diff options
Diffstat (limited to 'src/graph/mutators/NodeFusionMutator.cpp')
-rw-r--r-- | src/graph/mutators/NodeFusionMutator.cpp | 300 |
1 files changed, 193 insertions, 107 deletions
diff --git a/src/graph/mutators/NodeFusionMutator.cpp b/src/graph/mutators/NodeFusionMutator.cpp index 1d47668cf2..998a4a05c7 100644 --- a/src/graph/mutators/NodeFusionMutator.cpp +++ b/src/graph/mutators/NodeFusionMutator.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018-2020 Arm Limited. + * Copyright (c) 2018-2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -23,15 +23,18 @@ */ #include "arm_compute/graph/mutators/NodeFusionMutator.h" +#include "arm_compute/core/utils/DataTypeUtils.h" +#include "arm_compute/graph/backends/BackendRegistry.h" #include "arm_compute/graph/GraphBuilder.h" #include "arm_compute/graph/Logger.h" -#include "arm_compute/graph/Utils.h" -#include "arm_compute/graph/backends/BackendRegistry.h" #include "arm_compute/graph/nodes/FusedConvolutionBatchNormalizationNode.h" #include "arm_compute/graph/nodes/Nodes.h" +#include "arm_compute/graph/Utils.h" +#include "src/graph/mutators/MutatorUtils.h" #include "support/Cast.h" +#include <list> #include <set> namespace arm_compute @@ -40,24 +43,60 @@ namespace graph { namespace detail { +void transfer_driving_nodes_and_remove_old_node(Graph &g, INode *new_node, INode *old_node, bool add_output_tensor) +{ + if (new_node == nullptr || old_node == nullptr) + { + return; + } + + // Get driving nodes of last fusable node + std::vector<NodeIdxPair> last_driving_nodes = get_driving_nodes(*old_node); + + // Extract last fusable node accessor if any + if (old_node->output(0) == nullptr) + { + return; + } + auto old_node_accessor = old_node->output(0)->extract_accessor(); + + // Remove node + g.remove_node(old_node->id()); + + // Update fused node outputs + for (auto &driving_node : last_driving_nodes) + { + g.add_connection(new_node->id(), 0, driving_node.node_id, driving_node.index); + if (add_output_tensor) + { + configure_tensor(new_node->output(0)); + } + } + + // Update accessor to fused node + new_node->output(0)->set_accessor(std::move(old_node_accessor)); +} + void fuse_convolution_with_batch_normalization(Graph &g, const Edge *output_edge) { ARM_COMPUTE_ERROR_ON(output_edge == nullptr); auto *conv_node = arm_compute::utils::cast::polymorphic_downcast<ConvolutionLayerNode *>(output_edge->producer()); - auto *bn_node = arm_compute::utils::cast::polymorphic_downcast<BatchNormalizationLayerNode *>(output_edge->consumer()); + auto *bn_node = + arm_compute::utils::cast::polymorphic_downcast<BatchNormalizationLayerNode *>(output_edge->consumer()); // Not fusing if number of groups is greater than 1 - if(conv_node->num_groups() > 1) + if (conv_node->num_groups() > 1) { return; } - ARM_COMPUTE_LOG_GRAPH_VERBOSE("Fusing convolution node with ID : " << output_edge->producer_id() - << " with BatchNormalization Layer node with ID : " << output_edge->consumer_id() << std::endl); + ARM_COMPUTE_LOG_GRAPH_VERBOSE("Fusing convolution node with ID : " + << output_edge->producer_id() << " with BatchNormalization Layer node with ID : " + << output_edge->consumer_id() << std::endl); // Prevent fusion if fused node has an output accessor - if(conv_node->output(0)->accessor() == nullptr) + if (conv_node->output(0)->accessor() == nullptr) { const Target assigned_target = conv_node->assigned_target(); @@ -77,9 +116,10 @@ void fuse_convolution_with_batch_normalization(Graph &g, const Edge *output_edge const auto epsilon = bn_node->epsilon(); // Create the fused node - const NodeID fused_id = g.add_node<FusedConvolutionBatchNormalizationNode>(epsilon, conv_info, num_groups, conv_method, fast_math_hint, act_info); + const NodeID fused_id = g.add_node<FusedConvolutionBatchNormalizationNode>( + epsilon, conv_info, num_groups, conv_method, fast_math_hint, act_info); - if(conv_node->input_edge(2) != nullptr) + if (conv_node->input_edge(2) != nullptr) { auto conv_bias_id = conv_node->input_edge(2)->producer_id(); g.add_connection(conv_bias_id, 0, fused_id, 2); @@ -91,45 +131,33 @@ void fuse_convolution_with_batch_normalization(Graph &g, const Edge *output_edge g.add_connection(bn_mean_id, 0, fused_id, 3); g.add_connection(bn_var_id, 0, fused_id, 4); - if(bn_node->input_edge(3) != nullptr) + if (bn_node->input_edge(3) != nullptr) { const auto bn_beta_id = bn_node->input_edge(3)->producer_id(); g.add_connection(bn_beta_id, 0, fused_id, 5); } - if(bn_node->input_edge(4) != nullptr) + if (bn_node->input_edge(4) != nullptr) { const auto bn_gamma_id = bn_node->input_edge(4)->producer_id(); g.add_connection(bn_gamma_id, 0, fused_id, 6); } - auto fused_node = g.node(fused_id); - std::vector<NodeIdxPair> bn_driving_nodes = get_driving_nodes(*bn_node); + auto fused_node = g.node(fused_id); + auto bn_node_name = bn_node->name(); - // Extract batch normalization node accessor if any - auto bn_node_accessor = bn_node->output(0)->extract_accessor(); - auto bn_node_name = bn_node->name(); + transfer_driving_nodes_and_remove_old_node(g, fused_node, bn_node, true); - // Remove batch normalization node - g.remove_node(bn_node->id()); - - // Get driving nodes of batch normalization node - for(auto &driving_node : bn_driving_nodes) - { - g.add_connection(fused_id, 0, driving_node.node_id, driving_node.index); - configure_tensor(fused_node->output(0)); - } - // Update fused node outputs - fused_node->output(0)->set_accessor(std::move(bn_node_accessor)); fused_node->set_assigned_target(assigned_target); - fused_node->set_common_node_parameters(NodeParams{ conv_node->name() + "+" + bn_node_name, assigned_target }); + fused_node->set_common_node_parameters(NodeParams{conv_node->name() + "+" + bn_node_name, assigned_target}); // Remove convolution node g.remove_node(conv_node->id()); } else { - ARM_COMPUTE_LOG_GRAPH_VERBOSE("Prevented fusion of convolution with batch normalization due to the presence of an output accessor\n"); + ARM_COMPUTE_LOG_GRAPH_VERBOSE( + "Prevented fusion of convolution with batch normalization due to the presence of an output accessor\n"); } } @@ -137,14 +165,17 @@ void fuse_depthwise_convolution_with_batch_normalization(Graph &g, const Edge *o { ARM_COMPUTE_ERROR_ON(output_edge == nullptr); - auto *depth_conv_node = arm_compute::utils::cast::polymorphic_downcast<DepthwiseConvolutionLayerNode *>(output_edge->producer()); - auto *bn_node = arm_compute::utils::cast::polymorphic_downcast<BatchNormalizationLayerNode *>(output_edge->consumer()); + auto *depth_conv_node = + arm_compute::utils::cast::polymorphic_downcast<DepthwiseConvolutionLayerNode *>(output_edge->producer()); + auto *bn_node = + arm_compute::utils::cast::polymorphic_downcast<BatchNormalizationLayerNode *>(output_edge->consumer()); - ARM_COMPUTE_LOG_GRAPH_VERBOSE("Fusing depthwise convolution node with ID : " << output_edge->producer_id() - << " with BatchNormalization Layer node with ID : " << output_edge->consumer_id() << std::endl); + ARM_COMPUTE_LOG_GRAPH_VERBOSE("Fusing depthwise convolution node with ID : " + << output_edge->producer_id() << " with BatchNormalization Layer node with ID : " + << output_edge->consumer_id() << std::endl); // Prevent fusion if fused node has an output accessor - if(depth_conv_node->output(0)->accessor() == nullptr) + if (depth_conv_node->output(0)->accessor() == nullptr) { const Target assigned_target = depth_conv_node->assigned_target(); @@ -164,9 +195,10 @@ void fuse_depthwise_convolution_with_batch_normalization(Graph &g, const Edge *o const auto epsilon = bn_node->epsilon(); // Create the fused node - const NodeID fused_id = g.add_node<FusedDepthwiseConvolutionBatchNormalizationNode>(epsilon, conv_info, depth_multiplier, depth_conv_method, act_info); + const NodeID fused_id = g.add_node<FusedDepthwiseConvolutionBatchNormalizationNode>( + epsilon, conv_info, depth_multiplier, depth_conv_method, act_info); - if(depth_conv_node->input_edge(2) != nullptr) + if (depth_conv_node->input_edge(2) != nullptr) { const auto conv_bias_id = depth_conv_node->input_edge(2)->producer_id(); g.add_connection(conv_bias_id, 0, fused_id, 2); @@ -180,38 +212,29 @@ void fuse_depthwise_convolution_with_batch_normalization(Graph &g, const Edge *o g.add_connection(bn_beta_id, 0, fused_id, 5); g.add_connection(bn_gamma_id, 0, fused_id, 6); - auto fused_node = g.node(fused_id); - std::vector<NodeIdxPair> bn_driving_nodes = get_driving_nodes(*bn_node); - - // Extract batch normalization node accessor if any - auto bn_node_accessor = bn_node->output(0)->extract_accessor(); - auto bn_node_name = bn_node->name(); + auto fused_node = g.node(fused_id); + auto bn_node_name = bn_node->name(); - // Remove batch normalization node - g.remove_node(bn_node->id()); + transfer_driving_nodes_and_remove_old_node(g, fused_node, bn_node, true); - // Get driving nodes of batch normalization node - for(auto &driving_node : bn_driving_nodes) - { - g.add_connection(fused_id, 0, driving_node.node_id, driving_node.index); - configure_tensor(fused_node->output(0)); - } - // Update fused node outputs - fused_node->output(0)->set_accessor(std::move(bn_node_accessor)); fused_node->set_assigned_target(assigned_target); - fused_node->set_common_node_parameters(NodeParams{ depth_conv_node->name() + "+" + bn_node_name, assigned_target }); + fused_node->set_common_node_parameters( + NodeParams{depth_conv_node->name() + "+" + bn_node_name, assigned_target}); // Remove convolution node g.remove_node(depth_conv_node->id()); } else { - ARM_COMPUTE_LOG_GRAPH_VERBOSE("Prevented fusion of depthwise convolution with batch normalization due to the presence of an output accessor\n"); + ARM_COMPUTE_LOG_GRAPH_VERBOSE("Prevented fusion of depthwise convolution with batch normalization due to the " + "presence of an output accessor\n"); } } template <typename N> -void fuse_node_with_activation(Graph &g, const Edge *output_edge, const std::set<Activation> &supported_fused_activations) +void fuse_node_with_activation(Graph &g, + const Edge *output_edge, + const std::set<Activation> &supported_fused_activations) { ARM_COMPUTE_ERROR_ON(output_edge == nullptr); @@ -221,70 +244,126 @@ void fuse_node_with_activation(Graph &g, const Edge *output_edge, const std::set ARM_COMPUTE_ERROR_ON(act_node->output(0) == nullptr || n_node->output(0) == nullptr); // Check if activation is supported for fusion - if(supported_fused_activations.count(act_node->activation_info().activation()) == 0) + if (supported_fused_activations.count(act_node->activation_info().activation()) == 0) { return; } // EltwiseLayerNode can only be fused when dataype is float - if(n_node->type() == NodeType::EltwiseLayer && !is_data_type_float(n_node->output(0)->desc().data_type)) + if (n_node->type() == NodeType::EltwiseLayer && !is_data_type_float(n_node->output(0)->desc().data_type)) { return; } ARM_COMPUTE_LOG_GRAPH_VERBOSE("Fusing node with ID : " << output_edge->producer_id() - << " with Activation Layer node with ID : " << output_edge->consumer_id() << std::endl); + << " with Activation Layer node with ID : " + << output_edge->consumer_id() << std::endl); // Prevent fusion if fused node has an output accessor - if(n_node->output(0)->accessor() == nullptr) + if (n_node->output(0)->accessor() == nullptr) { - // Get driving nodes of activation node - std::vector<NodeIdxPair> act_driving_nodes = get_driving_nodes(*act_node); - // Set activation info to fused node n_node->set_fused_activation(act_node->activation_info()); - // Extract activation node accessor if any - auto act_node_accessor = act_node->output(0)->extract_accessor(); + transfer_driving_nodes_and_remove_old_node(g, n_node, act_node, false); + } + else + { + ARM_COMPUTE_LOG_GRAPH_VERBOSE( + "Prevented fusion of node with activation due to the presence of an output accessor\n"); + } +} + +template <typename N> +void fuse_pad_with_convolution(Graph &g, const Edge *output_edge) +{ + auto *pad_node = arm_compute::utils::cast::polymorphic_downcast<PadLayerNode *>(output_edge->producer()); + auto *conv_node = arm_compute::utils::cast::polymorphic_downcast<N *>(output_edge->consumer()); + + const Edge *input_edge = pad_node->input_edge(0); + if (input_edge != nullptr && input_edge->tensor() != nullptr && pad_node->output(0)->accessor() == nullptr && + pad_node->pad_value().get<float>() == 0.0) + { + const DataLayout layout = input_edge->tensor()->desc().layout; + const PaddingList padding_list = pad_node->padding(); + + const unsigned int height_index = get_dimension_idx(layout, DataLayoutDimension::HEIGHT); + const unsigned int width_index = get_dimension_idx(layout, DataLayoutDimension::WIDTH); - // Remove activation node - g.remove_node(act_node->id()); + const PaddingInfo pad_w = width_index < padding_list.size() ? padding_list[width_index] : PaddingInfo(0, 0); + const PaddingInfo pad_h = height_index < padding_list.size() ? padding_list[height_index] : PaddingInfo(0, 0); - // Update fused node outputs - for(auto &driving_node : act_driving_nodes) + if (is_padding_in_height_or_width(layout, padding_list)) { - g.add_connection(n_node->id(), 0, driving_node.node_id, driving_node.index); + // Add paddings to the convolution node + const PadStrideInfo conv_info = conv_node->convolution_info(); + const PadStrideInfo new_conv_info(conv_info.stride().first, conv_info.stride().second, + conv_info.pad_left() + pad_w.first, conv_info.pad_right() + pad_w.second, + conv_info.pad_top() + pad_h.first, conv_info.pad_bottom() + pad_h.second, + conv_info.round()); + conv_node->set_convolution_info(new_conv_info); + + // Update drivers of the convolution node + std::vector<NodeIdxPair> pad_driver_nodes = get_driver_nodes(*pad_node); + g.remove_node(pad_node->id()); + + // Update fused node inputs + for (auto &driver_node : pad_driver_nodes) + { + g.add_connection(driver_node.node_id, driver_node.index, conv_node->id(), 0); + } } - - // Update accessor to fused node - n_node->output(0)->set_accessor(std::move(act_node_accessor)); - } - else - { - ARM_COMPUTE_LOG_GRAPH_VERBOSE("Prevented fusion of node with activation due to the presence of an output accessor\n"); } } template <typename N1, typename N2, typename F, typename... Args> -void fuse_layer(Graph &g, std::function<bool(INode &)> const &prec, const F fuse_fcn, Args &&... optional_arguments) +void fuse_layer(Graph &g, std::function<bool(INode &)> const &prec, const F fuse_fcn, Args &&...optional_arguments) { - // Not interested in the order of nodes - for(auto &node : g.nodes()) + // Note that fused nodes may be added to the end of the node list. + // Instead of only looping over the original list of nodes, we loop over the current node list which could be growing. + // This is intentional as it probes the newly added fused nodes for further fusing opportunities. + for (unsigned int i = 0; i < g.nodes().size(); ++i) { - // Check if the node is of type N and not a branching node - if(node && node->type() == N1::node_type && node->output_edges().size() == 1) + auto node = g.node(i); + // Check if the node is of type N1 and not a branching node + if (node && node->type() == N1::node_type && node->output_edges().size() == 1) { const auto output_edge_id = *node->output_edges().begin(); const auto output_edge = g.edge(output_edge_id); - // Check if following node is an activation layer node - if((output_edge != nullptr) && (output_edge->consumer() != nullptr) && (output_edge->consumer()->type() == N2::node_type) && prec(*output_edge->producer())) + // Check if following node is a type N2 node + if ((output_edge != nullptr) && (output_edge->consumer() != nullptr) && + (output_edge->consumer()->type() == N2::node_type) && prec(*output_edge->producer())) { fuse_fcn(g, output_edge, optional_arguments...); } } } } + +template <typename N1, typename F, typename... Args> +void fuse_layer(Graph &g, std::function<bool(INode &)> const &prec, const F fuse_fcn, Args &&...optional_arguments) +{ + // Note that fused nodes may be added to the end of the node list. + // Instead of only looping over the original list of nodes, we loop over the current node list which could be growing. + // This is intentional as it probes the newly added fused nodes for further fusing opportunities. + for (unsigned int i = 0; i < g.nodes().size(); ++i) + { + auto node = g.node(i); + // Check if the node is of type N1 and not a branching node + if (node && node->type() == N1::node_type && node->output_edges().size() == 1) + { + const auto output_edge_id = *node->output_edges().begin(); + const auto output_edge = g.edge(output_edge_id); + + // Check if it's the correct target + if ((output_edge != nullptr) && (output_edge->consumer() != nullptr) && prec(*output_edge->producer())) + { + fuse_fcn(g, output_edge, i, optional_arguments...); + } + } + } +} } // namespace detail const char *NodeFusionMutator::name() @@ -300,43 +379,50 @@ IGraphMutator::MutationType NodeFusionMutator::type() const void NodeFusionMutator::mutate(Graph &g) { // Supported activations when fusing - const std::set<Activation> supported_fused_activations = { Activation::ABS, Activation::BOUNDED_RELU, Activation::ELU, - Activation::HARD_SWISH, Activation::IDENTITY, Activation::LEAKY_RELU, - Activation::LINEAR, Activation::LOGISTIC, Activation::LU_BOUNDED_RELU, - Activation::RELU, Activation::SOFT_RELU, Activation::SQRT, - Activation::SQUARE, Activation::TANH - }; + const std::set<Activation> supported_fused_activations = { + Activation::ABS, Activation::BOUNDED_RELU, Activation::ELU, + Activation::HARD_SWISH, Activation::IDENTITY, Activation::LEAKY_RELU, + Activation::LINEAR, Activation::LOGISTIC, Activation::LU_BOUNDED_RELU, + Activation::RELU, Activation::SOFT_RELU, Activation::SQRT, + Activation::SQUARE, Activation::TANH}; // Preconditions - auto empty_prec = [](INode &) - { - return true; - }; - auto cl_target_prec = [](INode & n) - { - return n.assigned_target() == Target::CL; - }; - auto qs8_prec = [&g](INode & n) + auto empty_prec = [](INode &) { return true; }; + auto cl_target_prec = [](INode &n) { return n.assigned_target() == Target::CL; }; + auto qs8_prec = [&g](INode &n) { ARM_COMPUTE_ERROR_ON(n.output(0) == nullptr); const auto output_edge_id = *n.output_edges().begin(); const auto output_edge = g.edge(output_edge_id); // To perform fusion the two nodes must have same output quantization information - const bool same_qinfo = n.output(0)->desc().quant_info == output_edge->producer()->output(0)->desc().quant_info; + const bool same_qinfo = n.output(0)->desc().quant_info == output_edge->producer()->output(0)->desc().quant_info; const bool output_qasymm8 = n.output(0)->desc().data_type == DataType::QASYMM8; return (output_qasymm8 && same_qinfo) || !output_qasymm8; }; // Fusion mutations - detail::fuse_layer<BatchNormalizationLayerNode, ActivationLayerNode>(g, empty_prec, detail::fuse_node_with_activation<BatchNormalizationLayerNode>, supported_fused_activations); - detail::fuse_layer<ConvolutionLayerNode, ActivationLayerNode>(g, empty_prec, detail::fuse_node_with_activation<ConvolutionLayerNode>, supported_fused_activations); - detail::fuse_layer<DepthwiseConvolutionLayerNode, ActivationLayerNode>(g, qs8_prec, detail::fuse_node_with_activation<DepthwiseConvolutionLayerNode>, supported_fused_activations); - detail::fuse_layer<FullyConnectedLayerNode, ActivationLayerNode>(g, empty_prec, detail::fuse_node_with_activation<FullyConnectedLayerNode>, supported_fused_activations); - detail::fuse_layer<EltwiseLayerNode, ActivationLayerNode>(g, cl_target_prec, detail::fuse_node_with_activation<EltwiseLayerNode>, supported_fused_activations); - detail::fuse_layer<ConvolutionLayerNode, BatchNormalizationLayerNode>(g, empty_prec, detail::fuse_convolution_with_batch_normalization); - detail::fuse_layer<DepthwiseConvolutionLayerNode, BatchNormalizationLayerNode>(g, empty_prec, detail::fuse_depthwise_convolution_with_batch_normalization); + + detail::fuse_layer<PadLayerNode, ConvolutionLayerNode>(g, empty_prec, + detail::fuse_pad_with_convolution<ConvolutionLayerNode>); + detail::fuse_layer<PadLayerNode, DepthwiseConvolutionLayerNode>( + g, empty_prec, detail::fuse_pad_with_convolution<DepthwiseConvolutionLayerNode>); + detail::fuse_layer<BatchNormalizationLayerNode, ActivationLayerNode>( + g, empty_prec, detail::fuse_node_with_activation<BatchNormalizationLayerNode>, supported_fused_activations); + detail::fuse_layer<ConvolutionLayerNode, ActivationLayerNode>( + g, empty_prec, detail::fuse_node_with_activation<ConvolutionLayerNode>, supported_fused_activations); + detail::fuse_layer<DepthwiseConvolutionLayerNode, ActivationLayerNode>( + g, qs8_prec, detail::fuse_node_with_activation<DepthwiseConvolutionLayerNode>, supported_fused_activations); + detail::fuse_layer<FullyConnectedLayerNode, ActivationLayerNode>( + g, empty_prec, detail::fuse_node_with_activation<FullyConnectedLayerNode>, supported_fused_activations); + detail::fuse_layer<EltwiseLayerNode, ActivationLayerNode>( + g, cl_target_prec, detail::fuse_node_with_activation<EltwiseLayerNode>, supported_fused_activations); + // The fusion of BatchNormalizationLayer must occur after the fusion of ActivationLayer. Because FusedConvolutionBatchNormalizationNode assumes the BatchNormalization is already fused with activation, if any + detail::fuse_layer<ConvolutionLayerNode, BatchNormalizationLayerNode>( + g, empty_prec, detail::fuse_convolution_with_batch_normalization); + detail::fuse_layer<DepthwiseConvolutionLayerNode, BatchNormalizationLayerNode>( + g, empty_prec, detail::fuse_depthwise_convolution_with_batch_normalization); } } // namespace graph } // namespace arm_compute |