[9fans] Radeon driver

[9fans] Radeon driver

[Venkatesh Srinivas]

Hi,

I've updated Philippe Anel's Radeon driver to work on a current
version of Plan 9. It compiles and should work, but I haven't been
able to test it on real hardware.

http://grex.org/~vsrinivas/radeon/

Take care,
-- vs

Re: [9fans] Radeon driver

[Balwinder S Dheeman]

On 05/30/2009 05:03 AM, Venkatesh Srinivas wrote:

> I've updated Philippe Anel's Radeon driver to work on a current
> version of Plan 9. It compiles and should work, but I haven't been
> able to test it on real hardware.
>
> http://grex.org/~vsrinivas/radeon/

Thanks, I have the following:
vid=0x1002 did=0x4c57	# [LW] Radeon Mobility 7500

on a DELL Inspiron 4150; shall definitely give it a try in a day or two.

Warm Regards,
--
Balwinder S "bdheeman" Dheeman        Registered Linux User: #229709
Anu'z Linux@HOME (Unix Shoppe)        Machines: #168573, 170593, 259192
Chandigarh, UT, 160062, India         Plan9, T2, Arch/Debian/FreeBSD/XP
Home: http://werc.homelinux.net/      Visit: http://counter.li.org/

Re: [9fans] Radeon driver

[Balwinder S Dheeman]

On 05/30/2009 05:03 AM, Venkatesh Srinivas wrote:
> I've updated Philippe Anel's Radeon driver to work on a current
> version of Plan 9. It compiles and should work, but I haven't been
> able to test it on real hardware.
>
> http://grex.org/~vsrinivas/radeon/

I, after following the steps in the radeon_howto, have compiled the
/sys/src/cmd/aux/vga utility and a 386 pcf kernel in /sys/src/9/pc/
successfully and booted off of newer kernel, which seems to be working fine.

Now, how do I test, benchmark and, or profile this radeon driver against
a generic one?

OTOH, I have edited the radeon_howto a bit, to make it somewhat handy
for the users of 'acme' in particular; one can simply right click on the
quoted file names and, or paths to edit and, or open the same in 'acme'.
Here at http://werc.homelinux.net/hacks/plan9/radeon_howto is modified
version of the same for your kind reference and a bundle of other
requisite files at http://werc.homelinux.net/hacks/plan9/radeon.tar.gz

--
Balwinder S "bdheeman" Dheeman        Registered Linux User: #229709
Anu'z Linux@HOME (Unix Shoppe)        Machines: #168573, 170593, 259192
Chandigarh, UT, 160062, India         Plan9, T2, Arch/Debian/FreeBSD/XP
Home: http://werc.homelinux.net/      Visit: http://counter.li.org/

Re: [9fans] Radeon driver

[Venkatesh Srinivas]

> Now, how do I test, benchmark and, or profile this radeon driver against
> a generic one?

IIRC, there was a port of some of xscreensaver's hacks to Plan 9. The
munch hack and hyperglenda should get a good bit faster. They might
have something to measure resulting fps-es. Also, watch the CPU load
monitor while they're running.

For even more speed, define RADEON_HW_ACCEL to use the blitter's
bltblt and fillrect features (inside vgaradeon.c).

Also, unlike the VESA driver, the Radeon driver doesn't use real mode,
so it should work on SMP systems.

Does anyone know why the Radeon driver never got into Plan 9? Did no
one submit it as a patch?

Thanks,
-- vs

Re: [9fans] Radeon driver

[erik quanstrom]

On Mon Jun  1 15:16:00 EDT 2009, me@acm.jhu.edu wrote:
> > Now, how do I test, benchmark and, or profile this radeon driver against
> > a generic one?
>
> IIRC, there was a port of some of xscreensaver's hacks to Plan 9. The
> munch hack and hyperglenda should get a good bit faster. They might
> have something to measure resulting fps-es. Also, watch the CPU load
> monitor while they're running.
>
> For even more speed, define RADEON_HW_ACCEL to use the blitter's
> bltblt and fillrect features (inside vgaradeon.c).
>
> Also, unlike the VESA driver, the Radeon driver doesn't use real mode,
> so it should work on SMP systems.
>
> Does anyone know why the Radeon driver never got into Plan 9? Did no
> one submit it as a patch?

is any of this code on the right track for modern integrated ati
graphics, for example with the sb[67]xx chipsets?

- erik

Re: [9fans] Radeon driver

[Venkatesh Srinivas]

As I understand, the sb600 series has the RS480 IGP, which is a
stripped-down r300 (missing some TnL h/w, vertex shaders). The 2D
parts of the Linux radeonfb look pretty similar between the r300 and
the rs480, so I'd go with yes.

No idea about anything newer... I can take a look if people are interested...

-- vs

Re: [9fans] Radeon driver

[Ori Bernstein]

On Mon, 1 Jun 2009 21:18:34 -0400
Venkatesh Srinivas <me@acm.jhu.edu> wrote:

> As I understand, the sb600 series has the RS480 IGP, which is a
> stripped-down r300 (missing some TnL h/w, vertex shaders). The 2D
> parts of the Linux radeonfb look pretty similar between the r300 and
> the rs480, so I'd go with yes.
>
> No idea about anything newer... I can take a look if people are interested...

The new stuff has full specs out (http://x.org/docs/AMD/), but as I
understand it, it essentially has no 2d engine at all. To get anything
going, you essentially have to initialize and use the 3d engine.

--
    Ori Bernstein

Re: [9fans] Radeon driver

[erik quanstrom]

On Mon Jun  1 21:20:38 EDT 2009, me@acm.jhu.edu wrote:
> As I understand, the sb600 series has the RS480 IGP, which is a
> stripped-down r300 (missing some TnL h/w, vertex shaders). The 2D
> parts of the Linux radeonfb look pretty similar between the r300 and
> the rs480, so I'd go with yes.
>
> No idea about anything newer... I can take a look if people are interested...

there needs to be some adjustment of how to get frequencies out
of rom:

radeon: Found sb600 guesswork
radeon: did:791e rid:00
radeon: mmio address: 0xdefef000 [size=0x10000]
radeon: frame buffer size=134217728 [128MB]
radeon: display type: CRT
radeon: bios=0x000c0000 offset=0x198
radeon: pll_info_block: 0x0
radeon: reference_freq: 0
radeon: reference_div:  0
radeon: min_pll_freq:   0
radeon: max_pll_freq:   29097984
radeon: xclk:           0
radeon: PPLL_CNTL=0x00000000
radeon: monitor type = '213t'
radeon: size = '1600x1200x32'
radeon: chan = ''
radeon: freq=150000000 deffreq=0 x=1600 y=1200 z=32
radeon: ht=2058 shb=1664 ehb=1856 shs=1664 ehs=1856 hsync='+'
radeon: vt=1250 vrs=1201 vre=1204 vsync='+'
mode->z = 32 (format = 6, bpp = 32, dac6bit = false)
hsync_start=1661 hsync_wid=24 hsync_fudge=5
radeon: initpll: freq=15000
8.out 80: suicide: sys: trap: divide error pc=0x0001af17

- erik

Re: [9fans] Radeon driver

[erik quanstrom]

On Mon Jun  1 21:20:38 EDT 2009, me@acm.jhu.edu wrote:
> As I understand, the sb600 series has the RS480 IGP, which is a
> stripped-down r300 (missing some TnL h/w, vertex shaders). The 2D
> parts of the Linux radeonfb look pretty similar between the r300 and
> the rs480, so I'd go with yes.
>
> No idea about anything newer... I can take a look if people are interested...

i've found a couple of issues when using the radeon driver with
current sb600-based ati controllers, and fixed them.  unfortunately
things are still wrong and there are just too many settings to guess.
worse, i don't know of any way i can tell if changes have made
things better or worse.

this is the output i currently get from aux/vga.  does anyone
have similar output from a working setup?

%  /sys/src/cmd/aux/vga/8.out -l $vgasize
radeon: Found sb600 guesswork
radeon: did:791e rid:00
radeon: mmio address: 0xdefef000 [size=0x10000]
radeon: frame buffer size=134217728 [128MB]
radeon: display type: CRT
radeon: atom bios
radeon: bios=0x000c0000 offset=0x198
radeon: pll_info_block: 0xaa58
radeon: reference_freq: 1432
radeon: reference_div:  12
radeon: min_pll_freq:   17235968
radeon: max_pll_freq:   120000
radeon: xclk:           40000
radeon: PPLL_CNTL 00000000
radeon: monitor type = 'XXXt'
radeon: size = '1600x1200x32'
radeon: chan = ''
radeon: freq=150000000 deffreq=0 x=1600 y=1200 z=32
radeon: ht=2160 shb=1664 ehb=1856 shs=1664 ehs=1856 hsync='+'
radeon: vt=1250 vrs=1201 vre=1204 vsync='+'
mode->z = 32 (format = 6, bpp = 32, dac6bit = false)
hsync_start=1661 hsync_wid=24 hsync_fudge=5
radeon: initpll: freq=15000
dc=1436330, of=17235960, fd=144435, pd=0
r300_workaround
Wrote: 0x0000000c 0x00023433 0x00000000 (0x00000000)
Wrote: rd=12, fd=1075, pd=2

- erik

[9fans] Radeon driver

[Alexander Sychev]

Hi, 9fans!

I have installed Plan9 on my IBM ThinkPad T40 and it works fine with vesa  
and wired networking.
But I can't compile Philippe Anel's radeon driver.

<----------------------------------------------->
term% mk 'CONF=pcf'
8c -FVw vgaradeon.c
vgaradeon.c:274 not a member of struct/union: aperture
vgaradeon.c:276 not a member of struct/union: isupamem
vgaradeon.c:290 not a member of struct/union: isupamem
vgaradeon.c:292 function args not checked: upamalloc
vgaradeon.c:294 function args not checked: upamalloc
vgaradeon.c:296 not a member of struct/union: isupamem
vgaradeon.c:299 not a member of struct/union: isupamem
vgaradeon.c:302 not a member of struct/union: isupamem
warning: vgaradeon.c:265 used and not set: wasupamem oaperture
vgaradeon.c:321 function args not checked: upamalloc
vgaradeon.c:332 not a member of struct/union: aperture
vgaradeon.c:446 not a member of struct/union: aperture
warning: vgaradeon.c:445 set and not used: storage
vgaradeon.c:753 initialization of incompatible pointers: vgaradeondev
IND FUNC(IND STRUCT VGAscr, INT, INT) VOID and IND FUNC(IND STRUCT VGAscr,  
IND INT, IND INT) ULONG
mk: 8c -FVw vgaradeon.c  : exit status=rc 234: 8c 236: error
<----------------------------------------------->

It seems struct VGAscr (screen.h) was changed a lot.

Does anybody maintain this driver?

-- 
Best regards,
   santucco

Re: [9fans] Radeon driver

[Philippe Anel]

Hello,

	I used to work on this driver, but these last days/months(years?), I
hadn't enough time to work on it. However, I
think I can help you here. I'll send you my latest vgaradeon.c tomorrow.
I hope the current aux/vga part is working with your card.

	Regards,
	Phil.

On Tue, 2006-10-17 at 17:44 +0400, Alexander Sychev wrote:
> Hi, 9fans!
> 
> I have installed Plan9 on my IBM ThinkPad T40 and it works fine with vesa  
> and wired networking.
> But I can't compile Philippe Anel's radeon driver.
> 
> <----------------------------------------------->
> term% mk 'CONF=pcf'
> 8c -FVw vgaradeon.c
> vgaradeon.c:274 not a member of struct/union: aperture
> vgaradeon.c:276 not a member of struct/union: isupamem
> vgaradeon.c:290 not a member of struct/union: isupamem
> vgaradeon.c:292 function args not checked: upamalloc
> vgaradeon.c:294 function args not checked: upamalloc
> vgaradeon.c:296 not a member of struct/union: isupamem
> vgaradeon.c:299 not a member of struct/union: isupamem
> vgaradeon.c:302 not a member of struct/union: isupamem
> warning: vgaradeon.c:265 used and not set: wasupamem oaperture
> vgaradeon.c:321 function args not checked: upamalloc
> vgaradeon.c:332 not a member of struct/union: aperture
> vgaradeon.c:446 not a member of struct/union: aperture
> warning: vgaradeon.c:445 set and not used: storage
> vgaradeon.c:753 initialization of incompatible pointers: vgaradeondev
> IND FUNC(IND STRUCT VGAscr, INT, INT) VOID and IND FUNC(IND STRUCT VGAscr,  
> IND INT, IND INT) ULONG
> mk: 8c -FVw vgaradeon.c  : exit status=rc 234: 8c 236: error
> <----------------------------------------------->
> 
> It seems struct VGAscr (screen.h) was changed a lot.
> 
> Does anybody maintain this driver?
> 

Re: [9fans] Radeon driver

[Alexander Sychev]

Hi, 9fans!

On Wed, 18 Oct 2006 10:06:02 +0400, Philippe Anel <xigh@free.fr> wrote:

> Hello,
>
> 	I used to work on this driver, but these last days/months(years?), I
> hadn't enough time to work on it. However, I
> think I can help you here. I'll send you my latest vgaradeon.c tomorrow.

Ok. I'll be waiting for it. Thanks!

> I hope the current aux/vga part is working with your card.

It seems, radeon.c in aux/vga can't be compiled too:
<----------------------------------------------->
8c -FVw radeon.c
radeon.c:393 incompatible types: "ULONG" and "IND VOID" for op "AS"
warning: radeon.c:357 used and not set: mmio
mk: 8c -FVw radeon.c  : exit status=rc 857: 8c 859: error
<----------------------------------------------->

>
> 	Regards,
> 	Phil.
>

-- 
Best regards,
   santucco

Re: [9fans] Radeon driver

[Alexander Sychev]

I think, it can be fixed by changing type of mmio from "ulong" to "ulong*".
To tell the truth, I don't quite understand how it could be compiled early.

>
> It seems, radeon.c in aux/vga can't be compiled too:
> <----------------------------------------------->
> 8c -FVw radeon.c
> radeon.c:393 incompatible types: "ULONG" and "IND VOID" for op "AS"
> warning: radeon.c:357 used and not set: mmio
> mk: 8c -FVw radeon.c  : exit status=rc 857: 8c 859: error
> <----------------------------------------------->


-- 
Best regards,
   santucco

Re: [9fans] Radeon driver

[Anel Philippe]

Hello,

You can download the following required files :

    http://xigh.free.fr/radeon.c          ->  /sys/src/cmd/aux/vga
    http://xigh.free.fr/radeon.h          ->  /sys/src/cmd/aux/vga

    http://xigh.free.fr/vgaradeon.c    ->  /sys/src/9/pc

    http://xigh.free.fr/vgadb             ->   merge with /lib/vgadb

Please note I couldn't test if this version fully works. Indeed the only
Radeon chipset I have (for the moment) is in my laptop (XPress 200M)
and I know for sure it doesn't work (I planned to work on the driver
but for lack of time, I couldn't). So ... pay carreful attention with these
files (and your computer).

    Regards,
    Phil;

Ps: I would suggest you to buy Intel/Nvidia's chipsets whenever you
    can. Indeed newer Ati's chipsets are not documented at all (I don't
    have docs and reading x11 drivers is ... a real drag ...) and only
    commercial drivers work (even for 2D only stuff) ... not to mention
    the fact Nvidia's chipsets seem to be much better designed than
    Ati one's. I mean that regarding api ... in order to add support for
    a newer Nvidia's chipset ... it seems you only have to edit /lib/vgadb.
    This not true with Ati.

Re: [9fans] Radeon driver

[Wes Kussmaul]

Anel Philippe wrote:
> Indeed newer Ati's chipsets are not documented at all (I don't
>    have docs and reading x11 drivers is ... a real drag ...) and only
>    commercial drivers work (even for 2D only stuff) ...
I suspect that may change now that AMD is buying ATI.

Re: [9fans] Radeon driver

[Sascha Retzki]

> Anel Philippe wrote:
>> Indeed newer Ati's chipsets are not documented at all (I don't
>>    have docs and reading x11 drivers is ... a real drag ...) and only
>>    commercial drivers work (even for 2D only stuff) ...
> I suspect that may change now that AMD is buying ATI.

Optimist :-)

Re: [9fans] Radeon driver

[Wes Kussmaul]

[-- Attachment #1: Type: text/html, Size: 988 bytes --]

Re: [9fans] Radeon driver

[erik quanstrom]

[-- Attachment #1: Type: text/plain, Size: 1629 bytes --]

the grass is always greener!  having spent quite a bit of time trying
to get nvidia's nforce 430 chipset with integrated graphics working,
so far with no success, i'm pretty sure that nvidia's documentation
is similar to ati's -- read the xfree86 driver.  unfortunately in the case
of the nvidia driver, there are several real bugs in the xf86 driver (like
stuffing a 0x480 into an 8-bit register with one bit of overflow and
stuffing 0x100 into a different 8-bit register with no overflow.)

by the way, i am having some trouble with the virge driver that seems
to have cropped up lately

; history vga
Sep 18 15:21:47 EDT 2006 vga 336305 [quanstro]
Sep 18 15:21:47 EDT 2006 /n/dump/2006/1019/386/bin/aux/vga 336305 [quanstro]
Feb 27 23:08:17 EST 2005 /n/dump/2006/0918/386/bin/aux/vga 301643 [quanstro]

the 1019 aux/vga gives the error "virge: pclk 79000000 out of range".
the vid/did are 5333/5631.  i'll attach aux/vga -p.

i scanned the source code changes but nothing jumped out at me.

- erik

On Thu Oct 19 04:30:59 EDT 2006, xigh@free.fr wrote:
> Ps: I would suggest you to buy Intel/Nvidia's chipsets whenever you
>     can. Indeed newer Ati's chipsets are not documented at all (I don't
>     have docs and reading x11 drivers is ... a real drag ...) and only
>     commercial drivers work (even for 2D only stuff) ... not to mention
>     the fact Nvidia's chipsets seem to be much better designed than
>     Ati one's. I mean that regarding api ... in order to add support for
>     a newer Nvidia's chipset ... it seems you only have to edit /lib/vgadb.
>     This not true with Ati.

[-- Attachment #2: vgap --]
[-- Type: text/plain, Size: 1980 bytes --]

vga->attr: vid=0x5333
vga->attr: did=0x5631
vga->attr: 0xC0043=Stealth 3D 2000 Vers. 1.04
vga misc             EF
vga feature          00
vga sequencer        03 01 0F 00 0A
vga crt              A3 7F 88 9A 86 1A 24 FD - 00 60 00 00 00 00 FF 00
                     03 29 FF 80 60 03 04 A3 - FF
vga graphics         00 00 00 00 00 50 05 0F - FF
vga attribute        00 01 02 03 04 05 06 07 - 08 09 0A 0B 0C 0D 0E 0F
                     41 FF 0F 00 00
vga virtual         0 0
vga panning         off
vga clock[1] f      135000000
vga clock[1] d i m          0        0       -      127
vga clock[1] n p q r       31        0       -        0        3
vga vmz               2097152
vga apz              67108864
vga linear                  1
virge Crt30          E1 8D 40 20 10 00 92 FF - 7B BD 15 95 00 00
virge Crt40          01 04 04 00 00 01 01 0E - 0A 84 00 FF F3 00 00 00
virge Crt50          00 00 90 08 A8 00 00 00 - 9A EC 00 00 43 00 40 40
virge Crt60          0F 0F 0F 00 00 00 89 00 - FE E0 C0 08 00 7D CF 1A
virge hde           400  1024
virge shb           448  1096
virge ehb           45A  1114
virge ht            540  1344
virge vde           300   768
virge vrs           303   771
virge vre           309   777
virge vt            326   806
virge Crt70          06 06 06 06 06 06 06 06 - 06 06 06 06 06 06 06 06
                     A3 7F 88 9A 86 1A 24 FD - 00 60 00 00 00 00 FF 00
                     03 29 FF 80 60 03 04 A3 - FF
virge Seq08          06 00 80 00 0C 03 0E 0F
virge Seq10          41 2C 29 77 00 02 02 9A - 40 00 41 1B 00 1D 1E 1F
                     41 2C 29 77 00 02 02 9A - 40 00 41 2B 00 2D 2E 2F
                     41 2C 29 77 00 02 02 9A - 40 00 41 3B 00 3D 3E 3F
                     03 01 0F 00 0A 05 06 07 - 06 00 80 00 0C 03 0E 00
virge Crt2D          56 31 06
virge dclk m n r     78749990      119       -        9        1
virge mclk m n r     54886356       44       -        1        2

Re: [9fans] Radeon driver

[Alexander Sychev]

Hi!

That is pity, but it doesn't work :-(.
I have got a "chatoyant" screen with a disappearing kernel output.
The system doesn't react to any commands and I can only poweroff the  
laptop.
I'm going to try to get some logs from aux/vga.

On Thu, 19 Oct 2006 12:30:04 +0400, Anel Philippe <xigh@free.fr> wrote:

>
> Hello,
>
> You can download the following required files :
>
>     http://xigh.free.fr/radeon.c          ->  /sys/src/cmd/aux/vga
>     http://xigh.free.fr/radeon.h          ->  /sys/src/cmd/aux/vga
>
>     http://xigh.free.fr/vgaradeon.c    ->  /sys/src/9/pc
>
>     http://xigh.free.fr/vgadb             ->   merge with /lib/vgadb
>
> Please note I couldn't test if this version fully works. Indeed the only
> Radeon chipset I have (for the moment) is in my laptop (XPress 200M)
> and I know for sure it doesn't work (I planned to work on the driver
> but for lack of time, I couldn't). So ... pay carreful attention with  
> these
> files (and your computer).
>
>     Regards,
>     Phil;
>
> Ps: I would suggest you to buy Intel/Nvidia's chipsets whenever you
>     can. Indeed newer Ati's chipsets are not documented at all (I don't
>     have docs and reading x11 drivers is ... a real drag ...) and only
>     commercial drivers work (even for 2D only stuff) ... not to mention
>     the fact Nvidia's chipsets seem to be much better designed than
>     Ati one's. I mean that regarding api ... in order to add support for
>     a newer Nvidia's chipset ... it seems you only have to edit  
> /lib/vgadb.
>     This not true with Ati.
>
>



-- 
Best regards,
   santucco