spookmineer wrote:I can hardly believe it is because they are better coders then professionals...?
You're making the assumption we're not professionals too.... coding has been my day job for quite a while now, and arQon even longer. Not sure how long qrealka has been at it, but I'm pretty sure coding is his day job too. Not games, but that makes no difference.
Its not that we're better. Its perhaps that we're more focused on making it faster, so have taken the time to profile it and look for bottlenecks and see if they're justified.
Its maybe a bit harsh to include it in this context, but there is definately code in q4 / d3 that shows to me they were learning C++ / OO principals, as there are some pretty ugly bits there, which I wouldnt really be happy with. Not noticed any of these causing perf issues tho (although the one I hated the most is rewritten in q4max anyway).
Does it have to do with enabling vertexlighting, or does the code really have that much flaws that when improving it, the results are so much better?
The changes that make .75 so much faster than .74 are not lighting related - they're just fixing some sloppy code which is doing a lot of work unnecessarily. On q4ctf1 we found it hitting a rather CPU intensive piece of code 3000 times per second unnecessarily - ie, doing nothing. (figure from memory, but pretty sure thats correct).
The changes arQon is referring to that we're looking at doing for .76 are again an area of code we believe to be significantly slower than we believe it needs to be, and are looking at ways it can be improved. Disabling this code gives the speed boosts arqon quotes with minimal effect on the game (although some, which is why it isnt done in .75).
These are both on top of vertex lighting, which obviously still speeds things up massively.
Oh, and server-IP - well theres always our public server, which is at: q4max.fragstore.net:28004 although theres another 270-odd servers, so you might wanna pick one near you.
Heres a
query.