Home
last modified time | relevance | path

Searched refs:dst_shape (Results 1 – 25 of 101) sorted by relevance

12345

/external/tensorflow/tensorflow/lite/delegates/gpu/common/
Dflops_util.cc21 uint64_t GetConvolutionFlops(const BHWC& dst_shape, const OHWI& weights_shape) { in GetConvolutionFlops() argument
22 uint64_t dst_elements = dst_shape.b * dst_shape.h * dst_shape.w * dst_shape.c; in GetConvolutionFlops()
27 uint64_t GetConvolutionWinograd4x4To6x6Flops(const BHWC& dst_shape, in GetConvolutionWinograd4x4To6x6Flops() argument
29 return GetConvolutionFlops(dst_shape, weights_shape) / 4u; in GetConvolutionWinograd4x4To6x6Flops()
39 uint64_t GetDepthwiseConvolutionFlops(const BHWC& dst_shape, in GetDepthwiseConvolutionFlops() argument
41 uint64_t dst_elements = dst_shape.b * dst_shape.h * dst_shape.w * dst_shape.c; in GetDepthwiseConvolutionFlops()
46 uint64_t GetFullyConnectedFlops(const BHWC& dst_shape, in GetFullyConnectedFlops() argument
48 uint64_t dst_elements = dst_shape.b * dst_shape.h * dst_shape.w * dst_shape.c; in GetFullyConnectedFlops()
Dwinograd_util.cc114 OHWI dst_shape; in RearrangeWeightsToWinograd4x4To6x6Weights() local
115 dst_shape.o = src_weights.shape.o; in RearrangeWeightsToWinograd4x4To6x6Weights()
116 dst_shape.h = 6; in RearrangeWeightsToWinograd4x4To6x6Weights()
117 dst_shape.w = 6; in RearrangeWeightsToWinograd4x4To6x6Weights()
118 dst_shape.i = src_weights.shape.i; in RearrangeWeightsToWinograd4x4To6x6Weights()
119 dst_weights->shape = dst_shape; in RearrangeWeightsToWinograd4x4To6x6Weights()
120 dst_weights->data.resize(dst_shape.DimensionsProduct()); in RearrangeWeightsToWinograd4x4To6x6Weights()
144 const int f_index = dst_shape.LinearIndex({d, y, x, s}); in RearrangeWeightsToWinograd4x4To6x6Weights()
Dflops_util.h26 uint64_t GetConvolutionFlops(const BHWC& dst_shape, const OHWI& weights_shape);
27 uint64_t GetConvolutionWinograd4x4To6x6Flops(const BHWC& dst_shape,
33 uint64_t GetDepthwiseConvolutionFlops(const BHWC& dst_shape,
36 uint64_t GetFullyConnectedFlops(const BHWC& dst_shape,
/external/tensorflow/tensorflow/lite/delegates/gpu/common/selectors/default/
Dconvolution_selector.cc36 const Convolution2DAttributes& attr, const BHWC& dst_shape, in SelectConvolutionAdreno() argument
43 ConvGeneric conv = CreateConvGeneric(gpu_info, op_def, attr, &dst_shape); in SelectConvolutionAdreno()
49 const Convolution2DAttributes& attr, const BHWC& dst_shape, in SelectConvolutionNVidia() argument
55 ConvGeneric conv = CreateConvGeneric(gpu_info, op_def, attr, &dst_shape); in SelectConvolutionNVidia()
61 const Convolution2DAttributes& attr, const BHWC& dst_shape, in SelectConvolutionApple() argument
66 IsGoodTaskSizeForAppleConvSimd(dst_shape, gpu_info)) { in SelectConvolutionApple()
68 CreateConvolutionMetalSimd(op_def, dst_shape, attr, gpu_info); in SelectConvolutionApple()
71 ConvGeneric conv = CreateConvGeneric(gpu_info, op_def, attr, &dst_shape); in SelectConvolutionApple()
79 const Convolution2DAttributes& attr, const BHWC& dst_shape, in SelectConvolution() argument
83 return SelectConvolutionApple(attr, dst_shape, gpu_info, op_def); in SelectConvolution()
[all …]
/external/ComputeLibrary/src/runtime/heuristics/direct_conv/
DClDirectConvDefaultConfigValhall.cpp81 …const TensorShape dst_shape = misc::shape_calculator::compute_deep_convolution_sh… in configure_G78_f32() local
84 const int32_t ofm = dst_shape[0]; in configure_G78_f32()
85 const int32_t m = dst_shape[1] * dst_shape[2]; in configure_G78_f32()
90 if(dst_shape[0] <= 4) in configure_G78_f32()
142 …const TensorShape dst_shape = misc::shape_calculator::compute_deep_convolution_sh… in configure_G78_f16() local
145 const int32_t ofm = dst_shape[0]; in configure_G78_f16()
146 const int32_t m = dst_shape[1] * dst_shape[2]; in configure_G78_f16()
152 if(dst_shape[0] <= 4) in configure_G78_f16()
185 desc.n0 = dst_shape[0]; in configure_G78_f16()
278 …const TensorShape dst_shape = misc::shape_calculator::compute_deep_convolution_sh… in configure_G57_f32() local
[all …]
/external/ComputeLibrary/tests/validation/reference/
DBatchToSpaceLayer.cpp38 …const SimpleTensor<T> &src, const SimpleTensor<int32_t> &block_shape, const TensorShape &dst_shape) in batch_to_space() argument
42 SimpleTensor<T> result(dst_shape, src.data_type()); in batch_to_space()
61 …nt out_pos = out_x + dst_shape[0] * out_y + z * dst_shape[0] * dst_shape[1] + (batch % r) * dst_sh… in batch_to_space()
71 … SimpleTensor<float> &src, const SimpleTensor<int32_t> &block_shape, const TensorShape &dst_shape);
72 …t SimpleTensor<half> &src, const SimpleTensor<int32_t> &block_shape, const TensorShape &dst_shape);
DDepthToSpaceLayer.cpp38 SimpleTensor<T> depth_to_space(const SimpleTensor<T> &src, const TensorShape &dst_shape, int32_t bl… in depth_to_space() argument
41 SimpleTensor<T> result(dst_shape, src.data_type()); in depth_to_space()
61 … int out_pos = out_x + dst_shape[0] * out_y + (z % r) * dst_shape[0] * dst_shape[1] + b * dst_shap… in depth_to_space()
71 …loat> depth_to_space(const SimpleTensor<float> &src, const TensorShape &dst_shape, int32_t block_s…
72 …<half> depth_to_space(const SimpleTensor<half> &src, const TensorShape &dst_shape, int32_t block_s…
DCol2Im.cpp38 SimpleTensor<T> col2im(const SimpleTensor<T> &src, const TensorShape &dst_shape, unsigned int num_g… in col2im() argument
40 SimpleTensor<T> dst{ dst_shape, src.data_type(), 1 }; in col2im()
43 …const size_t batches = dst_shape.total_size() / (dst_shape.x() * dst_shape.y() * dst_shape.z()); in col2im()
88 template SimpleTensor<float> col2im(const SimpleTensor<float> &src, const TensorShape &dst_shape, u…
89 template SimpleTensor<half> col2im(const SimpleTensor<half> &src, const TensorShape &dst_shape, uns…
90 …or<uint8_t> col2im(const SimpleTensor<uint8_t> &src, const TensorShape &dst_shape, unsigned int nu…
DDepthwiseConvolutionLayer.cpp54 …leTensor<T> &weights, const SimpleTensor<T> &biases, const TensorShape &dst_shape, const PadStride… in depthwise_convolution_fp() argument
59 SimpleTensor<T> dst{ dst_shape, src.data_type(), 1 }; in depthwise_convolution_fp()
84 const int maximum_x = (conv_info.stride().first * (dst_shape[0] - 1)); in depthwise_convolution_fp()
85 const int maximum_y = (conv_info.stride().second * (dst_shape[1] - 1)); in depthwise_convolution_fp()
138 …const SimpleTensor<TW> &weights, const SimpleTensor<int32_t> &biases, const TensorShape &dst_shape, in depthwise_convolution_quantized() argument
143 SimpleTensor<T> dst{ dst_shape, src.data_type(), 1, dst_qinfo }; in depthwise_convolution_quantized()
177 const int maximum_x = (conv_info.stride().first * (dst_shape[0] - 1)); in depthwise_convolution_quantized()
178 const int maximum_y = (conv_info.stride().second * (dst_shape[1] - 1)); in depthwise_convolution_quantized()
238 …onst SimpleTensor<float> &weights, const SimpleTensor<float> &biases, const TensorShape &dst_shape, in depthwise_convolution() argument
241 …return depthwise_convolution_fp(src, weights, biases, dst_shape, conv_info, depth_multiplier, dila… in depthwise_convolution()
[all …]
DReductionOperation.cpp184 …pute_reduction_operation(const SimpleTensor<T> &src, const TensorShape &dst_shape, unsigned int ax… in compute_reduction_operation() argument
189 SimpleTensor<OT> dst{ dst_shape, output_data_type, 1, src.quantization_info() }; in compute_reduction_operation()
278 SimpleTensor<OT> reduction_operation(const SimpleTensor<T> &src, const TensorShape &dst_shape, unsi… in reduction_operation() argument
281 return compute_reduction_operation<T, OT>(src, dst_shape, axis, op, policy); in reduction_operation()
285 …eduction_operation(const SimpleTensor<uint8_t> &src, const TensorShape &dst_shape, unsigned int ax… in reduction_operation() argument
292 return compute_reduction_operation<uint8_t, uint8_t>(src, dst_shape, axis, op, policy); in reduction_operation()
297 …mpleTensor<float> dst_f = reference::reduction_operation<float, float>(src_f, dst_shape, axis, op); in reduction_operation()
303 return compute_reduction_operation<uint8_t, uint8_t>(src, dst_shape, axis, op, policy); in reduction_operation()
308 …reduction_operation(const SimpleTensor<int8_t> &src, const TensorShape &dst_shape, unsigned int ax… in reduction_operation() argument
315 return compute_reduction_operation<int8_t, int8_t>(src, dst_shape, axis, op, policy); in reduction_operation()
[all …]
DSpaceToDepth.cpp38 SimpleTensor<T> space_to_depth(const SimpleTensor<T> &src, const TensorShape &dst_shape, const int … in space_to_depth() argument
40 SimpleTensor<T> result(dst_shape, src.data_type()); in space_to_depth()
42 const auto width_out = static_cast<int>(dst_shape[0]); in space_to_depth()
43 const auto height_out = static_cast<int>(dst_shape[1]); in space_to_depth()
44 const auto channel_out = static_cast<int>(dst_shape[2]); in space_to_depth()
79 …loat> space_to_depth(const SimpleTensor<float> &src, const TensorShape &dst_shape, const int block…
80 …<half> space_to_depth(const SimpleTensor<half> &src, const TensorShape &dst_shape, const int block…
DSpaceToBatch.cpp38 …eTensor<int32_t> &block_shape, const SimpleTensor<int32_t> &paddings, const TensorShape &dst_shape) in space_to_batch() argument
40 SimpleTensor<T> result(dst_shape, src.data_type(), 1, src.quantization_info()); in space_to_batch()
42 const auto width_out = static_cast<int>(dst_shape[0]); in space_to_batch()
43 const auto height_out = static_cast<int>(dst_shape[1]); in space_to_batch()
44 const auto batch_out = static_cast<int>(dst_shape[3]); in space_to_batch()
94 …Tensor<int32_t> &block_shape, const SimpleTensor<int32_t> &paddings, const TensorShape &dst_shape);
95 …Tensor<int32_t> &block_shape, const SimpleTensor<int32_t> &paddings, const TensorShape &dst_shape);
96 …Tensor<int32_t> &block_shape, const SimpleTensor<int32_t> &paddings, const TensorShape &dst_shape);
DWeightsReshape.cpp38 …SimpleTensor<T> &src, const SimpleTensor<T> &biases, const TensorShape &dst_shape, const unsigned … in weights_reshape() argument
40 SimpleTensor<T> dst{ dst_shape, src.data_type(), 1 }; in weights_reshape()
68 …nsor<float> &src, const SimpleTensor<float> &biases, const TensorShape &dst_shape, const unsigned …
69 …Tensor<half> &src, const SimpleTensor<half> &biases, const TensorShape &dst_shape, const unsigned …
70 …<uint8_t> &src, const SimpleTensor<uint8_t> &biases, const TensorShape &dst_shape, const unsigned …
DFullyConnectedLayer.cpp115 …pleTensor<T> &weights, const SimpleTensor<TB> &bias, const TensorShape &dst_shape, QuantizationInf… in fully_connected_layer() argument
124 SimpleTensor<T> dst{ TensorShape{ dst_shape }, src.data_type(), 1, out_quant_info }; in fully_connected_layer()
127 …const int num_batch_dimensions = std::max(0, static_cast<int>(dst_shape.num_dimensions())… in fully_connected_layer()
141 const int num_batches = dst_shape.total_size_upper(1); in fully_connected_layer()
161 … const SimpleTensor<float> &weights, const SimpleTensor<float> &bias, const TensorShape &dst_shape,
163 …c, const SimpleTensor<half> &weights, const SimpleTensor<half> &bias, const TensorShape &dst_shape,
165 …st SimpleTensor<uint8_t> &weights, const SimpleTensor<int32_t> &bias, const TensorShape &dst_shape,
167 …nst SimpleTensor<int8_t> &weights, const SimpleTensor<int32_t> &bias, const TensorShape &dst_shape,
/external/tensorflow/tensorflow/lite/delegates/gpu/common/tasks/
Dconv_generic.cc175 const GpuInfo& gpu_info, const BHWC* dst_shape) in ConvGeneric() argument
181 conv_params_(GuessBestParams(gpu_info, definition, attr, dst_shape)) { in ConvGeneric()
202 const BHWC* dst_shape) in ConvGeneric() argument
209 dst_shape)) {} in ConvGeneric()
213 const GpuInfo& gpu_info, const BHWC* dst_shape) in ConvGeneric() argument
219 conv_params_(GuessBestParams(gpu_info, definition, attr, dst_shape)) {} in ConvGeneric()
238 const GpuInfo& gpu_info, const BHWDC* dst_shape) in ConvGeneric() argument
246 conv_params_(GuessBestParams(gpu_info, definition, attr, dst_shape)) {} in ConvGeneric()
1136 int GetGroupsCount(const BHWC& dst_shape, const int3& wg_size, in GetGroupsCount() argument
1138 const int dst_slices = DivideRoundUp(dst_shape.c, 4); in GetGroupsCount()
[all …]
Dconv_generic.h111 const BHWC* dst_shape = nullptr);
114 const GpuInfo& gpu_info, const BHWC* dst_shape = nullptr);
117 const BHWC* dst_shape = nullptr);
121 const BHWDC* dst_shape = nullptr);
144 const BHWC* dst_shape);
149 const BHWC* dst_shape);
153 const OHWI& weights_shape, const BHWC* dst_shape);
158 const BHWC* dst_shape);
162 const Convolution2DAttributes& attr, const BHWC* dst_shape);
167 const BHWDC* dst_shape);
[all …]
Dconv_metal_simd.cc419 int Get2dGroupsCount(const BHWC& dst_shape, const int2 group_size) { in Get2dGroupsCount() argument
420 int x_groups = DivideRoundUp(dst_shape.w * dst_shape.b, group_size.x); in Get2dGroupsCount()
421 int y_groups = DivideRoundUp(dst_shape.h, group_size.y); in Get2dGroupsCount()
425 int2 GetOptimalGroupSize(const BHWC& dst_shape) { in GetOptimalGroupSize() argument
427 int min_2d_work_groups = Get2dGroupsCount(dst_shape, base_work_groups[0]); in GetOptimalGroupSize()
430 int groups_count = Get2dGroupsCount(dst_shape, base_work_groups[i]); in GetOptimalGroupSize()
455 const OperationDef& definition, const BHWC& dst_shape, in CreateConvolutionMetalSimd() argument
458 const int2 optimal_2d_group_size = GetOptimalGroupSize(dst_shape); in CreateConvolutionMetalSimd()
459 const int groups2d_count = Get2dGroupsCount(dst_shape, optimal_2d_group_size); in CreateConvolutionMetalSimd()
461 DivideRoundUp(dst_shape.w * dst_shape.b * dst_shape.h, 32); in CreateConvolutionMetalSimd()
[all …]
/external/ComputeLibrary/src/dynamic_fusion/sketch/gpu/components/cl/
DClComponentElementwiseBinary.cpp68 const auto &dst_shape = dst->tensor_shape(); in validate() local
71 …il::have_different_dimensions(lhs_shape, dst_shape, 0) && detail::have_different_dimensions(rhs_sh… in validate()
78 …(lhs_shape[1] != dst_shape[1] || rhs_shape[1] != dst_shape[1]) != (lhs_shape[2] != dst_shape[2] ||… in validate()
82 detail::have_different_dimensions(lhs_shape, dst_shape, 3), in validate()
86 detail::have_different_dimensions(rhs_shape, dst_shape, 3), in validate()
/external/tensorflow/tensorflow/python/ops/ragged/
Dragged_tensor_shape.py515 def _broadcast_to_ragged_shape(rt_input, dst_shape, broadcast_inner_dimensions): argument
519 rt_input.row_splits.dtype != dst_shape.dim_size_dtype):
526 dst_shape = dst_shape.with_dim_size_dtype(dtypes.int64)
529 if rt_input.shape.ndims is None or dst_shape.rank is None:
531 if rt_input.shape.ndims > dst_shape.rank:
534 rt_input.ragged_rank >= dst_shape.num_partitioned_dimensions):
538 src_shape = src_shape.broadcast_to_rank(dst_shape.rank)
541 if dst_shape.rank > rt_input.shape.ndims:
542 if rt_input.shape.ndims < dst_shape.num_inner_dimensions + 1:
544 rt_input, array_ops.concat([[-1], dst_shape.inner_dim_sizes], axis=0))
[all …]
/external/ComputeLibrary/src/runtime/heuristics/indirect_conv/
DClIndirectConvDefaultConfigValhall.cpp66 …const TensorShape dst_shape = misc::shape_calculator::compute_deep_convolution_sh… in configure_G77_f32() local
70 const int32_t ofm = dst_shape[0]; in configure_G77_f32()
71 const int32_t m = (dst_shape[1]/ stride_x) * (dst_shape[2] / stride_y); in configure_G77_f32()
110 …const TensorShape dst_shape = misc::shape_calculator::compute_deep_convolution_sh… in configure_G77_f16() local
113 const int32_t ofm = dst_shape[0]; in configure_G77_f16()
114 const int32_t m = dst_shape[1] * dst_shape[2]; in configure_G77_f16()
/external/ComputeLibrary/src/core/
DHelpers.cpp28 ValidRegion calculate_valid_region_scale(const ITensorInfo &src_info, const TensorShape &dst_shape, in calculate_valid_region_scale() argument
35 …const float scale_x = static_cast<float>(dst_shape[idx_width]) / src_info.tensor_shape()[id… in calculate_valid_region_scale()
36 …const float scale_y = static_cast<float>(dst_shape[idx_height]) / src_info.tensor_shape()[i… in calculate_valid_region_scale()
48 … auto valid_end_out_x = std::min<int>(std::ceil(valid_end_in_x * scale_x), dst_shape[idx_width]); in calculate_valid_region_scale()
49 …auto valid_end_out_y = std::min<int>(std::ceil(valid_end_in_y * scale_y), dst_shape[idx_height]); in calculate_valid_region_scale()
93 ValidRegion valid_region{ Coordinates(), dst_shape, dst_shape.num_dimensions() }; in calculate_valid_region_scale()
98 ….shape.set(idx_width, std::min<size_t>(valid_end_out_x - valid_start_out_x, dst_shape[idx_width])); in calculate_valid_region_scale()
99 …hape.set(idx_height, std::min<size_t>(valid_end_out_y - valid_start_out_y, dst_shape[idx_height])); in calculate_valid_region_scale()
/external/tensorflow/tensorflow/lite/delegates/gpu/metal/kernels/
Dconv_test.mm69 auto dst_shape = CalculateOutputShape(src_shape, attr);
73 conv_shape.b = dst_shape.b;
76 conv_shape.c = dst_shape.c;
95 auto gpu_op0 = CreateConvGeneric(env->GetGpuInfo(), op_def, attr, &dst_shape);
98 env->ExecuteGPUOperation(src_tensor, std::move(op0_ptr), dst_shape, &output0));
121 BHWC output3_shape = dst_shape;
159 auto dst_shape = BHWC(1, 1, 1, 8);
160 ConvGeneric operation = CreateConvGeneric(env->GetGpuInfo(), op_def, attr, &dst_shape);
162 src_tensor, std::make_unique<ConvGeneric>(std::move(operation)), dst_shape, &dst_tensor));
174 const BHWC dst_shape(1, 32, 32, 1024);
[all …]
/external/ComputeLibrary/src/gpu/cl/kernels/
DClPermuteKernel.cpp50 TensorShape dst_shape = src->tensor_shape(); in get_dst_shape() local
51 permute(dst_shape, perm); in get_dst_shape()
52 return dst_shape; in get_dst_shape()
71 …const TensorShape dst_shape = misc::shape_calculator::compute_permutation_output_shape(*src, perm); in validate_arguments() local
72 ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(dst->tensor_shape(), dst_shape); in validate_arguments()
89 const TensorShape dst_shape = get_dst_shape(src, perm); in configure() local
91 auto_init_if_empty(*dst, src->clone()->set_tensor_shape(dst_shape)); in configure()
/external/tensorflow/tensorflow/compiler/xla/service/
Dbitcast_decomposer.cc90 const Shape& dst_shape = instr->shape(); in Run() local
146 if (Rank(operand) == Rank(dst_shape)) { in Run()
191 ComposePermutations(ReverseIota(Rank(dst_shape)), in Run()
192 InversePermutation(LayoutPerm(dst_shape))); in Run()
199 elem_ty, ComposePermutations(dst_shape.dimensions(), in Run()
206 if (!ShapeUtil::Equal(operand->shape(), dst_shape)) { in Run()
207 operand = create_tib(dst_shape, operand, transpose_perm); in Run()
/external/tensorflow/tensorflow/lite/delegates/gpu/common/transformations/
Dglobal_pooling_to_reduce_op.cc35 const BHWC& dst_shape) { in IsGlobalPooling() argument
36 return dst_shape.w == 1 && dst_shape.h == 1 && attr.kernel.w == src_shape.w && in IsGlobalPooling()
43 const BHWC& src_shape, const BHWC& dst_shape) { in IsGlobalAveragePooling() argument
46 IsGlobalPooling(attr, src_shape, dst_shape); in IsGlobalAveragePooling()

12345