mirror of
https://github.com/lightningnetwork/lnd.git
synced 2026-08-19 13:17:32 +02:00
380d17ffad
In this commit, we wire the scheduler up to the scenario file. The scored batch now goes through it whether or not the file asks for concurrency: an absent section is a window of one, and running the sequential batch through the same loop is what makes the byte-identity proof mean anything. The warmup phase stays sequential, since what it exists to measure is the value of knowledge a node was handed rather than the scheduling of the payments that bought it. The five reported fields are pointers so that a file with no section emits none of them and stays byte identical to every run before this stage, while a file that has one emits all of them including the zeroes. A zero self_contention_failures is a result and must not vanish, which is the same discipline the stage A and B counters shipped with. max_concurrent and mean_concurrent are the manipulation check, the fourth stage in a row to need one: a tier whose payments never overlap is testing nothing, and every score it produced would look perfectly reasonable. mean_concurrent is time weighted over the virtual time in which anything was live, so a sequential batch reads exactly 1.0 and anything above it is real overlap. makespan_sec is reported and does not enter the objective. It is a new axis trading against success in an unmeasured way, and the program's rule is one change at a time. |
||
|---|---|---|
| .. | ||
| commands | ||
| lncli | ||
| lnd | ||
| routesim | ||