I'd like to introduce a device 'hpcfb', which you can find at
sys/arch/hpcmips/dev/hpcfb.c. I think the device has some features
which you mention though the device is still under development.

Following lines are comes from arch/hpcmips/conf/GENERIC,
which tell you how the device is configured.

	mqvideo0	at	vrip?		# MQ200 video controller
	hpcfb*          at      mqvideo?
	ite8181video0	at	vrip?		# ITE8181 video controller
	hpcfb*          at      ite8181video?
	wsdisplay*      at      hpcfb?

Hpcfb has virtual text vram in it and works as a charater display
for wsdipslay while lower video drivers are simple bitmap display.
Video drivers can also provide some accelerations including 
bitbliting for hpcfb if they are available.

Date: Tue, 30 Jan 2001 10:07:05 -0800
From: Jason R Thorpe <thorpej@zembu.com>
 >  > Given that frame buffer memory access tends to be significantly slower
 >  > than regular memory access, it may make sense to re-render the text
 >  > into the new frame buffer location rather than do a frame-buffer to
 >  > frame-buffer copy.
 > 
 > That requires storing the character cells somewhere.  Now, this is arguably
 > the right thing to do (as it would give us virtual terminal support on ALL
 > framebuffers) -- Roland Dowdeswell was going to work on this, but I think
 > his job got in the way...

Takemura