joe-degs / software-machines Goto Github PK

This is a code along of Eleanor McHugh's talk on building components of virtual machines in software(how vms are built). Basically abstract virtual machines. I do not really know anything about stuff like this, but reading and listening to stuff about the intricate parts of machines always fascinates me, so i'm trying out some of the code to understand more of it and how all this things work.

The codes are in c and go, go I am a little bit comfortable with but c i do not know anything about. But because go is close to c i feel like i know some things about it. So i'm testing the c code too.

Lets start with the components of computer machines;

There are few models of memory around. Most, i have not heard of before. One i actually know is the Von-Neumann memory model but, turns out there is a bunch of models around. The basic function of the memory is storing data and instructions. It is sometimes used for protection too.

• Von Neumannn model: This one is just like an array. One big contigous chunk of addressable space, separated/partitioned so we have different parts of the chunk for stack, heap anbd the rest. The basic one everybody who hasnt taken a computer science course on the subject knows about.
• Harvard model: Separate ROM and RAM, this means instructions and data are not stored in the same contigous chunk of memory. This potentially prevents buffer overruns attacks. so code will be stored in the ROM and data in the RAM.
• Forth model: Complex model with two diffent stacks, not really a standard heap but some lookup map. And some other complicated details
• Hybrid model: This one combines both forth and von neumann model to create a more complicated model that probably has more than two stacks.

• the heap is word aligned, byte addressable chunk of contigous memory. It is basically the same accross all the memory models. And the implementation in go might not be that different the implementation in c

• The stack is a slightly different beast, depending on the memory model, it might be trivial to implement. The implementation will be of the type linux processes use.
• There are a few types of stacks. We have the spaghetti/cactus stack and the array stack(regular linux process stack)
• The stack is a fixed array of memory, can be grown and shrinked and has a program counter that is an offset into the array.
• Whereas the array stack is just one contigous chunk, the cactus/spaghetti stack is list of memory cells linked together, basically a linked list. It is a functional data type especially if you make it immutable.

• maps are basically just key value pairs. This is the holy grail of datastructures. If you in the computer programming business, every turn you take you will hear somebody talking about hash maps, lookup tables, hash tables etc which are all maps. In distributed computing we have consistent hashing algorithms which use maps too. So yeah its a great deal for me, finally getting to play with map implementations is a great deal for me.
• Implementation Details: There are three basic aspects to maps.

1. where the (key, value) pairs are stored. this can be done using associative arrays
2. a way to search for (key, value) pairs.basically i nice search algorithm will do
3. a map interface to be able to set and get (key, value) pairs easily. most people use buckets