[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: clearing the screen -> framebuffer insanity

On Sat, 2002-12-07 at 07:51, Elimar Riesebieter wrote:
> On Fri, 06 Dec 2002 the mental interface of 
> Mark L. Kahnt told:
> > On Fri, 2002-12-06 at 20:31, Nathan E Norman wrote:
> > > On Fri, Dec 06, 2002 at 11:54:52AM -0500, Jason Wojciechowski wrote:
> > > > Nathan E Norman (nnorman@incanus.net) wrote:
> > > > 
> > > > > Note that some people (like Linus) say that anyone who runs a
> > > > > framebuffer console is insane.
>         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> stupid!  
> > 
> > Linus (from a quote and various other observations I've seen) considers
> > framebuffer on i386, umm, "less than ideal" (my words) as a general
> > configuration choice as, from my understanding of the words I've seen
> > and how it exists in the kernel, it is an incomplete and inconsistent
> > implementation on i386 - some graphics cards are not so well supported
> > as others, and simple SVGA functions are capable of the significant
> > range of framebuffer functions with vastly less overhead than running
> > the console strictly in graphics mode.
> > 
> > Framebuffer is needed on some systems that don't have text modes
> > comparable to the PC-style systems, but unless you *need* framebuffer
> > functions on an i386-style PC, you are not necessarily doing things the
> > most efficient way if you are doing most of your work on the console
> > strictly with text and no modified fonts, and as various graphics cards
> > only have *experimental* code implementing framebuffer, you are playing
> > with potential buggy code in the kernel for not necessarily any
> > performance benefit over a currently more reliable user-mode solution.
> I am running framebuffer at vt's since 2 years now on i386 without
> problems. I.e to read mails with mutt on a vt is quite more comfortable
> to read and to handle as on a X-Window. I view pictures and pdf`s
> with fbi (fbgs), my vt's are running with fbgetty. Never had any
> problems, so why it is insane?

You are using the aspects of framebuffer that justify it - if the only
graphics a system ever does is in X11 and it isn't running on
framebuffer, using framebuffer to effectively *emulate* text mode on a
graphic screen is a good bit of relative overhead and making the
effective text mode far more hardware specific than if it had been
handled through SVGA support.
> Never found a bug in the *experimental* drivers.

I had problems for a good stretch of kernels with atyfb - about the only
chances of hanging my console was in scrolling a framebuffer vt while
X11 was also running - not always, but enough times. Full screen
presentation of text also often didn't display correctly unless I
switched to another vt and then back. I'd call that buggy (although it
works fine now,) and I've heard rumblings that problems exist with
numerous other graphic chips currently.

Part of the underlying question is whether you want to commit your
cpu(s) to framebuffer, setiathome, halting, or other demands,
particularly for users that use no part of framebuffer other than
compiling it into the kernel (ie. 80x25, no fbgetty, no fbgs, no fbi, no
special font, no fbtv - I have used it with my tv tuner card and a vcr
to play back very sharp images of a kid's birthday party.) In that case,
on i386 PC-style platforms, it is doing basic text mode functions "the
hard way around."

Me, I use framebuffer, the Sun 12x22 font which makes the display vastly
sharper for me to read (and as I've stated previously on this topic,
I've worked in graphic design, and pay strong attention to typography,
and the Sun 12x22 font is a very handsome character set, similar to
Palatino,) fbi and fbgs when X11 is not available (ie. dingy me broke
some setting) and fbtv. I've never gotten fbgetty to work in a way that
would warrant my switching from getty at this point, but I suspect that
is as much a shortcoming in documentation as it is in my time to think
out what I want to do with it (and no, I don't have the time to think
that out now for a sudden eduacation through this mailing list.)

For me, framebuffer is a plus, but only very rarely does it come close
to being essential for me, and if atyfb had stayed buggy, or I had a
different graphics card where the framebuffer was still not working, I
wouldn't be running it. I can understand Linus' position: it has been a
lot of work for the kernel developers in an area that, while essential
for a handful of users, is not much beyond an aesthetic benefit for most
others, if they even actually use it, and given that still much work is
needed. It also isn't in one of the areas of computing aesthetics that I
get the impression he finds most interesting, as it is not an area of
innovative logic, but rather basic functional code needing to focus on
matching the performance needs of hardware - if he was more interested
in this sort of code, we'd have X11 in the kernel ;) I could be wrong -
he could love the concept and be frustrated with its slow completion or
greater complexity due to the breadth of cards and functions to be
implemented, but I've long found that what I read of what Linus says, he
is usually clear about what he is saying and what he means.
> Elimar
> -- 
>     .~.
>     /V\   L   I   N   U   X
>    /( )\ >Phear the Penguin<
>    ^^-^^
ML Kahnt New Markets Consulting
Tel: (613) 531-8684 / (613) 539-0935
Email: kahnt@hosehead.dyndns.org

Attachment: signature.asc
Description: This is a digitally signed message part

Reply to: