Peter wrote:V2 is completely wrong! Getting a Coldfire V2 to run QL code natively is impossible, even with hugest software efforts. Misimplemented instructions do not trap out, and for the V2 that problem is far worse than e.g. for the much later V4e.
Nasta intended to use the first generation Coldfire, the one before the V2. Which was not a Coldfire internally, but a cut-down 68EC040 with multiplexed bus and only 1 KB data cache. The chip you probably mean was the MCF5102. It should not be confused with a post-V4 microcontroller from Freescale, which was misleadingly also named V1.
Aye, their naming conventions were bonkers. I say V2 because they released a "Coldfire" family member before the version Nasta used. So yes, the MCF5102. I suspect have the reason for the broad non-acceptance of the platform was that the ICs were too incompatible with each other and required too much rework if you wanted to go up or down within the performance range at all.
As for the Dragonball, programmatically it is identical to a 68SEC000. It does require about 40 bytes of config to turn off all the optional features, deliver all the signals needed to external pins, configure the clock multipliers and bus size domains/chip selects. After that, it's just a really fast 000.
One neat trick I've read is that you can use the chip selects as interrupts to trigger code to reconfigure the clock, so accesses to CS0 can be "8 bit, 7.5MHz" while everything else CS1..3 is "16 bit, 66MHz"