by
Falco Girgis July 12, 2011
Originally posted by Tvspelsfreak (Anton Norgren) in our Private Development forum on 7.12.11.
Note: This was originally posted during the initial development of the Dreamcast version of libGyro. This is probably an advanced Dreamcast topic, but we thought it may come in handy for low-level developers looking to get more performance out of the DC. Direct rendering submits vertex data to the GPU immediately when rendering. Using DMA rendering, you essentially store your scene's vertices in main RAM until your scene is complete. Then you initiate an asynchronous DMA batch transfer to copy these from main RAM to the PVR's VRAM fo rendering. One giant benefit of this is that you no longer have to order your primitives by transparency type when rendering (TR, PT, OP). Store queues are a hardware feature that allows you to do extremely fast burst-writes of this vertex data directly into your vertex buffer in main RAM and bypass the cache. -- Falco Girgis
Ok, the first thing you wanna do is switch to DMA rendering.
To do this, you need to enable the "dma_enabled" flag in the params you pass to pvr_init().
You also need to set up a vertex buffer in main RAM for every list you've enabled.
Read More
share this post: