BTW, another gem of a Tim Sweeney commentary on CLR GC pauses, admittedly old from 2002 (though I don't think this has changed much since then, nor will it until C4)
http://floodyberry.com/sweeney/posting.html#s2002072102410 [C#] seems very capable in most regards and has many attributes of an > excellent game scripting language, but it's garbage collection pauses seem > to be a showstopper flaw. In prototype code that does an amount of realtime > object creation that is reasonable for a current generation game, i.e. > around 100 object allocations per frame at 60 fps, I'm seeing pauses of 2-3 > seconds that occur once every 5-10 minutes. > > Is this a feature or a bug? > > Obviously one can't ship an action game that hits a 5 second pause every > 5-10 minutes! > > Note that in this scenario the amortized time going to garbage collection > (under 3% of execution time) is totally acceptable. What's unacceptable is > that it all happens at once and makes the application appear to be locked > up. In a realtime app such as a game that attempts to run at 60 fps, it is > necessary to establish a realtime expectation on garbage collection > overhead (not a hard guarantee, but an expectation that we can count on > being broken only very infrequently), such as "the garbage collector will > never use more than 1 msec within any 10 msec period of time". More Tim Sweeney gems at... http://floodyberry.com/sweeney/posting.html#s1999071705210
_______________________________________________ bitc-dev mailing list bitc-dev@coyotos.org http://www.coyotos.org/mailman/listinfo/bitc-dev
