Without wishing to sound rude but you've not played with Bilinear Spatial Partitioning or any other similar method have you?
LFS is using a render tree of some description, most games do, specifically it is using the route of the track and a pre-rendered list of track segments visible along the route. If you leave it to the "engine" to decide what to draw (and bearing in mind here LFS' engine was written by Scawern specifically for LFS) then I pressume you are using a generic 3D engine that isn't optimised for a particular game, or you are just drawing everthing within camera occlusion - and consequently paying the performance hit for what I would describe as a sub-freeware quality 3D system - because yes, freeware programmers are doing stuff way more advanced than that to cram more pollies into their games.
Some form of visual occlusion is needed, whilst camera bearing position and range is ok for a quick mock up, most 3D games will go beyond this and optimise the render tree specificallly with the gameplay in mind - this can hugely boost performance, and is one reason why LFS' high polygon graphics can run on such low specification machines... If LFS did not do this our cars would look considerably more square and are tracks would be less detailed.
yes i have but since you are not a fool you will use BSP for view frustum culling, NOT for rendering because since the invention of graphic cards, this is a task performed by the hardware via graphic drivers + directx/opengl
[sarcasm on]i'm glad to know you know for sure how lfs is implemented...you must be the fourth misterious dev of scavier, so it must be scavierbe then!!![sarcasm off] <- don't take this too seriously
please explain me how can you prerender a 3D scene and still being a 3D game the way you depict it seem like the old arcadish 2D race games simulating a 3D visual...
i used opengl + my code (like scawen is using directx + his code), i didn't developed "games" but rather "demos" for special engineering purposes.
I'm also glad to see you know very well my code and so deeply that you can judge even its quality...
yes detecting objects into the viewport is a standard tecnique and yes it can be optimized for specific games (i.e. lfs and paths), but the latter point does not mean that lfs cannot run a "free run" mode infact that's my own quote:
as you and some others describe "autocross" is not possible in lfs and this is not true. Paths my help to speed up the view frustum culling but but they are not a must and this was true since S1 and more true with the latest test patch and i add that even multipath is not a problem since it only overload a bit the bsp-tree (or whatever) in the neighborhood of two or more paths joining...again why saying "it's not possible" while even now it's possible? why reading scawen's words with a so close mind? i cannot understand, really!
Actually Scawens words were 'It seems possible', and the rest is basically 'but we can't at the moment'.
The two together form, as far as I can tell: It seems possible, but we can't at the moment'. In other words LFS does not support multiple routes yet, although it might be a feature in the future (though I really hope it isn't, as this is a race sim, and in racing you don't tend to have route choices as you'll end up with cars going in opposite directions).
I'm sure there is a better post from Scawen, but RSC is so slow and requires a months wait between searches to save it's precious (read like Gollum) bandwidth which is rather annoying.
I wont, I am merely refering to Scawen's post where he explained how LFS did this. I believe he has mentioned it before also. And as Tristan says above (and if I recall without checking in this thread Scawen too said (or maybe it was another)) Scawen has also said that LFS could be made to do branching, but that it currently does not.
You really don't understand what I said did you ! I understand what you are saying, but it's about as closely related to what I said as modern human is to a T-Rex.
If you are leaving the "engine" (ie: OpenGL or Ogre or whatever you use) to decide what to render as you describe then you would indeed need to learn more before attempting a realtime 3D application such as a game unless you only want users with very powerful computers. One of the first things hobbyist game coders learn after producing their first few simple games is how to optimise the render list to increase performance of their games. It's one of the first skills starting indi coders learn, without it games have much lower poly counts or very low framerates. Even I do it and I dont confess to be good*.
*I leave that for others to judge in the smug knowledge that i'm flippin' brilliant.
yes he said that paths are a problem, yes he said he could work on multiple paths but not urgent now, you forgot that he also said that as it is it can work by removing the path from track, just like in autocross, but that means no AI and no wrong-way detection, so we are saying the same thing except the autocross thing
scawen also said that right now he could remove paths but that means no AI and no "wrong-way" -> but who really cares if we can have open tracks right now?
you said "prerendering" and that is a nosensebecause "rendering" is the process where a 3D scene is drawn into a 2D picture/frame
obviously you didnt read my post i said 2 times that view frustum culling is a standard and common tecnique to speed up things i never said the contrary, but i have also said that bsp-trees are used by game devs for view frustum culling and not rendering, once you know what objects you have in you bounding box/sphere, then the rendering is a task for your graphic card + opengl/directx, otherwise what is the purpose to have 3d accelerated graphic cards if you do software rendering? only old games like quake and unreal had software renderers because at that time 3d acceleration was still young, but nowadays what's the purpose to reimplement opengl or directx like things?
again fot the n-th time: i don't say to pass all objects to directx/opengl!!!!! but that's a different thing from rendering (the visible scene) -> this is a task for directx/opengl!
btw i dont understand why you brought up those things into discussion, i only said that the myth "lfs can only render what's ahead and what's behind, so it cannot support open tracks" is a clearly false statement for a 3D application either for a graphic developer geek who knows all the underneath precesses involved, but either for a common sense guy despite of more indeep knowledge.
iagree that my first post was too concise and without any real tech explanation, but teh post before meine said
directx/opengl can render anything despite of it is in front, behind or at your side! since scawen is not a mad to reimplement directx from scratch and we also know lfs uses the directx rendering engine, that quote is obviously false! guess what? autocross has no routes, but lfs still render autocross track....so how could it be possible? omg, omg, how could it be?
PS for the (n+1)-th time: i'm not claiming that lfs is not doing view frustum culling of the scene...having a path can speed up this task, but it's not that "if we don't have a single path, lfs cannot exist"
okay, just stop and think for a second, if you were to ask scawen 4 years ago what he thought LFS would be like, i bet he wouldn't think it is what it is right now, so i wouldn't doubt the devs; atleast we know its possible to have open courses, since EA and Rockstar Games proved to the world that its possible
so i say enough with this thread, because this isn't going to get anywhere, and it is possible, it just may be that scawen doesn't know how... there.. everyone happy now?
the first one is much less technical and much more boring, the original one let more skilled people to gain some tenths and make "safe but still difficult" passes... -> -1
the second definately +1, but i dont' have the feeling we can get it soon, i cannot say why...it's just a feeling...
FYI Autocross doesn't use the path system in quite the same way, it's a not culling as much (only stuff off to the sides based on how far along the 'route' you are. However doing this in a track with a lot more polygons 'at this time' wouldn't work, unless Scawen adapts the path system to support multiple routes. The autocross arena isn't using multiple routes, it's just a low polygon area in the first place so doesn't need much culling.
I quite agree, and so has Scawen when he's spoken on the subject before. All I did was explain why we dont have multiple routes 'at this time'. Of course it is possible to change this, and Scawen even expressed interest in doing so. Don't shoot the messenger please.
as long as we know it is possible and it will be implemented (even thou it's not the very forthcoming thing), i'm happy with this.
i really don't care of the deepest tachnical details of lfs, i just look at the results
[added in patch 0.6B] 'Open' circuit option for practice
Most circuit maps have several routes for different courses, but they are always blocked off. It would be good to be able to drive a circuit without being boxed in like this, even if its just in practice mode.
With the increasing popularity of the various cruise servers, which are great fun if you just fancy driving without the pressure of racing it would make sense to be able to choose a course that has all the routes open. Where they lead off the map obviously they'd need to remain closed off, but it would add a new level to the cruise servers if you had a few different paths to take.
try searching next time, old suggestion, Scawen has commented that the graphics engine renders ahead down the path, so with multiple paths he'd have to change it to render all paths, which would incur a pretty substantial performance hit.