LINT: Add cpplint, clang-format and pre-commit support
Change-Id: I3cbb787ef27d90486b212dfb1a8c77c460acc2ac
Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>
[ROCm/rdc commit: 434e40305d]
This commit is contained in:
Galantsev, Dmitrii
@@ -20,154 +20,139 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
||||
THE SOFTWARE.
|
||||
*/
|
||||
|
||||
#include <unistd.h>
|
||||
#include <string.h>
|
||||
#include <iostream>
|
||||
#include <unistd.h>
|
||||
|
||||
#include <iomanip>
|
||||
#include <iostream>
|
||||
#include <map>
|
||||
#include <string>
|
||||
|
||||
#include "rdc/rdc.h"
|
||||
|
||||
static std::string get_test_name(rdc_diag_test_cases_t test_case) {
|
||||
const std::map<rdc_diag_test_cases_t, std::string> test_desc = {
|
||||
{RDC_DIAG_COMPUTE_PROCESS, "No compute process"},
|
||||
{RDC_DIAG_COMPUTE_QUEUE, "Compute Queue ready"},
|
||||
{RDC_DIAG_SYS_MEM_CHECK, "System memory check"},
|
||||
{RDC_DIAG_NODE_TOPOLOGY, "Node topology check"},
|
||||
{RDC_DIAG_GPU_PARAMETERS, "GPU parameters check"},
|
||||
{RDC_DIAG_TEST_LAST, "Unknown"}
|
||||
};
|
||||
const std::map<rdc_diag_test_cases_t, std::string> test_desc = {
|
||||
{RDC_DIAG_COMPUTE_PROCESS, "No compute process"},
|
||||
{RDC_DIAG_COMPUTE_QUEUE, "Compute Queue ready"},
|
||||
{RDC_DIAG_SYS_MEM_CHECK, "System memory check"},
|
||||
{RDC_DIAG_NODE_TOPOLOGY, "Node topology check"},
|
||||
{RDC_DIAG_GPU_PARAMETERS, "GPU parameters check"},
|
||||
{RDC_DIAG_TEST_LAST, "Unknown"}};
|
||||
|
||||
auto test_name = test_desc.find(test_case);
|
||||
if (test_name == test_desc.end()) {
|
||||
return "Unknown Test";
|
||||
}
|
||||
return test_name->second;
|
||||
auto test_name = test_desc.find(test_case);
|
||||
if (test_name == test_desc.end()) {
|
||||
return "Unknown Test";
|
||||
}
|
||||
return test_name->second;
|
||||
}
|
||||
|
||||
int main(int, char **) {
|
||||
rdc_status_t result;
|
||||
rdc_handle_t rdc_handle;
|
||||
bool standalone = false;
|
||||
char hostIpAddress[] = {"127.0.0.1:50051"};
|
||||
char group_name[] = {"diag_group"};
|
||||
int main(int, char**) {
|
||||
rdc_status_t result;
|
||||
rdc_handle_t rdc_handle;
|
||||
bool standalone = false;
|
||||
char hostIpAddress[] = {"127.0.0.1:50051"};
|
||||
char group_name[] = {"diag_group"};
|
||||
|
||||
// Select the embedded mode and standalone mode dynamically.
|
||||
std::cout << "Start rdci in: \n";
|
||||
std::cout << "0 - Embedded mode \n";
|
||||
std::cout << "1 - Standalone mode \n";
|
||||
while (!(std::cin >> standalone)) {
|
||||
std::cout << "Invalid input.\n";
|
||||
std::cin.clear();
|
||||
std::cin.ignore();
|
||||
}
|
||||
std::cout << std::endl;
|
||||
std::cout << (standalone?
|
||||
"Standalone mode selected.\n":"Embedded mode selected.\n");
|
||||
// Select the embedded mode and standalone mode dynamically.
|
||||
std::cout << "Start rdci in: \n";
|
||||
std::cout << "0 - Embedded mode \n";
|
||||
std::cout << "1 - Standalone mode \n";
|
||||
while (!(std::cin >> standalone)) {
|
||||
std::cout << "Invalid input.\n";
|
||||
std::cin.clear();
|
||||
std::cin.ignore();
|
||||
}
|
||||
std::cout << std::endl;
|
||||
std::cout << (standalone ? "Standalone mode selected.\n" : "Embedded mode selected.\n");
|
||||
|
||||
// Init the rdc
|
||||
result = rdc_init(0);
|
||||
// Init the rdc
|
||||
result = rdc_init(0);
|
||||
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error initializing RDC. Return: " << rdc_status_string(result) << std::endl;
|
||||
goto cleanup;
|
||||
} else {
|
||||
std::cout << "RDC Initialized.\n";
|
||||
}
|
||||
|
||||
if (standalone) { // standalone
|
||||
result = rdc_connect(hostIpAddress, &rdc_handle, nullptr, nullptr, nullptr);
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error initializing RDC. Return: " <<
|
||||
rdc_status_string(result) << std::endl;
|
||||
goto cleanup;
|
||||
} else {
|
||||
std::cout << "RDC Initialized.\n";
|
||||
std::cout << "Error connecting to remote rdcd. Return: " << rdc_status_string(result)
|
||||
<< std::endl;
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
if (standalone) { // standalone
|
||||
result = rdc_connect(hostIpAddress, &rdc_handle,
|
||||
nullptr, nullptr, nullptr);
|
||||
if ( result != RDC_ST_OK ) {
|
||||
std::cout << "Error connecting to remote rdcd. Return: "
|
||||
<< rdc_status_string(result) << std::endl;
|
||||
goto cleanup;
|
||||
}
|
||||
} else { // embedded
|
||||
result = rdc_start_embedded(RDC_OPERATION_MODE_AUTO, &rdc_handle);
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error starting embedded RDC engine. Return: "
|
||||
<< rdc_status_string(result) << std::endl;
|
||||
goto cleanup;
|
||||
}
|
||||
}
|
||||
|
||||
// Now we can use the same API for both standalone and embedded
|
||||
// (1) create group for all GPUs
|
||||
rdc_gpu_group_t group_id;
|
||||
result = rdc_group_gpu_create(rdc_handle, RDC_GROUP_DEFAULT,
|
||||
group_name, &group_id);
|
||||
} else { // embedded
|
||||
result = rdc_start_embedded(RDC_OPERATION_MODE_AUTO, &rdc_handle);
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error creating group. Return: "
|
||||
<< rdc_status_string(result);
|
||||
goto cleanup;
|
||||
std::cout << "Error starting embedded RDC engine. Return: " << rdc_status_string(result)
|
||||
<< std::endl;
|
||||
goto cleanup;
|
||||
}
|
||||
}
|
||||
|
||||
// (2) start to run short diagnostic.
|
||||
rdc_diag_response_t response;
|
||||
result = rdc_diagnostic_run(rdc_handle, group_id,
|
||||
RDC_DIAG_LVL_SHORT, &response);
|
||||
// Now we can use the same API for both standalone and embedded
|
||||
// (1) create group for all GPUs
|
||||
rdc_gpu_group_t group_id;
|
||||
result = rdc_group_gpu_create(rdc_handle, RDC_GROUP_DEFAULT, group_name, &group_id);
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error creating group. Return: " << rdc_status_string(result);
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error run RDC_DIAG_LVL_SHORT diagnostic. Return: "
|
||||
<< rdc_status_string(result);
|
||||
goto cleanup;
|
||||
}
|
||||
// (2) start to run short diagnostic.
|
||||
rdc_diag_response_t response;
|
||||
result = rdc_diagnostic_run(rdc_handle, group_id, RDC_DIAG_LVL_SHORT, &response);
|
||||
|
||||
// (3) Check diagnostic results
|
||||
for (uint32_t i=0 ; i < response.results_count; i++) {
|
||||
const rdc_diag_test_result_t& test_result =
|
||||
response.diag_info[i];
|
||||
std::cout << std::setw(22) << std::left
|
||||
<< get_test_name(test_result.test_case) + ":"
|
||||
<< rdc_diagnostic_result_string(test_result.status) << "\n";
|
||||
}
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error run RDC_DIAG_LVL_SHORT diagnostic. Return: " << rdc_status_string(result);
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
// (4) diagnostic detail information
|
||||
std::cout <<" =============== Diagnostic Details ==================\n";
|
||||
for (uint32_t i=0 ; i < response.results_count; i++) {
|
||||
const rdc_diag_test_result_t& test_result =
|
||||
response.diag_info[i];
|
||||
if (test_result.info[0] != '\0') {
|
||||
std::cout << std::setw(22) << std::left
|
||||
<< get_test_name(test_result.test_case) + ":"
|
||||
// (3) Check diagnostic results
|
||||
for (uint32_t i = 0; i < response.results_count; i++) {
|
||||
const rdc_diag_test_result_t& test_result = response.diag_info[i];
|
||||
std::cout << std::setw(22) << std::left << get_test_name(test_result.test_case) + ":"
|
||||
<< rdc_diagnostic_result_string(test_result.status) << "\n";
|
||||
}
|
||||
|
||||
// (4) diagnostic detail information
|
||||
std::cout << " =============== Diagnostic Details ==================\n";
|
||||
for (uint32_t i = 0; i < response.results_count; i++) {
|
||||
const rdc_diag_test_result_t& test_result = response.diag_info[i];
|
||||
if (test_result.info[0] != '\0') {
|
||||
std::cout << std::setw(22) << std::left << get_test_name(test_result.test_case) + ":"
|
||||
<< test_result.info << "\n";
|
||||
}
|
||||
for (uint32_t j=0; j < test_result.per_gpu_result_count; j++) {
|
||||
const rdc_diag_per_gpu_result_t& gpu_result
|
||||
= test_result.gpu_results[j];
|
||||
if (strlen(gpu_result.gpu_result.msg) > 0) {
|
||||
std::cout << " GPU " << gpu_result.gpu_index
|
||||
<< " " << gpu_result.gpu_result.msg << "\n";
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// (5) run one test case
|
||||
std::cout <<" ============== Run individual diagnostic test ===========\n";
|
||||
rdc_diag_test_result_t test_result;
|
||||
result = rdc_test_case_run(rdc_handle, group_id,
|
||||
RDC_DIAG_COMPUTE_PROCESS, &test_result);
|
||||
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error run RDC_DIAG_COMPUTE_PROCESS diagnostic. Return: "
|
||||
<< rdc_status_string(result);
|
||||
goto cleanup;
|
||||
for (uint32_t j = 0; j < test_result.per_gpu_result_count; j++) {
|
||||
const rdc_diag_per_gpu_result_t& gpu_result = test_result.gpu_results[j];
|
||||
if (strlen(gpu_result.gpu_result.msg) > 0) {
|
||||
std::cout << " GPU " << gpu_result.gpu_index << " " << gpu_result.gpu_result.msg << "\n";
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
std::cout << std::setw(22) << std::left
|
||||
<< get_test_name(RDC_DIAG_COMPUTE_PROCESS) + ":"
|
||||
<< test_result.info << "\n";
|
||||
// (5) run one test case
|
||||
std::cout << " ============== Run individual diagnostic test ===========\n";
|
||||
rdc_diag_test_result_t test_result;
|
||||
result = rdc_test_case_run(rdc_handle, group_id, RDC_DIAG_COMPUTE_PROCESS, &test_result);
|
||||
|
||||
if (result != RDC_ST_OK) {
|
||||
std::cout << "Error run RDC_DIAG_COMPUTE_PROCESS diagnostic. Return: "
|
||||
<< rdc_status_string(result);
|
||||
goto cleanup;
|
||||
}
|
||||
|
||||
// Cleanup consists of shutting down RDC.
|
||||
cleanup:
|
||||
std::cout << "Cleaning up.\n";
|
||||
if (standalone)
|
||||
rdc_disconnect(rdc_handle);
|
||||
else
|
||||
rdc_stop_embedded(rdc_handle);
|
||||
rdc_shutdown();
|
||||
return result;
|
||||
std::cout << std::setw(22) << std::left << get_test_name(RDC_DIAG_COMPUTE_PROCESS) + ":"
|
||||
<< test_result.info << "\n";
|
||||
|
||||
// Cleanup consists of shutting down RDC.
|
||||
cleanup:
|
||||
std::cout << "Cleaning up.\n";
|
||||
if (standalone)
|
||||
rdc_disconnect(rdc_handle);
|
||||
else
|
||||
rdc_stop_embedded(rdc_handle);
|
||||
rdc_shutdown();
|
||||
return result;
|
||||
}
|
||||
|
||||
مرجع در شماره جدید
Block a user