TheRock/rocm-systems/projects/cuid/example/main.cc

/*
 * Copyright (c) Advanced Micro Devices, Inc. All rights reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 * THE SOFTWARE.
 */

#include "include/amd_cuid.h"
#include <climits>
#include <iostream>
#include <vector>

int main() {
  amdcuid_status_t err;
  uint32_t handle_count = 0;
  uint32_t available_handle_count = 0;
  std::vector<amdcuid_id_t> handles;

  // Retry until the available_handle_count matches the handle_count
  do {
    handle_count = available_handle_count;
    handles.resize(handle_count);
    err = amdcuid_get_all_handles(handles.data(), &available_handle_count);
    if (err != AMDCUID_STATUS_SUCCESS &&
        err != AMDCUID_STATUS_INSUFFICIENT_SIZE) {
      std::cerr << "Failed to get device handles. Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      return 1;
    }
  } while (handle_count != available_handle_count);

  // sort handles for GPUs and CPUs separately
  std::vector<amdcuid_id_t> cpu_handles;
  std::vector<amdcuid_id_t> gpu_handles;
  for (uint32_t i = 0; i < handles.size(); ++i) {
    amdcuid_device_type_t device_type;
    uint32_t device_type_size = sizeof(device_type);
    err = amdcuid_query_device_property(handles[i], AMDCUID_QUERY_DEVICE_TYPE,
                                        &device_type, &device_type_size);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get device type for handle #" << i
                << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      continue;
    }

    if (device_type == AMDCUID_DEVICE_TYPE_GPU) {
      // copy the GPU handle to gpu_handles vector
      gpu_handles.push_back(handles[i]);
    } else if (device_type == AMDCUID_DEVICE_TYPE_CPU) {
      // copy the CPU handle to cpu_handles vector
      cpu_handles.push_back(handles[i]);
    }
  }

  std::cout << "Discovered " << gpu_handles.size() << " GPU(s):" << std::endl;
  for (uint32_t i = 0; i < gpu_handles.size(); ++i) {
    uint16_t vendor_id = 0;
    uint32_t data_size = sizeof(vendor_id);
    err = amdcuid_query_device_property(gpu_handles[i], AMDCUID_QUERY_VENDOR_ID,
                                        &vendor_id, &data_size);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get vendor_id for GPU #" << i
                << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
    }

    char device_node[128] = {0};
    uint32_t device_node_len = sizeof(device_node);
    err =
        amdcuid_query_device_property(gpu_handles[i], AMDCUID_QUERY_DEVICE_PATH,
                                      device_node, &device_node_len);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get device node for GPU #" << i
                << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      device_node[0] = '\0';
    }

    amdcuid_id_t derived_id = {};
    uint32_t derived_id_size = sizeof(derived_id);
    err = amdcuid_query_device_property(gpu_handles[i],
                                        AMDCUID_QUERY_DERIVED_CUID, &derived_id,
                                        &derived_id_size);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get derived CUID for GPU #" << i
                << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
    }

    std::cout << "GPU #" << i << std::dec << " vendor_id: " << vendor_id
              << " DeviceNode: " << device_node
              << "  CUID: " << amdcuid_id_to_string(derived_id) << std::endl;
    std::cout << " Handle: " << amdcuid_id_to_string(gpu_handles[i])
              << std::endl;
    std::cout << std::endl;
  }

  // filter handles for CPU devices

  std::cout << "Discovered " << cpu_handles.size() << " CPU(s):" << std::endl;

  for (uint32_t i = 0; i < cpu_handles.size(); ++i) {
    uint16_t vendor_id = 0;
    uint32_t data_size = sizeof(vendor_id);
    err = amdcuid_query_device_property(cpu_handles[i], AMDCUID_QUERY_VENDOR_ID,
                                        &vendor_id, &data_size);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get vendor ID for CPU #" << i
                << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      vendor_id = 0;
    }

    uint16_t core = 0;
    data_size = sizeof(core);
    err = amdcuid_query_device_property(cpu_handles[i], AMDCUID_QUERY_CORE_ID,
                                        &core, &data_size);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get core for CPU #" << i
                << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      core = 0;
    }

    amdcuid_id_t derived_id = {};
    uint32_t derived_id_size = sizeof(derived_id);
    err = amdcuid_query_device_property(cpu_handles[i],
                                        AMDCUID_QUERY_DERIVED_CUID, &derived_id,
                                        &derived_id_size);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get derived CUID for CPU #" << i
                << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
    }

    std::cout << "CPU #" << i << std::dec << " Core: " << core
              << " VendorID: " << vendor_id
              << "  CUID: " << amdcuid_id_to_string(derived_id) << std::endl;
    std::cout << " Handle: " << amdcuid_id_to_string(cpu_handles[i])
              << std::endl;
    std::cout << std::endl;
  }
  // Example handle lookup by device path (only if at least one GPU exists).
  if (!gpu_handles.empty()) {
    char example_device_path[PATH_MAX] = {0};
    uint32_t path_length = sizeof(example_device_path);
    err =
        amdcuid_query_device_property(gpu_handles[0], AMDCUID_QUERY_DEVICE_PATH,
                                      example_device_path, &path_length);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get device path for GPU #0. Error code: " << err
                << " (" << amdcuid_status_to_string(err) << ")" << std::endl;
      return 1;
    }

    amdcuid_id_t device_handle = {};
    err = amdcuid_get_handle_by_dev_path(
        example_device_path, AMDCUID_DEVICE_TYPE_GPU, &device_handle);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get device handle for path "
                << example_device_path << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      return 1;
    }

    amdcuid_id_t derived_id = {};
    uint32_t derived_id_size = sizeof(derived_id);
    err =
        amdcuid_query_device_property(device_handle, AMDCUID_QUERY_DERIVED_CUID,
                                      &derived_id, &derived_id_size);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get derived CUID for device at path "
                << example_device_path << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      return 1;
    }
    std::cout << "Device at path " << example_device_path
              << " has derived CUID: " << amdcuid_id_to_string(derived_id)
              << std::endl;

    std::string example_bdf;
    uint32_t bdf_length =
        13; // typical length of BDF string "0000:00:00.0" + null terminator
    char bdf_buffer[13] = {0};
    err = amdcuid_query_device_property(gpu_handles[0], AMDCUID_QUERY_BDF,
                                        bdf_buffer, &bdf_length);
    if (err == AMDCUID_STATUS_SUCCESS) {
      example_bdf = bdf_buffer;
    } else {
      std::cerr << "Failed to get BDF query input from GPU #0. Error code: "
                << err << " (" << amdcuid_status_to_string(err) << ")"
                << std::endl;
      return 1;
    }

    // example for getting a specific device handle by BDF and querying its
    // properties
    amdcuid_id_t bdf_device_handle = {};
    err = amdcuid_get_handle_by_bdf(
        example_bdf.c_str(), AMDCUID_DEVICE_TYPE_GPU, &bdf_device_handle);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get device handle for BDF " << example_bdf
                << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      return 1;
    }

    amdcuid_id_t derived_id_bdf = {};
    uint32_t derived_id_size2 = sizeof(derived_id_bdf);
    err = amdcuid_query_device_property(bdf_device_handle,
                                        AMDCUID_QUERY_DERIVED_CUID,
                                        &derived_id_bdf, &derived_id_size2);
    if (err != AMDCUID_STATUS_SUCCESS) {
      std::cerr << "Failed to get derived CUID for device at BDF "
                << example_bdf << ". Error code: " << err << " ("
                << amdcuid_status_to_string(err) << ")" << std::endl;
      return 1;
    }
    std::cout << "Device at BDF " << example_bdf
              << " has derived CUID: " << amdcuid_id_to_string(derived_id_bdf)
              << std::endl;

  } else {
    std::cout << "No GPU devices found; skipping GPU path/BDF lookup examples."
              << std::endl;
  }

  return 0;
}