diff options
Diffstat (limited to 'src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp')
| -rw-r--r-- | src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp | 76 |
1 files changed, 51 insertions, 25 deletions
diff --git a/src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp b/src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp index fe3ea6a767..c8f65d2fd9 100644 --- a/src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp +++ b/src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp @@ -27,6 +27,7 @@ #include "arm_compute/core/Utils.h" #include "arm_compute/core/Validate.h" #include "arm_compute/runtime/Tensor.h" + #include "src/core/helpers/MemoryHelpers.h" #include "src/cpu/operators/CpuGemmConv2d.h" @@ -36,17 +37,18 @@ namespace arm_compute { struct NEGEMMConvolutionLayer::Impl { - const ITensor *weights{ nullptr }; - std::unique_ptr<cpu::CpuGemmConv2d> op{ nullptr }; + const ITensor *weights{nullptr}; + std::unique_ptr<cpu::CpuGemmConv2d> op{nullptr}; ITensorPack run_pack{}; MemoryGroup memory_group{}; - IWeightsManager *weights_manager{ nullptr }; + IWeightsManager *weights_manager{nullptr}; MemoryRequirements aux_mem_req{}; WorkspaceData<Tensor> workspace_tensors{}; - bool is_prepared{ false }; + bool is_prepared{false}; }; -NEGEMMConvolutionLayer::NEGEMMConvolutionLayer(const std::shared_ptr<IMemoryManager> &memory_manager, IWeightsManager *weights_manager) +NEGEMMConvolutionLayer::NEGEMMConvolutionLayer(const std::shared_ptr<IMemoryManager> &memory_manager, + IWeightsManager *weights_manager) : _impl(std::make_unique<Impl>()) { _impl->weights_manager = weights_manager; @@ -54,37 +56,61 @@ NEGEMMConvolutionLayer::NEGEMMConvolutionLayer(const std::shared_ptr<IMemoryMana } NEGEMMConvolutionLayer::~NEGEMMConvolutionLayer() = default; -void NEGEMMConvolutionLayer::configure(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info, - const Size2D &dilation, const ActivationLayerInfo &act_info, bool enable_fast_math, unsigned int num_groups) +void NEGEMMConvolutionLayer::configure(const ITensor *input, + const ITensor *weights, + const ITensor *biases, + ITensor *output, + const PadStrideInfo &conv_info, + const WeightsInfo &weights_info, + const Size2D &dilation, + const ActivationLayerInfo &act_info, + bool enable_fast_math, + unsigned int num_groups) { ARM_COMPUTE_ERROR_ON_NULLPTR(input, weights, output); _impl->weights = weights; _impl->op = std::make_unique<cpu::CpuGemmConv2d>(); - _impl->op->configure(input->info(), weights->info(), (biases != nullptr ? biases->info() : nullptr), output->info(), conv_info, weights_info, dilation, act_info, enable_fast_math, num_groups); + _impl->op->configure(input->info(), weights->info(), (biases != nullptr ? biases->info() : nullptr), output->info(), + conv_info, weights_info, dilation, act_info, enable_fast_math, num_groups); - _impl->run_pack = - { - { TensorType::ACL_SRC_0, input }, - { TensorType::ACL_SRC_1, weights }, - { TensorType::ACL_SRC_2, biases }, - { TensorType::ACL_DST, output } - }; - _impl->aux_mem_req = _impl->op->workspace(); - _impl->workspace_tensors = manage_workspace<Tensor>(_impl->aux_mem_req, _impl->memory_group, _impl->run_pack, _impl->run_pack); + _impl->run_pack = {{TensorType::ACL_SRC_0, input}, + {TensorType::ACL_SRC_1, weights}, + {TensorType::ACL_SRC_2, biases}, + {TensorType::ACL_DST, output}}; + _impl->aux_mem_req = _impl->op->workspace(); + _impl->workspace_tensors = + manage_workspace<Tensor>(_impl->aux_mem_req, _impl->memory_group, _impl->run_pack, _impl->run_pack); } -Status NEGEMMConvolutionLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info, - const WeightsInfo &weights_info, const Size2D &dilation, const ActivationLayerInfo &act_info, bool enable_fast_math, unsigned int num_groups) +Status NEGEMMConvolutionLayer::validate(const ITensorInfo *input, + const ITensorInfo *weights, + const ITensorInfo *biases, + const ITensorInfo *output, + const PadStrideInfo &conv_info, + const WeightsInfo &weights_info, + const Size2D &dilation, + const ActivationLayerInfo &act_info, + bool enable_fast_math, + unsigned int num_groups) { - return cpu::CpuGemmConv2d::validate(input, weights, biases, output, conv_info, weights_info, dilation, act_info, enable_fast_math, num_groups); + return cpu::CpuGemmConv2d::validate(input, weights, biases, output, conv_info, weights_info, dilation, act_info, + enable_fast_math, num_groups); } -Status NEGEMMConvolutionLayer::has_opt_impl(arm_compute::WeightFormat &expected_weight_format, const ITensorInfo *src, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *dst, - const PadStrideInfo &conv_info, - const WeightsInfo &weights_info, const Size2D &dilation, const ActivationLayerInfo &act_info, const bool enable_fast_math) +Status NEGEMMConvolutionLayer::has_opt_impl(arm_compute::WeightFormat &expected_weight_format, + const ITensorInfo *src, + const ITensorInfo *weights, + const ITensorInfo *biases, + const ITensorInfo *dst, + const PadStrideInfo &conv_info, + const WeightsInfo &weights_info, + const Size2D &dilation, + const ActivationLayerInfo &act_info, + const bool enable_fast_math) { - return cpu::CpuGemmConv2d::has_opt_impl(expected_weight_format, src, weights, biases, dst, conv_info, weights_info, dilation, act_info, enable_fast_math); + return cpu::CpuGemmConv2d::has_opt_impl(expected_weight_format, src, weights, biases, dst, conv_info, weights_info, + dilation, act_info, enable_fast_math); } void NEGEMMConvolutionLayer::run() @@ -96,7 +122,7 @@ void NEGEMMConvolutionLayer::run() void NEGEMMConvolutionLayer::prepare() { - if(!_impl->is_prepared) + if (!_impl->is_prepared) { _impl->op->prepare(_impl->run_pack); |