Subversion Repository Public Repository


This repository has no backups
This repository's network speed is throttled to 100KB/sec

File is too big or a binary file

Commits for Divide-Framework/trunk/Source Code/Hardware/Video/OpenGL/Buffers/VertexBuffer/glVertexArray.cpp

Diff revisions: vs.
Revision Author Commited Message
331 Diff Diff IonutCava picture IonutCava Sat 06 Dec, 2014 20:53:45 +0000

- Limited line length to 132 characters to improve readability and diff-comparisons
- Refactored memory allocation/deallocation functions
- Fixed a few compatibility issues with HashMap.h
- Fixed a bug in GPU Skinning shaders (cast a float to int)

330 Diff Diff IonutCava picture IonutCava Fri 10 Oct, 2014 17:19:11 +0000

- New rendering algorithm :
— Perform second pass for visible nodes gathering all render commands and uploading them to an indirect draw buffer
— Render geometry with indirect draw commands
— Removed per-node range binding of the node buffer. Instead, bind the entire buffer once and index data in shaders
— Use “baseInstance” parameter from IndirectDrawCommand as a node buffer index
— Prefer DSA extensions for updating buffer data where applicable
- Moved all rendering specific code from SceneGraphNode to a new RenderingComponent
- Optimized “toString” utility function

318 Diff Diff IonutCava picture IonutCava Sat 30 Aug, 2014 17:35:53 +0000

- Wrapped the entire code in a “Divide” namespace
- VertexBuffers now call “shrink_to_fit” on all internal data storage
- Improved some vector performance by preferring “emplace_back” instead of “push_back” + proepr usage of reserve / resize
- Wrapped OIS specific types and classes in Divide::Input namespace
- Added the messageBox.layout file (forgot it in the previous few commits)

307 Diff Diff IonutCava picture IonutCava Tue 12 Aug, 2014 19:45:12 +0000

- More editor work:
— Per object skeleton rendering, bounding box rendering and shadow map toggling
— Full screen depth preview toggle
— Text entry based transforms

302 Diff Diff IonutCava picture IonutCava Wed 06 Aug, 2014 22:09:55 +0000

- GenericVertexData buffer size (for read/write buffering) is now configurable instead of a fixed 3x and also applies to non-persistently mapped buffers
- Moved OpenGL enum tables to Divide::GLUtil::GL_ENUM_TABLE
- More code cleanups (glResources, glError, glVertexArray, glGenericVertexData, glIMPrimitive)

295 Diff Diff k1ngp1n picture k1ngp1n Fri 01 Aug, 2014 20:10:44 +0000

- If a shader fails to bind, skip draw call. A shader will fail to bind if it wasn’t linked properly, as well
- Moved ShaderManager to the Hardware project due to its tight coupling
- More code cleanup and comments:
— OpenGL wrapper code : GLWrapper.h, GLWrapper.cpp, GLStates.cpp, glShaderProgram.cpp
- Cleaned up Shader.cpp
- Binary shader dump also saves binary format in .fmt file for each binary shader
- Renamed immediate mode primitive files

293 Diff Diff k1ngp1n picture k1ngp1n Tue 29 Jul, 2014 20:01:28 +0000

- Code cleanup and comments in GLFWWrapper (for starters. more to follow)
- Reworked main loop logic (no more delegates and rendering api querying)
- Replaced all GLvoid with regular void
- replaced all boost::this_thread::sleep (deprecated calls) with boost::this_thread::sleep_for

286 Diff Diff IonutCava picture IonutCava Fri 18 Jul, 2014 16:32:03 +0000

[Ionut] [[BR]]

    • very buggy commit ** [[BR]]

- New rendering pipeline: [[BR]]
— Bumped OpenGL version to 4.4 [[BR]]
— Command based rendering. GFXDevice dispatches rendering commands to their respective buffers [[BR]]
— State changing is exclusively controlled by the GFXDevice class [[BR]]
— Material properties and transform matrices are stored in shader storage buffers that are updated after culling and uploaded once per RenderPass. Objects use drawIDs to index data in shaders [[BR]]
— Removed all custom matrices and other, per shader, bind data [[BR]]
— Removed per node light data. Light culling and shader buffers will replace the old light management system [[BR]]
— Framebuffers now use the Texture class for attachments [[BR]]
— Sampler objects now work like RenderStateBlocks. GL_API handles them and stores them in a hash indexed map [[BR]]
— Removed “prepareMaterial” calls. Pre-draw material setup is now only limited to texture binding [[BR]]
— Moved immediate mode emulation and primitive rendering from API level to GFXDevice level [[BR]]
— Framebuffer rendering must be completed with a call to end() to ensure proper MSAA resolve (asserts, otherwise) [[BR]]
- Fixed hash code generation for RenderStateBlock (and now, glSamplerObject) using boost::hash_combine [[BR]]
- Temporarily disabled XML caching for materials [[BR]]

280 Diff Diff IonutCava picture IonutCava Sat 28 Jun, 2014 19:11:37 +0000

[Ionut] [[BR]]
- Improve ClipPlane management (bunch equations into the matrix shader buffer) [[BR]]
- Immediate mode emulation primitives set their own state based on specified hash instead of relying on state functions to be defined [[BR]]
- updateStates() function is now called right before the actual draw call to facilitate future batching work [[BR]]
- added tracked bool values (similar to those in ShaderInfo) to SGN class to fix selection bug. [[BR]]
— will probably create a “stateTracker” class to allow multiple classes to use tracked variables without all of the code duplication [[BR]]

279 IonutCava picture IonutCava Sat 28 Jun, 2014 16:20:46 +0000

- Unified draw command system into GenericDrawCommand [[BR]]
- Fixed double VB creation in Object3D [[BR]]
- Reworked triangle data computation (now Object3D’s responsibility) [[BR]]
- VertexArrays can now render using the primitive type specified in the draw command instead of the fixed type specified at creation [[BR]]
- Optimized terrain shader [[BR]]