact: <eigen@codeplay.com>
9 //
10 // This Source Code Form is subject to the terms of the Mozilla
11 // Public License v. 2.0. If a copy of the MPL was not distributed
12 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
13 
14 #define EIGEN_TEST_NO_LONGDOUBLE
15 #define EIGEN_TEST_NO_COMPLEX
16 
17 #define EIGEN_DEFAULT_DENSE_INDEX_TYPE int64_t
18 #define EIGEN_USE_SYCL
19 
20 #include <iostream>
21 #include <chrono>
22 #include <ctime>
23 
24 #include "main.h"
25 #include <unsupported/Eigen/CXX11/Tensor>
26 #include <iomanip>
27 
28 using Eigen::array;
29 using Eigen::SyclDevice;
30 using Eigen::Tensor;
31 using Eigen::TensorMap;
32 static const float error_threshold =1e-4f;
33 
34 
35 template <typename DataType, int DataLayout, typename IndexType>
36 static void test_larg_expr1D(const Eigen::SyclDevice& sycl_device)
37 {
38  IndexType indim0 =53;
39  IndexType indim1= 55;
40  IndexType indim2= 51;
41  IndexType outdim0=50;
42  IndexType outdim1=55;
43  IndexType outdim2=51;
44  Eigen::array<IndexType, 3> input_dims = {{indim0, indim1, indim2}};
45  Eigen::array<IndexType, 1> kernel_dims = {{4}};
46  Eigen::array<IndexType, 3> result_dims = {{outdim0, outdim1, outdim2}};
47 
48  Tensor<DataType, 3, DataLayout, IndexType> input(input_dims);
49  Tensor<DataType, 1, DataLayout,IndexType> kernel(kernel_dims);
50  Tensor<DataType, 3, DataLayout,IndexType> result(result_dims);
51  Tensor<DataType, 3, DataLayout,IndexType> result_host(result_dims);
52 
53  Eigen::array<IndexType, 1> dims3{{0}};
54 
55  input.setRandom();
56  kernel.setRandom();
57  result.setZero();
58  result_host.setZero();
59 
60  std::size_t input_bytes = input.size() * sizeof(DataType);
61  std::size_t kernel_bytes = kernel.size() * sizeof(DataType);
62  std::size_t result_bytes = result.size() * sizeof(DataType);
63 
64  DataType * d_input = static_cast<DataType*>(sycl_device.allocate(input_bytes));
65  DataType * d_kernel = static_cast<DataType*>(sycl_device.allocate(kernel_bytes));
66  DataType * d_result = static_cast<DataType*>(sycl_device.allocate(result_bytes));
67 
68  Eigen::TensorMap<Eigen::Tensor<DataType, 3, DataLayout, IndexType> > gpu_input(d_input, input_dims);
69  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout, IndexType> > gpu_kernel(d_kernel, kernel_dims);
70  Eigen::TensorMap<Eigen::Tensor<DataType, 3, DataLayout, IndexType> > gpu_result(d_result, result_dims);
71  sycl_device.memcpyHostToDevice(d_input, input.data(), input_bytes);
72  sycl_device.memcpyHostToDevice(d_kernel, kernel.data(), kernel_bytes);
73 
74  gpu_result.device(sycl_device)=gpu_input.convolve(gpu_kernel, dims3);
75  sycl_device.memcpyDeviceToHost(result.data(), d_result, result_bytes);
76 
77  result_host=input.convolve(kernel, dims3);
78 
79 for(IndexType i=0; i< outdim0; i++ ){
80  for(IndexType j=0; j< outdim1; j++ ){
81  for(IndexType k=0; k< outdim2; k++ ){
82  if (!(Eigen::internal::isApprox(result(i,j,k), result_host(i,j,k), error_threshold))) {
83  std::cout <<std::setprecision(16)<< "mismatch detected at index ( "<< i << " , " << j << ", " << k << " ) " << " \t " << result(i,j,k) << " vs "<< result_host(i,j,k) << std::endl;
84  assert(false);
85  }
86  }
87  }
88 }
89  sycl_device.deallocate(d_input);
90  sycl_device.deallocate(d_kernel);
91  sycl_device.deallocate(d_result);
92 
93 }
94 
95 
96 template <typename DataType, int DataLayout, typename IndexType>
97 static void test_larg_expr2D(const Eigen::SyclDevice& sycl_device)
98 {
99  IndexType indim0 =53;
100  IndexType indim1= 55;
101  IndexType indim2= 51;
102  IndexType outdim0=50;
103  IndexType outdim1=51;
104  IndexType outdim2=51;
105  Eigen::array<IndexType, 3> input_dims = {{indim0, indim1, indim2}};
106  Eigen::array<IndexType, 2> kernel_dims = {{4,5}};
107  Eigen::array<IndexType, 3> result_dims = {{outdim0, outdim1, outdim2}};
108 
109  Tensor<DataType, 3, DataLayout, IndexType> input(input_dims);
110  Tensor<DataType, 2, DataLayout,IndexType> kernel(kernel_dims);
111  Tensor<DataType, 3, DataLayout,IndexType> result(result_dims);
112  Tensor<DataType, 3, DataLayout,IndexType> result_host(result_dims);
113 
114  Eigen::array<IndexType, 2> dims3{{0,1}};
115 
116  input.setRandom();
117  kernel.setRandom();
118  result.setZero();
119  result_host.setZero();
120 
121  std::size_t input_bytes = input.size() * sizeof(DataType);
122  std::size_t kernel_bytes = kernel.size() * sizeof(DataType);
123  std::size_t result_bytes = result.size() * sizeof(DataType);
124 
125  DataType * d_input = static_cast<DataType*>(sycl_device.allocate(input_bytes));
126  DataType * d_kernel = static_cast<DataType*>(sycl_device.allocate(kernel_bytes));
127  DataType * d_result = static_cast<DataType*>(sycl_device.allocate(result_bytes));
128 
129  Eigen::TensorMap<Eigen::Tensor<DataType, 3, DataLayout, IndexType> > gpu_input(d_input, input_dims);
130  Eigen::TensorMap<Eigen::Tensor<DataType, 2, DataLayout, IndexType> > gpu_kernel(d_kernel, kernel_dims);
131  Eigen::TensorMap<Eigen::Tensor<DataType, 3, DataLayout, IndexType> > gpu_result(d_result, result_dims);
132  sycl_device.memcpyHostToDevice(d_input, input.data(), input_bytes);
133  sycl_device.memcpyHostToDevice(d_kernel, kernel.data(), kernel_bytes);
134 
135  gpu_result.device(sycl_device)=gpu_input.convolve(gpu_kernel, dims3);
136  sycl_device.memcpyDeviceToHost(result.data(), d_result, result_bytes);
137 
138  result_host=input.convolve(kernel, dims3);
139 
140 for(IndexType i=0; i< outdim0; i++ ){
141  for(IndexType j=0; j< outdim1; j++ ){
142  for(IndexType k=0; k< outdim2; k++ ){
143  if (!(Eigen::internal::isApprox(result(i,j,k), result_host(i,j,k), error_threshold))) {
144  std::cout <<std::setprecision(16)<< "mismatch detected at index ( "<< i << " , " << j << ", " << k << " ) " << " \t " << result(i,j,k) << " vs "<< result_host(i,j,k) << std::endl;
145  assert(false);
146  }
147  }
148  }
149 }
150  sycl_device.deallocate(d_input);
151  sycl_device.deallocate(d_kernel);
152  sycl_device.deallocate(d_result);
153 
154 }
155 
156 
157 template <typename DataType, int DataLayout, typename IndexType>
158 static void test_larg_expr3D(const Eigen::SyclDevice& sycl_device)
159 {
160  IndexType indim0 =53;
161  IndexType indim1= 55;
162  IndexType indim2= 51;
163  IndexType outdim0=50;
164  IndexType outdim1=51;
165  IndexType outdim2=49;
166  Eigen::array<IndexType, 3> input_dims = {{indim0, indim1, indim2}};
167  Eigen::array<IndexType, 3> kernel_dims = {{4,5,3}};
168  Eigen::array<IndexType, 3> result_dims = {{outdim0, outdim1, outdim2}};
169 
170  Tensor<DataType, 3, DataLayout, IndexType> input(input_dims);
171  Tensor<DataType, 3, DataLayout,IndexType> kernel(kernel_dims);
172  Tensor<DataType, 3, DataLayout,IndexType> result(result_dims);
173  Tensor<DataType, 3, DataLayout,IndexType> result_host(result_dims);
174 
175  Eigen::array<IndexType, 3> dims3{{0,1,2}};
176 
177  input.setRandom();
178  kernel.setRandom();
179  result.setZero();
180  result_host.setZero();
181 
182  std::size_t input_bytes = input.size() * sizeof(DataType);
183  std::size_t kernel_bytes = kernel.size() * sizeof(DataType);
184  std::size_t result_bytes = result.size() * sizeof(DataType);
185 
186  DataType * d_input = static_cast<DataType*>(sycl_device.allocate(input_bytes));
187  DataType * d_kernel = static_cast<DataType*>(sycl_device.allocate(kernel_bytes));
188  DataType * d_result = static_cast<DataType*>(sycl_device.allocate(result_bytes));
189 
190  Eigen::TensorMap<Eigen::Tensor<DataType, 3, DataLayout, IndexType> > gpu_input(d_input, input_dims);
191  Eigen::TensorMap<Eigen::Tensor<DataType, 3, DataLayout, IndexType> > gpu_kernel(d_kernel, kernel_dims);
192  Eigen::TensorMap<Eigen::Tensor<DataType, 3, DataLayout, IndexType> > gpu_result(d_result, result_dims);
193  sycl_device.memcpyHostToDevice(d_input, input.data(), input_bytes);
194  sycl_device.memcpyHostToDevice(d_kernel, kernel.data(), kernel_bytes);
195 
196  gpu_result.device(sycl_device)=gpu_input.convolve(gpu_kernel, dims3);
197  sycl_device.memcpyDeviceToHost(result.data(), d_result, result_bytes);
198 
199  result_host=input.convolve(kernel, dims3);
200 
201 for(IndexType i=0; i< outdim0; i++ ){
202  for(IndexType j=0; j< outdim1; j++ ){
203  for(IndexType k=0; k< outdim2; k++ ){
204  if (!(Eigen::internal::isApprox(result(i,j,k), result_host(i,j,k), error_threshold))) {
205  std::cout <<std::setprecision(16)<< "mismatch detected at index ( "<< i << " , " << j << ", " << k << " ) " << " \t " << result(i,j,k) << " vs "<< result_host(i,j,k) << std::endl;
206  assert(false);
207  }
208  }
209  }
210 }
211  sycl_device.deallocate(d_input);
212  sycl_device.deallocate(d_kernel);
213  sycl_device.deallocate(d_result);
214 
215 }
216 
217 
218 template <typename DataType, int DataLayout, typename IndexType>
219 static void test_evals(const Eigen::SyclDevice& sycl_device)
220 {
221  Eigen::array<IndexType, 2> input_dims = {{3, 3}};
222  Eigen::array<IndexType, 1> kernel_dims = {{2}};
223  Eigen::array<IndexType, 2> result_dims = {{2, 3}};
224 
225  Tensor<DataType, 2, DataLayout, IndexType> input(input_dims);
226  Tensor<DataType, 1, DataLayout,IndexType> kernel(kernel_dims);
227  Tensor<DataType, 2, DataLayout,IndexType> result(result_dims);
228 
229  Eigen::array<IndexType, 1> dims3{{0}};
230 
231  input.setRandom();
232  kernel.setRandom();
233  result.setZero();
234 
235  std::size_t input_bytes = input.size() * sizeof(DataType);
236  std::size_t kernel_bytes = kernel.size() * sizeof(DataType);
237  std::size_t result_bytes = result.size() * sizeof(DataType);
238 
239  DataType * d_input = static_cast<DataType*>(sycl_device.allocate(input_bytes));
240  DataType * d_kernel = static_cast<DataType*>(sycl_device.allocate(kernel_bytes));
241  DataType * d_result = static_cast<DataType*>(sycl_device.allocate(result_bytes));
242 
243  Eigen::TensorMap<Eigen::Tensor<DataType, 2, DataLayout, IndexType> > gpu_input(d_input, input_dims);
244  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout, IndexType> > gpu_kernel(d_kernel, kernel_dims);
245  Eigen::TensorMap<Eigen::Tensor<DataType, 2, DataLayout, IndexType> > gpu_result(d_result, result_dims);
246  sycl_device.memcpyHostToDevice(d_input, input.data(), input_bytes);
247  sycl_device.memcpyHostToDevice(d_kernel, kernel.data(), kernel_bytes);
248 
249  gpu_result.device(sycl_device)=gpu_input.convolve(gpu_kernel, dims3);
250  sycl_device.memcpyDeviceToHost(result.data(), d_result, result_bytes);
251 
252  VERIFY_IS_APPROX(result(0,0), input(0,0)*kernel(0) + input(1,0)*kernel(1)); // index 0
253  VERIFY_IS_APPROX(result(0,1), input(0,1)*kernel(0) + input(1,1)*kernel(1)); // index 2
254  VERIFY_IS_APPROX(result(0,2), input(0,2)*kernel(0) + input(1,2)*kernel(1)); // index 4
255  VERIFY_IS_APPROX(result(1,0), input(1,0)*kernel(0) + input(2,0)*kernel(1)); // index 1
256  VERIFY_IS_APPROX(result(1,1), input(1,1)*kernel(0) + input(2,1)*kernel(1)); // index 3
257  VERIFY_IS_APPROX(result(1,2), input(1,2)*kernel(0) + input(2,2)*kernel(1)); // index 5
258 
259  sycl_device.deallocate(d_input);
260  sycl_device.deallocate(d_kernel);
261  sycl_device.deallocate(d_result);
262 }
263 
264 template <typename DataType, int DataLayout, typename IndexType>
265 static void test_expr(const Eigen::SyclDevice& sycl_device)
266 {
267  Eigen::array<IndexType, 2> input_dims = {{3, 3}};
268  Eigen::array<IndexType, 2> kernel_dims = {{2, 2}};
269  Eigen::array<IndexType, 2> result_dims = {{2, 2}};
270 
271  Tensor<DataType, 2, DataLayout, IndexType> input(input_dims);
272  Tensor<DataType, 2, DataLayout, IndexType> kernel(kernel_dims);
273  Tensor<DataType, 2, DataLayout, IndexType> result(result_dims);
274 
275  input.setRandom();
276  kernel.setRandom();
277  Eigen::array<IndexType, 2> dims;
278  dims[0] = 0;
279  dims[1] = 1;
280 
281  std::size_t input_bytes = input.size() * sizeof(DataType);
282  std::size_t kernel_bytes = kernel.size() * sizeof(DataType);
283  std::size_t result_bytes = result.size() * sizeof(DataType);
284 
285  DataType * d_input = static_cast<DataType*>(sycl_device.allocate(input_bytes));
286  DataType * d_kernel = static_cast<DataType*>(sycl_device.allocate(kernel_bytes));
287  DataType * d_result = static_cast<DataType*>(sycl_device.allocate(result_bytes));
288 
289  Eigen::TensorMap<Eigen::Tensor<DataType, 2, DataLayout,IndexType> > gpu_input(d_input, input_dims);
290  Eigen::TensorMap<Eigen::Tensor<DataType, 2, DataLayout,IndexType> > gpu_kernel(d_kernel, kernel_dims);
291  Eigen::TensorMap<Eigen::Tensor<DataType, 2, DataLayout,IndexType> > gpu_result(d_result, result_dims);
292  sycl_device.memcpyHostToDevice(d_input, input.data(), input_bytes);
293  sycl_device.memcpyHostToDevice(d_kernel, kernel.data(), kernel_bytes);
294 
295  gpu_result.device(sycl_device)=gpu_input.convolve(gpu_kernel, dims);
296  sycl_device.memcpyDeviceToHost(result.data(), d_result, result_bytes);
297 
298  VERIFY_IS_APPROX(result(0,0), input(0,0)*kernel(0,0) + input(0,1)*kernel(0,1) +
299  input(1,0)*kernel(1,0) + input(1,1)*kernel(1,1));
300  VERIFY_IS_APPROX(result(0,1), input(0,1)*kernel(0,0) + input(0,2)*kernel(0,1) +
301  input(1,1)*kernel(1,0) + input(1,2)*kernel(1,1));
302  VERIFY_IS_APPROX(result(1,0), input(1,0)*kernel(0,0) + input(1,1)*kernel(0,1) +
303  input(2,0)*kernel(1,0) + input(2,1)*kernel(1,1));
304  VERIFY_IS_APPROX(result(1,1), input(1,1)*kernel(0,0) + input(1,2)*kernel(0,1) +
305  input(2,1)*kernel(1,0) + input(2,2)*kernel(1,1));
306 
307  sycl_device.deallocate(d_input);
308  sycl_device.deallocate(d_kernel);
309  sycl_device.deallocate(d_result);
310 }
311 
312 
313 template <typename DataType, int DataLayout, typename IndexType>
314 static void test_modes(const Eigen::SyclDevice& sycl_device){
315 
316 Eigen::array<IndexType, 1> input_dims = {{3}};
317 Eigen::array<IndexType, 1> kernel_dims = {{3}};
318 
319 Tensor<DataType, 1, DataLayout, IndexType> input(input_dims);
320 Tensor<DataType, 1, DataLayout, IndexType> kernel(kernel_dims);
321 
322 input.setRandom();
323 kernel.setRandom();
324 Eigen::array<IndexType, 1> dims;
325 dims[0] = 0;
326 
327  input(0) = 1.0f;
328  input(1) = 2.0f;
329  input(2) = 3.0f;
330  kernel(0) = 0.5f;
331  kernel(1) = 1.0f;
332  kernel(2) = 0.0f;
333 
334  Eigen::array<std::pair<IndexType, IndexType>, 1> padding;
335 
336  // Emulate VALID mode (as defined in
337  // http://docs.scipy.org/doc/numpy/reference/generated/numpy.convolve.html).
338  padding[0] = std::make_pair(0, 0);
339  Tensor<DataType, 1, DataLayout, IndexType> valid(1);
340 
341  std::size_t input_bytes = input.size() * sizeof(DataType);
342  std::size_t kernel_bytes = kernel.size() * sizeof(DataType);
343  std::size_t valid_bytes = valid.size() * sizeof(DataType);
344 
345  DataType * d_input = static_cast<DataType*>(sycl_device.allocate(input_bytes));
346  DataType * d_kernel = static_cast<DataType*>(sycl_device.allocate(kernel_bytes));
347  DataType * d_valid = static_cast<DataType*>(sycl_device.allocate(valid_bytes));
348 
349  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout,IndexType> > gpu_input(d_input, input_dims);
350  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout,IndexType> > gpu_kernel(d_kernel, kernel_dims);
351  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout,IndexType> > gpu_valid(d_valid, valid.dimensions());
352  sycl_device.memcpyHostToDevice(d_input, input.data(), input_bytes);
353  sycl_device.memcpyHostToDevice(d_kernel, kernel.data(), kernel_bytes);
354 
355  gpu_valid.device(sycl_device)=gpu_input.pad(padding).convolve(gpu_kernel, dims);
356  sycl_device.memcpyDeviceToHost(valid.data(), d_valid, valid_bytes);
357 
358  VERIFY_IS_EQUAL(valid.dimension(0), 1);
359  VERIFY_IS_APPROX(valid(0), 2.5f);
360 
361  // Emulate SAME mode (as defined in
362  // http://docs.scipy.org/doc/numpy/reference/generated/numpy.convolve.html).
363  padding[0] = std::make_pair(1, 1);
364  Tensor<DataType, 1, DataLayout, IndexType> same(3);
365  std::size_t same_bytes = same.size() * sizeof(DataType);
366  DataType * d_same = static_cast<DataType*>(sycl_device.allocate(same_bytes));
367  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout,IndexType> > gpu_same(d_same, same.dimensions());
368  gpu_same.device(sycl_device)=gpu_input.pad(padding).convolve(gpu_kernel, dims);
369  sycl_device.memcpyDeviceToHost(same.data(), d_same, same_bytes);
370 
371  VERIFY_IS_EQUAL(same.dimension(0), 3);
372  VERIFY_IS_APPROX(same(0), 1.0f);
373  VERIFY_IS_APPROX(same(1), 2.5f);
374  VERIFY_IS_APPROX(same(2), 4.0f);
375 
376  // Emulate FULL mode (as defined in
377  // http://docs.scipy.org/doc/numpy/reference/generated/numpy.convolve.html).
378  padding[0] = std::make_pair(2, 2);
379 
380  Tensor<DataType, 1, DataLayout, IndexType> full(5);
381  std::size_t full_bytes = full.size() * sizeof(DataType);
382  DataType * d_full = static_cast<DataType*>(sycl_device.allocate(full_bytes));
383  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout,IndexType> > gpu_full(d_full, full.dimensions());
384  gpu_full.device(sycl_device)=gpu_input.pad(padding).convolve(gpu_kernel, dims);
385  sycl_device.memcpyDeviceToHost(full.data(), d_full, full_bytes);
386 
387  VERIFY_IS_EQUAL(full.dimension(0), 5);
388  VERIFY_IS_APPROX(full(0), 0.0f);
389  VERIFY_IS_APPROX(full(1), 1.0f);
390  VERIFY_IS_APPROX(full(2), 2.5f);
391  VERIFY_IS_APPROX(full(3), 4.0f);
392  VERIFY_IS_APPROX(full(4), 1.5f);
393 
394  sycl_device.deallocate(d_input);
395  sycl_device.deallocate(d_kernel);
396  sycl_device.deallocate(d_valid);
397  sycl_device.deallocate(d_same);
398  sycl_device.deallocate(d_full);
399 
400 }
401 
402 template <typename DataType, int DataLayout, typename IndexType>
403 static void test_strides(const Eigen::SyclDevice& sycl_device){
404 
405  Eigen::array<IndexType, 1> input_dims = {{13}};
406  Eigen::array<IndexType, 1> kernel_dims = {{3}};
407 
408  Tensor<DataType, 1, DataLayout, IndexType> input(input_dims);
409  Tensor<DataType, 1, DataLayout, IndexType> kernel(kernel_dims);
410  Tensor<DataType, 1, DataLayout, IndexType> result(2);
411 
412  input.setRandom();
413  kernel.setRandom();
414  Eigen::array<IndexType, 1> dims;
415  dims[0] = 0;
416 
417  Eigen::array<IndexType, 1> stride_of_3;
418  stride_of_3[0] = 3;
419  Eigen::array<IndexType, 1> stride_of_2;
420  stride_of_2[0] = 2;
421 
422  std::size_t input_bytes = input.size() * sizeof(DataType);
423  std::size_t kernel_bytes = kernel.size() * sizeof(DataType);
424  std::size_t result_bytes = result.size() * sizeof(DataType);
425 
426  DataType * d_input = static_cast<DataType*>(sycl_device.allocate(input_bytes));
427  DataType * d_kernel = static_cast<DataType*>(sycl_device.allocate(kernel_bytes));
428  DataType * d_result = static_cast<DataType*>(sycl_device.allocate(result_bytes));
429 
430  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout,IndexType> > gpu_input(d_input, input_dims);
431  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout,IndexType> > gpu_kernel(d_kernel, kernel_dims);
432  Eigen::TensorMap<Eigen::Tensor<DataType, 1, DataLayout,IndexType> > gpu_result(d_result, result.dimensions());
433  sycl_device.memcpyHostToDevice(d_input, input.data(), input_bytes);
434  sycl_device.memcpyHostToDevice(d_kernel, kernel.data(), kernel_bytes);
435 
436  gpu_result.device(sycl_device)=gpu_input.stride(stride_of_3).convolve(gpu_kernel, dims).stride(stride_of_2);
437  sycl_device.memcpyDeviceToHost(result.data(), d_result, result_bytes);
438 
439  VERIFY_IS_EQUAL(result.dimension(0), 2);
440  VERIFY_IS_APPROX(result(0), (input(0)*kernel(0) + input(3)*kernel(1) +
441  input(6)*kernel(2)));
442  VERIFY_IS_APPROX(result(1), (input(6)*kernel(0) + input(9)*kernel(1) +
443  input(12)*kernel(2)));
444 }
445 
446 template <typename Dev_selector> void tensorConvolutionPerDevice(Dev_selector& s){
447  QueueInterface queueInterface(s);
448  auto sycl_device=Eigen::SyclDevice(&queueInterface);
449  test_larg_expr1D<float, RowMajor, int64_t>(sycl_device);
450  test_larg_expr1D<float, ColMajor, int64_t>(sycl_device);
451  test_larg_expr2D<float, RowMajor, int64_t>(sycl_device);
452  test_larg_expr2D<float, ColMajor, int64_t>(sycl_device);
453  test_larg_expr3D<float, RowMajor, int64_t>(sycl_device);
454  test_larg_expr3D<float, ColMajor, int64_t>(sycl_device);
455  test_evals<float, ColMajor, int64_t>(sycl_device);
456  test_evals<float, RowMajor, int64_t>(sycl_device);
457  test_expr<float, ColMajor, int64_t>(sycl_device);
458  test_expr<float, RowMajor, int64_t>(sycl_device);
459  test_modes<float, ColMajor, int64_t>(sycl_device);
460  test_modes<float, RowMajor, int64_t>(sycl_device);
461  test_strides<float, ColMajor, int64_t>(sycl_device);
462  test_strides<float, RowMajor, int64_t>(sycl_device);
463 }
464 
465 EIGEN_DECLARE_TEST(cxx11_tensor_convolution_sycl) {
466  for (const auto& device :Eigen::get_sycl_supported_devices()) {
467  CALL_SUBTEST(tensorConvolutionPerDevice(device));
468  }
469 }
Eigen::Tensor::dimension
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index dimension(std::size_t n) const
Definition: Tensor.h:101
Eigen::Tensor
The tensor class.
Definition: Tensor.h:63
array
int array[24]
Definition: Map_general_stride.cpp:1
s
RealScalar s
Definition: level1_cplx_impl.h:126
e
Array< double, 1, 3 > e(1./3., 0.5, 2.)
test_larg_expr2D
static void test_larg_expr2D(const Eigen::SyclDevice &sycl_device)
Definition: cxx11_tensor_convolution_sycl.cpp:97
Eigen::array
Definition: EmulateArray.h:21
VERIFY_IS_EQUAL
#define VERIFY_IS_EQUAL(a, b)
Definition: main.h:386
test_expr
static void test_expr(const Eigen::SyclDevice &sycl_device)
Definition: cxx11_tensor_convolution_sycl.cpp:265
Eigen::internal::isApprox
EIGEN_DEVICE_FUNC bool isApprox(const Scalar &x, const Scalar &y, const typename NumTraits< Scalar >::Real &precision=NumTraits< Scalar >::dummy_precision())
Definition: Eigen/src/Core/MathFunctions.h:1947
EIGEN_DECLARE_TEST
EIGEN_DECLARE_TEST(cxx11_tensor_convolution_sycl)
Definition: cxx11_tensor_convolution_sycl.cpp:465
tensorConvolutionPerDevice
void tensorConvolutionPerDevice(Dev_selector &s)
Definition: cxx11_tensor_convolution_sycl.cpp:446
result
Values result
Definition: OdometryOptimize.cpp:8
test_larg_expr3D
static void test_larg_expr3D(const Eigen::SyclDevice &sycl_device)
Definition: cxx11_tensor_convolution_sycl.cpp:158
test_larg_expr1D
static void test_larg_expr1D(const Eigen::SyclDevice &sycl_device)
Definition: cxx11_tensor_convolution_sycl.cpp:36
j
std::ptrdiff_t j
Definition: tut_arithmetic_redux_minmax.cpp:2
Eigen::Tensor::dimensions
EIGEN_DEVICE_FUNC const EIGEN_STRONG_INLINE Dimensions & dimensions() const
Definition: Tensor.h:102
Eigen::TensorMap
A tensor expression mapping an existing array of data.
Definition: TensorForwardDeclarations.h:52
test_evals
static void test_evals(const Eigen::SyclDevice &sycl_device)
Definition: cxx11_tensor_convolution_sycl.cpp:219
size_t
std::size_t size_t
Definition: wrap/pybind11/include/pybind11/detail/common.h:490
tree::f
Point2(* f)(const Point3 &, OptionalJacobian< 2, 3 >)
Definition: testExpression.cpp:218
VERIFY_IS_APPROX
#define VERIFY_IS_APPROX(a, b)
Definition: integer_types.cpp:15
Eigen::TensorBase< Tensor< Scalar_, NumIndices_, Options_, IndexType_ > >::setRandom
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Tensor< Scalar_, NumIndices_, Options_, IndexType_ > & setRandom()
Definition: TensorBase.h:996
test_strides
static void test_strides(const Eigen::SyclDevice &sycl_device)
Definition: cxx11_tensor_convolution_sycl.cpp:403
main.h
Eigen::Tensor::data
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar * data()
Definition: Tensor.h:104
error_threshold
static const float error_threshold
Definition: cxx11_tensor_convolution_sycl.cpp:32
Eigen::Tensor::size
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index size() const
Definition: Tensor.h:103
test_modes
static void test_modes(const Eigen::SyclDevice &sycl_device)
Definition: cxx11_tensor_convolution_sycl.cpp:314
Eigen::TensorBase< Tensor< Scalar_, NumIndices_, Options_, IndexType_ > >::setZero
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Tensor< Scalar_, NumIndices_, Options_, IndexType_ > & setZero()
Definition: TensorBase.h:988
i
int i
Definition: BiCGSTAB_step_by_step.cpp:9
CALL_SUBTEST
#define CALL_SUBTEST(FUNC)
Definition: main.h:399

gtsam
Author(s):
autogenerated on Sat Nov 16 2024 04:02:08