Jump to content

Radeon Xpress 200M

thedopefishlives

By thedopefishlives,
in ATi

 Share
483 posts in this topic

Recommended Posts

dasepsilon
dasepsilon
Posted
Posted

back from holidays, and very good news indeed.

i wish i had more time to tinker with the drivers..

for those with 5955 who do not get "core image",

maybe you need a new ATIRadeon9700.kext

my one wasn't compiled for my 10.4.9 kernel.

(or was it 10.4.8??) so had a missing symbol

proc_pid. i downloaded this one here

 

http://forum.insanelymac.com/index.php?sho...p;hl=_proc_pid#

 

and took the kext from original/..

please copy also the ATI_Info.plist, and rename it.

or maybe you want to use the scripts?

i also did patch the kext with my 5955-id.

dont know if this is needed at all.

did so with sed, and not the script, maybe you try

out for yourselves.

 

i'm out of luck here: the first boot it said "core image

supported", but the overall performance was very poor

(as if i'm using no driver at all), on second boot i now

get that 20sec-freeze-thing from above... i'd like to

try a third and fourth boot but am running out of time..

enzobelmont
enzobelmont
Posted
Posted

pasting results (first system profiler):

 

ATI Radeon XPRESS 200M (RS400) (PCIE):

 Chipset Model:      ATI Radeon XPRESS 200M (RS400) (PCIE)

 Type:       Display

 Bus:       PCI

 VRAM (Total):      64 MB

 Vendor:       ATI (0x1002)

 Device ID:      0x5955

 Revision ID:      0x0000

 Displays:

Display:

 Resolution:       1024 x 768

 Depth:       32-bit Color

 Core Image:      Not Supported

 Main Display:      Yes

 Mirror:       Off

 Online:       Yes

 Quartz Extreme:      Not Supported

 

 

 

here is my dmesg (ati part):

 

VGA: family specific matching fails

Ati: Probing.

Ati: RADio address d0100000

Ati: Setting AAPL,gart-width=0x00000020

Ati: Setting device_type=display

Ati: Setting GARTCacheSize=0x00000010

Ati: Setting F8,size=0x00300000

Ati: F8 written 00300000

Ati: Setting fboffset=0x00020000

Ati: Setting GARTReclaimInterval=0x000000fa

Ati: Setting GARTReclaimTimeout=0x000000f0

Ati: Set model=ATI Radeon XPRESS 200M (RS400) (PCIE)

Ati: Corrected F8 size=00300000 VMEM = 02000000

Ati: Start with VGA

Ati: trace before Radeon.kext

(000000f8)=00300000  (00000100)=d4000000  (00000104)=d6000000  (00000108)=02000000  

(00000148)=1fff1c00  (0000014c)=003f0000  (0000015c)=1fff1c00  (00000170)=00000000  

(00000174)=00000000  (0000017c)=00000000  (00000198)=001d2080  (0000019c)=003d0000  

(000001a0)=00000000  (000001c0)=00ff0000  (000001d8)=00000000  (000001dc)=00000000  

(000001e0)=00000000  (000001e4)=00000000  (0000023c)=1c000000  (0000033c)=00000000  

(0000043c)=00000000  (00000700)=cdcdcdcc  (00000704)=00040000  (00000770)=00000000  

(00000774)=cdcdcdc0  (00000910)=00000000  (00000ab0)=00000000  (00000ab4)=00000000  

(00000e40)=00000140  (000015e0)=cdcdcdcd  (000019e4)=00000000  (00004018)=aa353c9b  

(000042c8)=bdc53a47  

ApplePS2Trackpad: Synaptics TouchPad v5.10

 

 

 

as you can see it says no QE/CI in system profiler;

 

i've moved out callisto from extension directory, and i've not changed nothing in atiradeon9700.kext

 

using 64mb UMA

 

any idea??

 

thanks in advance.

GoldenChaos
GoldenChaos
Posted
Posted

Tried ATISlice.kext without Callisto and with patched koverg ATIRadeon9700.kext. I experienced, on reboot, the "slowness" that other people have mentioned. It took an hour to put Callisto and my old ATIRadeon9700.kext back and get them working. Anything that produced any change in imagery on the screen made the system lag, especially dock animations.

 

I did, however, wait patiently enough for the System Profiler to open up and witnessed that there was still no QE/CI.

 

I don't think I'll be trying that combination again.

Slice
Slice
Posted
Posted
proc_pid. i downloaded this one here

 

http://forum.insanelymac.com/index.php?sho...p;hl=_proc_pid#

i'm out of luck here: the first boot it said "core image

supported", but the overall performance was very poor

(as if i'm using no driver at all), on second boot i now

get that 20sec-freeze-thing from above... i'd like to

try a third and fourth boot but am running out of time..

OK, this link give you patched 9700.kext. Koverg patch makes driver to recognize your card as chip RV380. May be it is not bad if you see CI supported

 

Ati: Setting AAPL,gart-width=0x00000020
-- try to write 0x40
Ati: Corrected F8 size=00300000 VMEM = 02000000

(0000014c)=003f0000

-- it is a main problem. This register non-initialized, so we can't use OpenGL memory larger then 0x003fffff=4Mb
as you can see it says no QE/CI in system profiler;

Didn't you forget to change info.plist of GA.plugin back to IOFramebuffer as IOProvider?

using 64mb UMA

 

any idea??

The main idea to make reg14c initialization in ATISlice.

 

Tried ATISlice.kext without Callisto and with patched koverg ATIRadeon9700.kext. I experienced, on reboot, the "slowness" that other people have mentioned. It took an hour to put Callisto and my old ATIRadeon9700.kext back and get them working. Anything that produced any change in imagery on the screen made the system lag, especially dock animations.

All testers including my own Radeon9000IGP found very slow system or scroll bug. What is it? After some investigation I think that it may be incorrect (14c) settings and may be we need drivers and frameworks (OpenGK and Quartz) from system 10.4.6 or below. As I found at the forum many successful report exists namely with old drivers even in Leo.

Look for example

http://forum.insanelymac.com/index.php?sho...1308&st=132

I don't think I'll be trying that combination again.

And what about new version of the combination? :censored2:

Slice
Slice
Posted
Posted

I propose a new combination to test

ATISlice104 here

ATIRadeon9700 v1.4.20 koverg patched

other drivers and framework from http://forum.insanelymac.com/index.php?showtopic=34409

first of all without Callisto

in GA.plugin.info.plist check

		<key>IOProviderClass</key>
	<string>IONDRVFramebuffer</string>

it may be replaced by CallistoFB if Callisto presents.

Why I hate Callisto? It writes in Radeon chip some values that I can't control.

When you got good graphic without Callisto you can undo it.

ole2
ole2
Posted
Posted

hello Everyone

 

here is mine installation steps (or how I managed to get QE/CI)

 

don't assume you'll have performance speedup, just create a statement inside System Profiler.

 

the work on ATISlice is not over yet and before that there will be lot's of underwater stones.

 

1. disabling callisto:

 

sudo mv /System/Library/Extensions/CallistoHAL.kext /System/Library/Extensions/CallistoHAL.kext_disabled

that's enough and you always can revert it back to live...

 

2. providing ATISlice:

 

sudo cp -R Desktop/ATISlice.kext /System/Library/Extensions/

that's enough, you not need even to fix any permissions, as cp will create new subfolder list in the name of root and group wheel

 

3. fixing PCIID's for ATISlice and ATIRadeon9700sudo vi /System/Library/Extensions/ATISlice.kext/Contents/Info.plist

 

sudo vi /System/Library/Extensions/ATIRadeon9700.kext/Contents/Info.plist

if you don't know VI - try to use nano instead

 

4. get ridding out of cache things

 

sudo rm /System/Library/Extensions.mkext

sudo rm /System/Library/Extensions.kextcache

 

5. now we can reboot and see the result

 

sudo reboot

 

remember, minimal slowdown will be only on UMA 128M mode, on all other modes you'll have to wait for a while on each operation

 

versioning question:

 

processor - AMD 64 3800+ with SSE3

 

ATIRadeon9700:

CFBundleGetInfoString = ATIRadeon9700 1.4.40.4 (3952)

 

ATIRadeon9700GLDriver:

CFBundleGetInfoString = ATIRadeon9700GLDriver 1.4.40.4 (3952)

 

kernel:8.9.1

 

special reply for BUSY people, who desperate to get back Callisto fast after all experiments with ATISlice:

 

1. you need to get Single User mode:

 

- reboot

- type Tab key, when you see "Darvin/x86 boot v5.0.xxx" message at the very begin of boot

- provide "-s" in command line, appeared after your Tab key press, then type Enter key

- wait until boot will finish

- type /sbin/fsck -fy

- type /sbin/mount -uw /

 

2. now you are in Single User mode and it's time to revert Callisto back:

 

- type mv /System/Library/Extentions/CallistoHAL.kext_disabled /System/Library/Extentions/CallistoHAL.kext

- type rm /System/Library/Extentions.mkext

- type rm /System/Library/Extentions.kextcache

- type reboot

 

now, your lovely Callisto is back

GoldenChaos
GoldenChaos
Posted
Posted

Argh, my system crashed and I had to reinstall. It might be a little bit longer (a day tops) before I try out your combination, Slice. I need to get all my stuff restored - Time Machine didn't work to restore anything. I'm a bit upset, but I can redo everything... it's just so time consuming...

 

Thank you ole2 for your installation instructions. Much appreciated.

 

Now back to work I go! Leopard crashes way too often for me... *grumble*

enzobelmont
enzobelmont
Posted
Posted
I propose a new combination to test

ATISlice104 here

ATIRadeon9700 v1.4.20 koverg patched

other drivers and framework from http://forum.insanelymac.com/index.php?showtopic=34409

first of all without Callisto

in GA.plugin.info.plist check

        <key>IOProviderClass</key>
        <string>IONDRVFramebuffer</string>

it may be replaced by CallistoFB if Callisto presents.

Why I hate Callisto? It writes in Radeon chip some values that I can't control.

When you got good graphic without Callisto you can undo it.

 

 

 

white screen.....

 

 

 

 

ahhggg!!!  ;)

 

see you tomorrow (some few hours... 3 AM here right now...  hehehe.... zzzzzzzzz

ole2
ole2
Posted
Posted
white screen.....

ahhggg!!! ;)

 

see you tomorrow (some few hours... 3 AM here right now... hehehe.... zzzzzzzzz

same for me, version 1.0.4 activate white screen at first boot

 

but pay attention, first boot means nothing

reboot again, you'll have a desktop, with other artifact:

 

no moving elements at all (no top menu as well as no lunch panel)

even clicking on finder window doesn't reach it (no feedback on click from finder)

 

however, mouse is moving fine, no hiding, no waiting wheels

treyx86
treyx86
Posted
Posted
Where can I find the "ATIRadeon9700 v1.4.20 koverg patched"

 

Use the ATI Installer from the post Slice made a few back.

 

 

OK, I finally got my test machine set up, installed all the kexts, from a clean slate also.

 

Result - white screen on boot, second boot I am able to see menu bar and mouse moves, but nothing responds to clicks. I waited approximately 20 minutes with no changes.

 

Slice, next step?

Slice
Slice
Posted
Posted
Result - white screen on boot, second boot I am able to see menu bar and mouse moves, but nothing responds to clicks. I waited approximately 20 minutes with no changes.

 

Slice, next step?

Yes, ATISlice104 is wrong.

Use next version http://forum.insanelymac.com/index.php?sho...mp;#entry571348

some recommendation at post #338 of those thread.

GoldenChaos
GoldenChaos
Posted
Posted

When somebody reports something that doesn't kill their system, I'll give that combination a go. It's not that I'm not willing to test stuff, it's just that now that I finally got my system back and more stable than ever before, I really don't want to sacrifice it. Time Machine doesn't restore properly and setting up a new system is such a pain. So, when configurations DON'T bring about white screens or kernel panics en masse, I'll be testing.

 

I realize that results will differ from computer to computer, but I'll feel better if I'm not the first one whose system crashes. :)

dasepsilon
dasepsilon
Posted
Posted

mmh, i tried that new one, but had only some minutes,

and hence no success :blink:

however, slice, i did an mmiotrace on my fglrx--

the values differ slightly from that on the dri-page.

i dont know if it helps at all, still i attach it.

and, keep up the excellent work! :D

mmio.trace.tar.gz

enzobelmont
enzobelmont
Posted
Posted

please slice, can you post more general and accurate instalation instructions??

 

some hints... how to get a patched version of X kexts and where to patch them, etcetera...

 

have been a lot of versions and posts lately... i'm dizzy :(

 

 

BTW, tried slice105... no success.

 

thanks in advance.

 

sorry my poor english.

dvavd
dvavd
Posted
Posted

__1040___1088___1093___1080___1074_.ziphi there!

i've tried ATISlice105 on tiger 10.4.6 JaS with gart-width set to 32 and CallistoHAL.kext removed. the result is

I am able to see menu bar and mouse moves, but nothing responds to clicks.

when setting gart-width to 64, system seems to hang up with unresponsive mouse pointer on blue screen.

also tried combo ATISlice+10.4.6 video kexts & frameworks+no callisto on leopard 10.5.1 and got only kernel panic.

anyway here are ioreg and dmesg from 10.4.6 with gart-width=32

Slice
Slice
Posted
Posted

2 all. Thnx for testing.

mmh, i tried that new one, but had only some minutes,

and hence no success ;)

however, slice, i did an mmiotrace on my fglrx--

the values differ slightly from that on the dri-page.

i dont know if it helps at all, still i attach it.

and, keep up the excellent work! :(

Very thanks for mmio.trace. I think I got new information.

Look, value 14c in fglrx differ from that I trace with ATISlice.kext. It is wrong system initializing - the HackOSX problem. You are not successful because I didn't find good values till now.

 

 

please slice, can you post more general and accurate instalation instructions??

I do not want to repeat FAQ.

1. Drag old kext to trash

2. Drag new kext to /System/Library/Extensions/

3. Repair permissions.

4. Reboot

some hints... how to get a patched version of X kexts and where to patch them, etcetera...

You means corrections into info.plist? I use XCode.

You may type in terminal

nano X.kext/Contents/info.plist

It is very ancient textEditor but it works.

Any another text editor? At your will.

About binary patches... There is no exact instruction. One must know what he do.

sorry my poor english.

As mine.

 

 

i've tried ATISlice105 on tiger 10.4.6 JaS with gart-width set to 32 and CallistoHAL.kext removed. the result is

when setting gart-width to 64, system seems to hang up with unresponsive mouse pointer on blue screen.

Try to increase F8,size value to 0x01000000

also tried combo ATISlice+10.4.6 video kexts & frameworks+no callisto on leopard 10.5.1 and got only kernel panic.

anyway here are ioreg and dmesg from 10.4.6 with gart-width=32

Something is wrong :) Needs to think.

Anyway thanks for report.

dvavd
dvavd
Posted
Posted
Excuse me, system hangs with 32 or 64, or both?

it hangs both ways. but in case of 32 it shows mouse pointer that i can move around. but mouse pointer has no shadow that i expect too see in accelerated mode. menu bar is unresponsive, dock is broken, no icons on desktop. that's it.

in 64 case, system just hangs... or keeps loading but way slowly. mouse pointer shows up then disappears. after couple minutes apple logo with progress bar shows up, then mouse pointer appears again, and so on. i can't get any response from pointer though.

ole2
ole2
Posted
Posted
it hangs both ways. but in case of 32 it shows mouse pointer that i can move around. but mouse pointer has no shadow that i expect too see in accelerated mode. menu bar is unresponsive, dock is broken, no icons on desktop. that's it.

in 64 case, system just hangs... or keeps loading but way slowly. mouse pointer shows up then disappears. after couple minutes apple logo with progress bar shows up, then mouse pointer appears again, and so on. i can't get any response from pointer though.

 

I guess, we could postpone 64bit investigation as certain kext's like Radeon9700 may not be ready for 64bit data-stream at all and that causing side effects including "system hang". Let's stay on 32bit for the current step.

 

no shadow for mouse mean it's in another layer, it also mean, that we didn't programmed/assigned/initialized all layers correctly. IMHO same story for visibility but absence of reaction on menu click as well as lunch-bar - they are drawn in another layer but positioning not match expected by GUI.

dvavd
dvavd
Posted
Posted
I guess, we could postpone 64bit investigation as certain kext's like Radeon9700 may not be ready for 64bit data-stream at all and that causing side effects including "system hang". Let's stay on 32bit for the current step.

i've just tried to increase F8,size value to 0x01000000 for both 32 and 64 bits and got interesting results. when setting gart width to 32, system simply freezes at gray apple logo phase. and when i assigned 64 to the variable, system showed me the same gray logo but with movable mouse pointer that changed its appearance from arrow to spinning rainbow wheel to spinning monochrome wheel (of my applescript sitting inside startup items) to arrow.

ioreg and dmesg outputs for both cases attached.

__1040___1088___1093___1080___1074_.zip

dasepsilon
dasepsilon
Posted
Posted

slice, sorry, i forgot to mention that with my fglrx it outputs

 

[ 63.796027] [fglrx] GART Table is not in FRAME_BUFFER range

[ 63.796040] [fglrx] Reserve Block - 0 offset = 0X0 length = 0X40000

[ 63.796044] [fglrx] Reserve Block - 1 offset = 0X7ff5000 length = 0Xb000

 

so i dont know if the datas are really that valid, though i have

no problems running compiz and DRI is activated indeed.

edit: damn, i am on 64bit linux-- does this change anything??

 

with the values and so on, i didnt get it straight unluckily,

there are different xserver-xorg-video-ati sources, i am in

the dark which driver does what, well, i dont think they have

any news for you, still, pasting the lines doesnt hurt..

 

static void RADEONInitMemoryMap(ScrnInfoPtr pScrn)

{

RADEONInfoPtr info = RADEONPTR(pScrn);

unsigned char *RADEONMMIO = info->MMIO;

CARD32 mem_size;

CARD32 aper_size;

 

/* Default to existing values */

info->mc_fb_location = INREG(RADEON_MC_FB_LOCATION);

info->mc_agp_location = INREG(RADEON_MC_AGP_LOCATION);

 

/* We shouldn't use info->videoRam here which might have been clipped

* but the real video RAM instead

*/

mem_size = INREG(RADEON_CONFIG_MEMSIZE);

aper_size = INREG(RADEON_CONFIG_APER_SIZE);

if (mem_size == 0)

mem_size = 0x800000;

 

/* Fix for RN50, M6, M7 with 8/16/32(??) MBs of VRAM -

Novell bug 204882 + along with lots of ubuntu ones */

if (aper_size > mem_size)

mem_size = aper_size;

 

#ifdef XF86DRI

/* Apply memory map limitation if using an old DRI */

if (info->directRenderingEnabled && !info->newMemoryMap) {

if (aper_size < mem_size)

mem_size = aper_size;

}

#endif

 

{

if (info->IsIGP)

info->mc_fb_location = INREG(RADEON_NB_TOM);

else

#ifdef XF86DRI

/* Old DRI has restrictions on the memory map */

if ( info->directRenderingEnabled &&

info->pKernelDRMVersion->version_minor < 10 )

info->mc_fb_location = (mem_size - 1) & 0xffff0000U;

else

#endif

{

CARD32 aper0_base = INREG(RADEON_CONFIG_APER_0_BASE);

 

/* Recent chips have an "issue" with the memory controller, the

* location must be aligned to the size. We just align it down,

* too bad if we walk over the top of system memory, we don't

* use DMA without a remapped anyway.

* Affected chips are rv280, all r3xx, and all r4xx, but not IGP

*/

if (info->ChipFamily == CHIP_FAMILY_RV280 ||

info->ChipFamily == CHIP_FAMILY_R300 ||

info->ChipFamily == CHIP_FAMILY_R350 ||

info->ChipFamily == CHIP_FAMILY_RV350 ||

info->ChipFamily == CHIP_FAMILY_RV380 ||

info->ChipFamily == CHIP_FAMILY_R420 ||

info->ChipFamily == CHIP_FAMILY_RV410)

aper0_base &= ~(mem_size - 1);

 

info->mc_fb_location = (aper0_base >> 16) |

((aper0_base + mem_size - 1) & 0xffff0000U);

}

}

info->fbLocation = (info->mc_fb_location & 0xffff) << 16;

 

/* Just disable the damn AGP apertures for now, it may be

* re-enabled later by the DRM

*/

info->mc_agp_location = 0xffffffc0;

}

Slice
Slice
Posted
Posted

Wow! You works as professinal programmer!

slice, sorry, i forgot to mention that with my fglrx it outputs

 

[ 63.796027] [fglrx] GART Table is not in FRAME_BUFFER range

Is it warning or error? Does fglrx work?

edit: damn, i am on 64bit linux-- does this change anything??

I think no. The difference will be with memory >4Gb

with the values and so on, i didnt get it straight unluckily,

there are different xserver-xorg-video-ati sources, i am in

the dark which driver does what, well, i dont think they have

any news for you, still, pasting the lines doesnt hurt..

I know this sources. Here is my comments.

/* Default to existing values */

info->mc_fb_location = INREG(RADEON_MC_FB_LOCATION);

//Slice - reg 148

info->mc_agp_location = INREG(RADEON_MC_AGP_LOCATION);

//Slice - reg 14c

/* We shouldn't use info->videoRam here which might have been clipped

* but the real video RAM instead

*/

mem_size = INREG(RADEON_CONFIG_MEMSIZE);

//Slice - reg F8

aper_size = INREG(RADEON_CONFIG_APER_SIZE);

//Slice - reg 108

if (mem_size == 0)

mem_size = 0x800000;

 

/* Fix for RN50, M6, M7 with 8/16/32(??) MBs of VRAM -

Novell bug 204882 + along with lots of ubuntu ones */

if (aper_size > mem_size)

mem_size = aper_size;

//Slice - as I found we can't boot with so large F8 value.

 

if (info->IsIGP)

info->mc_fb_location = INREG(RADEON_NB_TOM);

//Slice - reg 15c

else

#ifdef XF86DRI

/* Old DRI has restrictions on the memory map */

if ( info->directRenderingEnabled &&

info->pKernelDRMVersion->version_minor

info->mc_fb_location = (mem_size - 1) & 0xffff0000U;

//Slice - not for us

CARD32 aper0_base = INREG(RADEON_CONFIG_APER_0_BASE);

//Slice - reg 100

 

info->mc_agp_location = 0xffffffc0;

//Slice - not good value too for shared memory

I dunno I am right or no so I begin the testing with customizable F8 register and in version 105 register 14c calculated on the base of 148 and F8.

What about other values of F8?

0x04000000

0x08000000

 Share
Go to topic listing
×
×
  • Create New...