Open thyrgle opened 8 years ago
Thanks, I will take a look
See comment in this commit. https://github.com/gandf/vqisosmall/commit/7037d346b3fae1d44f0cc8c43d716a3dcb30a8a8 Texture3D doesn't work without parameter. I wrote "//***" at end lines to modify in GenerateVolume.c and f00300_singleton.hpp. Thanks.
Some tests on windows 10 :
I think this is a problem with an old OpenGL versions I've mentioned here.
AMD Driver give log (not INTEL).
Validation failed! - Different sampler types for same sample texture unit in fragment shader.
Error is because we can only use one type of sampler. We can use sampler2D or sampler3D.
uniform sampler3D Texture0; //volume texture nearest uniform sampler3D Texture1; //volume texture linear uniform sampler2D Texture2; //terrain heightmap uniform sampler2D Texture3; //terrainID // no longer used uniform sampler3D Texture4; //ter tex
Since all texture3D have been disable, only texture2D are used and work on all GPU. I think we need to do two shader.
@gandf - its possible that upping the shader version would allow texture3D where it is failing, but that would involve a bit of other work as the OpenGL pipeline on the c++ end changes as well (no more use of old commands like glBegin, etc)
Currently, this game engine is not enought quick to be used for game. I think what we need to do upgrade version to GLSL 330 to have most of current hardware working with this engine. This will increase performance and fluidity. Work already done is significant and switch to version 330 needs to change code structure and how code works. With new version, we can't use direct access to write if we want to improve performance.
What do you want us to do ?
1) nothing : only nVidia GPU can run slowly game. 2) upgrade to version 140 : A lot of work most GPU compatible with run slow game. 3) upgrade to version 330 : most GPU compatible with better performance but greatest work.
I'm not openGL developer and I can't say what we should and may to do.
@gandf - upgrading to 330 is probably a good idea for compatibility purposes, but I don't think you will get much in the way of performance gains unless you actually change the underlying algorithms. There are plenty of ways to make it faster, but you will probably have to rethink the core algorithms. On high end hardware (gtx 980/1080) it is playable but the chunk picking system sucks (and thus much time is wasted in loading empty chunks and chunks that are underground). In future versions I have better chunk picking/loading (such as the following perspective version).
On my Mac:
Furthermore:
Full information: