[ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[Jakub Jelinek]

Hi!

Myself and David Miller would like to announce a test release
of our Xsun24 rewrite of the Creator/Creator3D/Elite3D driver.
We've spent on it quite a few months and would like to offer it for public
use if no showstoppers are found.
Basically, your Creator/3D/Elite3D should run now much faster, because we
try to get as much as possible from the chip.

Xmark93 went from 23808 to 39247, Xstones from 86842 to 466236 (both x11perf
and xbench are benchmarks measuring speed of most commonly used X
operations). All these numbers are in 32bpp (most of the PCI cards might
have good numbers in 8bpp (but I doubt it could have nearly 500kXstones),
but will be slow as pig in 32bpp).

For exact numbers and rpms see
http://ultra.linux.cz/ftp/OS/Linux/Sparc/local/XFree86/
announcement can be found at
http://ultra.linux.cz/ftp/OS/Linux/Sparc/local/XFree86/creator.html

We'd like to get some feedback on it, fix last few bugs if some appear and
then put this into distributions.

Once we integrate it into XFree86 4.0, we'll add support for mixed 32/8bpp
windows on the same screen and extensive 3D support (even for the 2D
models), so there are really things to improve.

BTW: The reported bug with mouse cursor disappearing near the left/top edges
is fixed.

Cheers,
    Jakub
___________________________________________________________________
Jakub Jelinek | jj@sunsite.mff.cuni.cz | http://sunsite.mff.cuni.cz
Administrator of SunSITE Czech Republic, MFF, Charles University
___________________________________________________________________
UltraLinux  |  http://ultra.linux.cz/  |  http://ultra.penguin.cz/
Linux version 2.2.3 on a sparc64 machine (3958.37 BogoMips)
___________________________________________________________________

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[Matthew Hawkins]

Hi Jakub!

On Wed, 17 Mar 1999, Jakub Jelinek wrote:

> Once we integrate it into XFree86 4.0, we'll add support for mixed 32/8bpp
> windows on the same screen
[--- snip ---]

Is this a chipset-specific feature, or is it possible for other X servers to
be similarly modified?  (I've always wanted to be able to fire up Xnest in a
depth different to the _real_ server running on :0, and individual windows
would be even better... I suppose one simply uses -visualid to choose?

-- 
Matt

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[David Miller]

   Date: 	Wed, 17 Mar 1999 23:59:12 +0000
   From: Matthew Hawkins <matt@mail.goldweb.com.au>

   Is this a chipset-specific feature, or is it possible for other X
   servers to be similarly modified?

Yes and no.  It can be done in software, but slowly.

Most modern cards support 8bpp and 32bpp by way of WIDs (Window IDs),
but mostly they do this for the sake of efficient direct rendering in
OpenGL etc.  See this magnificant paper by the SGI OpenGL people to
see how this all works:

http://reality.sgi.com/opengl/direct

Essentially the WIDs are there to provide advanced and efficient
clipping facilities.  A WID specifies who can write into that window,
and what his depth/colormap setting look like, and on most Sun cards
it also can specify the buffer (when using double buffering) that is
currently displayed for that WID's pixels.

So say window 1 has WID 1 and window 2 has WID 2.  The visible pixels
of window 1 are written with the 8-bit WID of 1.  So if window 2
partially is behind window 1, when window 2 tries to write pixels into
that hidden area, the graphics card just throws away the written
pixels because it sees the WID 1 pixels there already.  So an OpenGL
application doing direct rendering need not check for any clips as it
renders into it's X window.

SGI splits out WIDs into CIDs and DIDs in their architecture.  How to
encode the WID or whatever information is a hardware tradeoff
decision, some methods allow you to run-length encode the buffer and
colormap id's for each pixel so that the retrace scan logic can work
with less information as it refreshes the screen.

Anyways, go read the SGI paper, it's cool technology.

Later,
David S. Miller
davem@redhat.com

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[Elliott Hughes]

ok, mixed results. on the one hand, the new stuff is _much_ faster,
and the mouse cursor no longer disappears at the edges of the
screen. unfortunately, i've been having crashes that kill my system (to
the extent that i can't Stop-A, never mind change virtual console
or even ping!).

it seems to happen with bitblts. i use 9term and wily a lot, and
i've now had both of these cause a crash. the first was clicking
on 9term's scrollbar; the window was partially redrawn, and then
the machine froze. with wily, i'd selected some text to overwrite.
i started to type, and the area to be overwritten was cleared, but
the text below this area wasn't bitbltted up. (come to think of
it, this was basically what happened with 9term too; i was
scrolling down, so what remained on the screen was bitbltted
up to the top, but this time not all of the area to be cleared
was cleared before the machine stiffed.)

i hope this is enough to get you on the right track. if you want
my binaries for 9term or wily, i can make them available.

thanks,
Elliott

-- 
"Plan 9? Ah, yes: Plan 9 deals with the resurrection of the dead."
	http://plan9.bell-labs.com/plan9/

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[David Miller]

What kernel are you running and where did you get it from?

Later,
David S. Miller
davem@redhat.com

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[Ryan Cresawn]

Elliott Hughes writes:
 > 
 > ok, mixed results. on the one hand, the new stuff is _much_ faster,
 > and the mouse cursor no longer disappears at the edges of the
 > screen. unfortunately, i've been having crashes that kill my system (to
 > the extent that i can't Stop-A, never mind change virtual console
 > or even ping!).
 > 
 > it seems to happen with bitblts. i use 9term and wily a lot, and
 > i've now had both of these cause a crash. the first was clicking
 > on 9term's scrollbar; the window was partially redrawn, and then
 > the machine froze. with wily, i'd selected some text to overwrite.
 > i started to type, and the area to be overwritten was cleared, but
 > the text below this area wasn't bitbltted up. (come to think of
 > it, this was basically what happened with 9term too; i was
 > scrolling down, so what remained on the screen was bitbltted
 > up to the top, but this time not all of the area to be cleared
 > was cleared before the machine stiffed.)

I've had very similar problems.  I use WindowMaker and sometimes when
I click on a title bar and begin to move a window the whole computer
will lock up.  Lock up prevents Stop-A, ping, etc.  I using kernel
2.2.0-pre5.

The speed increase is incredible, however the lock ups are a pain.  If 
there is some information you need just let me know.  I will help if I 
can.  Oh yeah, I have an Ultra2 w/Creator 3D.

-- 
Ryan

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[David Miller]

   From: Ryan Cresawn <cresawn@arbs1.larc.nasa.gov>
   Date: 	Sun, 21 Mar 1999 22:43:35 -0500 (EST)

   The speed increase is incredible, however the lock ups are a pain.
   If there is some information you need just let me know.  I will
   help if I can.  Oh yeah, I have an Ultra2 w/Creator 3D.

Please grab the latest kernel and try to reproduce.  I can almost
guarentee that you will not be able to as I have been doing an
enormous amount of stress testing of things.

There are two ways to get the latest at the moment:

1) http://cvs.on.openprojects.net/

   Follow the instructions for the vger kernel mirror repository
   and checkout the cvs module named "linux"

2) Grab the latest 2.2.3 linux sources and apply the latest
   2.2.4 pre-patch on top, this will be essentially the same
   thing as what you'd get from method #1.

   ftp://ftp.xx.kernel.org/pub/linux/kernel/v2.2/linux-2.2.3.tar.bz2
   ftp://ftp.xx.kernel.org/pub/linux/kernel/testing/pre-patch-2.2.4-6.gz

   Replace "xx" with your country code, it may take a little while for
   the pre-patch to reach the mirrors.

Please report back to me asap as Linus is going to release the real
2.2.4 some time tomorrow and if this is a kernel bug I need to fix it
before tomorrow morning.

Thanks.

Later,
David S. Miller
davem@redhat.com

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[Elliott Hughes]

> What kernel are you running and where did you get it from?

i was running 2.2.0-pre5, but i've since switched to 2.2.4 as
you suggested, and i've had no trouble since. i've been doing
the same kind of stuff as i was doing yesterday -- i tried to
repeat exactly what i did yesterday -- and haven't been able
to provoke a crash.

i'll shout if anything goes wrong, but so far it looks as if you
can assume it's fixed.

thanks,
Elliott

-- 
"Plan 9? Ah, yes: Plan 9 deals with the resurrection of the dead."
	http://plan9.bell-labs.com/plan9/

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[Ryan Cresawn]

David Miller writes:

 > Please grab the latest kernel and try to reproduce.  I can almost
 > guarentee that you will not be able to as I have been doing an
 > enormous amount of stress testing of things.
 >
 > There are two ways to get the latest at the moment:
 > 
 > 1) http://cvs.on.openprojects.net/
 > 
 >    Follow the instructions for the vger kernel mirror repository
 >    and checkout the cvs module named "linux"
 > 
 > 2) Grab the latest 2.2.3 linux sources and apply the latest
 >    2.2.4 pre-patch on top, this will be essentially the same
 >    thing as what you'd get from method #1.
 > 
 >    ftp://ftp.xx.kernel.org/pub/linux/kernel/v2.2/linux-2.2.3.tar.bz2
 >    ftp://ftp.xx.kernel.org/pub/linux/kernel/testing/pre-patch-2.2.4-6.gz
 > 
 >    Replace "xx" with your country code, it may take a little while for
 >    the pre-patch to reach the mirrors.
 > 
 > Please report back to me asap as Linus is going to release the real
 > 2.2.4 some time tomorrow and if this is a kernel bug I need to fix it
 > before tomorrow morning.

Dave,

I have the kernel source, however I don't know how to compile/install
it.  I am very familiar with the Intel version but not the Sparc
version.  Is there a HOWTO?

-- 
Ryan

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[Jakub Jelinek]

> Dave,
> 
> I have the kernel source, however I don't know how to compile/install
> it.  I am very familiar with the Intel version but not the Sparc
> version.  Is there a HOWTO?

If you have the sparc64-linux-gcc compiler and recent binutils (UltraPenguin
1.1.9 e.g. has such packages), then just cd /usr/src/linux; make config (if
you don't want to bother with selecting config options, do make oldconfig
instead, it will have reasonable defaults). Then
make dep && make && make modules && make modules_install
cp vmlinux /boot/vmlinux-2.2.4
echo "disk /boot/vmlinux-2.2.4" > /proc/sys/kernel/reboot-cmd
/sbin/reboot

Cheers,
    Jakub
___________________________________________________________________
Jakub Jelinek | jj@sunsite.mff.cuni.cz | http://sunsite.mff.cuni.cz
Administrator of SunSITE Czech Republic, MFF, Charles University
___________________________________________________________________
UltraLinux  |  http://ultra.linux.cz/  |  http://ultra.penguin.cz/
Linux version 2.2.3 on a sparc64 machine (3958.37 BogoMips)
___________________________________________________________________

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[Ryan Cresawn]

Jakub Jelinek writes:

 > If you have the sparc64-linux-gcc compiler and recent binutils (UltraPenguin
 > 1.1.9 e.g. has such packages), then just cd /usr/src/linux; make config (if
 > you don't want to bother with selecting config options, do make oldconfig
 > instead, it will have reasonable defaults). Then
 > make dep && make && make modules && make modules_install
 > cp vmlinux /boot/vmlinux-2.2.4
 > echo "disk /boot/vmlinux-2.2.4" > /proc/sys/kernel/reboot-cmd
 > /sbin/reboot

Jakub,

Thanks for the information.  I tried your instructions, however it
didn't complete.  Here's where it stopped.

By the way, I have a really hard time understanding the version
numbering within the UltraPenguin distribution.  It seems to me that
the UltraPenguin-1.1.9 has been evolving under the same name for
months.  I wonder if it would be wise to change to something simple
based on the date like WINE.  What do you think?


##### this is the result of "make" ######


sparc64-linux-gcc -D__KERNEL__ -I/usr/src/linux/include -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -m64 -pipe -mno-fpu -mcpu=ultrasparc -mcmodel=medlow -ffixed-g4 -fcall-used-g5 -fcall-used-g7 -Wno-sign-compare   -c -o fonts.o fonts.c
fonts.c:61: #error No fonts configured.
fonts.c: In function `fbcon_find_font':
fonts.c:72: warning: comparison is always 0 due to limited range of data type
make[3]: *** [fonts.o] Error 1
make[3]: Leaving directory `/usr/src/linux-2.2.4/drivers/video'
make[2]: *** [first_rule] Error 2
make[2]: Leaving directory `/usr/src/linux-2.2.4/drivers/video'
make[1]: *** [_subdir_video] Error 2
make[1]: Leaving directory `/usr/src/linux-2.2.4/drivers'
make: *** [_dir_drivers] Error 2

-- 
Ryan

Re: [ANNOUNCE] Test release of rewritten Creator/Creator3D/Elite3D support

[David Miller]

   Date: Mon, 22 Mar 1999 13:38:17 +0100 (CET)
   From: Elliott Hughes <Elliott.Hughes@genedata.com>

   i'll shout if anything goes wrong, but so far it looks as if you
   can assume it's fixed.

Thanks so much for testing this.

Later,
David S. Miller
davem@redhat.com