On Tue, 22 Nov 2005 01:12:32 +0000, "Andrew Sayers" <[EMAIL PROTECTED]> said: > I've just rewritten the gradient algorithm in the CVS head. It's the > same basic algorithm, but using pointers now instead of integers. I've > been working on this stuff a lot in the map-rewrite branch so I *should* > know what I'm doing, but I'd appreciate people looking over > updateGlobalGradient() in Map.cpp. >
Thats great. I knew that the gradient system "could" be faster but i still think a mixed system (mixing gradient, pheromones and flags) could work even faster. I was thinking about a system with pheromonesnad flags, as a continuation of my last email about the network protocol, but i am still not sure of how to implement some details. > With that and -O2, I've managed to get the only 512x512 map I could find > (RMHM_no_Rush_Migration_Huge_Map_by_moa3333) running at less than 100% > processor load. > I never was able to play this map. > As well as being a significant speed increase, this goes a long way to > proving the assumption I've been working on for a while: that most of > the processor time is being used up updating gradients. > Yes, in general with finding the path of units. But also the checksums consume about 0.3 of CPU or so. > - Andrew > > > _______________________________________________ > glob2-devel mailing list > [email protected] > http://lists.nongnu.org/mailman/listinfo/glob2-devel -- http://www.fastmail.fm - A fast, anti-spam email service. _______________________________________________ glob2-devel mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/glob2-devel
