2025-3-2:
The 8086 Decoder was written as a part of Casey Muratori's Performance Aware Programming Course. We were given only the basic problem of reading an decoding an 8086 instruction stream, without any general suggestions at all as to how the software should look or what technology to use beyond this simple requirement. As of this writing, the program will decode all MOV, ADD, SUB, CMP, and conditional JMP and LOOP instructions. This software is written in something close to idiomatic C89, compiled with the MSVC C++ compiler.
The purpose behind the project is to learn intimately how instruction encoding works, and having built this over the course of many hours I can say that I won't soon forget what mod reg r/m means or the byte layout for various kinds of MOVs. Intel instruction encoding being variable length means that you have no real idea how long an instruction will be until you start decoding it. Based on certain bits which are usually in the first two bytes you will learn whether and what size displacement and data values follow. Going into this, I hadn't spent a ton of time dealing directly with binary files but by the end I had a fairly good sense for reading and understanding hexadecimal without needing recourse to a conversion tool, like in days of yore.
I briefly considered things like jump tables or more up front infrastructure for representing instructions in memory, but ultimately I went with pretty straightforward decoding and a general skeleton which saves some but not all of the data which will be needed for the next section of the course: writing an 8086 simulator on top of the decoder.
Given that this is a performance aware programming course, I will rewrite the decoder portion in the near future and eliminate as much branching as possible, as well as potentially including the entire spec as a separate file for use in generating aformentioned jump table if I decide to go that direction. It will be instructive to profile the various forms that the program takes.