/**
 * Copyright (c) 2026 Huawei Technologies Co., Ltd.
 * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
 * CANN Open Software License Agreement Version 2.0 (the "License").
 * Please refer to the License for details. You may not use this file except in compliance with the License.
 * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
 * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
 * See LICENSE in the root of the software repository for the full text of the License.
 */

/*!
 * \file test_geir_pdist.cpp
 * \brief GE IR (graph mode) RunGraph test case for Pdist.
 *
 * Note: on current CANN 9.2.0 the raw RunGraph path dispatches Pdist to the pypto
 * online-compile fallback (dynamic/pdist.py wrapper); that path compiles the kernel
 * but the runtime kernel-store lookup of the online-compiled fat binary fails
 * (empty kernelList in the online-compile json). The verified graph-mode entry for
 * pypto ops in this repo is the torch_extension binding (torch.library registration
 * + torch.compiler.allow_in_graph, which calls aclnnPdist and the prebuilt binary);
 * see torch_extension/pdist.py.
 */

#include <iostream>
#include <fstream>
#include <string.h>
#include <stdint.h>
#include <vector>
#include <string>
#include <map>
#include "assert.h"

#include "graph.h"
#include "types.h"
#include "tensor.h"
#include "ge_error_codes.h"
#include "ge_api_types.h"
#include "ge_api.h"
#include "array_ops.h"
#include "ge_ir_build.h"
#include "graph/operator_reg.h"
#include "graph/types.h"

// Local GE IR registration for Pdist (the hand-written op_graph/pdist_proto.h was removed;
// the built-in libopsproto.so already registers ge::op::Pdist, but ships no public proto
// header, so the registration is inlined here for standalone compilation).
namespace ge {
REG_OP(Pdist)
    .INPUT(x, TensorType({DT_FLOAT, DT_FLOAT16}))
    .OUTPUT(y, TensorType({DT_FLOAT, DT_FLOAT16}))
    .ATTR(p, Float, 2.0)
    .OP_END_FACTORY_REG(Pdist)
} // namespace ge

#define FAILED -1
#define SUCCESS 0

using namespace ge;
using std::map;
using std::string;
using std::vector;

#define ADD_INPUT(inputIndex, inputName, inputDtype, inputShape)                                                  \
    vector<int64_t> placeholder##inputIndex##_shape = inputShape;                                                 \
    auto placeholder##inputIndex = op::Data("placeholder" + inputIndex).set_attr_index(0);                        \
    TensorDesc placeholder##inputIndex##_desc = TensorDesc(ge::Shape(placeholder##inputIndex##_shape), FORMAT_ND, \
                                                           inputDtype);                                           \
    placeholder##inputIndex##_desc.SetPlacement(ge::kPlacementHost);                                              \
    placeholder##inputIndex##_desc.SetFormat(FORMAT_ND);                                                          \
    Tensor tensor_placeholder##inputIndex;                                                                        \
    ret = GenInputData(placeholder##inputIndex##_shape, tensor_placeholder##inputIndex,                           \
                       placeholder##inputIndex##_desc, inputDtype, inputIndex);                                   \
    if (ret != SUCCESS) {                                                                                         \
        printf("%s - ERROR - [XIR]: Generate input data failed\n", GetTime().c_str());                            \
        return FAILED;                                                                                            \
    }                                                                                                             \
    placeholder##inputIndex.update_input_desc_x(placeholder##inputIndex##_desc);                                  \
    input.push_back(tensor_placeholder##inputIndex);                                                              \
    graph.AddOp(placeholder##inputIndex);                                                                         \
    pdist_op.set_input_##inputName(placeholder##inputIndex);                                                      \
    inputs.push_back(placeholder##inputIndex);

#define ADD_INPUT_ATTR(attrName, attrValue) pdist_op.set_attr_##attrName(attrValue);

#define ADD_OUTPUT(outputIndex, outputName, outputDtype, outputShape)                                       \
    TensorDesc outputName##outputIndex##_desc = TensorDesc(ge::Shape(outputShape), FORMAT_ND, outputDtype); \
    pdist_op.update_output_desc_##outputName(outputName##outputIndex##_desc);

#define LOG_PRINT(message, ...)         \
    do {                                \
        printf(message, ##__VA_ARGS__); \
    } while (0)

string GetTime()
{
    time_t timep;
    time(&timep);
    char tmp[64];
    strftime(tmp, sizeof(tmp), "%Y-%m-%d %H:%M:%S,000", localtime(&timep));
    return tmp;
}

uint32_t GetDataTypeSize(DataType dt)
{
    if (dt == ge::DT_FLOAT) {
        return 4;
    }
    if (dt == ge::DT_FLOAT16 || dt == ge::DT_BF16) {
        return 2;
    }
    if (dt == ge::DT_INT64 || dt == ge::DT_UINT64) {
        return 8;
    }
    return 4;
}

// Deterministic fp32/fp16 input data; seed varies per input index.
int32_t GenInputData(vector<int64_t> shapes, Tensor& input_tensor, TensorDesc& input_tensor_desc, DataType data_type,
                     int32_t seed)
{
    input_tensor_desc.SetRealDimCnt(shapes.size());
    size_t size = 1;
    for (uint32_t i = 0; i < shapes.size(); i++) {
        size *= shapes[i];
    }
    uint32_t data_len = size * GetDataTypeSize(data_type);
    float* pData = new (std::nothrow) float[size];
    for (size_t i = 0; i < size; ++i) {
        pData[i] = static_cast<float>(((i * 37 + seed * 101) % 19) - 9) * 0.25f;
    }
    input_tensor = Tensor(input_tensor_desc, reinterpret_cast<uint8_t*>(pData), data_len);
    return SUCCESS;
}

int32_t WriteDataToFile(string bin_file, uint64_t data_size, uint8_t* inputData)
{
    FILE* fp;
    fp = fopen(bin_file.c_str(), "w");
    fwrite(inputData, sizeof(uint8_t), data_size, fp);
    fclose(fp);
    return SUCCESS;
}

int CreateOppInGraph(DataType inDtype, std::vector<ge::Tensor>& input, std::vector<Operator>& inputs,
                     std::vector<Operator>& outputs, Graph& graph)
{
    Status ret = SUCCESS;
    // Single Pdist op in the graph.
    auto pdist_op = op::Pdist("test_geir_pdist");

    // x: [N, M] fp32; y: [N*(N-1)/2]; p = 2.0 (Euclidean).
    const int64_t N = 64;
    const int64_t M = 16;
    std::vector<int64_t> xShape = {N, M};
    std::vector<int64_t> yShape = {N * (N - 1) / 2};

    ADD_INPUT(1, x, DT_FLOAT, xShape);
    ADD_OUTPUT(1, y, DT_FLOAT, yShape);
    ADD_INPUT_ATTR(p, 2.0f);

    outputs.push_back(pdist_op);
    return SUCCESS;
}

int main(int argc, char* argv[])
{
    const char* graph_name = "tc_ge_irrun_test";
    Graph graph(graph_name);
    std::vector<ge::Tensor> input;

    printf("%s - INFO - [XIR]: Start to initialize ge using ge global options\n", GetTime().c_str());
    std::map<AscendString, AscendString> global_options = {{"ge.exec.deviceId", "0"}, {"ge.graphRunMode", "1"}};
    Status ret = ge::GEInitialize(global_options);
    if (ret != SUCCESS) {
        printf("%s - INFO - [XIR]: Initialize ge using ge global options failed\n", GetTime().c_str());
        return FAILED;
    }
    printf("%s - INFO - [XIR]: Initialize ge using ge global options success\n", GetTime().c_str());

    std::vector<Operator> inputs{};
    std::vector<Operator> outputs{};

    DataType inDtype = DT_FLOAT;
    std::cout << inDtype << std::endl;

    ret = CreateOppInGraph(inDtype, input, inputs, outputs, graph);
    if (ret != SUCCESS) {
        printf("%s - ERROR - [XIR]: Create graph failed\n", GetTime().c_str());
        return FAILED;
    }

    if (!inputs.empty() && !outputs.empty()) {
        graph.SetInputs(inputs).SetOutputs(outputs);
    }

    std::map<AscendString, AscendString> build_options = {};
    printf("%s - INFO - [XIR]: Start to create ir session using build options\n", GetTime().c_str());
    ge::Session* session = new Session(build_options);
    if (session == nullptr) {
        printf("%s - ERROR - [XIR]: Create ir session using build options failed\n", GetTime().c_str());
        return FAILED;
    }
    printf("%s - INFO - [XIR]: Create ir session using build options success\n", GetTime().c_str());
    printf("%s - INFO - [XIR]: Start to add compute graph to ir session\n", GetTime().c_str());

    std::map<AscendString, AscendString> graph_options = {};
    uint32_t graph_id = 0;
    ret = session->AddGraph(graph_id, graph, graph_options);
    if (ret != SUCCESS) {
        printf("%s - ERROR - [XIR]: AddGraph failed\n", GetTime().c_str());
        delete session;
        GEFinalize();
        return FAILED;
    }

    printf("%s - INFO - [XIR]: dump graph to txt\n", GetTime().c_str());
    std::string file_path = "./dump";
    aclgrphDumpGraph(graph, file_path.c_str(), file_path.length());
    printf("%s - INFO - [XIR]: Start to run ir compute graph\n", GetTime().c_str());
    std::vector<ge::Tensor> output;
    ret = session->RunGraph(graph_id, input, output);
    if (ret != SUCCESS) {
        printf("%s - INFO - [XIR]: Run graph failed\n", GetTime().c_str());
        delete session;
        GEFinalize();
        return FAILED;
    }
    printf("%s - INFO - [XIR]: Session run ir compute graph success\n", GetTime().c_str());

    int input_num = input.size();
    for (int i = 0; i < input_num; i++) {
        std::cout << "input " << i << " dtype :  " << input[i].GetTensorDesc().GetDataType() << std::endl;
        string input_file = "./tc_ge_irrun_test_npu_input_" + std::to_string(i) + ".bin";
        uint8_t* input_data_i = input[i].GetData();
        int64_t input_shape = input[i].GetTensorDesc().GetShape().GetShapeSize();
        std::cout << "this is " << i << "th input, input shape size =" << input_shape << std::endl;
        uint32_t data_size = input_shape * GetDataTypeSize(input[i].GetTensorDesc().GetDataType());
        WriteDataToFile((const char*)input_file.c_str(), data_size, input_data_i);
    }

    int output_num = output.size();
    for (int i = 0; i < output_num; i++) {
        std::cout << "output " << i << " dtype :  " << output[i].GetTensorDesc().GetDataType() << std::endl;
        string output_file = "./tc_ge_irrun_test_npu_output_" + std::to_string(i) + ".bin";
        uint8_t* output_data_i = output[i].GetData();
        int64_t output_shape = output[i].GetTensorDesc().GetShape().GetShapeSize();
        std::cout << "this is " << i << "th output, output shape size =" << output_shape << std::endl;
        uint32_t data_size = output_shape * GetDataTypeSize(output[i].GetTensorDesc().GetDataType());
        WriteDataToFile((const char*)output_file.c_str(), data_size, output_data_i);
        float* outFloat = reinterpret_cast<float*>(output_data_i);
        for (int64_t j = 0; j < 4 && j < output_shape; j++) {
            LOG_PRINT("y[%ld] is: %f\n", j, outFloat[j]);
        }
    }
    ge::AscendString error_msg = ge::GEGetErrorMsgV2();
    std::string error_str(error_msg.GetString());
    std::cout << "Error message: " << error_str << std::endl;
    ge::AscendString warning_msg = ge::GEGetWarningMsgV2();
    std::string warning_str(warning_msg.GetString());
    std::cout << "Warning message: " << warning_str << std::endl;
    printf("%s - INFO - [XIR]: Start to finalize ir graph session\n", GetTime().c_str());
    ret = ge::GEFinalize();
    if (session != nullptr) {
        delete session;
    }
    (void)ret;
    return SUCCESS;
}