April 24, 2019 10:40 AM, "Henri Verbeet" hverbeet@gmail.com wrote:
On Wed, 24 Apr 2019 at 19:52, Ken Thomases ken@codeweavers.com wrote:
Right, but with automatic graphics switching there are two GPUs that could potentially drive the main display.
This may be a slight tangent, but while early switchable GPUs had both GPUs connected to the display through a multiplexer, on current setups only the integrated/weak/low-power GPU is actually connected to the display, and the other GPU is actually headless. The GPU connected to the display is then responsible for getting the contents of the other GPU onto the screen. I'm not sure how accurately that topology is reflected by the relevant MacOS APIs.
I'm not so sure that is in fact how it works on Mac. I remember a change to MoltenVK that was made to use MTLCreateSystemDefaultDevice() in order to force the system to switch to presenting with the discrete GPU, because that performed better when also using it for rendering.
Chip