Subject: Re: X for hp300?
To: None <carrel@cisco.com, mike@cs.utah.edu>
From: Mike Hibler <mike@cs.utah.edu>
List: port-hp300
Date: 04/04/1994 11:45:28
> To: mike@cs (Mike Hibler)
> Subject: Re: X for hp300? 
> Date: Mon, 04 Apr 1994 10:30:07 -0700
> From: David Carrel <carrel@cisco.com>
> 
...
> > As for speed, as I recall text drawing was a little slow (I've gotten used
> > to it now so I can't say for sure).  Mark knew what was wrong but didn't have
> > time to fix it.  Maybe he can elaborate.
> 
> The biggest thing I noticed is that input is done by read()-ing the HIL
> instead of using the shared memory queues (they are documented in hil(4) ).
> Anyway, I'm not concerned with that right now.  The speed is really good
> enough.  I really just want to be able to type.  I turned the machine off
> last night and _thought_ about the problem so maybe I'll do better tonight.
> 

reading vs. queue mapping has little effect in practice.  We originally did
the mapped interface and hacked the hell out of the X10 server to use it.
The read interface was there for HP-UX compatibility.  Then X11 came out and
we didn't want to go through the hassles of re-adding the queue mapping
code to HPs contributed server so we punted and added the magic HILIOCHPUX
ioctl to enable the read interface.  It might be worth revisiting this
decision in the future.  You will notice that mapped queues still use the
ioctl interface and not mmap.  It wasn't clear how I could do things like
map multiple devices to the same queue using the mmap interface.

> > Oops!  Thats a Utah thing that wound up in the MAKEDEV file.  You can pick
> > it up from jaguar.cs.utah.edu in ~ftp/pub/hilinfo.c.
> 
> Thanks, I got it.  I'll try it out tonight.  ;-)
> 

I also put "grfinfo.c" out there.  It may also be used by MAKEDEV.

------------------------------------------------------------------------------