master

EEngin ManapRaise model limit to 65536 with a 2D transform texture

55375c51创建于 9 天前历史提交

文件	最后提交记录	最后更新时间
OpenGLESGraphics.cpp	Raise model limit to 65536 with a 2D transform texture The model transform texture was 4 * NR_MAX_MODELS wide, so the model limit was really a texture width limit (the Pi missing objects bug). It is now 1024x512, a 4x2 block per model in 2-row bands, each matrix in its own 2x2 quad. Depth passes don't read the normal matrix, so they touch one quad per model. The texture width reaches shaders through getShaderHeader as MODEL_TRANSFORM_TEXTURE_WIDTH. The bone texture height is now NR_MAX_RIGS (4096) instead of NR_MAX_MODELS, and the model index batch is capped by NR_MAX_MODEL_INDEX_BATCH (4096) instead of NR_MAX_MODELS. The texture size warning checks both dimensions of both textures.	9 天前
OpenGLESGraphics.h	Raise model limit to 65536 with a 2D transform texture The model transform texture was 4 * NR_MAX_MODELS wide, so the model limit was really a texture width limit (the Pi missing objects bug). It is now 1024x512, a 4x2 block per model in 2-row bands, each matrix in its own 2x2 quad. Depth passes don't read the normal matrix, so they touch one quad per model. The texture width reaches shaders through getShaderHeader as MODEL_TRANSFORM_TEXTURE_WIDTH. The bone texture height is now NR_MAX_RIGS (4096) instead of NR_MAX_MODELS, and the model index batch is capped by NR_MAX_MODEL_INDEX_BATCH (4096) instead of NR_MAX_MODELS. The texture size warning checks both dimensions of both textures.	9 天前
OpenGLGraphics.cpp	Raise model limit to 65536 with a 2D transform texture The model transform texture was 4 * NR_MAX_MODELS wide, so the model limit was really a texture width limit (the Pi missing objects bug). It is now 1024x512, a 4x2 block per model in 2-row bands, each matrix in its own 2x2 quad. Depth passes don't read the normal matrix, so they touch one quad per model. The texture width reaches shaders through getShaderHeader as MODEL_TRANSFORM_TEXTURE_WIDTH. The bone texture height is now NR_MAX_RIGS (4096) instead of NR_MAX_MODELS, and the model index batch is capped by NR_MAX_MODEL_INDEX_BATCH (4096) instead of NR_MAX_MODELS. The texture size warning checks both dimensions of both textures.	9 天前
OpenGLGraphics.h	Raise model limit to 65536 with a 2D transform texture The model transform texture was 4 * NR_MAX_MODELS wide, so the model limit was really a texture width limit (the Pi missing objects bug). It is now 1024x512, a 4x2 block per model in 2-row bands, each matrix in its own 2x2 quad. Depth passes don't read the normal matrix, so they touch one quad per model. The texture width reaches shaders through getShaderHeader as MODEL_TRANSFORM_TEXTURE_WIDTH. The bone texture height is now NR_MAX_RIGS (4096) instead of NR_MAX_MODELS, and the model index batch is capped by NR_MAX_MODEL_INDEX_BATCH (4096) instead of NR_MAX_MODELS. The texture size warning checks both dimensions of both textures.	9 天前
Std140Layout.h	Use multi buffering for UBOs too This is a way to workaround rpi drivers problems. It blocks on every write, and doesn't have much internal buffering.	9 天前