mikmykWS
Posts: 11524
Joined: 3/22/2005 Status: offline
|
quote:
ORIGINAL: Sunburn Okay, looked at this with the profiler at hand. While I did not experience any noticeable slowdown or OOM exception (and on a rig much weaker than the OP's hotrod), the profiler clearly showed a massive CPU spike at the ship-pathfinding code. So I ran the save in step-debug mode and discovered that at least one of the civilian ships (the 135.000-ton cruiseliner) was mercilessly hammering the navigator code with pathfinding requests. (Pathfinding is easily one of the most expensive areas of the code, both in CPU & RAM). This meant that the ship had real trouble following its plotted course. Curious, I traced the manual-plotted course of the ship, and noticed that at least one of the waypoints was located in an location which should be inaccessible to the ship (too shallow, just 2m deep). I checked our related UI code and confirmed that we prohibit adding a new course leg to a non-accessible location......BUUUUUT.... we currently don't prohibit _moving_ an existing waypoint to such a location. So what I presume happened is that you first plotted a rough course for the ship, and while subsequently manually refining it, you dragged the waypoint on the too-shallow location. Which led to the ship's navigator going completely bonkers. (The nav code has no concept of "we tried going through that point a thousand times, just let it go already!"). So, fixed this for the next version by adding an extra UI check that prevents dragging a waypoint to an inaccessible location. This will not automatically solve the problem in your specific scenario. I would suggest re-plotting the courses for the civilian ships, this time paying attention to depths. Thanks. After this winter we ought to rent that profiler to the MBTA(Boston Trains, Buses and Subways) profiler for a few. Just saying Nice work D
< Message edited by mikmyk -- 4/28/2015 12:39:47 AM >
_____________________________
|