Oh no. I wasn't suggesting getting the source from Take 2. There are thousands of games out there, many which look/play exactly/mostly the same and very few weren't coded from scratch. If someone can make Wolfenstein/Doom/Quake/Unreal/Halo from scratch (ok, _mostly_ from scratch) I don't see why I can't make another Real Time Tactical game from scratch. I'm just looking for discussion from programmers (preferably ones who haven't seen the Myth Source code) about the reasons the Myth engine fails to scale.
As for the features, mipmapping I think would have actually made the Myth Engine faster, as it would allow distant units to use a pre-filtered lower level of detail image, and I am surprised that Bungie didn't use it in their Engine, but I'm guessing since the additional images would have to be included in the unit collections somehow adding one third to their size and alot to their complexity, they just said to hell with it.
The 3d audio chat was on the original list of things that was supposed to be in the game, since 3d audio was already in there. I think they just realized that most of there customers/players were going to be on 56k modems at best and opted to skip this feature too.
The no skybox thing was just plain lazy. If there was a completely free z (or y depending on how you label the axis) the camera could fly over the units wrecking the illusion of the unit/scenery billboards. By fixing z to the C/V arc, like they did, the units and scenery only had to be rendered from one particular elevation to look good at any Zoom setting, another Engine compromise/shortcut, that as a side effect eliminated the need for a skybox. Why have the engine render a skybox when the player can't ever stop looking at the ground?
The multiple, modifyable media types would be a nice-to-have.
I also need more than two meshes to load after the current one. Ideally you should be able to have as many "next meshes" as you want but at least 4 would have been nice.
Frenzy Released - about time too
-
iron
- Site Admin
- Posts: 2006
- Joined: Thu Feb 26, 2004 1:21 am
- Location: diving out of the Sun at 10 o'clock high!
- Contact:
There's several reasons why Myth won't scale up to more than 400ish units. Firstly its use of sprites (as much as I love sprites) means a lot heavier texture load on the graphics card than if 3D units were used. Next, there's the CPU load in processing the status, pathfinding, projectile collisions and so on for each and every unit every 30th of a second. Also there's the consideration of having to keep up to 16 machines running in sync in a multiplayer game. Besides, its a small squad tactics game - you really don't have the controls needed to efficiently manage large groups.
It could probably be pushed higher than 400 anyway if you didn't have to worry about the game running on older machines, however we never have that luxury with Myth.
You're right about the massive linked list & lack of a scene graph. Along with that, I've experimented with tilting the camera to different angles - if you make it more horizontal then the fps plummets. There's something in the mesh rendering that simply can't handle too many cells being visible at once. I'd say that more than anything explains why the camera is looking downwards all the time. And yes, if your sprites are rendered from a particular angle then you can't really go changing the view aspect too much.
Really if you wanted to make significant enhancements with the rendering you'd be better looking at a brand new engine rather than trying to hack them into Myth. Its something we've discussed at length.
Edited By iron on 1124934191
It could probably be pushed higher than 400 anyway if you didn't have to worry about the game running on older machines, however we never have that luxury with Myth.
You're right about the massive linked list & lack of a scene graph. Along with that, I've experimented with tilting the camera to different angles - if you make it more horizontal then the fps plummets. There's something in the mesh rendering that simply can't handle too many cells being visible at once. I'd say that more than anything explains why the camera is looking downwards all the time. And yes, if your sprites are rendered from a particular angle then you can't really go changing the view aspect too much.
Really if you wanted to make significant enhancements with the rendering you'd be better looking at a brand new engine rather than trying to hack them into Myth. Its something we've discussed at length.
Edited By iron on 1124934191
-
William Wallet
- Posts: 1494
- Joined: Tue Mar 30, 2004 9:40 am
- Location: Perth Australia
- Contact:
I really didn't mean to bring up the 401st unit thing, I just meant that I wanted to work on a new engine and wanted to have an open discussion on the design details so I didn't start something that was fatally flawed. (but I do have an idea on how to handle the extra units.) Say all 400 units use the same collection, say warrior. Isn't the texture load going be the same whether its 1, 10 or 401? I would think the real killer would be on the CPU doing the z-buffer calculations on all the individual units...
The OpenGL meshviewer or plugin viewer or whatever its called can fly over SB meshes with an open z and has no problem rendering the mesh from top down. (in retrospect I wish I'd never used it).
The OpenGL meshviewer or plugin viewer or whatever its called can fly over SB meshes with an open z and has no problem rendering the mesh from top down. (in retrospect I wish I'd never used it).