broadcast_node.h

Go to the documentation of this file.

#pragma once
 
#include "sdfg/data_flow/library_nodes/math/tensor/tensor_node.h"
 
#include "sdfg/serializer/json_serializer.h"
 
namespace sdfg {
namespace math {
namespace tensor {
 
inline data_flow::LibraryNodeCode LibraryNodeType_Broadcast("ml::Broadcast");
 
class BroadcastNode : public TensorNode {
private:
    std::vector<symbolic::Expression> input_shape_;
    std::vector<symbolic::Expression> output_shape_;
    // true: leading-alignment (linalg.broadcast); false: trailing-alignment (NumPy/PyTorch).
    bool padded_;
 
public:
    BroadcastNode(
        size_t element_id,
        const DebugInfo& debug_info,
        const graph::Vertex vertex,
        data_flow::DataFlowGraph& parent,
        const std::vector<symbolic::Expression>& input_shape,
        const std::vector<symbolic::Expression>& output_shape,
        bool padded = true
    );
 
    static auto constexpr RESULT_PTR_IDX = 0;
    static auto constexpr X_INPUT_IDX = 1;
 
    void validate(const Function& function) const override;
 
    const std::vector<symbolic::Expression>& input_shape() const {
        return input_shape_;
    }
    const std::vector<symbolic::Expression>& output_shape() const {
        return output_shape_;
    }
    bool padded() const {
        return padded_;
    }
 
    symbolic::SymbolSet symbols() const override;
 
    void replace(const symbolic::Expression old_expression, const symbolic::Expression new_expression) override;
 
    void replace(const symbolic::ExpressionMapping& replacements) override;
 
    bool supports_integer_types() const override {
        return true;
    }
 
    passes::LibNodeExpander::ExpandOutcome
    expand(passes::LibNodeExpander::ExpandContext& context, structured_control_flow::Block& block) override;
 
    std::unique_ptr<data_flow::DataFlowNode>
    clone(size_t element_id, const graph::Vertex vertex, data_flow::DataFlowGraph& parent) const override;
 
    data_flow::PointerAccessType pointer_access_type(int input_idx) const override;
};
 
class BroadcastNodeSerializer : public serializer::LibraryNodeSerializer {
public:
    nlohmann::json serialize(const data_flow::LibraryNode& library_node) override {
        const BroadcastNode& reduce_node = static_cast<const BroadcastNode&>(library_node);
        nlohmann::json j;
 
        j["code"] = reduce_node.code().value();
 
        serializer::JSONSerializer serializer;
        j["input_shape"] = nlohmann::json::array();
        for (auto& dim : reduce_node.input_shape()) {
            j["input_shape"].push_back(serializer.expression(dim));
        }
 
        j["output_shape"] = nlohmann::json::array();
        for (auto& dim : reduce_node.output_shape()) {
            j["output_shape"].push_back(serializer.expression(dim));
        }
 
        j["padded"] = reduce_node.padded();
 
        return j;
    }
 
    data_flow::LibraryNode& deserialize(
        const nlohmann::json& j, builder::StructuredSDFGBuilder& builder, structured_control_flow::Block& parent
    ) override {
        // Assertions for required fields
        assert(j.contains("element_id"));
        assert(j.contains("code"));
        assert(j.contains("debug_info"));
        assert(j.contains("input_shape"));
        assert(j.contains("output_shape"));
 
        auto code = j["code"].get<std::string>();
 
        std::vector<symbolic::Expression> input_shape;
        for (const auto& dim : j["input_shape"]) {
            input_shape.push_back(symbolic::parse(dim.get<std::string>()));
        }
 
        std::vector<symbolic::Expression> output_shape;
        for (const auto& dim : j["output_shape"]) {
            output_shape.push_back(symbolic::parse(dim.get<std::string>()));
        }
 
        bool padded = j.value("padded", true);
 
        // Extract debug info using JSONSerializer
        sdfg::serializer::JSONSerializer serializer;
        DebugInfo debug_info = serializer.json_to_debug_info(j["debug_info"]);
 
        return builder.add_library_node<BroadcastNode>(parent, debug_info, input_shape, output_shape, padded);
    }
};
 
} // namespace tensor
} // namespace math
} // namespace sdfg