CPU Usage. in BYOND Discussion

ID:135586
 
Jun 1 2004, 2:13 am
Why has Dreamseeker been resently taking over half my CPU Usage PER dreamseeker? It used to take like 10% for 2 put together. Can someone make sense of this?
#1 Jun 1 2004, 2:24 am
Metroid wrote:
Can someone make sense of this?

Poor programming.
#2 Jun 1 2004, 3:15 am
In response to Goku72 (#1)
Regular ticker with spawned() stuff. If your ticker event takes longer than your ticker delay, you get a build up.

50%? Bah. I once saw a game eating a full 98% of a CPU. And that was through a Linux shell command, not the odd CPU var.
#3 Jun 1 2004, 3:26 am
In response to Hazman (#2)
Yeah, Aaiko hates hosting Seika 2 because his CPU Usage jumps to 100% and doesn't wanna move from it. And when someone joins, you can't do anything for about 10 minutes.
#4 Jun 1 2004, 3:30 am
In response to Zaole (#3)
That's actually painful. If it jumps to 100%, it's probablary because of several infinate loops, several NPCs and an inefficient character generation and world/New().
#5 Jun 4 2004, 4:31 pm
In response to Hazman (#4)
Just wondering, could it possibly because I have loads of NPCs that wander randomly?
#6 Jun 6 2004, 7:09 am
In response to Metroid (#5)
Yes, just use step_rand().
#7 Jun 6 2004, 9:32 am
In response to Hazman (#6)
I do.
#8 Jun 6 2004, 10:06 am
In response to Metroid (#7)
Then, theoretically, there shouldn't be a problem. Just as long as you don't have a while(1) loop or anything like that.
#9 Jun 6 2004, 10:09 am
In response to Hazman (#8)
well actually I recently changed it to
proc/met_step_rand(atom/movable/a)
    step(a,pick(NORTH,SOUTH,WEST,EAST))

mob
    monster
        proc
            RandWalk()
                spawn while(1)
                    met_step_rand(src)
                    var/B = rand(20,60)
                    sleep(B)


But that should basically do the same as step_rand without the diagnal walking.
#10 Jun 6 2004, 6:46 pm
In response to Metroid (#9)
Metroid wrote:
well actually I recently changed it to

proc/met_step_rand(atom/movable/a)
step(a,pick(NORTH,SOUTH,WEST,EAST))

mob
monster
proc
RandWalk()
spawn while(1)
met_step_rand(src)
var/B = rand(20,60)
sleep(B)

But that should basically do the same as step_rand without the diagnal walking.

Nope, not quite. step_rand() uses an edge-following system which makes the AI only change direction at semi-random intervals or when the next step would take it into a dense object. This usually produces a more "aimless" random movement system, instead of a "mindless" random movement system, but has a disadvantage that, unless you force the AI to pause and change its direction manually, it can -- possibly -- eventually wind up just walking in a fairly continuous loop around the edges of the map.

Deadron has his own routine, dd_step_rand(), which provides an excellent mix between random mindlessness and deterministic aimlessness -- and offers more parameters than the typical step_rand() proc, allowing you more flexibility. I'm not sure which library it's in, however, so you might have to do some digging.
#11 Jun 8 2004, 8:49 am
In response to Spuzzum (#10)
Actually, I think its more of a Virus, for the CPU problem, Because recently, ive actually ADDED more monsters that use my rand_walk, and the CPU has gone down 20%. Either way, wehn I get home ill check more...