Home
last modified time | relevance | path

Searched refs:work_group_size_ (Results 1 – 25 of 28) sorted by relevance

12

/external/tensorflow/tensorflow/lite/delegates/gpu/common/tasks/
Dfully_connected.cc45 work_group_size_ = int3(16, 4, 1); in FullyConnected()
47 work_group_size_ = int3(32, 4, 1); in FullyConnected()
49 work_group_size_ = int3(32, 4, 1); in FullyConnected()
53 work_group_size_ = int3(8, 4, 1); in FullyConnected()
55 work_group_size_ = int3(16, 4, 1); in FullyConnected()
78 const int wg_total_size = work_group_size_.x * work_group_size_.y; in GetFullyConnectedKernelCode()
99 c += "#define WG_X " + std::to_string(work_group_size_.x) + "\n"; in GetFullyConnectedKernelCode()
100 c += "#define WG_Y " + std::to_string(work_group_size_.y) + "\n"; in GetFullyConnectedKernelCode()
142 for (int i = 1; i < work_group_size_.y; ++i) { in GetFullyConnectedKernelCode()
Dmean_stddev_normalization.cc140 work_group_size_.x = desired_work_group_size; in MeanStdDevNormalization()
141 work_group_size_.y = 1; // Required in MeanStdDevNormalization()
142 work_group_size_.z = 1; // Required in MeanStdDevNormalization()
158 c += GetReduceCode(gpu_info, work_group_size_.x); in GetNormalizationCode()
162 std::to_string(work_group_size_.x) + ", 1, 1)))\n"; in GetNormalizationCode()
170 std::to_string(work_group_size_.x) + R"(]; in GetNormalizationCode()
211 const int grid_x = work_group_size_.x; in GetGridSize()
Dreduce.cc142 work_group_size_ = current_wg_size; in Reduce()
144 code_ = GetReduceKernelCode(definition_, work_group_size_, in Reduce()
435 work_group_size_.x * work_group_size_.y * work_group_size_.z; in BindArguments()
456 grid_x *= work_group_size_.x; in GetGridSize()
457 grid_y *= work_group_size_.y; in GetGridSize()
458 grid_z *= work_group_size_.z; in GetGridSize()
468 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
Dconv_powervr.cc330 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
433 work_group_size_.x * work_group_size_.y * work_group_size_.z; in GenerateConv()
476 std::to_string(work_group_size_.x) + ", " + in GenerateConv()
477 std::to_string(work_group_size_.y) + ", " + in GenerateConv()
478 std::to_string(work_group_size_.z) + ")))\n"; in GenerateConv()
524 c += " int lid = LOCAL_ID_1 * " + std::to_string(work_group_size_.x) + in GenerateConv()
904 work_group_size_.x * work_group_size_.y * work_group_size_.z; in GenerateConv()
1073 work_group_size_ = int3(32, 1, 1); in GuessBestParams()
1078 work_group_size_ = int3(32, 1, 1); in GuessBestParams()
1118 work_group_size_ = int3(32, 1, 1); in GuessBestParams()
[all …]
Dconvolution_transposed_3x3.cc30 work_group_size_ = int3(8, 4, 1); in ConvolutionTransposed3x3()
123 work_group_size_.x * work_group_size_.y * work_group_size_.z; in GenerateConvolutionTransposedCode()
357 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
Dconvolution_transposed_4x4.cc52 work_group_size_ = int3(8, 4, 1); in ConvolutionTransposed4x4()
109 work_group_size_.x * work_group_size_.y * work_group_size_.z; in GenerateConvolutionTransposedCode()
Dwinograd.h38 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
105 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
Ddepthwise_conv_3x3_stride_h2.cc201 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
217 desc.work_group_size_ = int3(8, 4, 1); in CreateDepthWiseConv3x3StrideH2()
Dmean_stddev_normalization.h38 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
Dsoftmax1x1.h33 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
Dwinograd.cc224 desc.work_group_size_ = int3(8, 4, 1); in CreateWinograd4x4To36()
232 work_group_size_ = int3(32, 1, 1); in Winograd4x4To36TileX6()
514 desc.work_group_size_ = int3(32, 1, 1); in CreateWinograd36To4x4()
521 work_group_size_ = int3(32, 1, 1); in Winograd36To4x4Tile4x1()
Dconvolution_transposed_4x4.h44 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
Ddepthwise_conv_3x3.cc32 work_group_size_ = int3(8, 4, 1); in DepthwiseConv3x3()
295 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
Dconv_metal.h61 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
Dsoftmax1x1.cc28 work_group_size_ = int3(32, 1, 1); in Softmax1x1()
Dsplit.cc27 work_group_size_ = int3(8, 4, 1); in Split()
Dstrided_slice.cc80 work_group_size_ = int3(8, 4, 1); in StridedSlice()
Dfully_connected.h123 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
/external/tensorflow/tensorflow/lite/delegates/gpu/common/tasks/special/
Dfc_fc_add.cc40 work_group_size_ = int3(16, 4, 1); in FCFCAdd()
42 work_group_size_ = int3(32, 4, 1); in FCFCAdd()
44 work_group_size_ = int3(32, 4, 1); in FCFCAdd()
47 work_group_size_ = int3(8, 4, 1); in FCFCAdd()
49 work_group_size_ = int3(8, 4, 1); in FCFCAdd()
51 work_group_size_ = int3(8, 4, 1); in FCFCAdd()
53 work_group_size_ = int3(16, 4, 1); in FCFCAdd()
92 c += "#define WG_X " + std::to_string(work_group_size_.x) + "\n"; in GetFCFCAddKernelCode()
93 c += "#define WG_Y " + std::to_string(work_group_size_.y) + "\n"; in GetFCFCAddKernelCode()
159 for (int i = 1; i < work_group_size_.y; ++i) { in GetFCFCAddKernelCode()
Dfc_fc_add.h99 work_groups->push_back(work_group_size_); in GetPossibleKernelWorkGroups()
/external/tensorflow/tensorflow/lite/delegates/gpu/metal/
Dcompute_task.cc213 operation_->work_group_size_, operation_->work_group_launch_order_); in UpdateParams()
224 groupsSize.width = operation_->work_group_size_.x; in Encode()
225 groupsSize.height = operation_->work_group_size_.y; in Encode()
226 groupsSize.depth = operation_->work_group_size_.z; in Encode()
253 operation_->work_group_size_ = possible_work_groups[0]; in Tune()
256 operation_->work_group_size_, operation_->work_group_launch_order_); in Tune()
/external/tensorflow/tensorflow/lite/delegates/gpu/cl/
Dcl_operation.cc179 operation_->work_group_size_, operation_->work_group_launch_order_); in UpdateParams()
218 operation_->work_group_size_ = possible_work_groups[0]; in Tune()
221 operation_->work_group_size_, operation_->work_group_launch_order_); in Tune()
235 operation_->work_group_size_ = possible_work_groups[best_work_group_index]; in Tune()
238 operation_->work_group_size_, operation_->work_group_launch_order_); in Tune()
Dcl_operation.h71 operation_->work_group_size_); in AddToQueue()
/external/tensorflow/tensorflow/lite/delegates/gpu/common/task/
Dgpu_operation.cc95 work_group_size_(operation.work_group_size_), in GPUOperation()
117 std::swap(work_group_size_, operation.work_group_size_); in operator =()
Dgpu_operation.h142 int3 work_group_size_ = int3(8, 4, 1); variable

12