Sonnet 7200 PCI programming

[jbenam]

Originally Posted by trixster

What processor library are you using? Have you tried mmulibs? What does BULI’s Config use?

I use Thor's mmulib on my config. Not sure about BULI's, I've already reflashes the CF with my image. I'll do some more testing with the beta library later.

EDIT: Same thing with the BETA library. I've also tried the diagnostic start without startup sequence as written in the wiki. Same error.

@Hedeon
I haven't got any logging or output - am I missing something?

[trixster]

Have you used the rexx buildconfig script to setup an mmu-configuration file in prefs:env-archive, as discussed in https://github.com/Sakura-IT/SonnetA...nd-MMU-library

[gsoravil]

Phew, this is a lot to keep up with! Some of this is really above my head (at least for today) but I will try my best to keep up. I never really played in Amiga MMU land.

I can confirm that like jbenam I do sometimes get an 80000004 with Killer NIC (with PMC I do not).

So maybe there is hope on our v1 Mediators?

I will try to play with the MMU libs (per Trixter) and Buli's alternate powerpc.library as soon as I have a chance.

Buli - A3640 is the only 68K CPU card I have. I was hoping to just upgrade it to 060 w/custom ROMS (e.g. poor man's "A3660") and this project would supplement it on the PPC side. At least that's my goal.

[Hedeon]

Not sure if that library has debugging enabled. But it is over serial. Or use sashimi to redirect.

[jbenam]

Originally Posted by trixster

Have you used the rexx buildconfig script to setup an mmu-configuration file in prefs:env-archive, as discussed in https://github.com/Sakura-IT/SonnetA...nd-MMU-library

Yep, nothing changed. I still get the same ramlib error, memory gets added, stuff freezes after opening a few windows.

Originally Posted by Hedeon

Not sure if that library has debugging enabled. But it is over serial. Or use sashimi to redirect.

Thanks, thought as much. I'll try tomorrow and see if I can get something useful out of it.

[BULI]

Originally Posted by trixster

What processor library are you using? Have you tried mmulibs? What does BULI’s Config use?

In this config I used 68040.library from Commodore (for 36040 CPU card) and Oxypatcher, to proper init PPC, that should be enought.


About ram library, you can try startup Workbench with pressed right ALt on keyboard (start without WBstartup and Oxypatcher), and you will see is there any change.

[jbenam]

@Hedeon
With Sashimi enabled (launching the executable without any flags, not sure if I need any) I don't even get the ramlib suspend/reboot prompt, it just freezes.

I can use a real serial cable with my lab PC if needed, but I'm not sure about baud rate, parity, etc. Any pointers?

[Hedeon]

Let's start over: please provide a pciinfo output and a MuScan output (from mmu lib package).

If no mmu lib installed. just try the following:

Copy the correct mmu package cpu lib and the mmu lib to libs:
reboot
MuScan (please show output)
InitPPC

[jbenam]

Originally Posted by Hedeon

Let's start over: please provide a pciinfo output and a MuScan output (from mmu lib package).

If no mmu lib installed. just try the following:

Copy the correct mmu package cpu lib and the mmu lib to libs:
reboot
MuScan (please show output)
InitPPC

I already had MMUlib installed, but I noticed there's a new version from April and so I've ran the Installer again to update my current install.

Rebooted, got a MUScan (attached to this post) and then launched StartPPC (I do not use InitPPC as I do not have powerpc.library in Libs, but in powerpc, as advised on the wiki), now it just freezes

EDIT: Attached PCIinfo too, might be useful. Anyway, after trying again, we're back at ramlib suspend/reboot as usual.

[Hedeon]

What happens if you make everything from 0x40000000 to 0x5FFFFFFF cache inhibited?

I'll post a test version of the library later today where debugging is for certain enabled.

Do you have stackattack2 installed?

[jbenam]

Originally Posted by Hedeon

What happens if you make everything from 0x40000000 to 0x5FFFFFFF cache inhibited?

I'll post a test version of the library later today where debugging is for certain enabled.

Do you have stackattack2 installed?

I've put this in MMU-Configuration:

SetCacheMode from 0x40000000 size 0x1FFFFFFF CacheInhibit

Is that right?

Yep, I do have stackattack2 - with argument 32768. Same ramlib error.

Alright, fingers crossed for the debugging enabled version. Thanks for your time, much appreciated!

[BULI]

Chellange: Sulaco by Encore on my Amiga 4000 PPC (Ragnarok), this time with correct sound
[ Show youtube player ]

Look at the frames rate.

Originally Posted by Hedeon

Seen this?

[ Show youtube player ]

Look at the numbers. That is 604e 420MHz with a CVPPC card.

At 6:22. I'm guessing it's the CVPPC. Would be nice to know how it does with a Voodoo3.

It does of course run AOS4.

[Hedeon]

It is limited to 50FPS sadly. The Mad Wizard demos are not.

[BULI]

Originally Posted by Hedeon

It is limited to 50FPS sadly. The Mad Wizard demos are not.

Yes, it's true. Sulaco without sound give me perfect 50FPS on every scene.

[johnnybarker]

This is on my a3000 with 420mhz CyberstormPPC and voodoo 2000. Nowhere near 50fps.

[trixster]

Yeah, BULI’s 800+MHz G3 is much faster than the 604e on a csppc

[Hedeon]

@johnnybarker

Thanks for sharing that. I guess memory and CPU to PCI access speeds is also a thing. The demos do not suffer THAT much when you turn off the second level cache of the G3/G4.

[johnnybarker]

Should add that it was on full screen mode and 70ns ram speed. It crashed when I tried 60ns so perhaps would get a little more with better simms but probably not much. I can try it on the apocalypse card @450mhz when I get my PSU reassembled on the a4000. I expect results will be somewhere in between but not to far off the G3.

[DarrenHD]

Originally Posted by Hedeon

@johnnybarker

Thanks for sharing that. I guess memory and CPU to PCI access speeds is also a thing. The demos do not suffer THAT much when you turn off the second level cache of the G3/G4.

I think my times are the same or slightly better with my (not overclocked) CS PPC, 233MHz, but with a GREX 4000. Voodoo 3. I can post them...

The GREX-4000 gets about 19MB/sec on long writes to graphics ram over the GREX, if the "MemTesting030+RTG" benchmark I'm using is anywhere near accurate.

A Mediator/Prometheus gets 8-9MB/sec in long writes to graphics ram.

Note: The GREX was advertised with 50MB/sec...:-/

[BULI]

Originally Posted by johnnybarker

This is on my a3000 with 420mhz CyberstormPPC and voodoo 2000. Nowhere near 50fps.

Try run this demo in Window Mode (like me in video from YT)- you should have little better performance.