Files
damn_simple_architecture/compiler/src/backend/dsa/codegen.rs
T

739 lines
27 KiB
Rust
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
use std::collections::HashMap;
use std::sync::atomic::AtomicU32;
use std::time::SystemTime;
use chrono::{DateTime, Local};
use super::registers::RegisterAllocator;
use crate::{block, comment, dsa};
use crate::model::{
BinaryOperator, CompilerError, ConstExpr, Declaration, Dependency, Expression,
Program, Statement, UnaryOperator, Variable,
};
pub struct CodeGenerator {
ast: Program,
imports: HashMap<String, String>,
globals: Vec<String>,
functions: Vec<String>,
symbols: Vec<String>,
allocator: RegisterAllocator,
}
fn import(name: &str, path: &str) -> String {
format!("include {name}: \"{}\"", path)
}
impl CodeGenerator {
const RET: &'static str = "\tjmp _ret";
pub fn new(ast: Program) -> Self {
CodeGenerator {
ast,
imports: HashMap::new(),
globals: Vec::new(),
functions: Vec::new(),
symbols: Vec::new(),
allocator: RegisterAllocator::new(),
}
}
pub fn include(&mut self, name: &str, path: &str) {
self.imports.insert(name.to_string(), path.to_string());
}
fn is_global(&self, name: &str) -> bool {
// Check if this variable is in the globals list
self.globals
.iter()
.any(|g| g.contains(&format!("dw {}:", name)))
}
pub fn generate(&mut self) -> Result<String, CompilerError> {
// always include the print library for debugging!
self.include("print", "./lib/io/print.dsa");
for block in self.ast.clone().declarations {
match block {
Declaration::Variable {
var: Variable { name, .. },
..
} => self.symbols.push(name),
Declaration::Function { name, .. } => self.symbols.push(name),
Declaration::Dependency(Dependency { name, .. }) => {
self.symbols.push(name)
}
}
}
for block in self.ast.clone().declarations {
self.generate_block(block.clone())?;
}
self.generate_layout()
}
fn generate_layout(&mut self) -> Result<String, CompilerError> {
let datetime: DateTime<Local> = SystemTime::now().into();
Ok(dsa![
"",
comment!("GENERATED BY DSC COMPILER"),
comment!(format!(
"Generated at {}",
datetime.format("%Y-%m-%d %H:%M:%S")
)),
"",
// imports
comment!("Imports"),
self.imports
.iter()
.map(|(k, v)| import(k, v))
.collect::<Vec<String>>()
.join("\n"),
"",
// reserved memory
comment!("Globals & Reserved Memory"),
self.globals.join("\n"),
"",
// entry point
comment!("Entry Point"),
"dw stack: 0x10000",
"db message: \"Process Exited with code:\"",
block! [ "_init"
dsa![ldw stack, bpr],
dsa![mov bpr, spr],
dsa![push zero],
dsa![call main],
dsa![call print::print_newline],
dsa![lwi message, rg0],
dsa![push rg0],
dsa![call print::print],
dsa![pop zero],
dsa![call print::print_hex_word],
dsa![pop zero],
dsa![hlt]
],
"",
comment!("Return"),
block! [ "_ret"
dsa![mov bpr, spr],
dsa![pop bpr],
dsa![return]
],
comment!("Compiled Code Starts..."),
// block! [ "main"
// dsa![push bpr],
// dsa![mov spr, bpr],
// dsa![lwi 67, rg1],
// dsa![stw rg1, spr, 8],
// dsa![mov bpr, spr],
// dsa![pop bpr],
// dsa![return]
// ],
self.functions.join("\n"),
])
}
fn generate_global(&mut self, name: &str, init: Option<ConstExpr>) {
self.globals.push(format!(
"dw {}: {}",
name,
init.unwrap_or(ConstExpr::Number(0))
))
}
fn generate_block(&mut self, block: Declaration) -> Result<(), CompilerError> {
match block {
Declaration::Variable { var, init, .. } => {
self.generate_global(&var.name, init)
}
Declaration::Function {
name, params, body, ..
} => {
let func = self.generate_function(&name, &params, &body).join("\n");
self.functions.push(format!("{func}\n"));
}
Declaration::Dependency(Dependency { name, path }) => {
self.imports.insert(name, path);
}
};
Ok(())
}
// Example: Generate code for a function
fn generate_function(
&mut self,
name: &str,
params: &[Variable],
body: &[Statement],
) -> Vec<String> {
let mut code = Vec::new();
// Reset allocator for new function
self.allocator.reset();
// Function prologue
code.push(format!("{}:", name));
code.push("\tpush bpr".to_string());
code.push("\tmov spr, bpr".to_string());
code.push(String::new());
// Allocate parameters to registers or stack locations
for (i, param) in params.iter().enumerate() {
let offset = 8 + (i as i32 * 4); // Parameters start at bpr+8
// Track that this parameter is at a stack location
let (reg, load_code) = self.allocator.alloc_var(&param.name).unwrap();
code.extend(load_code);
code.push(format!("\tldw bpr, {}, {}", reg, offset));
}
// Generate code for function body
for stmt in body {
let stmt_code = self.generate_statement(stmt).unwrap();
code.extend(stmt_code);
}
// automatically return at function end
if let Some(x) = code.last()
&& x == Self::RET
{
} else {
code.push(Self::RET.to_string());
}
code
}
// Example: Generate code for a statement
fn generate_statement(
&mut self,
stmt: &Statement,
) -> Result<Vec<String>, CompilerError> {
let mut code = Vec::new();
match stmt {
Statement::Declaration { var, value } => {
if let Some(expr) = value {
// Evaluate expression
let (result_reg, expr_code) = self.generate_expression(expr, true)?;
code.extend(expr_code);
// Store result in variable
let store_code = self.allocator.store_var(&var.name, &result_reg);
code.extend(store_code);
// Free temporary register
self.allocator.free_temp(&result_reg);
} else {
// Just declaring variable without initialization
self.allocator.alloc_var(&var.name)?;
}
}
Statement::Break => unimplemented!(),
Statement::Continue => unimplemented!(),
Statement::PtrWrite { ptr, value } => {
let (result_reg, expr_code) = self.generate_expression(value, true)?;
code.extend(expr_code);
let (ptr_reg, ptr_code) = self.generate_expression(ptr, true)?;
code.extend(ptr_code);
code.push(format!("\tstw {}, {}", result_reg, ptr_reg));
self.allocator.free_temp(&result_reg);
self.allocator.free_temp(&ptr_reg);
}
Statement::Assign { varname, value } => {
// Evaluate expression
let (result_reg, expr_code) = self.generate_expression(value, true)?;
code.extend(expr_code);
// Check if this is a global variable
if self.is_global(varname) {
// Store to global label
code.push(format!("\tstw {}, {}", result_reg, varname));
} else {
// Store result in local variable
let store_code = self.allocator.store_var(varname, &result_reg);
code.extend(store_code);
}
// Free temporary register
self.allocator.free_temp(&result_reg);
}
Statement::Return(expr) => {
if let Some(e) = expr {
let (result_reg, expr_code) = self.generate_expression(e, true)?;
code.extend(expr_code);
code.push(format!("\tstw {}, bpr, 8", result_reg));
code.push(format!("\tjmp _ret"));
self.allocator.free_temp(&result_reg);
}
}
Statement::If {
condition,
then_stmt,
else_stmt,
} => {
// Generate condition
let (cond_reg, cond_code) = self.generate_expression(condition, true)?;
code.extend(cond_code);
// Compare with zero
code.push(format!("\tcmp {}, zero", cond_reg));
self.allocator.free_temp(&cond_reg);
// Generate unique labels
let then_label = format!("_then_{}", self.get_unique_label());
let else_label = format!("_else_{}", self.get_unique_label());
let end_label = format!("_end_{}", self.get_unique_label());
// Jump to else if condition is false (equal to zero)
code.push(format!("\tjeq {}", else_label));
// Then block
code.push(format!("{}:", then_label));
for s in then_stmt {
code.extend(self.generate_statement(s)?);
}
if then_stmt.len() == 0 {
code.push("\tnop".to_string());
}
code.push(format!("\tjmp {}", end_label));
// Else block
code.push(format!("{}:", else_label));
for s in else_stmt {
code.extend(self.generate_statement(s)?);
}
if else_stmt.len() == 0 {
code.push("\tnop".to_string());
}
code.push(format!("{}:", end_label));
}
Statement::While { condition, body } => {
let loop_start = format!("_while_start_{}", self.get_unique_label());
let loop_end = format!("_while_end_{}", self.get_unique_label());
code.push(format!("{}:", loop_start));
// Generate condition
let (cond_reg, cond_code) = self.generate_expression(condition, true)?;
code.extend(cond_code);
code.push(format!("\tcmp {}, zero", cond_reg));
self.allocator.free_temp(&cond_reg);
code.push(format!("\tjeq {}", loop_end));
// Loop body
for s in body {
code.extend(self.generate_statement(s)?);
}
code.push(format!("\tjmp {}", loop_start));
code.push(format!("{}:", loop_end));
}
Statement::Loop(body) => {
let loop_start = format!("_loop_start_{}", self.get_unique_label());
code.push(format!("{}:", loop_start));
for s in body {
code.extend(self.generate_statement(s)?);
}
code.push(format!("\tjmp {}", loop_start));
}
Statement::Expression { expr } => {
let (result_reg, expr_code) = self.generate_expression(expr, false)?;
code.extend(expr_code);
self.allocator.free_temp(&result_reg);
}
Statement::Block(statements) => {
for s in statements {
code.extend(self.generate_statement(s)?);
}
}
}
Ok(code)
}
// Example: Generate code for an expression
// Returns (register containing result, assembly code)
fn generate_expression(
&mut self,
expr: &Expression,
use_result: bool,
) -> Result<(String, Vec<String>), CompilerError> {
let mut code = Vec::new();
// optimisation to prevent generating dead code!
if expr.is_pure() && !use_result {
return Ok((String::new(), code));
}
match expr {
Expression::StringLiteral(value) => {
let (reg, alloc_code) = self.allocator.alloc_temp()?;
code.extend(alloc_code);
// write string into memory
let uuid = self.get_unique_label();
code.push(format!("\tdb str_{uuid}: \"{value}\""));
// Load pointer to string
code.push(format!("\tlwi str_{uuid}, {reg}"));
Ok((reg, code))
}
Expression::CharLiteral(value) => {
let (reg, alloc_code) = self.allocator.alloc_temp()?;
code.extend(alloc_code);
// Load immediate value
code.push(format!("\tlli {}, {} // '{value}'", *value as u8, reg));
Ok((reg, code))
}
Expression::Number(value) => {
let (reg, alloc_code) = self.allocator.alloc_temp()?;
code.extend(alloc_code);
// Load immediate value
code.push(format!("\tlli {}, {}", value & 0xFFFF, reg));
if *value > 0xFFFF || *value < 0 {
code.push(format!("\tlui {}, {}", (value >> 16) & 0xFFFF, reg));
}
Ok((reg, code))
}
Expression::Variable { name, .. } => {
if self.is_global(&name.name) {
// Allocate a temporary register for the global
let (reg, alloc_code) = self.allocator.alloc_temp()?;
code.extend(alloc_code);
// Load from global label
code.push(format!("\tldw {}, {}", name.name, reg));
Ok((reg, code))
} else {
// Local variable - use existing allocator logic
let (reg, load_code) = self.allocator.load_var(&name.name)?;
code.extend(load_code);
Ok((reg, code))
}
}
Expression::Binary { op, left, right } => {
// Evaluate left operand
let (left_reg, left_code) = self.generate_expression(left, true)?;
code.extend(left_code);
// Evaluate right operand
let (right_reg, right_code) = self.generate_expression(right, true)?;
code.extend(right_code);
// Allocate result register
let (result_reg, result_alloc) = self.allocator.alloc_temp()?;
code.extend(result_alloc);
// Generate operation
match op {
BinaryOperator::Add => {
code.push(format!(
"\tadd {}, {}, {}",
left_reg, right_reg, result_reg
));
}
BinaryOperator::Sub => {
code.push(format!(
"\tsub {}, {}, {}",
left_reg, right_reg, result_reg
));
}
BinaryOperator::Mul => {
self.include("maths", "./lib/maths/core.dsa");
// Call multiply function
code.push(format!("\tpush {}", right_reg));
code.push(format!("\tpush {}", left_reg));
code.push("\tcall maths::multiply".to_string());
code.push(format!("\tpop {}", result_reg));
code.push("\tpop zero".to_string());
}
// Comparison operators - return 1 (true) or 0 (false)
BinaryOperator::Eq => {
code.push(format!("\tcmp {}, {}", left_reg, right_reg));
code.push(format!("\tlli 0, {}", result_reg));
let end_label = format!("_cmp_end_{}", self.get_unique_label());
code.push(format!("\tjne {}", end_label)); // If not equal, skip setting to 1
code.push(format!("\tlli 1, {}", result_reg));
code.push(format!("{}:", end_label));
}
BinaryOperator::Ne => {
code.push(format!("\tcmp {}, {}", left_reg, right_reg));
code.push(format!("\tlli 0, {}", result_reg));
let end_label = format!("_cmp_end_{}", self.get_unique_label());
code.push(format!("\tjeq {}", end_label)); // If equal, skip setting to 1
code.push(format!("\tlli 1, {}", result_reg));
code.push(format!("{}:", end_label));
}
BinaryOperator::Lt => {
code.push(format!("\tcmp {}, {}", left_reg, right_reg));
code.push(format!("\tlli 0, {}", result_reg));
let end_label = format!("_cmp_end_{}", self.get_unique_label());
code.push(format!("\tjge {}", end_label)); // If greater or equal, skip setting to 1
code.push(format!("\tlli 1, {}", result_reg));
code.push(format!("{}:", end_label));
}
BinaryOperator::Le => {
code.push(format!("\tcmp {}, {}", left_reg, right_reg));
code.push(format!("\tlli 0, {}", result_reg));
let end_label = format!("_cmp_end_{}", self.get_unique_label());
code.push(format!("\tjgt {}", end_label)); // If greater than, skip setting to 1
code.push(format!("\tlli 1, {}", result_reg));
code.push(format!("{}:", end_label));
}
BinaryOperator::Gt => {
code.push(format!("\tcmp {}, {}", left_reg, right_reg));
code.push(format!("\tlli 0, {}", result_reg));
let end_label = format!("_cmp_end_{}", self.get_unique_label());
code.push(format!("\tjle {}", end_label)); // If less or equal, skip setting to 1
code.push(format!("\tlli 1, {}", result_reg));
code.push(format!("{}:", end_label));
}
BinaryOperator::Ge => {
code.push(format!("\tcmp {}, {}", left_reg, right_reg));
code.push(format!("\tlli 0, {}", result_reg));
let end_label = format!("_cmp_end_{}", self.get_unique_label());
code.push(format!("\tjlt {}", end_label)); // If less than, skip setting to 1
code.push(format!("\tlli 1, {}", result_reg));
code.push(format!("{}:", end_label));
}
_ => unimplemented!(),
}
// Free operand registers (allocator will protect variables)
self.allocator.free_temp(&left_reg);
self.allocator.free_temp(&right_reg);
Ok((result_reg, code))
}
Expression::Call { name, args } => {
// first evaluate all the args we're going to need
let mut arg_regs = Vec::new();
for arg in args.iter().rev() {
let (arg_reg, arg_code) = self.generate_expression(arg, true)?;
code.extend(arg_code);
arg_regs.push(arg_reg);
}
// Save caller-saved registers and track which ones we saved
// old method, inefficient.
// let saved_regs = self.allocator.get_caller_saved_registers();
// for reg in &saved_regs {
// code.push(format!("\tpush {}", reg));
// }
// Save caller-saved registers and track which ones we saved
let saved_regs = self.allocator.get_caller_saved_registers();
for reg in &saved_regs {
// spill variables to stack
code.extend(self.allocator.spill_register(reg).unwrap());
}
// Evaluate and push arguments in reverse order
for (i, arg_reg) in arg_regs.iter().enumerate() {
code.push(format!(
"\tpush {} // push arg {}",
arg_reg,
args.len() - 1 - i
));
}
// if GLOBAL_METHODS.contains_key(name.name.as_str()) {
// code.push(format!("\tcall {}",
// GLOBAL_METHODS[name.name.as_str()])); } else
if self.symbols.contains(&name.name) {
// Call local function
code.push(format!("\tcall {}", name));
} else if let Some(ns) = name.namespace.clone()
&& self.imports.contains_key(&ns)
{
code.push(format!("\tcall {}", name));
} else {
return Err(CompilerError::Undefined(name.clone()));
}
let result_reg: String;
if use_result {
let (temp_result_reg, result_alloc) = self.allocator.alloc_temp()?;
result_reg = temp_result_reg;
code.extend(result_alloc);
code.push(format!("\tpop {}", result_reg));
// Clean up arguments
if args.len() > 1 {
for _ in 0..(args.len() - 1) {
code.push("\tpop zero".to_string());
}
}
} else {
result_reg = "zero".to_string();
// Clean up arguments
if args.len() > 0 {
for _ in 0..(args.len()) {
code.push("\tpop zero".to_string());
}
}
}
// Restore caller-saved registers in reverse order (LIFO)
// for reg in saved_regs.iter().rev() {
// code.push(format!("\tpop {}", reg));
// }
// Free argument registers
for reg in arg_regs {
self.allocator.free_temp(&reg);
}
Ok((result_reg, code))
}
Expression::Unary { op, operand } => {
let (operand_reg, operand_code) =
self.generate_expression(operand, true)?;
code.extend(operand_code);
let (result_reg, result_alloc) = self.allocator.alloc_temp()?;
code.extend(result_alloc);
match op {
UnaryOperator::Minus => {
// Negate: result = 0 - operand
code.push(format!("\tsub zero, {}, {}", operand_reg, result_reg));
}
UnaryOperator::Plus => {
// Just move
code.push(format!("\tmov {}, {}", operand_reg, result_reg));
}
UnaryOperator::Dereference => {
code.push(format!("\tldw {}, {}", operand_reg, result_reg));
}
UnaryOperator::Reference => {
code.extend(self.allocator.spill_register(&operand_reg)?);
code.push(format!(
"\tsubi bpr {} {}",
-(4 + self.allocator.get_stack_offset()),
result_reg
))
}
}
self.allocator.free_temp(&operand_reg);
Ok((result_reg, code))
}
Expression::Empty => Ok(("zero".to_string(), code)),
}
}
// Helper for generating unique labels
fn get_unique_label(&mut self) -> String {
// You'd implement a counter here
static COUNTER: AtomicU32 = AtomicU32::new(0);
let val = COUNTER.fetch_add(1, std::sync::atomic::Ordering::SeqCst);
(val + 1).to_string()
}
}
/// Build a single string from any number of arguments.
/// Each argument must implement `Display` or be convertible to a string.
#[macro_export]
macro_rules! dsa {
($($arg:expr),* $(,)?) => {{
// Start with an empty String well grow it as we go.
use std::fmt::Write;
let mut s = ::std::string::String::new();
$(
// `write!` is cheaper than `format!` for each element
// because it reuses the same buffer.
write!(s, "{}\n", $arg).expect("write to String failed");
)*
s
}};
}
// ──────────────────────── dsa! ────────────────────────
// A tiny helper that just turns its tokenstream into a string.
// The trailing comma is kept its part of the syntax you want.
#[macro_export]
macro_rules! cmd {
($($tokens:tt)*) => {{
// Well just stringify the tokens and return a String.
format!("{}", concat!(stringify!($tokens), "\n"))
}};
}
// ──────────────────────── block! ────────────────────────
// Usage:
//
// let asm = block![ "name"
// dsa![mov rg0, rg1],
// dsa![add rg1, rg1]
// ];
//
// `asm` is a `&'static str` containing:
//
// name:
// mov rg0, rg1
// add rg1, rg1
//
#[macro_export]
macro_rules! block {
// The first token must be a string literal thats the label.
($label:literal $(dsa![$($ins:tt)*]),* ) => {{
// Build a single string at compile time.
const CODE: &str = concat!(
$label, ":\n",
// Each `dsa!` call yields a string like `"mov rg0, rg1"`.
// We add a newline after each one to get the desired layout.
$(concat!("\t", stringify!($($ins)*), "\n")),*
);
CODE
}};
}
#[macro_export]
macro_rules! comment {
($text:expr) => {{ format!("// {}", $text) }};
}