Is there any working example of the above in OpenGL? I’m using a 970 (and occasionally a Titan or Titan X), and all these should have 2 copy engines, and therefore should be able to do async uploads well, even while rendering.
The I’ve tried several methods and mapping approaches, but I haven’t gotten anywhere really… I’ve tried to modify the code from chapter 29 in GPU insights, but the demo is a complete mess and it doesn’t even render anything.
How do you know it is not working ? Do not follow marketing schemes and terminology. Asynchronous uploads have been in OpenGL via PBOs for years. Whether or not these are backed by hardware is another thing. DMA is nothing new and I"m assuming these ‘copy engines’ are nothing more than dedicated DMA controllers used by the driver. As such they are not exposed, an it should just be a matter of jumping through a few hoops to figure out what flags to used when creating and using the buffer.