News for hackers. Stuff that matters.
Re:One-pass Compiler by traes on Saturday May 23, 2020 @02:30PM (#23285224)
For anyone wondering, the NOPs are so that jumps don't get messed up. Assembly level GOTOs generally work on byte/word offsets, so if something intended to jump back before the calculation but was expecting two PUSH_INTs and an OP_PLUS (5 words: 3 opcodes, 2 arguments) but there was only a single PUSH_INT (2 words), it would jump back 3 words too far.
Re:One-pass Compiler by stevekemp on Saturday May 23, 2020 @02:43PM (#23285337)
Good clarification, and exactly right!

Later I do walk back over the bytecode and remove the nops. But I have to update all JMP/CALL instructions to cope with the changed destination offsets. Not a hard job in my case, as there are only a couple of instructions which refer to byte-offsets.

(If I allowed "LD A,[BC]", or similar permuations I'd have a lot more work to do.)