some start for homemade regalloc

Cargo.toml

@@ -23,8 +23,8 @@ hbjit = { path = "jit" }
 
 [profile.release]
 lto = true
-debug = true
+#debug = true
-#strip = true
+strip = true
 codegen-units = 1
 panic = "abort"
 
@@ -32,7 +32,7 @@ panic = "abort"
 rustflags = ["-Zfmt-debug=none", "-Zlocation-detail=none"]
 inherits = "release"
 opt-level = "z"
-strip = true
+strip = "debuginfo"
 lto = true
 codegen-units = 1
 panic = "abort"

lang/Cargo.toml

@@ -20,9 +20,9 @@ log = "0.4.22"
 [dependencies.regalloc2]
 git = "https://github.com/jakubDoka/regalloc2"
 branch = "reuse-allocations"
-features = ["trace-log"]
+default-features = false
 
 [features]
-default = ["std"]
+default = ["std", "regalloc2/trace-log"]
 std = []
 no_log = ["log/max_level_off"]

lang/src/lib.rs

@@ -68,6 +68,7 @@ pub mod fs;
 pub mod fuzz;
 pub mod lexer;
 pub mod parser;
+pub mod regalloc;
 pub mod son;
 
 mod utils;

lang/src/regalloc.rs

@@ -0,0 +1,150 @@
+use {crate::reg::Reg, alloc::vec::Vec, core::ops::Range};
+
+type Nid = u16;
+
+pub trait Ctx {
+    fn uses_of(&self, nid: Nid) -> impl Iterator<Item = Nid>;
+    fn params_of(&self, nid: Nid) -> impl Iterator<Item = Nid>;
+    fn args_of(&self, nid: Nid) -> impl Iterator<Item = Nid>;
+    fn dom_of(&self, nid: Nid) -> Nid;
+}
+
+pub struct Env<'a, C: Ctx> {
+    ctx: &'a C,
+    func: &'a Func,
+    res: &'a mut Res,
+}
+
+impl<'a, C: Ctx> Env<'a, C> {
+    pub fn new(ctx: &'a C, func: &'a Func, res: &'a mut Res) -> Self {
+        Self { ctx, func, res }
+    }
+
+    pub fn run(&mut self) {
+        self.res.reg_to_node.clear();
+        self.res.reg_to_node.resize(self.func.instrs.len(), 0);
+
+        let mut bundle = Bundle::new(self.func.instrs.len());
+        for &inst in &self.func.instrs {
+            for uinst in self.ctx.uses_of(inst) {
+                let mut cursor = self.ctx.dom_of(uinst);
+                while cursor != self.ctx.dom_of(inst) {
+                    let mut range = self.func.blocks
+                        [self.func.id_to_block[cursor as usize] as usize]
+                        .range
+                        .clone();
+                    range.start = range.start.max(inst as usize);
+                    range.end = range.end.min(uinst as usize);
+                    bundle.add(range);
+                    cursor = self.ctx.dom_of(cursor);
+                }
+            }
+
+            match self.res.bundles.iter_mut().enumerate().find(|(_, b)| !b.overlaps(&bundle)) {
+                Some((i, other)) => {
+                    other.merge(&bundle);
+                    bundle.clear();
+                    self.res.reg_to_node[inst as usize] = i as Reg;
+                }
+                None => {
+                    self.res.reg_to_node[inst as usize] = self.res.bundles.len() as Reg;
+                    self.res.bundles.push(bundle);
+                    bundle = Bundle::new(self.func.instrs.len());
+                }
+            }
+        }
+    }
+}
+
+pub struct Res {
+    bundles: Vec<Bundle>,
+    pub reg_to_node: Vec<Reg>,
+}
+
+pub struct Bundle {
+    //unit_range: Range<usize>,
+    //set: BitSet,
+    taken: Vec<bool>,
+}
+
+impl Bundle {
+    fn new(size: usize) -> Self {
+        Self { taken: vec![false; size] }
+    }
+
+    fn add(&mut self, range: Range<usize>) {
+        self.taken[range].fill(true);
+    }
+
+    fn overlaps(&self, other: &Self) -> bool {
+        self.taken.iter().zip(other.taken.iter()).any(|(a, b)| a & b)
+    }
+
+    fn merge(&mut self, other: &Self) {
+        debug_assert!(!self.overlaps(other));
+        self.taken.iter_mut().zip(other.taken.iter()).for_each(|(a, b)| *a = *b);
+    }
+
+    fn clear(&mut self) {
+        self.taken.fill(false);
+    }
+
+    //fn overlaps(&self, other: &Self) -> bool {
+    //    if self.unit_range.start >= other.unit_range.end
+    //        || self.unit_range.end <= other.unit_range.start
+    //    {
+    //        return false;
+    //    }
+
+    //    let [mut a, mut b] = [self, other];
+    //    if a.unit_range.start > b.unit_range.start {
+    //        mem::swap(&mut a, &mut b);
+    //    }
+    //    let [mut tmp_a, mut tmp_b] = [0; 2];
+    //    let [units_a, units_b] = [a.set.units(&mut tmp_a), b.set.units(&mut tmp_b)];
+    //    let len = a.unit_range.len().min(b.unit_range.len());
+    //    let [units_a, units_b] =
+    //        [&units_a[b.unit_range.start - a.unit_range.start..][..len], &units_b[..len]];
+    //    units_a.iter().zip(units_b).any(|(&a, &b)| a & b != 0)
+    //}
+
+    //fn merge(mut self, mut other: Self) -> Self {
+    //    debug_assert!(!self.overlaps(&other));
+
+    //    if self.unit_range.start > other.unit_range.start {
+    //        mem::swap(&mut self, &mut other);
+    //    }
+
+    //    let final_range = self.unit_range.start..self.unit_range.end.max(other.unit_range.end);
+
+    //    self.set.reserve(final_range.len());
+
+    //    let mut tmp = 0;
+    //    let other_units = other.set.units(&mut tmp);
+
+    //    match self.set.units_mut() {
+    //        Ok(units) => {
+    //            units[other.unit_range.start - self.unit_range.start..]
+    //                .iter_mut()
+    //                .zip(other_units)
+    //                .for_each(|(a, b)| *a |= b);
+    //        }
+    //        Err(view) => view.add_mask(tmp),
+    //    }
+
+    //    self
+    //}
+}
+
+pub struct Func {
+    pub blocks: Vec<Block>,
+    pub instrs: Vec<Nid>,
+    pub id_to_instr: Vec<Nid>,
+    pub id_to_block: Vec<Nid>,
+}
+
+pub struct Block {
+    pub range: Range<usize>,
+    pub start_id: Nid,
+    pub eld_id: Nid,
+}

lang/src/son/hbvm.rs

@@ -9,10 +9,9 @@ use {
         utils::{BitSet, Vc},
         HashMap, Offset, PLoc, Reloc, Sig, Size, TypedReloc, Types,
     },
-    alloc::{borrow::ToOwned, string::String, vec::Vec},
+    alloc::{borrow::ToOwned, boxed::Box, collections::BTreeMap, string::String, vec::Vec},
     core::{assert_matches::debug_assert_matches, mem},
     hbbytecode::{self as instrs, *},
-    std::{boxed::Box, collections::BTreeMap},
 };
 
 impl Types {

lang/src/utils.rs

@@ -1,3 +1,4 @@
+#![expect(dead_code)]
 use {
     alloc::alloc,
     core::{
@@ -176,15 +177,47 @@ impl BitSet {
     }
 
     pub fn clear(&mut self, len: usize) {
-        if len > self.data_and_len().1 {
+        self.reserve(len);
-            self.grow(len.next_power_of_two().max(4 * Self::UNIT));
-        }
         if self.is_inline() {
             unsafe { self.inline &= Self::FLAG };
         } else {
             self.data_mut_and_len().0.fill(0);
         }
     }
+
+    pub fn units<'a>(&'a self, slot: &'a mut usize) -> &'a [usize] {
+        if self.is_inline() {
+            *slot = unsafe { self.inline } & !Self::FLAG;
+            core::slice::from_ref(slot)
+        } else {
+            self.data_and_len().0
+        }
+    }
+
+    pub fn reserve(&mut self, len: usize) {
+        if len > self.data_and_len().1 {
+            self.grow(len.next_power_of_two().max(4 * Self::UNIT));
+        }
+    }
+
+    pub fn units_mut(&mut self) -> Result<&mut [usize], &mut InlineBitSetView> {
+        if self.is_inline() {
+            Err(unsafe {
+                core::mem::transmute::<&mut usize, &mut InlineBitSetView>(&mut self.inline)
+            })
+        } else {
+            Ok(self.data_mut_and_len().0)
+        }
+    }
+}
+
+pub struct InlineBitSetView(usize);
+
+impl InlineBitSetView {
+    pub(crate) fn add_mask(&mut self, tmp: usize) {
+        debug_assert!(tmp & BitSet::FLAG == 0);
+        self.0 |= tmp;
+    }
 }
 
 pub struct BitSetIter<'a> {