Sonnet 7200 PCI programming

[grelbfarlk]

Originally Posted by Wrangler

@grelbfarlk: Yep AmigaAMP v2.25 works fine. Thanks for the tip!

@Magic: Have you sent the environment variable VoodooMemOS to Yes?

@All: Could someone who has got Warp3D working with a Sonnet on a Voodoo 3 please post their env variables (Mediator, Warp3D, PowerPC, anything else)? Warp3D works for me on 68K but I get recoverable or unrecoverable alerts on PPC depending on the variables I set. I think I am missing something on the PPC side but I'm not sure what.

ENV:Mediator
Buster=11
FastEthernet=0
MMU=No
RadeonInt=No
RadeonMem=32
RadeonMemOS=Yes
SpiderBuf=0
VoodooInt=Yes (Important for Warp3D)
VoodooMem=13
VoodooMemOS=Yes
Warp3D=No

[Hedeon]

Originally Posted by Wrangler

@All: Could someone who has got Warp3D working with a Sonnet on a Voodoo 3 please post their env variables (Mediator, Warp3D, PowerPC, anything else)? Warp3D works for me on 68K but I get recoverable or unrecoverable alerts on PPC depending on the variables I set. I think I am missing something on the PPC side but I'm not sure what.

The PPC cannot give alerts, only the 68k. The ppc can give a crash output window. I know Warp3D can have a stack overflow. Did you install something of a stack enlarger?

[Hedeon]

@magic: could you check the output of snoop dos when you start initppc? Maybe remove ppc.library if you have it in libs?

[Wrangler]

@grelbfarlk/hedeon: Thanks for the env settings and libs - really appreciate it. Unfortunately, I am still getting the same problems.

Gears full screen PPC version hard crashes with a 80000003 guru = address error. Gears window and gears small PPC version first have a very brief recoverable alert 01000009 = freeing memory that is already freed (?) followed immediately by a task failure in gearsppc.exe 80000008 = privilege violation

The 68K version full screen works fine and other PPC progs (getinfo, cyberpi, cybermand, AmigaAMP/WUP decoder) are fine. As I'm using the same libs/envs as you this looks like something deeper. I'm attaching my MedReport for more detail. This is using powerpc.library build 252. Any help gratefully received!

[Hedeon]

Same question here. Do you use something like stack attack? Is mmu.library running? If so could you attach the output of MuScan?

[Wrangler]

No, no stack software running. Should I be doing that? MuScan output attached...

[wXR]

@Hedeon

Apologies for the slight diversion again. What other PowerPC-on-PCI cards do you imagine would be one day possible to support in SonnetAmiga?

http://www.everymac.com/upgrade_card...tem/index.html

[grelbfarlk]

Originally Posted by wXR

@Hedeon

Apologies for the slight diversion again. What other PowerPC-on-PCI cards do you imagine would be one day possible to support in SonnetAmiga?

http://www.everymac.com/upgrade_card...tem/index.html

The Sonnet Crescendo 7200 is the only one of out that bunch that is a PCI card. However there are other PowerPC on PCI cards that are possibilities.

As mentioned in the thread there is a currently an effort to reproduce this card which is pretty much a matter of reproducing the PCB and sourcing parts, there are no missing CPLDs as a roadblock.

@Wrangler have you tried setting a larger stack for the Gear launch script, like
stack 100000
gears -width 640

[Wrangler]

So, progress of a sort. Gave it a big stack and made sure I was in 640x480@16 in Workbench before launching gearsppc.exe (I recalled Warp3D is fussy about screen depth?). No restart this time - instead I was presented with a blank grey screen. A quick Amiga-M showed the attached Sonnet PPC exception. Does this give a clue what is going on?

[Hedeon]

It looks like you only have 32 mb installed? 53000000 to 60000000?

[grelbfarlk]

Also is your 640x480x16bpp workbench mode a PC mode?

Also remove this from your loadmodule:
DEVS:card.resource

This is just for A1200 and will cause all sorts of trouble for A4000, this is incorrectly added by BB3or4.

[Wrangler]

Card.resource has now been sent to silicon heaven. Same result.

Yes, 640x480x16 is a PC mode.

@Hedeon: not even 32MB, I think only 24 which is as much as I have until some DIMMs arrive from the good old US of A next week, hopefully. Maybe that will do the trick.

[Hedeon]

I see that the processor is in supervisor mode before it crashes (seen in srr1). Could you dump memory from the address shown in srr0 plus and minus $100 next time it happens? With a program like filex? I can then disassemble it with vdappc.... Maybe I should integrate it in the library..

[Hedeon]

I also noticed I forgot to implement disl2flush....

[amigoun]

I tried mess emu.
Amidogs MESS works ok (but it is quite slow, possibly unoptimized, and so I had to use frameskip 3 for Sonic).
But I can still not get WarpMESS to work. I used the same commandline as advised by grelbfarlk, and it starts, but after selecting the screenmode (rtgmaster selection) all I get is unfortunatelly a black screen.

[grelbfarlk]

Originally Posted by amigoun

I tried mess emu.
Amidogs MESS works ok (but it is quite slow, possibly unoptimized, and so I had to use frameskip 3 for Sonic).
But I can still not get WarpMESS to work. I used the same commandline as advised by grelbfarlk, and it starts, but after selecting the screenmode (rtgmaster selection) all I get is unfortunatelly a black screen.

Amidog's MESS yeah isn't fast for Genesis emulation, I picked testing Genesis because I thought (maybe wrongly) that it is the most demanding platform in MESS. C64 or the like should run quite fast but there are better C64 emulators I imagine.

I'll check WES MESS tomorrow, I haven't tried it in a long time, though I don't think I get an RTGMaster requester. If you haven't yet try setting a bigger stack.

[wXR]

@grelbfarlk

I'm excited to get ahold of one of those reproductions, and I hope they do come. Meanwhile, what are the names/model numbers of some of those other cards? Are there any developers who might want to take a stab at hacking support in for any of them? Will hardware adapters be needed as well?

I was demonstrating Bountysource in another thread, and added this:

https://www.bountysource.com/issues/...200tx-mediator

If it would be helpful at this stage, we could make some bounties for other cards as well, or perhaps bounties to accelerate the development of this reproduction.

Anyway, keep up the good work!

[strim]

Originally Posted by wXR

Meanwhile, what are the names/model numbers of some of those other cards?

There's for example:

https://www.xes-inc.com/products/io-modules/xport1013/

Which looks like maybe it could boot from the memory present on PCI bus (a requirement for SonnetAmiga). I wrote them (XES) asking about it and wanting to buy one for evaluation purposes, but never got reply from them.

[grelbfarlk]

Originally Posted by wXR

@grelbfarlk

I'm excited to get ahold of one of those reproductions, and I hope they do come. Meanwhile, what are the names/model numbers of some of those other cards? Are there any developers who might want to take a stab at hacking support in for any of them? Will hardware adapters be needed as well?

I was demonstrating Bountysource in another thread, and added this:

https://www.bountysource.com/issues/...200tx-mediator

If it would be helpful at this stage, we could make some bounties for other cards as well, or perhaps bounties to accelerate the development of this reproduction.

Anyway, keep up the good work!

I think a bounty to just design a new card would be the best thing to do. There's no secret magic here to the design, which is easy for me to say. The folks working on the reproduction are humble folks that aren't playing the hype+more hype+broken promises thing us Amigans love and hate.

Whether they would like to collaborate or even pursue a bounty is another story, it would be unfortunate that we have two teams working on the same thing.
A basic outline of requirements would go:
1) 5V and 3.3V Universal PCI card
2) SDRAM or better memory interface, SODIMM might be a good idea
3) MPC107 bridge or other PCI/PowerPC bridge that operates in Agent mode, with option to run in host mode (?)
4) 600MHz or faster MPC750 code compatible
5) Quantities of required parts in the wild to exceed 1000+

Maybe that last one is a little ridiculous or not possible.

[wXR]

@grelbfarlk

I'm certain you could sell a few hundred of them... Well, I suppose that last part depends on how many PCI bridgeboards there are out in the wild. That surely isn't a common item, right? Do we have any sense at all about those numbers?

Are the people working on the reproduction even interested in/motivated by bounties, is maybe a better starting question. And then, if so, is the SonnetAmiga repo an appropriate place to attach the bounties, or someplace else? Do you have any idea if the board design will be open source? If so, and if it is already started, then clearly that repo would be the best place.

@strim

I know you don't have a lot of time, but if you've already done some research on this, maybe it would be helpful to compile a list of alternative cards (and requirements for them to work) on the SonnetAmiga repo wiki?