Hi, On 12/23/19 7:55 PM, Daniele Ceraolo Spurio wrote:
If you only care about gen9, a possible solution would be to forward-port just the old GuC submission (instead of the whole i915) from a know working kernel. It shouldn't be too bad since the GuC code is relatively self-contained, but given the speed at which our submission code evolves there might be issue in the interactions between the old GuC code and the other parts of the submission flow.
Right, we may have to look at that. We're currently looking into a non-GuC workaround as there doesn't seem to be any OpenCL kernels for the Surface Pro 7/Surface Laptop 3 devices (that's another issue...), but we will keep that in mind for the older devices.
No idea about the details of what goes on on Windows. The firmware is OS-agnostic, but they might be using a different version compared to us, especially on older platforms.
Yeah, I suspect they're using older firmware on the older devices. Just wanted to know if there's any special stuff on Windows. Given that the newer devices (Pro 7/Laptop 3) don't seem to have any IPTS firmware at all (OpenCL kernels etc.) could suggest that they might use a different implementation (maybe non-GuC/no GPU processing), altough the IPTS driver still seems to be in use. Thank you for sharing your insights, this has really helped us. Regards, Maximilian _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
