[Coco] CoCoRX

Frank Pittel fwp at deepthought.com
Wed Jun 3 23:19:18 EDT 2015


On Tue, Jun 02, 2015 at 06:08:06PM -0400, Robert Gault wrote:
> Frank Pittel wrote:
> >Sorry for responding so late but I got behind in reading email.
> >
> >Isn't the idea of the cartridge to test actual cocos? Unless seriously broken
> >in ways I can't imagine possible a 6809 or 6309 is always going to clock cycle
> >accurate. In the case of an emulator running on a PC or FPGA it should be up
> >to the author(s) to indicate if it's clock cycle accurate. It's well known that
> >the "cpu" on the coco3fpga isn't cycle accurate and neither will anything based
> >on that core.
> >
> >The Other Frank
> >
> >
> >
> >On Sat, May 30, 2015 at 02:43:25PM -0500, Melanie and John Mark Mobley wrote:
> >>We need a diagnostic test that will prove out the cycle accuracy timing of
> >>the CPU so that it can be used to test the FPGA design.
> >>
> >>John Mark Mobley
> 
> To add a comment on this request, it would be impossible for any
> software to verify a "clock cycle" as it would run using the clock
> cycle in question. Unless the PAK contained its own crystal and
> timing circuits, there would not be any basis for measuring the Coco
> or emulator timing. If the PAK did include a clock circuit, how
> would that timing be verified?
> 
> Perhaps if the PAK included a realtime clock, the crystal in the
> clock could be used to validate the Coco/emulator timing by
> measuring the number of software loop cycles per some unit time. I
> have doubts about the precision of such measurements and the added
> cost to the PAK would probably be prohibitive.
> 
> A high precision counter would be needed to measure the
> Coco/emulator at various test points and the counter would need to
> be certified as to its accuracy to provide meaningful results.
> Better just to assume if the unit under question produces reasonable
> pictures on a monitor, the timing is adequate.

My initial thoughts were to take advantage of the hardware interrupts for video.
I didn't give it nearly as much thought as you have. :) The reason is simple An
actual 6809 or 6309 is going to be clock cycle accurate unless it has some serious
issues and should fail on a host of other tests. ( I hope anyhow )




More information about the Coco mailing list