I’m generating correctly-executable code for an arbitrarily large number of processors from an architecture description and ‘hand-assembled’ program (using generated field macros)

And now I begin to see the lacunae….

For example, in thinking about how to incrementally add instructions, I realized that there were no checks in place to check the uniqueness of instruction decode. Bah, humbug…

Now to think hard about how either change the instruction encoding data structures; or to faff with a collection of instructions and discern the decode tree after the fact (as I did with the very first version of this stuff, dealing with instructions presented in more or less alphabetical order of mnemonics); or to work out how to insert things into the current structure.

This last seems most attractive at the moment.

It doesn’t fix the ‘uniqueness’ thingy, but that might fall out, or be amenable to a simple ’symbol table’ solution,,,

© kiva design groupe • 2017