Closed valpackett closed 5 years ago
@johalun any opinions on this?
If it looks good to you, I'm happy to test it for the port.
No makefiles in the drivers/ subfolders are used so no need changing them. We have to remember to pull this in to 4.17 later also.
@myfreeweb Can you revert the changes to makefiles in drivers/ so we don't get merge fail on those later?
I don't see the point of merging something that isn't tested. Loading module is one thing, using them is another thing (trust me on this).
Is there anyone with an arm64 with a radeon or amdgpu graphics card to test? Otherwise I'm inclined to agree with @evadot I didn't push my power changes until I had confirmed that the driver at least attached.
@myfreeweb Can we close this? The POWER support changes should make this easier to add in the future if someone has appropriate hardware.
Yes, I'll redo this on top of the POWER changes when I get hardware to test on. (btw, did you find anything from this PR useful for the POWER changes?)
This PR adds some basic portability and makes radeon and amdgpu build on aarch64. Does not include openfirmware stuff (damn that's a lot of API surface), only works with the existing PCIe support.
radeon loads and says this:
amdgpu doesn't load (this is a soft FPU function, I guess we need a flag other than
-mhard-float
??):Whether it actually works is not tested… porting rkpcie would be the first requirement on my hardware (RK3399), and then there's the issue of whether u-boot configures pci correctly. Marvell Armada 8k (MACCHIATObin) is at least known to work with GPUs on Linux, but we don't yet have pcie on that either. I guess if someone has physical access to a ThunderX box, they can test it :D Jetson TX1 might have PCIe working too?