| /external/tensorflow/tensorflow/core/kernels/ |
| D | spacetobatch_functor_gpu.cu.cc | 67 for (int block_dim = NUM_BLOCK_DIMS - 1; block_dim >= 0; --block_dim) { in S2B() local
69 if (block_dim > 0) { in S2B()
70 offset %= args.block_shape[block_dim]; in S2B()
73 batch_tensor_pos[block_dim + 1] * args.block_shape[block_dim] + in S2B()
74 offset - args.pad_start[block_dim]; in S2B()
76 space_tensor_pos >= args.space_tensor_spatial_shape[block_dim]) { in S2B()
84 space_tensor_stride *= args.space_tensor_spatial_shape[block_dim]; in S2B()
85 if (block_dim == 0) { in S2B()
95 remaining_block_idx /= args.block_shape[block_dim]; in S2B()
117 for (int block_dim = 0; block_dim < NUM_BLOCK_DIMS; ++block_dim) { in operator ()() local
[all …]
|
| D | spacetobatch_functor.cc | 107 for (int block_dim = 0; block_dim < NUM_BLOCK_DIMS; ++block_dim) { in operator ()() local
108 pad_start[block_dim] = paddings_tensor[block_dim * 2]; in operator ()()
109 block_shape[block_dim] = block_shape_tensor[block_dim]; in operator ()()
110 space_tensor_shape[block_dim] = space_tensor.dimension(block_dim + 1); in operator ()()
111 batch_tensor_shape[block_dim] = batch_tensor.dimension(block_dim + 1); in operator ()()
135 for (int block_dim = NUM_BLOCK_DIMS - 1; block_dim >= 0; --block_dim) { in operator ()() local
137 block_offsets[block_dim] = in operator ()()
138 block_dim > 0 ? block_index % block_shape[block_dim] : block_index; in operator ()()
139 block_index /= block_shape[block_dim]; in operator ()()
|
| D | spacetobatch_op.cc | 101 for (int block_dim = 0; block_dim < block_dims; ++block_dim) { in SpaceToBatchOpCompute() local
102 block_shape_product *= block_shape[block_dim]; in SpaceToBatchOpCompute()
138 for (int block_dim = 0; block_dim < removed_prefix_block_dims; ++block_dim) { in SpaceToBatchOpCompute() local
139 const int64 size = orig_input_tensor.dim_size(block_dim + 1); in SpaceToBatchOpCompute()
146 for (int block_dim = removed_prefix_block_dims; in SpaceToBatchOpCompute() local
147 block_dim < block_dims - removed_suffix_block_dims; ++block_dim) { in SpaceToBatchOpCompute()
148 const int64 pad_start = paddings[2 * block_dim], in SpaceToBatchOpCompute()
149 pad_end = paddings[2 * block_dim + 1]; in SpaceToBatchOpCompute()
153 const int64 input_size = orig_input_tensor.dim_size(block_dim + 1); in SpaceToBatchOpCompute()
154 const int64 block_shape_value = block_shape[block_dim]; in SpaceToBatchOpCompute()
[all …]
|
| D | batchtospace_op.cc | 98 for (int block_dim = 0; block_dim < block_dims; ++block_dim) { in BatchToSpaceOpCompute() local
99 block_shape_product *= block_shape[block_dim]; in BatchToSpaceOpCompute()
140 for (int block_dim = 0; block_dim < removed_prefix_block_dims; ++block_dim) { in BatchToSpaceOpCompute() local
141 const int64 size = orig_input_tensor.dim_size(block_dim + 1); in BatchToSpaceOpCompute()
148 for (int block_dim = removed_prefix_block_dims; in BatchToSpaceOpCompute() local
149 block_dim < block_dims - removed_suffix_block_dims; ++block_dim) { in BatchToSpaceOpCompute()
150 const int64 crop_start = crops[2 * block_dim], in BatchToSpaceOpCompute()
151 crop_end = crops[2 * block_dim + 1]; in BatchToSpaceOpCompute()
154 const int64 input_size = orig_input_tensor.dim_size(block_dim + 1); in BatchToSpaceOpCompute()
155 const int64 block_shape_value = block_shape[block_dim]; in BatchToSpaceOpCompute()
[all …]
|
| D | spacetobatch_benchmark_test.cc | 32 for (int block_dim = 0; block_dim < num_block_dims; ++block_dim) { in ConstructSpaceToBatchGraph() local
33 paddings_eigen_tensor(block_dim, 0) = paddings[block_dim].first; in ConstructSpaceToBatchGraph()
34 paddings_eigen_tensor(block_dim, 1) = paddings[block_dim].second; in ConstructSpaceToBatchGraph()
|
| D | depthwise_conv_op_gpu.h | 605 dim3 block_dim; in LaunchDepthwiseConv2dGPUSmall() local
610 block_dim = dim3(kBlockDepth, args.in_cols, block_height); in LaunchDepthwiseConv2dGPUSmall()
619 block_dim = dim3(args.in_cols, block_height, kBlockDepth); in LaunchDepthwiseConv2dGPUSmall()
641 block_dim.x * block_dim.y * block_dim.z); in LaunchDepthwiseConv2dGPUSmall()
642 kernel<<<config.block_count, block_dim, shared_memory_size, in LaunchDepthwiseConv2dGPUSmall()
1600 dim3 block_dim;
1605 block_dim = dim3(kBlockDepth, args.in_cols, block_height);
1613 block_dim = dim3(args.in_cols, block_height, kBlockDepth);
1627 block_dim.x * block_dim.y * block_dim.z);
1628 kernel<<<config.block_count, block_dim, shared_memory_size,
|
| D | reduction_gpu_kernels.cu.h | 621 dim3 block_dim(32, std::min(Eigen::divup(extent_x, rows_per_warp), 32), 1);
624 rows_per_warp * block_dim.y),
636 ColumnReduceMax16ColumnsKernel<IN_T, OUT_T, Op>, grid_dim, block_dim, 0,
646 grid_dim, block_dim, 0, cu_stream, in,
663 dim3 block_dim(32, std::min(extent_x, 32), 1);
675 block_dim, 0, cu_stream, in, out, extent_x,
686 ColumnReduceKernel<IN_T, T*, Op>, grid_dim, block_dim, 0, cu_stream, in,
692 block_dim, 0, cu_stream,
|
| /external/tensorflow/tensorflow/contrib/model_pruning/python/ |
| D | pruning_utils_test.py | 73 def _compare_expand_tensor_with_kronecker_product(self, tensor, block_dim): argument
76 expanded_tensor = pruning_utils.expand_tensor(tensor, block_dim)
78 tensor, array_ops.ones(block_dim))
103 def testExpandTensor(self, input_shape, block_dim): argument
105 self._compare_expand_tensor_with_kronecker_product(weights, block_dim)
|
| D | pruning_utils.py | 136 def _generate_indices(num_rows, block_dim): argument
137 indices = np.zeros(shape=[num_rows * block_dim, 1], dtype=np.int32)
138 for k in range(block_dim):
140 indices[k * num_rows + r] = r * block_dim + k
|
| /external/tensorflow/tensorflow/stream_executor/ |
| D | device_description.cc | 73 const BlockDim &block_dim = block_dim_limit(); in ToMap() local
75 absl::StrCat(block_dim.x, ",", block_dim.y, ",", block_dim.z); in ToMap()
|
| /external/tensorflow/tensorflow/examples/android/jni/object_tracking/ |
| D | flow_cache.h | 271 int block_dim = kNumCacheLevels; in BlockDimForCacheLevel() local
274 block_dim *= kCacheBranchFactor; in BlockDimForCacheLevel()
276 return block_dim; in BlockDimForCacheLevel()
|
| /external/tensorflow/tensorflow/compiler/tests/ |
| D | spacetobatch_op_test.py | 54 for block_dim, block_shape_value in enumerate(block_shape):
55 reduced_size = padded.shape[block_dim + 1] // block_shape_value
|
| /external/tensorflow/tensorflow/core/util/ |
| D | cuda_launch_config.h | 331 Status CudaLaunchKernel(void (*function)(Ts...), dim3 grid_dim, dim3 block_dim, in CudaLaunchKernel() argument
340 auto result = cudaLaunchKernel(func_ptr, grid_dim, block_dim, arg_ptrs.data(), in CudaLaunchKernel()
|
| /external/tensorflow/tensorflow/contrib/rnn/kernels/ |
| D | lstm_ops_gpu.cu.cc | 242 const int block_dim = 128; in LSTMBlockCellFpropWithCUDA() local
244 Eigen::divup(batch_size * (cell_size + input_size), block_dim); in LSTMBlockCellFpropWithCUDA()
245 TF_CHECK_OK(CudaLaunchKernel(concat_xh<T>, grid_dim, block_dim, 0, cu_stream, in LSTMBlockCellFpropWithCUDA()
|
| /external/tensorflow/tensorflow/python/kernel_tests/ |
| D | spacetobatch_op_test.py | 58 for block_dim, block_shape_value in enumerate(block_shape):
59 reduced_size = padded.shape[block_dim + 1] // block_shape_value
|
| /external/vulkan-validation-layers/layers/ |
| D | buffer_validation.cpp | 4837 auto block_dim = FormatTexelBlockExtent(image_state->createInfo.format); in ValidateBufferBounds() local
4838 buffer_width = (buffer_width + block_dim.width - 1) / block_dim.width; in ValidateBufferBounds()
4839 buffer_height = (buffer_height + block_dim.height - 1) / block_dim.height; in ValidateBufferBounds()
4841 copy_extent.width = (copy_extent.width + block_dim.width - 1) / block_dim.width; in ValidateBufferBounds()
4842 copy_extent.height = (copy_extent.height + block_dim.height - 1) / block_dim.height; in ValidateBufferBounds()
4843 copy_extent.depth = (copy_extent.depth + block_dim.depth - 1) / block_dim.depth; in ValidateBufferBounds()
|
| /external/tensorflow/tensorflow/contrib/model_pruning/ |
| D | README.md | 78 …trix is partitioned into non-overlapping blocks of size *[block_height, block_dim]* and the either…
|