AI/rocm-systems
2
0
Форкнуть 0
Код Задачи Запросы на слияние Действия Пакеты Проекты Релизы Вики Активность

Support standard deviation and json output for job stats

Просмотр исходного кода 
In the job stats, in addition to the max, min and average,
it will also display the standard deviation.

A new option --json is added to the rdci to output the results
in json format.

In the job stats, using the GMT time instead of timestamp
for start and end time.

Change-Id: If245c4fc4854a1dc867f97ff5aa9112af7962eca
Этот коммит содержится в:
Bill(Shuzhou) Liu
2020-07-14 15:20:08 -04:00
коммит произвёл Chris Freehill
родитель 9efb55b06f
Коммит e6d910f67a
17 изменённых файлов: 515 добавлений и 112 удалений
Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы
rdci/src/RdciGroupSubSystem.cc
+97 -26
Просмотреть файл
@@ -37,6 +37,7 @@ RdciGroupSubSystem::RdciGroupSubSystem():
void RdciGroupSubSystem::parse_cmd_opts(int argc, char ** argv) {
const int HOST_OPTIONS = 1000;
const int JSON_OPTIONS = 1001;
const struct option long_options[] = {
{"host", required_argument, nullptr, HOST_OPTIONS },
{"help", optional_argument, nullptr, 'h' },
@@ -47,6 +48,7 @@ void RdciGroupSubSystem::parse_cmd_opts(int argc, char ** argv) {
{"add", required_argument, nullptr, 'a' },
{"info", optional_argument, nullptr, 'i' },
{"delete", required_argument, nullptr, 'd' },
{"json", optional_argument, nullptr, JSON_OPTIONS },
{ nullptr, 0 , nullptr, 0 }
};
@@ -59,6 +61,9 @@ void RdciGroupSubSystem::parse_cmd_opts(int argc, char ** argv) {
case HOST_OPTIONS:
ip_port_ = optarg;
break;
case JSON_OPTIONS:
set_json_output(true);
break;
case 'h':
group_ops_ = GROUP_HELP;
return;
@@ -116,18 +121,22 @@ void RdciGroupSubSystem::parse_cmd_opts(int argc, char ** argv) {
}
void RdciGroupSubSystem::show_help() const {
if (is_json_output()) return;
std::cout << " group -- Used to create and maintain groups of GPUs.\n\n";
std::cout << "Usage\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [-u] -l\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [-u] -c <groupName> "
<< "[-a <entityId>]\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [-u] -g <groupId> "
<< "[-a <entityId>]\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [-u] "
std::cout << " rdci group [--host <IP/FQDN>:port] [--json] [-u] -l\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [--json] [-u]"
<< " -c <groupName> [-a <entityId>]\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [--json] [-u]"
<< " -g <groupId> [-a <entityId>]\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [--json] [-u] "
<< "-g <groupId> [-i]\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [-u] -d <groupId>\n";
std::cout << " rdci group [--host <IP/FQDN>:port] [--json] [-u] "
<< "-d <groupId>\n";
std::cout << "\nFlags:\n";
show_common_usage();
std::cout << " --json "
<< "Output using json.\n";
std::cout << " -l --list "
<< "List the groups that currently exist for a host.\n";
std::cout << " -g --group groupId "
@@ -148,6 +157,7 @@ void RdciGroupSubSystem::process() {
std::vector<std::string> gpu_ids;
rdc_group_info_t group_info;
uint32_t count = 0;
std::string json_group_ids = "\"gpu_groups\": [";
switch (group_ops_) {
case GROUP_HELP:
show_help();
@@ -181,8 +191,13 @@ void RdciGroupSubSystem::process() {
}
if (result == RDC_ST_OK) {
std::cout << "Successfully created group with a group ID "
<< group_id << std::endl;
if (is_json_output()) {
std::cout << "\"group_id\": \"" << group_id
<<"\", \"status\": \"ok\"";
} else {
std::cout << "Successfully created group with a group ID "
<< group_id << std::endl;
}
return;
}
break;
@@ -194,8 +209,13 @@ void RdciGroupSubSystem::process() {
}
result = rdc_group_gpu_destroy(rdc_handle_, group_id_);
if (result == RDC_ST_OK) {
std::cout << "Successfully deleted the group "
<< group_id_ << std::endl;
if (is_json_output()) {
std::cout << "\"group_id\": \"" << group_id_
<<"\", \"status\": \"ok\"";
} else {
std::cout << "Successfully deleted the group "
<< group_id_ << std::endl;
}
return;
}
break;
@@ -204,8 +224,10 @@ void RdciGroupSubSystem::process() {
result = rdc_group_get_all_ids(rdc_handle_, group_id_list, &count);
if ( result != RDC_ST_OK) break;
std::cout << count << " group found.\n";
std::cout << "GroupID\t" << "GroupName\t" << "GPUIndex\n";
if (!is_json_output()) {
std::cout << count << " group found.\n";
std::cout << "GroupID\t" << "GroupName\t" << "GPUIndex\n";
}
for (uint32_t i = 0; i < count; i++) {
result = rdc_group_gpu_get_info(rdc_handle_,
group_id_list[i], &group_info);
@@ -215,15 +237,43 @@ void RdciGroupSubSystem::process() {
+ std::to_string(group_id_list[i]));
}
std::cout << group_id_list[i] << "\t"
<< group_info.group_name << "\t\t";
if (!is_json_output()) {
std::cout << group_id_list[i] << "\t"
<< group_info.group_name << "\t\t";
} else {
json_group_ids += "{\"group_id\": \"";
json_group_ids += std::to_string(group_id_list[i]);
json_group_ids += "\", \"group_name\": \"";
json_group_ids += group_info.group_name;
json_group_ids += "\", \"gpu_indexes\": [";
}
for (uint32_t j = 0; j < group_info.count; j++) {
std::cout << group_info.entity_ids[j];
if (!is_json_output()) {
std::cout << group_info.entity_ids[j];
} else {
json_group_ids +=
std::to_string(group_info.entity_ids[j]);
}
if (j < group_info.count -1) {
std::cout << ",";
if (!is_json_output()) {
std::cout << ",";
} else {
json_group_ids += ",";
}
}
}
std::cout << std::endl;
if (!is_json_output()) {
std::cout << std::endl;
} else {
json_group_ids += "]}";
if (i != count -1) {
json_group_ids += ",";
}
}
}
if (is_json_output()) {
json_group_ids += "], \"status\": \"ok\"";
std::cout << json_group_ids;
}
break;
case GROUP_ADD_GPUS:
@@ -247,8 +297,13 @@ void RdciGroupSubSystem::process() {
}
}
if (result == RDC_ST_OK) {
std::cout << "Successfully added the GPU " << gpu_ids_
<< " to group "<< group_id_ << std::endl;
if (is_json_output()) {
std::cout << "\"group_id\": \"" << group_id_
<<"\", \"status\": \"ok\"";
} else {
std::cout << "Successfully added the GPU " << gpu_ids_
<< " to group "<< group_id_ << std::endl;
}
return;
}
break;
@@ -261,13 +316,29 @@ void RdciGroupSubSystem::process() {
result = rdc_group_gpu_get_info(rdc_handle_,
group_id_, &group_info);
if (result == RDC_ST_OK) {
std::cout << "Group name: "
<< group_info.group_name << std::endl;
std::cout << "Gpu indexes: ";
for (uint32_t i = 0; i < group_info.count; i++) {
std::cout << group_info.entity_ids[i] << " ";
if (is_json_output()) {
std::cout << "\"group_name\": \"" << group_info.group_name
<< "\", \"gpu_indexes\": [";
} else {
std::cout << "Group name: "
<< group_info.group_name << std::endl;
std::cout << "Gpu indexes: ";
}
for (uint32_t i = 0; i < group_info.count; i++) {
if (is_json_output()) {
std::cout << group_info.entity_ids[i];
if ( i != group_info.count-1 ) {
std::cout << ",";
}
} else {
std::cout << group_info.entity_ids[i] << " ";
}
}
if (is_json_output()) {
std::cout << "], \"status\": \"ok\"";
} else {
std::cout << std::endl;
}
std::cout << std::endl;
return;
}
break;