A1200 clockport, what can it be used for?

[Karlos]

I heard tale of a device that you can attach to the port that keeps your machine abreast of the current time and date. Incredible, eh?

[Trev]

Why would you use a clockport for that? If the common name were userport, I might consider attaching a clock to it.

[Karlos]

Why would you use a clockport for that? If the common name were userport, I might consider attaching a clock to it.

Exactly :lol:

[Cammy]

Add a Subway and a USB Ethernet Adaptor....and here we go.

This isn't a solution for 8MB A600/A1200 users, it's a slow option for accelerated Amigas at best. If there was a more effective way of connecting the RR-Net and suitable drivers it might be usable, but really we just need a real, dedicated ethernet or wireless card for the Amiga clockport. Maybe it could have an extra clockport on the card to daisy chain it to another expansion without requiring a splitter.

[Trev]

The CS8900 series (the chip used on the RRNet) doesn't reliably generate interrupts in 8-bit mode (per the datasheets), so it's up to the CPU to push the data around. (Those lines aren't wired on the RRNet, regardless.) Assuming the CPU could keep up with the traffic, a chip that does generate interrupts might produce suitable results. In either case, though, the card would be functional, and the higher-level protocols were designed with dropped frames in mind.

So really, the hardware isn't really the problem. It's the stack. AmiTCP and its derivatives have pretty hefty requirements, and a new, lightweight stack would need its own bsdsocket.library implementation to be useful to existing software. Open source applications could be modified to fit.

[platon42]

This isn't a solution for 8MB A600/A1200 users, it's a slow option for accelerated Amigas at best. If there was a more effective way of connecting the RR-Net and suitable drivers it might be usable, but really we just need a real, dedicated ethernet or wireless card for the Amiga clockport. Maybe it could have an extra clockport on the card to daisy chain it to another expansion without requiring a splitter.

The A1200 clockport is an 8 bit databus, 4 bit address bus connector with extreme and artificial wait states to allow slow clock chips to work. 4 bit address bus means that there are a total of 16 possible addresses, each one byte wide. Each address is four bytes away from eachother, so there's no way to combine word or long word accesses to get faster access.

There was exactly one (!) card that allowed daisy chaining with exactly one other card and AFAIR it was the Melody+Twister combination. It was specifically designed to allow this kind of thing, and obviously, because each card only used part of the address space.

It is nearly magic that a USB card like the Subway actually works on that kind of bus. The Subway performs much faster on Zorro card clockports due to the better timing though.

The RRNet could work on the Amiga clockport, but it lacks an IRQ which probably makes it awkward software wise (polling for interrupt bits is a bit too retro). Anyway, the speed of such a device (or a redesigned version with IRQ support) on an original A1200 clockport probably won't be much higher than a Subway + USB Ethernet adapter solution, because in the end, it comes down to copying memory across the bottleneck of the clockport.

[Trev]

Polling works well on the RRNet, but deciding when and where to poll quickly becomes a problem. I was using vblank and a low priority idle loop, which worked well in a test harness. Periodically forcing a softint to boost priority improved network performance, but the scheme I was using didn't play well with others. :-/

EDIT: I should add that network interface polling is (or was) in vogue: http://info.iet.unipi.it/~luigi/polling/

[Cammy]

I guess daisy chaining and splitting on the A1200 clockport is out of the question then.

Please don't think I was blaming Poseidon for the slow speed of the Subway/USB ethernet solution, I'm aware it's the clockport that is the real bottleneck.

But surely a new, dedicated clockport ethernet card would work with less RAM than using a USB ethernet card through Poseidon. I wouldn't have expected significant speed increase, but at least a card like this would be of use to people with un-accelerated Amigas with clockports and extra RAM, especially since PCMCIA ethernet cards will reduce most 8MB cards to 4MB, and 4MB isn't fun to use online.

[jj]

How do cards like the Z4 busboard work with multiple clockports ( never used mine ).  Do they run throught the zorro space ?

[platon42]

I guess daisy chaining and splitting on the A1200 clockport is out of the question then.

Please don't think I was blaming Poseidon for the slow speed of the Subway/USB ethernet solution, I'm aware it's the clockport that is the real bottleneck.

I wasn't thinking that;)

But surely a new, dedicated clockport ethernet card would work with less RAM than using a USB ethernet card through Poseidon. I wouldn't have expected significant speed increase, but at least a card like this would be of use to people with un-accelerated Amigas with clockports and extra RAM, especially since PCMCIA ethernet cards will reduce most 8MB cards to 4MB, and 4MB isn't fun to use online.

You don't have to load all class drivers with Poseidon. You need the main library (120 KB), the subwayusb.device (13 KB + 16 (?) KB), the hub.class (12 KB) and the ethernet driver (18 KB + 4 (?) KB), the Poseidon overhead can probably be less than 200 KB of memory. You don't even have to load the MUI stuff... But sure, having a dedicated driver only for a clockport ethernet module would probably only need about 10-12 KB ram.

Anyway, Poseidon was not meant to be used on unaccelerated A1200 machines without additional RAM.

[platon42]

How do cards like the Z4 busboard work with multiple clockports ( never used mine ).  Do they run throught the zorro space ?

No. They use multiple address spaces at an offset of $4000 for each clockport (base address is 0xd80001, which clashes with some chip area on A4000D/T machines -- peeking and poking in that area on an A4000 machine causes all kinds of havok to the IDE/SCSI port). Programmers that want to support these additional boards need to adapt their drivers for it and detection of multiple clockports that are not mere mirrors of the first one, is not trivial.

[delshay]

i wonder how the QUADDRAPORT works.

[Akiko]

i wonder how the QUADDRAPORT works.

Also Jens is planning an optional clockport expander, for his IDE-Fix Express (rev 2)
 
I wonder how this works, and if it mean a faster clockport and use of a second clockport expansion!?

http://amigakit.leamancomputing.com/catalog/product_info.php?cPath=42&products_id=425

[mikeymike]

I heard tale of a device that you can attach to the port that keeps your machine abreast of the current time and date. Incredible, eh?

I just blu-tack a watch to my Amiga!

[mongo]

Clockport splitter -

http://www.robthenerd.com/show.php?sect=projects&sub=clockport

[kolla]

Hopefully someone will have the vision to create a clockport Ethernet card that works with Amigas so all those A600 and A1200 users with 8MB Fast RAM can use it all online and not be limited to 4MB/5.5MB because of a PCMCIA network card. Then all we'd need is a 8MB/RTC/IDE/Clockport card for the CD32 and we'd be in expansion heaven.

Hear hear!!! :hammer:

Also, dont forget that TCP/IP is not the only thing to use ethernet for, there's also Envoy and various other "pre internet" protocols that use SANA2. I want my CD32 machines online primarly to avoid "floppy net" and hazardous (and with SX32, incompaible) pronet/parnet/serial cables.

Clockport adapter for CD32 and clockport ethernet has been on my wish list forr years