.
worth a read. Using formal methods and the RISC-V testsuite. Tackling a simple machine first, but with some thinking ahead, with the aim of hitting reasonable performance:

https://tomverbeure.github.io/risc-v/20 ... ation.html

A very different kind of debugging progress compared to the usual approach using simulation - and overall more time-efficient, I reckon. But, you do need the testsuite, which in the case of RISC-V has been written for you, and in the case of a homebrew CPU that would be a major task with its own kind of expertise.

How ever formal logic still cannot stop simple typo or logic errors.
Working a on cross compiler with two versions, I had a output print out of
130 pages with only two typo errors. 1 #define BPW 4 rather than the
correct #define BPW 2 for that machine and forgotton & for &p->name[0]
Ben.