[Dri-devel] Re: Radeon: lockup on state change

[Keith W. <ke...@tu...>]

Felix K=FChling wrote:
> On Sun, 8 Dec 2002 14:24:58 +0100
> Felix K=FChling <fe...@vi...> wrote:
>=20
>=20
>>Hi,
>>
>>as I reported earlier there seems to be a race condition in the Radeon
>>driver when state is emitted while the card is processing vertices. Now
>=20
>=20
> Keith, I just read your other message. Ok, so let's call it "failure" ;=
-)
>=20
>=20
>>I narrowed it down to tex[0], tex[1] and tcl states that appear to
>>require 3D to be idle. Can someone with Radeon specs confirm this? Othe=
r
>=20
>=20
> That's exactly what you asked for :). More precisely, if TCL was
> disabled I had to wait before emitting texture changes. If TCL was
> enabled it was enough to wait before emitting TCL state. But this may b=
e
> due to the order of state changes. If TCL state is always emitted befor=
e
> texture state, then waiting before the TCL state will also ensure that
> 3D is idle before the texture state change.
>

Felix,

I've committed the first version for now, because
	1) it's cleaner
	2) it emits at most 1 wait per statechange
	3) I'm not convinced that there are many statechanges that don't touch  =
any=20
of tcl, tex0 or tex1 -- ie I think the second version would emit a wait o=
n=20
*almost* every statechange.

If you want to narrow the lockup conditions down further, feel free & I'l=
l=20
commit the results.

Keith