Add bytecode vm interpreter

This commit is contained in:
Daniel M 2022-01-31 22:14:05 +01:00
parent e28b3c4f37
commit 5eae0712bf
4 changed files with 359 additions and 2 deletions

191
src/bytecode.rs Normal file
View File

@ -0,0 +1,191 @@
use std::collections::HashMap;
use crate::ast::{Ast, Expr, Stmt, BinOpType};
#[repr(u32)]
#[derive(Debug, Clone, Copy)]
pub enum OP {
Push,
Pop,
Load,
Store,
Add,
Subtract,
Multiply,
Divide,
Modulo,
Eq,
Neq,
Gt,
Lt,
Jump,
JumpTrue,
JumpFalse,
Value(u32),
Print,
}
#[derive(Debug, Default)]
pub struct Compiler {
ops: Vec<OP>,
global_vars: HashMap<String, u64>,
}
impl Compiler {
pub fn new() -> Self {
Compiler::default()
}
pub fn compile(&mut self, ast: &Ast) {
for stmt in &ast.prog {
match stmt {
Stmt::Expr(expr) => {
self.compile_expr(expr);
self.ops.push(OP::Pop);
}
Stmt::Let(name, rhs) => {
let id = self.global_vars.len() as u64;
self.global_vars.insert(name.clone(), id);
self.compile_expr(rhs);
self.gen_store(id);
}
Stmt::While(cond, body) => {
let idx_start = self.ops.len();
self.compile_expr(cond);
self.ops.push(OP::JumpFalse);
let idx_jmp = self.ops.len();
self.gen_i64(0);
self.compile(body);
self.ops.push(OP::Jump);
self.gen_i64(idx_start as i64);
self.overwrite_i64(idx_jmp, self.ops.len() as i64);
}
Stmt::For(_, _, _, _) => todo!(),
Stmt::If(cond, if_block, else_block) => {
self.compile_expr(cond);
self.ops.push(OP::JumpFalse);
let idx_if = self.ops.len();
self.gen_i64(0);
self.compile(if_block);
self.ops.push(OP::Jump);
let idx_else = self.ops.len();
self.gen_i64(0);
self.overwrite_i64(idx_if, self.ops.len() as i64);
self.compile(else_block);
self.overwrite_i64(idx_else, self.ops.len() as i64);
},
Stmt::DbgPrint(_) => todo!(),
Stmt::Print(expr) => {
self.compile_expr(expr);
self.ops.push(OP::Print);
}
}
}
}
pub fn into_ops(self) -> Vec<OP> {
self.ops
}
pub fn compile_expr(&mut self, expr: &Expr) {
match expr {
Expr::I64(val) => {
self.ops.push(OP::Push);
self.gen_i64(*val)
}
Expr::Ident(name) => {
match self.global_vars.get(name).copied() {
Some(addr) => self.gen_load(addr),
None => panic!("Variable '{}' used before declaration", name),
}
},
Expr::BinOp(bo, lhs, rhs) => self.compile_binop(bo, lhs, rhs),
Expr::UnOp(_, _) => todo!(),
Expr::Str(_) => todo!(),
}
}
fn compile_binop(&mut self, bo: &BinOpType, lhs: &Expr, rhs: &Expr) {
if matches!(bo, BinOpType::Assign) {
self.compile_expr(rhs);
if let Expr::Ident(name) = lhs {
let addr = *self.global_vars.get(name).expect("Trying to assign var before decl");
self.gen_store(addr);
} else {
panic!("Trying to assign value to rvalue");
}
return;
}
self.compile_expr(lhs);
self.compile_expr(rhs);
match bo {
BinOpType::Add => self.ops.push(OP::Add),
BinOpType::Sub => self.ops.push(OP::Subtract),
BinOpType::Mul => self.ops.push(OP::Multiply),
BinOpType::Div => self.ops.push(OP::Divide),
BinOpType::Mod => self.ops.push(OP::Modulo),
BinOpType::BOr => todo!(),
BinOpType::BAnd => todo!(),
BinOpType::BXor => todo!(),
BinOpType::Shl => todo!(),
BinOpType::Shr => todo!(),
BinOpType::Equ => self.ops.push(OP::Eq),
BinOpType::Neq => self.ops.push(OP::Neq),
BinOpType::Gt => self.ops.push(OP::Gt),
BinOpType::Ge => todo!(),
BinOpType::Lt => self.ops.push(OP::Lt),
BinOpType::Le => todo!(),
BinOpType::Assign => todo!(),
}
}
fn gen_i64(&mut self, val: i64) {
self.ops.push(OP::Value((val & u32::MAX as i64) as u32));
self.ops.push(OP::Value((val >> 32) as u32));
}
fn overwrite_i64(&mut self, idx: usize, val: i64) {
self.ops[idx] = OP::Value((val & u32::MAX as i64) as u32);
self.ops[idx+1] = OP::Value((val >> 32) as u32);
}
fn gen_load(&mut self, addr: u64) {
self.ops.push(OP::Load);
self.gen_i64(addr as i64)
}
fn gen_store(&mut self, addr: u64) {
self.ops.push(OP::Store);
self.gen_i64(addr as i64)
}
}
pub fn compile(ast: &Ast) -> Vec<OP> {
let mut compiler = Compiler::new();
compiler.compile(ast);
compiler.into_ops()
}

View File

@ -3,3 +3,5 @@ pub mod parser;
pub mod interpreter;
pub mod token;
pub mod ast;
pub mod bytecode;
pub mod vm;

View File

@ -1,6 +1,6 @@
use std::{env::args, io::Write};
use nek_lang::interpreter::Interpreter;
use nek_lang::{interpreter::Interpreter, lexer::lex, parser::parse, bytecode::compile, vm::Vm};
#[derive(Debug, Default)]
struct CliConfig {
@ -27,7 +27,18 @@ fn main() {
if let Some(file) = &cfg.file {
let code = std::fs::read_to_string(file).expect(&format!("File not found: '{}'", file));
interpreter.run_text(&code, cfg.print_tokens, cfg.print_ast);
let tokens = lex(&code);
let ast = parse(tokens);
let prog = compile(&ast);
// println!("{:?}", prog);
let mut vm = Vm::new(prog);
vm.run();
// interpreter.run_text(&code, cfg.print_tokens, cfg.print_ast);
}
if cfg.interactive || cfg.file.is_none() {

153
src/vm.rs Normal file
View File

@ -0,0 +1,153 @@
use crate::{bytecode::OP, interpreter::Value};
#[derive(Debug, Default)]
pub struct Vm {
prog: Vec<OP>,
ip: usize,
stack: Vec<Value>,
/// This isn't actually a heap. It's actually still more of a f*cked up stack
heap: Vec<Value>,
}
impl Vm {
pub fn new(prog: Vec<OP>) -> Self {
Self {
prog,
..Default::default()
}
}
pub fn run(&mut self) {
while let Some(op) = self.prog.get(self.ip).copied() {
self.ip += 1;
match op {
OP::Push => {
let val = self.read_i64();
self.stack.push(Value::I64(val));
}
OP::Pop => {
self.stack.pop();
}
OP::Load => {
let addr = self.read_i64() as usize;
if let Some(val) = self.heap.get(addr) {
self.stack.push(val.clone());
} else {
panic!("Trying to load from uninitialized heap");
}
}
OP::Store => {
let val = self
.stack
.pop()
.expect("Trying to pop value from stack for storing");
let addr = self.read_i64() as usize;
if self.heap.len() == addr {
self.heap.push(val);
} else {
self.heap[addr] = val;
}
}
OP::Print => {
let val = self
.stack
.pop()
.expect("Trying to pop value from stack for printing");
print!("{}", val);
}
OP::Add => {
let vals = self.pop2_i64();
self.stack.push(Value::I64(vals.0 + vals.1))
}
OP::Subtract => {
let vals = self.pop2_i64();
self.stack.push(Value::I64(vals.0 - vals.1))
}
OP::Multiply => {
let vals = self.pop2_i64();
self.stack.push(Value::I64(vals.0 * vals.1))
}
OP::Divide => {
let vals = self.pop2_i64();
self.stack.push(Value::I64(vals.0 / vals.1))
}
OP::Modulo => {
let vals = self.pop2_i64();
self.stack.push(Value::I64(vals.0 % vals.1))
}
OP::Eq => {
let vals = self.pop2_i64();
self.stack
.push(Value::I64(if vals.0 == vals.1 { 1 } else { 0 }))
}
OP::Neq => {
let vals = self.pop2_i64();
self.stack
.push(Value::I64(if vals.0 != vals.1 { 1 } else { 0 }))
}
OP::Gt => {
let vals = self.pop2_i64();
self.stack
.push(Value::I64(if vals.0 > vals.1 { 1 } else { 0 }))
}
OP::Lt => {
let vals = self.pop2_i64();
self.stack
.push(Value::I64(if vals.0 < vals.1 { 1 } else { 0 }))
}
OP::Value(_) => {
panic!("This is not an instruction, but data. This should never be evaluated")
}
OP::Jump => {
self.ip = self.read_i64() as usize;
}
OP::JumpTrue => {
let jmp_target = self.read_i64() as usize;
if !matches!(self.stack.pop(), Some(Value::I64(0))) {
self.ip = jmp_target;
}
}
OP::JumpFalse => {
let jmp_target = self.read_i64() as usize;
if matches!(self.stack.pop(), Some(Value::I64(0))) {
self.ip = jmp_target;
}
}
}
}
}
fn pop2_i64(&mut self) -> (i64, i64) {
let rhs = self.stack.pop();
let lhs = self.stack.pop();
match (lhs, rhs) {
(Some(Value::I64(lhs)), Some(Value::I64(rhs))) => (lhs, rhs),
_ => panic!("Invalid data for add"),
}
}
fn read_i64(&mut self) -> i64 {
let mut val = if let Some(OP::Value(val)) = self.prog.get(self.ip).copied() {
val
} else {
panic!("Expected Value as next OP")
} as i64;
self.ip += 1;
val |= (if let Some(OP::Value(val)) = self.prog.get(self.ip).copied() {
val
} else {
panic!("Expected Value as next OP")
} as i64)
<< 32;
self.ip += 1;
val
}
}