Rudl
#4420 posted by inertia on 2008/02/17 07:09:14
Shot #2 reminds me of the factory from the movie "Joe and The Volcano."
CDA Fullvis Runtime
#4421 posted by JPL on 2008/02/17 08:54:44
Just curious JPL, what was the processor?
I have an AMD Athlon 2600+ with 768 MB RAM... it is an old configuration bought 3 years ago, and CDA was released in October 2005... So the gap I could obtained with a brand new processor is not that obvious...
I don't think it is insane to wait for 1218 hours to see your "baby" out, when you spent 7 months building it... It was worth to do, I don't regret, cause as of today, I consider CDA as my best map (visulally at least...)
Also, aguirRe was very happy that I used his vis tool, and that CDA proved it was supporting loooooong run, and was not buggy at all at the end ;)
#4422 posted by rudl on 2008/02/17 10:05:31
I think I have to wait for a multithread vis tool to vis it.
#4423 posted by JneeraZ on 2008/02/17 11:26:20
Well, I posted this before but here is my code for multithreaded LIGHT and VIS that run on OSX.
http://wantonhubris.com/SrcForJim/
If someone wants to make that work on Win32, go for it! I'm sure the community would love you to pieces for it.
#4424 posted by JneeraZ on 2008/02/17 11:28:17
I know that multithreading VIS gives you about a 30% speed up with a second processor in your machine so even with a dual core machine, you'd cut that 1218 hours down to 852. Still an assload, but much less of one.
And if you could find someone with an 8-core machine, well, that should give an improvement that would almost bring it into the realm of reasonability. :)
AguirRe's Vis Tools
#4425 posted by than on 2008/02/17 11:42:52
Automatically save the vis data every now and then, so you can stop vis and resume later. I've never had to use that particular feature, but I can't imagine JPL or Kinn not taking advantage of it... imagine if a thunderstorm made the power go out 45 days in!
My maps have never taken more than an hour or so to vis, and my old maps took around the same time to vis on a p200 with 64mb ram. I, however, am not throwing around these big open spaces quite so much as other people seem to be doing these days. If you look at Marcher, you will discover that the main outside area of the map takes up half the available area for making a map in Quake. It's fucking huge.
I wonder how long czg_hate was taking. Care to enlighten us, czg?
Ch, Chek...
#4426 posted by distrans on 2008/02/17 11:54:27
...skill set beta locked and loaded. Now we wait :)
Rudl...
#4427 posted by distrans on 2008/02/17 11:55:55
...computer panels as exterior structural elements just doesn't make sense, even in Quake.
Wow - Good Discussion!
#4428 posted by RickyT33 on 2008/02/17 16:15:02
Several things:
1 - Somebody PLEASE do a multithread vis for windows. All of you Mac users are getting off on this and rubbing your cocks in glee. We need to make a stand!!! I would do it, but I aint no computer science student, I cant even do QuakeC, and I dont know where to start with compiling from Willems source code. (Did you create that youself BTW Willem?)
2 - Than, your too smart to need a long compile time. Your maps are soo fucking good that you dont need to resort to trying to blow peoples mind with the biggest wide open space. Your maps are just classy.
3 - JPL, Sickbase took 6 and a half days on an Athlon XP1700! CDA would have taken longer on my machine! What did you do for 2 months, aside from tear your hair out, and try to block out the sound of the CPU fans?
4 - Willem, Sickbase was not nice to Quake. Its not nice for anyone! For starters you cant play it, cause it only runs in AguirRes modified GLQuake, or his NehWarp engine is better. I dont think its OSX compatible ATM. Also, the only person who managed to do a Nightmare Hundred Percent run was Sielwolf, who, from what I can gather, has been in hiding ever since. And I have the demos on my hard drive(s), but he told me not to post them. He didnt say anything about not emailing them though....
What I mean is its too hard for you, Willem.
5 - Rudl how long is Fast vis? What is your processor and what vis tool do you use?
#4429 posted by czg on 2008/02/17 16:46:40
All of you Mac users
All two of them?
RickyT23
#4430 posted by JPL on 2008/02/17 18:06:49
JPL, Sickbase took 6 and a half days on an Athlon XP1700! CDA would have taken longer on my machine! What did you do for 2 months, aside from tear your hair out, and try to block out the sound of the CPU fans?
What I did ? nothing with my home PC (except internet and email accesses), but fortunately I installed QuArK / Quake / etc... on my office laptop, in order to start something else.... Did you think I was just waiting for 1218 hours in front of my PC screen doing nothing ? Come on, be serious ...
#4431 posted by JneeraZ on 2008/02/17 18:27:50
" (Did you create that youself BTW Willem?) "
I didn't do a whole lot in reality. I just fixed up the function calls and type names that Carmack already had in the code so that they worked on OSX. I don't know/think it would be that easy on Windows but it would certainly be worth someone's time to do it! Many people would benefit.
I Dunno
#4432 posted by ijed on 2008/02/17 19:30:48
But if a map takes more than 24 hours to compile there should be a good reason for it. Marcher had probably the biggest area yet in Q1, outside of a test or beta map and it took a week.
Noclipping around you can see how well optimised it is, whereas eg. Sickbase wasn't so much. I'm not having a go at you Ricky, because the improvement in the Hand was obvious, in all ways.
I think its worth remembering that good mapping isn't good visuals and gameplay alone, but also the technical side of using the tools efficiently to produce the best results.
I know alot of the old Quake sites are 404 but even so a Google for bsp tutorials will still turn up a lot of the old stuff which was thought through back when people were compiling on 486's. Granted we've moved on in the amount of processing power available, but that's not to say that the old methods are now invalid.
#4433 posted by JneeraZ on 2008/02/17 20:22:00
I agree. It's still the same engine at its core and since nobody is rewriting QBSP or VIS, the old techniques and optimization rules still apply.
#4434 posted by rudl on 2008/02/17 21:14:43
5 - Rudl how long is Fast vis? What is your processor and what vis tool do you use?
Several minutes.
Processor is an AMD 6000+ X2
and I use aguirRe's tools
?
#4435 posted by distrans on 2008/02/18 00:04:17
Marcher had probably the biggest area yet in Q1, outside of a test or beta map and it took a week.
*cough* qte1m1 and m2 *cough*
Aardappel
#4436 posted by aguirRe on 2008/02/18 00:58:00
made already in 1998 a DM level Kasteel that allegedly took five weeks to do a level 2 vis on a sun ultrasparc 300MHz.
Some years later I did a level 4 vis of the same map in three days on a PIII 600. On JPL's system it'd probably take less than a day now ...
#4437 posted by gb on 2008/02/18 01:06:56
Bambuz: Thanks, I got gimp running fine now. Xnview is rather good, but the convert part of it is badly documented. With gimp, at least I know how to use it.
The problem is, even when I raise brightness by 50 and contrast by 35 percent, you can't really see any more in the shots. :-/ A tiny bit more, perhaps. It may have to do with the fact that I'm using software rendering with its smack-dark shadows. GLquake looks almost drastically different.
I'm also using tyrlite which seems to enhance the shadows, not the brightness, which I like for this map... using Lord Havoc's light tool for example, it's radically different, almost a radiosity effect like in Quake 2. I could play with different falloff settings etc I guess, but if I did just for the screenshots, they would look nothing like the actual map, which kinda defies the purpose :-)
When it's time, I'll organize some good-looking shots.
Screenshots
#4438 posted by Spirit on 2008/02/18 09:15:18
Do not tinker around with Brightness and Contrast, that will look like poo (or take too much time). Instead use the layers -> curves tool. You can save the curve for later usage to save more time.
Multithreaded Vis
#4439 posted by inertia on 2008/02/18 09:54:57
Seriously, if you guys code up an EM64T-compatible linux multithreaded executable, I'll run it on our supercomputer. I'm curious to see compile times!
Sickbase Had Really Bad Speeds.
#4440 posted by RickyT33 on 2008/02/18 11:43:25
Practically 0 vis-blocking.
I'm learning still. New map also hs some high detail wide open bits, but I'm trying to bend all of the corridors leading away from it. In Fitz I type r_showtris 1 then run around the map, seeing where the slow spots are.
Rudl - I think my 6 and a half day spout was a 6-7 minute fast vis. I think.
Inertia - trying to get people to code things for us is not easy. I ask all of the time. The ones who are capable just dont listen. I guess they think I'm lazy or ignorant or something. I'd like to think that some coders are working on some new stuff which we would like, but are saving it for a surprise.
I'd like to hear what a supercomputer can do to vis times too!!
AguirRe - I'm glad I didnt make Sickbase on a p133 with 16mb ram.
Distrans
#4441 posted by ijed on 2008/02/18 13:01:23
I forgot - need to play again.
#4442 posted by JneeraZ on 2008/02/18 18:23:33
Hey would someone be willing to send me one of these ridiculously huge VIS time maps so I could fool with it in ToeTag? I would need the WADs as well of course.
If not, that's cool, but I'd be curious to see what could be done to improve VIS times without huge restructurings.
Willem
#4443 posted by Orl on 2008/02/18 18:31:31
I have a map that takes over 30 hours to complete. Do you want just the .bsp and wad file?
#4444 posted by JneeraZ on 2008/02/18 18:37:27
Well, I would need the MAP file so I could look at the brush work. If you're willing to send that and the WAD files to me, that would be awesome.
willem at wantonhubris
|