Created attachment 1075 [details] [review]
p_i810-accel.diff

Needs to be discussed outside of Bugzilla first.  
 

reopen to close as FIXED. 

committed to CVS. 

Updating to the latest CVS (which includes your patch) produces weird artefacts
with acceleration on my i815 based systems, e.g. incompletely drawn gtk widgets.
The problem seems to be in this change:

https://freedesktop.org/bugzilla/attachment.cgi?id=1075&action=diff#programs/Xserver/hw/xfree86/drivers/i810_old/i810_accel.c_sec1

which apparently prohibits the allocation of an LpRing at Xserver startup.  I
don't claim complete understanding of what's going on there, but reverting this
bit restores the correct behavior.  Tested on Intel D815EEA and GigaByte GA-6IEM
i815-based mobos.

Created attachment 1167 [details] [review]
Patch reverting to the original behavior

Patch on top of the previous one to revert the single offending bit.

BTW, the artefacts I was talking about are very similar to those people
complained about in bug #1232.

Yes I agree, the patch was wrong. It should have been (... && pScrn->vtSema).
Would you please try that?
I don't know why it came out wrong. I'm going to change this in CVS.

I've thrown the stupid vtSema fix out now. 
I don't thin we need it any more anyway.
Thanks Roman!