C&T acceleration support, needs testers.

C&T acceleration support, needs testers.

[Ryuichi Oikawa]

Hi,

I'm trying to implement C&T 6555x HW acceleration to the X servers.
I put a Xpmac binary:
 ftp://ppc.linux.or.jp/pub/users/oikawa/Xserver/hw/powermac/Xpmac-19990731-glibc2.1.gz
XF68_FBDev version hasn't been compiled, yet.

Because I don't have any C&T 6555x hardware, I need testers for debugging.
Thank you for your help.

Ryuichi Oikawa
roikawa@rr.iij4u.or.jp

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[Vinai]

On Sun, 1 Aug 1999, Ryuichi Oikawa wrote:

}I'm trying to implement C&T 6555x HW acceleration to the X servers.
}I put a Xpmac binary:
} ftp://ppc.linux.or.jp/pub/users/oikawa/Xserver/hw/powermac/Xpmac-
}19990731-glibc2.1.gz
}XF68_FBDev version hasn't been compiled, yet.
}
}Because I don't have any C&T 6555x hardware, I need testers for
}debugging.  Thank you for your help.

Hi,

I grabbed your Xpmac binary this morning and tried it out on a 3400/240
running R5.  It seemed to work well for the most part.  Very noticeable
acceleration, especailly over the XF86 server I was currently using.  I
did run into a couple problems though.

- If windows are moved around too quickly, vertical bands appear.  If I
  explicitly request that part of the screen to be redrawn (e.g. by my
  closing the xterm and reopening it), the lines disappear.

- When I passed it the wrong option (I used "startx --depth 16" instead
  of the correct "startx -- -depth 16", it off course ignored the wrong
  swtich and started up in 8-bit colour).  But on exiting X, it locked
  up the machine completely.  I couldn't even switch to another virtual
  console to kill the process or reboot the machine, so I ended up with
  a hard reboot).  For a window managers - I am using an old version of
  AfterStep that I recompiled from the src.rpm for R5.  It is version
  1.0.7a (Yes - I know it's old, but it is small, I know how to set it
  up, and does all I need ;-)

Otherwise, I have not found any other problems, and the person that uses
this machine is actually pretty happy right now, as am I ;-)  Thanks for
the great work !!

cheers
vinai




[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[cesli]

On Sat, Jul 31, 1999 at 03:12:24PM -0500, Vinai supposed:

 Vinai }  - If windows are moved around too quickly, vertical bands appear.  If I
 Vinai }    explicitly request that part of the screen to be redrawn (e.g. by my
 Vinai }    closing the xterm and reopening it), the lines disappear.
same phenomenon on my 3400, but no problem.
 Vinai }  Otherwise, I have not found any other problems, and the person that uses
 Vinai }  this machine is actually pretty happy right now, as am I ;-)  Thanks for
 Vinai }  the great work !!
Dito
THX to Ryuichi
Claus

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[Don Lunder]

I got to try out this X server, and it seems to work quite well.  Here is 
the output on starting it:

Xpmac / X Window System (X11R6.3)
Release Date: July 31, 1999

Running in mode=10, depth=16; req mode=0 depth=0
Using VT number 7
Mapped frame buffer (physaddr 0x81800000 size 0xea610) at 0x30057000.
C&T 6555x detected. Mapping MMIO/IO regsiter...
Mapped physaddr 0x81400000 size 0x20000 at 0x30014000.
Mapped physaddr 0xf2000000 size 0x1000 at 0x30034000.
C&T 6555x successfully mapped. Starting acceleration...


Running WindowMaker-0.60.0-4, on a PowerBook 2400/180

The only thing I saw out of the ordinary, were the verticle bars that would 
appear when moving windows fast.  (Altho moving the xmms (mp3 player) 
window didn't seem to leave any artifacts)  Also, if I drug windows between 
desktops, verticle bars or other artifacts would appear in the window being 
moved.  Usually Yellow or Blue it seems (this could just be for me tho)

I booted in Linux with it running a video driver (not the OF based one), 
does this make a diffrence, would you like to know of any other setups?

On a side note, I noticed that when I run with the video driver on the 
2400, the first time I start X, instead of a "grey" pixely screen, it has a 
blue hue to it, and it will seg fault, but if I try running it again, it 
starts up correctly, and goes into WindowMaker.

On my machine, I was able to exit WindowMaker just fine with out any 
errors, or hard reboots.

I'll keep playing with it, and tell you what I find, if you have any 
specific requests, please do ask!

Thanks for your hard work, time and effort!!!!

Don Lunder

PS After Spell checking the email, I see that register has a typo in the 
Xpmac out put. <shrug>


[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[Vinai]

On Sat, 31 Jul 1999, Don Lunder wrote:

}On a side note, I noticed that when I run with the video driver on the 
}2400, the first time I start X, instead of a "grey" pixely screen, it  
}has a blue hue to it, and it will seg fault, but if I try running it
}again, it starts up correctly, and goes into WindowMaker.

I had this problem with the original XF86 server, that I replaced with
the Xpmac that was posted this morning ...  It seemed that MacOS would
set the bit depth of the screen to one the XF86 could not recognize or
handle, so that is why it would be that weird colour and segfault. When
that happened, and the X-server exited, it said it was trying to set a
24-bit screen.  But the second time (when it would run correctly) it'd
say 15-bit screen.  So it looks like X set the bit depth correctly on
the first launch and was able to use it the second time.

But with the new X-server, I don't have this problem...

Cheers
vinai




[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[Don Lunder]

Okay, I did find something strange with this new X server.

I have my f11 and f12 keys set up for the diffrent mouse clicks, by passing 
the kernel an adb_buttons command within bootX.

For some reason, all of my 2/3rd button mouse clicks go to the root window, 
to bring up the various windows (instead of pasting, etc etc)  even when i 
click on the dock (window maker, still) and xterms, etc etc, you name it... 
it sends it to the root window (or something of that idea)

Well thanks, keep up the great work!

Don Lunder


[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[Ryuichi Oikawa]

Hi, 
Thank you for responses.

From: Don Lunder <speedy@carboncode.com>
Subject: Re: C&T acceleration support, needs testers.
> I'll keep playing with it, and tell you what I find, if you have any 
> specific requests, please do ask!

I'd like to further know,

- Difference between 8bp and 16bpp. 8bpp doesn't cause endiannes problem but
 may be more unstable.

- Vertical band problem, more precisely(ex. both 8 and 16bpp? happens
 at what X command? appearing location and size?, etc.)

- Difference between chipsfb and offb(BootX' "no video driver").
  When kernel booted with "no video driver", Xpmac with -ct6555 option
  does BitBLT origin offset calculation, color palette handling and aperture
  setting. I have report that reading color palette immediately causes
  machine check exception resulting in kernel panic(why??). And if aperture
  setting is failed, Xpmac goes into infinite loop.

> PS After Spell checking the email, I see that register has a typo in the 
> Xpmac out put. <shrug>
Oops! :-)

Anyway, I'll soon upload the source code(at the same place, please check)
so that anyone can try and fix it, before I'm on a vacation.

In the mean time, I had a advice that I should end it and spend time
for new XFree(4.0). I'd be glad if someone continued to work on this,
though I haven't decided anything yet.


Regards,

Ryuichi Oikawa
roikawa@rr.iij4u.or.jp

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[Geert Uytterhoeven]

On Mon, 2 Aug 1999, Ryuichi Oikawa wrote:
> In the mean time, I had a advice that I should end it and spend time
> for new XFree(4.0). I'd be glad if someone continued to work on this,
> though I haven't decided anything yet.

Yes, working on XFree86 4.0 (3.9.15 these days) will benefit all platforms.

Greetings,

						Geert

--
Geert Uytterhoeven                     Geert.Uytterhoeven@cs.kuleuven.ac.be
Wavelets, Linux/{m68k~Amiga,PPC~CHRP}  http://www.cs.kuleuven.ac.be/~geert/
Department of Computer Science -- Katholieke Universiteit Leuven -- Belgium


[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[Ryuichi Oikawa]

From: Ryuichi Oikawa <roikawa@rr.iij4u.or.jp>
Subject: C&T acceleration support, needs testers.
> I'm trying to implement C&T 6555x HW acceleration to the X servers.
> I put a Xpmac binary:
>  ftp://ppc.linux.or.jp/pub/users/oikawa/Xserver/hw/powermac/Xpmac-19990731-glibc2.1.gz
> XF68_FBDev version hasn't been compiled, yet.
 It's just been compiled:
ftp://ppc.linux.or.jp/pub/users/oikawa/Xserver/hw/xfree68/XF68_FBDev-19990801-glibc2.1.gz

 Because FB_ACCEL code isn't defined for C&T 65550 and it requires to accesss
both MMIO and IO space, it scans PCI configuration space through /proc/bus/pci.
I'm not quite sure if it can detect C&T 6555x properly (just write and compile,
not checked, yet). I hope Geert or someone accept the patch...


Regards,

Ryuichi Oikawa
roikawa@rr.iij4u.or.jp

[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]

Re: C&T acceleration support, needs testers.

[Geert Uytterhoeven]

On Sun, 1 Aug 1999, Ryuichi Oikawa wrote:
> From: Ryuichi Oikawa <roikawa@rr.iij4u.or.jp>
> Subject: C&T acceleration support, needs testers.
> > I'm trying to implement C&T 6555x HW acceleration to the X servers.
> > I put a Xpmac binary:
> >  ftp://ppc.linux.or.jp/pub/users/oikawa/Xserver/hw/powermac/Xpmac-19990731-glibc2.1.gz
> > XF68_FBDev version hasn't been compiled, yet.
>  It's just been compiled:
> ftp://ppc.linux.or.jp/pub/users/oikawa/Xserver/hw/xfree68/XF68_FBDev-19990801-glibc2.1.gz
> 
>  Because FB_ACCEL code isn't defined for C&T 65550 and it requires to accesss
> both MMIO and IO space, it scans PCI configuration space through /proc/bus/pci.
> I'm not quite sure if it can detect C&T 6555x properly (just write and compile,
> not checked, yet). I hope Geert or someone accept the patch...

The Right Way(TM) to make XF68_FBDev detect a C&T 65550 is to request one or
more FB_ACCEL_* IDs from me.

Greetings,

						Geert

--
Geert Uytterhoeven                     Geert.Uytterhoeven@cs.kuleuven.ac.be
Wavelets, Linux/{m68k~Amiga,PPC~CHRP}  http://www.cs.kuleuven.ac.be/~geert/
Department of Computer Science -- Katholieke Universiteit Leuven -- Belgium


[[ This message was sent via the linuxppc-dev mailing list.  Replies are ]]
[[ not  forced  back  to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. Please check http://lists.linuxppc.org/ ]]
[[ and http://www.linuxppc.org/ for useful information before posting.   ]]