Sonnet 7200 PCI programming

[kgc210]

Is that purely running the PPC program in VGA memory or in combination with the PPC Memory?

Would it be possible to use VGA memory as secondary / page file setup. Or is that beyond the drivers ability / hardware possible?

As a side thought I wonder if it would be easier to add larger RAM chips on the K1 / M1 as long as there are enough address lines / a way to update the firmware?

Originally Posted by Hedeon

Radeon cannot be in 1x256mb mode. And that's all I can say, as the powerpc.library is not active yet, it's a pci.library problem.

Yes, definitely pci.library problem.


Tested with Voodoo3 and the same thing happens with it. Also pciinfo hangs machine as Wrangler told.

[Hedeon]

Originally Posted by kgc210

Is that purely running the PPC program in VGA memory or in combination with the PPC Memory?

Would it be possible to use VGA memory as secondary / page file setup. Or is that beyond the drivers ability / hardware possible?

As a side thought I wonder if it would be easier to add larger RAM chips on the K1 / M1 as long as there are enough address lines / a way to update the firmware?

Not sure if anyone has had success by adding memory.

And sure, a page file is possible. The cost versus benefit I'm not really sure of.

[Hedeon]

Originally Posted by jvaltane

Yes, definitely pci.library problem.


Tested with Voodoo3 and the same thing happens with it. Also pciinfo hangs machine as Wrangler told.

Is there ANYTHING in the $20000000-$40000000 range like some expansion or something?

[grelbfarlk]

Originally Posted by kgc210

Is that purely running the PPC program in VGA memory or in combination with the PPC Memory?

Would it be possible to use VGA memory as secondary / page file setup. Or is that beyond the drivers ability / hardware possible?

As a side thought I wonder if it would be easier to add larger RAM chips on the K1 / M1 as long as there are enough address lines / a way to update the firmware?

There seem to be enough address lines available and I say this comparing the chips on a K1 versus the next size up in RAM chips. I did try replacing them thinking at the least it would work but only set up the current 64MB but it didn't work. But this was probably because I didn't solder them properly and eventually popcorned the space between the chips.


The other idea for not updating the firmware is that there is a way to put the board into a hardware defined default mode which would ignore the firmware on the board and it could be set up via software somewhat like how the Sonnet and other PPC boards function. This would take a non-trivial amount of work on Hedeon's part.

[nexus]

actually i dont believe he can without taking the firmware off the card as he cannot do a HARD reset
so you either have to take it off or disable manually or flash with new
he cant do a power on reset with that firmware on at boot after it has been loaded you cant do another thats the only way to config those settings is at power on

CFG_RESET_SOURCE, CFG_CLKIN_DIV
is only set during PORESET
Although the configuration reset words are loaded during hard reset flows, the clocks and PLL
modes are reset only when PORESET is asserted, during a power-on reset flow

Default Reset Configuration Words
If the device is configured not to load the reset configuration words from a local bus EEPROM or I 2 C
EEPROM, it can also be initialized with one of five hard-coded default options, selected by the reset
configuration input signals, CFG_RESET_SOURCE[0:2]. In this mode, the device is assumed to be a PCI
agent, and therefore only clock modes differ among the four options

so you have to have it on local bus in eprom or set by jumpers
those hardware bits are trump over the firmware

[Hedeon]

Originally Posted by nexus

actually i dont believe he can without taking the firmware off the card as he cannot do a HARD reset
so you either have to take it off or disable manually or flash with new
he cant do a power on reset with that firmware on at boot after it has been loaded you cant do another thats the only way to config those settings is at power on

CFG_RESET_SOURCE, CFG_CLKIN_DIV
is only set during PORESET
Although the configuration reset words are loaded during hard reset flows, the clocks and PLL
modes are reset only when PORESET is asserted, during a power-on reset flow

Default Reset Configuration Words
If the device is configured not to load the reset configuration words from a local bus EEPROM or I 2 C
EEPROM, it can also be initialized with one of five hard-coded default options, selected by the reset
configuration input signals, CFG_RESET_SOURCE[0:2]. In this mode, the device is assumed to be a PCI
agent, and therefore only clock modes differ among the four options

so you have to have it on local bus in eprom or set by jumpers
those hardware bits are trump over the firmware

He was talking about setting RAM size. What you are talking about (the clocks) can only be set once, that's true. For everything else I can do a soft reset.

[indigolemon]

Originally Posted by Hedeon

@All: I might have broken Harrier support (not yet tested) and the unofficial A1200 stuff. Please check.

397 installed and working exactly the same as previously on my A1200 with the K1. Nothing broken so far (based on what previously worked that is!) I shall play around a bit more

[edit] One initial thing I've noticed, the system is a LOT more stable after the powerpc.library is loaded and InitPPC has been called.

[Hedeon]

Compared to what? Previous versions or compared to a system without the library loaded?

[indigolemon]

Originally Posted by Hedeon

Compared to what? Previous versions or compared to a system without the library loaded?

Sorry, I meant compared to previous versions. I only set it up to be loaded on demand as it made the system unstable to have it start with the machine. It would run a few things fine, but if I ran multiple apps that used the PPC it would eventually crap out. For example, QuakeWOS could be run once, then I'd need to reboot if I wished to start it again without the machine falling over.

With this version, those issues appear to be no more which is great!

[esel]

Originally Posted by Hedeon

No. I asked for specific info.

here it the info using the latest lib (397):

Pci Device 1:
   VendorID: $1002
   VendorName: ATI Technologies Inc. / Advanced Micro Devices, Inc.
   DeviceID: $5960
   DeviceName: RV280 Radeon 9200 Pro
   ClassCode: $30000
   RevisionID: $1
   InterruptPin: 2
   SubsystemVendorID: $174b
   SubsystemID: $250
   MemSpace0: $40000000 - $47ffffff
   IOSpace1: $60c06000 - $60c060ff
   MemSpace2: $48000000 - $4800ffff
   RomSpace: $48100000 - $4811ffff

Pci Device 2:
   VendorID: $1002
   VendorName: ATI Technologies Inc. / Advanced Micro Devices, Inc.
   DeviceID: $5940
   DeviceName: RV280 Radeon 9200 Pro Secondary
   ClassCode: $38000
   RevisionID: $1
   InterruptPin: 0
   SubsystemVendorID: $174b
   SubsystemID: $251
   MemSpace0: $50000000 - $57ffffff
   MemSpace1: $48200000 - $4820ffff

Pci Device 3:
   VendorID: $1957
   VendorName: Unknown
   DeviceID: $86
   DeviceName: Unknown
   ClassCode: $b2000
   RevisionID: $30
   InterruptPin: 3
   SubsystemVendorID: $1a56
   SubsystemID: $201
   MemSpace0: $48300000 - $483fffff
   MemSpace1: $4c000000 - $4fffffff

Pci Device 4:
   VendorID: $10ec
   VendorName: Realtek Audio/Lan　Maker
   DeviceID: $8139
   DeviceName: RTL-8139/8139C/8139D Realtek RTL8139 Family PCI Fast Ethernet NIC
   ClassCode: $20000
   RevisionID: $10
   InterruptPin: 4
   SubsystemVendorID: $1259
   SubsystemID: $2503
   IOSpace0: $60c06100 - $60c061ff
   MemSpace1: $48400000 - $484000ff

MuScan 40.3 (28.11.99) © THOR

68060 MMU detected.
MMU page size is 0x1000 bytes.

Memory map:
0x00000000 - 0x001FFFFF CacheInhibit Imprecise NonSerial 
0x00200000 - 0x00BBFFFF Blank 
0x00BC0000 - 0x00BFFFFF CacheInhibit I/O space 
0x00C00000 - 0x00D7FFFF Blank 
0x00D80000 - 0x00DFFFFF CacheInhibit I/O space 
0x00E00000 - 0x00EFFFFF Blank 
0x00F00000 - 0x00F7FFFF CacheInhibit 
0x00F80000 - 0x00FFFFFF ROM CopyBack Remapped to 0x07F60000 
0x01000000 - 0x03FFFFFF Blank 
0x04000000 - 0x07F5FFFF CopyBack 
0x07F60000 - 0x07FDFFFF ROM CopyBack 
0x07FE0000 - 0x07FFFFFF CopyBack 
0x08000000 - 0x3FFFFFFF Blank 
0x40000000 - 0x60FFFFFF CacheInhibit I/O space 
0x61000000 - 0xFFFF7FFF Blank 
0xFFFF8000 - 0xFFFFFFFF CopyBack Remapped to 0x04AB8000

is there a way i can test the PPC ? Maybe the card is faulty ?

[Hedeon]

Looks correct. Was the pciinfo taken after initppc? Where did you put powerpc.library? Does getinfo work?

[esel]

Yes, the pci info was used after ppcinit.
Getinfo is working but getinfo_ppc only shows all values 0.

[Hedeon]

Originally Posted by esel

Yes, the pci info was used after ppcinit.
Getinfo is working but getinfo_ppc only shows all values 0.

That would be correct. Getinfo_ppc was just a proof of concept. So how about Warprace and Voxelspace? Those things running? How about ADoom and QuakeWOS (software renderer) that working? What kind of error do you get? Lock up? PPC exception window? How fast? After starting the game? After a few minutes?

[esel]

yes warprace, voxelspace, adoom and quake are runnning but it will lock up after a few minutes. sometime after a few seconds sometimes after a few minutes.
most of the time it will lock the computer but two times it crashed and will show the guru i posted a site back.

The only app i can not get to work because its crashing at start is dnetc_wos.
This will show a PPC Execption window.

[Wrangler]

DNETC_WOS is compiled with altivec instructions I believe so only works on a G4.


The crashes you are getting could be a number of things. Based on my own experience, the following are candidates:

• PCI USB card - caused crashes for me, but others have them working fine. Took me a long time to track this down. If you have one, pull it out and see if there is an improvement.
• Overheating - particularly if it takes time to get a crash (like you do). Have you got enough cooling in the case? Leave the case off and turn a desk fan on the PCI/Zorro area as a test to see if that improves stability.
• Power problems. Could be a weak PSU or poor power distribution around the system. Try a different PSU and clean all connectors very carefully with isopropyl alcohol.

It's much more likely the problem you have is hardware than software.

[Hedeon]

Originally Posted by Wrangler

DNETC_WOS is compiled with altivec instructions

There are three versions. The original one which doesn't have Altivec and is readily available on distributed.net. And the other two which were a proof of concept which both contain Altivec but in one of them I messed up Altivec detection and hence crashes on a G2/G3. These two were only sent to a couple of people, don't think he has one of those versions.

[grelbfarlk]

Found a funny new error state, that I caused. I was testing out some modifications to one of the Raptures, and got no System VGA Memory detected (PCIDMA) turns out it was from setting Spiderbuf too high. Just throwing this out there in case someone gets a similar error.

[starship]

Hi guys.
I need to ask something because I couldnt found any solution.
When I open fpse_wos or neomame colors are wrong. I open palette settings on neomame and found right colors from rasters but when I exit it is the same. I am using radeon and tried on all screen modes. Is it big endian issue or what I am doing wrong?

[grelbfarlk]

Originally Posted by starship

Hi guys.
I need to ask something because I couldnt found any solution.
When I open fpse_wos or neomame colors are wrong. I open palette settings on neomame and found right colors from rasters but when I exit it is the same. I am using radeon and tried on all screen modes. Is it big endian issue or what I am doing wrong?

Yes unfortunately Radeon and Voodoo 4 are Little Endian only and Amidog's stuff is Big Endian. Not much to be done about that other than port those applications (which shouldn't be too hard).