GPU Architecture Introduction 笔记 2019-11-24

https://developer.samsung.com/game/gpu-architecture

modern GPU architectures can be coarsely split into two categores: IMRs and TBRs

Immediate-Mode Renderers IMRs

IMR-style atchitecturea are mominant in desktops, laptops and consoles.
the key charactericstic of an IMR is that primitvies are prcessed immmediately throught the graphics pipeline.
this means that each primitive travles through the entire GPU pipeline without any knowledge of the primitives that came beore or after it.

IMRs_graphic_pipeline.png

左侧是GPU，右侧是system memory，
Geometry data应该是定点坐标和索引，Texture data也需要copy到video memory才能用。
图中有两个Z test阶段，

Tile-Based Renderers TBRs

TBR-style architerctures are domniant in mobile and embedded devices.
the key chararcteristic of TBR is taht scene is divided into fixed-sized, screen-space regions.
the pixel colors in the scene are calculated a tile at a time.Once a tile render is complete, its attachments are written out to system memory.
TBR的特点是，把一个画面分割成多个固定大小的块，一次计算一个块的pixel color， tile render完成后，把他的attachments写到system memory。

Tile-Based Render pipeline.png

benefit of TBRs is that a tile-sized cache can be used to store intermediate attachment data during the rendering process.
Caching this information on-chip avoids costly system meory bandwidth transfers.
the downside of TBRs is that entire scene's geometry needs to be transformed and binned into tiled regions before per-tile rasterization can begin.
this increases the complexity of GPU and increatesed memroy requiremebnts, as an intermediate buffer to store the transformed primitives and binning information that the per-tile operations can read later in GPU execution process.

TBR的优点是，render期间，tile-sized cache可以保存到临时attachment data中，这样cache在video memory可以节省system memory的带宽。
TBR的缺点是，在per-tile光栅化前，所有的geometry都需要转换和绑定到tiled regions，这就提高了GPU的复杂度和内存需求，作为保存transformed primitives和binning info的intermediate buffer，per-tile操作可以在以后的GPU处理阶段读取这些信息。