Subject: Re: Retina Z3 hardware cursor problem
To: Andreas Johansson <email@example.com>
From: None <firstname.lastname@example.org>
Date: 11/24/1994 09:19:54
> I have been using the Xamiga24 Xserver for my machine a few days, and it works
> pretty well except for the mouse pointer. It is positioned correctly in x-led,
> but in y-led it will only use 5 different locations on my 1280x1024 screen.
> That is, only the most significant 8 bits of the y cursor positioning actually
I *guess* this is yet another flavor of malfunction of the 64x64 bit
cursor.. I'm "sorry" to say that on my machine, only the inverting of the
sprite at y=0 happens, other than that, it works just fine. For those that
get nastier behavior, you'll have to change the driver to use 32x32bit sprites.
Xamiga24 should handle this ok, since it queries the maximum dimension of
the sprite before setting it up.
> Now, on the topic of the RetinaZ3 kernel driver routines, I have looked at them
> a little (actually I have ported the Retina Z3 console to Amiga Linux), and
> it strikes me as a little unnecessary to busy wait for the blitter to finish.
> Or are those
> while ((blablabla & 1) == 0);
> something else than I think? :) Ok, what I would like to know is this:
It might be unnecessary to do this 95% of the time, but since the blitter is
started like this:
- stop any possible running operation
- start new operation
you get into quite a mess if you don't wait for the blitter to finish and
start a new op before the old one has finished. Nice example to trigger this
behavior is if you set the "DragWindow" (or similar..) resource for olwm so
it moves entire windows, not just window frames.
> Is the only reason for these busywaits and the not-so-clever-retina-memory-
> management (that is, only framebuffer, pointer and temporary blitplace) that
Nah, not at all.
> noone has taken the time to enhance the routines? I surely would appreciate
> if the machine would not have to swap portions of the display against
> harddrive when there is >2.5Mb free in the Retina memory. Those 12Mb of mine
> are not much when it comes to NetBSD-megs :)
Huh? Eh, swapping display memory? You have a vivid imagination :-)) Besides,
those busy-loops -as you stated- look like
while ((blablabla & 1) == 0);
so if the enginge is already finished, this is a test on one bit, nothing
more, no loop.
> On second thought, this is pretty odd. I belive PC-Linux is able to run X
> without swapping with only 8Mb ram. My machine fills up all my 12 Mb as I
> start X, xclock, xbiff, xload, four xterms and fvwm. Is linux really so much
> more efficient (in ram usage) or is something wrong set up in my system?
Does Linux also use 1280x1024 virtual screen size? If you're concerned about
memory use, you might want to cut on this parameter (in the monitor
definitions in the driver), since X tends to back up lots of the display area
somewhere, display size indirectly influences X memory consumption.
EUnet Switzerland Tel: +41 1 291 45 80 Markus Wild
Zweierstrasse 35 Hotline: +41 1 291 45 60 email@example.com
CH-8004 Zuerich Fax: +41 1 291 46 42 S=mw;P=EUnet;A=EUnet;C=CH
>Solaris 2 is not an upgrade from Solaris 1. They just want you to THINK it is.