New AGPGart

Slice · March 18, 2008

Slice: what about merging AGPGart and Natit? Is it a huge work?
In this way you would have better control on registry settings and no overwrite of the same registers.

AGPgart works before creating VGA nub while Natit works after it. It is not problem for me to write new natit as I do with ATILead. But I dunno needed properties. I am collecting all information now. One problem - I don't know nVidia as developer, only as user of PCIe card.

Slice · March 18, 2008

Sorry for Intel users. There was one mistake with AGP_Base.

Test new version.

It is corrected for MrRoper too

Try to change info,plist to

			<key>AGP_Base</key>
		<integer>0xe0000000</integer>

or 0xf0000000

ole2 · March 18, 2008

Sorry for Intel users. There was one mistake with AGP_Base.
Test new version.

It is corrected for MrRoper too

Try to change info,plist to
			<key>AGP_Base</key>
		<integer>0xe0000000</integer>
or 0xf0000000

I think, same value was mentioned someday to ATI x1250 in version delivered for Intel processors (maybe due to Intel specs on AGP_Base)

I would recommend to test this version everybody having this ATI chipset also.

JaS · March 18, 2008

AGPGart 265b2 intel i865 agp bus

Mar 18 15:13:50 localhost kernel[0]: ACPI: Button driver prevents system sleep
Mar 18 15:13:50 localhost kernel[0]: mbinit: done
Mar 18 15:13:50 localhost kernel[0]: Security auditing service present
Mar 18 15:13:50 localhost kernel[0]: BSM auditing present
Mar 18 15:13:50 localhost kernel[0]: AGP: Found Intel 82865 host to AGP bridge
Mar 18 15:13:50 localhost kernel[0]: AGPBridge buses: pri 0 secBus 1 subBus 1 BridgeDev 0
Mar 18 15:13:50 localhost kernel[0]: AGPgart: BridgeMemoryRange [fa900000, fe9fffff]
Mar 18 15:13:50 localhost kernel[0]: AGPgart: BridgePrefetchableMemoryRange [ca800000, ea7fffff]
Mar 18 15:13:50 localhost kernel[0]: AGPgart: BridgeIORange [0000f000, 00000fff]
Mar 18 15:13:50 localhost kernel[0]: AGP: saveBridgeState OK
Mar 18 15:13:50 localhost kernel[0]: AGP create nub for 022110de bus=1 cap=68
Mar 18 15:13:50 localhost kernel[0]: AGPGart: publishNub display
Mar 18 15:13:51 localhost kernel[0]: NVinject: Probing.
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting NVPM=<data not shown>
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting NVCAP=<data not shown>
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting rom-revision=NVinject 0.1.5
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @0,compatible=NVDA,NVMac
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @1,name=NVDA,Display-B
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @2,#adress-cells=<data not shown>
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @2,hwsensor-params-version=<data not shown>
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @0,name=NVDA,Display-A
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @1,compatible=NVDA,NVMac
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @2,#size-cells=<data not shown>
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @2,compatible=NVDA,sensor-parent
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @2,device_type=NVDA,gpu-diode
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @2,reg=<data not shown>
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @1,device_type=display
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @0,device_type=display
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting name=NVDA,Parent
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting device_type=NVDA,GeForce
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @2,hwctrl-params-version=<data not shown>
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting @2,name=sensor-parent
Mar 18 15:13:51 localhost kernel[0]: NVinject: Setting model=GeForce 6200A
Mar 18 15:16:30 localhost kernel[0]: AGP: Coherence support: no
Mar 18 15:16:30 localhost kernel[0]: AGP: GART is 32 bit capable
Mar 18 15:16:30 localhost kernel[0]: AGP: Found an AGP 3.0 compliant device.
Mar 18 15:16:30 localhost kernel[0]: AGPINTEL: aperture [036c9460, 04000000]
Mar 18 15:16:30 localhost kernel[0]: AGPINTEL trace PCI space
Mar 18 15:16:30 localhost kernel[0]: (00)=25708086   (04)=20900106   (08)=06000002   (0c)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (10)=00000008   (14)=00000000   (18)=00000000   (1c)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (20)=00000000   (24)=00000000   (28)=00000000   (2c)=25708086   
Mar 18 15:16:30 localhost kernel[0]: (30)=00000000   (34)=000000e4   (38)=00000000   (3c)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (40)=00000000   (44)=00000000   (48)=00000000   (4c)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (50)=004a0200   (54)=001c8040   (58)=00000000   (5c)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (60)=00000500   (64)=00000000   (68)=00000000   (6c)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (70)=00000000   (74)=00000000   (78)=00000000   (7c)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (80)=00000000   (84)=00000000   (88)=00000000   (8c)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (90)=00111110   (94)=00333300   (98)=00000000   (9c)=00381a00   
Mar 18 15:16:30 localhost kernel[0]: (a0)=00300002   (a4)=1f004a1b   (a8)=00000b00   (ac)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (b0)=00000080   (b4)=00000030   (b8)=03bac000   (bc)=00001010   
Mar 18 15:16:30 localhost kernel[0]: (c0)=00000000   (c4)=240e5000   (c8)=00000000   (cc)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (d0)=0e042802   (d4)=00000d0b   (d8)=00000000   (dc)=41400000   
Mar 18 15:16:30 localhost kernel[0]: (e0)=00000000   (e4)=0106a009   (e8)=00000200   (ec)=00000000   
Mar 18 15:16:30 localhost kernel[0]: (f0)=00000000   (f4)=00000000   (f8)=00030f68   (fc)=00000000   
Mar 18 15:16:30 localhost kernel[0]: AGP: Setting 08 data rate
Mar 18 15:16:30 localhost kernel[0]: AGP: command written target=00000b12 master=1f00e312
Mar 18 15:16:30 localhost kernel[0]: AGP memory 00f27000 length 00001000 offset 00000000 cnt 00000000
Mar 18 15:16:30 localhost kernel[0]: AGP memory 00f27000 length 00001000 offset 00001000 cnt 00000001
Mar 18 15:16:30 localhost kernel[0]: AGP memory 00f27000 length 00001000 offset 00002000 cnt 0000000

Slice in your post you said that agpgart should be set up before any nvinject or natit, yet is seems to load and start setup, then nvinject loads, then agpgart finishes .. could this be an issue, also in the pci trace I see my device id twice , could this be an issue ?

Btw this version of the agpgart acts the same way all other versions have for me over the last two years, so no need to say what happens when I boot.

Keep up the good work Slice, I know you can crack this case and get this agpbus working, could it be because i865 has 2571 and 2570 for video , 2571 is onboard, could this be causing an issue ? maybe agp mem is being treated as shared vram for the onboard gfx ?

I am attaching the pdf for intel's agp 3.0 specification for anyone that wants to read it. I will be doing some reading myself. Just to get a better handle on what is happening here with the driver.

I also noticed that you said gart with was 40 but agpgart is 32 ... in my ioreg without agpgart loaded I have

without agpgart = "NVDA,gart-width" = 64

with agpgart = "NVDA,gart-width" = 32

About this :

kernel[0]: NVChannel(GL): Graphics channel exception! status = 0xffff info32 = 0x6 = Fifo: Parse Error kernel[0]: 0000000b

I have found this:

http://discussions.apple.com/thread.jspa?t...74&tstart=0

I am also going to be reading this

http://developer.apple.com/cgi-bin/search....ault_collection

Tho this was happening to be before the leo gfx update.

agp30.pdf.zip

MrRoper · March 18, 2008

Sorry for Intel users. There was one mistake with AGP_Base.
Test new version.

It is corrected for MrRoper too

excellent, thank you. guess that worked i am unable to boot normally though.

with the -x -v flags i can get to desktop...

I know your a busy man slice... I've ended up in disaster each of my attempts so far, so i'm trying to start again fresh. Can you lend me a hand with getting my 9800 pro working?

Darth Raizzar · March 19, 2008

I loaded the lastest Slice AGPgart in my Hackintosh running iAtkos Leo, and it finally see's the card as AGP in profiler!!!

Now I just need to find a driver for the card that wasn't written for PCIe (ATI HD2400 AGP) :wacko:

GREAT WORK SLICE!! = :angel:

Asus P4SD-LA Mobo (HP Pavillion a562n)

3.0 ghz P4 HT

1 GB DDR

Realtek RTL 8139/810x Network

Realtek AC 97 Audio

Intel 82865G/PE/P2848P/Processor - AGP 2571

Intel 82865G/PE/P2848P/Processor -IO 2570

Slice · March 19, 2008

Thank you JaS for the links!

As I see the same problem with nVidia is on Native Macs! It is Apple driver's mistake!

YAY, all fixed.

Apple service guy came today and replaced my graphics card.

Everything is now working sweetly again.

For reference, the ROM version of my old card was: 3021

and the new card shows: 3022

It is bad ROM? Or Apple's driver works good with new ROM?

May be use previous version of NVDA*.kext?

And again for JaS and MrRoper I didn't see your attempts to change AGP_Base. If you see black screen then aperture address is not good.

As I previously said I can boot into desktop with only one value

VRAM_Address+VRAM_SIZE

but in JaS report I see again (10)=00000008

and ... what is this? AGPINTEL: aperture [036c9460, 04000000] -- my mistake!?

EDIT:

don't understand why

		pciwrite32(hostSpace, APBASE, newApbase | (tmp & 0x0000000f));
}


systemBase			= newApbase;
initial_aperture	= systemLength;
//entries				= initial_aperture >> AGP_PAGE_SHIFT;

if( !systemLength)
{
	IOLog("AGP: ERROR! Could not find aperture size from registers\n");
	return -1;
}
else if(systemLength < 0x02000000 || systemLength > 0x08000000)	 //Slice && -> ||
	IOLog("AGP: WARNING! Please set your aperture to 64 or 128MB in your BIOS for more compatibility\n");
	*address = systemBase;
	*length = systemLength;
	IOLog("AGPINTEL: aperture [%08lx, %08lx]\n", systemBase, systemLength);

So newApbase=0

systemBase = 0

why aperture = 036c9460 ? Leo rumor?

enoch · March 19, 2008

Thank you JaS for the links!
As I see the same problem with nVidia is on Native Macs! It is Apple driver's mistake!

It is bad ROM? Or Apple's driver works good with new ROM?

May be use previous version of NVDA*.kext?

And again for JaS and MrRoper I didn't see your attempts to change AGP_Base. If you see black screen then aperture address is not good.

As I previously said I can boot into desktop with only one value

VRAM_Address+VRAM_SIZE

but in JaS report I see again (10)=00000008

and ... what is this? AGPINTEL: aperture [036c9460, 04000000] -- my mistake!?

EDIT:

don't understand why
		pciwrite32(hostSpace, APBASE, newApbase | (tmp & 0x0000000f));
}


systemBase			= newApbase;
initial_aperture	= systemLength;
//entries				= initial_aperture >> AGP_PAGE_SHIFT;

if( !systemLength)
{
	IOLog("AGP: ERROR! Could not find aperture size from registers\n");
	return -1;
}
else if(systemLength < 0x02000000 || systemLength > 0x08000000)	 //Slice && -> ||
	IOLog("AGP: WARNING! Please set your aperture to 64 or 128MB in your BIOS for more compatibility\n");
	*address = systemBase;
	*length = systemLength;
	IOLog("AGPINTEL: aperture [%08lx, %08lx]\n", systemBase, systemLength);
So newApbase=0

systemBase = 0

why aperture = 036c9460 ? Leo rumor?

so... how much time do you think i have to wait to see agpgart working on my hacked leo ????

Slice · March 19, 2008

so... how much time do you think i have to wait to see agpgart working on my hacked leo ????

Are you propose to live long life?

ole2 · March 19, 2008

so... how much time do you think i have to wait to see agpgart working on my hacked leo ????

hello enoch

don't you worry, all you need is this:

http://forum.insanelymac.com/index.php?showtopic=14169

Slice · March 19, 2008

For Natit users and probably for all who understand me.

That's what I found in Apple's IONDRV driver

	prop = OSDynamicCast( OSData, nub->getProperty("AAPL,vram-memory"));
if (prop && (prop->getLength() >= (2 * sizeof(IOByteCount))))
{
	IOByteCount * lengths;

	lengths = (IOByteCount *) prop->getBytesNoCopy();
	length = lengths[1];
	vramBase = lengths[0];
}

It means AAPL,vram-memory property must be and contains two value

vramBase and length

For example for me

vramBase=0x38000000

length=0x08000000

Using Property List Editor I write the values and got

				<key>AAPL,vram-memory</key>
			<data>
			OAAAAAgAAAA=
			</data>

But for AGP we probably must divide memory on two half, one half for Framebuffer, other for AGP memory.

It is a fresh news. I do not experiments yet.

Any thoughts?

MrRoper · March 19, 2008

And again for JaS and MrRoper I didn't see your attempts to change AGP_Base. If you see black screen then aperture address is not good.

sorry about that, i misunderstood your post. I will try again today with AGPGart.

Regarding graphics driver, at this point i'm not sure exactly which kexts, frameworks, bundles & instructions i need to be using?

should i be following these instructions ?

from post 236

1. remove ATIRadeon9700.kext, ATIRadeon9700GA.plugin

2. install old 10.4.8 ATIRadeon9700.Kext and ATIRadeon9700GA.plugin

3. install callisto, with callisto instructions that come with it

a. follow callisto faq QE/CI for ATIRadeon9700GA.plugin!!!

4. remove ATIRadeon9700GLDriver inside ATIRadeon9700GLDriver.bundle. The path is Contents/MacOS

5. copy ATIRadeonX100GLDriver (version 10.5.1) in the directory where you have delete the ATIRadeon9700GLDriver and rename as ATIRadeon9700GLDriver ( the internal path is the same Contents/MacOS )

6. remove ATINDRV.kext and copy Natit (Natit_ATIauto.kext.zip)

7. set the attribs in new files (meybe it's not important because afer you must repair a permissions)

a. sudo chown -R root:wheel /System/Library/Extensions/file name

b. sudo chmod -R 755 /System/Library/Extensions/file name

8. repair a permissions

9. reboot with -f parameter or delete Extensions.mkext

If there's a better way, kindly point me to it if you would be so kind

Slice · March 19, 2008

sorry about that, i misunderstood your post. I will try again today with AGPGart.

Regarding graphics driver, at this point i'm not sure exactly which kexts, frameworks, bundles & instructions i need to be using?

should i be following these instructions ?

from post 236

1. remove ATIRadeon9700.kext, ATIRadeon9700GA.plugin

2. install old 10.4.8 ATIRadeon9700.Kext and ATIRadeon9700GA.plugin

3. install callisto, with callisto instructions that come with it

a. follow callisto faq QE/CI for ATIRadeon9700GA.plugin!!!

4. remove ATIRadeon9700GLDriver inside ATIRadeon9700GLDriver.bundle. The path is Contents/MacOS

5. copy ATIRadeonX100GLDriver (version 10.5.1) in the directory where you have delete the ATIRadeon9700GLDriver and rename as ATIRadeon9700GLDriver ( the internal path is the same Contents/MacOS )

6. remove ATINDRV.kext and copy Natit (Natit_ATIauto.kext.zip)

7. set the attribs in new files (meybe it's not important because afer you must repair a permissions)

a. sudo chown -R root:wheel /System/Library/Extensions/file name

b. sudo chmod -R 755 /System/Library/Extensions/file name

8. repair a permissions

9. reboot with -f parameter or delete Extensions.mkext

If there's a better way, kindly point me to it if you would be so kind

Very strange propositions and result

For my card the bad thing was that things got very sloppy, lot of tearing (even Mouse locator did not help much, iTunes totally 'teared away'), no cover flow, 90%transparant menu, I didn;t even try to play a movie.

As I know you have Radeon DevID=4e48 that is directly suppoted by ATIRadeon9700*.* so you need no any patch for it. May be use older version but I am not sure.

New reseach.

Does Leo supports AGP? Yes!

		if ((data = OSDynamicCast(OSData, device->getProperty(kIOAGPCommandValueKey))))
		accessFlags |= (kIOAGPFastWrite & (*((UInt32 *) data->getBytesNoCopy())))
					   ? kFramebufferAGPFastWriteAccess : 0;

It is sources of IOGraphics for Leo.

EDITED:

One note

Basically we came to the same conclusion-- the Radeon 9700 GL drivers in Leopard are compiled PPC-only. It would be useful if anyone out there with a copy of an earlier developer build of Leopard could check the following file to see if it's universal or not.

JaS · March 19, 2008

here is what I get with lspci in os x for my gfx card w/o agpgart

01:00.0 VGA compatible controller: nVidia Corporation NV44A [GeForce 6200] (rev a1) (prog-if 00 [VGA controller])
Subsystem: XFX Pine Group Inc. Unknown device 2145
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR+ FastB2B- DisINTx-
Status: Cap+ 66MHz+ UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 248 (1250ns min, 250ns max)
Interrupt: pin A routed to IRQ 16
Region 0: Memory at fd000000 (32-bit, non-prefetchable)
Region 1: Memory at e0000000 (32-bit, prefetchable)
Region 2: Memory at fc000000 (32-bit, non-prefetchable)
Capabilities: [60] Power Management version 2
	Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
	Status: D0 PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [44] AGP version 3.0
	Status: RQ=256 Iso- ArqSz=0 Cal=3 SBA+ ITACoh- GART64- HTrans- 64bit- FW+ AGP3+ Rate=x4,x8
	Command: RQ=1 ArqSz=0 Cal=0 SBA- AGP- GART64- 64bit- FW- Rate=<none>Mar 19 15:11:33

w/ agpgart

01:00.0 VGA compatible controller: nVidia Corporation NV44A [GeForce 6200] (rev a1) (prog-if 00 [VGA controller])
Subsystem: XFX Pine Group Inc. Unknown device 2145
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR+ FastB2B- DisINTx-
Status: Cap+ 66MHz+ UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 248 (1250ns min, 250ns max)
Interrupt: pin A routed to IRQ 16
Region 0: Memory at fd000000 (32-bit, non-prefetchable)
Region 1: Memory at c0000000 (32-bit, prefetchable)
Region 2: Memory at fc000000 (32-bit, non-prefetchable)
Capabilities: [60] Power Management version 2
	Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
	Status: D0 PME-Enable- DSel=0 DScale=0 PME-
Capabilities: [44] AGP version 3.0
	Status: RQ=256 Iso- ArqSz=0 Cal=3 SBA+ ITACoh- GART64- HTrans- 64bit- FW+ AGP3+ Rate=x4,x8
	Command: RQ=32 ArqSz=7 Cal=0 SBA+ AGP+ GART64- 64bit- FW+ Rate=x8

strange that I have my latency set to 32 in the bios ....... think it could be a latancy issue not getting set proper on intel i865 mobos ?

Also Region 1 changes ?

Any idea why agpgart does the pci trace / setup twice ?

localhost kernel[0]: AGP: GART is 32 bit capable
Mar 19 15:11:33 localhost kernel[0]: AGP: Found an AGP 3.0 compliant device.
Mar 19 15:11:33 localhost kernel[0]: AGP: WARNING! Please set your aperture to 64 or 128MB in your BIOS for more compatibility
Mar 19 15:11:33 localhost kernel[0]: AGPINTEL: aperture [02995840, 10000000]
Mar 19 15:11:33 localhost kernel[0]: AGPINTEL trace PCI space
Mar 19 15:11:33 localhost kernel[0]: (00)=25708086   (04)=20900106   (08)=06000002   (0c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (10)=00000008   (14)=00000000   (18)=00000000   (1c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (20)=00000000   (24)=00000000   (28)=00000000   (2c)=25708086   
Mar 19 15:11:33 localhost kernel[0]: (30)=00000000   (34)=000000e4   (38)=00000000   (3c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (40)=00000000   (44)=00000000   (48)=00000000   (4c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (50)=004a0200   (54)=001c8040   (58)=00000000   (5c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (60)=00000500   (64)=00000000   (68)=00000000   (6c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (70)=00000000   (74)=00000000   (78)=00000000   (7c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (80)=00000000   (84)=00000000   (88)=00000000   (8c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (90)=00111110   (94)=00333300   (98)=00000000   (9c)=00381a00   
Mar 19 15:11:33 localhost kernel[0]: (a0)=00300002   (a4)=1f004a1b   (a8)=00000b00   (ac)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (b0)=00000080   (b4)=00000000   (b8)=03f28000   (bc)=00001010   
Mar 19 15:11:33 localhost kernel[0]: (c0)=00000000   (c4)=280e5000   (c8)=00000000   (cc)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (d0)=0e042802   (d4)=00000d0b   (d8)=00000000   (dc)=41400000   
Mar 19 15:11:33 localhost kernel[0]: (e0)=00000000   (e4)=0106a009   (e8)=00000200   (ec)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (f0)=00000000   (f4)=00000000   (f8)=00030f68   (fc)=00000000   
Mar 19 15:11:33 localhost kernel[0]: AGP: Setting 08 data rate
Mar 19 15:11:33 localhost kernel[0]: AGP: command written target=00000b12 master=1f00e312
Mar 19 15:11:33 localhost kernel[0]: AGP: Coherence support: no
Mar 19 15:11:33 localhost kernel[0]: AGP: GART is 32 bit capable
Mar 19 15:11:33 localhost kernel[0]: AGP: Found an AGP 3.0 compliant device.
Mar 19 15:11:33 localhost kernel[0]: AGP: WARNING! Please set your aperture to 64 or 128MB in your BIOS for more compatibility
Mar 19 15:11:33 localhost kernel[0]: AGPINTEL: aperture [40000000, 10000000]
Mar 19 15:11:33 localhost kernel[0]: AGPINTEL trace PCI space
Mar 19 15:11:33 localhost kernel[0]: (00)=25708086   (04)=20900106   (08)=06000002   (0c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (10)=40000008   (14)=00000000   (18)=00000000   (1c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (20)=00000000   (24)=00000000   (28)=00000000   (2c)=25708086   
Mar 19 15:11:33 localhost kernel[0]: (30)=00000000   (34)=000000e4   (38)=00000000   (3c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (40)=00000000   (44)=00000000   (48)=00000000   (4c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (50)=004a0200   (54)=001c8040   (58)=00000000   (5c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (60)=00000500   (64)=00000000   (68)=00000000   (6c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (70)=00000000   (74)=00000000   (78)=00000000   (7c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (80)=00000000   (84)=00000000   (88)=00000000   (8c)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (90)=00111110   (94)=00333300   (98)=00000000   (9c)=00381a00   
Mar 19 15:11:33 localhost kernel[0]: (a0)=00300002   (a4)=1f004a1b   (a8)=00000b12   (ac)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (b0)=00000080   (b4)=00000000   (b8)=041fd000   (bc)=00001010   
Mar 19 15:11:33 localhost kernel[0]: (c0)=00000000   (c4)=280e5000   (c8)=00000000   (cc)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (d0)=0e042802   (d4)=00000d0b   (d8)=00000000   (dc)=41400000   
Mar 19 15:11:33 localhost kernel[0]: (e0)=00000000   (e4)=0106a009   (e8)=00000200   (ec)=00000000   
Mar 19 15:11:33 localhost kernel[0]: (f0)=00000000   (f4)=00000000   (f8)=00030f68   (fc)=00000000   
Mar 19 15:11:33 localhost kernel[0]: AGP: Setting 08 data rate
Mar 19 15:11:33 localhost kernel[0]: AGP: command written target=00000b12 master=

Slice · March 20, 2008

Hi JaS!

1. About latency I don't know. I have the calculations in AGPGart

	// MIN GRANT
mingnt		=	AGPBridge->configRead8( PCIR_MINGNT );
//LATENCY
lattimer	=	AGPBridge->configRead8( PCIR_LATTIMER );
maxlat		=	AGPBridge->configRead8( PCIR_MAXLAT );
.....
IOLog("LATENCY : TIMER = 0x%02x, MAX = 0x%02x\n", lattimer, maxlat);
....
//Slice - apply bridge parameters to master
master->setProperty("devsel-speed", &bridgeDev->devsel, sizeof(bridgeDev->devsel));
if(bridgeDev->fastback)
master->setProperty("fast-back-to-back", &bridgeDev->fastback, sizeof(bridgeDev->fastback));
master->setProperty("max-latency", &bridgeDev->maxlat, sizeof(bridgeDev->maxlat));
master->setProperty("min-grant", &bridgeDev->mingnt, sizeof(bridgeDev->mingnt));

It is my experimenting. No one said me it is good or wrong.

2. Region 1 changed? It is video memory address! The value that I try to catch changed by himself?

I don't know yet where is my mistake:

AGPINTEL: aperture [02995840, 10000000]
(10)=00000008

AGPINTEL: aperture [40000000, 10000000]

(10)=40000008

Method mem = master->getDeviceMemoryWithIndex(i) fails in Leo? Or master is not the same as in Tiger?

3. Why AGPGart::createAGPspace called twice from IOAccelerator clients? May be you have two clients? Or first time calls was unsuccessful so Accelerator made second attempt?

4. I have a trace for real AGP transaction restricted to 3 count. So I see first 3 trasactions of AGP data

AGP: command written target=00000312 master=1f00e312
AGP memory 00600000 length 00001000 offset 00000000 cnt 00000000

AGP memory 00600000 length 00001000 offset 00001000 cnt 00000001

AGP memory 00600000 length 00001000 offset 00002000 cnt 00000002

As you don't see the messages in your dmesg (or system.log) so AGPGart is not switched on. Why?

AGP: WARNING! Please set your aperture to 64 or 128MB in your BIOS for more compatibility

Yes! I know exactly that ATIRadeon8500.kext don't work with AGP if aperture > 0x08000000. Dunno about nVidia.

heng2006 · March 22, 2008

To Slice,

My Mainboard: Asrock 4CoreDual-VSTA (VIA Chipset, P880Ultra)

AGP Card: ATI 9600Pro (Brand Name: PowerColor)

I can install successfully with OSX XxX_OS_x86_10.4.11_Intel.iso

But no CI/QE

I want to have CI/QE

Thank you very much

heng2006

Slice · March 22, 2008

To Slice,

My Mainboard: Asrock 4CoreDual-VSTA (VIA Chipset, P880Ultra)

AGP Card: ATI 9600Pro (Brand Name: PowerColor)

I can install successfully with OSX XxX_OS_x86_10.4.11_Intel.iso

But no CI/QE

I want to have CI/QE

Thank you very much

heng2006

Read topic please

If anyone have trouble with it then report as
ioreg -l -x -w 1024 >ioreg.txt

zip this file and upload here if you want I help you.

heng2006 · March 23, 2008

To Slice,

As your requests.

Thank you very for your help.

heng2006

ioreg.txt

cyclonefr · March 23, 2008

my contribution to the project : my ioreg

slice i think u already have it...let's hope it helps

i did the agp base parameter. Still no go on i865

cybland · March 24, 2008

To Slice

Thanks for your efforts.

Here are some files for your archive. I've been able to achieve CI/QE by using NVinject (256Mb and 512Mb VRAM size settings) and also by manipulating EFI by adding GFX string to my com.apple.Boot.plist and did some tests w/o AGPGart using Cinebench95, OpenMark, Xbench and OpenGL Extensions Viewer. All results with corresponding ioreg.txt are in attached zip. Also supplied is my dmesg.txt when using NVinject from which I saw that NVinject tries and fails to activate acceleration (if I'm interpreting correctly the VGAG lines). I don't think that this is new information for you but anyhow if I'm right about NVinject activating acceleration that would mean that to make AGPGart being responsible for acceleration one should avoid using NVinject. Tell me if you want me to try something for you. My specs are in my sig.

cybland.zip

Slice · March 24, 2008

To Slice,
As your requests.

Thank you very for your help.

heng2006

You have those chipset that I have no experience.

	| |   | |   "IOPCIExpressLinkStatus" = 0x1
| |   | |   "IOPCIExpressLinkCapabilities" = 0x100c41
| |   | |   "class-code" = <00040600>
| |   | |   "compatible" = <"pci1106,a208","pciclass,060400">
| |   | |   "IODTPersist" = <a6e8c235dee8c235>
| |   | |   "#address-cells" = <03000000>
| |   | |   "reg" = <0010000000000000000000000000000000000000>
| |   | |   "ranges" = <0000008200000000000040ff0000008200000000000040ff0000000000001000>
| |   | | }
| |   | | 
| |   | +-o AGPGart  <class AGPGart, registered, matched, active, busy 0, retain count 6>
| |   |	 {

Why PCIExpress?

Please use my IOPCIFamily 2.4.4 and show me dmesg

(sudo dmesg >dmesgHeng.txt)

my contribution to the project : my ioreg
slice i think u already have it...let's hope it helps

i did the agp base parameter. Still no go on i865

Your comment to my post #390?

Thanks for your efforts.
Here are some files for your archive. I've been able to achieve CI/QE by using NVinject (256Mb and 512Mb VRAM size settings) and also by manipulating EFI by adding GFX string to my com.apple.Boot.plist and did some tests w/o AGPGart using Cinebench95, OpenMark, Xbench and OpenGL Extensions Viewer. All results with corresponding ioreg.txt are in attached zip. Also supplied is my dmesg.txt when using NVinject from which I saw that NVinject tries and fails to activate acceleration (if I'm interpreting correctly the VGAG lines). I don't think that this is new information for you but anyhow if I'm right about NVinject activating acceleration that would mean that to make AGPGart being responsible for acceleration one should avoid using NVinject. Tell me if you want me to try something for you. My specs are in my sig.

You have 9 fps in OpenMark - too bad result!

Your AMD64 might good work in Tiger as you can see from Common Sense report. But in Leo we can't achieve success.

NVinject don't activate acceleration. It activate correct name matching to other driver may work. But some other keys raise me questions

NVinject: Setting @1,VRAM,memsize=<data not shown>
NVinject: Setting NVCAP=<data not shown>

memsize might me divided to two half as I see in real MAC with AGP bus.

NVCAP - don't know at all.

cybland · March 24, 2008

Ok Slice,

Tomorrow I’ll reconfigure my rig and install Tubgirl’s Tiger (10.4.10) and Zephyroth’s Leopard (10.5.2). When I finish I’ll make a post here so if you want to test anything on AMD64 just let me know.

On a side note the memsize thing made me try with 512Mb setting (256Mb BIOS aperture + 256Mb VGA VRAM) but I haven’t done anything with it. To be honest I’m not that good with driver customization.

junoir · March 25, 2008

to slice,

running zephyroth's amd 10.5.2

thx for your help

ioreg.txt

JaS · March 25, 2008

Okay tested beta2 ov agpgart 265 on leo 9c7010 and without editing the plist at all I get a scrambled / mixed up screen when I should be at the desktop, the screen has on it the last thing that was on the display before the system was rebooted.

3/25/08 1:46:04 AM kernel NVinject: Probing. 
3/25/08 1:46:04 AM kernel NVinject: Setting NVPM=<data not shown> 
3/25/08 1:46:04 AM kernel NVinject: Setting NVCAP=<data not shown> 
3/25/08 1:46:04 AM kernel NVinject: Setting @0,compatible=NVDA,NVMac 
3/25/08 1:46:04 AM kernel NVinject: Setting @1,name=NVDA,Display-B 
3/25/08 1:46:04 AM kernel NVinject: Setting @2,#adress-cells=<data not shown> 
3/25/08 1:46:04 AM kernel NVinject: Setting @2,hwsensor-params-version=<data not shown> 
3/25/08 1:46:04 AM kernel NVinject: Setting @0,name=NVDA,Display-A 
3/25/08 1:46:04 AM kernel NVinject: Setting @1,compatible=NVDA,NVMac 
3/25/08 1:46:04 AM kernel NVinject: Setting @2,#size-cells=<data not shown> 
3/25/08 1:46:04 AM kernel NVinject: Setting @2,compatible=NVDA,sensor-parent 
3/25/08 1:46:04 AM kernel NVinject: Setting @2,device_type=NVDA,gpu-diode 
3/25/08 1:46:04 AM kernel NVinject: Setting @2,reg=<data not shown> 
3/25/08 1:46:04 AM kernel NVinject: Setting @1,device_type=display 
3/25/08 1:46:04 AM kernel NVinject: Setting @0,device_type=display 
3/25/08 1:46:04 AM kernel NVinject: Setting name=NVDA,Parent 
3/25/08 1:46:04 AM kernel NVinject: Setting device_type=NVDA,GeForce 
3/25/08 1:46:04 AM kernel NVinject: Setting @2,hwctrl-params-version=<data not shown> 
3/25/08 1:46:04 AM kernel NVinject: Setting VRAM,totalsize=<data not shown> 
3/25/08 1:46:04 AM kernel NVinject: Setting @2,name=sensor-parent 
3/25/08 1:46:04 AM kernel NVinject: Setting model=NV44A GeForce 6200A 
3/25/08 1:46:04 AM kernel AGP: Coherence support: no 
3/25/08 1:46:04 AM kernel AGP: GART is 32 bit capable 
3/25/08 1:46:04 AM kernel AGP: Found an AGP 3.0 compliant device. 
3/25/08 1:46:04 AM kernel AGP: VRAM=[d0000000, 10000000]AGPINTEL: aperture [d8000000, 08000000] 
3/25/08 1:46:04 AM kernel AGPINTEL trace PCI space 
3/25/08 1:46:04 AM kernel (00)=25708086   (04)=20900106   (08)=06000002   (0c)=00000000	
3/25/08 1:46:04 AM kernel (10)=d8000008   (14)=00000000   (18)=00000000   (1c)=00000000	
3/25/08 1:46:04 AM kernel (20)=00000000   (24)=00000000   (28)=00000000   (2c)=25708086	
3/25/08 1:46:04 AM kernel (30)=00000000   (34)=000000e4   (38)=00000000   (3c)=00000000	
3/25/08 1:46:04 AM kernel (40)=00000000   (44)=00000000   (48)=00000000   (4c)=00000000	
3/25/08 1:46:04 AM kernel (50)=004a0200   (54)=001c8040   (58)=00000000   (5c)=00000000	
3/25/08 1:46:04 AM kernel (60)=00000500   (64)=00000000   (68)=00000000   (6c)=00000000	
3/25/08 1:46:04 AM kernel (70)=00000000   (74)=00000000   (78)=00000000   (7c)=00000000	
3/25/08 1:46:04 AM kernel (80)=00000000   (84)=00000000   (88)=00000000   (8c)=00000000	
3/25/08 1:46:04 AM kernel (90)=00111110   (94)=00333300   (98)=00000000   (9c)=00381a00	
3/25/08 1:46:05 AM kernel (a0)=00300002   (a4)=1f004a1b   (a8)=00000b00   (ac)=00000000	
3/25/08 1:46:05 AM kernel (b0)=00000080   (b4)=00000020   (b8)=049f1000   (bc)=00001010	
3/25/08 1:46:05 AM kernel (c0)=00000000   (c4)=240e5000   (c8)=00000000   (cc)=00000000	
3/25/08 1:46:05 AM kernel (d0)=0e042802   (d4)=00000d0b   (d8)=00000000   (dc)=41400000	
3/25/08 1:46:05 AM kernel (e0)=00000000   (e4)=0106a009   (e8)=00000200   (ec)=00000000	
3/25/08 1:46:05 AM kernel (f0)=00000000   (f4)=00000000   (f8)=00030f68   (fc)=00000000	
3/25/08 1:46:05 AM kernel AGP: Setting 08 data rate 
3/25/08 1:46:05 AM kernel AGP: command written target=00000b12 master=1f00e312 
3/25/08 1:46:05 AM kernel AGP memory 0a099000 length 00001000 offset 00000000 cnt 00000000 
3/25/08 1:46:05 AM kernel AGP memory 0a099000 length 00001000 offset 00001000 cnt 00000001 
3/25/08 1:46:05 AM kernel AGP memory 0a099000 length 00001000 offset 00002000 cnt 00000002 
3/25/08 1:46:16 AM kernel NVChannel(GL): Graphics channel exception!  status = 0xffff info32 = 0x6 = Fifo: Parse Error 
3/25/08 1:46:16 AM kernel 0000000b

Slice · March 25, 2008

Hi JaS!

Now I see that you fulfil all my requirements.

aperture [d8000000, 08000000]

It is good address (as I think) and good size = 128Mb.

And now you have transactions

3/25/08 1:46:05 AM kernel AGP memory 0a099000 length 00001000 offset 00000000 cnt 00000000 
3/25/08 1:46:05 AM kernel AGP memory 0a099000 length 00001000 offset 00001000 cnt 00000001 
3/25/08 1:46:05 AM kernel AGP memory 0a099000 length 00001000 offset 00002000 cnt 00000002

But I don't understand exactly what do you see on the screen.

Text screen of verbose mode? Or picture of previous boot?

I can suppose that address=d8000000 choosed by default is not good. You'd better edit info.plist to write

AGP_Base

0xe0000000

I make experiments with my ATI chipset and found that I can boot into desktop with only one value (not default). My value = 0x3c000000 cause I have shared memory.

New AGPGart

Recommended Posts

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites