moving codegen to instrs alongside disasm

This commit is contained in:
mlokr 2024-09-19 18:27:25 +02:00
parent 31c501c643
commit ede18f86f8
No known key found for this signature in database
GPG key ID: DEA147DDEE644993
16 changed files with 881 additions and 578 deletions

5
.gitignore vendored
View file

@ -1,7 +1,4 @@
/target
/hbbytecode/src/opcode.rs
/hbbytecode/src/ops.rs
/hblang/src/instrs.rs
/hblang/src/disasm.rs
/hbbytecode/src/instrs.rs
/.rgignore
rust-ice-*

295
Cargo.lock generated Normal file
View file

@ -0,0 +1,295 @@
# This file is automatically @generated by Cargo.
# It is not intended for manual editing.
version = 3
[[package]]
name = "ahash"
version = "0.8.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e89da841a80418a9b391ebaea17f5c112ffaaa96f621d2c285b5174da76b9011"
dependencies = [
"cfg-if",
"once_cell",
"version_check",
"zerocopy",
]
[[package]]
name = "arbitrary"
version = "1.3.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7d5a26814d8dcb93b0e5a0ff3c6d80a8843bafb21b39e8e18a6f05471870e110"
[[package]]
name = "bumpalo"
version = "3.16.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "79296716171880943b8470b5f8d03aa55eb2e645a4874bdbb28adb49162e012c"
[[package]]
name = "cfg-if"
version = "1.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
[[package]]
name = "cranelift-bforest"
version = "0.113.0"
source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
dependencies = [
"cranelift-entity",
]
[[package]]
name = "cranelift-bitset"
version = "0.113.0"
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#a27ac7aed78fae886e45f119139a1e54b20dd523"
dependencies = [
"bumpalo",
"cranelift-bforest",
"cranelift-bitset",
"cranelift-codegen-meta",
"cranelift-codegen-shared",
"cranelift-control",
"cranelift-entity",
"cranelift-isle",
"hashbrown",
"log",
"regalloc2 0.10.2 (registry+https://github.com/rust-lang/crates.io-index)",
"rustc-hash",
"smallvec",
"target-lexicon",
]
[[package]]
name = "cranelift-codegen-meta"
version = "0.113.0"
source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
dependencies = [
"cranelift-codegen-shared",
]
[[package]]
name = "cranelift-codegen-shared"
version = "0.113.0"
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#a27ac7aed78fae886e45f119139a1e54b20dd523"
dependencies = [
"arbitrary",
]
[[package]]
name = "cranelift-entity"
version = "0.113.0"
source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
dependencies = [
"cranelift-bitset",
]
[[package]]
name = "cranelift-isle"
version = "0.113.0"
source = "git+https://github.com/jakubDoka/wasmtime.git#a27ac7aed78fae886e45f119139a1e54b20dd523"
[[package]]
name = "hashbrown"
version = "0.14.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e5274423e17b7c9fc20b6e7e208532f9b19825d82dfd615708b70edd83df41f1"
dependencies = [
"ahash",
]
[[package]]
name = "hbbytecode"
version = "0.1.0"
[[package]]
name = "hbcb"
version = "0.1.0"
dependencies = [
"cranelift-codegen",
"cranelift-codegen-meta",
"cranelift-control",
"cranelift-isle",
"log",
"regalloc2 0.10.2 (registry+https://github.com/rust-lang/crates.io-index)",
"smallvec",
"target-lexicon",
]
[[package]]
name = "hbjit"
version = "0.1.0"
[[package]]
name = "hblang"
version = "0.1.0"
dependencies = [
"hbbytecode",
"hbvm",
"regalloc2 0.10.2 (git+https://github.com/jakubDoka/regalloc2)",
]
[[package]]
name = "hbvm"
version = "0.1.0"
dependencies = [
"hbbytecode",
]
[[package]]
name = "hbxrt"
version = "0.1.0"
dependencies = [
"hbvm",
"memmap2",
]
[[package]]
name = "libc"
version = "0.2.158"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d8adc4bb1803a324070e64a98ae98f38934d91957a99cfb3a43dcbc01bc56439"
[[package]]
name = "log"
version = "0.4.22"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a7a70ba024b9dc04c27ea2f0c0548feb474ec5c54bba33a7f72f873a39d07b24"
[[package]]
name = "memmap2"
version = "0.9.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fd3f7eed9d3848f8b98834af67102b720745c4ec028fcd0aa0239277e7de374f"
dependencies = [
"libc",
]
[[package]]
name = "once_cell"
version = "1.19.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3fdb12b2476b595f9358c5161aa467c2438859caa136dec86c26fdd2efe17b92"
[[package]]
name = "proc-macro2"
version = "1.0.86"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5e719e8df665df0d1c8fbfd238015744736151d4445ec0836b8e628aae103b77"
dependencies = [
"unicode-ident",
]
[[package]]
name = "quote"
version = "1.0.37"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b5b9d34b8991d19d98081b46eacdd8eb58c6f2b201139f7c5f643cc155a633af"
dependencies = [
"proc-macro2",
]
[[package]]
name = "regalloc2"
version = "0.10.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "12908dbeb234370af84d0579b9f68258a0f67e201412dd9a2814e6f45b2fc0f0"
dependencies = [
"hashbrown",
"log",
"rustc-hash",
"slice-group-by",
"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"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "583034fd73374156e66797ed8e5b0d5690409c9226b22d87cb7f19821c05d152"
[[package]]
name = "slice-group-by"
version = "0.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "826167069c09b99d56f31e9ae5c99049e932a98c9dc2dac47645b08dbbf76ba7"
[[package]]
name = "smallvec"
version = "1.13.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3c5e1a9a646d36c3599cd173a41282daf47c44583ad367b8e6837255952e5c67"
[[package]]
name = "syn"
version = "2.0.77"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9f35bcdf61fd8e7be6caf75f429fdca8beb3ed76584befb503b1569faee373ed"
dependencies = [
"proc-macro2",
"quote",
"unicode-ident",
]
[[package]]
name = "target-lexicon"
version = "0.12.16"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "61c41af27dd6d1e27b1b16b489db798443478cef1f06a660c96db617ba5de3b1"
[[package]]
name = "unicode-ident"
version = "1.0.13"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e91b56cd4cadaeb79bbf1a5645f6b4f8dc5bde8834ad5894a8db35fda9efa1fe"
[[package]]
name = "version_check"
version = "0.9.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
[[package]]
name = "xtask"
version = "0.1.0"
[[package]]
name = "zerocopy"
version = "0.7.35"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1b9b4fd18abc82b8136838da5d50bae7bdea537c574d8dc1a34ed098d6c166f0"
dependencies = [
"zerocopy-derive",
]
[[package]]
name = "zerocopy-derive"
version = "0.7.35"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fa4f8080344d4671fb4e831a13ad1e68092748387dfc4f55e356242fae12ce3e"
dependencies = [
"proc-macro2",
"quote",
"syn",
]

View file

@ -2,3 +2,8 @@
name = "hbbytecode"
version = "0.1.0"
edition = "2018"
[features]
default = ["disasm"]
std = []
disasm = ["std"]

View file

@ -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')
}

View file

@ -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(())
}

View file

@ -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"

View file

@ -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,

View file

@ -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();

View file

@ -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.

View file

@ -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]

View file

@ -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()
}

View file

@ -8,5 +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" }

View file

@ -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')
}

View file

@ -27,6 +27,7 @@ use {
son::reg,
ty::ArrayLen,
},
hbbytecode as instrs,
parser::Ast,
std::{
collections::{hash_map, BTreeMap, VecDeque},
@ -52,7 +53,6 @@ pub mod codegen;
pub mod parser;
pub mod son;
mod instrs;
mod lexer;
mod task {
@ -715,7 +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()
@ -744,7 +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 {
@ -857,181 +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;
}
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.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>>,
}

View file

@ -9,4 +9,4 @@ alloc = []
nightly = []
[dependencies]
hbbytecode = { path = "../hbbytecode" }
hbbytecode = { path = "../hbbytecode", default-features = false }

View file

@ -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)),
}
}