<sirspudd>
top of the morning peeps; pardon my n00bness, but is scatter gather a point of failure people common see/encounter?
<sirspudd>
I know I have had to disable this in amdgpu in recent memory due to artifacts, I am wondering about how problematic this is in the context of etnaviv (running on a gc2000)
<sirspudd>
another point of interest, I see the upstream mainline imx6 device trees reserve no explicit memory for cma, opting for a cma region whos size is defined in the defconfig instead
<sirspudd>
I am seeing an exceptionally low level of CMA usage, which historically (Linux 4.9 historically) I required 320M of memory, I now see 10M used out of my allocated CMA region, and then scatter gather being broadly used