Revisit the ISA, decide finally and debug/fix the existing software

[sy2002]

There are open questions that came up while Michael created the great and comprehensive new test_programs/cpu_test.asm. Some of them are described here: #53 (comment) and some of them are described here: #50, #40 and #37

The compromise solution for now is: On the emulator, Bernd already implemented the "non forgetting" SR register and we should also change the hardware accordingly, because this does not seem to break things: #53 (comment)

Still open are topics around the V flag, CMP semantics and other topics.

So this issue is about

• Revisiting the ISA
• Finally deciding the route
• (Here comes the monster loose end): Debug/fix all the existing software, that relies on the "wrong" behaviour, for example of the V flag which is today simply implented as: "adding/subtracting two positive numbers and getting a negative result or adding/subtracting two negative numbers and getting a positive result"

As for the time of creating this issue, we already foresee, that the following might turn out into a large undertaking. So it is labled with "V1.7" as a placeholder for "not in V1.6 but later"

[sy2002]

As decided in our videochat, here is the way forward:

About Overflow

Change overflow handling to "correct behaviour", which is: Is written in the comment of the emulator and what is written here:

http://www.righto.com/2012/12/the-6502-overflow-flag-explained.html

Branch subroutine calls should not modify the status register (just as the hardware is doing right now in V1.6)

About Compare

Change compare to implement a SUB instruction without writing back the results.

**CAUTION: THESE CHANGES WILL BREAK ALL EXISTING SOFTWARE IN SUBTLE AND NOT SO SUBTLE WAYS - A LOT OF DEBUGGING REQUIRED **

[sy2002]

Without having a debugging infrastructure, this one needs to go to 1.8