holey-bytes/lang/src/lib.rs

1698 lines
50 KiB
Rust
Raw Normal View History

2024-07-08 03:13:50 -05:00
#![feature(
assert_matches,
2024-07-08 03:13:50 -05:00
let_chains,
if_let_guard,
macro_metavar_expr,
anonymous_lifetime_in_impl_trait,
core_intrinsics,
never_type,
unwrap_infallible,
slice_partition_dedup,
portable_simd,
iter_collect_into,
2024-09-13 12:30:47 -05:00
new_uninit,
2024-07-08 03:13:50 -05:00
ptr_metadata,
2024-09-03 15:34:17 -05:00
slice_ptr_get,
slice_take,
2024-09-06 09:11:57 -05:00
map_try_insert,
2024-09-12 11:42:21 -05:00
extract_if,
2024-09-30 12:09:17 -05:00
ptr_internals,
2024-10-01 08:28:18 -05:00
iter_intersperse,
2024-10-10 01:35:17 -05:00
str_from_raw_parts,
ptr_sub_ptr,
2024-10-12 06:07:49 -05:00
slice_from_ptr_range,
is_sorted,
2024-10-27 07:57:00 -05:00
iter_next_chunk,
2024-10-30 12:42:25 -05:00
pointer_is_aligned_to,
maybe_uninit_fill
2024-07-08 03:13:50 -05:00
)]
2024-09-30 12:09:17 -05:00
#![warn(clippy::dbg_macro)]
2024-10-27 07:57:00 -05:00
#![expect(stable_features, internal_features)]
2024-09-30 12:09:17 -05:00
#![no_std]
2024-05-17 12:53:59 -05:00
2024-09-30 12:09:17 -05:00
#[cfg(feature = "std")]
pub use fs::*;
2024-07-08 00:22:53 -05:00
use {
2024-09-13 08:12:20 -05:00
self::{
lexer::TokenKind,
parser::{idfl, CommentOr, Expr, ExprRef, FileId, Pos},
2024-09-13 08:12:20 -05:00
ty::ArrayLen,
},
2024-10-27 08:29:14 -05:00
alloc::{string::String, vec::Vec},
core::{cell::Cell, fmt::Display, ops::Range},
2024-09-30 12:09:17 -05:00
hashbrown::hash_map,
2024-05-20 07:11:58 -05:00
};
2024-09-30 12:09:17 -05:00
#[macro_use]
extern crate alloc;
#[cfg(any(feature = "std", test))]
extern crate std;
2024-09-28 09:34:08 -05:00
#[cfg(test)]
const README: &str = include_str!("../README.md");
2024-05-09 16:41:59 -05:00
#[macro_export]
macro_rules! run_tests {
2024-09-28 09:34:08 -05:00
($runner:path: $($name:ident;)*) => {$(
2024-05-09 16:41:59 -05:00
#[test]
fn $name() {
2024-09-30 12:09:17 -05:00
$crate::run_test(core::any::type_name_of_val(&$name), stringify!($name), $crate::README, $runner);
2024-05-09 16:41:59 -05:00
}
)*};
}
2024-10-04 14:44:29 -05:00
pub mod fmt;
2024-09-30 12:09:17 -05:00
#[cfg(any(feature = "std", test))]
pub mod fs;
2024-10-27 07:57:00 -05:00
pub mod fuzz;
pub mod lexer;
2024-06-23 06:55:48 -05:00
pub mod parser;
2024-10-31 08:56:55 -05:00
pub mod regalloc;
2024-09-02 17:07:20 -05:00
pub mod son;
2024-06-23 06:55:48 -05:00
2024-10-30 12:42:25 -05:00
mod utils;
2024-06-23 06:55:48 -05:00
2024-10-24 06:25:30 -05:00
mod debug {
pub fn panicking() -> bool {
#[cfg(feature = "std")]
{
std::thread::panicking()
}
#[cfg(not(feature = "std"))]
{
false
}
}
#[cfg(all(debug_assertions, feature = "std"))]
pub type Trace = std::rc::Rc<std::backtrace::Backtrace>;
#[cfg(not(all(debug_assertions, feature = "std")))]
pub type Trace = ();
pub fn trace() -> Trace {
#[cfg(all(debug_assertions, feature = "std"))]
{
std::rc::Rc::new(std::backtrace::Backtrace::capture())
}
#[cfg(not(all(debug_assertions, feature = "std")))]
{}
}
}
2024-10-10 01:35:17 -05:00
pub mod reg {
pub const STACK_PTR: Reg = 254;
pub const ZERO: Reg = 0;
pub const RET: Reg = 1;
pub const RET_ADDR: Reg = 31;
pub type Reg = u8;
}
2024-10-01 14:33:30 -05:00
mod ctx_map {
use core::hash::BuildHasher;
pub type Hash = u64;
pub type HashBuilder = core::hash::BuildHasherDefault<IdentityHasher>;
#[derive(Default)]
pub struct IdentityHasher(u64);
impl core::hash::Hasher for IdentityHasher {
fn finish(&self) -> u64 {
self.0
}
fn write(&mut self, _: &[u8]) {
unimplemented!()
}
fn write_u64(&mut self, i: u64) {
self.0 = i;
}
}
2024-10-21 11:57:23 -05:00
#[derive(Clone)]
2024-10-01 14:33:30 -05:00
pub struct Key<T> {
pub value: T,
pub hash: Hash,
}
impl<T> core::hash::Hash for Key<T> {
fn hash<H: core::hash::Hasher>(&self, state: &mut H) {
state.write_u64(self.hash);
}
}
pub trait CtxEntry {
type Ctx: ?Sized;
type Key<'a>: Eq + core::hash::Hash;
fn key<'a>(&self, ctx: &'a Self::Ctx) -> Self::Key<'a>;
}
2024-10-21 11:57:23 -05:00
#[derive(Clone)]
2024-10-01 14:33:30 -05:00
pub struct CtxMap<T> {
inner: hashbrown::HashMap<Key<T>, (), HashBuilder>,
}
impl<T> Default for CtxMap<T> {
fn default() -> Self {
Self { inner: Default::default() }
}
}
impl<T: CtxEntry> CtxMap<T> {
pub fn entry<'a, 'b>(
&'a mut self,
2024-10-01 15:53:03 -05:00
key: T::Key<'b>,
2024-10-01 14:33:30 -05:00
ctx: &'b T::Ctx,
) -> (hashbrown::hash_map::RawEntryMut<'a, Key<T>, (), HashBuilder>, Hash) {
2024-10-01 15:53:03 -05:00
let hash = crate::FnvBuildHasher::default().hash_one(&key);
(self.inner.raw_entry_mut().from_hash(hash, |k| k.value.key(ctx) == key), hash)
2024-10-01 14:33:30 -05:00
}
2024-10-01 15:53:03 -05:00
pub fn get<'a>(&self, key: T::Key<'a>, ctx: &'a T::Ctx) -> Option<&T> {
let hash = crate::FnvBuildHasher::default().hash_one(&key);
2024-10-01 14:33:30 -05:00
self.inner
.raw_entry()
2024-10-01 15:53:03 -05:00
.from_hash(hash, |k| k.value.key(ctx) == key)
2024-10-01 14:33:30 -05:00
.map(|(k, _)| &k.value)
}
pub fn clear(&mut self) {
self.inner.clear();
}
2024-10-01 15:53:03 -05:00
pub fn remove(&mut self, value: &T, ctx: &T::Ctx) -> Option<T> {
2024-10-01 14:33:30 -05:00
let (entry, _) = self.entry(value.key(ctx), ctx);
match entry {
hashbrown::hash_map::RawEntryMut::Occupied(o) => Some(o.remove_entry().0.value),
hashbrown::hash_map::RawEntryMut::Vacant(_) => None,
}
}
2024-10-01 15:53:03 -05:00
pub fn insert<'a>(&mut self, key: T::Key<'a>, value: T, ctx: &'a T::Ctx) {
let (entry, hash) = self.entry(key, ctx);
match entry {
hashbrown::hash_map::RawEntryMut::Occupied(_) => unreachable!(),
hashbrown::hash_map::RawEntryMut::Vacant(v) => {
_ = v.insert(Key { hash, value }, ())
}
}
}
pub fn get_or_insert<'a>(
&mut self,
key: T::Key<'a>,
ctx: &'a mut T::Ctx,
with: impl FnOnce(&'a mut T::Ctx) -> T,
) -> &mut T {
let (entry, hash) = self.entry(key, unsafe { &mut *(&mut *ctx as *mut _) });
match entry {
hashbrown::hash_map::RawEntryMut::Occupied(o) => &mut o.into_key_value().0.value,
hashbrown::hash_map::RawEntryMut::Vacant(v) => {
&mut v.insert(Key { hash, value: with(ctx) }, ()).0.value
}
}
}
2024-10-01 14:33:30 -05:00
}
}
mod task {
use super::Offset;
pub fn unpack(offset: Offset) -> Result<Offset, usize> {
if offset >> 31 != 0 {
Err((offset & !(1 << 31)) as usize)
} else {
Ok(offset)
}
}
pub fn is_done(offset: Offset) -> bool {
unpack(offset).is_ok()
}
pub fn id(index: usize) -> Offset {
1 << 31 | index as u32
}
}
2024-10-27 07:57:00 -05:00
#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Default, Debug)]
pub struct Ident(u32);
2024-10-27 07:57:00 -05:00
impl Ident {
pub const INVALID: Self = Self(u32::MAX);
const LEN_BITS: u32 = 6;
2024-10-27 07:57:00 -05:00
pub fn len(self) -> u32 {
self.0 & ((1 << Self::LEN_BITS) - 1)
}
pub fn is_empty(self) -> bool {
self.len() == 0
}
2024-10-27 07:57:00 -05:00
pub fn is_null(self) -> bool {
(self.0 >> Self::LEN_BITS) == 0
}
2024-10-27 07:57:00 -05:00
pub fn pos(self) -> u32 {
(self.0 >> Self::LEN_BITS).saturating_sub(1)
}
2024-10-27 07:57:00 -05:00
pub fn new(pos: u32, len: u32) -> Option<Self> {
(len < (1 << Self::LEN_BITS)).then_some(((pos + 1) << Self::LEN_BITS) | len).map(Self)
}
2024-10-27 07:57:00 -05:00
pub fn range(self) -> core::ops::Range<usize> {
let (len, pos) = (self.len() as usize, self.pos() as usize);
pos..pos + len
}
2024-10-27 07:57:00 -05:00
fn builtin(builtin: u32) -> Ident {
debug_assert!(Self(builtin).is_null());
Self(builtin)
}
}
2024-09-13 08:12:20 -05:00
mod ty {
use {
crate::{
lexer::TokenKind,
2024-10-27 07:57:00 -05:00
parser::{self, FileId, Pos},
Ident, Size, Types,
2024-09-13 08:12:20 -05:00
},
2024-10-24 06:25:30 -05:00
core::{num::NonZeroU32, ops::Range},
2024-09-13 08:12:20 -05:00
};
pub type ArrayLen = u32;
pub type Builtin = u32;
pub type Struct = u32;
2024-10-30 14:20:03 -05:00
pub type Opt = u32;
2024-09-13 08:12:20 -05:00
pub type Ptr = u32;
pub type Func = u32;
pub type Global = u32;
pub type Module = u32;
pub type Slice = u32;
2024-10-21 11:57:23 -05:00
pub const ECA: Func = Func::MAX;
2024-10-24 02:43:07 -05:00
#[derive(Clone, Copy, PartialEq, Eq, Hash, Debug, Default, PartialOrd, Ord)]
2024-09-13 08:12:20 -05:00
pub struct Tuple(pub u32);
impl Tuple {
const LEN_BITS: u32 = 5;
const LEN_MASK: usize = Self::MAX_LEN - 1;
const MAX_LEN: usize = 1 << Self::LEN_BITS;
pub fn new(pos: usize, len: usize) -> Option<Self> {
if len >= Self::MAX_LEN {
return None;
}
Some(Self((pos << Self::LEN_BITS | len) as u32))
}
pub fn range(self) -> Range<usize> {
let start = self.0 as usize >> Self::LEN_BITS;
start..start + self.len()
}
pub fn len(self) -> usize {
self.0 as usize & Self::LEN_MASK
}
pub fn empty() -> Self {
Self(0)
}
2024-10-24 02:43:07 -05:00
pub fn args(self) -> ArgIter {
ArgIter(self.range())
}
}
pub struct ArgIter(Range<usize>);
pub enum Arg {
Type(Id),
Value(Id),
}
impl ArgIter {
pub(crate) fn next(&mut self, tys: &Types) -> Option<Arg> {
let ty = tys.ins.args[self.0.next()?];
if ty == Id::TYPE {
return Some(Arg::Type(tys.ins.args[self.0.next().unwrap()]));
}
Some(Arg::Value(ty))
}
pub(crate) fn next_value(&mut self, tys: &Types) -> Option<Id> {
loop {
match self.next(tys)? {
Arg::Type(_) => continue,
Arg::Value(id) => break Some(id),
}
}
}
2024-09-13 08:12:20 -05:00
}
#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug, Hash)]
pub struct Id(NonZeroU32);
2024-10-01 14:33:30 -05:00
impl crate::ctx_map::CtxEntry for Id {
2024-10-01 15:53:03 -05:00
type Ctx = crate::TypeIns;
2024-10-01 14:33:30 -05:00
type Key<'a> = crate::SymKey<'a>;
fn key<'a>(&self, ctx: &'a Self::Ctx) -> Self::Key<'a> {
2024-10-01 15:53:03 -05:00
match self.expand() {
Kind::Struct(s) => {
let st = &ctx.structs[s as usize];
debug_assert_ne!(st.pos, Pos::MAX);
2024-10-22 03:17:16 -05:00
crate::SymKey::Struct(st.file, st.pos, st.captures)
2024-10-01 15:53:03 -05:00
}
Kind::Ptr(p) => crate::SymKey::Pointer(&ctx.ptrs[p as usize]),
2024-10-30 14:20:03 -05:00
Kind::Opt(p) => crate::SymKey::Optional(&ctx.opts[p as usize]),
2024-10-01 15:53:03 -05:00
Kind::Func(f) => {
let fc = &ctx.funcs[f as usize];
if let Some(base) = fc.base {
crate::SymKey::FuncInst(base, fc.sig.unwrap().args)
} else {
crate::SymKey::Decl(fc.file, fc.name)
}
}
Kind::Global(g) => {
let gb = &ctx.globals[g as usize];
crate::SymKey::Decl(gb.file, gb.name)
}
2024-10-19 12:37:02 -05:00
Kind::Slice(s) => crate::SymKey::Array(&ctx.slices[s as usize]),
2024-10-27 07:57:00 -05:00
Kind::Module(_) | Kind::Builtin(_) => {
crate::SymKey::Decl(FileId::MAX, Ident::INVALID)
}
2024-10-01 15:53:03 -05:00
}
2024-10-01 14:33:30 -05:00
}
}
2024-09-13 08:12:20 -05:00
impl Default for Id {
fn default() -> Self {
Self(unsafe { NonZeroU32::new_unchecked(UNDECLARED) })
}
}
impl Id {
2024-10-25 07:51:33 -05:00
pub const DEFAULT_INT: Self = Self::UINT;
2024-10-27 07:57:00 -05:00
pub fn bin_ret(self, op: TokenKind) -> Id {
use TokenKind as T;
match op {
T::Lt | T::Gt | T::Le | T::Ge | T::Ne | T::Eq => BOOL.into(),
_ => self,
}
}
2024-10-29 07:36:12 -05:00
pub fn is_float(self) -> bool {
matches!(self.repr(), F32 | F64) || self.is_never()
}
2024-09-13 08:12:20 -05:00
pub fn is_signed(self) -> bool {
2024-10-27 07:57:00 -05:00
matches!(self.repr(), I8..=INT) || self.is_never()
2024-09-13 08:12:20 -05:00
}
pub fn is_unsigned(self) -> bool {
2024-10-27 07:57:00 -05:00
matches!(self.repr(), U8..=UINT) || self.is_never()
2024-09-13 08:12:20 -05:00
}
pub fn is_integer(self) -> bool {
2024-10-27 07:57:00 -05:00
matches!(self.repr(), U8..=INT) || self.is_never()
2024-10-25 08:14:32 -05:00
}
pub fn is_never(self) -> bool {
self == Self::NEVER
2024-09-13 08:12:20 -05:00
}
pub fn strip_pointer(self) -> Self {
match self.expand() {
Kind::Ptr(_) => Kind::Builtin(UINT).compress(),
_ => self,
}
}
pub fn is_pointer(self) -> bool {
2024-10-25 08:14:32 -05:00
matches!(self.expand(), Kind::Ptr(_)) || self.is_never()
2024-09-13 08:12:20 -05:00
}
2024-10-30 14:20:03 -05:00
pub fn is_optional(self) -> bool {
matches!(self.expand(), Kind::Opt(_)) || self.is_never()
}
2024-10-27 07:57:00 -05:00
pub fn try_upcast(self, ob: Self) -> Option<Self> {
2024-09-13 08:12:20 -05:00
let (oa, ob) = (Self(self.0.min(ob.0)), Self(self.0.max(ob.0)));
let (a, b) = (oa.strip_pointer(), ob.strip_pointer());
Some(match () {
2024-09-15 13:14:56 -05:00
_ if oa == Self::from(NEVER) => ob,
_ if ob == Self::from(NEVER) => oa,
2024-09-13 08:12:20 -05:00
_ if oa == ob => oa,
2024-10-31 04:36:18 -05:00
_ if ob.is_optional() => ob,
2024-09-13 08:12:20 -05:00
_ if oa.is_pointer() && ob.is_pointer() => return None,
_ if a.is_signed() && b.is_signed() || a.is_unsigned() && b.is_unsigned() => ob,
_ if a.is_unsigned() && b.is_signed() && a.repr() - U8 < b.repr() - I8 => ob,
2024-10-25 08:45:00 -05:00
_ if a.is_unsigned() && b.is_signed() && a.repr() - U8 > b.repr() - I8 => oa,
2024-10-27 07:57:00 -05:00
_ if oa.is_integer() && ob.is_pointer() => ob,
2024-09-13 08:12:20 -05:00
_ => return None,
})
}
pub fn expand(self) -> Kind {
Kind::from_ty(self)
}
pub const fn repr(self) -> u32 {
self.0.get()
}
pub(crate) fn simple_size(&self) -> Option<Size> {
Some(match self.expand() {
Kind::Ptr(_) => 8,
Kind::Builtin(VOID) => 0,
Kind::Builtin(NEVER) => 0,
2024-10-29 07:36:12 -05:00
Kind::Builtin(INT | UINT | F64) => 8,
Kind::Builtin(I32 | U32 | TYPE | F32) => 4,
Kind::Builtin(I16 | U16) => 2,
Kind::Builtin(I8 | U8 | BOOL) => 1,
_ => return None,
})
}
pub(crate) fn extend(self) -> Self {
if self.is_signed() {
Self::INT
} else if self.is_pointer() {
self
} else {
Self::UINT
}
}
2024-10-24 02:43:07 -05:00
pub(crate) fn loc(&self, tys: &Types) -> Loc {
match self.expand() {
2024-10-30 14:20:03 -05:00
Kind::Opt(o)
if let ty = tys.ins.opts[o as usize].base
&& ty.loc(tys) == Loc::Reg
&& (ty.is_pointer() || tys.size_of(ty) < 8) =>
{
Loc::Reg
}
Kind::Ptr(_) | Kind::Builtin(_) => Loc::Reg,
Kind::Struct(_) if tys.size_of(*self) == 0 => Loc::Reg,
2024-10-30 14:20:03 -05:00
Kind::Struct(_) | Kind::Slice(_) | Kind::Opt(_) => Loc::Stack,
Kind::Func(_) | Kind::Global(_) | Kind::Module(_) => unreachable!(),
}
}
2024-10-24 05:28:18 -05:00
pub(crate) fn has_pointers(&self, tys: &Types) -> bool {
match self.expand() {
Kind::Struct(s) => tys.struct_fields(s).iter().any(|f| f.ty.has_pointers(tys)),
Kind::Ptr(_) => true,
Kind::Slice(s) => tys.ins.slices[s as usize].len == ArrayLen::MAX,
_ => false,
}
}
}
#[derive(PartialEq, Eq, Clone, Copy)]
pub enum Loc {
Reg,
Stack,
2024-09-13 08:12:20 -05:00
}
impl From<u64> for Id {
fn from(id: u64) -> Self {
Self(unsafe { NonZeroU32::new_unchecked(id as _) })
}
}
impl From<u32> for Id {
fn from(id: u32) -> Self {
Kind::Builtin(id).compress()
}
}
const fn array_to_lower_case<const N: usize>(array: [u8; N]) -> [u8; N] {
let mut result = [0; N];
let mut i = 0;
while i < N {
result[i] = array[i].to_ascii_lowercase();
i += 1;
}
result
}
// const string to lower case
macro_rules! builtin_type {
($($name:ident;)*) => {
$(pub const $name: Builtin = ${index(0)} + 1;)*
mod __lc_names {
use super::*;
2024-10-10 01:35:17 -05:00
$(pub const $name: &str = unsafe {
const LCL: &[u8] = unsafe {
&array_to_lower_case(
*(stringify!($name).as_ptr() as *const [u8; stringify!($name).len()])
)
};
core::str::from_utf8_unchecked(LCL)
};)*
2024-09-13 08:12:20 -05:00
}
2024-09-30 12:09:17 -05:00
impl Id {
$(pub const $name: Self = Kind::Builtin($name).compress();)*
}
2024-09-13 08:12:20 -05:00
pub fn from_str(name: &str) -> Option<Builtin> {
2024-10-10 01:35:17 -05:00
match name {
2024-09-13 08:12:20 -05:00
$(__lc_names::$name => Some($name),)*
_ => None,
}
}
pub fn to_str(ty: Builtin) -> &'static str {
match ty {
2024-10-10 01:35:17 -05:00
$($name => __lc_names::$name,)*
2024-09-13 08:12:20 -05:00
v => unreachable!("invalid type: {}", v),
}
}
};
}
builtin_type! {
UNDECLARED;
2024-10-19 03:17:36 -05:00
LEFT_UNREACHABLE;
RIGHT_UNREACHABLE;
2024-09-13 08:12:20 -05:00
NEVER;
VOID;
TYPE;
BOOL;
U8;
U16;
U32;
UINT;
I8;
I16;
I32;
INT;
2024-10-29 07:36:12 -05:00
F32;
F64;
2024-09-13 08:12:20 -05:00
}
macro_rules! type_kind {
($(#[$meta:meta])* $vis:vis enum $name:ident {$( $variant:ident, )*}) => {
$(#[$meta])*
$vis enum $name {
$($variant($variant),)*
}
impl $name {
const FLAG_BITS: u32 = (${count($variant)} as u32).next_power_of_two().ilog2();
2024-09-30 12:09:17 -05:00
const FLAG_OFFSET: u32 = core::mem::size_of::<Id>() as u32 * 8 - Self::FLAG_BITS;
2024-09-13 08:12:20 -05:00
const INDEX_MASK: u32 = (1 << (32 - Self::FLAG_BITS)) - 1;
$vis fn from_ty(ty: Id) -> Self {
let (flag, index) = (ty.repr() >> Self::FLAG_OFFSET, ty.repr() & Self::INDEX_MASK);
match flag {
$(${index(0)} => Self::$variant(index),)*
i => unreachable!("{i}"),
}
}
$vis const fn compress(self) -> Id {
let (index, flag) = match self {
$(Self::$variant(index) => (index, ${index(0)}),)*
};
Id(unsafe { NonZeroU32::new_unchecked((flag << Self::FLAG_OFFSET) | index) })
}
$vis const fn inner(self) -> u32 {
match self {
$(Self::$variant(index) => index,)*
}
}
}
};
}
type_kind! {
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Kind {
Builtin,
Struct,
Ptr,
2024-10-31 04:36:18 -05:00
Slice,
Opt,
2024-09-13 08:12:20 -05:00
Func,
Global,
Module,
}
}
impl Default for Kind {
fn default() -> Self {
Self::Builtin(UNDECLARED)
}
}
pub struct Display<'a> {
tys: &'a super::Types,
files: &'a [parser::Ast],
ty: Id,
}
impl<'a> Display<'a> {
2024-10-20 03:37:48 -05:00
pub fn new(tys: &'a super::Types, files: &'a [parser::Ast], ty: Id) -> Self {
2024-09-13 08:12:20 -05:00
Self { tys, files, ty }
}
2024-10-20 03:37:48 -05:00
pub fn rety(&self, ty: Id) -> Self {
2024-09-13 08:12:20 -05:00
Self::new(self.tys, self.files, ty)
}
}
2024-10-12 08:04:58 -05:00
impl core::fmt::Display for Display<'_> {
2024-09-30 12:09:17 -05:00
fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
2024-09-13 08:12:20 -05:00
use Kind as TK;
match TK::from_ty(self.ty) {
2024-10-12 06:07:49 -05:00
TK::Module(idx) => {
f.write_str("@use(\"")?;
self.files[idx as usize].path.fmt(f)?;
f.write_str(")[")?;
idx.fmt(f)?;
f.write_str("]")
2024-09-13 08:12:20 -05:00
}
2024-10-12 06:07:49 -05:00
TK::Builtin(ty) => f.write_str(to_str(ty)),
2024-10-30 14:20:03 -05:00
TK::Opt(ty) => {
f.write_str("?")?;
self.rety(self.tys.ins.opts[ty as usize].base).fmt(f)
}
2024-10-13 06:30:00 -05:00
TK::Ptr(ty) => {
f.write_str("^")?;
self.rety(self.tys.ins.ptrs[ty as usize].base).fmt(f)
}
2024-09-13 08:12:20 -05:00
TK::Struct(idx) => {
2024-10-01 14:36:23 -05:00
let record = &self.tys.ins.structs[idx as usize];
2024-10-27 07:57:00 -05:00
if record.name.is_null() {
2024-10-12 06:07:49 -05:00
f.write_str("[")?;
idx.fmt(f)?;
f.write_str("]{")?;
2024-09-30 15:17:54 -05:00
for (i, &super::Field { name, ty }) in
2024-09-30 12:09:17 -05:00
self.tys.struct_fields(idx).iter().enumerate()
{
if i != 0 {
2024-10-12 06:07:49 -05:00
f.write_str(", ")?;
}
2024-10-12 06:07:49 -05:00
f.write_str(self.tys.names.ident_str(name))?;
f.write_str(": ")?;
self.rety(ty).fmt(f)?;
2024-09-13 08:12:20 -05:00
}
2024-10-12 06:07:49 -05:00
f.write_str("}")
} else {
let file = &self.files[record.file as usize];
2024-10-12 06:07:49 -05:00
f.write_str(file.ident_str(record.name))
2024-09-13 08:12:20 -05:00
}
}
2024-10-12 06:07:49 -05:00
TK::Func(idx) => {
f.write_str("fn")?;
idx.fmt(f)
}
TK::Global(idx) => {
f.write_str("global")?;
idx.fmt(f)
}
2024-09-13 08:12:20 -05:00
TK::Slice(idx) => {
2024-10-19 12:37:02 -05:00
let array = self.tys.ins.slices[idx as usize];
2024-10-12 06:07:49 -05:00
f.write_str("[")?;
2024-10-19 12:37:02 -05:00
self.rety(array.elem).fmt(f)?;
2024-10-12 06:07:49 -05:00
if array.len != ArrayLen::MAX {
f.write_str("; ")?;
array.len.fmt(f)?;
2024-09-13 08:12:20 -05:00
}
2024-10-12 06:07:49 -05:00
f.write_str("]")
2024-09-13 08:12:20 -05:00
}
}
}
}
}
type Offset = u32;
type Size = u32;
2024-10-01 15:53:03 -05:00
#[derive(PartialEq, Eq, Hash, Clone, Copy)]
pub enum SymKey<'a> {
2024-10-01 14:33:30 -05:00
Pointer(&'a Ptr),
2024-10-30 14:20:03 -05:00
Optional(&'a Opt),
2024-10-22 03:17:16 -05:00
Struct(FileId, Pos, ty::Tuple),
FuncInst(ty::Func, ty::Tuple),
Decl(FileId, Ident),
2024-10-01 14:33:30 -05:00
Array(&'a Array),
2024-09-13 08:12:20 -05:00
}
#[derive(Clone, Copy)]
struct Sig {
args: ty::Tuple,
ret: ty::Id,
}
struct Func {
file: FileId,
2024-10-01 15:53:03 -05:00
name: Ident,
base: Option<ty::Func>,
2024-09-13 08:12:20 -05:00
expr: ExprRef,
sig: Option<Sig>,
offset: Offset,
// TODO: change to indices into common vec
relocs: Vec<TypedReloc>,
code: Vec<u8>,
}
impl Default for Func {
fn default() -> Self {
Self {
file: u32::MAX,
2024-10-27 07:57:00 -05:00
name: Default::default(),
2024-10-01 15:53:03 -05:00
base: None,
2024-09-13 08:12:20 -05:00
expr: Default::default(),
sig: None,
offset: u32::MAX,
relocs: Default::default(),
code: Default::default(),
}
}
}
2024-10-21 11:57:23 -05:00
#[derive(Clone, Copy)]
2024-09-13 08:12:20 -05:00
struct TypedReloc {
target: ty::Id,
reloc: Reloc,
}
2024-10-13 08:33:57 -05:00
#[derive(Clone)]
2024-09-13 08:12:20 -05:00
struct Global {
file: FileId,
name: Ident,
ty: ty::Id,
offset: Offset,
data: Vec<u8>,
}
impl Default for Global {
fn default() -> Self {
Self {
ty: Default::default(),
offset: u32::MAX,
data: Default::default(),
file: u32::MAX,
2024-10-27 07:57:00 -05:00
name: Default::default(),
2024-09-13 08:12:20 -05:00
}
}
}
// TODO: make into bit struct (width: u2, sub_offset: u3, offset: u27)
#[derive(Clone, Copy, Debug)]
struct Reloc {
offset: Offset,
sub_offset: u8,
width: u8,
}
impl Reloc {
fn new(offset: usize, sub_offset: u8, width: u8) -> Self {
Self { offset: offset as u32, sub_offset, width }
}
fn apply_jump(mut self, code: &mut [u8], to: u32, from: u32) -> i64 {
self.offset += from;
let offset = to as i64 - self.offset as i64;
self.write_offset(code, offset);
offset
}
fn write_offset(&self, code: &mut [u8], offset: i64) {
let bytes = offset.to_ne_bytes();
let slice = &mut code[self.offset as usize + self.sub_offset as usize..];
slice[..self.width as usize].copy_from_slice(&bytes[..self.width as usize]);
}
}
struct Field {
2024-09-30 12:09:17 -05:00
name: Ident,
2024-09-13 08:12:20 -05:00
ty: ty::Id,
}
2024-09-30 12:09:17 -05:00
#[derive(Default)]
2024-09-13 08:12:20 -05:00
struct Struct {
name: Ident,
2024-10-01 15:53:03 -05:00
pos: Pos,
file: FileId,
2024-09-30 12:09:17 -05:00
size: Cell<Size>,
align: Cell<u8>,
2024-10-22 03:17:16 -05:00
captures: ty::Tuple,
2024-09-30 12:09:17 -05:00
explicit_alignment: Option<u8>,
field_start: u32,
2024-09-13 08:12:20 -05:00
}
2024-10-30 14:20:03 -05:00
#[derive(PartialEq, Eq, Hash, Clone, Copy)]
pub struct Opt {
base: ty::Id,
}
#[derive(PartialEq, Eq, Hash, Clone, Copy)]
2024-10-01 15:53:03 -05:00
pub struct Ptr {
2024-09-13 08:12:20 -05:00
base: ty::Id,
}
2024-10-01 14:33:30 -05:00
#[derive(Clone, Copy, PartialEq, Eq, Hash)]
2024-10-01 15:53:03 -05:00
pub struct Array {
2024-10-19 12:37:02 -05:00
elem: ty::Id,
2024-09-13 08:12:20 -05:00
len: ArrayLen,
}
2024-10-19 12:37:02 -05:00
impl Array {
fn len(&self) -> Option<usize> {
(self.len != ArrayLen::MAX).then_some(self.len as usize)
}
}
2024-09-13 08:12:20 -05:00
enum PLoc {
Reg(u8, u16),
WideReg(u8, u16),
Ref(u8, u32),
}
2024-09-13 08:12:20 -05:00
struct ParamAlloc(Range<u8>);
impl ParamAlloc {
2024-10-24 02:43:07 -05:00
pub fn next(&mut self, ty: ty::Id, tys: &Types) -> Option<PLoc> {
Some(match tys.size_of(ty) {
0 => return None,
size @ 1..=8 => PLoc::Reg(self.0.next().unwrap(), size as _),
size @ 9..=16 => PLoc::WideReg(self.0.next_chunk::<2>().unwrap()[0], size as _),
size @ 17.. => PLoc::Ref(self.0.next().unwrap(), size),
2024-10-24 02:43:07 -05:00
})
2024-09-13 08:12:20 -05:00
}
}
2024-10-01 14:33:30 -05:00
impl ctx_map::CtxEntry for Ident {
type Ctx = str;
type Key<'a> = &'a str;
2024-09-30 12:09:17 -05:00
2024-10-01 14:33:30 -05:00
fn key<'a>(&self, ctx: &'a Self::Ctx) -> Self::Key<'a> {
2024-10-27 07:57:00 -05:00
unsafe { ctx.get_unchecked(self.range()) }
2024-09-30 12:09:17 -05:00
}
}
#[derive(Default)]
struct IdentInterner {
2024-10-01 14:33:30 -05:00
lookup: ctx_map::CtxMap<Ident>,
2024-09-30 12:09:17 -05:00
strings: String,
}
impl IdentInterner {
fn intern(&mut self, ident: &str) -> Ident {
2024-10-01 14:33:30 -05:00
let (entry, hash) = self.lookup.entry(ident, &self.strings);
match entry {
hash_map::RawEntryMut::Occupied(o) => o.get_key_value().0.value,
2024-09-30 12:09:17 -05:00
hash_map::RawEntryMut::Vacant(v) => {
2024-10-27 07:57:00 -05:00
let id = Ident::new(self.strings.len() as _, ident.len() as _).unwrap();
2024-09-30 12:09:17 -05:00
self.strings.push_str(ident);
2024-10-01 14:33:30 -05:00
v.insert(ctx_map::Key { hash, value: id }, ());
2024-09-30 12:09:17 -05:00
id
}
}
}
fn ident_str(&self, ident: Ident) -> &str {
2024-10-27 07:57:00 -05:00
&self.strings[ident.range()]
2024-09-30 12:09:17 -05:00
}
fn project(&self, ident: &str) -> Option<Ident> {
2024-10-01 14:33:30 -05:00
self.lookup.get(ident, &self.strings).copied()
2024-09-30 12:09:17 -05:00
}
fn clear(&mut self) {
self.lookup.clear();
self.strings.clear()
}
2024-09-30 12:09:17 -05:00
}
2024-09-13 08:12:20 -05:00
#[derive(Default)]
2024-10-01 14:33:30 -05:00
struct TypesTmp {
fields: Vec<Field>,
frontier: Vec<ty::Id>,
globals: Vec<ty::Global>,
funcs: Vec<ty::Func>,
2024-10-12 15:21:20 -05:00
args: Vec<ty::Id>,
2024-10-01 14:33:30 -05:00
}
#[derive(Default)]
2024-10-01 15:53:03 -05:00
pub struct TypeIns {
2024-09-13 08:12:20 -05:00
funcs: Vec<Func>,
args: Vec<ty::Id>,
globals: Vec<Global>,
structs: Vec<Struct>,
2024-09-30 12:09:17 -05:00
fields: Vec<Field>,
2024-09-13 08:12:20 -05:00
ptrs: Vec<Ptr>,
2024-10-30 14:20:03 -05:00
opts: Vec<Opt>,
2024-10-19 12:37:02 -05:00
slices: Vec<Array>,
2024-10-01 14:33:30 -05:00
}
2024-10-01 08:28:18 -05:00
struct FTask {
file: FileId,
id: ty::Func,
}
struct StringRef(ty::Global);
impl ctx_map::CtxEntry for StringRef {
type Ctx = [Global];
type Key<'a> = &'a [u8];
fn key<'a>(&self, ctx: &'a Self::Ctx) -> Self::Key<'a> {
&ctx[self.0 as usize].data
}
}
2024-10-01 14:33:30 -05:00
#[derive(Default)]
struct Types {
syms: ctx_map::CtxMap<ty::Id>,
names: IdentInterner,
strings: ctx_map::CtxMap<StringRef>,
2024-10-01 14:33:30 -05:00
ins: TypeIns,
tmp: TypesTmp,
tasks: Vec<Option<FTask>>,
2024-09-13 08:12:20 -05:00
}
trait TypeParser {
fn tys(&mut self) -> &mut Types;
fn on_reuse(&mut self, existing: ty::Id);
fn find_local_ty(&mut self, name: Ident) -> Option<ty::Id>;
2024-10-20 09:43:25 -05:00
fn eval_const(&mut self, file: FileId, expr: &Expr, ty: ty::Id) -> u64;
fn eval_global(&mut self, file: FileId, name: Ident, expr: &Expr) -> ty::Id;
fn infer_type(&mut self, expr: &Expr) -> ty::Id;
fn report(&self, file: FileId, pos: Pos, msg: impl Display) -> ty::Id;
2024-09-30 12:09:17 -05:00
fn find_type(
&mut self,
pos: Pos,
from_file: FileId,
file: FileId,
id: Result<Ident, &str>,
files: &[parser::Ast],
) -> ty::Id {
let ty = if let Ok(id) = id
&& let Some(ty) = self.find_local_ty(id)
{
ty
} else if let Ok(id) = id
&& let tys = self.tys()
&& let Some(&ty) = tys.syms.get(SymKey::Decl(file, id), &tys.ins)
{
self.on_reuse(ty);
ty
} else {
let f = &files[file as usize];
let Some((Expr::BinOp { left, right, .. }, name)) = f.find_decl(id) else {
return match id {
2024-11-03 03:15:03 -06:00
Ok(_) => ty::Id::NEVER,
Err("main") => self.report(
from_file,
pos,
format_args!(
"missing main function in '{}', compiler can't \
2024-11-03 03:15:03 -06:00
emmit libraries since such concept is not defined \
(minimal main function: `main := fn(): void {{}}`)",
f.path
),
),
Err(name) => {
self.report(from_file, pos, format_args!("undefined indentifier: {name}"))
}
};
};
let tys = self.tys();
if let Some(&ty) = tys.syms.get(SymKey::Decl(file, name), &tys.ins) {
ty
} else {
let ty = left
.find_pattern_path(name, right, |right| {
self.parse_ty(file, right, Some(name), files)
})
.unwrap_or_else(|_| unreachable!());
let tys = self.tys();
let nm = match ty.expand() {
ty::Kind::Struct(s) => &mut tys.ins.structs[s as usize].name,
ty::Kind::Func(s) => &mut tys.ins.funcs[s as usize].name,
ty::Kind::Global(s) => &mut tys.ins.globals[s as usize].name,
2024-10-27 07:57:00 -05:00
_ => &mut Ident::default(),
};
2024-10-27 07:57:00 -05:00
if nm.is_null() {
*nm = name;
}
tys.syms.insert(SymKey::Decl(file, name), ty, &tys.ins);
ty
}
};
let tys = self.tys();
if let ty::Kind::Global(g) = ty.expand() {
let g = &tys.ins.globals[g as usize];
if g.ty == ty::Id::TYPE {
return ty::Id::from(
u32::from_ne_bytes(g.data.as_slice().try_into().unwrap()) as u64
);
}
}
ty
}
/// returns none if comptime eval is required
fn parse_ty(
&mut self,
file: FileId,
expr: &Expr,
name: Option<Ident>,
files: &[parser::Ast],
) -> ty::Id {
match *expr {
Expr::Mod { id, .. } => ty::Kind::Module(id).compress(),
Expr::UnOp { op: TokenKind::Xor, val, .. } => {
let base = self.parse_ty(file, val, None, files);
self.tys().make_ptr(base)
}
2024-10-30 14:20:03 -05:00
Expr::UnOp { op: TokenKind::Que, val, .. } => {
let base = self.parse_ty(file, val, None, files);
self.tys().make_opt(base)
}
2024-10-27 07:57:00 -05:00
Expr::Ident { id, .. } if id.is_null() => id.len().into(),
Expr::Ident { id, pos, .. } => self.find_type(pos, file, file, Ok(id), files),
2024-10-20 09:43:25 -05:00
Expr::Field { target, pos, name }
if let ty::Kind::Module(inside) =
2024-10-20 09:43:25 -05:00
self.parse_ty(file, target, None, files).expand() =>
{
self.find_type(pos, file, inside, Err(name), files)
}
2024-10-20 09:43:25 -05:00
Expr::Directive { name: "TypeOf", args: [expr], .. } => self.infer_type(expr),
2024-10-01 07:00:41 -05:00
Expr::Slice { size: None, item, .. } => {
let ty = self.parse_ty(file, item, None, files);
self.tys().make_array(ty, ArrayLen::MAX)
2024-10-01 07:00:41 -05:00
}
Expr::Slice { size: Some(&Expr::Number { value, .. }), item, .. } => {
let ty = self.parse_ty(file, item, None, files);
self.tys().make_array(ty, value as _)
}
2024-10-20 09:43:25 -05:00
Expr::Slice { size, item, .. } => {
let ty = self.parse_ty(file, item, None, files);
let len = size
.map_or(ArrayLen::MAX, |expr| self.eval_const(file, expr, ty::Id::U32) as _);
self.tys().make_array(ty, len)
}
Expr::Struct { pos, fields, packed, captured, .. } => {
2024-10-22 03:17:16 -05:00
let captures_start = self.tys().tmp.args.len();
for &cp in captured {
let ty = self.find_local_ty(cp).expect("TODO");
self.tys().tmp.args.push(ty);
}
let captured = self.tys().pack_args(captures_start).expect("TODO");
let sym = SymKey::Struct(file, pos, captured);
let tys = self.tys();
if let Some(&ty) = tys.syms.get(sym, &tys.ins) {
return ty;
}
let prev_tmp = self.tys().tmp.fields.len();
2024-09-30 12:09:17 -05:00
for field in fields.iter().filter_map(CommentOr::or) {
let ty = self.parse_ty(file, &field.ty, None, files);
let field = Field { name: self.tys().names.intern(field.name), ty };
self.tys().tmp.fields.push(field);
2024-09-30 12:09:17 -05:00
}
let tys = self.tys();
tys.ins.structs.push(Struct {
file,
2024-10-01 15:53:03 -05:00
pos,
2024-10-27 07:57:00 -05:00
name: name.unwrap_or_default(),
field_start: tys.ins.fields.len() as _,
explicit_alignment: packed.then_some(1),
2024-09-30 12:09:17 -05:00
..Default::default()
});
tys.ins.fields.extend(tys.tmp.fields.drain(prev_tmp..));
let ty = ty::Kind::Struct(tys.ins.structs.len() as u32 - 1).compress();
tys.syms.insert(sym, ty, &tys.ins);
ty
}
Expr::Closure { pos, args, ret, .. } if let Some(name) = name => {
let func = Func {
file,
name,
sig: 'b: {
let arg_base = self.tys().tmp.args.len();
for arg in args {
let sym = parser::find_symbol(&files[file as usize].symbols, arg.id);
if sym.flags & idfl::COMPTIME != 0 {
self.tys().tmp.args.truncate(arg_base);
break 'b None;
}
let ty = self.parse_ty(file, &arg.ty, None, files);
self.tys().tmp.args.push(ty);
}
let Some(args) = self.tys().pack_args(arg_base) else {
return self.report(file, pos, "function has too many argumnets");
};
let ret = self.parse_ty(file, ret, None, files);
Some(Sig { args, ret })
},
expr: ExprRef::new(expr),
..Default::default()
};
let id = self.tys().ins.funcs.len() as _;
self.tys().ins.funcs.push(func);
ty::Kind::Func(id).compress()
}
2024-10-20 09:43:25 -05:00
_ if let Some(name) = name => self.eval_global(file, name, expr),
_ => ty::Id::from(self.eval_const(file, expr, ty::Id::TYPE)),
}
}
}
impl Types {
fn struct_field_range(&self, strct: ty::Struct) -> Range<usize> {
let start = self.ins.structs[strct as usize].field_start as usize;
let end = self
.ins
.structs
.get(strct as usize + 1)
.map_or(self.ins.fields.len(), |s| s.field_start as usize);
start..end
}
fn pack_args(&mut self, arg_base: usize) -> Option<ty::Tuple> {
let base = self.ins.args.len();
self.ins.args.extend(self.tmp.args.drain(arg_base..));
let needle = &self.ins.args[base..];
if needle.is_empty() {
return Some(ty::Tuple::empty());
}
let len = needle.len();
// FIXME: maybe later when this becomes a bottleneck we use more
// efficient search (SIMD?, indexing?)
let sp = self.ins.args.windows(needle.len()).position(|val| val == needle).unwrap();
self.ins.args.truncate((sp + needle.len()).max(base));
ty::Tuple::new(sp, len)
}
fn struct_fields(&self, strct: ty::Struct) -> &[Field] {
&self.ins.fields[self.struct_field_range(strct)]
}
2024-10-12 06:07:49 -05:00
fn reassemble(&mut self, buf: &mut Vec<u8>) {
self.ins.funcs.iter_mut().for_each(|f| f.offset = u32::MAX);
self.ins.globals.iter_mut().for_each(|g| g.offset = u32::MAX);
self.assemble(buf)
}
2024-10-24 02:43:07 -05:00
fn parama(&self, ret: impl Into<ty::Id>) -> (Option<PLoc>, ParamAlloc) {
let mut iter = ParamAlloc(1..12);
let ret = iter.next(ret.into(), self);
2024-10-24 02:43:07 -05:00
iter.0.start += ret.is_none() as u8;
(ret, iter)
2024-09-13 08:12:20 -05:00
}
2024-10-30 14:20:03 -05:00
fn make_opt(&mut self, base: ty::Id) -> ty::Id {
self.make_generic_ty(
Opt { base },
|ins| &mut ins.opts,
|e| SymKey::Optional(e),
ty::Kind::Opt,
)
2024-09-13 08:12:20 -05:00
}
2024-10-30 14:20:03 -05:00
fn make_ptr(&mut self, base: ty::Id) -> ty::Id {
self.make_generic_ty(
Ptr { base },
|ins| &mut ins.ptrs,
|e| SymKey::Pointer(e),
ty::Kind::Ptr,
)
2024-09-13 08:12:20 -05:00
}
2024-10-30 14:20:03 -05:00
fn make_array(&mut self, elem: ty::Id, len: ArrayLen) -> ty::Id {
self.make_generic_ty(
Array { elem, len },
|ins| &mut ins.slices,
|e| SymKey::Array(e),
ty::Kind::Slice,
)
2024-09-13 08:12:20 -05:00
}
2024-10-30 14:20:03 -05:00
fn make_generic_ty<T: Copy>(
&mut self,
ty: T,
get_col: fn(&mut TypeIns) -> &mut Vec<T>,
key: fn(&T) -> SymKey,
kind: fn(u32) -> ty::Kind,
) -> ty::Id {
*self.syms.get_or_insert(key(&{ ty }), &mut self.ins, |ins| {
get_col(ins).push(ty);
kind(get_col(ins).len() as u32 - 1).compress()
})
2024-09-13 08:12:20 -05:00
}
fn size_of(&self, ty: ty::Id) -> Size {
match ty.expand() {
ty::Kind::Slice(arr) => {
2024-10-19 12:37:02 -05:00
let arr = &self.ins.slices[arr as usize];
2024-09-13 08:12:20 -05:00
match arr.len {
0 => 0,
ArrayLen::MAX => 16,
2024-10-19 12:37:02 -05:00
len => self.size_of(arr.elem) * len,
2024-09-13 08:12:20 -05:00
}
}
ty::Kind::Struct(stru) => {
2024-10-01 14:36:23 -05:00
if self.ins.structs[stru as usize].size.get() != 0 {
return self.ins.structs[stru as usize].size.get();
2024-09-30 12:09:17 -05:00
}
let mut oiter = OffsetIter::new(stru, self);
while oiter.next(self).is_some() {}
2024-10-01 14:36:23 -05:00
self.ins.structs[stru as usize].size.set(oiter.offset);
oiter.offset
2024-09-13 08:12:20 -05:00
}
2024-10-31 04:36:18 -05:00
ty::Kind::Opt(opt) => {
let base = self.ins.opts[opt as usize].base;
if self.nieche_of(base).is_some() {
self.size_of(base)
} else {
self.size_of(base) + self.align_of(base)
}
}
_ if let Some(size) = ty.simple_size() => size,
2024-09-13 08:12:20 -05:00
ty => unimplemented!("size_of: {:?}", ty),
}
}
fn align_of(&self, ty: ty::Id) -> Size {
match ty.expand() {
2024-09-22 11:17:30 -05:00
ty::Kind::Struct(stru) => {
2024-10-01 14:36:23 -05:00
if self.ins.structs[stru as usize].align.get() != 0 {
return self.ins.structs[stru as usize].align.get() as _;
2024-09-30 12:09:17 -05:00
}
2024-10-01 14:36:23 -05:00
let align = self.ins.structs[stru as usize].explicit_alignment.map_or_else(
2024-09-30 12:09:17 -05:00
|| {
self.struct_fields(stru)
.iter()
.map(|&Field { ty, .. }| self.align_of(ty))
.max()
.unwrap_or(1)
},
|a| a as _,
);
2024-10-01 14:36:23 -05:00
self.ins.structs[stru as usize].align.set(align.try_into().unwrap());
2024-09-30 12:09:17 -05:00
align
2024-09-22 11:17:30 -05:00
}
2024-09-13 08:12:20 -05:00
ty::Kind::Slice(arr) => {
2024-10-19 12:37:02 -05:00
let arr = &self.ins.slices[arr as usize];
2024-09-13 08:12:20 -05:00
match arr.len {
ArrayLen::MAX => 8,
2024-10-19 12:37:02 -05:00
_ => self.align_of(arr.elem),
2024-09-13 08:12:20 -05:00
}
}
_ => self.size_of(ty).max(1),
}
}
2024-09-28 08:13:32 -05:00
fn base_of(&self, ty: ty::Id) -> Option<ty::Id> {
match ty.expand() {
2024-10-01 14:36:23 -05:00
ty::Kind::Ptr(p) => Some(self.ins.ptrs[p as usize].base),
2024-09-28 08:13:32 -05:00
_ => None,
}
}
2024-09-30 12:09:17 -05:00
2024-10-30 14:20:03 -05:00
fn inner_of(&self, ty: ty::Id) -> Option<ty::Id> {
match ty.expand() {
ty::Kind::Opt(o) => Some(self.ins.opts[o as usize].base),
_ => None,
}
}
2024-10-31 04:36:18 -05:00
fn opt_layout(&self, inner_ty: ty::Id) -> OptLayout {
match self.nieche_of(inner_ty) {
Some((_, flag_offset, flag_ty)) => {
OptLayout { flag_ty, flag_offset, payload_offset: 0 }
}
None => OptLayout {
flag_ty: ty::Id::BOOL,
flag_offset: 0,
payload_offset: self.align_of(inner_ty),
},
}
}
fn nieche_of(&self, ty: ty::Id) -> Option<(bool, Offset, ty::Id)> {
2024-10-30 14:20:03 -05:00
match ty.expand() {
2024-10-31 04:36:18 -05:00
ty::Kind::Ptr(_) => Some((false, 0, ty::Id::UINT)),
// TODO: cache this
ty::Kind::Struct(s) => OffsetIter::new(s, self).into_iter(self).find_map(|(f, off)| {
self.nieche_of(f.ty).map(|(uninit, o, ty)| (uninit, o + off, ty))
}),
_ => None,
2024-10-30 14:20:03 -05:00
}
}
2024-09-30 12:09:17 -05:00
fn find_struct_field(&self, s: ty::Struct, name: &str) -> Option<usize> {
2024-10-01 14:33:30 -05:00
let name = self.names.project(name)?;
2024-09-30 12:09:17 -05:00
self.struct_fields(s).iter().position(|f| f.name == name)
}
fn clear(&mut self) {
self.syms.clear();
self.names.clear();
self.strings.clear();
self.ins.funcs.clear();
self.ins.args.clear();
self.ins.globals.clear();
self.ins.structs.clear();
self.ins.fields.clear();
self.ins.ptrs.clear();
self.ins.slices.clear();
debug_assert_eq!(self.tmp.fields.len(), 0);
debug_assert_eq!(self.tmp.frontier.len(), 0);
debug_assert_eq!(self.tmp.globals.len(), 0);
debug_assert_eq!(self.tmp.funcs.len(), 0);
debug_assert_eq!(self.tmp.args.len(), 0);
debug_assert_eq!(self.tasks.len(), 0);
}
2024-09-13 08:12:20 -05:00
}
2024-10-31 04:36:18 -05:00
struct OptLayout {
flag_ty: ty::Id,
flag_offset: Offset,
payload_offset: Offset,
}
struct OffsetIter {
strct: ty::Struct,
offset: Offset,
2024-09-30 12:09:17 -05:00
fields: Range<usize>,
}
impl OffsetIter {
2024-09-30 12:09:17 -05:00
fn new(strct: ty::Struct, tys: &Types) -> Self {
Self { strct, offset: 0, fields: tys.struct_field_range(strct) }
}
fn offset_of(tys: &Types, idx: ty::Struct, field: &str) -> Option<(Offset, ty::Id)> {
2024-10-01 14:33:30 -05:00
let field_id = tys.names.project(field)?;
2024-09-30 12:09:17 -05:00
OffsetIter::new(idx, tys)
.into_iter(tys)
.find(|(f, _)| f.name == field_id)
.map(|(f, off)| (off, f.ty))
}
fn next<'a>(&mut self, tys: &'a Types) -> Option<(&'a Field, Offset)> {
2024-10-01 14:36:23 -05:00
let stru = &tys.ins.structs[self.strct as usize];
2024-10-01 15:53:03 -05:00
let field = &tys.ins.fields[self.fields.next()?];
2024-09-30 12:09:17 -05:00
let align = stru.explicit_alignment.map_or_else(|| tys.align_of(field.ty), |a| a as u32);
self.offset = (self.offset + align - 1) & !(align - 1);
let off = self.offset;
self.offset += tys.size_of(field.ty);
Some((field, off))
}
fn next_ty(&mut self, tys: &Types) -> Option<(ty::Id, Offset)> {
let (field, off) = self.next(tys)?;
Some((field.ty, off))
}
fn into_iter(mut self, tys: &Types) -> impl Iterator<Item = (&Field, Offset)> {
2024-09-30 12:09:17 -05:00
core::iter::from_fn(move || self.next(tys))
2024-05-17 12:53:59 -05:00
}
}
2024-09-30 12:09:17 -05:00
type HashMap<K, V> = hashbrown::HashMap<K, V, FnvBuildHasher>;
type FnvBuildHasher = core::hash::BuildHasherDefault<FnvHasher>;
2024-05-20 07:11:58 -05:00
struct FnvHasher(u64);
2024-09-30 12:09:17 -05:00
impl core::hash::Hasher for FnvHasher {
2024-05-20 07:11:58 -05:00
fn finish(&self) -> u64 {
self.0
}
fn write(&mut self, bytes: &[u8]) {
self.0 = bytes.iter().fold(self.0, |hash, &byte| {
let mut hash = hash;
hash ^= byte as u64;
hash = hash.wrapping_mul(0x100000001B3);
hash
});
}
}
impl Default for FnvHasher {
fn default() -> Self {
Self(0xCBF29CE484222325)
}
}
2024-06-23 06:55:48 -05:00
#[cfg(test)]
pub fn run_test(
name: &'static str,
ident: &'static str,
input: &'static str,
test: fn(&'static str, &'static str, &mut String),
) {
2024-09-30 12:09:17 -05:00
use std::{io::Write, path::PathBuf, string::ToString};
2024-06-23 06:55:48 -05:00
let filter = std::env::var("PT_FILTER").unwrap_or_default();
if !filter.is_empty() && !name.contains(&filter) {
return;
}
let mut output = String::new();
2024-09-04 16:46:32 -05:00
{
struct DumpOut<'a>(&'a mut String);
impl Drop for DumpOut<'_> {
fn drop(&mut self) {
if std::thread::panicking() {
2024-09-30 12:09:17 -05:00
std::println!("{}", self.0);
2024-09-04 16:46:32 -05:00
}
}
}
let dump = DumpOut(&mut output);
test(ident, input, dump.0);
}
2024-06-23 06:55:48 -05:00
let mut root = PathBuf::from(
std::env::var("PT_TEST_ROOT")
.unwrap_or(concat!(env!("CARGO_MANIFEST_DIR"), "/tests").to_string()),
);
2024-07-08 00:22:53 -05:00
root.push(name.replace("::", "_").replace(concat!(env!("CARGO_PKG_NAME"), "_"), ""));
2024-06-23 06:55:48 -05:00
root.set_extension("txt");
let expected = std::fs::read_to_string(&root).unwrap_or_default();
if output == expected {
return;
}
if std::env::var("PT_UPDATE").is_ok() {
std::fs::write(&root, output).unwrap();
return;
}
if !root.exists() {
std::fs::create_dir_all(root.parent().unwrap()).unwrap();
std::fs::write(&root, vec![]).unwrap();
}
let mut proc = std::process::Command::new("diff")
.arg("-u")
.arg("--color")
.arg(&root)
.arg("-")
.stdin(std::process::Stdio::piped())
.stdout(std::process::Stdio::inherit())
.spawn()
.unwrap();
2024-07-08 00:22:53 -05:00
proc.stdin.as_mut().unwrap().write_all(output.as_bytes()).unwrap();
2024-06-23 06:55:48 -05:00
proc.wait().unwrap();
panic!("test failed");
}
2024-09-13 07:30:23 -05:00
#[cfg(test)]
fn test_parse_files(
ident: &str,
input: &str,
ctx: &mut parser::Ctx,
) -> (Vec<parser::Ast>, Vec<Vec<u8>>) {
2024-10-13 08:22:16 -05:00
use {
self::parser::FileKind,
std::{borrow::ToOwned, string::ToString},
};
2024-09-30 12:09:17 -05:00
fn find_block<'a>(mut input: &'a str, test_name: &str) -> &'a str {
2024-09-13 07:30:23 -05:00
const CASE_PREFIX: &str = "#### ";
const CASE_SUFFIX: &str = "\n```hb";
loop {
let Some(pos) = input.find(CASE_PREFIX) else {
unreachable!("test {test_name} not found");
};
input = unsafe { input.get_unchecked(pos + CASE_PREFIX.len()..) };
if !input.starts_with(test_name) {
continue;
}
input = unsafe { input.get_unchecked(test_name.len()..) };
if !input.starts_with(CASE_SUFFIX) {
continue;
}
input = unsafe { input.get_unchecked(CASE_SUFFIX.len()..) };
let end = input.find("```").unwrap_or(input.len());
break unsafe { input.get_unchecked(..end) };
}
}
let input = find_block(input, ident);
let mut module_map = Vec::new();
2024-10-13 13:01:18 -05:00
let mut embed_map = Vec::new();
2024-09-13 07:30:23 -05:00
let mut last_start = 0;
2024-10-13 13:01:18 -05:00
let mut last_module_name = "test.hb";
2024-09-13 07:30:23 -05:00
for (i, m) in input.match_indices("// in module: ") {
2024-10-13 13:01:18 -05:00
if last_module_name.ends_with(".hb") {
fmt::test::format(ident, input[last_start..i].trim());
module_map.push((last_module_name, &input[last_start..i]));
} else {
embed_map.push((last_module_name, &input[last_start..i]));
}
2024-09-13 07:30:23 -05:00
let (module_name, _) = input[i + m.len()..].split_once('\n').unwrap();
last_module_name = module_name;
last_start = i + m.len() + module_name.len() + 1;
}
2024-10-13 13:01:18 -05:00
if last_module_name.ends_with(".hb") {
fmt::test::format(ident, input[last_start..].trim());
module_map.push((last_module_name, &input[last_start..]));
} else {
embed_map.push((last_module_name, &input[last_start..]));
}
let mut loader = |path: &str, _: &str, kind| match kind {
FileKind::Module => module_map
2024-09-13 07:30:23 -05:00
.iter()
.position(|&(name, _)| name == path)
.map(|i| i as parser::FileId)
2024-10-13 13:01:18 -05:00
.ok_or("Module Not Found".to_string()),
FileKind::Embed => embed_map
.iter()
.position(|&(name, _)| name == path)
.map(|i| i as parser::FileId)
.ok_or("Embed Not Found".to_string()),
2024-09-13 07:30:23 -05:00
};
2024-10-13 13:01:18 -05:00
(
module_map
.iter()
.map(|&(path, content)| parser::Ast::new(path, content.to_owned(), ctx, &mut loader))
2024-10-13 13:01:18 -05:00
.collect(),
embed_map.iter().map(|&(_, content)| content.to_owned().into_bytes()).collect(),
)
2024-09-13 07:30:23 -05:00
}
2024-10-20 05:22:28 -05:00
fn endoce_string(
literal: &str,
str: &mut Vec<u8>,
report: impl Fn(&core::str::Bytes, &str),
) -> Option<()> {
let report = |bytes: &core::str::Bytes, msg: &_| {
report(bytes, msg);
None::<u8>
};
let decode_braces = |str: &mut Vec<u8>, bytes: &mut core::str::Bytes| {
while let Some(b) = bytes.next()
&& b != b'}'
{
let c = bytes.next().or_else(|| report(bytes, "incomplete escape sequence"))?;
let decode = |b: u8| {
Some(match b {
b'0'..=b'9' => b - b'0',
b'a'..=b'f' => b - b'a' + 10,
b'A'..=b'F' => b - b'A' + 10,
_ => report(bytes, "expected hex digit or '}'")?,
})
};
str.push(decode(b)? << 4 | decode(c)?);
}
Some(())
};
let mut bytes = literal.bytes();
while let Some(b) = bytes.next() {
if b != b'\\' {
str.push(b);
continue;
}
let b = match bytes.next().or_else(|| report(&bytes, "incomplete escape sequence"))? {
b'n' => b'\n',
b'r' => b'\r',
b't' => b'\t',
b'\\' => b'\\',
b'\'' => b'\'',
b'"' => b'"',
b'0' => b'\0',
b'{' => {
decode_braces(str, &mut bytes);
continue;
}
_ => report(&bytes, "unknown escape sequence, expected [nrt\\\"'{0]")?,
};
str.push(b);
}
if str.last() != Some(&0) {
report(&bytes, "string literal must end with null byte (for now)");
}
2024-10-20 05:22:28 -05:00
Some(())
}
2024-10-12 06:07:49 -05:00
pub fn quad_sort<T>(mut slice: &mut [T], mut cmp: impl FnMut(&T, &T) -> core::cmp::Ordering) {
while let Some(it) = slice.take_first_mut() {
for ot in &mut *slice {
if cmp(it, ot) == core::cmp::Ordering::Greater {
core::mem::swap(it, ot);
}
}
}
debug_assert!(slice.is_sorted_by(|a, b| cmp(a, b) != core::cmp::Ordering::Greater));
}