Bikerdude: - questions

[Bikerdude]

i found with rift that 'over monster clipping' make it larger

I have have been getting a bit carried away doing that thinking it would help...so i shall prune down the amount ajnd see how I go..

[LEGION]

Well sweet Jesus

 

The 4GB file patched thing appeared to work - D3 peaked at 1.35GB and completed a single Dmap/Map compile session...!!! So next Im gonna have DR loaded(with said map@1Gb footprint) at the same time and see if it crashes...

 

Wow, I wasn´t aware of what is going on in here! It seems my post about that patch some time ago wasn´t in vain, if you, Nils, used that post as a source?

Just in case I already patched DarkRadiant.exe.

 

I don´t wanna read all here, I´m too afraid of spoilers, but it seems Bikerdude, that your map is gonna be the biggest beast the (TDM-)world has ever seen or what?

 

Remember you can set some AI neverDormant 0 which means they will freeze if not alert and they are outside the player area. This is suitable for things like static sentries but also many patrols. You just need to test them and if any look silly because they are always in the same place when you enter then remove it or possibly modify the path. It's surprisingly unnoticeable in many cases unless the player is deliberately looking for it. There were seven "neverdormant" "0" in Heart and nobody noticed - or at least, nobody commented.

Fidcal, a nice side-affect was in the basement-area there were sometimes no AI´s to be seen, that raised my impression that they are going extra random long patrols, also that added more variety and made it harder to forecast patterns. Of course the player could also think: "Damn, they must be stuck somewhere".

[nbohr1more]

LEGION: ungoliant quoted you, then I quoted ungoliant

 

I'm a triple thief (if you count my TDM activity).

 

 

If Bikerdude makes his Halloween deadline the TDM players will owe you a little thanks for this one as well

[Baddcog]

I have have been getting a bit carried away doing that thinking it would help...so i shall prune down the amount ajnd see how I go..

 

 

yeah, it's fine to do just a wall to seperate off limit decor areas. filling that whole area with clip just bloats the size.

[Bikerdude]

Question:

Why does the FPS drop all over the map when both maps are combined..? eg I am seeing a 10+ fps drop in the cathedral.

On the file sizes -

 

Ok, did the "killMonster" command and the fps remained unchanged, so what else could be easting the FPS npw that I have merged the two maps..?

[Baddcog]

probably dumb question but what's the visportal situation like?

 

i assume they are all proper/working... there's a tight corridor between map segments? or a large portalled skybox area?

[nbohr1more]

Biker that's "killMonsters"

 

Note the "s"...

 

That's what you meant right?

[Bikerdude]

yup

[Bikerdude]

i assume they are all proper/working... there's a tight corridor between map segments? or a large portalled skybox area?

All VPs are working correctly and I have optomizes as much as poss in that dept, but... there is one big outdoor area that has VPs, namly the town. Wither map was fine on thier own, its only when I merged to the two did I see a massive drop..

[nbohr1more]

Did you check your caulk where the maps merge?

 

...and leak check?

 

(yeah obvious stuff but that stuff keeps stumping folks).

[Bikerdude]

they merge via tunnel, I will check with "r_showtris" to see if anything is being rendered outside my closed VP's

[Fidcal]

Probably lots of things going on that are multiplied with larger maps. For instance, if you have a proximity trigger like a trigger_once, waiting for the player to enter that brush volume then the code must presumably keep testing that say every 1/N second. Same with speakers; is the player within the max distance? Same with objective 'let player be in EscZone' and 'place object x at position y'. A torch, candle, etc has a fire stim with a certain radius so the code must test if anything is within that range and if so, test if it has a response to that fire stim. Probably dozens of other things. I don't know to what extent these are switched off if the player is in another portal area. But I'm thinking probably not because you could have some event taking place anywhere in the map. It might be an AI walking along (even if you use killmonsters, the code that is testing a damage stim doesn't know that no AI can come along because there aren't any - it just keeps testing. Maybe a trigger timer or pendulum swinging a lit candle near another unlit candle. The second must still ignite even if the player is a mile away. What I mean is, the candle might be static but the code has to keep testing because it doesn't know that another candle might come close for any reason; it just keeps blind testing.)

 

All kinds of things are done in the code every fraction of a second. Maybe that sound location thing keeps testing I don't know. Each one is only a tiny test but they all add up and the bigger the map the more time they all take. It's not just a dead map out there and only the part where the player is active is 'alive'. Unless you have a big custom lagger you can think of like a big s & r or objectives brush there might not be much you can do about the routine stuff.

[LEGION]

LEGION: ungoliant quoted you, then I quoted ungoliant

 

I'm a triple thief (if you count my TDM activity).

 

 

If Bikerdude makes his Halloween deadline the TDM players will owe you a little thanks for this one as well

 

From the looks of it I´m not sure with a Halloween-release. Anyway, I take the liberty in advance to demand THE CAT, she wasn´t there in St.Alban, wasn´t she, Biker?

[Bikerdude]

• For instance, if you have a proximity trigger like a trigger_once,
• it might be an AI walking along (even if you use killmonsters)
• pendulum swinging a lit candle near another unlit candle.

• Ok removed all my triggers
• Ok, so essentially the killMosnters command is a bl;oody waste of time, I will have to export and remove Ai to check it its Ai then..
• Hmm i have some in the cave, will remove them and see if thats the case

[Fidcal]

Perhaps I didn't explain very well. If you have put in a dozen triggers it's probably irrelevant. Triggers are just an example of hundreds of events taking place. Try deleting all your lights (except ambient_world) if you want to test some of them. Try deleting all moveables. I'm just guessing here but it seems to me that the code must keep testing all these things many times a second. Why? Because suppose a timer, a script, whatever, deletes a crate with a moveable bottle on top. How does that bottle 'know' to fall? And if it falls, how does it 'know' when to stop? I'm guessing that the code tests every moveable over and over continuously to check for collision.

 

Trying killmonsters was not a waste of time. It seems to prove that your AI are so well optimised that removing them made no noticeable difference. This is presumably because of reduced distance thinking and dormancy.

[Bikerdude]

hey Fids

 

Well as the saying goes "shit on a stick, it works"

 

Turns out it was just the triggers (has some in the two gardens in the mission when when the player walked through etc), so as they werent really need I have left them out..

 

My fps in the cathedral is backup to what it was and it should I imagine improve fps on the town part of the map also :-)

Edited October 15, 2010 by Bikerdude

[Baddcog]

curios, how many triggers are we talking about?

 

they can be very useful so it would be good to know limits.

[Bikerdude]

curios, how many triggers are we talking about? they can be very useful so it would be good to know limits.

 

thats the thing, there were only 6 in the twon part of the map (in the gardens) and Im sure I set the check times for no lower than 0.5-1 second.

[Baddcog]

hmm, 6 triggers in a large map killing fps doesn't sound good at all. My map will probably use at least that many.

[Serpentine]

Interesting, any specific type of trigger/spawnargs? performance cost seems a bit insane

[Bikerdude]

hmm, 6 triggers in a large map killing fps doesn't sound good at all. My map will probably use at least that many.

Interesting, any specific type of trigger/spawnargs? performance cost seems a bit insane

 

Ok I got my figures a little wrong, the FPS is only up by 5 fps not 10.

 

The triggers were the same ones used in the training mission where the player climes down a wall of ivy.

Edited October 15, 2010 by Bikerdude

[Baddcog]

still, would be good if we didn't get fps drp due to triggers.

 

not really sure what can be done though. Obviously larger think times help, and it seems no matter what those timings have to be there. I guess even S&R is gonna use some timing to check.

 

I guess options would be a 'frob' trigger. ie: door or object.

[Bikerdude]

good the sooner we get the source code so shit like triggers can be offloaded on to other cores..

 

But that said there are a buck load of triggers of all type in the training map and that wasnt effected. Its the styles and size of my map, it has large open spaces with lots of VP's which dosent help me one bit.

 

[Tels]

Biker, could you please create a testmap with your triggers in it? (the exact same type and settings)

 

I have a few ideas why the triggers are slow, would be easier to test and then give you tips what kind of triggers to use instead.

[Bikerdude]

Biker, could you please create a testmap with your triggers in it? (the exact same type and settings)

 

Sure, I`ll pm you a link with an older copy of the map.