
This window provides a Performance Profiler that lets you visually monitor the load balancing performance of parallel simulations.
To take full advantage of this option, you must create an empty file profile.cfg in the directory where the Autodyn model resides.
If you do this, the following type of output will be displayed in the scrolling box at the foot of the window:
Cycle/Time 1 / 1.637809225157053E-4 clock time (msec) 32856 (update) Cycle/Time 2 / 3.439399411878227E-4 clock time (msec) 17030 Cycle/Time 3 / 5.421148660224775E-4 clock time (msec) 17056 Cycle/Time 4 / 7.601072880654561E-4 clock time (msec) 17029 Cycle/Time 5 / 9.998989575100770E-4 clock time (msec) 17054 Cycle/Time 6 / 1.263669799616239E-3 clock time (msec) 17016 Cycle/Time 7 / 1.553817732221804E-3 clock time (msec) 17081 Cycle/Time 8 / 1.872980465005591E-3 clock time (msec) 16994 Cycle/Time 9 / 2.224059478677190E-3 clock time (msec) 17116 Cycle/Time 10 / 2.610246402086324E-3 clock time (msec) 17024 Cycle/Time 11 / 3.035052027043786E-3 clock time (msec) 17075 Cycle/Time 12 / 3.502338224625149E-3 clock time (msec) 17070 Cycle/Time 13 / 4.016353053105620E-3 clock time (msec) 17053 Cycle/Time 14 / 4.581769376689206E-3 clock time (msec) 17019 Cycle/Time 15 / 5.203727346111727E-3 clock time (msec) 17070 Cycle/Time 16 / 5.887881127305133E-3 clock time (msec) 17011 Cycle/Time 17 / 6.640450302929377E-3 clock time (msec) 17085 Cycle/Time 18 / 7.468276414058693E-3 clock time (msec) 34853 (update) Cycle/Time 19 / 8.378885156037852E-3 clock time (msec) 17017 Cycle/Time 20 / 9.380554793925532E-3 clock time (msec) 17098
Each line gives the cycle number, the cycle time, and the wall clock time that the calculation used to compute the cycle (in milliseconds). If the line ends with "(update)", this indicates that work unit information for the interaction logic was updated on that cycle. Such cycles usually take longer to compute.
The profiler also shows you visually how well balanced your calculation is. In the example shown above, Balance is a measure of the relative CPU time spent on each task for each cycle in the simulation. So a balance of 1.0 for each task is the ideal / most efficient goal. A balance of less than 1.0 means that a task is under utilized (does not have enough work to do). A balance of greater than 1.0 means that a task is over utilized (has too much work to do).