On Mon, 26 Dec 2016 20:44:40 +0100 "Enrico Weigelt, metux IT consult" <enrico.weig...@gr13.net> wrote: > On 25.12.2016 22:58, Michael T. Pope wrote: > > There is usually only zero, one, or two game objects (client + server). > > Are client and server both running in the same process ?
Yes in single player, no in multiplayer. For game play purposes there is no need for them to be in the same process in single player either. OTOH for development purposes, debug mode is pretty much useless without access to the server in the client, and FreeCol is *not* at a stage where we can do without debug mode yet. > In which ways do these two instances differ ? None AFAICT. > Another open question: what about the Specification class ? Can there > be multiple instances at the same time ? Specification is a field in Game. It should follow the same pattern as Game. Sometimes the Specification field in the client or server Game is null, sometimes it is changed, and then messaging occurs to synchronize it with the other party and the other clients. IIRC this can only happen independently of Game at the pre-game stage (see Game + MapGenerator Options). Once the Game starts, Specification should not change. > Ah, AI players are basicly just usual players ? Not yet. AIs still essentially live in the server, and exploit the extra information available to them mercilessly. This has been one of my long term projects --- slowly decouple the AIs to the point that they *could* be ordinary players. The major obstacle that remains though is that the AI is just not smart enough, and still relies on blatant cheating. > Could they, in theory, > run in a different process, or on a different machine ? Not yet, but I have been pushing them in that direction, step by step. The comms portion of this should all "just work". The next *big* step is to stop the AIs from directly accessing the server Map. > > As usual I have to ask what problem you are trying to solve? It sounds > > like you are worried that the getGame calls are inefficient. > > Well, they probably do some overhead, as they seem to be called quite > often, but I don't know much actually take. > > I'll yet have to find out, whether > a) these functions get inlined > b) they get jit-compiled at all (according to official docs, a method > has to be called 1k times before jit gets active) > c) references to long-lived objects or final refs also get inlined I think you will find the answer is both Yes and No to all of the above:-). It is complex and depends on a whole range of compiler, run-time and system parameters. Consider also, if you find out what *your* machine does, should we make optimization decisions based on just that one configuration? Remember that FreeCol is intended to run on three reasonably distinct operating systems, on <whatever> hardware (within reason), and is sometimes compiled on site by our users so we can not just ship .jar files and think we know exactly how they will behave. > I'm not so confident w/ java profiling, Neither am I, but I have used various GUI tools and JRE hacks from time to time when we had a serious bug that was causing a hang or a massive slow down. I do not recall much detail, but it was not difficult to work out. Here is a chance for you to learn:-). > > we need to put our effort into stuff that is *broken*. > > Okay, which ones do you suggest ? I can not improve on wintertime's suggestions. - If you want to work on optimization the display code has *always* been the major bottleneck every time I checked (modulo the outright bugs and hangs). The next big category is almost always due to the client-server I/O with the hottest routines being low in the Java libraries where the bytes are being shovelled or where we are waiting for input --- there is not much we can do about that (not our code), but what we could do is try to read/write less data. There has not been much effort in that direction, because we struggle just to get to correctness. - Yes, GUI.createMiniMapThumbNail is a mess. I think it should be completely rewritten in the MapEditorController (only place it is called) as just a simple image creation routine. There is no reason for it to mess with MiniMaps. This problem is blocking a newly contributed map from being added, so it is important work. Cheers, Mike Pope
pgpwIAddeS6ZX.pgp
Description: OpenPGP digital signature
------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________ Freecol-developers mailing list Freecol-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freecol-developers
