moving codegen to instrs alongside disasm

merge
reducing dependencies
2024-09-19 18:25:28 +02:00 · 2024-09-19 17:15:03 +02:00 · 2024-09-19 17:13:51 +02:00 · 2024-09-19 13:40:03 +02:00 · 2024-09-18 10:34:07 +02:00 · 2024-09-18 10:14:17 +02:00
36 changed files with 2703 additions and 1778 deletions
--- a/Cargo.lock
+++ b/Cargo.lock
@ -35,7 +35,7 @@ checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
 [[package]]
 name = "cranelift-bforest"
 version = "0.113.0"
-source = "git+https://github.com/jakubDoka/wasmtime.git#969c85a481985be2fa2d555cdb338454f77213c6"
+source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
 dependencies = [
 "cranelift-entity",
 ]
@ -43,12 +43,12 @@ dependencies = [
 [[package]]
 name = "cranelift-bitset"
 version = "0.113.0"
-source = "git+https://github.com/jakubDoka/wasmtime.git#969c85a481985be2fa2d555cdb338454f77213c6"
+source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"

 [[package]]
 name = "cranelift-codegen"
 version = "0.113.0"
-source = "git+https://github.com/jakubDoka/wasmtime.git#969c85a481985be2fa2d555cdb338454f77213c6"
+source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
 dependencies = [
 "bumpalo",
 "cranelift-bforest",
@ -60,7 +60,7 @@ dependencies = [
 "cranelift-isle",
 "hashbrown",
 "log",
- "regalloc2",
+ "regalloc2 0.10.2 (registry+https://github.com/rust-lang/crates.io-index)",
 "rustc-hash",
 "smallvec",
 "target-lexicon",
@ -69,7 +69,7 @@ dependencies = [
 [[package]]
 name = "cranelift-codegen-meta"
 version = "0.113.0"
-source = "git+https://github.com/jakubDoka/wasmtime.git#969c85a481985be2fa2d555cdb338454f77213c6"
+source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
 dependencies = [
 "cranelift-codegen-shared",
 ]
@ -77,12 +77,12 @@ dependencies = [
 [[package]]
 name = "cranelift-codegen-shared"
 version = "0.113.0"
-source = "git+https://github.com/jakubDoka/wasmtime.git#969c85a481985be2fa2d555cdb338454f77213c6"
+source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"

 [[package]]
 name = "cranelift-control"
 version = "0.113.0"
-source = "git+https://github.com/jakubDoka/wasmtime.git#969c85a481985be2fa2d555cdb338454f77213c6"
+source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
 dependencies = [
 "arbitrary",
 ]
@ -90,7 +90,7 @@ dependencies = [
 [[package]]
 name = "cranelift-entity"
 version = "0.113.0"
-source = "git+https://github.com/jakubDoka/wasmtime.git#969c85a481985be2fa2d555cdb338454f77213c6"
+source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
 dependencies = [
 "cranelift-bitset",
 ]
@ -98,7 +98,7 @@ dependencies = [
 [[package]]
 name = "cranelift-isle"
 version = "0.113.0"
-source = "git+https://github.com/jakubDoka/wasmtime.git#969c85a481985be2fa2d555cdb338454f77213c6"
+source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"

 [[package]]
 name = "hashbrown"
@ -122,7 +122,7 @@ dependencies = [
 "cranelift-control",
 "cranelift-isle",
 "log",
- "regalloc2",
+ "regalloc2 0.10.2 (registry+https://github.com/rust-lang/crates.io-index)",
 "smallvec",
 "target-lexicon",
 ]
@ -135,7 +135,9 @@ version = "0.1.0"
 name = "hblang"
 version = "0.1.0"
 dependencies = [
+ "hbbytecode",
 "hbvm",
+ "regalloc2 0.10.2 (git+https://github.com/jakubDoka/regalloc2)",
 ]

 [[package]]
@ -211,6 +213,16 @@ dependencies = [
 "smallvec",
 ]

+[[package]]
+name = "regalloc2"
+version = "0.10.2"
+source = "git+https://github.com/jakubDoka/regalloc2#7e74b2fde4f022816cded93ab5685e46f8e3a159"
+dependencies = [
+ "hashbrown",
+ "rustc-hash",
+ "smallvec",
+]
+
 [[package]]
 name = "rustc-hash"
 version = "2.0.0"
--- a/hbbytecode/Cargo.toml
+++ b/hbbytecode/Cargo.toml
@ -2,3 +2,8 @@
 name = "hbbytecode"
 version = "0.1.0"
 edition = "2018"
+
+[features]
+default = ["disasm"]
+std = []
+disasm = ["std"]
--- a/hbbytecode/build.rs
+++ b/hbbytecode/build.rs
@ -1,58 +1,208 @@
 #![feature(iter_next_chunk)]

+use std::{collections::HashSet, fmt::Write};
+
 fn main() -> Result<(), Box<dyn std::error::Error>> {
    println!("cargo:rerun-if-changed=build.rs");
    println!("cargo:rerun-if-changed=instructions.in");

    let mut generated = String::new();
-    gen_op_structs(&mut generated)?;
-    std::fs::write("src/ops.rs", generated)?;
-
-    let mut generated = String::new();
-    gen_op_codes(&mut generated)?;
-    std::fs::write("src/opcode.rs", generated)?;
+    gen_instrs(&mut generated)?;
+    std::fs::write("src/instrs.rs", generated)?;

    Ok(())
 }

-fn gen_op_structs(generated: &mut String) -> std::fmt::Result {
-    use std::fmt::Write;
-    let mut seen = std::collections::HashSet::new();
+fn gen_instrs(generated: &mut String) -> Result<(), Box<dyn std::error::Error>> {
+    writeln!(generated, "#![allow(dead_code)] #![allow(clippy::upper_case_acronyms)]")?;
    writeln!(generated, "use crate::*;")?;
-    for [.., args, _] in instructions() {
-        if !seen.insert(args) {
-            continue;
-        }

-        writeln!(generated, "#[derive(Clone, Copy, Debug)]")?;
-        writeln!(generated, "#[repr(packed)]")?;
-        write!(generated, "pub struct Ops{args}(")?;
-        let mut first = true;
-        for ch in args.chars().filter(|&ch| ch != 'N') {
-            if !std::mem::take(&mut first) {
-                write!(generated, ",")?;
+    '_opcode_structs: {
+        let mut seen = HashSet::new();
+        for [.., args, _] in instructions() {
+            if !seen.insert(args) {
+                continue;
            }
-            write!(generated, "pub Op{ch}")?;
+
+            writeln!(generated, "#[derive(Clone, Copy, Debug)]")?;
+            writeln!(generated, "#[repr(packed)]")?;
+            write!(generated, "pub struct Ops{args}(")?;
+            let mut first = true;
+            for ch in args.chars().filter(|&ch| ch != 'N') {
+                if !std::mem::take(&mut first) {
+                    write!(generated, ",")?;
+                }
+                write!(generated, "pub Op{ch}")?;
+            }
+            writeln!(generated, ");")?;
+            writeln!(generated, "unsafe impl BytecodeItem for Ops{args} {{}}")?;
        }
-        writeln!(generated, ");")?;
-        writeln!(generated, "unsafe impl BytecodeItem for Ops{args} {{}}")?;
    }

+    '_max_size: {
+        let max = instructions()
+            .map(
+                |[_, _, ty, _]| {
+                    if ty == "N" {
+                        1
+                    } else {
+                        iter_args(ty).map(arg_to_width).sum::<usize>() + 1
+                    }
+                },
+            )
+            .max()
+            .unwrap();
+
+        writeln!(generated, "pub const MAX_SIZE: usize = {max};")?;
+    }
+
+    '_encoders: {
+        for [op, name, ty, doc] in instructions() {
+            writeln!(generated, "/// {}", doc.trim_matches('"'))?;
+            let name = name.to_lowercase();
+            let args = comma_sep(
+                iter_args(ty)
+                    .enumerate()
+                    .map(|(i, c)| format!("{}{i}: {}", arg_to_name(c), arg_to_type(c))),
+            );
+            writeln!(generated, "pub fn {name}({args}) -> (usize, [u8; MAX_SIZE]) {{")?;
+            let arg_names =
+                comma_sep(iter_args(ty).enumerate().map(|(i, c)| format!("{}{i}", arg_to_name(c))));
+            writeln!(generated, "    unsafe {{ crate::encode({ty}({op}, {arg_names})) }}")?;
+            writeln!(generated, "}}")?;
+        }
+    }
+
+    '_structs: {
+        let mut seen = std::collections::HashSet::new();
+        for [_, _, ty, _] in instructions() {
+            if !seen.insert(ty) {
+                continue;
+            }
+            let types = comma_sep(iter_args(ty).map(arg_to_type).map(|s| s.to_string()));
+            writeln!(generated, "#[repr(packed)] pub struct {ty}(u8, {types});")?;
+        }
+    }
+
+    '_name_list: {
+        writeln!(generated, "pub const NAMES: [&str; {}] = [", instructions().count())?;
+        for [_, name, _, _] in instructions() {
+            writeln!(generated, "    \"{}\",", name.to_lowercase())?;
+        }
+        writeln!(generated, "];")?;
+    }
+
+    let instr = "Instr";
+    let oper = "Oper";
+
+    '_instr_enum: {
+        writeln!(generated, "#[derive(Debug, Clone, Copy, PartialEq, Eq)] #[repr(u8)]")?;
+        writeln!(generated, "pub enum {instr} {{")?;
+        for [id, name, ..] in instructions() {
+            writeln!(generated, "    {name} = {id},")?;
+        }
+        writeln!(generated, "}}")?;
+    }
+
+    '_arg_kind: {
+        writeln!(generated, "#[derive(Debug, Clone, Copy, PartialEq, Eq)]")?;
+        writeln!(generated, "pub enum {oper} {{")?;
+        let mut seen = HashSet::new();
+        for ty in instructions().flat_map(|[.., ty, _]| iter_args(ty)) {
+            if !seen.insert(ty) {
+                continue;
+            }
+            writeln!(generated, "    {ty}({}),", arg_to_type(ty))?;
+        }
+        writeln!(generated, "}}")?;
+    }
+
+    '_parse_opers: {
+        writeln!(
+            generated,
+            "/// This assumes the instruction byte is still at the beginning of the buffer"
+        )?;
+        writeln!(generated, "#[cfg(feature = \"disasm\")]")?;
+        writeln!(generated, "pub fn parse_args(bytes: &mut &[u8], kind: {instr}, buf: &mut std::vec::Vec<{oper}>) -> Option<()> {{")?;
+        writeln!(generated, "    match kind {{")?;
+        let mut instrs = instructions().collect::<Vec<_>>();
+        instrs.sort_unstable_by_key(|&[.., ty, _]| ty);
+        for group in instrs.chunk_by(|[.., a, _], [.., b, _]| a == b) {
+            let ty = group[0][2];
+            for &[_, name, ..] in group {
+                writeln!(generated, "        | {instr}::{name}")?;
+            }
+            generated.pop();
+            writeln!(generated, " => {{")?;
+            if iter_args(ty).count() != 0 {
+                writeln!(generated, "            let data = crate::decode::<{ty}>(bytes)?;")?;
+                writeln!(
+                    generated,
+                    "            buf.extend([{}]);",
+                    comma_sep(
+                        iter_args(ty).zip(1u32..).map(|(t, i)| format!("{oper}::{t}(data.{i})"))
+                    )
+                )?;
+            } else {
+                writeln!(generated, "            crate::decode::<{ty}>(bytes)?;")?;
+            }
+
+            writeln!(generated, "        }}")?;
+        }
+        writeln!(generated, "    }}")?;
+        writeln!(generated, "    Some(())")?;
+        writeln!(generated, "}}")?;
+    }
+
+    std::fs::write("src/instrs.rs", generated)?;
    Ok(())
 }

-fn gen_op_codes(generated: &mut String) -> std::fmt::Result {
-    use std::fmt::Write;
-    for [op, name, _, comment] in instructions() {
-        writeln!(generated, "#[doc = {comment}]")?;
-        writeln!(generated, "pub const {name}: u8 = {op};")?;
-    }
-    Ok(())
+fn comma_sep(items: impl Iterator<Item = String>) -> String {
+    items.map(|item| item.to_string()).collect::<Vec<_>>().join(", ")
 }

 fn instructions() -> impl Iterator<Item = [&'static str; 4]> {
    include_str!("../hbbytecode/instructions.in")
        .lines()
-        .map(|line| line.strip_suffix(';').unwrap())
+        .filter_map(|line| line.strip_suffix(';'))
        .map(|line| line.splitn(4, ',').map(str::trim).next_chunk().unwrap())
 }
+
+fn arg_to_type(arg: char) -> &'static str {
+    match arg {
+        'R' | 'B' => "u8",
+        'H' => "u16",
+        'W' => "u32",
+        'D' | 'A' => "u64",
+        'P' => "i16",
+        'O' => "i32",
+        _ => panic!("unknown type: {}", arg),
+    }
+}
+
+fn arg_to_width(arg: char) -> usize {
+    match arg {
+        'R' | 'B' => 1,
+        'H' => 2,
+        'W' => 4,
+        'D' | 'A' => 8,
+        'P' => 2,
+        'O' => 4,
+        _ => panic!("unknown type: {}", arg),
+    }
+}
+
+fn arg_to_name(arg: char) -> &'static str {
+    match arg {
+        'R' => "reg",
+        'B' | 'H' | 'W' | 'D' => "imm",
+        'P' | 'O' => "offset",
+        'A' => "addr",
+        _ => panic!("unknown type: {}", arg),
+    }
+}
+
+fn iter_args(ty: &'static str) -> impl Iterator<Item = char> {
+    ty.chars().filter(|c| *c != 'N')
+}
--- a/hbbytecode/src/instrs.rs
+++ b/hbbytecode/src/instrs.rs
--- a/hbbytecode/src/lib.rs
+++ b/hbbytecode/src/lib.rs
@ -1,10 +1,12 @@
 #![no_std]

-pub use crate::ops::*;
+#[cfg(feature = "std")]
+extern crate std;
+
+pub use crate::instrs::*;
 use core::convert::TryFrom;

-pub mod opcode;
-mod ops;
+mod instrs;

 type OpR = u8;

@ -22,6 +24,38 @@ type OpD = u64;
 pub unsafe trait BytecodeItem {}
 unsafe impl BytecodeItem for u8 {}

+impl TryFrom<u8> for Instr {
+    type Error = u8;
+
+    #[inline]
+    fn try_from(value: u8) -> Result<Self, Self::Error> {
+        #[cold]
+        fn failed(value: u8) -> Result<Instr, u8> {
+            Err(value)
+        }
+
+        if value < NAMES.len() as u8 {
+            unsafe { Ok(std::mem::transmute::<u8, Instr>(value)) }
+        } else {
+            failed(value)
+        }
+    }
+}
+
+#[inline]
+unsafe fn encode<T>(instr: T) -> (usize, [u8; instrs::MAX_SIZE]) {
+    let mut buf = [0; instrs::MAX_SIZE];
+    core::ptr::write(buf.as_mut_ptr() as *mut T, instr);
+    (core::mem::size_of::<T>(), buf)
+}
+
+#[inline]
+fn decode<T>(binary: &mut &[u8]) -> Option<T> {
+    let (front, rest) = std::mem::take(binary).split_at_checked(core::mem::size_of::<T>())?;
+    *binary = rest;
+    unsafe { Some(core::ptr::read(front.as_ptr() as *const T)) }
+}
+
 /// Rounding mode
 #[derive(Clone, Copy, Debug, PartialEq, Eq)]
 #[repr(u8)]
@ -39,3 +73,181 @@ impl TryFrom<u8> for RoundingMode {
        (value <= 3).then(|| unsafe { core::mem::transmute(value) }).ok_or(())
    }
 }
+
+#[cfg(feature = "disasm")]
+#[derive(Clone, Copy)]
+pub enum DisasmItem {
+    Func,
+    Global,
+}
+
+#[cfg(feature = "disasm")]
+pub fn disasm(
+    binary: &mut &[u8],
+    functions: &std::collections::BTreeMap<u32, (&str, u32, DisasmItem)>,
+    out: &mut impl std::io::Write,
+    mut eca_handler: impl FnMut(&mut &[u8]),
+) -> std::io::Result<()> {
+    use {
+        self::instrs::Instr,
+        std::{
+            collections::{hash_map::Entry, HashMap},
+            convert::TryInto,
+            vec::Vec,
+        },
+    };
+
+    fn instr_from_byte(b: u8) -> std::io::Result<Instr> {
+        if b as usize >= instrs::NAMES.len() {
+            return Err(std::io::ErrorKind::InvalidData.into());
+        }
+        Ok(unsafe { std::mem::transmute::<u8, Instr>(b) })
+    }
+
+    let mut labels = HashMap::<u32, u32>::default();
+    let mut buf = Vec::<instrs::Oper>::new();
+    let mut has_cycle = false;
+    let mut has_oob = false;
+
+    '_offset_pass: for (&off, &(_name, len, kind)) in functions.iter() {
+        if matches!(kind, DisasmItem::Global) {
+            continue;
+        }
+
+        let prev = *binary;
+
+        *binary = &binary[..off as usize];
+
+        let mut label_count = 0;
+        while let Some(&byte) = binary.first() {
+            let offset: i32 = (prev.len() - binary.len()).try_into().unwrap();
+            if offset as u32 == off + len {
+                break;
+            }
+            let Ok(inst) = instr_from_byte(byte) else { break };
+            instrs::parse_args(binary, inst, &mut buf).ok_or(std::io::ErrorKind::OutOfMemory)?;
+
+            for op in buf.drain(..) {
+                let rel = match op {
+                    instrs::Oper::O(rel) => rel,
+                    instrs::Oper::P(rel) => rel.into(),
+                    _ => continue,
+                };
+
+                has_cycle |= rel == 0;
+
+                let global_offset: u32 = (offset + rel).try_into().unwrap();
+                if functions.get(&global_offset).is_some() {
+                    continue;
+                }
+                label_count += match labels.entry(global_offset) {
+                    Entry::Occupied(_) => 0,
+                    Entry::Vacant(entry) => {
+                        entry.insert(label_count);
+                        1
+                    }
+                }
+            }
+
+            if matches!(inst, Instr::ECA) {
+                eca_handler(binary);
+            }
+        }
+
+        *binary = prev;
+    }
+
+    let mut ordered = functions.iter().collect::<Vec<_>>();
+    ordered.sort_unstable_by_key(|(_, (name, _, _))| name);
+
+    '_dump: for (&off, &(name, len, kind)) in ordered {
+        if matches!(kind, DisasmItem::Global) {
+            continue;
+        }
+        let prev = *binary;
+
+        writeln!(out, "{name}:")?;
+
+        *binary = &binary[..off as usize];
+        while let Some(&byte) = binary.first() {
+            let offset: i32 = (prev.len() - binary.len()).try_into().unwrap();
+            if offset as u32 == off + len {
+                break;
+            }
+            let Ok(inst) = instr_from_byte(byte) else {
+                writeln!(out, "invalid instr {byte}")?;
+                break;
+            };
+            instrs::parse_args(binary, inst, &mut buf).unwrap();
+
+            if let Some(label) = labels.get(&offset.try_into().unwrap()) {
+                write!(out, "{:>2}: ", label)?;
+            } else {
+                write!(out, "    ")?;
+            }
+
+            write!(out, "{inst:<8?} ")?;
+
+            'a: for (i, op) in buf.drain(..).enumerate() {
+                if i != 0 {
+                    write!(out, ", ")?;
+                }
+
+                let rel = 'b: {
+                    match op {
+                        instrs::Oper::O(rel) => break 'b rel,
+                        instrs::Oper::P(rel) => break 'b rel.into(),
+                        instrs::Oper::R(r) => write!(out, "r{r}")?,
+                        instrs::Oper::B(b) => write!(out, "{b}b")?,
+                        instrs::Oper::H(h) => write!(out, "{h}h")?,
+                        instrs::Oper::W(w) => write!(out, "{w}w")?,
+                        instrs::Oper::D(d) if (d as i64) < 0 => write!(out, "{}d", d as i64)?,
+                        instrs::Oper::D(d) => write!(out, "{d}d")?,
+                        instrs::Oper::A(a) => write!(out, "{a}a")?,
+                    }
+
+                    continue 'a;
+                };
+
+                let global_offset: u32 = (offset + rel).try_into().unwrap();
+                if let Some(&(name, ..)) = functions.get(&global_offset) {
+                    if name.contains('\0') {
+                        write!(out, ":{name:?}")?;
+                    } else {
+                        write!(out, ":{name}")?;
+                    }
+                } else {
+                    let local_has_oob = global_offset < off
+                        || global_offset > off + len
+                        || instr_from_byte(prev[global_offset as usize]).is_err()
+                        || prev[global_offset as usize] == 0;
+                    has_oob |= local_has_oob;
+                    let label = labels.get(&global_offset).unwrap();
+                    if local_has_oob {
+                        write!(out, "!!!!!!!!!{rel}")?;
+                    } else {
+                        write!(out, ":{label}")?;
+                    }
+                }
+            }
+
+            writeln!(out)?;
+
+            if matches!(inst, Instr::ECA) {
+                eca_handler(binary);
+            }
+        }
+
+        *binary = prev;
+    }
+
+    if has_oob {
+        return Err(std::io::ErrorKind::InvalidInput.into());
+    }
+
+    if has_cycle {
+        return Err(std::io::ErrorKind::TimedOut.into());
+    }
+
+    Ok(())
+}
--- a/hbcb/Cargo.toml
+++ b/hbcb/Cargo.toml
@ -4,7 +4,7 @@ version = "0.1.0"
 edition = "2021"

 [dependencies]
-cranelift-codegen = { git = "https://github.com/jakubDoka/wasmtime.git", default-features = false }
+cranelift-codegen = { git = "https://github.com/jakubDoka/wasmtime.git", default-features = false, features = ["std"] }
 cranelift-control = { git = "https://github.com/jakubDoka/wasmtime.git", default-features = false }
 log = "0.4.22"
 regalloc2 = "0.10.2"
--- a/hbcb/build.rs
+++ b/hbcb/build.rs
@ -15,7 +15,6 @@
 // current directory is used to find the sources.

 use {
-    core::panic,
    cranelift_codegen_meta::{self as meta, isle::IsleCompilations},
    cranelift_isle::error::Errors,
    meta::isle::IsleCompilation,
--- a/hbcb/src/abi.rs
+++ b/hbcb/src/abi.rs
@ -1,13 +1,15 @@
 //! Implementation of a standard Riscv64 ABI.

 use {
-    crate::inst::*,
+    crate::{
+        inst::*,
+        settings::{self, Flags as RiscvFlags},
+    },
    alloc::{boxed::Box, vec::Vec},
    cranelift_codegen::{
        ir::{self, types::*, LibCall, Signature},
        isa::{self, unwind::UnwindInst, CallConv},
        machinst::*,
-        settings::{self, Flags as RiscvFlags},
        CodegenError, CodegenResult,
    },
    regalloc2::{MachineEnv, PReg, PRegSet},
@ -21,11 +23,6 @@ pub(crate) type Riscv64Callee = Callee<Riscv64MachineDeps>;
 /// Support for the Riscv64 ABI from the caller side (at a callsite).
 pub(crate) type Riscv64ABICallSite = CallSite<Riscv64MachineDeps>;

-/// This is the limit for the size of argument and return-value areas on the
-/// stack. We place a reasonable limit here to avoid integer overflow issues
-/// with 32-bit arithmetic: for now, 128 MB.
-static STACK_ARG_RET_SIZE_LIMIT: u32 = 128 * 1024 * 1024;
-
 /// Riscv64-specific ABI behavior. This struct just serves as an implementation
 /// point for the trait; it is never actually instantiated.
 pub struct Riscv64MachineDeps;
@ -70,6 +67,11 @@ impl ABIMachineSpec for Riscv64MachineDeps {
    type F = RiscvFlags;
    type I = Inst;

+    /// This is the limit for the size of argument and return-value areas on the
+    /// stack. We place a reasonable limit here to avoid integer overflow issues
+    /// with 32-bit arithmetic: for now, 128 MB.
+    const STACK_ARG_RET_SIZE_LIMIT: u32 = 128 * 1024 * 1024;
+
    fn word_bits() -> u32 {
        64
    }
@ -651,8 +653,12 @@ impl ABIMachineSpec for Riscv64MachineDeps {
    }
 }

-impl Riscv64ABICallSite {
-    pub fn emit_return_call(mut self, ctx: &mut Lower<Inst>, args: isle::ValueSlice) {
+pub trait EmitReturnCall {
+    fn emit_return_call(mut self, ctx: &mut Lower<Inst>, args: isle::ValueSlice);
+}
+
+impl EmitReturnCall for Riscv64ABICallSite {
+    fn emit_return_call(mut self, ctx: &mut Lower<Inst>, args: isle::ValueSlice) {
        let new_stack_arg_size =
            u32::try_from(self.sig(ctx.sigs()).sized_stack_arg_space()).unwrap();

--- a/hbcb/src/inst/encode.rs
+++ b/hbcb/src/inst/encode.rs
@ -6,12 +6,14 @@
 //! Some instructions especially in extensions have slight variations from
 //! the base RISC-V specification.

-use super::*;
-use crate::lower::isle::generated_code::{
-    COpcodeSpace, CaOp, CbOp, CiOp, CiwOp, ClOp, CrOp, CsOp, CssOp, CsznOp, FpuOPWidth,
-    VecAluOpRImm5, VecAluOpRR, VecAluOpRRRImm5, VecAluOpRRRR, VecOpCategory, ZcbMemOp,
+use {
+    super::*,
+    crate::lower::isle::generated_code::{
+        COpcodeSpace, CaOp, CbOp, CiOp, CiwOp, ClOp, CrOp, CsOp, CssOp, CsznOp, FpuOPWidth,
+        VecAluOpRImm5, VecAluOpRR, VecAluOpRRRImm5, VecAluOpRRRR, VecOpCategory, ZcbMemOp,
+    },
+    cranelift_codegen::machinst::isle::WritableReg,
 };
-use crate::machinst::isle::WritableReg;

 fn unsigned_field_width(value: u32, width: u8) -> u32 {
    debug_assert_eq!(value & (!0 << width), 0);
@ -199,14 +201,7 @@ pub fn encode_valu_rr(op: VecAluOpRR, vd: WritableReg, vs: Reg, masking: VecOpMa
        (reg_to_gpr_num(vs), op.aux_encoding())
    };

-    encode_r_type_bits(
-        op.opcode(),
-        reg_to_gpr_num(vd.to_reg()),
-        op.funct3(),
-        vs1,
-        vs2,
-        funct7,
-    )
+    encode_r_type_bits(op.opcode(), reg_to_gpr_num(vd.to_reg()), op.funct3(), vs1, vs2, funct7)
 }

 pub fn encode_valu_r_imm(
@ -222,14 +217,7 @@ pub fn encode_valu_r_imm(
    let vs1 = imm.bits() as u32;
    let vs2 = op.aux_encoding();

-    encode_r_type_bits(
-        op.opcode(),
-        reg_to_gpr_num(vd.to_reg()),
-        op.funct3(),
-        vs1,
-        vs2,
-        funct7,
-    )
+    encode_r_type_bits(op.opcode(), reg_to_gpr_num(vd.to_reg()), op.funct3(), vs1, vs2, funct7)
 }

 /// Encodes a Vector CFG Imm instruction.
--- a/hbcb/src/inst/regs.rs
+++ b/hbcb/src/inst/regs.rs
@ -1,12 +1,11 @@
 //! Riscv64 ISA definitions: registers.
 //!

-use crate::machinst::{Reg, Writable};
-
-use alloc::vec;
-use alloc::vec::Vec;
-
-use regalloc2::{PReg, RegClass, VReg};
+use {
+    alloc::{vec, vec::Vec},
+    cranelift_codegen::machinst::{Reg, Writable},
+    regalloc2::{PReg, RegClass, VReg},
+};

 // first argument of function call
 #[inline]
--- a/hbcb/src/inst/vector.rs
+++ b/hbcb/src/inst/vector.rs
@ -1,19 +1,20 @@
-use crate::lower::isle::generated_code::VecAluOpRRRR;
-use crate::lower::isle::generated_code::{
-    VecAMode, VecAluOpRImm5, VecAluOpRR, VecAluOpRRImm5, VecAluOpRRR, VecAluOpRRRImm5, VecAvl,
-    VecElementWidth, VecLmul, VecMaskMode, VecOpCategory, VecOpMasking, VecTailMode,
+use {
+    super::{Type, UImm5},
+    crate::{
+        lower::isle::generated_code::{
+            VecAMode, VecAluOpRImm5, VecAluOpRR, VecAluOpRRImm5, VecAluOpRRR, VecAluOpRRRImm5,
+            VecAluOpRRRR, VecAvl, VecElementWidth, VecLmul, VecMaskMode, VecOpCategory,
+            VecOpMasking, VecTailMode,
+        },
+        Reg,
+    },
+    core::fmt,
+    cranelift_codegen::machinst::{OperandVisitor, RegClass},
 };
-use crate::machinst::{OperandVisitor, RegClass};
-use crate::Reg;
-use core::fmt;
-
-use super::{Type, UImm5};

 impl VecAvl {
    pub fn _static(size: u32) -> Self {
-        VecAvl::Static {
-            size: UImm5::maybe_from_u8(size as u8).expect("Invalid size for AVL"),
-        }
+        VecAvl::Static { size: UImm5::maybe_from_u8(size as u8).expect("Invalid size for AVL") }
    }

    pub fn is_static(&self) -> bool {
@ -178,11 +179,7 @@ impl VType {

 impl fmt::Display for VType {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        write!(
-            f,
-            "{}, {}, {}, {}",
-            self.sew, self.lmul, self.tail_mode, self.mask_mode
-        )
+        write!(f, "{}, {}, {}, {}", self.sew, self.lmul, self.tail_mode, self.mask_mode)
    }
 }

@ -255,6 +252,7 @@ impl VecAluOpRRRR {
        // Vector Opcode
        0x57
    }
+
    pub fn funct3(&self) -> u32 {
        self.category().encode()
    }
@ -323,6 +321,7 @@ impl VecAluOpRRRImm5 {
        // Vector Opcode
        0x57
    }
+
    pub fn funct3(&self) -> u32 {
        self.category().encode()
    }
@ -369,9 +368,11 @@ impl VecAluOpRRR {
        // Vector Opcode
        0x57
    }
+
    pub fn funct3(&self) -> u32 {
        self.category().encode()
    }
+
    pub fn funct6(&self) -> u32 {
        // See: https://github.com/riscv/riscv-v-spec/blob/master/inst-table.adoc
        match self {
@ -658,6 +659,7 @@ impl VecAluOpRRImm5 {
        // Vector Opcode
        0x57
    }
+
    pub fn funct3(&self) -> u32 {
        self.category().encode()
    }
@ -1016,6 +1018,7 @@ impl VecAluOpRImm5 {
        // Vector Opcode
        0x57
    }
+
    pub fn funct3(&self) -> u32 {
        self.category().encode()
    }
--- a/hblang/Cargo.toml
+++ b/hblang/Cargo.toml
@ -8,4 +8,6 @@ name = "hbc"
 path = "src/main.rs"

 [dependencies]
+hbbytecode = { version = "0.1.0", path = "../hbbytecode" }
 hbvm = { path = "../hbvm", features = ["nightly"] }
+regalloc2 = { git = "https://github.com/jakubDoka/regalloc2" }
--- a/hblang/README.md
+++ b/hblang/README.md
@ -237,7 +237,7 @@ main := fn(): int {
 	size_of_Type_in_bytes := @sizeof(foo.Type)
 	align_of_Type_in_bytes := @alignof(foo.Type)
 	hardcoded_pointer := @as(^u8, @bitcast(10))
-	ecall_that_returns_int := @eca(int, 1, foo.Type.(10, 20), 5, 6)
+	ecall_that_returns_int := @as(int, @eca(1, foo.Type.(10, 20), 5, 6))
 	return @inline(foo.foo)
 }

@ -356,6 +356,29 @@ main := fn(): int {
 }
 ```

+#### wide_ret
+```hb
+OemIdent := struct {
+	dos_version: [u8; 8],
+	dos_version_name: [u8; 8],
+}
+
+Stru := struct {
+	a: u16,
+	b: u16,
+}
+
+small_struct := fn(): Stru {
+	return .{a: 0, b: 0}
+}
+
+main := fn(major: int, minor: int): OemIdent {
+	small_struct()
+	ver := [u8].(0, 0, 0, 0, 0, 0, 0, 0)
+	return OemIdent.(ver, ver)
+}
+```
+
 ### Incomplete Examples

 #### comptime_pointers
@ -376,8 +399,8 @@ modify := fn($num: ^int): void {
 MALLOC_SYS_CALL := 69
 FREE_SYS_CALL := 96

-malloc := fn(size: uint, align: uint): ^void return @eca(^void, MALLOC_SYS_CALL, size, align)
-free := fn(ptr: ^void, size: uint, align: uint): void return @eca(void, FREE_SYS_CALL, ptr, size, align)
+malloc := fn(size: uint, align: uint): ^void return @eca(MALLOC_SYS_CALL, size, align)
+free := fn(ptr: ^void, size: uint, align: uint): void return @eca(FREE_SYS_CALL, ptr, size, align)

 Vec := fn($Elem: type): type {
 	return struct {
@ -750,7 +773,7 @@ screenidx := fn(orange: int): int {
 // in module: random.hb

 integer := fn(min: int, max: int): int {
-	rng := @eca(int, 3, 4)
+	rng := @as(int, @eca(3, 4))

 	if min != 0 | max != 0 {
 		return rng % (max - min + 1) + min
@ -781,7 +804,7 @@ main := fn(): void {

 // in module: random.hb
 integer_range := fn(min: uint, max: int): uint {
-	return @eca(uint, 3, 4) % (@bitcast(max) - min + 1) + min
+	return @eca(3, 4) % (@bitcast(max) - min + 1) + min
 }
 ```

@ -876,7 +899,7 @@ request_page := fn(page_count: u8): ^u8 {
 	msg := "\{00}\{01}xxxxxxxx\0"
 	msg_page_count := msg + 1;
 	*msg_page_count = page_count
-	return @eca(^u8, 3, 2, msg, 12)
+	return @eca(3, 2, msg, 12)
 }

 create_back_buffer := fn(total_pages: int): ^u32 {
--- a/hblang/build.rs
+++ b/hblang/build.rs
@ -1,183 +0,0 @@
-#![feature(iter_next_chunk)]
-use std::{collections::HashSet, fmt::Write};
-
-fn main() -> Result<(), Box<dyn std::error::Error>> {
-    println!("cargo:rerun-if-changed=build.rs");
-    println!("cargo:rerun-if-changed=../hbbytecode/instructions.in");
-
-    gen_instrs()?;
-
-    Ok(())
-}
-
-fn gen_instrs() -> Result<(), Box<dyn std::error::Error>> {
-    let mut generated = String::new();
-
-    writeln!(generated, "#![allow(dead_code)] #![allow(clippy::upper_case_acronyms)]")?;
-
-    '_max_size: {
-        let max = instructions()
-            .map(
-                |[_, _, ty, _]| {
-                    if ty == "N" {
-                        1
-                    } else {
-                        iter_args(ty).map(arg_to_width).sum::<usize>() + 1
-                    }
-                },
-            )
-            .max()
-            .unwrap();
-
-        writeln!(generated, "pub const MAX_SIZE: usize = {max};")?;
-    }
-
-    '_encoders: {
-        for [op, name, ty, doc] in instructions() {
-            writeln!(generated, "/// {}", doc.trim_matches('"'))?;
-            let name = name.to_lowercase();
-            let args = comma_sep(
-                iter_args(ty)
-                    .enumerate()
-                    .map(|(i, c)| format!("{}{i}: {}", arg_to_name(c), arg_to_type(c))),
-            );
-            writeln!(generated, "pub fn {name}({args}) -> (usize, [u8; MAX_SIZE]) {{")?;
-            let arg_names =
-                comma_sep(iter_args(ty).enumerate().map(|(i, c)| format!("{}{i}", arg_to_name(c))));
-            writeln!(generated, "    unsafe {{ crate::encode({ty}({op}, {arg_names})) }}")?;
-            writeln!(generated, "}}")?;
-        }
-    }
-
-    '_structs: {
-        let mut seen = std::collections::HashSet::new();
-        for [_, _, ty, _] in instructions() {
-            if !seen.insert(ty) {
-                continue;
-            }
-            let types = comma_sep(iter_args(ty).map(arg_to_type).map(|s| s.to_string()));
-            writeln!(generated, "#[repr(packed)] pub struct {ty}(u8, {types});")?;
-        }
-    }
-
-    '_name_list: {
-        writeln!(generated, "pub const NAMES: [&str; {}] = [", instructions().count())?;
-        for [_, name, _, _] in instructions() {
-            writeln!(generated, "    \"{}\",", name.to_lowercase())?;
-        }
-        writeln!(generated, "];")?;
-    }
-
-    let instr = "Instr";
-    let oper = "Oper";
-
-    '_instr_enum: {
-        writeln!(generated, "#[derive(Debug, Clone, Copy, PartialEq, Eq)] #[repr(u8)]")?;
-        writeln!(generated, "pub enum {instr} {{")?;
-        for [id, name, ..] in instructions() {
-            writeln!(generated, "    {name} = {id},")?;
-        }
-        writeln!(generated, "}}")?;
-    }
-
-    '_arg_kind: {
-        writeln!(generated, "#[derive(Debug, Clone, Copy, PartialEq, Eq)]")?;
-        writeln!(generated, "pub enum {oper} {{")?;
-        let mut seen = HashSet::new();
-        for ty in instructions().flat_map(|[.., ty, _]| iter_args(ty)) {
-            if !seen.insert(ty) {
-                continue;
-            }
-            writeln!(generated, "    {ty}({}),", arg_to_type(ty))?;
-        }
-        writeln!(generated, "}}")?;
-    }
-
-    '_parse_opers: {
-        writeln!(
-            generated,
-            "/// This assumes the instruction byte is still at the beginning of the buffer"
-        )?;
-        writeln!(generated, "pub fn parse_args(bytes: &mut &[u8], kind: {instr}, buf: &mut Vec<{oper}>) -> Option<()> {{")?;
-        writeln!(generated, "    match kind {{")?;
-        let mut instrs = instructions().collect::<Vec<_>>();
-        instrs.sort_unstable_by_key(|&[.., ty, _]| ty);
-        for group in instrs.chunk_by(|[.., a, _], [.., b, _]| a == b) {
-            let ty = group[0][2];
-            for &[_, name, ..] in group {
-                writeln!(generated, "        | {instr}::{name}")?;
-            }
-            generated.pop();
-            writeln!(generated, " => {{")?;
-            if iter_args(ty).count() != 0 {
-                writeln!(generated, "            let data = crate::decode::<{ty}>(bytes)?;")?;
-                writeln!(
-                    generated,
-                    "            buf.extend([{}]);",
-                    comma_sep(
-                        iter_args(ty).zip(1u32..).map(|(t, i)| format!("{oper}::{t}(data.{i})"))
-                    )
-                )?;
-            } else {
-                writeln!(generated, "            crate::decode::<{ty}>(bytes)?;")?;
-            }
-
-            writeln!(generated, "        }}")?;
-        }
-        writeln!(generated, "    }}")?;
-        writeln!(generated, "    Some(())")?;
-        writeln!(generated, "}}")?;
-    }
-
-    std::fs::write("src/instrs.rs", generated)?;
-    Ok(())
-}
-
-fn comma_sep(items: impl Iterator<Item = String>) -> String {
-    items.map(|item| item.to_string()).collect::<Vec<_>>().join(", ")
-}
-
-fn instructions() -> impl Iterator<Item = [&'static str; 4]> {
-    include_str!("../hbbytecode/instructions.in")
-        .lines()
-        .filter_map(|line| line.strip_suffix(';'))
-        .map(|line| line.splitn(4, ',').map(str::trim).next_chunk().unwrap())
-}
-
-fn arg_to_type(arg: char) -> &'static str {
-    match arg {
-        'R' | 'B' => "u8",
-        'H' => "u16",
-        'W' => "u32",
-        'D' | 'A' => "u64",
-        'P' => "i16",
-        'O' => "i32",
-        _ => panic!("unknown type: {}", arg),
-    }
-}
-
-fn arg_to_width(arg: char) -> usize {
-    match arg {
-        'R' | 'B' => 1,
-        'H' => 2,
-        'W' => 4,
-        'D' | 'A' => 8,
-        'P' => 2,
-        'O' => 4,
-        _ => panic!("unknown type: {}", arg),
-    }
-}
-
-fn arg_to_name(arg: char) -> &'static str {
-    match arg {
-        'R' => "reg",
-        'B' | 'H' | 'W' | 'D' => "imm",
-        'P' | 'O' => "offset",
-        'A' => "addr",
-        _ => panic!("unknown type: {}", arg),
-    }
-}
-
-fn iter_args(ty: &'static str) -> impl Iterator<Item = char> {
-    ty.chars().filter(|c| *c != 'N')
-}
--- a/hblang/src/codegen.rs
+++ b/hblang/src/codegen.rs
@ -898,13 +898,21 @@ impl Codegen {
            E::Directive { name: "TypeOf", args: [expr], .. } => {
                Some(Value::ty(self.infer_type(expr)))
            }
-            E::Directive { name: "eca", args: [ret_ty, args @ ..], .. } => {
-                let ty = self.ty(ret_ty);
+            E::Directive { name: "eca", args, pos } => {
+                let Some(ty) = ctx.ty else {
+                    self.report(
+                        pos,
+                        "type to return form eca is unknown, use `@as(<type>, @eca(...<expr>))`",
+                    );
+                };

                let mut parama = self.tys.parama(ty);
                let base = self.pool.arg_locs.len();
                for arg in args {
                    let arg = self.expr(arg)?;
+                    if arg.ty == ty::Id::from(ty::TYPE) {
+                        self.report(pos, "na na na nana, no passing types to ecas");
+                    }
                    self.pass_arg(&arg, &mut parama);
                    self.pool.arg_locs.push(arg.loc);
                }
@ -932,7 +940,7 @@ impl Codegen {
                let Some(ty) = ctx.ty else {
                    self.report(
                        expr.pos(),
-                        "type to cast to is unknown, use `@as(<type>, <expr>)`",
+                        "type to cast to is unknown, use `@as(<type>, @intcast(<expr>))`",
                    );
                };
                let mut val = self.expr(val)?;
@ -1086,7 +1094,7 @@ impl Codegen {
                Some(Value::new(self.tys.make_ptr(ty::U8.into()), reg))
            }
            E::Ctor { pos, ty, fields, .. } => {
-                let (ty, loc) = self.prepare_struct_ctor(pos, ctx, ty, fields.len());
+                let (ty, loc) = self.prepare_struct_ctor(pos, &mut ctx, ty, fields.len());

                let ty::Kind::Struct(stru) = ty.expand() else {
                    self.report(
@ -1103,10 +1111,16 @@ impl Codegen {
                    let value = self.expr_ctx(value, Ctx::default().with_loc(loc).with_ty(ty))?;
                    self.ci.free_loc(value.loc);
                }
-                return Some(Value { ty, loc });
+
+                if let Some(dst_loc) = ctx.loc {
+                    self.store_typed(loc, &dst_loc, ty);
+                    return Some(Value { ty, loc: dst_loc });
+                } else {
+                    return Some(Value { ty, loc });
+                }
            }
            E::Tupl { pos, ty, fields, .. } => {
-                let (ty, loc) = self.prepare_struct_ctor(pos, ctx, ty, fields.len());
+                let (ty, loc) = self.prepare_struct_ctor(pos, &mut ctx, ty, fields.len());

                match ty.expand() {
                    ty::Kind::Struct(stru) => {
@ -1141,7 +1155,12 @@ impl Codegen {
                    ),
                }

-                return Some(Value { ty, loc });
+                if let Some(dst_loc) = ctx.loc {
+                    self.store_typed(loc, &dst_loc, ty);
+                    return Some(Value { ty, loc: dst_loc });
+                } else {
+                    return Some(Value { ty, loc });
+                }
            }
            E::Field { target, name: field } => {
                let checkpoint = self.ci.snap();
@ -1369,24 +1388,24 @@ impl Codegen {
                tk => Some(Value::ty(tk.compress())),
            },
            E::Return { pos, val, .. } => {
-                let ty = if let Some(val) = val {
-                    let size = self.ci.ret.map_or(17, |ty| self.tys.size_of(ty));
-                    let loc = match size {
-                        _ if self.ci.inline_ret_loc != Loc::default() => {
-                            Some(self.ci.inline_ret_loc.as_ref())
-                        }
-                        0 => None,
-                        1..=16 => Some(Loc::reg(1)),
-                        _ => Some(Loc::reg(self.ci.ret_reg.as_ref()).into_derefed()),
-                    };
-                    self.expr_ctx(val, Ctx { ty: self.ci.ret, loc })?.ty
+                let size = self.ci.ret.map_or(17, |ty| self.tys.size_of(ty));
+                let loc = match size {
+                    _ if self.ci.inline_ret_loc != Loc::default() => {
+                        Some(self.ci.inline_ret_loc.as_ref())
+                    }
+                    0 => None,
+                    1..=16 => Some(Loc::reg(1)),
+                    _ => Some(Loc::reg(self.ci.ret_reg.as_ref()).into_derefed()),
+                };
+                let value = if let Some(val) = val {
+                    self.expr_ctx(val, Ctx { ty: self.ci.ret, loc })?
                } else {
-                    ty::VOID.into()
+                    Value::void()
                };

                match self.ci.ret {
-                    None => self.ci.ret = Some(ty),
-                    Some(ret) => _ = self.assert_ty(pos, ty, ret, "return type"),
+                    None => self.ci.ret = Some(value.ty),
+                    Some(ret) => _ = self.assert_ty(pos, value.ty, ret, "return type"),
                }

                self.ci.ret_relocs.push(Reloc::new(self.ci.code.len(), 1, 4));
@ -1891,7 +1910,7 @@ impl Codegen {
    fn prepare_struct_ctor(
        &mut self,
        pos: Pos,
-        ctx: Ctx,
+        ctx: &mut Ctx,
        ty: Option<&Expr>,
        field_len: usize,
    ) -> (ty::Id, Loc) {
@ -1927,8 +1946,11 @@ impl Codegen {
        }

        let size = self.tys.size_of(ty);
-        let loc = ctx.loc.unwrap_or_else(|| Loc::stack(self.ci.stack.allocate(size)));
-        (ty, loc)
+        if ctx.loc.as_ref().map_or(true, |l| l.is_reg()) {
+            (ty, Loc::stack(self.ci.stack.allocate(size)))
+        } else {
+            (ty, ctx.loc.take().unwrap_or_else(|| Loc::stack(self.ci.stack.allocate(size))))
+        }
    }

    fn struct_op(
@ -2278,6 +2300,25 @@ impl Codegen {
                    self.ci.emit(cp(dst.get(), src.get()));
                }
            }
+            //(lpat!(false, src, 0, None), lpat!(false, dst, off, None)) => {
+            //    assert!(size <= 8);
+            //    let off_rem = 8 * (off % 8);
+            //    let freg = dst.get() + (off / 8) as u8;
+            //    if size < 8 {
+            //        let mask = !(((1u64 << (8 * size)) - 1) << off_rem);
+            //        self.ci.emit(andi(freg, freg, mask));
+            //        if off_rem == 0 {
+            //            self.ci.emit(or(freg, freg, src.get()));
+            //        } else {
+            //            let tmp = self.ci.regs.allocate();
+            //            self.ci.emit(slui64(tmp.get(), src.get(), off_rem as _));
+            //            self.ci.emit(or(freg, freg, src.get()));
+            //            self.ci.regs.free(tmp);
+            //        }
+            //    } else {
+            //        self.ci.emit(cp(freg, src.get()));
+            //    }
+            //}
            (lpat!(true, src, soff, ref ssta), lpat!(false, dst, 0, None)) => {
                if size < 8 {
                    self.ci.emit(cp(dst.get(), 0));
@ -2740,5 +2781,6 @@ mod tests {
        writing_into_string => README;
        request_page => README;
        tests_ptr_to_ptr_copy => README;
+        wide_ret => README;
    }
 }
--- a/hblang/src/lexer.rs
+++ b/hblang/src/lexer.rs
@ -371,6 +371,10 @@ impl<'a> Lexer<'a> {
        Self { pos, bytes: input.as_bytes() }
    }

+    pub fn source(&self) -> &'a str {
+        unsafe { std::str::from_utf8_unchecked(self.bytes) }
+    }
+
    pub fn slice(&self, tok: std::ops::Range<usize>) -> &'a str {
        unsafe { std::str::from_utf8_unchecked(&self.bytes[tok]) }
    }
@ -511,10 +515,6 @@ impl<'a> Lexer<'a> {
            false
        }
    }
-
-    pub fn line_col(&self, pos: u32) -> (usize, usize) {
-        line_col(self.bytes, pos)
-    }
 }

 pub fn line_col(bytes: &[u8], pos: u32) -> (usize, usize) {
--- a/hblang/src/lib.rs
+++ b/hblang/src/lib.rs
@ -27,10 +27,11 @@ use {
        son::reg,
        ty::ArrayLen,
    },
+    hbbytecode as instrs,
    parser::Ast,
    std::{
        collections::{hash_map, BTreeMap, VecDeque},
-        io::{self, Read},
+        io,
        ops::Range,
        path::{Path, PathBuf},
        rc::Rc,
@ -52,7 +53,6 @@ pub mod codegen;
 pub mod parser;
 pub mod son;

-mod instrs;
 mod lexer;

 mod task {
@ -601,6 +601,19 @@ impl ParamAlloc {
    }
 }

+#[repr(packed)]
+#[allow(dead_code)]
+struct AbleOsExecutableHeader {
+    magic_number: [u8; 3],
+    executable_version: u32,
+
+    code_length: u64,
+    data_length: u64,
+    debug_length: u64,
+    config_length: u64,
+    metadata_length: u64,
+}
+
 #[derive(Default)]
 struct Types {
    syms: HashMap<SymKey, ty::Id>,
@ -613,15 +626,24 @@ struct Types {
    arrays: Vec<Array>,
 }

+const HEADER_SIZE: usize = std::mem::size_of::<AbleOsExecutableHeader>();
+
 impl Types {
    fn assemble(&mut self, to: &mut Vec<u8>) {
+        to.extend([0u8; HEADER_SIZE]);
+
        emit(to, instrs::jal(reg::RET_ADDR, reg::ZERO, 0));
        emit(to, instrs::tx());
-        self.dump_reachable(0, to);
-        Reloc::new(0, 3, 4).apply_jump(to, self.funcs[0].offset, 0);
+        let exe = self.dump_reachable(0, to);
+        Reloc::new(HEADER_SIZE, 3, 4).apply_jump(to, self.funcs[0].offset, 0);
+
+        unsafe { *to.as_mut_ptr().cast::<AbleOsExecutableHeader>() = exe }
    }

-    fn dump_reachable(&mut self, from: ty::Func, to: &mut Vec<u8>) {
+    fn dump_reachable(&mut self, from: ty::Func, to: &mut Vec<u8>) -> AbleOsExecutableHeader {
+        let mut used_funcs = vec![];
+        let mut used_globals = vec![];
+
        let mut frontier = vec![ty::Kind::Func(from).compress()];

        while let Some(itm) = frontier.pop() {
@ -631,8 +653,8 @@ impl Types {
                    if task::is_done(fuc.offset) {
                        continue;
                    }
-                    fuc.offset = to.len() as _;
-                    to.extend(&fuc.code);
+                    fuc.offset = 0;
+                    used_funcs.push(func);
                    frontier.extend(fuc.relocs.iter().map(|r| r.target));
                }
                ty::Kind::Global(glob) => {
@ -640,18 +662,31 @@ impl Types {
                    if task::is_done(glb.offset) {
                        continue;
                    }
-                    glb.offset = to.len() as _;
-                    to.extend(&glb.data);
+                    glb.offset = 0;
+                    used_globals.push(glob);
                }
                _ => unreachable!(),
            }
        }

-        for fuc in &self.funcs {
-            if !task::is_done(fuc.offset) {
-                continue;
-            }
+        for &func in &used_funcs {
+            let fuc = &mut self.funcs[func as usize];
+            fuc.offset = to.len() as _;
+            to.extend(&fuc.code);
+        }

+        let code_length = to.len();
+
+        for &global in &used_globals {
+            let global = &mut self.globals[global as usize];
+            global.offset = to.len() as _;
+            to.extend(&global.data);
+        }
+
+        let data_length = to.len() - code_length;
+
+        for func in used_funcs {
+            let fuc = &self.funcs[func as usize];
            for rel in &fuc.relocs {
                let offset = match rel.target.expand() {
                    ty::Kind::Func(fun) => self.funcs[fun as usize].offset,
@ -661,6 +696,16 @@ impl Types {
                rel.reloc.apply_jump(to, offset, fuc.offset);
            }
        }
+
+        AbleOsExecutableHeader {
+            magic_number: [0x15, 0x91, 0xD2],
+            executable_version: 0,
+            code_length: (code_length - HEADER_SIZE) as _,
+            data_length: data_length as _,
+            debug_length: 0,
+            config_length: 0,
+            metadata_length: 0,
+        }
    }

    pub fn disasm(
@ -670,7 +715,7 @@ impl Types {
        output: &mut impl std::io::Write,
        eca_handler: impl FnMut(&mut &[u8]),
    ) -> std::io::Result<()> {
-        use crate::DisasmItem;
+        use instrs::DisasmItem;
        let functions = self
            .funcs
            .iter()
@ -699,7 +744,7 @@ impl Types {
                (g.offset, (name, g.data.len() as Size, DisasmItem::Global))
            }))
            .collect::<BTreeMap<_, _>>();
-        crate::disasm(&mut sluce, &functions, output, eca_handler)
+        instrs::disasm(&mut sluce, &functions, output, eca_handler)
    }

    fn parama(&self, ret: impl Into<ty::Id>) -> ParamAlloc {
@ -812,178 +857,6 @@ impl Types {
    }
 }

-#[inline]
-unsafe fn encode<T>(instr: T) -> (usize, [u8; instrs::MAX_SIZE]) {
-    let mut buf = [0; instrs::MAX_SIZE];
-    std::ptr::write(buf.as_mut_ptr() as *mut T, instr);
-    (std::mem::size_of::<T>(), buf)
-}
-
-#[inline]
-fn decode<T>(binary: &mut &[u8]) -> Option<T> {
-    unsafe { Some(std::ptr::read(binary.take(..std::mem::size_of::<T>())?.as_ptr() as *const T)) }
-}
-
-#[derive(Clone, Copy)]
-enum DisasmItem {
-    Func,
-    Global,
-}
-
-fn disasm(
-    binary: &mut &[u8],
-    functions: &BTreeMap<u32, (&str, u32, DisasmItem)>,
-    out: &mut impl std::io::Write,
-    mut eca_handler: impl FnMut(&mut &[u8]),
-) -> std::io::Result<()> {
-    use self::instrs::Instr;
-
-    fn instr_from_byte(b: u8) -> std::io::Result<Instr> {
-        if b as usize >= instrs::NAMES.len() {
-            return Err(std::io::ErrorKind::InvalidData.into());
-        }
-        Ok(unsafe { std::mem::transmute::<u8, Instr>(b) })
-    }
-
-    let mut labels = HashMap::<u32, u32>::default();
-    let mut buf = Vec::<instrs::Oper>::new();
-    let mut has_cycle = false;
-    let mut has_oob = false;
-
-    '_offset_pass: for (&off, &(_name, len, kind)) in functions.iter() {
-        if matches!(kind, DisasmItem::Global) {
-            continue;
-        }
-
-        let prev = *binary;
-
-        binary.take(..off as usize).unwrap();
-
-        let mut label_count = 0;
-        while let Some(&byte) = binary.first() {
-            let offset: i32 = (prev.len() - binary.len()).try_into().unwrap();
-            if offset as u32 == off + len {
-                break;
-            }
-            let Ok(inst) = instr_from_byte(byte) else { break };
-            instrs::parse_args(binary, inst, &mut buf).ok_or(std::io::ErrorKind::OutOfMemory)?;
-
-            for op in buf.drain(..) {
-                let rel = match op {
-                    instrs::Oper::O(rel) => rel,
-                    instrs::Oper::P(rel) => rel.into(),
-                    _ => continue,
-                };
-
-                has_cycle |= rel == 0;
-
-                let global_offset: u32 = (offset + rel).try_into().unwrap();
-                if functions.get(&global_offset).is_some() {
-                    continue;
-                }
-                label_count += labels.try_insert(global_offset, label_count).is_ok() as u32;
-            }
-
-            if matches!(inst, Instr::ECA) {
-                eca_handler(binary);
-            }
-        }
-
-        *binary = prev;
-    }
-
-    '_dump: for (&off, &(name, len, kind)) in functions.iter() {
-        if matches!(kind, DisasmItem::Global) {
-            continue;
-        }
-        let prev = *binary;
-
-        writeln!(out, "{name}:")?;
-
-        binary.take(..off as usize).unwrap();
-        while let Some(&byte) = binary.first() {
-            let offset: i32 = (prev.len() - binary.len()).try_into().unwrap();
-            if offset as u32 == off + len {
-                break;
-            }
-            let Ok(inst) = instr_from_byte(byte) else {
-                writeln!(out, "invalid instr {byte}")?;
-                break;
-            };
-            instrs::parse_args(binary, inst, &mut buf).unwrap();
-
-            if let Some(label) = labels.get(&offset.try_into().unwrap()) {
-                write!(out, "{:>2}: ", label)?;
-            } else {
-                write!(out, "    ")?;
-            }
-
-            write!(out, "{inst:<8?} ")?;
-
-            'a: for (i, op) in buf.drain(..).enumerate() {
-                if i != 0 {
-                    write!(out, ", ")?;
-                }
-
-                let rel = 'b: {
-                    match op {
-                        instrs::Oper::O(rel) => break 'b rel,
-                        instrs::Oper::P(rel) => break 'b rel.into(),
-                        instrs::Oper::R(r) => write!(out, "r{r}")?,
-                        instrs::Oper::B(b) => write!(out, "{b}b")?,
-                        instrs::Oper::H(h) => write!(out, "{h}h")?,
-                        instrs::Oper::W(w) => write!(out, "{w}w")?,
-                        instrs::Oper::D(d) if (d as i64) < 0 => write!(out, "{}d", d as i64)?,
-                        instrs::Oper::D(d) => write!(out, "{d}d")?,
-                        instrs::Oper::A(a) => write!(out, "{a}a")?,
-                    }
-
-                    continue 'a;
-                };
-
-                let global_offset: u32 = (offset + rel).try_into().unwrap();
-                if let Some(&(name, ..)) = functions.get(&global_offset) {
-                    if name.contains('\0') {
-                        write!(out, ":{name:?}")?;
-                    } else {
-                        write!(out, ":{name}")?;
-                    }
-                } else {
-                    let local_has_oob = global_offset < off
-                        || global_offset > off + len
-                        || instr_from_byte(prev[global_offset as usize]).is_err()
-                        || prev[global_offset as usize] == 0;
-                    has_oob |= local_has_oob;
-                    let label = labels.get(&global_offset).unwrap();
-                    if local_has_oob {
-                        write!(out, "!!!!!!!!!{rel}")?;
-                    } else {
-                        write!(out, ":{label}")?;
-                    }
-                }
-            }
-
-            writeln!(out)?;
-
-            if matches!(inst, Instr::ECA) {
-                eca_handler(binary);
-            }
-        }
-
-        *binary = prev;
-    }
-
-    if has_oob {
-        return Err(std::io::ErrorKind::InvalidInput.into());
-    }
-
-    if has_cycle {
-        return Err(std::io::ErrorKind::TimedOut.into());
-    }
-
-    Ok(())
-}
-
 struct TaskQueue<T> {
    inner: Mutex<TaskQueueInner<T>>,
 }
@ -1070,107 +943,24 @@ impl<T> TaskQueueInner<T> {
 }

 pub fn parse_from_fs(extra_threads: usize, root: &str) -> io::Result<Vec<Ast>> {
-    const GIT_DEPS_DIR: &str = "git-deps";
+    fn resolve(path: &str, from: &str) -> Result<PathBuf, CantLoadFile> {
+        let path = match Path::new(from).parent() {
+            Some(parent) => PathBuf::from_iter([parent, Path::new(path)]),
+            None => PathBuf::from(path),
+        };

-    enum Chk<'a> {
-        Branch(&'a str),
-        Rev(&'a str),
-        Tag(&'a str),
-    }
-
-    enum ImportPath<'a> {
-        Rel { path: &'a str },
-        Git { link: &'a str, path: &'a str, chk: Option<Chk<'a>> },
-    }
-
-    impl<'a> TryFrom<&'a str> for ImportPath<'a> {
-        type Error = ParseImportError;
-
-        fn try_from(value: &'a str) -> Result<Self, Self::Error> {
-            let (prefix, path) = value.split_once(':').unwrap_or(("", value));
-
-            match prefix {
-                "rel" | "" => Ok(Self::Rel { path }),
-                "git" => {
-                    let (link, path) =
-                        path.split_once(':').ok_or(ParseImportError::ExpectedPath)?;
-                    let (link, params) = link.split_once('?').unwrap_or((link, ""));
-                    let chk = params.split('&').filter_map(|s| s.split_once('=')).find_map(
-                        |(key, value)| match key {
-                            "branch" => Some(Chk::Branch(value)),
-                            "rev" => Some(Chk::Rev(value)),
-                            "tag" => Some(Chk::Tag(value)),
-                            _ => None,
-                        },
-                    );
-                    Ok(Self::Git { link, path, chk })
-                }
-                _ => Err(ParseImportError::InvalidPrefix),
-            }
-        }
-    }
-
-    fn preprocess_git(link: &str) -> &str {
-        let link = link.strip_prefix("https://").unwrap_or(link);
-        link.strip_suffix(".git").unwrap_or(link)
-    }
-
-    impl<'a> ImportPath<'a> {
-        fn resolve(&self, from: &str) -> Result<PathBuf, CantLoadFile> {
-            let path = match self {
-                Self::Rel { path } => match Path::new(from).parent() {
-                    Some(parent) => PathBuf::from_iter([parent, Path::new(path)]),
-                    None => PathBuf::from(path),
-                },
-                Self::Git { path, link, .. } => {
-                    let link = preprocess_git(link);
-                    PathBuf::from_iter([GIT_DEPS_DIR, link, path])
-                }
-            };
-
-            path.canonicalize().map_err(|source| CantLoadFile {
-                path,
-                from: PathBuf::from(from),
-                source,
-            })
-        }
-    }
-
-    #[derive(Debug)]
-    enum ParseImportError {
-        ExpectedPath,
-        InvalidPrefix,
-    }
-
-    impl std::fmt::Display for ParseImportError {
-        fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
-            match self {
-                Self::ExpectedPath => "expected path".fmt(f),
-                Self::InvalidPrefix => "invalid prefix, expected one of rel, \
-                    git or none followed by colon"
-                    .fmt(f),
-            }
-        }
-    }
-
-    impl std::error::Error for ParseImportError {}
-
-    impl From<ParseImportError> for io::Error {
-        fn from(e: ParseImportError) -> Self {
-            io::Error::new(io::ErrorKind::InvalidInput, e)
-        }
+        path.canonicalize().map_err(|source| CantLoadFile { path, source })
    }

    #[derive(Debug)]
    struct CantLoadFile {
        path: PathBuf,
-        from: PathBuf,
        source: io::Error,
    }

    impl std::fmt::Display for CantLoadFile {
        fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
-            write!(f, "can't load file: {} (from: {})", self.path.display(), self.from.display(),)
+            write!(f, "can't load file: {}", parser::display_rel_path(&self.path),)
        }
    }

@ -1186,37 +976,22 @@ pub fn parse_from_fs(extra_threads: usize, root: &str) -> io::Result<Vec<Ast>> {
        }
    }

-    #[derive(Debug)]
-    struct InvalidFileData(std::str::Utf8Error);
-
-    impl std::fmt::Display for InvalidFileData {
-        fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
-            write!(f, "invalid file data")
-        }
-    }
-
-    impl std::error::Error for InvalidFileData {
-        fn source(&self) -> Option<&(dyn std::error::Error + 'static)> {
-            Some(&self.0)
-        }
-    }
-
-    impl From<InvalidFileData> for io::Error {
-        fn from(e: InvalidFileData) -> Self {
-            io::Error::new(io::ErrorKind::InvalidData, e)
-        }
-    }
-
-    type Task = (u32, PathBuf, Option<std::process::Command>);
+    type Task = (u32, PathBuf);

    let seen = Mutex::new(HashMap::<PathBuf, u32>::default());
    let tasks = TaskQueue::<Task>::new(extra_threads + 1);
    let ast = Mutex::new(Vec::<io::Result<Ast>>::new());

    let loader = |path: &str, from: &str| {
-        let path = ImportPath::try_from(path)?;
+        if path.starts_with("rel:") {
+            return Err(io::Error::new(
+                io::ErrorKind::Other,
+                "`rel:` prefix was removed and is now equivalent to no prefix (remove it)"
+                    .to_string(),
+            ));
+        }

-        let physiscal_path = path.resolve(from)?;
+        let physiscal_path = resolve(path, from)?;

        let id = {
            let mut seen = seen.lock().unwrap();
@ -1232,63 +1007,30 @@ pub fn parse_from_fs(extra_threads: usize, root: &str) -> io::Result<Vec<Ast>> {
            }
        };

-        let command = if !physiscal_path.exists() {
-            let ImportPath::Git { link, chk, .. } = path else {
-                return Err(io::Error::new(
-                    io::ErrorKind::NotFound,
-                    format!("can't find file: {}", physiscal_path.display()),
-                ));
-            };
+        if !physiscal_path.exists() {
+            return Err(io::Error::new(
+                io::ErrorKind::NotFound,
+                format!("can't find file: {}", parser::display_rel_path(&physiscal_path)),
+            ));
+        }

-            let root = PathBuf::from_iter([GIT_DEPS_DIR, preprocess_git(link)]);
-
-            let mut command = std::process::Command::new("git");
-            command.args(["clone", "--depth", "1"]);
-            if let Some(chk) = chk {
-                command.args(match chk {
-                    Chk::Branch(b) => ["--branch", b],
-                    Chk::Tag(t) => ["--tag", t],
-                    Chk::Rev(r) => ["--rev", r],
-                });
-            }
-            command.arg(link).arg(root);
-            Some(command)
-        } else {
-            None
-        };
-
-        tasks.push((id, physiscal_path, command));
+        tasks.push((id, physiscal_path));
        Ok(id)
    };

-    let execute_task = |(_, path, command): Task, buffer: &mut Vec<u8>| {
-        if let Some(mut command) = command {
-            let output = command.output()?;
-            if !output.status.success() {
-                let msg =
-                    format!("git command failed: {}", String::from_utf8_lossy(&output.stderr));
-                return Err(io::Error::new(io::ErrorKind::Other, msg));
-            }
-        }
-
+    let execute_task = |(_, path): Task| {
        let path = path.to_str().ok_or_else(|| {
            io::Error::new(
                io::ErrorKind::InvalidData,
-                format!("path contains invalid characters: {}", path.display()),
+                format!("path contains invalid characters: {}", parser::display_rel_path(&path)),
            )
        })?;
-        let mut file = std::fs::File::open(path)?;
-        file.read_to_end(buffer)?;
-        let src = std::str::from_utf8(buffer).map_err(InvalidFileData)?;
-        Ok(Ast::new(path, src.to_owned(), &loader))
+        Ok(Ast::new(path, std::fs::read_to_string(path)?, &loader))
    };

    let thread = || {
-        let mut buffer = Vec::new();
        while let Some(task @ (indx, ..)) = tasks.pop() {
-            let res = execute_task(task, &mut buffer);
-            buffer.clear();
-
+            let res = execute_task(task);
            let mut ast = ast.lock().unwrap();
            let len = ast.len().max(indx as usize + 1);
            ast.resize_with(len, || Err(io::ErrorKind::InvalidData.into()));
@ -1298,7 +1040,7 @@ pub fn parse_from_fs(extra_threads: usize, root: &str) -> io::Result<Vec<Ast>> {

    let path = Path::new(root).canonicalize()?;
    seen.lock().unwrap().insert(path.clone(), 0);
-    tasks.push((0, path, None));
+    tasks.push((0, path));

    if extra_threads == 0 {
        thread();
@ -1467,7 +1209,7 @@ fn test_run_vm(out: &[u8], output: &mut String) {
    let mut vm = unsafe {
        hbvm::Vm::<_, { 1024 * 100 }>::new(
            LoggedMem::default(),
-            hbvm::mem::Address::new(out.as_ptr() as u64),
+            hbvm::mem::Address::new(out.as_ptr() as u64).wrapping_add(HEADER_SIZE),
        )
    };

@ -1506,7 +1248,7 @@ fn test_run_vm(out: &[u8], output: &mut String) {
        }
    };

-    writeln!(output, "code size: {}", out.len()).unwrap();
+    writeln!(output, "code size: {}", out.len() - HEADER_SIZE).unwrap();
    writeln!(output, "ret: {:?}", vm.read_reg(1).0).unwrap();
    writeln!(output, "status: {:?}", stat).unwrap();
 }
--- a/hblang/src/parser.rs
+++ b/hblang/src/parser.rs
@ -5,8 +5,10 @@ use {
    },
    std::{
        cell::{Cell, UnsafeCell},
+        ffi::OsStr,
        fmt, io,
        ops::{Deref, Not},
+        path::PathBuf,
        ptr::NonNull,
        sync::atomic::AtomicUsize,
    },
@ -95,20 +97,20 @@ impl<'a, 'b> Parser<'a, 'b> {
        let f = self.collect_list(TokenKind::Semi, TokenKind::Eof, |s| s.expr_low(true));

        self.pop_scope(0);
-        let has_undeclared = !self.idents.is_empty();
+        let mut errors = String::new();
        for id in self.idents.drain(..) {
-            let (line, col) = self.lexer.line_col(ident::pos(id.ident));
-            eprintln!(
-                "{}:{}:{} => undeclared identifier: {}",
+            report_to(
+                self.lexer.source(),
                self.path,
-                line,
-                col,
-                self.lexer.slice(ident::range(id.ident))
+                id.ident,
+                format_args!("undeclared identifier: {}", self.lexer.slice(ident::range(id.ident))),
+                &mut errors,
            );
        }

-        if has_undeclared {
+        if !errors.is_empty() {
            // TODO: we need error recovery
+            eprintln!("{errors}");
            unreachable!();
        }

@ -193,7 +195,7 @@ impl<'a, 'b> Parser<'a, 'b> {
                    self.declare_rec(value, top_level)
                }
            }
-            _ => self.report_pos(expr.pos(), "cant declare this shit (yet)"),
+            _ => self.report(expr.pos(), "cant declare this shit (yet)"),
        }
    }

@ -201,7 +203,7 @@ impl<'a, 'b> Parser<'a, 'b> {
        if let Some(index) = index_to_check
            && index != 0
        {
-            self.report_pos(
+            self.report(
                pos,
                format_args!(
                    "out of order declaration not allowed: {}",
@ -212,7 +214,7 @@ impl<'a, 'b> Parser<'a, 'b> {

        let index = self.idents.binary_search_by_key(&id, |s| s.ident).expect("fck up");
        if std::mem::replace(&mut self.idents[index].declared, true) {
-            self.report_pos(
+            self.report(
                pos,
                format_args!("redeclaration of identifier: {}", self.lexer.slice(ident::range(id))),
            )
@ -276,7 +278,9 @@ impl<'a, 'b> Parser<'a, 'b> {
                    path: self.arena.alloc_str(path),
                    id: match (self.loader)(path, self.path) {
                        Ok(id) => id,
-                        Err(e) => self.report(format_args!("error loading dependency: {e:#}")),
+                        Err(e) => {
+                            self.report(str.start, format_args!("error loading dependency: {e:#}"))
+                        }
                    },
                }
            }
@ -408,7 +412,7 @@ impl<'a, 'b> Parser<'a, 'b> {
                    pos,
                    value: match u64::from_str_radix(slice, radix as u32) {
                        Ok(value) => value,
-                        Err(e) => self.report(format_args!("invalid number: {e}")),
+                        Err(e) => self.report(token.start, format_args!("invalid number: {e}")),
                    } as i64,
                    radix,
                }
@ -419,7 +423,7 @@ impl<'a, 'b> Parser<'a, 'b> {
                expr
            }
            T::Comment => Expr::Comment { pos, literal: self.move_str(token) },
-            tok => self.report(format_args!("unexpected token: {tok:?}")),
+            tok => self.report(token.start, format_args!("unexpected token: {tok:?}")),
        };

        loop {
@ -497,7 +501,10 @@ impl<'a, 'b> Parser<'a, 'b> {
        if matches!(self.token.kind, TokenKind::Ident | TokenKind::CtIdent) {
            self.next()
        } else {
-            self.report(format_args!("expected identifier, found {:?}", self.token.kind))
+            self.report(
+                self.token.start,
+                format_args!("expected identifier, found {:?}", self.token.kind),
+            )
        }
    }

@ -552,20 +559,19 @@ impl<'a, 'b> Parser<'a, 'b> {

    fn expect_advance(&mut self, kind: TokenKind) -> Token {
        if self.token.kind != kind {
-            self.report(format_args!("expected {:?}, found {:?}", kind, self.token.kind));
+            self.report(
+                self.token.start,
+                format_args!("expected {:?}, found {:?}", kind, self.token.kind),
+            );
        }
        self.next()
    }

    #[track_caller]
-    fn report(&self, msg: impl fmt::Display) -> ! {
-        self.report_pos(self.token.start, msg)
-    }
-
-    #[track_caller]
-    fn report_pos(&self, pos: Pos, msg: impl fmt::Display) -> ! {
-        let (line, col) = self.lexer.line_col(pos);
-        eprintln!("{}:{}:{} => {}", self.path, line, col, msg);
+    fn report(&self, pos: Pos, msg: impl fmt::Display) -> ! {
+        let mut str = String::new();
+        report_to(self.lexer.source(), self.path, pos, msg, &mut str);
+        eprintln!("{str}");
        unreachable!();
    }

@ -1231,19 +1237,34 @@ impl AstInner<[Symbol]> {
    }

    pub fn report_to(&self, pos: Pos, msg: impl fmt::Display, out: &mut impl fmt::Write) {
-        let str = &self.file;
-        let (line, mut col) = lexer::line_col(str.as_bytes(), pos);
-        _ = writeln!(out, "{}:{}:{}: {}", self.path, line, col, msg);
-
-        let line = &str[str[..pos as usize].rfind('\n').map_or(0, |i| i + 1)
-            ..str[pos as usize..].find('\n').unwrap_or(str.len()) + pos as usize];
-        col += line.matches('\t').count() * 3;
-
-        _ = writeln!(out, "{}", line.replace("\t", "    "));
-        _ = writeln!(out, "{}^", " ".repeat(col - 1));
+        report_to(&self.file, &self.path, pos, msg, out);
    }
 }

+pub fn display_rel_path(path: &(impl AsRef<OsStr> + ?Sized)) -> std::path::Display {
+    static CWD: std::sync::LazyLock<PathBuf> =
+        std::sync::LazyLock::new(|| std::env::current_dir().unwrap_or_default());
+    std::path::Path::new(path).strip_prefix(&*CWD).unwrap_or(std::path::Path::new(path)).display()
+}
+
+pub fn report_to(
+    file: &str,
+    path: &str,
+    pos: Pos,
+    msg: impl fmt::Display,
+    out: &mut impl fmt::Write,
+) {
+    let (line, mut col) = lexer::line_col(file.as_bytes(), pos);
+    _ = writeln!(out, "{}:{}:{}: {}", display_rel_path(path), line, col, msg);
+
+    let line = &file[file[..pos as usize].rfind('\n').map_or(0, |i| i + 1)
+        ..file[pos as usize..].find('\n').unwrap_or(file.len()) + pos as usize];
+    col += line.matches('\t').count() * 3;
+
+    _ = writeln!(out, "{}", line.replace("\t", "    "));
+    _ = writeln!(out, "{}^", " ".repeat(col - 1));
+}
+
 #[derive(PartialEq, Eq, Hash)]
 pub struct Ast(NonNull<AstInner<[Symbol]>>);

--- a/hblang/src/son.rs
+++ b/hblang/src/son.rs
--- a/hblang/tests/codegen_tests_comments.txt
+++ b/hblang/tests/codegen_tests_comments.txt
@ -1,3 +1,9 @@
+foo:
+    ADDI64 r254, r254, -8d
+    ST r31, r254, 0a, 8h
+    LD r31, r254, 0a, 8h
+    ADDI64 r254, r254, 8d
+    JALA r0, r31, 0a
 main:
    ADDI64 r254, r254, -8d
    ST r31, r254, 0a, 8h
@ -6,12 +12,6 @@ main:
    LD r31, r254, 0a, 8h
    ADDI64 r254, r254, 8d
    JALA r0, r31, 0a
-foo:
-    ADDI64 r254, r254, -8d
-    ST r31, r254, 0a, 8h
-    LD r31, r254, 0a, 8h
-    ADDI64 r254, r254, 8d
-    JALA r0, r31, 0a
 code size: 143
 ret: 0
 status: Ok(())
--- a/hblang/tests/codegen_tests_functions.txt
+++ b/hblang/tests/codegen_tests_functions.txt
@ -1,3 +1,19 @@
+add_one:
+    ADDI64 r254, r254, -16d
+    ST r31, r254, 0a, 16h
+    CP r32, r2
+    ADDI64 r1, r32, 1d
+    LD r31, r254, 0a, 16h
+    ADDI64 r254, r254, 16d
+    JALA r0, r31, 0a
+add_two:
+    ADDI64 r254, r254, -16d
+    ST r31, r254, 0a, 16h
+    CP r32, r2
+    ADDI64 r1, r32, 2d
+    LD r31, r254, 0a, 16h
+    ADDI64 r254, r254, 16d
+    JALA r0, r31, 0a
 main:
    ADDI64 r254, r254, -24d
    ST r31, r254, 0a, 24h
@ -11,22 +27,6 @@ main:
    LD r31, r254, 0a, 24h
    ADDI64 r254, r254, 24d
    JALA r0, r31, 0a
-add_two:
-    ADDI64 r254, r254, -16d
-    ST r31, r254, 0a, 16h
-    CP r32, r2
-    ADDI64 r1, r32, 2d
-    LD r31, r254, 0a, 16h
-    ADDI64 r254, r254, 16d
-    JALA r0, r31, 0a
-add_one:
-    ADDI64 r254, r254, -16d
-    ST r31, r254, 0a, 16h
-    CP r32, r2
-    ADDI64 r1, r32, 1d
-    LD r31, r254, 0a, 16h
-    ADDI64 r254, r254, 16d
-    JALA r0, r31, 0a
 code size: 257
 ret: 33
 status: Ok(())
--- a/hblang/tests/codegen_tests_generic_functions.txt
+++ b/hblang/tests/codegen_tests_generic_functions.txt
@ -1,20 +1,3 @@
-main:
-    ADDI64 r254, r254, -32d
-    ST r31, r254, 0a, 32h
-    LI64 r2, 2d
-    LI64 r3, 2d
-    JAL r31, r0, :add
-    CP r32, r1
-    LI64 r2, 1d
-    LI64 r3, 3d
-    JAL r31, r0, :add
-    CP r33, r1
-    CP r34, r32
-    SXT32 r34, r34
-    SUB64 r1, r34, r33
-    LD r31, r254, 0a, 32h
-    ADDI64 r254, r254, 32d
-    JALA r0, r31, 0a
 add:
    ADDI64 r254, r254, -24d
    ST r31, r254, 0a, 24h
@ -33,6 +16,23 @@ add:
    LD r31, r254, 0a, 24h
    ADDI64 r254, r254, 24d
    JALA r0, r31, 0a
+main:
+    ADDI64 r254, r254, -32d
+    ST r31, r254, 0a, 32h
+    LI64 r2, 2d
+    LI64 r3, 2d
+    JAL r31, r0, :add
+    CP r32, r1
+    LI64 r2, 1d
+    LI64 r3, 3d
+    JAL r31, r0, :add
+    CP r33, r1
+    CP r34, r32
+    SXT32 r34, r34
+    SUB64 r1, r34, r33
+    LD r31, r254, 0a, 32h
+    ADDI64 r254, r254, 32d
+    JALA r0, r31, 0a
 code size: 275
 ret: 0
 status: Ok(())
--- a/hblang/tests/codegen_tests_generic_types.txt
+++ b/hblang/tests/codegen_tests_generic_types.txt
@ -1,3 +1,33 @@
+deinit:
+    ADDI64 r254, r254, -24d
+    ST r31, r254, 0a, 24h
+    CP r32, r2
+    LD r2, r32, 0a, 8h
+    LD r33, r32, 16a, 8h
+    MULI64 r33, r33, 8d
+    CP r3, r33
+    LI64 r4, 8d
+    JAL r31, r0, :free
+    CP r1, r32
+    JAL r31, r0, :new
+    LD r31, r254, 0a, 24h
+    ADDI64 r254, r254, 24d
+    JALA r0, r31, 0a
+free:
+    ADDI64 r254, r254, -40d
+    ST r31, r254, 0a, 40h
+    CP r32, r2
+    CP r33, r3
+    CP r34, r4
+    LRA r35, r0, :FREE_SYS_CALL
+    LD r2, r35, 0a, 8h
+    CP r3, r32
+    CP r4, r33
+    CP r5, r34
+    ECA 
+    LD r31, r254, 0a, 40h
+    ADDI64 r254, r254, 40d
+    JALA r0, r31, 0a
 main:
    ADDI64 r254, r254, -48d
    ST r31, r254, 24a, 24h
@ -17,20 +47,18 @@ main:
    LD r31, r254, 24a, 24h
    ADDI64 r254, r254, 48d
    JALA r0, r31, 0a
-deinit:
-    ADDI64 r254, r254, -24d
-    ST r31, r254, 0a, 24h
+malloc:
+    ADDI64 r254, r254, -32d
+    ST r31, r254, 0a, 32h
    CP r32, r2
-    LD r2, r32, 0a, 8h
-    LD r33, r32, 16a, 8h
-    MULI64 r33, r33, 8d
-    CP r3, r33
-    LI64 r4, 8d
-    JAL r31, r0, :free
-    CP r1, r32
-    JAL r31, r0, :new
-    LD r31, r254, 0a, 24h
-    ADDI64 r254, r254, 24d
+    CP r33, r3
+    LRA r34, r0, :MALLOC_SYS_CALL
+    LD r2, r34, 0a, 8h
+    CP r3, r32
+    CP r4, r33
+    ECA 
+    LD r31, r254, 0a, 32h
+    ADDI64 r254, r254, 32d
    JALA r0, r31, 0a
 new:
    ADDI64 r254, r254, -24d
@ -45,21 +73,6 @@ new:
    LD r31, r254, 0a, 24h
    ADDI64 r254, r254, 24d
    JALA r0, r31, 0a
-free:
-    ADDI64 r254, r254, -40d
-    ST r31, r254, 0a, 40h
-    CP r32, r2
-    CP r33, r3
-    CP r34, r4
-    LRA r35, r0, :FREE_SYS_CALL
-    LD r2, r35, 0a, 8h
-    CP r3, r32
-    CP r4, r33
-    CP r5, r34
-    ECA 
-    LD r31, r254, 0a, 40h
-    ADDI64 r254, r254, 40d
-    JALA r0, r31, 0a
 push:
    ADDI64 r254, r254, -72d
    ST r31, r254, 0a, 72h
@ -124,19 +137,6 @@ push:
 4: LD r31, r254, 0a, 72h
    ADDI64 r254, r254, 72d
    JALA r0, r31, 0a
-malloc:
-    ADDI64 r254, r254, -32d
-    ST r31, r254, 0a, 32h
-    CP r32, r2
-    CP r33, r3
-    LRA r34, r0, :MALLOC_SYS_CALL
-    LD r2, r34, 0a, 8h
-    CP r3, r32
-    CP r4, r33
-    ECA 
-    LD r31, r254, 0a, 32h
-    ADDI64 r254, r254, 32d
-    JALA r0, r31, 0a
 code size: 1201
 ret: 69
 status: Ok(())
--- a/hblang/tests/codegen_tests_if_statements.txt
+++ b/hblang/tests/codegen_tests_if_statements.txt
@ -1,11 +1,3 @@
-main:
-    ADDI64 r254, r254, -8d
-    ST r31, r254, 0a, 8h
-    LI64 r2, 10d
-    JAL r31, r0, :fib
-    LD r31, r254, 0a, 8h
-    ADDI64 r254, r254, 8d
-    JALA r0, r31, 0a
 fib:
    ADDI64 r254, r254, -24d
    ST r31, r254, 0a, 24h
@ -27,6 +19,14 @@ fib:
 1: LD r31, r254, 0a, 24h
    ADDI64 r254, r254, 24d
    JALA r0, r31, 0a
+main:
+    ADDI64 r254, r254, -8d
+    ST r31, r254, 0a, 8h
+    LI64 r2, 10d
+    JAL r31, r0, :fib
+    LD r31, r254, 0a, 8h
+    ADDI64 r254, r254, 8d
+    JALA r0, r31, 0a
 code size: 231
 ret: 55
 status: Ok(())
--- a/hblang/tests/codegen_tests_inline_test.txt
+++ b/hblang/tests/codegen_tests_inline_test.txt
@ -1,45 +1,3 @@
-main:
-    ADDI64 r254, r254, -64d
-    ST r31, r254, 48a, 16h
-    LI64 r32, 0d
-    ST r32, r254, 0a, 8h
-    LI64 r32, 0d
-    ST r32, r254, 8a, 8h
-    LD r2, r254, 0a, 16h
-    LI64 r32, 0d
-    ST r32, r254, 16a, 8h
-    LI64 r32, 0d
-    ST r32, r254, 24a, 8h
-    LD r4, r254, 16a, 16h
-    LI64 r32, 0d
-    ST r32, r254, 32a, 8h
-    LI64 r32, 0d
-    ST r32, r254, 40a, 8h
-    LD r6, r254, 32a, 16h
-    LI64 r8, 10d
-    JAL r31, r0, :line
-    LI64 r32, 0d
-    ST r32, r254, 0a, 8h
-    LI64 r32, 0d
-    ST r32, r254, 8a, 8h
-    LD r2, r254, 0a, 16h
-    LI64 r32, 0d
-    ST r32, r254, 16a, 8h
-    LI64 r32, 0d
-    ST r32, r254, 24a, 8h
-    LD r4, r254, 16a, 16h
-    LI64 r32, 0d
-    ST r32, r254, 32a, 8h
-    LI64 r32, 0d
-    ST r32, r254, 40a, 8h
-    LD r6, r254, 32a, 16h
-    LI64 r8, 10d
-    JAL r31, r0, :rect_line
-    JAL r31, r0, :example
-    LI64 r1, 0d
-    LD r31, r254, 48a, 16h
-    ADDI64 r254, r254, 64d
-    JALA r0, r31, 0a
 example:
    ADDI64 r254, r254, -48d
    ST r31, r254, 0a, 48h
@ -103,6 +61,69 @@ integer:
 1: LD r31, r254, 0a, 56h
    ADDI64 r254, r254, 56d
    JALA r0, r31, 0a
+line:
+    ADDI64 r254, r254, -80d
+    ST r31, r254, 48a, 32h
+    ST r2, r254, 16a, 16h
+    ST r4, r254, 0a, 16h
+    ST r6, r254, 32a, 16h
+    CP r32, r8
+    LI64 r33, 1d
+    JEQ r33, r0, :0
+    LD r33, r254, 16a, 8h
+    LD r34, r254, 0a, 8h
+    JGTS r33, r34, :1
+    JMP :1
+ 1: JMP :2
+ 0: LD r34, r254, 24a, 8h
+    LD r33, r254, 8a, 8h
+    JGTS r34, r33, :2
+    JMP :2
+ 2: LD r31, r254, 48a, 32h
+    ADDI64 r254, r254, 80d
+    JALA r0, r31, 0a
+main:
+    ADDI64 r254, r254, -64d
+    ST r31, r254, 48a, 16h
+    LI64 r32, 0d
+    ST r32, r254, 0a, 8h
+    LI64 r32, 0d
+    ST r32, r254, 8a, 8h
+    LD r2, r254, 0a, 16h
+    LI64 r32, 0d
+    ST r32, r254, 16a, 8h
+    LI64 r32, 0d
+    ST r32, r254, 24a, 8h
+    LD r4, r254, 16a, 16h
+    LI64 r32, 0d
+    ST r32, r254, 32a, 8h
+    LI64 r32, 0d
+    ST r32, r254, 40a, 8h
+    LD r6, r254, 32a, 16h
+    LI64 r8, 10d
+    JAL r31, r0, :line
+    LI64 r32, 0d
+    ST r32, r254, 0a, 8h
+    LI64 r32, 0d
+    ST r32, r254, 8a, 8h
+    LD r2, r254, 0a, 16h
+    LI64 r32, 0d
+    ST r32, r254, 16a, 8h
+    LI64 r32, 0d
+    ST r32, r254, 24a, 8h
+    LD r4, r254, 16a, 16h
+    LI64 r32, 0d
+    ST r32, r254, 32a, 8h
+    LI64 r32, 0d
+    ST r32, r254, 40a, 8h
+    LD r6, r254, 32a, 16h
+    LI64 r8, 10d
+    JAL r31, r0, :rect_line
+    JAL r31, r0, :example
+    LI64 r1, 0d
+    LD r31, r254, 48a, 16h
+    ADDI64 r254, r254, 64d
+    JALA r0, r31, 0a
 rect_line:
    ADDI64 r254, r254, -112d
    ST r31, r254, 48a, 64h
@ -135,27 +156,6 @@ rect_line:
 1: LD r31, r254, 48a, 64h
    ADDI64 r254, r254, 112d
    JALA r0, r31, 0a
-line:
-    ADDI64 r254, r254, -80d
-    ST r31, r254, 48a, 32h
-    ST r2, r254, 16a, 16h
-    ST r4, r254, 0a, 16h
-    ST r6, r254, 32a, 16h
-    CP r32, r8
-    LI64 r33, 1d
-    JEQ r33, r0, :0
-    LD r33, r254, 16a, 8h
-    LD r34, r254, 0a, 8h
-    JGTS r33, r34, :1
-    JMP :1
- 1: JMP :2
- 0: LD r34, r254, 24a, 8h
-    LD r33, r254, 8a, 8h
-    JGTS r34, r33, :2
-    JMP :2
- 2: LD r31, r254, 48a, 32h
-    ADDI64 r254, r254, 80d
-    JALA r0, r31, 0a
 code size: 1400
 ret: 0
 status: Ok(())
--- a/hblang/tests/codegen_tests_integer_inference_issues.txt
+++ b/hblang/tests/codegen_tests_integer_inference_issues.txt
@ -1,13 +1,3 @@
-main:
-    ADDI64 r254, r254, -16d
-    ST r31, r254, 0a, 16h
-    LI64 r2, 0d
-    LI64 r3, 1000d
-    JAL r31, r0, :integer_range
-    CP r32, r1
-    LD r31, r254, 0a, 16h
-    ADDI64 r254, r254, 16d
-    JALA r0, r31, 0a
 integer_range:
    ADDI64 r254, r254, -32d
    ST r31, r254, 0a, 32h
@ -24,6 +14,16 @@ integer_range:
    LD r31, r254, 0a, 32h
    ADDI64 r254, r254, 32d
    JALA r0, r31, 0a
+main:
+    ADDI64 r254, r254, -16d
+    ST r31, r254, 0a, 16h
+    LI64 r2, 0d
+    LI64 r3, 1000d
+    JAL r31, r0, :integer_range
+    CP r32, r1
+    LD r31, r254, 0a, 16h
+    ADDI64 r254, r254, 16d
+    JALA r0, r31, 0a
 code size: 210
 ret: 42
 status: Ok(())
--- a/hblang/tests/codegen_tests_loops.txt
+++ b/hblang/tests/codegen_tests_loops.txt
@ -1,11 +1,3 @@
-main:
-    ADDI64 r254, r254, -8d
-    ST r31, r254, 0a, 8h
-    LI64 r2, 10d
-    JAL r31, r0, :fib
-    LD r31, r254, 0a, 8h
-    ADDI64 r254, r254, 8d
-    JALA r0, r31, 0a
 fib:
    ADDI64 r254, r254, -40d
    ST r31, r254, 0a, 40h
@ -25,6 +17,14 @@ fib:
    LD r31, r254, 0a, 40h
    ADDI64 r254, r254, 40d
    JALA r0, r31, 0a
+main:
+    ADDI64 r254, r254, -8d
+    ST r31, r254, 0a, 8h
+    LI64 r2, 10d
+    JAL r31, r0, :fib
+    LD r31, r254, 0a, 8h
+    ADDI64 r254, r254, 8d
+    JALA r0, r31, 0a
 code size: 218
 ret: 55
 status: Ok(())
--- a/hblang/tests/codegen_tests_pointers.txt
+++ b/hblang/tests/codegen_tests_pointers.txt
@ -1,3 +1,10 @@
+drop:
+    ADDI64 r254, r254, -16d
+    ST r31, r254, 0a, 16h
+    CP r32, r2
+    LD r31, r254, 0a, 16h
+    ADDI64 r254, r254, 16d
+    JALA r0, r31, 0a
 main:
    ADDI64 r254, r254, -40d
    ST r31, r254, 8a, 32h
@ -14,13 +21,6 @@ main:
    LD r31, r254, 8a, 32h
    ADDI64 r254, r254, 40d
    JALA r0, r31, 0a
-drop:
-    ADDI64 r254, r254, -16d
-    ST r31, r254, 0a, 16h
-    CP r32, r2
-    LD r31, r254, 0a, 16h
-    ADDI64 r254, r254, 16d
-    JALA r0, r31, 0a
 modify:
    ADDI64 r254, r254, -24d
    ST r31, r254, 0a, 24h
--- a/hblang/tests/codegen_tests_request_page.txt
+++ b/hblang/tests/codegen_tests_request_page.txt
@ -1,12 +1,3 @@
-main:
-    ADDI64 r254, r254, -16d
-    ST r31, r254, 0a, 16h
-    LI64 r2, 400d
-    JAL r31, r0, :create_back_buffer
-    CP r32, r1
-    LD r31, r254, 0a, 16h
-    ADDI64 r254, r254, 16d
-    JALA r0, r31, 0a
 create_back_buffer:
    ADDI64 r254, r254, -32d
    ST r31, r254, 0a, 32h
@ -38,6 +29,15 @@ create_back_buffer:
 1: LD r31, r254, 0a, 32h
    ADDI64 r254, r254, 32d
    JALA r0, r31, 0a
+main:
+    ADDI64 r254, r254, -16d
+    ST r31, r254, 0a, 16h
+    LI64 r2, 400d
+    JAL r31, r0, :create_back_buffer
+    CP r32, r1
+    LD r31, r254, 0a, 16h
+    ADDI64 r254, r254, 16d
+    JALA r0, r31, 0a
 request_page:
    ADDI64 r254, r254, -32d
    ST r31, r254, 0a, 32h
--- a/hblang/tests/codegen_tests_struct_patterns.txt
+++ b/hblang/tests/codegen_tests_struct_patterns.txt
@ -1,43 +1,3 @@
-main:
-    ADDI64 r254, r254, -26d
-    ST r31, r254, 2a, 24h
-    LI64 r32, 10d
-    ST r32, r254, 0a, 1h
-    LI64 r32, 10d
-    ST r32, r254, 1a, 1h
-    CP r32, r0
-    LD r32, r254, 0a, 1h
-    CP r33, r0
-    LD r33, r254, 1a, 1h
-    CP r2, r32
-    JAL r31, r0, :fib
-    CP r32, r1
-    CP r2, r33
-    JAL r31, r0, :fib_iter
-    CP r33, r1
-    SUB64 r1, r32, r33
-    LD r31, r254, 2a, 24h
-    ADDI64 r254, r254, 26d
-    JALA r0, r31, 0a
-fib_iter:
-    ADDI64 r254, r254, -40d
-    ST r31, r254, 0a, 40h
-    CP r32, r2
-    LI64 r33, 0d
-    LI64 r34, 1d
- 2: LI64 r35, 0d
-    JNE r32, r35, :0
-    JMP :1
- 0: CP r35, r33
-    ADD64 r35, r35, r34
-    CP r33, r34
-    CP r34, r35
-    ADDI64 r32, r32, -1d
-    JMP :2
- 1: CP r1, r33
-    LD r31, r254, 0a, 40h
-    ADDI64 r254, r254, 40d
-    JALA r0, r31, 0a
 fib:
    ADDI64 r254, r254, -32d
    ST r31, r254, 0a, 32h
@ -60,6 +20,46 @@ fib:
 1: LD r31, r254, 0a, 32h
    ADDI64 r254, r254, 32d
    JALA r0, r31, 0a
+fib_iter:
+    ADDI64 r254, r254, -40d
+    ST r31, r254, 0a, 40h
+    CP r32, r2
+    LI64 r33, 0d
+    LI64 r34, 1d
+ 2: LI64 r35, 0d
+    JNE r32, r35, :0
+    JMP :1
+ 0: CP r35, r33
+    ADD64 r35, r35, r34
+    CP r33, r34
+    CP r34, r35
+    ADDI64 r32, r32, -1d
+    JMP :2
+ 1: CP r1, r33
+    LD r31, r254, 0a, 40h
+    ADDI64 r254, r254, 40d
+    JALA r0, r31, 0a
+main:
+    ADDI64 r254, r254, -26d
+    ST r31, r254, 2a, 24h
+    LI64 r32, 10d
+    ST r32, r254, 0a, 1h
+    LI64 r32, 10d
+    ST r32, r254, 1a, 1h
+    CP r32, r0
+    LD r32, r254, 0a, 1h
+    CP r33, r0
+    LD r33, r254, 1a, 1h
+    CP r2, r32
+    JAL r31, r0, :fib
+    CP r32, r1
+    CP r2, r33
+    JAL r31, r0, :fib_iter
+    CP r33, r1
+    SUB64 r1, r32, r33
+    LD r31, r254, 2a, 24h
+    ADDI64 r254, r254, 26d
+    JALA r0, r31, 0a
 code size: 452
 ret: 0
 status: Ok(())
--- a/hblang/tests/codegen_tests_struct_return_from_module_function.txt
+++ b/hblang/tests/codegen_tests_struct_return_from_module_function.txt
@ -1,3 +1,16 @@
+foo:
+    ADDI64 r254, r254, -32d
+    ST r31, r254, 16a, 16h
+    LI64 r32, 3d
+    ST r32, r254, 0a, 8h
+    LI64 r32, 2d
+    ST r32, r254, 8a, 4h
+    LI64 r32, 2d
+    ST r32, r254, 12a, 4h
+    LD r1, r254, 0a, 16h
+    LD r31, r254, 16a, 16h
+    ADDI64 r254, r254, 32d
+    JALA r0, r31, 0a
 main:
    ADDI64 r254, r254, -40d
    ST r31, r254, 16a, 24h
@ -21,17 +34,6 @@ main:
    LD r31, r254, 16a, 24h
    ADDI64 r254, r254, 40d
    JALA r0, r31, 0a
-foo:
-    ADDI64 r254, r254, -8d
-    ST r31, r254, 0a, 8h
-    LI64 r1, 3d
-    ANDI r2, r2, -4294967296d
-    ORI r2, r2, 2d
-    ANDI r2, r2, 4294967295d
-    ORI r2, r2, 8589934592d
-    LD r31, r254, 0a, 8h
-    ADDI64 r254, r254, 8d
-    JALA r0, r31, 0a
-code size: 313
+code size: 341
 ret: 0
 status: Ok(())
--- a/hblang/tests/codegen_tests_structs.txt
+++ b/hblang/tests/codegen_tests_structs.txt
@ -21,16 +21,6 @@ main:
 1: LD r31, r254, 24a, 24h
    ADDI64 r254, r254, 48d
    JALA r0, r31, 0a
-pass:
-    ADDI64 r254, r254, -32d
-    ST r31, r254, 0a, 32h
-    CP r32, r2
-    LD r33, r32, 0a, 8h
-    LD r34, r32, 8a, 8h
-    SUB64 r1, r33, r34
-    LD r31, r254, 0a, 32h
-    ADDI64 r254, r254, 32d
-    JALA r0, r31, 0a
 odher_pass:
    ADDI64 r254, r254, -32d
    ST r31, r254, 0a, 32h
@ -41,6 +31,16 @@ odher_pass:
    LD r31, r254, 0a, 32h
    ADDI64 r254, r254, 32d
    JALA r0, r31, 0a
+pass:
+    ADDI64 r254, r254, -32d
+    ST r31, r254, 0a, 32h
+    CP r32, r2
+    LD r33, r32, 0a, 8h
+    LD r34, r32, 8a, 8h
+    SUB64 r1, r33, r34
+    LD r31, r254, 0a, 32h
+    ADDI64 r254, r254, 32d
+    JALA r0, r31, 0a
 code size: 394
 ret: 3
 status: Ok(())
--- a/hblang/tests/codegen_tests_wide_ret.txt
+++ b/hblang/tests/codegen_tests_wide_ret.txt
@ -0,0 +1,45 @@
+main:
+    ADDI64 r254, r254, -48d
+    ST r31, r254, 16a, 32h
+    CP r32, r3
+    CP r33, r4
+    JAL r31, r0, :small_struct
+    CP r34, r1
+    LI64 r34, 0d
+    ST r34, r254, 8a, 1h
+    LI64 r34, 0d
+    ST r34, r254, 9a, 1h
+    LI64 r34, 0d
+    ST r34, r254, 10a, 1h
+    LI64 r34, 0d
+    ST r34, r254, 11a, 1h
+    LI64 r34, 0d
+    ST r34, r254, 12a, 1h
+    LI64 r34, 0d
+    ST r34, r254, 13a, 1h
+    LI64 r34, 0d
+    ST r34, r254, 14a, 1h
+    LI64 r34, 0d
+    ST r34, r254, 15a, 1h
+    LD r34, r254, 8a, 8h
+    ST r34, r254, 0a, 8h
+    ST r34, r254, 8a, 8h
+    LD r1, r254, 0a, 16h
+    LD r31, r254, 16a, 32h
+    ADDI64 r254, r254, 48d
+    JALA r0, r31, 0a
+small_struct:
+    ADDI64 r254, r254, -20d
+    ST r31, r254, 4a, 16h
+    LI64 r32, 0d
+    ST r32, r254, 0a, 2h
+    LI64 r32, 0d
+    ST r32, r254, 2a, 2h
+    CP r1, r0
+    LD r1, r254, 0a, 4h
+    LD r31, r254, 4a, 16h
+    ADDI64 r254, r254, 20d
+    JALA r0, r31, 0a
+code size: 440
+ret: 0
+status: Ok(())
--- a/hblang/tests/codegen_tests_writing_into_string.txt
+++ b/hblang/tests/codegen_tests_writing_into_string.txt
@ -1,3 +1,10 @@
+inl:
+    ADDI64 r254, r254, -16d
+    ST r31, r254, 0a, 16h
+    LRA r32, r0, :"luhahah\0"
+    LD r31, r254, 0a, 16h
+    ADDI64 r254, r254, 16d
+    JALA r0, r31, 0a
 main:
    ADDI64 r254, r254, -8d
    ST r31, r254, 0a, 8h
@ -6,13 +13,6 @@ main:
    LD r31, r254, 0a, 8h
    ADDI64 r254, r254, 8d
    JALA r0, r31, 0a
-inl:
-    ADDI64 r254, r254, -16d
-    ST r31, r254, 0a, 16h
-    LRA r32, r0, :"luhahah\0"
-    LD r31, r254, 0a, 16h
-    ADDI64 r254, r254, 16d
-    JALA r0, r31, 0a
 outl:
    ADDI64 r254, r254, -16d
    ST r31, r254, 0a, 16h
--- a/hbvm/Cargo.toml
+++ b/hbvm/Cargo.toml
@ -9,4 +9,4 @@ alloc   = []
 nightly = []

 [dependencies]
-hbbytecode = { path = "../hbbytecode" }
+hbbytecode = { path = "../hbbytecode", default-features = false }
--- a/hbvm/src/vmrun.rs
+++ b/hbvm/src/vmrun.rs
@ -34,7 +34,7 @@ where
    /// Program can return [`VmRunError`] if a trap handling failed
    #[cfg_attr(feature = "nightly", repr(align(4096)))]
    pub fn run(&mut self) -> Result<VmRunOk, VmRunError> {
-        use hbbytecode::opcode::*;
+        use hbbytecode::Instr as I;
        loop {
            // Big match
            //
@ -56,105 +56,112 @@ where
            // - Yes, we assume you run 64 bit CPU. Else ?conradluget a better CPU
            //   sorry 8 bit fans, HBVM won't run on your Speccy :(
            unsafe {
-                match self.memory.prog_read::<u8>(self.pc as _) {
-                    UN => {
+                match self
+                    .memory
+                    .prog_read::<u8>(self.pc as _)
+                    .try_into()
+                    .map_err(VmRunError::InvalidOpcode)?
+                {
+                    I::UN => {
                        self.bump_pc::<OpsN>();
                        return Err(VmRunError::Unreachable);
                    }
-                    TX => {
+                    I::TX => {
                        self.bump_pc::<OpsN>();
                        return Ok(VmRunOk::End);
                    }
-                    NOP => handler!(self, |OpsN()| ()),
-                    ADD8 => self.binary_op(u8::wrapping_add),
-                    ADD16 => self.binary_op(u16::wrapping_add),
-                    ADD32 => self.binary_op(u32::wrapping_add),
-                    ADD64 => self.binary_op(u64::wrapping_add),
-                    SUB8 => self.binary_op(u8::wrapping_sub),
-                    SUB16 => self.binary_op(u16::wrapping_sub),
-                    SUB32 => self.binary_op(u32::wrapping_sub),
-                    SUB64 => self.binary_op(u64::wrapping_sub),
-                    MUL8 => self.binary_op(u8::wrapping_mul),
-                    MUL16 => self.binary_op(u16::wrapping_mul),
-                    MUL32 => self.binary_op(u32::wrapping_mul),
-                    MUL64 => self.binary_op(u64::wrapping_mul),
-                    AND => self.binary_op::<u64>(ops::BitAnd::bitand),
-                    OR => self.binary_op::<u64>(ops::BitOr::bitor),
-                    XOR => self.binary_op::<u64>(ops::BitXor::bitxor),
-                    SLU8 => self.binary_op_shift::<u8>(u8::wrapping_shl),
-                    SLU16 => self.binary_op_shift::<u16>(u16::wrapping_shl),
-                    SLU32 => self.binary_op_shift::<u32>(u32::wrapping_shl),
-                    SLU64 => self.binary_op_shift::<u64>(u64::wrapping_shl),
-                    SRU8 => self.binary_op_shift::<u8>(u8::wrapping_shr),
-                    SRU16 => self.binary_op_shift::<u16>(u16::wrapping_shr),
-                    SRU32 => self.binary_op_shift::<u32>(u32::wrapping_shr),
-                    SRU64 => self.binary_op_shift::<u64>(u64::wrapping_shr),
-                    SRS8 => self.binary_op_shift::<i8>(i8::wrapping_shr),
-                    SRS16 => self.binary_op_shift::<i16>(i16::wrapping_shr),
-                    SRS32 => self.binary_op_shift::<i32>(i32::wrapping_shr),
-                    SRS64 => self.binary_op_shift::<i64>(i64::wrapping_shr),
-                    CMPU => handler!(self, |OpsRRR(tg, a0, a1)| self.cmp(
+                    I::NOP => handler!(self, |OpsN()| ()),
+                    I::ADD8 => self.binary_op(u8::wrapping_add),
+                    I::ADD16 => self.binary_op(u16::wrapping_add),
+                    I::ADD32 => self.binary_op(u32::wrapping_add),
+                    I::ADD64 => self.binary_op(u64::wrapping_add),
+                    I::SUB8 => self.binary_op(u8::wrapping_sub),
+                    I::SUB16 => self.binary_op(u16::wrapping_sub),
+                    I::SUB32 => self.binary_op(u32::wrapping_sub),
+                    I::SUB64 => self.binary_op(u64::wrapping_sub),
+                    I::MUL8 => self.binary_op(u8::wrapping_mul),
+                    I::MUL16 => self.binary_op(u16::wrapping_mul),
+                    I::MUL32 => self.binary_op(u32::wrapping_mul),
+                    I::MUL64 => self.binary_op(u64::wrapping_mul),
+                    I::AND => self.binary_op::<u64>(ops::BitAnd::bitand),
+                    I::OR => self.binary_op::<u64>(ops::BitOr::bitor),
+                    I::XOR => self.binary_op::<u64>(ops::BitXor::bitxor),
+                    I::SLU8 => self.binary_op_shift::<u8>(u8::wrapping_shl),
+                    I::SLU16 => self.binary_op_shift::<u16>(u16::wrapping_shl),
+                    I::SLU32 => self.binary_op_shift::<u32>(u32::wrapping_shl),
+                    I::SLU64 => self.binary_op_shift::<u64>(u64::wrapping_shl),
+                    I::SRU8 => self.binary_op_shift::<u8>(u8::wrapping_shr),
+                    I::SRU16 => self.binary_op_shift::<u16>(u16::wrapping_shr),
+                    I::SRU32 => self.binary_op_shift::<u32>(u32::wrapping_shr),
+                    I::SRU64 => self.binary_op_shift::<u64>(u64::wrapping_shr),
+                    I::SRS8 => self.binary_op_shift::<i8>(i8::wrapping_shr),
+                    I::SRS16 => self.binary_op_shift::<i16>(i16::wrapping_shr),
+                    I::SRS32 => self.binary_op_shift::<i32>(i32::wrapping_shr),
+                    I::SRS64 => self.binary_op_shift::<i64>(i64::wrapping_shr),
+                    I::CMPU => handler!(self, |OpsRRR(tg, a0, a1)| self.cmp(
                        tg,
                        a0,
                        self.read_reg(a1).cast::<u64>()
                    )),
-                    CMPS => handler!(self, |OpsRRR(tg, a0, a1)| self.cmp(
+                    I::CMPS => handler!(self, |OpsRRR(tg, a0, a1)| self.cmp(
                        tg,
                        a0,
                        self.read_reg(a1).cast::<i64>()
                    )),
-                    DIRU8 => self.dir::<u8>(),
-                    DIRU16 => self.dir::<u16>(),
-                    DIRU32 => self.dir::<u32>(),
-                    DIRU64 => self.dir::<u64>(),
-                    DIRS8 => self.dir::<i8>(),
-                    DIRS16 => self.dir::<i16>(),
-                    DIRS32 => self.dir::<i32>(),
-                    DIRS64 => self.dir::<i64>(),
-                    NEG => handler!(self, |OpsRR(tg, a0)| {
+                    I::DIRU8 => self.dir::<u8>(),
+                    I::DIRU16 => self.dir::<u16>(),
+                    I::DIRU32 => self.dir::<u32>(),
+                    I::DIRU64 => self.dir::<u64>(),
+                    I::DIRS8 => self.dir::<i8>(),
+                    I::DIRS16 => self.dir::<i16>(),
+                    I::DIRS32 => self.dir::<i32>(),
+                    I::DIRS64 => self.dir::<i64>(),
+                    I::NEG => handler!(self, |OpsRR(tg, a0)| {
                        // Bit negation
                        self.write_reg(tg, self.read_reg(a0).cast::<u64>().wrapping_neg())
                    }),
-                    NOT => handler!(self, |OpsRR(tg, a0)| {
+                    I::NOT => handler!(self, |OpsRR(tg, a0)| {
                        // Logical negation
                        self.write_reg(tg, u64::from(self.read_reg(a0).cast::<u64>() == 0));
                    }),
-                    SXT8 => handler!(self, |OpsRR(tg, a0)| {
+                    I::SXT8 => handler!(self, |OpsRR(tg, a0)| {
                        self.write_reg(tg, self.read_reg(a0).cast::<i8>() as i64)
                    }),
-                    SXT16 => handler!(self, |OpsRR(tg, a0)| {
+                    I::SXT16 => handler!(self, |OpsRR(tg, a0)| {
                        self.write_reg(tg, self.read_reg(a0).cast::<i16>() as i64)
                    }),
-                    SXT32 => handler!(self, |OpsRR(tg, a0)| {
+                    I::SXT32 => handler!(self, |OpsRR(tg, a0)| {
                        self.write_reg(tg, self.read_reg(a0).cast::<i32>() as i64)
                    }),
-                    ADDI8 => self.binary_op_imm(u8::wrapping_add),
-                    ADDI16 => self.binary_op_imm(u16::wrapping_add),
-                    ADDI32 => self.binary_op_imm(u32::wrapping_add),
-                    ADDI64 => self.binary_op_imm(u64::wrapping_add),
-                    MULI8 => self.binary_op_imm(u8::wrapping_mul),
-                    MULI16 => self.binary_op_imm(u16::wrapping_mul),
-                    MULI32 => self.binary_op_imm(u32::wrapping_mul),
-                    MULI64 => self.binary_op_imm(u64::wrapping_mul),
-                    ANDI => self.binary_op_imm::<u64>(ops::BitAnd::bitand),
-                    ORI => self.binary_op_imm::<u64>(ops::BitOr::bitor),
-                    XORI => self.binary_op_imm::<u64>(ops::BitXor::bitxor),
-                    SLUI8 => self.binary_op_ims::<u8>(u8::wrapping_shl),
-                    SLUI16 => self.binary_op_ims::<u16>(u16::wrapping_shl),
-                    SLUI32 => self.binary_op_ims::<u32>(u32::wrapping_shl),
-                    SLUI64 => self.binary_op_ims::<u64>(u64::wrapping_shl),
-                    SRUI8 => self.binary_op_ims::<u8>(u8::wrapping_shr),
-                    SRUI16 => self.binary_op_ims::<u16>(u16::wrapping_shr),
-                    SRUI32 => self.binary_op_ims::<u32>(u32::wrapping_shr),
-                    SRUI64 => self.binary_op_ims::<u64>(u64::wrapping_shr),
-                    SRSI8 => self.binary_op_ims::<i8>(i8::wrapping_shr),
-                    SRSI16 => self.binary_op_ims::<i16>(i16::wrapping_shr),
-                    SRSI32 => self.binary_op_ims::<i32>(i32::wrapping_shr),
-                    SRSI64 => self.binary_op_ims::<i64>(i64::wrapping_shr),
-                    CMPUI => handler!(self, |OpsRRD(tg, a0, imm)| { self.cmp(tg, a0, imm) }),
-                    CMPSI => handler!(self, |OpsRRD(tg, a0, imm)| { self.cmp(tg, a0, imm as i64) }),
-                    CP => handler!(self, |OpsRR(tg, a0)| self.write_reg(tg, self.read_reg(a0))),
-                    SWA => handler!(self, |OpsRR(r0, r1)| {
+                    I::ADDI8 => self.binary_op_imm(u8::wrapping_add),
+                    I::ADDI16 => self.binary_op_imm(u16::wrapping_add),
+                    I::ADDI32 => self.binary_op_imm(u32::wrapping_add),
+                    I::ADDI64 => self.binary_op_imm(u64::wrapping_add),
+                    I::MULI8 => self.binary_op_imm(u8::wrapping_mul),
+                    I::MULI16 => self.binary_op_imm(u16::wrapping_mul),
+                    I::MULI32 => self.binary_op_imm(u32::wrapping_mul),
+                    I::MULI64 => self.binary_op_imm(u64::wrapping_mul),
+                    I::ANDI => self.binary_op_imm::<u64>(ops::BitAnd::bitand),
+                    I::ORI => self.binary_op_imm::<u64>(ops::BitOr::bitor),
+                    I::XORI => self.binary_op_imm::<u64>(ops::BitXor::bitxor),
+                    I::SLUI8 => self.binary_op_ims::<u8>(u8::wrapping_shl),
+                    I::SLUI16 => self.binary_op_ims::<u16>(u16::wrapping_shl),
+                    I::SLUI32 => self.binary_op_ims::<u32>(u32::wrapping_shl),
+                    I::SLUI64 => self.binary_op_ims::<u64>(u64::wrapping_shl),
+                    I::SRUI8 => self.binary_op_ims::<u8>(u8::wrapping_shr),
+                    I::SRUI16 => self.binary_op_ims::<u16>(u16::wrapping_shr),
+                    I::SRUI32 => self.binary_op_ims::<u32>(u32::wrapping_shr),
+                    I::SRUI64 => self.binary_op_ims::<u64>(u64::wrapping_shr),
+                    I::SRSI8 => self.binary_op_ims::<i8>(i8::wrapping_shr),
+                    I::SRSI16 => self.binary_op_ims::<i16>(i16::wrapping_shr),
+                    I::SRSI32 => self.binary_op_ims::<i32>(i32::wrapping_shr),
+                    I::SRSI64 => self.binary_op_ims::<i64>(i64::wrapping_shr),
+                    I::CMPUI => handler!(self, |OpsRRD(tg, a0, imm)| { self.cmp(tg, a0, imm) }),
+                    I::CMPSI => {
+                        handler!(self, |OpsRRD(tg, a0, imm)| { self.cmp(tg, a0, imm as i64) })
+                    }
+                    I::CP => handler!(self, |OpsRR(tg, a0)| self.write_reg(tg, self.read_reg(a0))),
+                    I::SWA => handler!(self, |OpsRR(r0, r1)| {
                        // Swap registers
                        match (r0, r1) {
                            (0, 0) => (),
@ -167,33 +174,33 @@ where
                            }
                        }
                    }),
-                    LI8 => handler!(self, |OpsRB(tg, imm)| self.write_reg(tg, imm)),
-                    LI16 => handler!(self, |OpsRH(tg, imm)| self.write_reg(tg, imm)),
-                    LI32 => handler!(self, |OpsRW(tg, imm)| self.write_reg(tg, imm)),
-                    LI64 => handler!(self, |OpsRD(tg, imm)| self.write_reg(tg, imm)),
-                    LRA => handler!(self, |OpsRRO(tg, reg, off)| self.write_reg(
+                    I::LI8 => handler!(self, |OpsRB(tg, imm)| self.write_reg(tg, imm)),
+                    I::LI16 => handler!(self, |OpsRH(tg, imm)| self.write_reg(tg, imm)),
+                    I::LI32 => handler!(self, |OpsRW(tg, imm)| self.write_reg(tg, imm)),
+                    I::LI64 => handler!(self, |OpsRD(tg, imm)| self.write_reg(tg, imm)),
+                    I::LRA => handler!(self, |OpsRRO(tg, reg, off)| self.write_reg(
                        tg,
                        self.pcrel(off).wrapping_add(self.read_reg(reg).cast::<i64>()).get(),
                    )),
                    // Load. If loading more than register size, continue on adjecent registers
-                    LD => handler!(self, |OpsRRAH(dst, base, off, count)| self
+                    I::LD => handler!(self, |OpsRRAH(dst, base, off, count)| self
                        .load(dst, base, off, count)?),
                    // Store. Same rules apply as to LD
-                    ST => handler!(self, |OpsRRAH(dst, base, off, count)| self
+                    I::ST => handler!(self, |OpsRRAH(dst, base, off, count)| self
                        .store(dst, base, off, count)?),
-                    LDR => handler!(self, |OpsRROH(dst, base, off, count)| self.load(
+                    I::LDR => handler!(self, |OpsRROH(dst, base, off, count)| self.load(
                        dst,
                        base,
                        self.pcrel(off).get(),
                        count
                    )?),
-                    STR => handler!(self, |OpsRROH(dst, base, off, count)| self.store(
+                    I::STR => handler!(self, |OpsRROH(dst, base, off, count)| self.store(
                        dst,
                        base,
                        self.pcrel(off).get(),
                        count
                    )?),
-                    BMC => {
+                    I::BMC => {
                        // Block memory copy
                        match if let Some(copier) = &mut self.copier {
                            // There is some copier, poll.
@ -227,7 +234,7 @@ where
                            core::task::Poll::Pending => (),
                        }
                    }
-                    BRC => handler!(self, |OpsRRB(src, dst, count)| {
+                    I::BRC => handler!(self, |OpsRRB(src, dst, count)| {
                        // Block register copy
                        if src.checked_add(count).is_none() || dst.checked_add(count).is_none() {
                            return Err(VmRunError::RegOutOfBounds);
@ -239,11 +246,11 @@ where
                            usize::from(count),
                        );
                    }),
-                    JMP => {
+                    I::JMP => {
                        let OpsO(off) = self.decode();
                        self.pc = self.pc.wrapping_add(off);
                    }
-                    JAL => {
+                    I::JAL => {
                        // Jump and link. Save PC after this instruction to
                        // specified register and jump to reg + relative offset.
                        let OpsRRO(save, reg, offset) = self.decode();
@ -251,7 +258,7 @@ where
                        self.write_reg(save, self.pc.next::<OpsRRO>());
                        self.pc = self.pcrel(offset).wrapping_add(self.read_reg(reg).cast::<i64>());
                    }
-                    JALA => {
+                    I::JALA => {
                        // Jump and link. Save PC after this instruction to
                        // specified register and jump to reg
                        let OpsRRA(save, reg, offset) = self.decode();
@ -261,8 +268,8 @@ where
                            Address::new(self.read_reg(reg).cast::<u64>().wrapping_add(offset));
                    }
                    // Conditional jumps, jump only to immediates
-                    JEQ => self.cond_jmp::<u64>(Ordering::Equal),
-                    JNE => {
+                    I::JEQ => self.cond_jmp::<u64>(Ordering::Equal),
+                    I::JNE => {
                        let OpsRRP(a0, a1, ja) = self.decode();
                        if self.read_reg(a0).cast::<u64>() != self.read_reg(a1).cast::<u64>() {
                            self.pc = self.pcrel(ja);
@ -270,11 +277,11 @@ where
                            self.bump_pc::<OpsRRP>();
                        }
                    }
-                    JLTS => self.cond_jmp::<i64>(Ordering::Less),
-                    JGTS => self.cond_jmp::<i64>(Ordering::Greater),
-                    JLTU => self.cond_jmp::<u64>(Ordering::Less),
-                    JGTU => self.cond_jmp::<u64>(Ordering::Greater),
-                    ECA => {
+                    I::JLTS => self.cond_jmp::<i64>(Ordering::Less),
+                    I::JGTS => self.cond_jmp::<i64>(Ordering::Greater),
+                    I::JLTU => self.cond_jmp::<u64>(Ordering::Less),
+                    I::JGTU => self.cond_jmp::<u64>(Ordering::Greater),
+                    I::ECA => {
                        // So we don't get timer interrupt after ECALL
                        if TIMER_QUOTIENT != 0 {
                            self.timer = self.timer.wrapping_add(1);
@ -283,33 +290,33 @@ where
                        self.bump_pc::<OpsN>();
                        return Ok(VmRunOk::Ecall);
                    }
-                    EBP => {
+                    I::EBP => {
                        self.bump_pc::<OpsN>();
                        return Ok(VmRunOk::Breakpoint);
                    }
-                    FADD32 => self.binary_op::<f32>(ops::Add::add),
-                    FADD64 => self.binary_op::<f64>(ops::Add::add),
-                    FSUB32 => self.binary_op::<f32>(ops::Sub::sub),
-                    FSUB64 => self.binary_op::<f64>(ops::Sub::sub),
-                    FMUL32 => self.binary_op::<f32>(ops::Mul::mul),
-                    FMUL64 => self.binary_op::<f64>(ops::Mul::mul),
-                    FDIV32 => self.binary_op::<f32>(ops::Div::div),
-                    FDIV64 => self.binary_op::<f64>(ops::Div::div),
-                    FMA32 => self.fma::<f32>(),
-                    FMA64 => self.fma::<f64>(),
-                    FINV32 => handler!(self, |OpsRR(tg, reg)| self
+                    I::FADD32 => self.binary_op::<f32>(ops::Add::add),
+                    I::FADD64 => self.binary_op::<f64>(ops::Add::add),
+                    I::FSUB32 => self.binary_op::<f32>(ops::Sub::sub),
+                    I::FSUB64 => self.binary_op::<f64>(ops::Sub::sub),
+                    I::FMUL32 => self.binary_op::<f32>(ops::Mul::mul),
+                    I::FMUL64 => self.binary_op::<f64>(ops::Mul::mul),
+                    I::FDIV32 => self.binary_op::<f32>(ops::Div::div),
+                    I::FDIV64 => self.binary_op::<f64>(ops::Div::div),
+                    I::FMA32 => self.fma::<f32>(),
+                    I::FMA64 => self.fma::<f64>(),
+                    I::FINV32 => handler!(self, |OpsRR(tg, reg)| self
                        .write_reg(tg, 1. / self.read_reg(reg).cast::<f32>())),
-                    FINV64 => handler!(self, |OpsRR(tg, reg)| self
+                    I::FINV64 => handler!(self, |OpsRR(tg, reg)| self
                        .write_reg(tg, 1. / self.read_reg(reg).cast::<f64>())),
-                    FCMPLT32 => self.fcmp::<f32>(Ordering::Less),
-                    FCMPLT64 => self.fcmp::<f64>(Ordering::Less),
-                    FCMPGT32 => self.fcmp::<f32>(Ordering::Greater),
-                    FCMPGT64 => self.fcmp::<f64>(Ordering::Greater),
-                    ITF32 => handler!(self, |OpsRR(tg, reg)| self
+                    I::FCMPLT32 => self.fcmp::<f32>(Ordering::Less),
+                    I::FCMPLT64 => self.fcmp::<f64>(Ordering::Less),
+                    I::FCMPGT32 => self.fcmp::<f32>(Ordering::Greater),
+                    I::FCMPGT64 => self.fcmp::<f64>(Ordering::Greater),
+                    I::ITF32 => handler!(self, |OpsRR(tg, reg)| self
                        .write_reg(tg, self.read_reg(reg).cast::<i64>() as f32)),
-                    ITF64 => handler!(self, |OpsRR(tg, reg)| self
+                    I::ITF64 => handler!(self, |OpsRR(tg, reg)| self
                        .write_reg(tg, self.read_reg(reg).cast::<i64>() as f64)),
-                    FTI32 => handler!(self, |OpsRRB(tg, reg, mode)| self.write_reg(
+                    I::FTI32 => handler!(self, |OpsRRB(tg, reg, mode)| self.write_reg(
                        tg,
                        crate::float::f32toint(
                            self.read_reg(reg).cast::<f32>(),
@ -317,7 +324,7 @@ where
                                .map_err(|()| VmRunError::InvalidOperand)?,
                        ),
                    )),
-                    FTI64 => handler!(self, |OpsRRB(tg, reg, mode)| self.write_reg(
+                    I::FTI64 => handler!(self, |OpsRRB(tg, reg, mode)| self.write_reg(
                        tg,
                        crate::float::f64toint(
                            self.read_reg(reg).cast::<f64>(),
@ -325,9 +332,9 @@ where
                                .map_err(|()| VmRunError::InvalidOperand)?,
                        ),
                    )),
-                    FC32T64 => handler!(self, |OpsRR(tg, reg)| self
+                    I::FC32T64 => handler!(self, |OpsRR(tg, reg)| self
                        .write_reg(tg, self.read_reg(reg).cast::<f32>() as f64)),
-                    FC64T32 => handler!(self, |OpsRRB(tg, reg, mode)| self.write_reg(
+                    I::FC64T32 => handler!(self, |OpsRRB(tg, reg, mode)| self.write_reg(
                        tg,
                        crate::float::conv64to32(
                            self.read_reg(reg).cast(),
@ -335,27 +342,26 @@ where
                                .map_err(|()| VmRunError::InvalidOperand)?,
                        ),
                    )),
-                    LRA16 => handler!(self, |OpsRRP(tg, reg, imm)| self.write_reg(
+                    I::LRA16 => handler!(self, |OpsRRP(tg, reg, imm)| self.write_reg(
                        tg,
                        (self.pc + self.read_reg(reg).cast::<u64>() + imm + 3_u16).get(),
                    )),
-                    LDR16 => handler!(self, |OpsRRPH(dst, base, off, count)| self.load(
+                    I::LDR16 => handler!(self, |OpsRRPH(dst, base, off, count)| self.load(
                        dst,
                        base,
                        self.pcrel(off).get(),
                        count
                    )?),
-                    STR16 => handler!(self, |OpsRRPH(dst, base, off, count)| self.store(
+                    I::STR16 => handler!(self, |OpsRRPH(dst, base, off, count)| self.store(
                        dst,
                        base,
                        self.pcrel(off).get(),
                        count
                    )?),
-                    JMP16 => {
+                    I::JMP16 => {
                        let OpsP(off) = self.decode();
                        self.pc = self.pcrel(off);
                    }
-                    op => return Err(VmRunError::InvalidOpcode(op)),
                }
            }
Author	SHA1	Message	Date
mlokr	29be2ba9c3	moving codegen to instrs alongside disasm	2024-09-19 18:25:28 +02:00
mlokr	31c501c643	merge	2024-09-19 17:15:03 +02:00
mlokr	0d118c17b2	reducing dependencies	2024-09-19 17:13:51 +02:00
mlokr	e4e7f8d5b5	implementing ableos executable format	2024-09-19 13:40:03 +02:00
mlokr	4849807353	removing git support and relative path prefix which did nothing anyway	2024-09-18 10:34:07 +02:00
mlokr	6e30968c54	improving one particular error message	2024-09-18 10:14:17 +02:00
mlokr	6fc0eb3498	triing to turn absolute to relative paths in error messages	2024-09-18 10:07:40 +02:00
mlokr	98dfd6b09c	improving parser error messages	2024-09-18 09:47:52 +02:00
mlokr	ece9bb8bf2	eca now infers the return type	2024-09-17 18:11:07 +02:00
mlokr	09fcbbc03b	forcing structs to always be on stack	2024-09-17 18:07:15 +02:00
mlokr	a7fda408ef	forgot update	2024-09-17 17:59:32 +02:00
mlokr	5d77ae93b4	forcing structs to always be on stack	2024-09-17 17:59:03 +02:00
mlokr	4a9b9de87f	nah, lets use dummer code	2024-09-17 15:50:45 +02:00
mlokr	bba3570788	adding wide return move for wider range of cases	2024-09-17 15:47:23 +02:00
mlokr	6852452f1a	fixing wide return value	2024-09-17 15:25:31 +02:00