Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
Grimm

Best Nodes Builder?

Recommended Posts

I always use Zennode. It's one of the fastest I know of, and it does rejects as well.

Share this post


Link to post

DeepBsp is very fast, which makes it good to use for use during editing.

I don't have much experience of nodes builders, and I don't know much about the tech, but for release builds, I'd possibly go with Zennode or BSP. (Vrack2b was compiled with Zennode).

Share this post


Link to post

The "best" one is the one that works with your level. Which means that more than one can be "best".

For Legacy, for whatever reason (I think because Marc worked on some issues for that port), Zennode should be used, only because a lot of levels won't look right (cracks and all that stuff caused by GL issues).

For GL ports it has to be GLBSP (if they are coded to look for the new lumps).

BSP usually offers no advantages at all, except historical hype. It's pretty slow. Perhaps the slowest of them all, haven't checked for a while. Since machines are so fast, speed is not much of an issue though, unless you are using an older relic.

WARM can work better in some cases and has it's devoted following.

All just depends on your audience and the level. A LOT of stuff is pure B.S., so try it out on your level and see what is more important for you. In some cases, (large levels, slow machine), it makes sense to use one for development and another for release. If one adds up the time difference it can be quite significant for larger levels + testing.

Vrack2 was DeePBSP and then someone convinced Fred that a REJECT was required<g>. The REJECT is only required for old ~P200 machines. On newer stuff (PIII450 and up), I never found any noticable speed difference at all - so the REJECT is a waste of time - even on Vrack2, most levels are much smaller. The newest ZDOOM will run without any REJECT at all:)

Share this post


Link to post

A REJECT is a good idea on big maps even nowadays, since, even if we have more poweful computers in general, it's not unusual to set up a weaker machine to run DOOM in one of its incarnations.

As for a good nodebuilder, I generally prefer zennode among the ones not built into an editor. I make plain WADs and it is quite suitable for that. I've tried BSP but it tends to cause more VPOs. Still it doesn't hurt to have more than one nodebuilder at hand (it's not like they effectively take up space in your HDD.) On a certain map one may occasionally have a better effect than the other for some reason. If one fails somehow, it can be a good idea to try another one before resorting to modifying the map...

Share this post


Link to post

I use Zennode with the reject builder turned off for development. If I notice any glitches I try other nodebuilders to eliminate them before release.

BTW, when Vrack2 was first released (early 2001), computers were *much* slower than they are today. Typical Doomers were not running something like a P3 500+ I can tell you that. If it were being released today, the reject issue wouldn't even be one. Back then it was though.

Share this post


Link to post
Ichor said:

I always use Zennode. It's one of the fastest I know of, and it does rejects as well.

And I use Warm as a backup too. :) Thinking about BSP though.

Share this post


Link to post

Warm was also built with Hexen in mind, while BSP as far as I remember, was not, which is why it doesn't screw up polyobjs as much.

BTW, reject maps only effect the monsters, it has nothing to do with the actual map at all so really there's little room for improvement on most machines in big maps such as vrack2 since it's just a huge map and will take up memory regardless.

Share this post


Link to post

Except that Fredrik's baby has got quite a bunch of monsters on UV.

Share this post


Link to post
myk said:

Except that Fredrik's baby has got quite a bunch of monsters on UV.


This is true. Is there a noticable speed increase if you run through it with -nomonsters on? I don't notice any, but my processor is 1.7ghz...

Share this post


Link to post
Cyb said:

Warm was also built with Hexen in mind,

There are only 2 HEXEN aware node builders (in the sense that they know about polyobjects). One is DeePBSP and the other is the one Randy just put into ZDOOM. The part that has to be "aware" attempts to avoid splitting subsectors around polyobjects. The rest of the code is exactly the same. Since Randy can directly relate his work to his code, the internal ZDOOM may be the best for this. Don't really know for sure.

To the other posts: A REJECT is a quick first pass "attack" detector for things. Without a REJECT, there's a bit of extra code executed to see if the objects are "attackable" so to speak. So all it means is that this bit of extra code is run - not significant on newer machines. Thus it makes sense that if there are a lot of enemies that's the best test case since that means this code is run the most. Hence Vrack2 is a good one to base my argument on.

Using Vrack2, I've found that a REJECT gives practically no speed improvement on today's machines. Actually that's true starting with the 1st Pentiums for average levels under about 2000 linedefs (which is larger than any stock DOOM2 level). A 486DX4 is no match for even a Pentium66.

In 2001 P200s were already history. A PIII450 was common and for sure the world was mostly PII233 on up (a huge jump from a P200). I went through the exact same exercise/discussion with stphrz - if he recalls - at NewDoom when Vrack2 was released! And if he recalls, it was discovered that it really did hardly anything for speed. Sure a frame or so, but hardly what people claim. All gossip, all word-of-mouth stuff without hardcore figures - almost like a superstitious belief:)

I think I also did the test on a P200 with a 3DVirge card (since I had a working P200 still) and even there it made little difference. There are other factors at work for Vrack2 that are much more significant then the REJECT - one being just the sheer level size and the amount of computation required to traverse it.

Share this post


Link to post

I remember that discussion back then. Granted, on a powerful machine it won't make a difference. But when Vrack2 was released my main PC was a P166 with a crappy video card (2 MB Trident card) and with the added REJECT it increased from about under 20 FPS to 25 FPS. My point being, there's nothing to be lost in adding a REJECT, so I'd always add it, especially on a map with lots of monsters. No point in not covering all the bases if it isn't a hassel to do so. I still have that P166, by the way. It's not uncommon for DOOM players to keep old processors around since they run DOOM quite well and are useful for a possible LAN or network testing (I already said this below, really.)

Share this post


Link to post

I built Vrack 2 on a Celeron 300 with 64MB RAM. Every time I made a physical change to the level and wanted to playtest, I had to wait 45-60 seconds for the nodes build to complete. When playing the level in ZDoom in 320x200, I'd get ~25 fps all-around, with Doom 3-like slowdown upon entering a detailed room.

After releasing Vrack 2, I soon started work with Vrack 3. Unfortunately, after having built only four-five rooms, I had the compile times that Vrack 2 had when it was near completion. I'm of such a nature, that every time I change the position of a line, I have to check it out in-game to see what it looks like. Building a whole level that way would have been extremely frustrating. Vrack 3 was one of the main reasons why I hurried to get a new computer.

So, I bought a new computer... a Thunderbird 1400 MHz, 512 MB DDR RAM, and other goodness. Vrack 2 ran with a fluent 35 fps in 640x480, rebuilding the nodes with DeepBSP now took something like 6.41 seconds.

The decision to release a second version of Vrack 2 wasn't really about improving speed with a REJECT map, I wanted to fix some bugs (mainly the sky setups in some rooms). But I did a REJECT build just for the hell of it - it took something in the order of 2-3 minutes - the only catch would be an additional zip size of a couple of 100 K.

I didn't find that the REJECT made any difference on my new computer, but other people reported that it did (though they could've been making that up). Anyhow, testing Vrack 2b on my old computer, I *did* notice a speedup, especially near the beginning of the level. 50% increases in performance were later reported, and while I have my doubts about such a figure, I don't think it was wasted to build the REJECT.

Share this post


Link to post

Since I mainly do maps that are compatable with DOOM2.EXE, I use BSP. The WADED internal nodebuilder is horrible, and likes to screw up large rooms. BSP has always been relatively fast for me (probably because I am still not yet to the point where I have a ton of linedefs) and has always yielded good results. My system specs, by the way, is an AMD 300 mhz, 64 megs of RAM, and I'm quite happy with the setup I have.

I even have a nifty batch file set up that BSP's the WAD, runs it in DOOM2.EXE -warp 01 and then goes back to WADED on that same map. Very nifty. ^_^

Of course, I do not mess around with source port specific features, and probably won't do so in the near future, so your milage may vary.

Share this post


Link to post
Fredrik said:

Anyhow, testing Vrack 2b on my old computer, I *did* notice a speedup, especially near the beginning of the level. 50% increases in performance were later reported, and while I have my doubts about such a figure, I don't think it was wasted to build the REJECT.

Thanks for clearing up the details. I have 100% doubt about those reports since I spent quite some time doing detailed benchmarks. What you have to remember (and most people may not realize), that once you test a level you have just "primed" your machine for playing that level. IOW, the 1st time you start playing it takes significantly longer since Windows actually allocates real memory, pages the other stuff out, loads DLLs and so on. The 2nd time this is not required. Easy to see. Load MS Word the 1st time from a boot. Now exit and load it again. See the difference?

I agree it's not a waste of time as such for large levels, although for small ones it is. I always disagree with the figures claimed:) Levels large enough to see a difference won't be played on machines that show the difference. IOW, Vrack2 is not fun to play on a P200, REJECT or no REJECT. I think it ran at 5fps or so (not sure I remember correctly though) - that's not playable. I never noticed much difference on the P200 between the 2 versions.

However, like I said, it's whatever works for one's level. If one wants to put a REJECT in do it - but don't expect miracles. I just know it's not 2 times faster<g>. Many people may not realize that build times (depending on how one creates levels) can be significantly reduced by choosing the right node builder and not doing a reject during development. Then in the "release" do whatever one thinks is appropriate. The time differences can be very large and add up to a big deal if the levels are large. IOW all this depends on one's machine, level and editing style.

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×