Zugg Software :: View topic

Newbie Joined: 11 Apr 2006 Posts: 9

I feel the automap calculation in CMUD is considerably slower and more CPU-consuming than in zMUD. A quick test shows the speed of retrieving a path is about 10% slower than zMUD721, 85% slower than zMUD555.

Could Zugg help check what is the reason?

Cheers

Wizard Joined: 03 Mar 2001 Posts: 1127 Location: London

The mapper hasn't yet been optimised for CMUD from what I understand. It's something that Zugg will work in later in the beta or after the first public release.

Posted: Tue Oct 31, 2006 5:31 pm

It might also be a problem with your map file. The zMUD 7.21 path finding is *much* faster than v5.55 normally. But your numbers seem to be saying that 5.55 is faster. Keep in mind that 5.55 had a kludged path finder that often would find the incorrect path. So I wouldn't really trust it. It also depends on things like zone links.

But CMUD uses the same mapper code as zMUD 7.21, so it should really be the same speed. 10% isn't much of a difference.

Newbie Joined: 11 Apr 2006 Posts: 9

Well I would also like to believe that but the test just keeps confusing me. I drew a simple map containing 20 rooms (1-20) in line and test the path retrieving speed like this:

#tel 1
initial=%secs
#loop 2000 {%walk(20)}
#say %eval(%secs-@initial)

zMUD555 gives the result of about 360ms while in zMUD721 it is over 6000ms (Cmud not test yet which should be the same with zmud721). Of cuz executing 2000 commands may cost some time itself but does not matter a lot in fact (just costs 100-200ms in total). ps the conditions of test: zmud offline, CPU P4 3Gb.

Sorry about mentioning this again but I hope some1 could help with this case.

Cheers

Posted: Wed Nov 08, 2006 8:52 pm

As I said, the zMUD 5.55 path finding was broken. For simple paths like that it's going to be faster, yes. But for more complicated maps, zMUD 5.55 is going to return the WRONG PATH, and sometimes even a different path each time. It depends upon the complexity of your exits, the presence of zones, etc. zMUD 5.55 used a very simplistic path algorithm that was fast but wrong. Later versions of zMUD use the proper path finding algorithm (which is a very known algorithm in computer science) and optimizes it.

So you are comparing Apples and Oranges. I can come up with an even faster algorithm than 5.55...an algorithm that just returns a constant result. Sure, its faster, but that doesn't make it right. It's more important to most people that zMUD calculates the CORRECT path.

Wizard Joined: 14 Aug 2004 Posts: 1269

Do you actually have any real-world examples of slow path-finding or is just an academic exercise?

Newbie Joined: 11 Apr 2006 Posts: 9

Thanks, Zugg. You did make this clear. Because I never met a wrong path in my map (about 3800 rooms) I had that problem.

Seb, yes my initial test was based on my real-world map, maybe it is just not big enough.

Posted: Thu Nov 09, 2006 5:23 pm

Rivo, that might be true, but trust me, this was a big issue back in the v5.55 days (which was many years ago). The path finding problem was solved several years ago, so this is really an old and dead issue at this point. As I mentioned, it really depended upon the complexity of the map, the number of zones, one-way directions, etc. But it was definitely a big problem that I used to get lots of email about.