geopm_agent_c(3) -- query information about available agents

SYNOPSIS

Link with -lgeopm

int geopm_agent_supported(const char *agent_name);

int geopm_agent_num_policy(const char *agent_name,
                           int *num_policy);

int geopm_agent_policy_name(const char *agent_name,
                            int policy_idx,
                            size_t policy_name_max,
                            char *policy_name);

int geopm_agent_policy_json(const char *agent_name,
                            const double *policy_array,
                            size_t json_string_max,
                            char *json_string);

int geopm_agent_policy_json_partial(const char *agent_name,
                                    size_t policy_array_size,
                                    const double *policy_array,
                                    size_t json_string_max,
                                    char *json_string);

int geopm_agent_num_sample(const char *agent_name,
                           int *num_sample);

int geopm_agent_sample_name(const char *agent_name,
                            int sample_idx,
                            size_t sample_name_max,
                            char *sample_name);

int geopm_agent_num_avail(int *num_agent);

int geopm_agent_name(int agent_idx,
                     size_t agent_name_max,
                     char *agent_name);

int geopm_agent_enforce_policy(void);

DESCRIPTION

The geopm_agent_c interface is used to query GEOPM about agents on a system and create static policies for those agents. The interface provides the ability to query about specific types of supported agents, the policy parameters required by an agent, and the sample parameters provided by an agent. In addition, this interface provides a way to create a JSON policy configuration file that will be parsed by the GEOPM runtime for statically enforcing a job policy.

All functions described in this man page return an error code on failure and zero upon success; see ERRORS section below for details.

geopm_agent_supported(): queries GEOPM to determine if agent_name is supported. Returns zero if an agent is supported, otherwise an error code is returned.
geopm_agent_num_policy(): queries GEOPM for the number of policy parameters required by agent_name. The result is placed in num_policy. Returns zero if an agent is supported, otherwise an error code is returned.
geopm_agent_policy_name(): queries GEOPM for the name of a policy parameter occurring at index policy_idx for agent_name. The result is stored in policy_name. The number of bytes reserved for the output string is specified in policy_name_max. Returns zero if an agent is supported, the policy_idx is in range, and the policy name can be stored in the output string (policy_name). Otherwise an error code is returned.
geopm_agent_policy_json(): creates a JSON file based on the provided policy_array that can be used for statically controlling the job policy for agent_name. The output is stored in the json_string buffer. The number of bytes reserved for the buffer is specified in json_string_max.
geopm_agent_policy_json_partial(): creates a JSON file based on the provided policy_array of length policy_array_size that can be used for statically controlling the job policy for agent_name. The output is stored in the json_string buffer. The number of bytes reserved for the buffer is specified in json_string_max. The policy_array may contain only a partial list of the policies accepted by the Agent. Missing policy values will be automatically interpreted as NAN, indicating that the Agent should use a default value.
geopm_agent_num_sample(): queries GEOPM for the number of sample parameters provided by agent_name. The result is stored in num_sample. Returns zero on success, otherwise an error code is returned.
geopm_agent_sample_name(): queries GEOPM for the name of a sample parameter occurring at index sample_idx for agent_name. The output is stored in the sample_name buffer. The number of bytes allocated for the output buffer is specified in sample_name_max. Returns zero if an agent is supported, the sample_idx is in range, and the sample name can be stored in the output string (sample_name). Otherwise an error code is returned.
geopm_agent_num_avail(): queries GEOPM for the number of currently available agents. The output is stored in num_agent. Returns zero on success, otherwise an error code is returned.
geopm_agent_name(): queries GEOPM for the name of the agent occurring at index agent_idx. The result is stored in agent_name. The number of bytes reserved for the output string is specified in agent_name_max. Returns zero if the agent_idx is in range, and the agent name can be stored in the output string (agent_name). Otherwise an error code is returned.
geopm_agent_enforce_policy(): queries the environment for the GEOPM_AGENT and GEOPM_POLICY and enforces the policy for the agent by writing controls to the platform (see geopm_pio_c(3)). A resource manager can use this function to enforce the GEOPM configured policy (see geopm(7)) prior to releasing compute nodes for a user allocation. More generally this function allows one-time use of an Agent outside of the context of an MPI runtime or the use of a GEOPM Controller. Note that the enforcement of the policy will not support any of the dynamic features of the agent. For instance, calling geopm_agent_enforce_policy() for the power_balancer will set all nodes to the same power limit, and those limits will not vary with time.

For functions dealing with GEOPM agent policy configuration, the JSON data follows this schema:

{
    "$schema": "http://json-schema.org/draft-04/schema#",
    "id": "https://geopm.github.io/geopmagent_policy.schema.json",
    "title": "GEOPM agent policy",
    "type": "object",
    "additionalProperties": {
      "type": ["integer", "string"]
    }
}

ERRORS

All functions described on this man page return an error code. See geopm_error(3) for a full description of the error numbers and how to convert them to strings.

geopm_agent_c(3) -- query information about available agents

SYNOPSIS

DESCRIPTION

ERRORS

SEE ALSO