[PATCH] drm/amdgpu: throttle buffer migrations at CS using a fixed MBps limit

[Marek Olšák]

On Wed, Aug 24, 2016 at 10:05 AM, Christian König
<deathsimple at vodafone.de> wrote:
>> +               struct mutex            lock;
>
> I haven't completely checked the code, but it looked like you just made a
> bunch of calculations while holding the lock (e.g. no sleeping).
>
> So it might be more appropriate to use a spinlock here (less overhead).
>
> Apart from that this looks really good to me, patch is Acked-by: Christian
> König <christian.koenig at amd.com>.
>
> BTW Regarding VRAM fragmentation: Could it be that most align restrictions
> Mesa forwards to the kernel became nonsense with VM support?
>
> I mean for tilling the memory must be aligned in the VM, but not physically
> any more. Doesn't it?

I think all allocations must be both physically and VM aligned. I
think that physical alignment is required because tile modes expect a
certain mapping of tiles to memory channels.

Marek