2d filters and speedup with fbo(frame buffer object):
-like hdri
-ssao
-relief maps
-etc-etc
shadows improvements:
-soft and alpha shadow pach bugfix and implementing
-large terrain shadow like Light Space Projective Shadow Mapping techinc or other
-sun lamp shadow same as Light Space Projective Shadow Mapping techinc?
tools stuffs:
-portals
-3d decals
-decimate modifier with preserve uv make easy for Lod http://www.wxtools.com/wp/?p=166
-dynamic 2d imposters
This thread doesn’t make much sense. You need someone interested in working first. And make a poll thread second. Not the contrary. It’s worthless for us to vote for this and don’t have anyone willing to do it.
If someone is willing to do what ideasman did. They will do it. All we need to do right now is let them know that we are willing to pay for improvements. But we can’t tell them what to do without knowing what they want to do.
I would vote to actually speed up the BGE renderer before adding new graphical bells and whistles (Ben2610 did some work here, but didn’t make a difference).
Also, assuming if Ben2610 hasn’t made the BGE multithreaded within the next few months like he mentioned when working with Entertainment Arts, multi-threaded physics.
@Cyborg Dragon,
between 2.48 and 2.49a YoFrankie whent from 50 -> 80fps. For an overall speedup IMHO thats impressive, most of it can be attributed to benoits work. (I did some py speedups too but dont think they changed overall performance much)
The problem with funding optimizations is that they are really hard to measure. they depends a lot on each dataset, and how to you measure when the BGE is optimized enough.
In short I dont think optimizations make for good community funded projects, unless people are happy to donate to general optimizations.
The performance could be increased indirectly. I mean just implemented LOD for terrain would increase total amount of polys in scene like 5-10 times. I think that most of people need this feature. “Normal” terrain nowadays is at least 1024*1024 ~ 1000K polys. If with LOD for regular terrain (triangular or quad mesh) could be render per frame less than 100k without much difference for visual that won’t kill the performance. User could specify terrain object, blender calcualtes LOD once, store it somewhere, and BGE loads it with dynamic obj script you are working on, and renders it.
Aslo, some automatic LOD system could be made for things like vegetation or trees. User just specifies LOD objects, camera distance, and BGE does the work itself. I know it could be done partly with python, but why not to optimise it?
I’m starting work on a mouse actuator, so far I’ve only changed the ui code, whether I change the actual game code is totally up to how much spare time I have.
Should reduce those noob threads saying “how do I rotate the camera with the mouse/make a fps?” or “how do I move things with the mouse?”
that’s a good thing noob_dest
also a slider for setting mouse acceleration would be nice.
If you’re done with that and still have time&motivation left, how about adding .ogg and midi support for the bge? I proposed that in another thread too but that’s something that can be mentioned several times imo
This is a good idea though if you can keep adding/editing things to the list to be voted on.
Nobody would want to donate for just the things one guy wants or is interested in,
but certain things would wind up striking a chord with everyone.
A couple of days ago I was thinking about trying to add another uv layer for the GE,
I almost asked on a scale from 1 to ten what would the difficulty in that be,
but I realized it would probably drop the framerate in half, and that I didn’t really
need another uv layer.
I think a built-in particle system would be a great new addition for the GE - by re-using the existing particle UI for placement / parameters etc, even a very basic system ( single spawn - no collisions ) would enable users to quickly add good looking particle effects to their games ( fire / water fountains, engine glows / streams ).
To do this currently requires adding a lot of billboard objects with logic blocks / script - by developing a built-in particle system, a lot of optimisations could be gained ( drawing all particles in a system at the same time, to ensure no 3D context swaps ).
hi this is the first try for poll and test what the reaction for members so i think that i will collect all information here and start a new poll for adding some other options and recreate the groups, but please note that very hard to fit for every members for create a good group for features, rather than please choice a group that near for your request , anyway i think any gourp of this features would be nice for GBE
-material:
-dds texture comprression
-TextureFace should work in GLSL
-antialising
-glow glsl map
-amb at Map To panel works with glsl
-Refl at Map To panel with Cube Env map
-new logic bricks:
-fps logic brick
-constrain to path support (actuator? or use the constraint)
-text object box multi line actuator? (left, center, right, align)
-timer sensor (like delay sensor): count back, reset, start, pause
-getIpoCurrentFrame value with python
-Add Property button - need List type?
-extras:
-gui system
-particle system
-network system
-ai system
-physics:
-python access to collision points
-tools second group:
-bitmap font creator build in
-normalmap baking with AA
-create uv unwarp on all active or selected objects and show/editing in on time the uv editor
-baking lightmap for instanced mesh (dupligroup or dulicated linked mesh)
-work tangent normalmap with mirrored uv (or with mirror modifier?)
-unique uv map for instanced object
-terminal feedback inside blender window
others:
-up down button for all visual logic brick blocks without need closing
- Implement a global and per-object event queue where both python and
(internal) c++ functions can be registered as event handlers
(trough a callback mechanism) that way the engine will be more
dynamic, and libraries created in python will have a lot more
power and will be easier to translate to c++ if they prove to
be useful and/or need more speed.
I think the GE only supports 2 layers, unless it’s just a 3 layer texture splat bug.
I don’t really need it though.
That’s a good one, Normal map baking with aa!
@locadax. great list, maybe you could start a wiki page listing features like this, and collect more?
its important they are feasible, not “make blender as good as cry-engine”
You mention getting collision points, I think this is not too hard, I briefly looked into this while doing reinstancePhysicsMesh. but I better not get too distracted from my current pledge
That’s his work with the scenegraph and logic, I recall his latest commits for improving performance was supposed to reduce overhead in the BGE, but he also stated in the mail that what he did to reduce overhead lost most of the benefit benefit because it was lost in the OpenGL. That was when he mentioned Nvidia’s concept of bindless OpenGL that could allow some real good benefit performance wise from reducing the amount of overhead.
What I really wanna see in blender are extended baking options like (FULL RENDER of all rexternal renders like indigo or yafray) ( or direct global illumination baking would help)
I would like the shadow improvements and 3d decals = (really good for inovative seam hiding)
=hope you do it
