gatos-devel Mailing List for GATOS

Great, I did not know that Rage128 cards needed this as well.
I'll merge this in as soon as I get some free time (hopefully next week).

                        Vladimir Dergachev

On Thu, 28 Feb 2002, Sebastian Ehrenreich wrote:

> Hello !
>
> As mentioned in a private email contact, there is a problem with some
> (European?) Rage Fury Pro Cards in relation to the definition of the
> composite and svideo input ports. Unfortunately, the options
>
> 	RageTheatreCompositePort
> 	RageTheatreSVideoPort
>
> introduced by Vladimir are not supported with r128 based cards. I hacked
> the r128_driver.c and r128.h code to include this very useful options
> too. I attached a diff I made over my version of these two files and the
> original XFree4.2.0 sources. I am not familiar with CVS and that stuff,
> so I can not check the changes in the development tree (Maybe there is
> somebody who can help me doing this...)
>
> Please do not wonder about the additional "noTV" option I introduced.
> This part is under development, because I have severe problems crashing
> the XServer at higher resolutions with a TV connected, even with the
> "not TV-Out capable" versions of ati.2 drivers. I am currently searching
> for the right place to fix that problem, therefore I inserted this new
> switch. It has no function yet.
>
> By the way: I have operating TVOut with my Rage Fury Pro card and PAL.
> Here is a short "cookbook" how it works for me (640x470x16 mode only):
>
> Boot Linux with TV connected. I use vga=788 in my lilo.conf. During
> bootup, the TV shows a blue picture. (ati r128 framebuffer is used as
> kernel option) Then I start an XFRee4.2.0 server with ati.devel drivers
> included. The tricky thing is the modline and/or XF86Config, here is an
> excerpt of mine:
>
> Section "Modes"
>    Identifier 	"Modes[0]"
> Modeline 	"640x480" 31.38 640 664 784 832 480 490 495 522
> EndSection
>
>
> Section "Screen"
>    DefaultDepth	16
>    SubSection "Display"
>      Depth	16
>      Modes  	"640x480"
>    EndSubSection
>    Device  	"Device[0]"
>    Identifier  	"Screen[0]"
>    Monitor  	"Monitor[0]"
> EndSection
>
>
> Section "Device"
>    BoardName  	"Rage128ProAGP4x(TMDS)"
>    BusID  	"1:0:0"
>    Driver  	"r128"
>    Option        "RageTheatreCompositePort" "2"
> #  Option	"NoTV"
>    Identifier  	"Device[0]"
>    VendorName  	"ATI"
> EndSection
>
> Unfortunately exiting XFree results in a black screen and a monitor with
> lost sync. But you can execute commands (normally shutdown -r now ;--))
> or you can restart XFree.
> A lot of things posted in the past don't work for me, e.g. activating
> TVOut without having a TV connected during boot or this console
> reconstruction hack. Especially this hack results in a totally
> inoperatable XFree installation.
>
> So, that's it so far.
>
> Sebastian
>
> P.S: Many thanks to everybody working so hard for the ATI drivers. Great
> job!
> P.P.S: Watching DVDs with xine on my TV now works perfectly. Now I can
> remove Windows (brrr) from my box! ;-))
>
>

Hello !

As mentioned in a private email contact, there is a problem with some 
(European?) Rage Fury Pro Cards in relation to the definition of the 
composite and svideo input ports. Unfortunately, the options 	

	RageTheatreCompositePort
	RageTheatreSVideoPort

introduced by Vladimir are not supported with r128 based cards. I hacked 
the r128_driver.c and r128.h code to include this very useful options 
too. I attached a diff I made over my version of these two files and the 
original XFree4.2.0 sources. I am not familiar with CVS and that stuff, 
so I can not check the changes in the development tree (Maybe there is 
somebody who can help me doing this...)

Please do not wonder about the additional "noTV" option I introduced. 
This part is under development, because I have severe problems crashing 
the XServer at higher resolutions with a TV connected, even with the 
"not TV-Out capable" versions of ati.2 drivers. I am currently searching 
for the right place to fix that problem, therefore I inserted this new 
switch. It has no function yet.

By the way: I have operating TVOut with my Rage Fury Pro card and PAL. 
Here is a short "cookbook" how it works for me (640x470x16 mode only):

Boot Linux with TV connected. I use vga=788 in my lilo.conf. During 
bootup, the TV shows a blue picture. (ati r128 framebuffer is used as 
kernel option) Then I start an XFRee4.2.0 server with ati.devel drivers 
included. The tricky thing is the modline and/or XF86Config, here is an 
excerpt of mine:

Section "Modes"
   Identifier 	"Modes[0]"
Modeline 	"640x480" 31.38 640 664 784 832 480 490 495 522
EndSection


Section "Screen"
   DefaultDepth	16
   SubSection "Display"
     Depth	16
     Modes  	"640x480"
   EndSubSection
   Device  	"Device[0]"
   Identifier  	"Screen[0]"
   Monitor  	"Monitor[0]"
EndSection


Section "Device"
   BoardName  	"Rage128ProAGP4x(TMDS)"
   BusID  	"1:0:0"
   Driver  	"r128"
   Option        "RageTheatreCompositePort" "2"
#  Option	"NoTV"
   Identifier  	"Device[0]"
   VendorName  	"ATI"
EndSection

Unfortunately exiting XFree results in a black screen and a monitor with 
lost sync. But you can execute commands (normally shutdown -r now ;--)) 
or you can restart XFree.
A lot of things posted in the past don't work for me, e.g. activating 
TVOut without having a TV connected during boot or this console 
reconstruction hack. Especially this hack results in a totally 
inoperatable XFree installation.

So, that's it so far.

Sebastian

P.S: Many thanks to everybody working so hard for the ATI drivers. Great 
job!
P.P.S: Watching DVDs with xine on my TV now works perfectly. Now I can 
remove Windows (brrr) from my box! ;-))

Vladimir Dergachev wrote:

>>>>
>>>>I now that
>>>>
>>>>outf(CAPTURE_CONFIG, ENABLE_ATI_CAPTURE_DOUBLEBUF | 0x100);
>>>>
>>>>should do it on  ATI Rage Mobility M1 chip - see the ZV- patch for
>>>>atixv.c (in the CVS ati.2
>>>>sources) at http://www.linuxtv.org. I tried that on the ATI 3D Rage LT
>>>>Pro chip
>>>>which resulted in a X-server crash. I guess that I have to use another
>>>>
>>>A crash or lockup ? It should not segfault because of this statement, if
>>>it does the error is someplace else.
>>>
>>This is the message I get at the X11 log:
>>
>>>
>>>Fatal server error:
>>>Caught signal 11.  Server aborting
>>>
>
>Signal 11 is a segmentation fault. So X is trying to access memory that
>does not exist - use diff to check your modifications versus unmodified
>ati.2 and check all the pointer usage in them.
>
OK. I found the bug. Now X is running - but what about the
registers that control ZV?

>>>
>>Do you have any info of the CAPTURE_CONFIG register? Is there any
>>other register that controls ZV? Ideally, I want to have a function
>>that checks if a ZV capable card is inserted before switching on
>>ZV.
>>
>>
>>Fredrik
>>
>>
>>
>

On Thu, 28 Feb 2002, Fredrik Lingvall wrote:

>
>
> Vladimir Dergachev wrote:
>
> >
> >On Thu, 28 Feb 2002, Fredrik Lingvall wrote:
> >
> >>I'm trying to get ZV to work on an ATI 3D Rage LT Pro chip.
> >>Which register do I use to switch on ZV?
> >>
> >>I now that
> >>
> >>outf(CAPTURE_CONFIG, ENABLE_ATI_CAPTURE_DOUBLEBUF | 0x100);
> >>
> >>should do it on  ATI Rage Mobility M1 chip - see the ZV- patch for
> >>atixv.c (in the CVS ati.2
> >>sources) at http://www.linuxtv.org. I tried that on the ATI 3D Rage LT
> >>Pro chip
> >>which resulted in a X-server crash. I guess that I have to use another
> >>
> >
> >A crash or lockup ? It should not segfault because of this statement, if
> >it does the error is someplace else.
> >
> This is the message I get at the X11 log:
>
> >
> >
> >Fatal server error:
> >Caught signal 11.  Server aborting
> >

Signal 11 is a segmentation fault. So X is trying to access memory that
does not exist - use diff to check your modifications versus unmodified
ati.2 and check all the pointer usage in them.

                      Vladimir Dergachev

> >
> >When reporting a problem related to a server crash, please send
> >the full server output, not just the last messages.
> >This can be found in the log file "/var/log/XFree86.0.log".
> >Please report problems to xf...@xf....
> >
> Do you have any info of the CAPTURE_CONFIG register? Is there any
> other register that controls ZV? Ideally, I want to have a function
> that checks if a ZV capable card is inserted before switching on
> ZV.
>
>
> Fredrik
>
>
>

Vladimir Dergachev wrote:

>
>On Thu, 28 Feb 2002, Fredrik Lingvall wrote:
>
>>I'm trying to get ZV to work on an ATI 3D Rage LT Pro chip.
>>Which register do I use to switch on ZV?
>>
>>I now that
>>
>>outf(CAPTURE_CONFIG, ENABLE_ATI_CAPTURE_DOUBLEBUF | 0x100);
>>
>>should do it on  ATI Rage Mobility M1 chip - see the ZV- patch for
>>atixv.c (in the CVS ati.2
>>sources) at http://www.linuxtv.org. I tried that on the ATI 3D Rage LT
>>Pro chip
>>which resulted in a X-server crash. I guess that I have to use another
>>
>
>A crash or lockup ? It should not segfault because of this statement, if
>it does the error is someplace else.
>
This is the message I get at the X11 log:

>
>
>Fatal server error:
>Caught signal 11.  Server aborting
>
>
>When reporting a problem related to a server crash, please send
>the full server output, not just the last messages.
>This can be found in the log file "/var/log/XFree86.0.log".
>Please report problems to xf...@xf....
>
Do you have any info of the CAPTURE_CONFIG register? Is there any
other register that controls ZV? Ideally, I want to have a function
that checks if a ZV capable card is inserted before switching on
ZV.


Fredrik

On Thu, 28 Feb 2002, Fredrik Lingvall wrote:

> I'm trying to get ZV to work on an ATI 3D Rage LT Pro chip.
> Which register do I use to switch on ZV?
>
> I now that
>
> outf(CAPTURE_CONFIG, ENABLE_ATI_CAPTURE_DOUBLEBUF | 0x100);
>
> should do it on  ATI Rage Mobility M1 chip - see the ZV- patch for
> atixv.c (in the CVS ati.2
> sources) at http://www.linuxtv.org. I tried that on the ATI 3D Rage LT
> Pro chip
> which resulted in a X-server crash. I guess that I have to use another

A crash or lockup ? It should not segfault because of this statement, if
it does the error is someplace else.

                         Vladimir Dergachev

> register than
> CAPTURE_CONFIG or use another value than 0x100.
>
> Fredrik
>
>
> _______________________________________________
> Gatos-devel mailing list
> Gat...@li...
> https://lists.sourceforge.net/lists/listinfo/gatos-devel
>

On 28 Feb 2002, Christian Krause wrote:

> On Sun, 24 Feb 2002 10:01:47 -0500 (EST), Vladimir Dergachev wrote:
>
> >> >> >> Some problems occured only when using 3D (using the newest drm
> >> >> >> kernel-driver from CVS):
> >> >> >>
> >> >> >> - glgears: no problems
> >> >> >> - the most xscreensaver gl savers ( bubble3d, lament, ..): no problems
> >> >> >> - but some of them (molecule, glplanet, pipes, starwars) segfault or
> >> >> >> lock the kernel
> >> >> >>
> >>
> >> I've tested the original XFree 4.2.0 with the drm kernel driver supplied
> >> with this XFree. (version 1.1.1).
> >>
> >> The problems I described gone away. No more segfaults in the above
> >> xscreensaver modules. The graphic is ok, too. (beside pipes, but this
> >> could be a xscreensaver problem...)
>
> > I still think this is something similar. The reason is that I have tested
> > glxgears, QuakeIII and descent -t with these drivers and they work fine.
> > So the error does not seem to be with vertex or texture operations.
>
> After a little debugging I can say, that this segfault (especially in
> glplanet), occurs in READ_RGBA
> (xc/lib/GL/mesa/src/drv/radeon/radeon_span.c):
>
> #define READ_RGBA( rgba, _x, _y )                                       \
>    do {                                                                 \
>       GLushort p;                                                       \
>         p = *(GLushort *)(read_buf + _x*2 + _y*pitch);                  \
>         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
> This assignment cause the segfault.
>
> Does this help? What could be the reason for this problem?

Hmm.. Could you put some "printf" (or whatever they use for printf)
statement in that would print the value of read_buf and then email the
entire X log and the value of read_buf to me.

                         Vladimir Dergachev

>
> regards,
> christian
>
>
> _______________________________________________
> Gatos-devel mailing list
> Gat...@li...
> https://lists.sourceforge.net/lists/listinfo/gatos-devel
>

On 28 Feb 2002 Tim...@t-... wrote:

> What i meant was, how should it work, in terms of what options to set and other
> things to do, apart from connecting the tv during boot time.

You just connect the TV before turning computer on and set X resolution to
800x600 16 bpp. If it does not work, well, then it does not work with your
card at the moment.

                           Vladimir Dergachev

>
> Tim Schroeder
>
>
> On 28 Feb 2002 Tim...@t-... wrote:
>
> > How does it work the ? Are there any option to set ? I tried it a few weeks
> ago
> > and i couldnt get it working on my radeon 7500.
>
> It is tested with plain radeons. Radeon 7500 is a little different..
>
>                      Vladimir Dergachev
>

On Thu, 28 Feb 2002, jammer wrote:

> I've been trying to get Teletext working using various
> software (alevtd/zapping) but have been unable to access
> the /dev/vbi device needed.

No, this is not supported yet. Robert, any comments ?

                      Vladimir Dergachev

>
> Is there any support for this in the ati.2 drivers ?
> If not, has anyone got any ideas about where I might start
> to try and add support for this. (I'm new to Linux but
> I'm a fairly experienced embedded programmer.)
>
> Thanks in advance.
> _____________________________________________________________________
> // free anonymous email || forums \\ subZINE || anonymous browsing
>             subDIMENSION -- http://www.subdimension.com
>
> _______________________________________________
> Gatos-devel mailing list
> Gat...@li...
> https://lists.sourceforge.net/lists/listinfo/gatos-devel
>

I'm trying to get ZV to work on an ATI 3D Rage LT Pro chip.
Which register do I use to switch on ZV?

I now that

outf(CAPTURE_CONFIG, ENABLE_ATI_CAPTURE_DOUBLEBUF | 0x100);

should do it on  ATI Rage Mobility M1 chip - see the ZV- patch for 
atixv.c (in the CVS ati.2
sources) at http://www.linuxtv.org. I tried that on the ATI 3D Rage LT 
Pro chip
which resulted in a X-server crash. I guess that I have to use another 
register than
CAPTURE_CONFIG or use another value than 0x100.

Fredrik

On Sun, 24 Feb 2002 10:01:47 -0500 (EST), Vladimir Dergachev wrote:

>> >> >> Some problems occured only when using 3D (using the newest drm
>> >> >> kernel-driver from CVS):
>> >> >>
>> >> >> - glgears: no problems
>> >> >> - the most xscreensaver gl savers ( bubble3d, lament, ..): no problems
>> >> >> - but some of them (molecule, glplanet, pipes, starwars) segfault or
>> >> >> lock the kernel
>> >> >>
>> 
>> I've tested the original XFree 4.2.0 with the drm kernel driver supplied
>> with this XFree. (version 1.1.1).
>> 
>> The problems I described gone away. No more segfaults in the above
>> xscreensaver modules. The graphic is ok, too. (beside pipes, but this
>> could be a xscreensaver problem...)

> I still think this is something similar. The reason is that I have tested
> glxgears, QuakeIII and descent -t with these drivers and they work fine.
> So the error does not seem to be with vertex or texture operations.

After a little debugging I can say, that this segfault (especially in
glplanet), occurs in READ_RGBA
(xc/lib/GL/mesa/src/drv/radeon/radeon_span.c):

#define READ_RGBA( rgba, _x, _y )                                       \
   do {                                                                 \
      GLushort p;                                                       \
        p = *(GLushort *)(read_buf + _x*2 + _y*pitch);                  \
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

This assignment cause the segfault.

Does this help? What could be the reason for this problem?

regards,
christian

On Wed, Feb 27, 2002 at 11:37:24PM -0500, Michael Leuchtenburg wrote:
> I noticed that a simple starfield program I wrote (teaching myself
> graphics programming) causes XFree86 to use about 30% CPU on my laptop,
> with a mach64, while on my workstation it uses only about 1%, with a
> radeon. I can't imagine that it actually needs that much CPU just to
> write a 640x480 image to the screen, so I want to find out where the CPU
> is getting used in the hopes of fixing it.
The problem here is that "top" is lying to you. While indeed this takes up to
30% on mach64, on radeon it takes perhaps half of it but "top" shows 1%
because the kernel measurement technique is unable to detect it everytime.

The cause is "memcpy". If you want to make it faster, use DRI (which
unfortunately has no DMA functions working yet on mach64, but I promise as
soon at it does I'll rewrite mach64PutImage to make a use of it).

Bye,

Peter Surda (Shurdeek) <shu...@pa...>, ICQ 10236103, +436505122023

--
Disclaimer: This E-mail, because of shabby security on the Internet, in no way
            reflects my thoughts or intentions. It may even not be from me!

I read in the page about options for the Rage Theater chip - but the nifty stuff like HW MC/iDCT isn't available yet due to ATI not releasing info, right?
So what are those RT options good for?

(BTW I'm currently not using the GATOS drivers cuz they messed up my DRI, parsec crashes really hard).

DF

I've been trying to get Teletext working using various 
software (alevtd/zapping) but have been unable to access
the /dev/vbi device needed.

Is there any support for this in the ati.2 drivers ?
If not, has anyone got any ideas about where I might start
to try and add support for this. (I'm new to Linux but 
I'm a fairly experienced embedded programmer.)

Thanks in advance.
_____________________________________________________________________
// free anonymous email || forums \\ subZINE || anonymous browsing 
            subDIMENSION -- http://www.subdimension.com

Spake Vladimir Dergachev:
> 
> 
> On Thu, 28 Feb 2002, Michael Leuchtenburg wrote:
> 
> > Spake Vladimir Dergachev:
> > >
> > > On Wed, 27 Feb 2002, Michael Leuchtenburg wrote:
> > >
> > > > I noticed that a simple starfield program I wrote (teaching myself
> > > > graphics programming) causes XFree86 to use about 30% CPU on my laptop,
> > > > with a mach64, while on my workstation it uses only about 1%, with a
> > > > radeon. I can't imagine that it actually needs that much CPU just to
> > > > write a 640x480 image to the screen, so I want to find out where the CPU
> > > > is getting used in the hopes of fixing it.
> > > >
> > > > However, I've run into a problem: building X with profiling. I can't
> > > > just profile the libs, then X won't start. Thus, I need to build X
> > > > profiling as well, or so I think. X seems to have some stuff in its
> > > > makefiles for building a profiled version, but I can't figure out how to
> > > > trigger it. Running make World CFLAGS=-pg makes it fail to compile
> > > > because it doesn't get the normal flags.
> > >
> > > I have never tried profiling X, but the reason radeon uses less cpu time
> > > might have to do with
> > >    a) radeon is faster. In particular AGP transfers are faster than PCI.
> >
> > 40x faster? And also, now that I run the test again on my workstation,
> > it's not even using 0.1% CPU.
> 
>      a) there is an issue with time counting - if your program takes very
>         little time in the time slice Linux will "forget" it.
>      b) If you are running your mach64 at high resolution you might have
>         bandwidth issues as PCI bus and RAMDAC are competing for bandwidth
>         to video ram. Radeons have quite a lot more bandwidth than mach64.

I currently run at 1280x768 at 16bpp. Seeing as the "Maximum clock" line
says 230.00 MHz (I'm assuming that means the RAMDAC), it should only
take about half of the bandwidth pulling out pixels to draw them
(118MB/sec, running at 63.1hz refresh rate). That's assuming that the
path is only 1byte, too; I suspect it's wider?

Interestingly, it does use a lot less CPU when I'm in a lower resolution
(854x480 at 16bpp). Maybe this is indeed the problem. That would be
annoying. Perhaps once DRI is working it will run faster.
 
> >
> > >    b) most likely you have dri working with radeon and thus image
> > >       transfers use DMA
> >
> > True. However, turning off DRI makes absolutely no difference.
> >
> > I may have figured out how to build X with profiling.. I'm not sure
> > which parts need profiling, though. I get that the symbol mcount is
> > unresolved when it tries to start with my previous compile. Hm. Also
> > RADEONOptions, though, and a few other symbols... Maybe I just need to
> > rebuild X or somesuch.
> >
> > I shall keep poking.
> 
> Did you try looking in xc/cf/.. ?
>
Yes, that's where I found it. There are defines you put in site.def for
individual libs.

-- 
Michael Leuchtenburg - mi...@sl...
1984 is not an instruction manual

What i meant was, how should it work, in terms of what options to set and other
things to do, apart from connecting the tv during boot time.

Tim Schroeder 


On 28 Feb 2002 Tim...@t-... wrote:

> How does it work the ? Are there any option to set ? I tried it a few weeks
ago
> and i couldnt get it working on my radeon 7500.

It is tested with plain radeons. Radeon 7500 is a little different..

                     Vladimir Dergachev

On 28 Feb 2002 Tim...@t-... wrote:

> How does it work the ? Are there any option to set ? I tried it a few weeks ago
> and i couldnt get it working on my radeon 7500.

It is tested with plain radeons. Radeon 7500 is a little different..

                     Vladimir Dergachev

>
> regards Tim
>
> On 28 Feb 2002 Tim...@t-... wrote:
>
> > Hi,
> >
> > I read about TV-out for the r128 in development branch. Is there something
> for
> > radeon also ?
>
> Yes.
>
>                     Vladimir Dergachev
>
> >
> > regard Tim
> >
> > _______________________________________________
> > Gatos-devel mailing list
> > Gat...@li...
> > https://lists.sourceforge.net/lists/listinfo/gatos-devel
> >
>

How does it work the ? Are there any option to set ? I tried it a few weeks ago
and i couldnt get it working on my radeon 7500. 

regards Tim

On 28 Feb 2002 Tim...@t-... wrote:

> Hi,
>
> I read about TV-out for the r128 in development branch. Is there something
for
> radeon also ?

Yes.

                    Vladimir Dergachev

>
> regard Tim
>
> _______________________________________________
> Gatos-devel mailing list
> Gat...@li...
> https://lists.sourceforge.net/lists/listinfo/gatos-devel
>

On Thu, 28 Feb 2002, Michael Leuchtenburg wrote:

> Spake Vladimir Dergachev:
> >
> > On Wed, 27 Feb 2002, Michael Leuchtenburg wrote:
> >
> > > I noticed that a simple starfield program I wrote (teaching myself
> > > graphics programming) causes XFree86 to use about 30% CPU on my laptop,
> > > with a mach64, while on my workstation it uses only about 1%, with a
> > > radeon. I can't imagine that it actually needs that much CPU just to
> > > write a 640x480 image to the screen, so I want to find out where the CPU
> > > is getting used in the hopes of fixing it.
> > >
> > > However, I've run into a problem: building X with profiling. I can't
> > > just profile the libs, then X won't start. Thus, I need to build X
> > > profiling as well, or so I think. X seems to have some stuff in its
> > > makefiles for building a profiled version, but I can't figure out how to
> > > trigger it. Running make World CFLAGS=-pg makes it fail to compile
> > > because it doesn't get the normal flags.
> >
> > I have never tried profiling X, but the reason radeon uses less cpu time
> > might have to do with
> >    a) radeon is faster. In particular AGP transfers are faster than PCI.
>
> 40x faster? And also, now that I run the test again on my workstation,
> it's not even using 0.1% CPU.

     a) there is an issue with time counting - if your program takes very
        little time in the time slice Linux will "forget" it.
     b) If you are running your mach64 at high resolution you might have
        bandwidth issues as PCI bus and RAMDAC are competing for bandwidth
        to video ram. Radeons have quite a lot more bandwidth than mach64.

>
> >    b) most likely you have dri working with radeon and thus image
> >       transfers use DMA
>
> True. However, turning off DRI makes absolutely no difference.
>
> I may have figured out how to build X with profiling.. I'm not sure
> which parts need profiling, though. I get that the symbol mcount is
> unresolved when it tries to start with my previous compile. Hm. Also
> RADEONOptions, though, and a few other symbols... Maybe I just need to
> rebuild X or somesuch.
>
> I shall keep poking.

Did you try looking in xc/cf/.. ?

                         Vladimir Dergachev

>
> --
> Michael Leuchtenburg - mi...@sl...
> 1984 is not an instruction manual
>
> _______________________________________________
> Gatos-devel mailing list
> Gat...@li...
> https://lists.sourceforge.net/lists/listinfo/gatos-devel
>

On 28 Feb 2002 Tim...@t-... wrote:

> Hi,
>
> I read about TV-out for the r128 in development branch. Is there something for
> radeon also ?

Yes.

                    Vladimir Dergachev

>
> regard Tim
>
> _______________________________________________
> Gatos-devel mailing list
> Gat...@li...
> https://lists.sourceforge.net/lists/listinfo/gatos-devel
>

Spake Vladimir Dergachev:
> 
> On Wed, 27 Feb 2002, Michael Leuchtenburg wrote:
> 
> > I noticed that a simple starfield program I wrote (teaching myself
> > graphics programming) causes XFree86 to use about 30% CPU on my laptop,
> > with a mach64, while on my workstation it uses only about 1%, with a
> > radeon. I can't imagine that it actually needs that much CPU just to
> > write a 640x480 image to the screen, so I want to find out where the CPU
> > is getting used in the hopes of fixing it.
> >
> > However, I've run into a problem: building X with profiling. I can't
> > just profile the libs, then X won't start. Thus, I need to build X
> > profiling as well, or so I think. X seems to have some stuff in its
> > makefiles for building a profiled version, but I can't figure out how to
> > trigger it. Running make World CFLAGS=-pg makes it fail to compile
> > because it doesn't get the normal flags.
> 
> I have never tried profiling X, but the reason radeon uses less cpu time
> might have to do with
>    a) radeon is faster. In particular AGP transfers are faster than PCI.

40x faster? And also, now that I run the test again on my workstation,
it's not even using 0.1% CPU.

>    b) most likely you have dri working with radeon and thus image
>       transfers use DMA

True. However, turning off DRI makes absolutely no difference.

I may have figured out how to build X with profiling.. I'm not sure
which parts need profiling, though. I get that the symbol mcount is
unresolved when it tries to start with my previous compile. Hm. Also
RADEONOptions, though, and a few other symbols... Maybe I just need to
rebuild X or somesuch.

I shall keep poking.

-- 
Michael Leuchtenburg - mi...@sl...
1984 is not an instruction manual

Hi,

I read about TV-out for the r128 in development branch. Is there something for
radeon also ?  

regard Tim  

On Wed, 27 Feb 2002, Michael Leuchtenburg wrote:

> I noticed that a simple starfield program I wrote (teaching myself
> graphics programming) causes XFree86 to use about 30% CPU on my laptop,
> with a mach64, while on my workstation it uses only about 1%, with a
> radeon. I can't imagine that it actually needs that much CPU just to
> write a 640x480 image to the screen, so I want to find out where the CPU
> is getting used in the hopes of fixing it.
>
> However, I've run into a problem: building X with profiling. I can't
> just profile the libs, then X won't start. Thus, I need to build X
> profiling as well, or so I think. X seems to have some stuff in its
> makefiles for building a profiled version, but I can't figure out how to
> trigger it. Running make World CFLAGS=-pg makes it fail to compile
> because it doesn't get the normal flags.

I have never tried profiling X, but the reason radeon uses less cpu time
might have to do with
   a) radeon is faster. In particular AGP transfers are faster than PCI.
   b) most likely you have dri working with radeon and thus image
      transfers use DMA

                         Vladimir Dergachev

>
> Anybody?
>
> --
> Michael Leuchtenburg - mi...@sl...
> 1984 is not an instruction manual
>
> _______________________________________________
> Gatos-devel mailing list
> Gat...@li...
> https://lists.sourceforge.net/lists/listinfo/gatos-devel
>

On Thu, 28 Feb 2002, Chris Jensen wrote:

> Hi,
> I've tried using avview and there's very little I can do that won't cause it
> to exit with a message similar to the following
> X Error of failed request:  BadMatch (invalid parameter attributes)
>   Major opcode of failed request:  140 (XVideo)
>   Minor opcode of failed request:  14 ()
>   Serial number of failed request:  2338
>   Current serial number in output stream:  2338
>
> All I've been able to do is select my AIW video card, select XVIDEO for
> encoding (anything else causes an exit) and select a channel table.
> I can use xawtv to look through the channels, but I can't get any audio out
> of it. I read somewhere on the list that you can unmute the volume with
> attributes, but if I try to choose attributes from the interface menu, avview
> exits!
> I've tried avview 0.9.0 and the CVS from 27/02/02 both gave the same results.
> Any help would be appreciated

Are you using 4.2.0 drivers or 4.1.0 ? (This seems as a result of using
4.1.0 drivers).

                   Vladimir Dergachev

>
> --
> Chris Jensen
> ch...@dr...
>
> Public Key: http://drspirograph.com/public_key/
>
> Wait: Did you know that there's a direct correlation between the decline of
> Spirograph and the rise in gang activity?  Think about it.
> - Dr Spirograph (The Simpsons)
>
>
> _______________________________________________
> Gatos-devel mailing list
> Gat...@li...
> https://lists.sourceforge.net/lists/listinfo/gatos-devel
>

errr... try using, implicitly, video_grab_device
( after loading km and starting your favorite Xv capable viewer )

ffmpeg -f video_grab_device -s 640x240 -r 60 -i /dev/video -f\
audio_device -i /dev/dsp -acodec mp2 -ar 44100 -ab 64 -vcodec\
mpeg1video -s 320x240 -r 29.97 -deinterlace -qscale 6 -g 400\
-me log -y test.mpg

jft...@eo... scribbled:

>>>I also tried:
>>>
>>>ffmpeg -f mpegvideo -s 640x240 -r 60 -i /dev/video -f audiodevice -i /dev/dsp -acodec mp2 -ar 44100 -ab 64 -vcodec mpeg1video -s 320x240 -r 29.97 -deinterlace -qscale 6 -g 400 -me log -y test.mpg
>>>
>>>libavcodec: CPU flags: mmx
>>>Segmentation fault
>>>
>>>Any idea what the problem might be?
>>>
>>The problem is that ffmpeg is written with bt84x card in mind. The
>>workaround is to tell ffmpeg that /dev/video0 is a file that contains raw
>>YUV422 data. You might want to supply deinterlace flag as well.
>>
>>
> 
> Grr....I am so close to getting this to work I can almost taste it!
> 
> I told ffmpeg to use /dev/video0 as you suggested.  Here is the command line I used:
> 
> ffmpeg -f mpegvideo -s 640x240 -r 60 -i /dev/video0 -f audiodevice -i /dev/dsp -acodec mp2 -ar 44100 -ab 64 -vcodec mpeg1video -s 320x240 -r 29.97 -deinterlace -qscale 6 -g 400 -me log -y test.mpg
> 
> At this point, nothing happens, but I don't get an error either.  Then I start
> up xawtv.
> 
> I get some CPU activity, and then start seeing a lot of memory getting used,
> and then still more, and then more...about 250 MB worth of physical RAM, then
> it starts eating into swap.
> 
> At no point during this process do I actually see a file named "test.mpg".
> 
> I can CTRL-C out of ffmpeg and instantly all my RAM and Swap are returned to
> their previous state.  I still don't have an output file.
> 
> What am I doing wrong?
> 
> And how do I tell ffmpeg that /dev/video0 is YUV422 data?
> 
> -Michael


-- 
  Issued as is without warranty of any kind; without even the
  implied warranty of merchantability or fitness for a
  particular purpose - GNU General Public License

  the unofficial official GATOS irc at:
  irc.openprojects.net #gatos-ati
	# Ross Coleman
-----