Need suggestions for Kickflash board.

[platon42]

ral-clan wrote:

Very cool.  But when Poseidon is running, is it quite a RAM/CPU hog?  I've heard that because USB devices are fairly dumb, USB transfers & stacks can really slow down older computers.

Poseidon itself is nearly idle if no activity is on the USB (massstorage.class for example will check every three seconds if a media is in the drive though). Depending on the capabilities of the host controller hardware, the driver however, will need to poll the bus for interrupt transfers (e.g. when using USB hubs, mice, keyboards, etc.)

In case of the Algor and Deneb, this causes no additional CPU load because the host controller will do this in a predefined schedule in hardware. In case of the Subway and Highway, the CPU must trigger the schedule and will therefore generate a very short interrupt every 1 ms (i.e. 1000 interrupts per second). This sounds a lot, but is not as bad as it seems. I didn't measure the difference, but I would be surprised if it took more than 2-5% of your CPU power away.

In comparison, AFAIK the A4000T scsi controller transfers data in 1 KB DMA transfers, and after every DMA transfer, an interrupt is caused, which then again wakes up a task, which in turn starts up the next DMA transfer. And at a rate of nearly 9,5 MB/sec, that's about 20000 interrupts per second.

[platon42]

ral-clan wrote:
Okay, the Kickflash has an A1200 style clockport connector.  I understand that the Amiga sees this as an extra Zorro slot.

No it doesn't. Commodore didn't provide a standard way for accessing clockport hardware and at that time, there only was *the* A1200 clockport. Not four of them, not extra clockports on Zorro cards. It would have been clever to add a clockport.resource to the system that would know about the available clockports back in 1993. But that didn't happen and now that most driver writers for clockport cards are gone (Melody, Silversurfer, Twister), building one now wouldn't help much.

So every driver must currently know about all clockports there are on every existing Zorro card and check them manually. The clockport is just a I/O region of 16 bytes somewhere. There is no ID for clockport hardware, hence the recognition of it depends on the hardware itself -- in theory a check of the one hardware with the other hardware connected could crash the machine and vice versa. The Subway for example, has a recognition mechanism that will only require multiple read accesses to one register, no writing is necessary. But this is not the case for other clockport hardware.

If a new zorro card is release with a clockport interface, the driver must be updated or somehow patched. I've done that for the Melody driver for example, which of course didn't know about the Highway at that time.

 But is that Zorro II or Zorro III?  But since my Kickflash is going to be plugged into an A2000 computer, which is Zorro II, does this mean that certain clockport devices which need Zorro III transfer rates should be used with a Kickflash attached to an A2000 Zorro II slot?

Except for the Deneb USB, there are currently no Zorro III cards with Clockports on it. The original Clockport on A1200 has a very very slow timing. Most clockport implementations on Zorro II cards have a faster timing, but AFAIK every clockport hardware also works with that timing.

The Deneb is the first Zorro III card with clockport. I think Michael Böhmer wants to implement configurable timing for that, so that it can go even faster than the Zorro II clockports available.

No clockport devices "need" Zorro III transfer rates. The Clockport is a 4 bit address, 8 bit data bus with slow bus timing (see specs).