Advances, Systems and Applications
Table 8 Accumulated exclusive time per process of SingleDGEMM in case of 16 threads
| Â | Â | x86 | OpenPOWER | ||
|---|---|---|---|---|---|
Processes | Interpretation | Exclusive Execution Time (Sec) | % of Total Execution Time | Exclusive Execution Time (Sec) | % of Total Execution Time |
MPIDI_CH3I_Progress | maintain process progress state for asynchronous communications | 1,436.507 | 65.74 | 1,991.025 | 45.11 |
MPID_nem_tcp_connpoll | tcp communication between processes | 0424.685 | 19.44 | 1,812.856 | 41.07 |
mkl_blas_avx2_dgemm_kernel_0 | dgemm function of the Math Kernel Library from Intel | 0127.769 | 05.85 | - | - |
MPID_nem_network_poll | network polling for communicating threads | - | - | 0339.154 | 00.08 |
Unknown | not reported | 0071.447 | 03.27 | 0149.722 | 00.03 |
MPIDU_Sched_are_pending | Check if all_schedules is empty | 0062.728 | 02.87 | 0121.134 | 00.03 |
Others | for rest | 0004.303 | 00.20 | - | - |