|
Dynamic Load Balance 3.6.1+32-59d1
|
#include <dlb_talp.h>
Data Fields | |
| char | name [DLB_MONITOR_NAME_MAX] |
| int | node_id |
| int | processes_per_node |
| int64_t | total_useful_time |
| int64_t | total_mpi_time |
| int64_t | max_useful_time |
| int64_t | max_mpi_time |
| float | parallel_efficiency |
| float | communication_efficiency |
| float | load_balance |
Node metrics (of one monitor) collected asynchronously from the shared memory
| char name[DLB_MONITOR_NAME_MAX] |
Name of the monitor
| int node_id |
Node identifier, only meaningful value if MPI is enabled
| int processes_per_node |
Number of processes per node
| int64_t total_useful_time |
Time (in nanoseconds) of the accumulated CPU time of useful computation of all processes in the node
| int64_t total_mpi_time |
Time (in nanoseconds) of the accumulated CPU time of communication of all processes in the node
| int64_t max_useful_time |
Time (in nanoseconds) of the accumulated CPU time of useful computation of the process with the highest value
| int64_t max_mpi_time |
Time (in nanoseconds) of the accumulated CPU time of communication of the process with the highest value
| float parallel_efficiency |
Node ParallelEfficiency = Communication Efficiency * LoadBalance
| float communication_efficiency |
Node efficiency lost due to transfer and serialization
| float load_balance |
Node Load Balance coefficient