According to CPLEX, times are elapsed wall-clock seconds, meaning a total elapsed time of ~319,963.5 sec. That amount coincides almost exactly with the AMPL's _solve_time=_solve_system_time+_solve_user_time=319,962.73, sec, so I assume that _solve_system are also wall-clock sec.
My concern is about AMPL's _solve_elapsed_time = 292948 . I though that this amount was the total elapsed time of the CPLEX's run, so I don't understand the reason for the large discrepancy between _solve_elapsed_time and both _solve_time and the elapsed time reported by CPLEX (a discrepancy greater than 27,000 sec).
So, the point is: if I do want to take the elapsed wall-clock running time of a solve command (in a multi-threading execution), what parameter I should take? _solve_elapsed_time or _solve_time ?
Thank you
Javier
Root node processing (before b&c):
Real time = 0.14 sec. (22.73 ticks)
Parallel b&c, 4 threads:
Real time = 119.88 sec. (47223.83 ticks)
Sync time (average) = 5.96 sec.
Wait time (average) = 0.02 sec.
------------
Total (root+branch&cut) = 120.02 sec. (47246.56 ticks)