#68490 - acox - Wed Jan 25, 2006 1:19 am
Looking around I?ve seen references to the rasterization hardware being line-oriented, with a 256x1 z-buffer and tris filled on scan-out.
Is there any evidence that this is the way it really works?
Given ARM?s links to Imagination, offering the latter?s PowerVR technology with their cores for SOC solutions, I cant help wondering if the 2048 tri limit might have another source. I was thinking that rasterization might be tiled. It could be backed by a non-addressable, hidden framebuffer (fullscreen 256x192) (hidden just for engineering reasons so it is on a highspeed dedicated bus, connected to the rasterizer). The 2048 tri limit would then be explained as the size of the tile-sorting buffer of the rendering core.
This setup would have the same limitations people are seeing: No access to z-buffer (because it is temporary and 32x32(say), as opposed to 256x1); Framebuffer can be captured to a background but only at extra cost.
Before I go too far with speculation I?ll just ask again: Is there any evidence for the ?standard model? of DS rasterization?
Any pointers to info on Display Capture would be most appreciated.
_________________
3D on GBA
Is there any evidence that this is the way it really works?
Given ARM?s links to Imagination, offering the latter?s PowerVR technology with their cores for SOC solutions, I cant help wondering if the 2048 tri limit might have another source. I was thinking that rasterization might be tiled. It could be backed by a non-addressable, hidden framebuffer (fullscreen 256x192) (hidden just for engineering reasons so it is on a highspeed dedicated bus, connected to the rasterizer). The 2048 tri limit would then be explained as the size of the tile-sorting buffer of the rendering core.
This setup would have the same limitations people are seeing: No access to z-buffer (because it is temporary and 32x32(say), as opposed to 256x1); Framebuffer can be captured to a background but only at extra cost.
Before I go too far with speculation I?ll just ask again: Is there any evidence for the ?standard model? of DS rasterization?
Any pointers to info on Display Capture would be most appreciated.
_________________
3D on GBA