Sonnet 7200 PCI programming

[Hedeon]

Ok, both the K1/M1 and the Apocalypse/Ragnarok show the same problem with the Prometheus: A response time of 45 milliseconds instead of 250 microseconds.

I need to see what the issue is. Maybe contact the developer of the Prometheus.

[Hedeon]

@BULI could you contact the man to see how I can best contact him? :-)

[starship]

some cooling again

now on notebook style, hot air directly to outside.

[grelbfarlk]

Originally Posted by starship

some cooling again

now on notebook style, hot air directly to outside.

It looks like you used some male/female PMC connectors to make it wider is that correct?
Did you swap in the PPC750GX@1GHz like you were supposed to?

[starship]

Yes I did it to make it wider.

I did not changed or touched cards.only buy 4 male and 4 female connector. With this connectors I made 4 connector. 1 side is male other side is female so I have a space for fan.

[starship]

And I made auto aga/rtg switch with pic16f628A. Works great with powerpc.library

[grelbfarlk]

Upgrade the PPC750FX to PPC750GX@1GHz, then you can have a unicorn. You get double the L2 cache and then we'll know if it works.

[starship]

you mixed my mind. Pin and voltage compatible? Hmm I will look

[jbenam]

Originally Posted by starship

And I made auto aga/rtg switch with pic16f628A. Works great with powerpc.library

Slight OT: Do you plan on release the source? I might be interested in building one

[starship]

I released. It is in my topic inside .rar file.

[BULI]

Originally Posted by Hedeon

@BULI could you contact the man to see how I can best contact him? :-)

Take a look at PM

[Hedeon]

Originally Posted by BULI

Take a look at PM

Thanks!

I ran a test program on the VBI of the voodoo3 driver (which is connected to INT2) and you guessed it, it is at 91ms (so like 11 Hz). I am not sure if this is standard for the Prometheus or that it is an error in the FireBird.

[Hedeon]

Message to the Prometheus HW guy:

Hi,
I am trying to write a driver for the prometheus PCI bridge (or actually the firestorm version with the E3B firmware) and I notice a very long delay from PCI interrupt to INT2 (or a wrong one).

The mediator is in the range of 350 microseconds that I can handle a PCI interrupt on the Amiga side. For the Firestorm it seems to be 91 milliseconds.
I even hacked the VBInt Prometheus.card interrupt to add a counter and also here. 2100 counts in 200 seconds (approx) which is 11 Hz or 91 ms. I would expect 50 or 60Hz.
Is this normal? Is this a Prometheus error? Is this a Firestorm firmware error?
I heard you are busy and that it was long ago (20 years), but I hope you can shine a light on which way to look. Or if this is normal behaviour, to sadly give up the project (PowerPC cards for the Firestorm).

Regards,

[Hedeon]

I am guessing that none of the PCI interrupt get through. The 11Hz stuff seems to come from the ciaa.resource.

A quick look to the ROM code (https://eab.abime.net/showthread.php?t=62830&page=3) seems to indicate that this interrupt always sets the Z flag so falls through to the next one in the server chain. Which is my case it the VBI Prometheus.card.

If I copy stuff (SCSI INT2) then the PPC is much faster to react. (as the SCSI interrupt prio is below the PPC and VBI prometheus.card interrupt)

[Hedeon]

Have put some questions out at different HW/SW people. Let's see what comes of it.

[Hedeon]

Found the issue :-)

[indigolemon]

Originally Posted by Hedeon

Found the issue :-)

Oooo! Do tell

[Hedeon]

Originally Posted by indigolemon

Oooo! Do tell

Error in the prometheus.library I have. Not sure if there are multiple versions out there. I have a debug version.

[Hedeon]

Also Warp3D now running

[torsti76]

Originally Posted by Hedeon

Also Warp3D now running

Wow! Great news! Have to finally assemble my AA3000+ for Christmas. :-)