diff options
Diffstat (limited to 'src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp')
-rw-r--r-- | src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp | 88 |
1 files changed, 55 insertions, 33 deletions
diff --git a/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp b/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp index c8c18f35db..aef7cddd7a 100644 --- a/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp +++ b/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp @@ -27,10 +27,11 @@ #include "arm_compute/core/PixelValue.h" #include "arm_compute/core/Size2D.h" #include "arm_compute/core/Utils.h" -#include "arm_compute/core/Validate.h" #include "arm_compute/core/utils/misc/ShapeCalculator.h" #include "arm_compute/core/utils/quantization/AsymmHelpers.h" +#include "arm_compute/core/Validate.h" #include "arm_compute/runtime/CL/CLScheduler.h" + #include "src/core/helpers/MemoryHelpers.h" #include "src/gpu/cl/operators/ClGemmConv2d.h" #include "support/Cast.h" @@ -47,18 +48,19 @@ using namespace arm_compute::experimental; struct CLGEMMConvolutionLayer::Impl { - const ITensor *weights{ nullptr }; - std::unique_ptr<opencl::ClGemmConv2d> op{ nullptr }; + const ITensor *weights{nullptr}; + std::unique_ptr<opencl::ClGemmConv2d> op{nullptr}; ITensorPack run_pack{}; ITensorPack prep_pack{}; MemoryGroup memory_group{}; - IWeightsManager *weights_manager{ nullptr }; + IWeightsManager *weights_manager{nullptr}; MemoryRequirements aux_mem_req{}; WorkspaceData<CLTensor> workspace_tensors{}; - bool is_prepared{ false }; + bool is_prepared{false}; }; -CLGEMMConvolutionLayer::CLGEMMConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager, IWeightsManager *weights_manager) +CLGEMMConvolutionLayer::CLGEMMConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager, + IWeightsManager *weights_manager) : _impl(std::make_unique<Impl>()) { _impl->memory_group = MemoryGroup(memory_manager); @@ -67,40 +69,60 @@ CLGEMMConvolutionLayer::CLGEMMConvolutionLayer(std::shared_ptr<IMemoryManager> m CLGEMMConvolutionLayer::~CLGEMMConvolutionLayer() = default; -void CLGEMMConvolutionLayer::configure(const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info, - const Size2D &dilation, const ActivationLayerInfo &act_info, unsigned int num_groups) +void CLGEMMConvolutionLayer::configure(const ICLTensor *input, + const ICLTensor *weights, + const ICLTensor *biases, + ICLTensor *output, + const PadStrideInfo &conv_info, + const WeightsInfo &weights_info, + const Size2D &dilation, + const ActivationLayerInfo &act_info, + unsigned int num_groups) { - configure(CLKernelLibrary::get().get_compile_context(), input, weights, biases, output, conv_info, weights_info, dilation, act_info, num_groups); + configure(CLKernelLibrary::get().get_compile_context(), input, weights, biases, output, conv_info, weights_info, + dilation, act_info, num_groups); } -void CLGEMMConvolutionLayer::configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, - const PadStrideInfo &conv_info, - const WeightsInfo &weights_info, const Size2D &dilation, const ActivationLayerInfo &act_info, unsigned int num_groups) +void CLGEMMConvolutionLayer::configure(const CLCompileContext &compile_context, + const ICLTensor *input, + const ICLTensor *weights, + const ICLTensor *biases, + ICLTensor *output, + const PadStrideInfo &conv_info, + const WeightsInfo &weights_info, + const Size2D &dilation, + const ActivationLayerInfo &act_info, + unsigned int num_groups) { ARM_COMPUTE_ERROR_ON_NULLPTR(input, weights, output); _impl->weights = weights; _impl->op = std::make_unique<opencl::ClGemmConv2d>(); const Conv2dInfo conv2d_info = Conv2dInfo(conv_info, dilation, act_info, false, num_groups); - _impl->op->configure(compile_context, input->info(), weights->info(), (biases != nullptr ? biases->info() : nullptr), output->info(), conv2d_info, weights_info); + _impl->op->configure(compile_context, input->info(), weights->info(), + (biases != nullptr ? biases->info() : nullptr), output->info(), conv2d_info, weights_info); - _impl->run_pack = - { - { TensorType::ACL_SRC_0, input }, - { TensorType::ACL_SRC_1, weights }, - { TensorType::ACL_SRC_2, biases }, - { TensorType::ACL_DST, output } - }; - _impl->prep_pack = - { - { TensorType::ACL_SRC_1, weights }, - { TensorType::ACL_SRC_2, biases }, + _impl->run_pack = {{TensorType::ACL_SRC_0, input}, + {TensorType::ACL_SRC_1, weights}, + {TensorType::ACL_SRC_2, biases}, + {TensorType::ACL_DST, output}}; + _impl->prep_pack = { + {TensorType::ACL_SRC_1, weights}, + {TensorType::ACL_SRC_2, biases}, }; - _impl->aux_mem_req = _impl->op->workspace(); - _impl->workspace_tensors = manage_workspace<CLTensor>(_impl->aux_mem_req, _impl->memory_group, _impl->run_pack, _impl->prep_pack); + _impl->aux_mem_req = _impl->op->workspace(); + _impl->workspace_tensors = + manage_workspace<CLTensor>(_impl->aux_mem_req, _impl->memory_group, _impl->run_pack, _impl->prep_pack); } -Status CLGEMMConvolutionLayer::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, unsigned int num_groups) +Status CLGEMMConvolutionLayer::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, + unsigned int num_groups) { const Conv2dInfo conv2d_info = Conv2dInfo(conv_info, dilation, act_info, false, num_groups); return opencl::ClGemmConv2d::validate(input, weights, biases, output, conv2d_info, weights_info); @@ -115,14 +137,14 @@ void CLGEMMConvolutionLayer::run() void CLGEMMConvolutionLayer::prepare() { - if(!_impl->is_prepared) + if (!_impl->is_prepared) { _impl->op->prepare(_impl->prep_pack); - auto has_reshape = std::find_if(_impl->aux_mem_req.begin(), - _impl->aux_mem_req.end(), - [](const MemoryInfo & m) -> bool { return m.lifetime == MemoryLifetime::Persistent; }); + auto has_reshape = + std::find_if(_impl->aux_mem_req.begin(), _impl->aux_mem_req.end(), + [](const MemoryInfo &m) -> bool { return m.lifetime == MemoryLifetime::Persistent; }); - if(has_reshape != std::end(_impl->aux_mem_req)) + if (has_reshape != std::end(_impl->aux_mem_req)) { _impl->weights->mark_as_unused(); } |