Re: 11-BETA, WebGL, modesetting and Radeon HD-5450: breaks X

matthew green <[email protected]> wrote:
>> Platform: 	Linux x86_64
>> Browser User Agent: 	Mozilla/5.0 (X11; NetBSD x86_64; rv:142.0)
>> Gecko/20100101 Firefox/142.0
[ ... ]
>>
>> and mentions llvmpipe. On the other hand it may be inaccurate since the
>> Platform value is wrong. On the test machine it shows "Radeon HD 3200
>> Graphics, or similar".
>
> this may just be the latest way firefox works.  it reports linux x86_64
> everywhere except a handful of places like windows or macos or mobile.
>
> one thing i've noticed the radeon driver still has issues with some
> sort of cache issue, where the display is incorrect some relatively
> large portion of the time, menus wrong, that clear up when you move
> the cursor near them, weird pixelated stuff.  it only happens really
> poorly on one specific system for me -- a "AMD E2-7110 APU with AMD
> Radeon R2 Graphics" laptop, that seems to be the last generation of
> APUs that used "radeon" instead of "amdgpu" (i have another laptop
> with a very slighty older CPU -- same micro-arch, bulldozer#4 -- but
> comes with amdgpu that works fine.
>
> anyway, i wonder if the ring timeouts are because values aren't 
> making it into the right place in time (bad cache flush type thing?)

Isn't that basically what we see with amdgpu?

A GPU on the other side of a PCIe bus works but those sharing memory
with the CPU don't.