Trident CyberBladeXP4 status?

[Jason L Tibbitts III]

Some may remember that I was desperate to get some form of
acceleration working on a Toshiba Portege R100 with a CyberBladeXP4m32
graphics chip.  Well, I've had little luck.  I spent quite some time
poking about trying to get accelerated region fills working with no
luck, so I switched to screen->screen copies and am faring slightly
better.

Basically, I know the chip is blitting regions around because I can
see the driver making the calls and I can see the engine go busy and
back to idle.  The busy time is correlated very strongly with the size
of the region being copied.

Unfortunately what's actually getting blitted around is garbage.
There is some correlation between what's being blitted and the garbage
that shows up, and from that I can derive that the weird reversed
coordinate packing is at least partially correct.  (The XPs seem to
pack coordinates into a long backwards from the way other Trident
chips do and if I switch things around I can tell that things are
moving in the wrong direction.)  Maybe it's more complicated than
that.

Anyone have any ideas or any tricks for debugging this kind of thing?
Is there any hook I can use to fill the entire framebuffer with some
pattern when the server starts so I can figure out where the blitter
is working?  Any chance that anyone has docs for this chipset and
could drop me a hint or a code snippet?

(BTW, if this is a matter of someone having docs under NDA but not
having hardware, I may be willing to buy another R100 and loan it
out.)

 - J<