#include <onnxruntime_cxx_api.h>
#include <iostream>
#include <fstream>
#include <vector>
#include <string>
#include <thread>
#include <chrono>

int main() {
    std::string ep_name = "MacaEP";
    Ort::Env env = Ort::Env(ORT_LOGGING_LEVEL_WARNING, "ONNXRT");
    OrtMACAProviderOptions maca_options;
    maca_options.device_id=0;
    Ort::SessionOptions sessionOptions;
    sessionOptions.AppendExecutionProvider_MACA(maca_options);
    Ort::Session session(env, "/local/zj_1213_bev/camera.backbone.onnx", sessionOptions);
    
    std::vector<float> input_data_1(1*6*3*256*704, 1.0f);
    std::vector<float> output_data_1(6*32*88*80);
    std::vector<float> output_data_2(6*118*32*88);
    Ort::MemoryInfo memory_info_handler = Ort::MemoryInfo::CreateCpu(OrtArenaAllocator, OrtMemTypeDefault);
    
    Ort::IoBinding binding(session);
    std::vector<std::int64_t> input_shapes_1 = session.GetInputTypeInfo(0).GetTensorTypeAndShapeInfo().GetShape();
    Ort::Value input_tensor_1 = Ort::Value::CreateTensor<float>(
        memory_info_handler, input_data_1.data(), input_data_1.size(),
        input_shapes_1.data(), input_shapes_1.size());
    binding.BindInput("img", input_tensor_1);
    
    std::vector<std::int64_t> output_shapes_1 = session.GetOutputTypeInfo(0).GetTensorTypeAndShapeInfo().GetShape();
    Ort::Value output_tensor_1 = Ort::Value::CreateTensor<float>(
        memory_info_handler, output_data_1.data(), output_data_1.size(),
        output_shapes_1.data(), output_shapes_1.size());
    binding.BindOutput("camera_feature", output_tensor_1);
    
    std::vector<std::int64_t> output_shapes_2 = session.GetOutputTypeInfo(1).GetTensorTypeAndShapeInfo().GetShape();
    Ort::Value output_tensor_2 = Ort::Value::CreateTensor<float>(
        memory_info_handler, output_data_2.data(), output_data_2.size(),
        output_shapes_2.data(), output_shapes_2.size());
    binding.BindOutput("camera_depth_weights", output_tensor_2);
    while (1) {
        auto start = std::chrono::high_resolution_clock::now();
    	session.Run(Ort::RunOptions{nullptr}, binding);
        auto end = std::chrono::high_resolution_clock::now();
        auto duration = std::chrono::duration_cast<std::chrono::microseconds>(end - start).count();
        std::cout << "consume: " << duration << " us" << std::endl;
        for (int i=0; i<20; i++) {
            std::cout << output_data_1[i] << " ";
        }
        std::cout << std::endl;
        std::cout << "-----------------------------------------------------------"<< std::endl;
    }
    return 0;
}