The WIP Thread
-
-
- Posts: 17934
- Joined: Fri Jul 06, 2007 3:22 pm
Re: The WIP Thread
I can't imagine performances would be good if the entire house is a model.
-
-
- Posts: 1384
- Joined: Sun Oct 14, 2012 1:43 am
- Location: Ukraine
Re: The WIP Thread
Give me a house that is a model, even simple one works I'll test and optimize then.
It took me few hours to find this bridge model. I was trying to find some piece of architecture that is:
1) not at turbosquid
2) not worth $$$
3) easily conversible to md3 (made of one texture)
Current version (http://www.mediafire.com/file/p99e18dtx ... lision.pk3) is a bit better than in the video (or than previously posted). Player can properly move on the bridge.
However, it certain situations he can still noclip through (easiest: run against a corner, or crouch, go below the bridge, and uncrouch) and "moving up" is not implemented yet (nor "moving down").
I, also, wouldn't hope for it to be easy. I can already say that for example hitscans would require manual reimplementing to be aware of this system. Monster AI... not sure, actually. Being a simple "walk at the player, then turn around if stuck" it might support models transparently with a bit of A_Chase hacking. Or not.
It took me few hours to find this bridge model. I was trying to find some piece of architecture that is:
1) not at turbosquid
2) not worth $$$
3) easily conversible to md3 (made of one texture)
Current version (http://www.mediafire.com/file/p99e18dtx ... lision.pk3) is a bit better than in the video (or than previously posted). Player can properly move on the bridge.
However, it certain situations he can still noclip through (easiest: run against a corner, or crouch, go below the bridge, and uncrouch) and "moving up" is not implemented yet (nor "moving down").
I, also, wouldn't hope for it to be easy. I can already say that for example hitscans would require manual reimplementing to be aware of this system. Monster AI... not sure, actually. Being a simple "walk at the player, then turn around if stuck" it might support models transparently with a bit of A_Chase hacking. Or not.
-
- Posts: 772
- Joined: Sun May 04, 2014 7:22 pm
Re: The WIP Thread
Threw together a house md3:
https://drive.google.com/open?id=1wSn89 ... mFkGQUawyt
No idea how you did this. It's really very fascinating though. Are those particles spawned at the vertices just for debugging or do they have anything to do with the actual collisions checks?
https://drive.google.com/open?id=1wSn89 ... mFkGQUawyt
No idea how you did this. It's really very fascinating though. Are those particles spawned at the vertices just for debugging or do they have anything to do with the actual collisions checks?
-
- Posts: 8196
- Joined: Sun Jan 28, 2007 3:55 pm
- Preferred Pronouns: He/Him
- Location: QZDoom Maintenance Team
Re: The WIP Thread
Man, that's insanely awesome ZZYZX. So is that a WIP boilerplate code on getting the vertices, or is this still under heavy refinement?
-
- Posts: 1337
- Joined: Tue Jul 15, 2003 4:18 pm
Re: The WIP Thread
So, the core of this is that you basically implemented an .md3 parser in ZScript? That is completely awesome! (even if slightly insane )ZZYZX wrote:Give me a house that is a model, even simple one works I'll test and optimize then.
It took me few hours to find this bridge model. I was trying to find some piece of architecture that is:
1) not at turbosquid
2) not worth $$$
3) easily conversible to md3 (made of one texture)
Current version (http://www.mediafire.com/file/p99e18dtx ... lision.pk3) is a bit better than in the video (or than previously posted). Player can properly move on the bridge.
However, it certain situations he can still noclip through (easiest: run against a corner, or crouch, go below the bridge, and uncrouch) and "moving up" is not implemented yet (nor "moving down").
I, also, wouldn't hope for it to be easy. I can already say that for example hitscans would require manual reimplementing to be aware of this system. Monster AI... not sure, actually. Being a simple "walk at the player, then turn around if stuck" it might support models transparently with a bit of A_Chase hacking. Or not.
A side effect of this may also be that, with some tweaking, we will also be able to use model tags, which was never implemented in GZDoom's .md3 implementation (basically, you can attach a model to another model at a certain point, and have the attached model move in sync with the attachment point - how Q3 weapons are attached to player models, for example).
-
-
- Posts: 1384
- Joined: Sun Oct 14, 2012 1:43 am
- Location: Ukraine
Re: The WIP Thread
I was testing that my code rotates/scales/translates the model as expected. They are purely visual.Jaxxoon R wrote:No idea how you did this. It's really very fascinating though. Are those particles spawned at the vertices just for debugging or do they have anything to do with the actual collisions checks?
It's seven megabytes o_O (forgot to trim excessive frames in Blender?)Jaxxoon R wrote:Threw together a house md3:
https://drive.google.com/open?id=1wSn89 ... mFkGQUawyt
Takes few seconds to load, but then works just fine, lol. Except corner bugs (literally, two triangles making up a corner results in really weird handling).
PK3 updated to look at it. http://www.mediafire.com/file/p99e18dtx ... lision.pk3
Yes.AFADoomer wrote:So, the core of this is that you basically implemented an .md3 parser in ZScript?
AFADoomer wrote:A side effect of this may also be that, with some tweaking, we will also be able to use model tags, which was never implemented in GZDoom's .md3 implementation (basically, you can attach a model to another model at a certain point, and have the attached model move in sync with the attachment point - how Q3 weapons are attached to player models, for example).
Also: currently box<->triangle collision is heavily broken there. I discovered it after trying to rotate the bridgeZZYZX wrote:Note that disregarding the result of this particular work, it can already be used with MD3 tags to add animated effect anchors to models, for example, by the same mechanism as the white dots in the video
Apparently triangle collision always returns true, and it was working with bounding boxes all this time
Vertices work. What doesn't work is collision itself. There is one bug and two things that are just not done yet.Major Cooke wrote:Man, that's insanely awesome ZZYZX. So is that a WIP boilerplate code on getting the vertices, or is this still under heavy refinement?
Last edited by ZZYZX on Sun Dec 17, 2017 3:33 am, edited 1 time in total.
-
- Posts: 13792
- Joined: Tue Jan 13, 2004 1:31 pm
- Preferred Pronouns: She/Her
Re: The WIP Thread
ZZYZX - once again you amaze us with your insane genius. This is awesome work.
-
-
- Posts: 1384
- Joined: Sun Oct 14, 2012 1:43 am
- Location: Ukraine
Re: The WIP Thread
Honestly, would be way better if we had someone who actually understands basic 3D physics sim. My code currently consists of hacks, more hacks, and gross hacks.
For example take a look at 3D triangle->box intersection test https://hastebin.com/ilovucuyen.cpp
Which is, basically, 3 2D triangle->rect intersection tests because 2D is way easier to code.
It's not insane genius, just no one else apparently wants it to be done enough (just like it was with portals, which were just another combination of hacks and more hacks, and only doing 50% of the job got Graf interested in it).
For example take a look at 3D triangle->box intersection test https://hastebin.com/ilovucuyen.cpp
Which is, basically, 3 2D triangle->rect intersection tests because 2D is way easier to code.
It's not insane genius, just no one else apparently wants it to be done enough (just like it was with portals, which were just another combination of hacks and more hacks, and only doing 50% of the job got Graf interested in it).
-
-
- Posts: 17465
- Joined: Mon Oct 27, 2003 12:07 am
- Location: Kuala Lumpur, Malaysia
Re: The WIP Thread
ZZYZX wrote:I recently learned that Graf actually implemented direct lump data access, so using ZScript properly here (that is, for implementing features formerly thought to require being implemented in the engine).
Still under development, player movement not yet implemented (player can already walk on it, but sliding along polygons is heavily broken; this is however a problem not with the approach, but with my knowledge of physics and physics simulation, asked for help with maths at Discord, still WIP :roll: )
White dots are debug output.
Note that disregarding the result of this particular work, it can already be used with MD3 tags to add animated effect anchors to models, for example, by the same mechanism as the white dots in the video :)
@ZZYZX - why not implement this stuff natively into the engine?
Also it's not that no one is interested (I am interested AS FUCK) but nobody has the technical know-how?
-
-
- Posts: 1384
- Joined: Sun Oct 14, 2012 1:43 am
- Location: Ukraine
Re: The WIP Thread
There are a lot of reasons.Nash wrote:@ZZYZX - why not implement this stuff natively into the engine?
First, I'm not in the right mood to create yet another partially-unimplemented branch that will go on and be 1500 commits behind gzdoom/master like it was with portals, then get abandoned until someone wants to continue.
Second, I don't think my code or algorithms used are production quality, and I really like the idea of giving ZScript as much control as possible, so that people can slightly edit it if they need.
Honestly the only reason to put this in the engine that I can see is hitscans. And only because Graf still hadn't exported Trace to ZScript (or he had? I don't know, I know that few months ago it wasn't there).
Pretty sure there are a lot of people who understand it better than I do. Example: Gutawer, dpJudas. Not sure about Graf. Plus there are a lot of obscure modders who've been achieving even crazier things (walking on polyobjects, getting carried by polyobjects) before there even was ZScript.Nash wrote:Also it's not that no one is interested (I am interested AS FUCK) but nobody has the technical know-how?
-
- Posts: 2059
- Joined: Tue Jan 13, 2009 4:13 pm
- Graphics Processor: ATI/AMD with Vulkan/Metal Support
- Location: Somewhere in the future
Re: The WIP Thread
Finished Visor's ability, it will crush all his foes making them taste the floor as giblets, the Gravity Hammer:
-
- Posts: 7402
- Joined: Fri Oct 22, 2004 9:22 am
- Graphics Processor: nVidia with Vulkan support
- Location: MAP33
-
- Posts: 163
- Joined: Thu Oct 31, 2013 3:14 pm
- Graphics Processor: nVidia (Modern GZDoom)
- Location: French Hell (literally)
Re: The WIP Thread
Ladies and Gentlemen, the Sequel of UACSecretFiles.wad
Spoiler:
-
-
- Posts: 16891
- Joined: Tue Oct 02, 2012 2:20 am
- Location: An ancient Escape Shuttle(No longer active here anymore)
Re: The WIP Thread
Aesthetically G r i m d a r kKinsie wrote:
-
- Posts: 1268
- Joined: Wed Jul 20, 2011 4:24 pm