Update homework 1 and 2

This commit is contained in:
Jeehoon Kang
2020-03-26 03:38:20 +09:00
parent 241a66fcc1
commit 8938a7ad8f
40 changed files with 5171 additions and 1504 deletions

View File

@@ -34,7 +34,7 @@ cargo test --release # release build test
cargo test <test-name> # run a particular test
```
`<test-name>` can be `test_ast_print`, `ir_smoke`, ...
`<test-name>` can be `test_examples_write_c`, `test_examples_irgen`, ...
## Fuzzing
@@ -44,10 +44,7 @@ cargo test <test-name> # run a particular test
```sh
# Ubuntu 18.04 or higher
apt install -y make cmake python3
# MacOS
xcode-select install
brew install cmake python3
apt install -y csmith libcsmith-dev creduce
```
### Run
@@ -59,5 +56,28 @@ python3 tests/fuzz.py --help # print options
python3 tests/fuzz.py --print -n10 # test C AST printer for 10 times
```
We use [Csmith](https://embed.cs.utah.edu/csmith/) to randomly generate C source codes. Csmith will
be automatically downloaded and built by the test script.
We use `csmith` to randomly generate C source codes. `csmith` will be automatically downloaded and
built by the test script. For more information, we refer to the
[Csmith](https://embed.cs.utah.edu/csmith/) homepage.
### Reduce
When the fuzzer finds a buggy input program for your compiler, it is highly likely that the input
program is too big to manually inspect. We use `creduce` that reduces the buggy input program as
much as possible.
Suppose `tests/test_polished.c` is the buggy input program. Then the following script reduces the
program to `tests/test_reduced.c`:
```sh
python3 tests/fuzz.py --reduce <fuzz-option>
```
`<fuzz-option>` can be `--print` or `--irgen`. It shall be the one used in [Run](#run).
### How it reduces test case?
The script performs unguided test-case reduction using `creduce`: given a buggy program, it randomly
reduces the program; check if the reduced program still fails on the test, and if so, replaces the
given program with the reduced one; repeat until you get a small enough buggy program. For more
information, we refer to the [Creduce](https://embed.cs.utah.edu/creduce/) homepage.

View File

@@ -6,6 +6,7 @@ use clap::{crate_authors, crate_description, crate_version, App};
extern crate kecc;
use kecc::{Parse, Translate};
use std::path::Path;
fn main() {
let yaml = load_yaml!("fuzz_cli.yml");
@@ -20,7 +21,10 @@ fn main() {
let unit = ok_or_exit!(Parse::default().translate(&input), 1);
if matches.is_present("print") {
kecc::write_c_test(&unit);
return;
kecc::test_write_c(&unit, Path::new(&input));
}
if matches.is_present("irgen") {
kecc::test_irgen(&unit, Path::new(&input));
}
}

View File

@@ -4,6 +4,10 @@ args:
short: p
long: print
help: Fuzzes C AST printer
- irgen:
short: i
long: irgen
help: Fuzzes irgen
- INPUT:
help: Sets the input file to use
required: true

View File

@@ -5,7 +5,7 @@ use clap::{crate_authors, crate_description, crate_version, App};
#[macro_use]
extern crate kecc;
use kecc::{Codegen, Irgen, Optimize, Parse, Translate, O1};
use kecc::{write, Asmgen, Irgen, Optimize, Parse, Translate, O1};
fn main() {
let yaml = load_yaml!("kecc_cli.yml");
@@ -27,7 +27,7 @@ fn main() {
};
if matches.is_present("print") {
kecc::write_c(&unit, &mut output).unwrap();
write(&unit, &mut output).unwrap();
return;
}
@@ -39,7 +39,7 @@ fn main() {
}
};
if matches.is_present("irgen") {
kecc::write_ir(&ir, &mut output).unwrap();
write(&ir, &mut output).unwrap();
return;
}
@@ -47,6 +47,6 @@ fn main() {
O1::default().optimize(&mut ir);
}
let asm = ok_or_exit!(Codegen::default().translate(&ir), 1);
kecc::write_asm(&asm, &mut output);
let asm = ok_or_exit!(Asmgen::default().translate(&ir), 1);
write(&asm, &mut output).unwrap();
}

1
examples/.gitignore vendored Normal file
View File

@@ -0,0 +1 @@
*.exe

3205
examples/hw1/complete_cond.c Normal file

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,13 @@
int main() {
int i;
int p = 2;
int q = 5;
int r = (0 ? ((p > q) ? (p -= 2) : (p += 2)) : (p + q));
for (i = 0; i < 11; ((i % 2) ? (i += 2) : ++i)) {
if (i % 2) { p += q; }
else { p += r; }
}
return p == 34;
}

15
examples/hw1/gcd.c Normal file
View File

@@ -0,0 +1,15 @@
int gcd(int a, int b) {
a = (a > 0) ? a : -a;
b = (b > 0) ? b : -b;
while(a != b) {
if(a > b) { a -= b; }
else { b -= a; }
}
return a;
}
int main() {
return gcd(18, 21) == 3;
}

View File

@@ -0,0 +1,53 @@
#!/usr/bin/env python3
"""Make c program which uses complicated conditional expression
To make c program, execute `python3 make_cond.py > file_name.c`
"""
import random
def eval_cond(arr_cond):
"""Evaluate conditional expression
"""
if len(arr_cond) == 1:
return arr_cond[0]
new_arr_cond = []
for cond_start in range(len(arr_cond) // 3):
cond_val = arr_cond[3*cond_start + 1] if arr_cond[3*cond_start] else arr_cond[3*cond_start + 2]
new_arr_cond.append(cond_val)
return eval_cond(new_arr_cond)
def make_func(i):
"""Make a function that contains a conditional expression
"""
func_signature = "int " + "func_" + str(i) + "()"
variables = "abcdefghijklmnopqrstuvwxyzA"
func_inner = []
val_bitmap = []
# Variable initializiation
for var in variables:
val = random.randint(0, 1)
val_bitmap.append(val)
decl = "\tint " + var + " = " + str(val) + ";"
func_inner.append(decl)
expr_val = eval_cond(val_bitmap)
func_inner.append("\treturn (((a ? b : c) ? (d ? e : f) : (g ? h : i)) ? ((j ? k : l) ? (m ? n : o) : (p ? q : r)) : ((s ? t : u) ? (v ? w : x) : (y ? z : A))) == " + str(expr_val) + ";")
return "\n".join([func_signature, "{"] + func_inner + ["}"])
if __name__ == "__main__":
src = ""
return_stmt = "\treturn ("
NUM_FUNC = 100
for i in range(NUM_FUNC):
src += make_func(i)
src += "\n\n"
return_stmt += "func_" + str(i) + "()"
return_stmt += " && " if i != (NUM_FUNC - 1) else ") == "
return_stmt += "1;"
src += "int main()\n{\n" + return_stmt + "\n}\n"
print(src)

10
examples/simple_cond.c Normal file
View File

@@ -0,0 +1,10 @@
int f(int x) {
return x + 8;
}
int main() {
int x = 0;
int y = (x++ == 1) ? 1 : 2;
return f((x < y) ? x : 2) == 9;
}

3
scripts/copy-to-kecc-public.sh Executable file
View File

@@ -0,0 +1,3 @@
#!/usr/bin/env bash
rsync --exclude=".git" --delete --archive ./ ../kecc-public

View File

@@ -11,8 +11,9 @@ import sys
import re
if __name__ == "__main__":
for fullname in os.listdir("src"):
(filename, ext) = os.path.splitext(fullname)
for root, dir, files in os.walk("src"):
for f in files:
(filename, ext) = os.path.splitext(f)
if ext == ".public":
os.rename(os.path.join("src", fullname), os.path.join("src", filename))
if ext == ".public":
os.rename(os.path.join(root, f), os.path.join(root, filename))

View File

@@ -1,2 +1,4 @@
mod write_asm;
/// TODO
pub struct Asm {}

10
src/asm/write_asm.rs Normal file
View File

@@ -0,0 +1,10 @@
use std::io::{Result, Write};
use crate::asm::Asm;
use crate::write_base::WriteLine;
impl WriteLine for Asm {
fn write_line(&self, _indent: usize, _write: &mut dyn Write) -> Result<()> {
todo!()
}
}

View File

@@ -3,13 +3,13 @@ use crate::ir;
use crate::Translate;
#[derive(Default)]
pub struct Codegen {}
pub struct Asmgen {}
impl Translate<ir::TranslationUnit> for Codegen {
impl Translate<ir::TranslationUnit> for Asmgen {
type Target = Asm;
type Error = ();
fn translate(&mut self, _source: &ir::TranslationUnit) -> Result<Self::Target, Self::Error> {
unimplemented!()
todo!()
}
}

View File

@@ -3,7 +3,7 @@
use lang_c::ast::*;
use lang_c::span::Node;
use std::ops::Deref;
use core::ops::Deref;
use itertools::izip;

6
src/c/mod.rs Normal file
View File

@@ -0,0 +1,6 @@
mod ast_equiv;
mod parse;
mod write_c;
pub use ast_equiv::assert_ast_equiv;
pub use parse::Parse;

View File

@@ -1,4 +1,4 @@
use std::ops::Deref;
use core::ops::Deref;
use std::path::Path;
use lang_c::ast::*;
@@ -10,6 +10,7 @@ use crate::Translate;
#[derive(Debug)]
pub enum Error {
ParseError(ParseError),
#[allow(dead_code)]
Unsupported,
}
@@ -101,7 +102,9 @@ impl AssertSupported for FunctionDefinition {
impl AssertSupported for DeclarationSpecifier {
fn assert_supported(&self) {
match self {
Self::StorageClass(_) => panic!("DeclarationSpecifier::StorageClass"),
Self::StorageClass(storage_class_specifier) => {
storage_class_specifier.assert_supported()
}
Self::TypeSpecifier(type_specifier) => type_specifier.assert_supported(),
Self::TypeQualifier(type_qualifier) => type_qualifier.assert_supported(),
Self::Function(_) => panic!("DeclarationSpecifier::Function"),
@@ -111,6 +114,15 @@ impl AssertSupported for DeclarationSpecifier {
}
}
impl AssertSupported for StorageClassSpecifier {
fn assert_supported(&self) {
match self {
Self::Typedef => (),
_ => panic!("StorageClassifier other than Typedef"),
}
}
}
impl AssertSupported for TypeSpecifier {
fn assert_supported(&self) {
match self {
@@ -128,7 +140,7 @@ impl AssertSupported for TypeSpecifier {
Self::Atomic(_) => panic!("TypeSpecifier::Atomic"),
Self::Struct(struct_type) => struct_type.assert_supported(),
Self::Enum(_) => panic!("TypeSpecifier::Enum"),
Self::TypedefName(_) => panic!("TypeSpecifier::TypedefName"),
Self::TypedefName(_) => (),
Self::TypeOf(_) => panic!("TypeSpecifier::TypeOf"),
Self::TS18661Float(_) => panic!("TypeSpecifier::TS18661Float"),
}
@@ -271,7 +283,7 @@ impl AssertSupported for ParameterDeclaration {
impl AssertSupported for DeclaratorKind {
fn assert_supported(&self) {
match self {
Self::Abstract => panic!("DeclaratorKind::Abstract"),
Self::Abstract => (),
Self::Identifier(_) => (),
Self::Declarator(decl) => decl.assert_supported(),
}
@@ -415,8 +427,8 @@ impl AssertSupported for Expression {
match self {
Self::Identifier(_) => (),
Self::Constant(constant) => constant.assert_supported(),
Self::StringLiteral(_) => panic!("Expression:StringLiteral"),
Self::GenericSelection(_) => panic!("Expression:GenericSelection"),
Self::StringLiteral(_) => panic!("Expression::StringLiteral"),
Self::GenericSelection(_) => panic!("Expression::GenericSelection"),
Self::Member(member) => member.assert_supported(),
Self::Call(call) => call.assert_supported(),
Self::CompoundLiteral(_) => panic!("Expression::CompoundLiteral"),
@@ -531,11 +543,7 @@ impl AssertSupported for FloatFormat {
}
impl AssertSupported for UnaryOperator {
fn assert_supported(&self) {
if let Self::SizeOf = self {
panic!("UnaryOperaotr::SizeOf")
}
}
fn assert_supported(&self) {}
}
impl AssertSupported for BinaryOperator {

View File

@@ -1,8 +1,8 @@
use lang_c::ast::*;
use lang_c::span::Node;
use core::ops::Deref;
use std::io::{Result, Write};
use std::ops::Deref;
use crate::write_base::*;
@@ -45,7 +45,3 @@ impl WriteLine for TranslationUnit {
todo!("homework 1")
}
}
pub fn write_c(unit: &TranslationUnit, write: &mut dyn Write) -> Result<()> {
unit.write_line(0, write)
}

1096
src/ir.rs

File diff suppressed because it is too large Load Diff

612
src/ir/dtype.rs Normal file
View File

@@ -0,0 +1,612 @@
use core::convert::TryFrom;
use core::fmt;
use core::ops::Deref;
use itertools::izip;
use lang_c::ast;
use lang_c::span::Node;
use std::hash::Hash;
use failure::Fail;
#[derive(Debug, PartialEq, Fail)]
pub enum DtypeError {
/// For uncommon error
#[fail(display = "{}", message)]
Misc { message: String },
}
pub trait HasDtype {
fn dtype(&self) -> Dtype;
}
#[derive(Default)]
struct BaseDtype {
scalar: Option<ast::TypeSpecifier>,
signed_option: Option<ast::TypeSpecifier>,
is_const: bool,
}
#[derive(Debug, PartialEq, Eq, Hash, Clone)]
pub enum Dtype {
Unit {
is_const: bool,
},
Int {
width: usize,
is_signed: bool,
is_const: bool,
},
Float {
width: usize,
is_const: bool,
},
Pointer {
inner: Box<Dtype>,
is_const: bool,
},
Function {
ret: Box<Dtype>,
params: Vec<Dtype>,
},
}
impl BaseDtype {
/// Apply `TypeSpecifier` to `BaseDtype`
///
/// let's say declaration is `const int a;`, if `self` represents `int`
/// and `type_specifier` represents `const`, `self` is transformed to
/// representing `const int` after function performs.
///
/// # Arguments
///
/// * `self` - Part that has been converted to 'BaseDtype' on the declaration
/// * `type_qualifier` - type qualifiers requiring apply to 'self' immediately
///
#[inline]
fn apply_type_specifier(
&mut self,
type_specifier: &ast::TypeSpecifier,
) -> Result<(), DtypeError> {
match type_specifier {
ast::TypeSpecifier::Unsigned | ast::TypeSpecifier::Signed => {
if self.signed_option.is_some() {
return Err(DtypeError::Misc {
message: "duplicate signed option".to_string(),
});
}
self.signed_option = Some(type_specifier.clone());
}
ast::TypeSpecifier::Void
| ast::TypeSpecifier::Char
| ast::TypeSpecifier::Int
| ast::TypeSpecifier::Float => {
if self.scalar.is_some() {
return Err(DtypeError::Misc {
message: "two or more scalar types in declaration specifiers".to_string(),
});
}
self.scalar = Some(type_specifier.clone());
}
_ => todo!("support more like `double` in the future"),
}
Ok(())
}
/// Apply `Typequalifier` to `BaseDtype`
///
/// let's say declaration is `const int a;`, if `self` represents `int`
/// and `type_qualifier` represents `const`, `self` is transformed to
/// representing `const int` after function performs.
///
/// # Arguments
///
/// * `self` - Part that has been converted to 'BaseDtype' on the declaration
/// * `type_qualifier` - type qualifiers requiring apply to 'self' immediately
///
#[inline]
fn apply_type_qualifier(
&mut self,
type_qualifier: &ast::TypeQualifier,
) -> Result<(), DtypeError> {
match type_qualifier {
ast::TypeQualifier::Const => {
// duplicate `const` is allowed
self.is_const = true;
}
_ => panic!("type qualifier is unsupported except `const`"),
}
Ok(())
}
pub fn apply_typename_specifier(
&mut self,
typename_specifier: &ast::SpecifierQualifier,
) -> Result<(), DtypeError> {
match typename_specifier {
ast::SpecifierQualifier::TypeSpecifier(type_specifier) => {
self.apply_type_specifier(&type_specifier.node)?
}
ast::SpecifierQualifier::TypeQualifier(type_qualifier) => {
self.apply_type_qualifier(&type_qualifier.node)?
}
}
Ok(())
}
pub fn apply_declaration_specifier(
&mut self,
declaration_specifier: &ast::DeclarationSpecifier,
) -> Result<(), DtypeError> {
match declaration_specifier {
// TODO: `dtype` must be defined taking into account all specifier information.
ast::DeclarationSpecifier::StorageClass(_storage_class_spec) => {
todo!("analyze storage class specifier keyword to create correct `dtype`")
}
ast::DeclarationSpecifier::TypeSpecifier(type_specifier) => {
self.apply_type_specifier(&type_specifier.node)?
}
ast::DeclarationSpecifier::TypeQualifier(type_qualifier) => {
self.apply_type_qualifier(&type_qualifier.node)?
}
_ => panic!("is_unsupported"),
}
Ok(())
}
/// Apply `PointerQualifier` to `BaseDtype`
///
/// let's say pointer declarator is `* const` of `const int * const a;`.
/// If `self` represents nothing, and `pointer_qualifier` represents `const`
/// between first and second asterisk, `self` is transformed to
/// representing `const` after function performs. This information is used later
/// when generating `Dtype`.
///
/// # Arguments
///
/// * `self` - Part that has been converted to 'BaseDtype' on the pointer declarator
/// * `pointer_qualifier` - Pointer qualifiers requiring apply to 'BaseDtype' immediately
///
pub fn apply_pointer_qualifier(
&mut self,
pointer_qualifier: &ast::PointerQualifier,
) -> Result<(), DtypeError> {
match pointer_qualifier {
ast::PointerQualifier::TypeQualifier(type_qualifier) => {
self.apply_type_qualifier(&type_qualifier.node)?;
}
ast::PointerQualifier::Extension(_) => {
panic!("ast::PointerQualifier::Extension is unsupported")
}
}
Ok(())
}
pub fn apply_typename_specifiers(
&mut self,
typename_specifiers: &[Node<ast::SpecifierQualifier>],
) -> Result<(), DtypeError> {
for ast_spec in typename_specifiers {
self.apply_typename_specifier(&ast_spec.node)?;
}
Ok(())
}
pub fn apply_declaration_specifiers(
&mut self,
declaration_specifiers: &[Node<ast::DeclarationSpecifier>],
) -> Result<(), DtypeError> {
for ast_spec in declaration_specifiers {
self.apply_declaration_specifier(&ast_spec.node)?;
}
Ok(())
}
}
impl TryFrom<BaseDtype> for Dtype {
type Error = DtypeError;
/// Derive a data type containing scalar type from specifiers.
///
/// # Example
///
/// For declaration is `const unsigned int * p`, `specifiers` is `const unsigned int`,
/// and the result is `Dtype::Int{ width: 32, is_signed: false, is_const: ture }`
fn try_from(spec: BaseDtype) -> Result<Self, DtypeError> {
assert!(
!(spec.scalar.is_none() && spec.signed_option.is_none() && !spec.is_const),
"BaseDtype is empty"
);
// Creates `dtype` from scalar.
let mut dtype = if let Some(t) = spec.scalar {
match t {
ast::TypeSpecifier::Void => Self::unit(),
ast::TypeSpecifier::Unsigned | ast::TypeSpecifier::Signed => {
panic!("Signed option to scalar is not supported")
}
ast::TypeSpecifier::Bool => Self::BOOL,
ast::TypeSpecifier::Char => Self::CHAR,
ast::TypeSpecifier::Short => Self::SHORT,
ast::TypeSpecifier::Int => Self::INT,
ast::TypeSpecifier::Long => Self::LONG,
ast::TypeSpecifier::Float => Self::FLOAT,
ast::TypeSpecifier::Double => Self::DOUBLE,
_ => panic!("Unsupported ast::TypeSpecifier"),
}
} else {
Dtype::default()
};
// Applies signedness.
if let Some(signed_option) = spec.signed_option {
let is_signed = match signed_option {
ast::TypeSpecifier::Signed => true,
ast::TypeSpecifier::Unsigned => false,
_ => panic!(
"`signed_option` must be `TypeSpecifier::Signed` or `TypeSpecifier::Unsigned`"
),
};
dtype = dtype.set_signed(is_signed);
}
// Applies constness.
assert!(!dtype.is_const());
dtype = dtype.set_const(spec.is_const);
Ok(dtype)
}
}
impl TryFrom<&ast::TypeName> for Dtype {
type Error = DtypeError;
/// Derive a data type from typename.
fn try_from(type_name: &ast::TypeName) -> Result<Self, Self::Error> {
let mut spec = BaseDtype::default();
BaseDtype::apply_typename_specifiers(&mut spec, &type_name.specifiers)?;
let mut dtype = Self::try_from(spec)?;
if let Some(declarator) = &type_name.declarator {
dtype = dtype.with_ast_declarator(&declarator.node)?;
}
Ok(dtype)
}
}
impl TryFrom<&ast::ParameterDeclaration> for Dtype {
type Error = DtypeError;
/// Generate `Dtype` based on parameter declaration
fn try_from(parameter_decl: &ast::ParameterDeclaration) -> Result<Self, Self::Error> {
let mut spec = BaseDtype::default();
BaseDtype::apply_declaration_specifiers(&mut spec, &parameter_decl.specifiers)?;
let mut dtype = Self::try_from(spec)?;
if let Some(declarator) = &parameter_decl.declarator {
dtype = dtype.with_ast_declarator(&declarator.node)?;
}
Ok(dtype)
}
}
impl Dtype {
pub const BOOL: Self = Self::int(1);
pub const CHAR: Self = Self::int(8);
pub const SHORT: Self = Self::int(16);
pub const INT: Self = Self::int(32);
pub const LONG: Self = Self::int(64);
pub const LONGLONG: Self = Self::int(64);
pub const FLOAT: Self = Self::float(32);
pub const DOUBLE: Self = Self::float(64);
const WIDTH_OF_BYTE: usize = 8;
// TODO: consider architecture dependency in the future
const WIDTH_OF_POINTER: usize = 32;
#[inline]
pub const fn unit() -> Self {
Self::Unit { is_const: false }
}
#[inline]
pub const fn int(width: usize) -> Self {
Self::Int {
width,
is_signed: true,
is_const: false,
}
}
#[inline]
pub const fn float(width: usize) -> Self {
Self::Float {
width,
is_const: false,
}
}
#[inline]
pub fn pointer(inner: Dtype) -> Self {
Self::Pointer {
inner: Box::new(inner),
is_const: false,
}
}
#[inline]
pub fn function(ret: Dtype, params: Vec<Dtype>) -> Self {
Self::Function {
ret: Box::new(ret),
params,
}
}
#[inline]
pub fn get_int_width(&self) -> Option<usize> {
if let Self::Int { width, .. } = self {
Some(*width)
} else {
None
}
}
#[inline]
pub fn get_float_width(&self) -> Option<usize> {
if let Self::Float { width, .. } = self {
Some(*width)
} else {
None
}
}
#[inline]
pub fn get_pointer_inner(&self) -> Option<&Dtype> {
if let Self::Pointer { inner, .. } = self {
Some(inner.deref())
} else {
None
}
}
#[inline]
pub fn get_function_inner(&self) -> Option<(&Dtype, &Vec<Dtype>)> {
if let Self::Function { ret, params } = self {
Some((ret.deref(), params))
} else {
None
}
}
#[inline]
pub fn is_scalar(&self) -> bool {
match self {
Self::Unit { .. } => todo!(),
Self::Int { .. } => true,
Self::Float { .. } => true,
_ => false,
}
}
#[inline]
pub fn is_int_signed(&self) -> bool {
match self {
Self::Int { is_signed, .. } => *is_signed,
_ => panic!("only `Dtype::Int` can be judged whether it is sigend"),
}
}
#[inline]
pub fn is_const(&self) -> bool {
match self {
Self::Unit { is_const } => *is_const,
Self::Int { is_const, .. } => *is_const,
Self::Float { is_const, .. } => *is_const,
Self::Pointer { is_const, .. } => *is_const,
Self::Function { .. } => {
panic!("there should be no case that check whether `Function` is `const`")
}
}
}
pub fn set_const(self, is_const: bool) -> Self {
match self {
Self::Unit { .. } => Self::Unit { is_const },
Self::Int {
width, is_signed, ..
} => Self::Int {
width,
is_signed,
is_const,
},
Self::Float { width, .. } => Self::Float { width, is_const },
Self::Pointer { inner, .. } => Self::Pointer { inner, is_const },
Self::Function { .. } => panic!("`const` cannot be applied to `Dtype::Function`"),
}
}
/// Return byte size of `Dtype`
pub fn size_of(&self) -> Result<usize, DtypeError> {
// TODO: consider complex type like array, structure in the future
match self {
Self::Unit { .. } => Ok(0),
Self::Int { width, .. } => Ok(*width / Self::WIDTH_OF_BYTE),
Self::Float { width, .. } => Ok(*width / Self::WIDTH_OF_BYTE),
Self::Pointer { .. } => Ok(Self::WIDTH_OF_POINTER / Self::WIDTH_OF_BYTE),
Self::Function { .. } => Err(DtypeError::Misc {
message: "`sizeof` cannot be used with function types".to_string(),
}),
}
}
/// Return alignment requirements of `Dtype`
pub fn align_of(&self) -> Result<usize, DtypeError> {
// TODO: consider complex type like array, structure in the future
// TODO: when considering complex type like a structure,
// the calculation method should be different from `Dtype::size_of`.
match self {
Self::Unit { .. } => Ok(0),
Self::Int { width, .. } => Ok(*width / Self::WIDTH_OF_BYTE),
Self::Float { width, .. } => Ok(*width / Self::WIDTH_OF_BYTE),
Self::Pointer { .. } => Ok(Self::WIDTH_OF_POINTER / Self::WIDTH_OF_BYTE),
Self::Function { .. } => Err(DtypeError::Misc {
message: "`alignof` cannot be used with function types".to_string(),
}),
}
}
pub fn set_signed(self, is_signed: bool) -> Self {
match self {
Self::Int {
width, is_const, ..
} => Self::Int {
width,
is_signed,
is_const,
},
_ => panic!("`signed` and `unsigned` only be applied to `Dtype::Int`"),
}
}
/// Derive a data type from declaration specifiers.
pub fn try_from_ast_declaration_specifiers(
specifiers: &[Node<ast::DeclarationSpecifier>],
) -> Result<Self, DtypeError> {
let mut spec = BaseDtype::default();
BaseDtype::apply_declaration_specifiers(&mut spec, specifiers)?;
Self::try_from(spec)
}
/// Generate `Dtype` based on declarator and `base_dtype` which has scalar type.
///
/// let's say declaration is `const int * const * const a;`.
/// In general `base_dtype` start with `const int` which has scalar type and
/// `declarator` represents `* const * const` with `ast::Declarator`
///
/// # Arguments
///
/// * `declarator` - Parts requiring conversion to 'Dtype' on the declaration
/// * `base_dtype` - Part that has been converted to 'Dtype' on the declaration
///
pub fn with_ast_declarator(mut self, declarator: &ast::Declarator) -> Result<Self, DtypeError> {
for derived_decl in &declarator.derived {
self = match &derived_decl.node {
ast::DerivedDeclarator::Pointer(pointer_qualifiers) => {
let mut specifier = BaseDtype::default();
for qualifier in pointer_qualifiers {
specifier.apply_pointer_qualifier(&qualifier.node)?;
}
Self::pointer(self).set_const(specifier.is_const)
}
ast::DerivedDeclarator::Array(_array_decl) => todo!(),
ast::DerivedDeclarator::Function(func_decl) => {
let params = func_decl
.node
.parameters
.iter()
.map(|p| Self::try_from(&p.node))
.collect::<Result<Vec<_>, _>>()?;
Self::function(self, params)
}
ast::DerivedDeclarator::KRFunction(kr_func_decl) => {
// K&R function is allowed only when it has no parameter
assert!(kr_func_decl.is_empty());
Self::function(self, Vec::new())
}
};
}
let declarator_kind = &declarator.kind;
match &declarator_kind.node {
ast::DeclaratorKind::Abstract => panic!("ast::DeclaratorKind::Abstract is unsupported"),
ast::DeclaratorKind::Identifier(_) => Ok(self),
ast::DeclaratorKind::Declarator(declarator) => {
self.with_ast_declarator(&declarator.node)
}
}
}
/// Check whether type conflict exists between the two `Dtype` objects.
///
/// let's say expression is `const int a = 0; int b = 0; int c = a + b`.
/// Although `const int` of `a` and `int` of `b` looks different, `Plus`(+) operations between
/// these two types are possible without any type-casting. There is no conflict between
/// `const int` and `int`.
///
/// However, only the outermost const is ignored.
/// If check equivalence between `const int *const` and `int *`, result is false. Because
/// the second `const` (means left most `const`) of the `const int *const` is missed in `int *`.
/// By the way, outermost `const` (means right most `const`) is not a consideration here.
pub fn is_compatible(&self, other: &Self) -> bool {
match (self, other) {
(Self::Unit { .. }, Self::Unit { .. })
| (Self::Int { .. }, Self::Int { .. })
| (Self::Float { .. }, Self::Float { .. })
| (Self::Pointer { .. }, Self::Pointer { .. }) => {
self.clone().set_const(false) == other.clone().set_const(false)
}
(
Self::Function { ret, params },
Self::Function {
ret: other_ret,
params: other_params,
},
) => {
ret == other_ret
&& params.len() == other_params.len()
&& izip!(params, other_params).all(|(l, r)| l.is_compatible(r))
}
_ => false,
}
}
}
impl fmt::Display for Dtype {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
Self::Unit { is_const } => write!(f, "{}unit", if *is_const { "const " } else { "" }),
Self::Int {
width,
is_signed,
is_const,
} => write!(
f,
"{}{}{}",
if *is_const { "const " } else { "" },
if *is_signed { "i" } else { "u" },
width
),
Self::Float { width, is_const } => {
write!(f, "{}f{}", if *is_const { "const " } else { "" }, width)
}
Self::Pointer { inner, is_const } => {
write!(f, "{}* {}", inner, if *is_const { "const" } else { "" })
}
Self::Function { ret, params } => write!(
f,
"{} ({})",
ret,
params
.iter()
.map(|p| p.to_string())
.collect::<Vec<_>>()
.join(", ")
),
}
}
}
impl Default for Dtype {
fn default() -> Self {
// default dtype is `int`(i32)
Self::INT
}
}

View File

@@ -1,12 +1,13 @@
use crate::ir::*;
use crate::*;
use core::fmt;
use core::mem;
use failure::Fail;
use std::collections::HashMap;
use std::mem;
use itertools::izip;
use crate::ir::*;
use crate::*;
// TODO: the variants of Value will be added in the future
#[derive(Debug, PartialEq, Clone)]
pub enum Value {
@@ -74,23 +75,26 @@ pub enum InterpreterError {
NoMainFunction,
#[fail(display = "ir has no function definition of {} function", func_name)]
NoFunctionDefinition { func_name: String },
#[fail(
display = "{}:{}:{} / Undef value cannot be used as an operand",
func_name, bid, iid
)]
#[fail(display = "{}:{} / {}", func_name, pc, msg)]
Misc {
func_name: String,
bid: BlockId,
iid: usize,
pc: Pc,
msg: String,
},
}
#[derive(Debug, PartialEq, Clone)]
struct Pc {
#[derive(Debug, PartialEq, Clone, Copy)]
pub struct Pc {
pub bid: BlockId,
pub iid: usize,
}
impl fmt::Display for Pc {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "{}:{}", self.bid, self.iid)
}
}
impl Pc {
fn new(bid: BlockId) -> Pc {
Pc { bid, iid: 0 }
@@ -101,16 +105,20 @@ impl Pc {
}
}
#[derive(Debug, PartialEq, Clone)]
#[derive(Default, Debug, PartialEq, Clone)]
struct RegisterMap {
inner: HashMap<RegisterId, Value>,
}
impl RegisterMap {
fn new() -> Self {
Self {
inner: HashMap::new(),
}
fn read(&self, rid: RegisterId) -> &Value {
self.inner
.get(&rid)
.expect("`rid` must be assigned before it can be used")
}
fn write(&mut self, rid: RegisterId, value: Value) {
let _ = self.inner.insert(rid, value);
}
}
@@ -164,7 +172,7 @@ impl<'i> StackFrame<'i> {
fn new(bid: BlockId, func_name: String, func_def: &'i FunctionDefinition) -> Self {
StackFrame {
pc: Pc::new(bid),
registers: RegisterMap::new(),
registers: Default::default(),
func_name,
func_def,
}
@@ -229,6 +237,39 @@ mod calculator {
}
}
#[derive(Default, Debug, PartialEq)]
struct Memory {
// TODO: memory type should change to Vec<Vec<Byte>>
inner: Vec<Vec<Value>>,
}
impl Memory {
fn alloc(&mut self, dtype: &Dtype) -> Result<usize, InterpreterError> {
let memory_block = match dtype {
Dtype::Unit { .. } => vec![],
Dtype::Int { width, .. } => match width {
32 => vec![Value::Undef],
_ => todo!(),
},
Dtype::Float { .. } => todo!(),
Dtype::Pointer { .. } => vec![Value::Undef],
Dtype::Function { .. } => vec![],
};
self.inner.push(memory_block);
Ok(self.inner.len() - 1)
}
fn load(&self, bid: usize, offset: usize) -> &Value {
&self.inner[bid][offset]
}
fn store(&mut self, bid: usize, offset: usize, value: Value) {
self.inner[bid][offset] = value;
}
}
// TODO: allocation fields will be added in the future
// TODO: program fields will be added in the future
#[derive(Debug, PartialEq)]
@@ -238,8 +279,7 @@ struct State<'i> {
pub global_map: GlobalMap,
pub stack_frame: StackFrame<'i>,
pub stack: Vec<StackFrame<'i>>,
// TODO: memory type should change to Vec<Vec<Byte>>
pub memory: Vec<Vec<Value>>,
pub memory: Memory,
pub ir: &'i TranslationUnit,
}
@@ -265,35 +305,35 @@ impl<'i> State<'i> {
global_map: GlobalMap::default(),
stack_frame: StackFrame::new(func_def.bid_init, func_name, func_def),
stack: Vec::new(),
memory: Vec::new(),
memory: Default::default(),
ir,
};
state.alloc_global_variable()?;
state.alloc_global_variables()?;
// Initialize state with main function and args
state.pass_arguments(args)?;
state.alloc_local_variable()?;
state.write_args(func_def.bid_init, args)?;
state.alloc_local_variables()?;
Ok(state)
}
fn alloc_global_variable(&mut self) -> Result<(), InterpreterError> {
fn alloc_global_variables(&mut self) -> Result<(), InterpreterError> {
for (name, decl) in &self.ir.decls {
// Memory allocation
let bid = self.alloc_memory(&decl.dtype())?;
let bid = self.memory.alloc(&decl.dtype())?;
self.global_map.insert(name.clone(), bid)?;
// Initialize allocated memory space
match decl {
Declaration::Variable { dtype, initializer } => {
let value = if let Some(constant) = initializer {
self.constant_to_value(constant.clone())
self.interp_constant(constant.clone())
} else {
Value::default_from_dtype(dtype)
};
self.memory[bid][0] = value;
self.memory.store(bid, 0, value);
}
// If functin declaration, skip initialization
Declaration::Function { .. } => (),
@@ -303,46 +343,70 @@ impl<'i> State<'i> {
Ok(())
}
fn pass_arguments(&mut self, args: Vec<Value>) -> Result<(), InterpreterError> {
for (i, value) in args.iter().enumerate() {
self.register_write(RegisterId::arg(i), value.clone());
}
Ok(())
}
fn alloc_local_variable(&mut self) -> Result<(), InterpreterError> {
fn alloc_local_variables(&mut self) -> Result<(), InterpreterError> {
// add alloc register
for (id, allocation) in self.stack_frame.func_def.allocations.iter().enumerate() {
let bid = self.alloc_memory(&allocation)?;
let bid = self.memory.alloc(&allocation)?;
let ptr = Value::pointer(Some(bid), 0);
let rid = RegisterId::local("".to_string(), id);
self.register_write(rid, ptr)
self.stack_frame.registers.write(rid, ptr)
}
Ok(())
}
fn alloc_memory(&mut self, dtype: &Dtype) -> Result<usize, InterpreterError> {
// TODO: memory block will be handled as Vec<Byte>
let memory_block = match dtype {
Dtype::Unit { .. } => vec![],
Dtype::Int { width, .. } => match width {
32 => vec![Value::Undef],
_ => todo!(),
},
Dtype::Float { .. } => todo!(),
Dtype::Pointer { .. } => vec![Value::Undef],
Dtype::Function { .. } => vec![],
};
fn write_args(&mut self, bid_init: BlockId, args: Vec<Value>) -> Result<(), InterpreterError> {
for (i, value) in args.iter().enumerate() {
self.stack_frame
.registers
.write(RegisterId::arg(bid_init, i), value.clone());
}
self.memory.push(memory_block);
Ok(self.memory.len() - 1)
Ok(())
}
fn preprocess_args(
fn step(&mut self) -> Result<Option<Value>, InterpreterError> {
let block = self
.stack_frame
.func_def
.blocks
.get(&self.stack_frame.pc.bid)
.expect("block matched with `bid` must be exist");
// If it's time to execute an instruction, do so.
if let Some(instr) = block.instructions.get(self.stack_frame.pc.iid) {
self.interp_instruction(instr)?;
return Ok(None);
}
// Execute a block exit.
let return_value = some_or!(self.interp_block_exit(&block.exit)?, return Ok(None));
// If it's returning from a function, pop the stack frame.
// TODO: free memory allocated in the callee
// restore previous state
let prev_stack_frame = some_or!(self.stack.pop(), return Ok(Some(return_value)));
self.stack_frame = prev_stack_frame;
// create temporary register to write return value
let register = RegisterId::temp(self.stack_frame.pc.bid, self.stack_frame.pc.iid);
self.stack_frame.registers.write(register, return_value);
self.stack_frame.pc.increment();
Ok(None)
}
fn run(&mut self) -> Result<Value, InterpreterError> {
loop {
if let Some(value) = self.step()? {
return Ok(value);
}
}
}
fn interp_args(
&self,
signature: &FunctionSignature,
args: &[Operand],
@@ -355,135 +419,111 @@ impl<'i> State<'i> {
}
args.iter()
.map(|a| self.get_value(a.clone()))
.map(|a| self.interp_operand(a.clone()))
.collect::<Result<Vec<_>, _>>()
}
fn step(&mut self) -> Result<Option<Value>, InterpreterError> {
fn interp_jump(&mut self, arg: &JumpArg) -> Result<Option<Value>, InterpreterError> {
let block = self
.stack_frame
.func_def
.blocks
.get(&self.stack_frame.pc.bid)
.expect("block matched with `bid` must be exist");
.get(&arg.bid)
.expect("block matched with `arg.bid` must be exist");
if block.instructions.len() == self.stack_frame.pc.iid {
self.interpret_block_exit(&block.exit)
} else {
let instr = block
.instructions
.get(self.stack_frame.pc.iid)
.expect("instruction matched with `iid` must be exist");
self.interpret_instruction(instr)
assert_eq!(arg.args.len(), block.phinodes.len());
for (a, d) in izip!(&arg.args, &block.phinodes) {
assert!(a.dtype().is_compatible(&d));
}
for (i, a) in arg.args.iter().enumerate() {
let v = self.interp_operand(a.clone()).unwrap();
self.stack_frame
.registers
.inner
.insert(RegisterId::arg(arg.bid, i), v)
.unwrap();
}
self.stack_frame.pc = Pc::new(arg.bid);
Ok(None)
}
fn run(&mut self) -> Result<Value, InterpreterError> {
loop {
if let Some(value) = self.step()? {
// TODO: Before return, free memory allocated in a function
// restore previous state
let prev_stack_frame = some_or!(self.stack.pop(), {
return Ok(value);
});
self.stack_frame = prev_stack_frame;
// create temporary register to write return value
let register = RegisterId::temp(self.stack_frame.pc.bid, self.stack_frame.pc.iid);
self.register_write(register, value);
self.stack_frame.pc.increment();
}
}
}
fn interpret_block_exit(
fn interp_block_exit(
&mut self,
block_exit: &BlockExit,
) -> Result<Option<Value>, InterpreterError> {
match block_exit {
BlockExit::Jump { bid } => {
self.stack_frame.pc = Pc::new(*bid);
Ok(None)
}
BlockExit::Jump { arg } => self.interp_jump(arg),
BlockExit::ConditionalJump {
condition,
bid_then,
bid_else,
arg_then,
arg_else,
} => {
let value = self.get_value(condition.clone())?;
let value = self.interp_operand(condition.clone())?;
let value = value.get_bool().expect("`condition` must be `Value::Bool`");
self.stack_frame.pc = Pc::new(if value { *bid_then } else { *bid_else });
Ok(None)
self.interp_jump(if value { arg_then } else { arg_else })
}
BlockExit::Switch {
value,
default,
cases,
} => {
let value = self.get_value(value.clone())?;
let value = self.interp_operand(value.clone())?;
// TODO: consider different integer `width` in the future
let bid_next = cases
let arg = cases
.iter()
.find(|(c, _)| value == self.constant_to_value(c.clone()))
.map(|(_, bid)| bid)
.find(|(c, _)| value == self.interp_constant(c.clone()))
.map(|(_, arg)| arg)
.unwrap_or_else(|| default);
self.stack_frame.pc = Pc::new(*bid_next);
Ok(None)
self.interp_jump(arg)
}
BlockExit::Return { value } => Ok(Some(self.get_value(value.clone())?)),
BlockExit::Return { value } => Ok(Some(self.interp_operand(value.clone())?)),
BlockExit::Unreachable => Err(InterpreterError::Unreachable),
}
}
fn interpret_instruction(
&mut self,
instruction: &Instruction,
) -> Result<Option<Value>, InterpreterError> {
fn interp_instruction(&mut self, instruction: &Instruction) -> Result<(), InterpreterError> {
let result = match instruction {
Instruction::BinOp { op, lhs, rhs, .. } => {
let lhs = self.get_value(lhs.clone())?;
let rhs = self.get_value(rhs.clone())?;
let lhs = self.interp_operand(lhs.clone())?;
let rhs = self.interp_operand(rhs.clone())?;
calculator::calculate_binary_operator_expression(&op, lhs, rhs).map_err(|_| {
InterpreterError::Misc {
func_name: self.stack_frame.func_name.clone(),
bid: self.stack_frame.pc.bid,
iid: self.stack_frame.pc.iid,
pc: self.stack_frame.pc,
msg: "calculate_binary_operator_expression".into(),
}
})?
}
Instruction::UnaryOp { op, operand, .. } => {
let operand = self.get_value(operand.clone())?;
let operand = self.interp_operand(operand.clone())?;
calculator::calculate_unary_operator_expression(&op, operand).map_err(|_| {
InterpreterError::Misc {
func_name: self.stack_frame.func_name.clone(),
bid: self.stack_frame.pc.bid,
iid: self.stack_frame.pc.iid,
pc: self.stack_frame.pc,
msg: "calculate_unary_operator_expression".into(),
}
})?
}
Instruction::Store { ptr, value, .. } => {
let ptr = self.get_value(ptr.clone())?;
let value = self.get_value(value.clone())?;
self.memory_store(ptr, value)?;
let ptr = self.interp_operand(ptr.clone())?;
let value = self.interp_operand(value.clone())?;
let (bid, offset) = self.interp_ptr(ptr)?;
self.memory.store(bid, offset, value);
Value::Unit
}
Instruction::Load { ptr, .. } => {
let ptr = self.get_value(ptr.clone())?;
self.memory_load(ptr)?
let ptr = self.interp_operand(ptr.clone())?;
let (bid, offset) = self.interp_ptr(ptr)?;
self.memory.load(bid, offset).clone()
}
Instruction::Call { callee, args, .. } => {
let ptr = self.get_value(callee.clone())?;
let ptr = self.interp_operand(callee.clone())?;
// Get function name from pointer
let (bid, _) = ptr.get_pointer().expect("`ptr` must be `Value::Pointer`");
@@ -508,48 +548,50 @@ impl<'i> State<'i> {
func_name: callee_name.clone(),
})?;
let args = self.preprocess_args(func_signature, args)?;
let args = self.interp_args(func_signature, args)?;
let stack_frame = StackFrame::new(func_def.bid_init, callee_name, func_def);
let prev_stack_frame = mem::replace(&mut self.stack_frame, stack_frame);
self.stack.push(prev_stack_frame);
// Initialize state with function obtained by callee and args
self.pass_arguments(args)?;
self.alloc_local_variable()?;
self.write_args(func_def.bid_init, args)?;
self.alloc_local_variables()?;
return Ok(None);
return Ok(());
}
Instruction::TypeCast {
value,
target_dtype,
} => {
let value = self.get_value(value.clone())?;
let value = self.interp_operand(value.clone())?;
calculator::calculate_typecast(&value, target_dtype).map_err(|_| {
InterpreterError::Misc {
func_name: self.stack_frame.func_name.clone(),
bid: self.stack_frame.pc.bid,
iid: self.stack_frame.pc.iid,
pc: self.stack_frame.pc,
msg: "calculate_typecast".into(),
}
})?
}
};
let register = RegisterId::temp(self.stack_frame.pc.bid, self.stack_frame.pc.iid);
self.register_write(register, result);
self.stack_frame.registers.write(register, result);
self.stack_frame.pc.increment();
Ok(None)
Ok(())
}
fn get_value(&self, operand: Operand) -> Result<Value, InterpreterError> {
fn interp_operand(&self, operand: Operand) -> Result<Value, InterpreterError> {
match &operand {
Operand::Constant(value) => Ok(self.constant_to_value(value.clone())),
Operand::Register { rid, .. } => Ok(self.register_read(rid.clone())),
Operand::Constant(value) => Ok(self.interp_constant(value.clone())),
Operand::Register { rid, .. } => {
Ok(self.stack_frame.registers.read(rid.clone()).clone())
}
}
}
fn constant_to_value(&self, value: Constant) -> Value {
fn interp_constant(&self, value: Constant) -> Value {
match value {
Constant::Unit => Value::Unit,
// TODO: consider `width` and `is_signed` in the future
@@ -570,43 +612,27 @@ impl<'i> State<'i> {
}
}
fn register_write(&mut self, rid: RegisterId, value: Value) {
let _ = self.stack_frame.registers.inner.insert(rid, value);
}
fn register_read(&self, rid: RegisterId) -> Value {
self.stack_frame
.registers
.inner
.get(&rid)
.cloned()
.expect("`rid` must be assigned before it can be used")
}
fn memory_store(&mut self, pointer: Value, value: Value) -> Result<(), InterpreterError> {
fn interp_ptr(&mut self, pointer: Value) -> Result<(usize, usize), InterpreterError> {
let (bid, offset) = pointer
.get_pointer()
.expect("`pointer` must be `Value::Pointer` to access memory");
.ok_or_else(|| InterpreterError::Misc {
func_name: self.stack_frame.func_name.clone(),
pc: self.stack_frame.pc,
msg: "Accessing memory with non-pointer".into(),
})?;
let bid = bid.expect("write to memory using constant value address is not allowed");
self.memory[bid][offset] = value;
let bid = bid.ok_or_else(|| InterpreterError::Misc {
func_name: self.stack_frame.func_name.clone(),
pc: self.stack_frame.pc,
msg: "Accessing memory with constant pointer".into(),
})?;
Ok(())
}
fn memory_load(&self, pointer: Value) -> Result<Value, InterpreterError> {
let (bid, offset) = pointer
.get_pointer()
.expect("`pointer` must be `Value::Pointer` to access memory");
let bid = bid.expect("read from memory using constant value address is not allowed");
Ok(self.memory[bid][offset].clone())
Ok((bid, offset))
}
}
#[inline]
pub fn run_ir(ir: &TranslationUnit, args: Vec<Value>) -> Result<Value, InterpreterError> {
pub fn interp(ir: &TranslationUnit, args: Vec<Value>) -> Result<Value, InterpreterError> {
let mut init_state = State::new(ir, args)?;
init_state.run()
}

578
src/ir/mod.rs Normal file
View File

@@ -0,0 +1,578 @@
mod dtype;
mod interp;
mod write_ir;
use core::convert::TryFrom;
use core::fmt;
use core::ops::Deref;
use lang_c::ast;
use std::collections::HashMap;
use std::hash::{Hash, Hasher};
pub use dtype::{Dtype, DtypeError, HasDtype};
pub use interp::{interp, Value};
#[derive(Debug, PartialEq)]
pub struct TranslationUnit {
pub decls: HashMap<String, Declaration>,
}
#[derive(Debug, PartialEq, Clone)]
pub enum Declaration {
Variable {
dtype: Dtype,
initializer: Option<Constant>,
},
Function {
signature: FunctionSignature,
definition: Option<FunctionDefinition>,
},
}
impl TryFrom<Dtype> for Declaration {
type Error = DtypeError;
/// Create an appropriate declaration according to `dtype`.
///
/// # Example
///
/// If `int g = 0;` is declared, `dtype` is
/// `ir::Dtype::Int{ width:32, is_signed:true, is_const:false }`.
/// In this case, `ir::Declaration::Variable{ dtype, initializer: Some(Constant::I32(1)) }`
/// is generated.
///
/// Conversely, if `int foo();` is declared, `dtype` is
/// `ir::Dtype::Function{ret: Scalar(Int), params: []}`.
/// Thus, in this case, `ir::Declaration::Function` is generated.
fn try_from(dtype: Dtype) -> Result<Self, Self::Error> {
match &dtype {
Dtype::Unit { .. } => Err(DtypeError::Misc {
message: "A variable of type `void` cannot be declared".to_string(),
}),
Dtype::Int { .. } | Dtype::Float { .. } | Dtype::Pointer { .. } => {
Ok(Declaration::Variable {
dtype,
initializer: None,
})
}
Dtype::Function { .. } => Ok(Declaration::Function {
signature: FunctionSignature::new(dtype),
definition: None,
}),
}
}
}
impl Declaration {
pub fn get_variable(&self) -> Option<(&Dtype, &Option<Constant>)> {
if let Self::Variable { dtype, initializer } = self {
Some((dtype, initializer))
} else {
None
}
}
pub fn get_function(&self) -> Option<(&FunctionSignature, &Option<FunctionDefinition>)> {
if let Self::Function {
signature,
definition,
} = self
{
Some((signature, definition))
} else {
None
}
}
pub fn get_function_mut(
&mut self,
) -> Option<(&mut FunctionSignature, &mut Option<FunctionDefinition>)> {
if let Self::Function {
signature,
definition,
} = self
{
Some((signature, definition))
} else {
None
}
}
/// Check if type is conflicting for pre-declared one
///
/// In case of `Variable`, need to check if the two types are exactly the same.
/// On the other hand, in the case of `Function`, outermost `const` of return type and
/// parameters one is not an issue of concern.
pub fn is_compatible(&self, other: &Declaration) -> bool {
match (self, other) {
(Self::Variable { dtype, .. }, Self::Variable { dtype: other, .. }) => dtype == other,
(
Self::Function { signature, .. },
Self::Function {
signature: other, ..
},
) => signature.dtype().is_compatible(&other.dtype()),
_ => false,
}
}
}
impl HasDtype for Declaration {
fn dtype(&self) -> Dtype {
match self {
Self::Variable { dtype, .. } => dtype.clone(),
Self::Function { signature, .. } => signature.dtype(),
}
}
}
#[derive(Debug, PartialEq, Clone)]
pub struct FunctionDefinition {
/// Memory allocations for local variables. The allocation is performed at the beginning of a
/// function invocation.
pub allocations: Vec<Dtype>,
/// Basic blocks.
pub blocks: HashMap<BlockId, Block>,
/// The initial block id.
pub bid_init: BlockId,
}
#[derive(Debug, PartialEq, Clone)]
pub struct FunctionSignature {
pub ret: Dtype,
pub params: Vec<Dtype>,
}
impl FunctionSignature {
pub fn new(dtype: Dtype) -> Self {
let (ret, params) = dtype
.get_function_inner()
.expect("function signature's dtype must be function type");
Self {
ret: ret.clone(),
params: params.clone(),
}
}
}
impl HasDtype for FunctionSignature {
fn dtype(&self) -> Dtype {
Dtype::function(self.ret.clone(), self.params.clone())
}
}
#[derive(Debug, Eq, Clone)]
pub enum RegisterId {
/// Registers holding pointers to local allocations.
///
/// # Fields
///
/// - `name`: only for debugging purposes.
/// - `id`: local allocation id.
Local { name: String, id: usize },
/// Registers holding block arguments.
///
/// # Fields
///
/// - `bid`: When it is the initial block id, then it holds a function argument; otherwise, it
/// holds a phinode value.
/// - `aid`: the argument index.
Arg { bid: BlockId, aid: usize },
/// Registers holding the results of instructions.
///
/// # Fields
///
/// - `bid`: the instruction's block id.
/// - `iid`: the instruction's id in the block.
Temp { bid: BlockId, iid: usize },
}
impl RegisterId {
pub fn local(name: String, id: usize) -> Self {
Self::Local { name, id }
}
pub fn arg(bid: BlockId, aid: usize) -> Self {
Self::Arg { bid, aid }
}
pub fn temp(bid: BlockId, iid: usize) -> Self {
Self::Temp { bid, iid }
}
}
impl fmt::Display for RegisterId {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
Self::Local { name, id } => write!(f, "%(local:{}:{})", name, id),
Self::Arg { bid, aid } => write!(f, "%(arg:{}:{})", bid, aid),
Self::Temp { bid, iid } => write!(f, "%({}:{})", bid, iid),
}
}
}
impl PartialEq<RegisterId> for RegisterId {
fn eq(&self, other: &RegisterId) -> bool {
match (self, other) {
(Self::Local { id, .. }, Self::Local { id: other_id, .. }) => id == other_id,
(
Self::Arg { bid, aid },
Self::Arg {
bid: other_bid,
aid: other_aid,
},
) => bid == other_bid && aid == other_aid,
(
Self::Temp { bid, iid },
Self::Temp {
bid: other_bid,
iid: other_iid,
},
) => bid == other_bid && iid == other_iid,
_ => false,
}
}
}
impl Hash for RegisterId {
fn hash<H: Hasher>(&self, state: &mut H) {
match self {
Self::Local { id, .. } => id.hash(state),
Self::Arg { bid, aid } => {
// TODO: needs to distinguish arg/temp?
bid.hash(state);
aid.hash(state);
}
Self::Temp { bid, iid } => {
bid.hash(state);
iid.hash(state);
}
}
}
}
#[derive(Debug, PartialEq, Clone)]
pub enum Constant {
Unit,
Int {
value: u128,
width: usize,
is_signed: bool,
},
Float {
/// `value` may be `f32`, but it is possible to consider it as `f64`.
///
/// * Casting from an f32 to an f64 is perfect and lossless (f32 -> f64)
/// * Casting from an f64 to an f32 will produce the closest possible value (f64 -> f32)
/// https://doc.rust-lang.org/stable/reference/expressions/operator-expr.html#type-cast-expressions
value: f64,
width: usize,
},
GlobalVariable {
name: String,
dtype: Dtype,
},
}
impl TryFrom<&ast::Constant> for Constant {
type Error = ();
fn try_from(constant: &ast::Constant) -> Result<Self, Self::Error> {
match constant {
ast::Constant::Integer(integer) => {
let is_signed = !integer.suffix.unsigned;
let dtype = match integer.suffix.size {
ast::IntegerSize::Int => Dtype::INT,
ast::IntegerSize::Long => Dtype::LONG,
ast::IntegerSize::LongLong => Dtype::LONGLONG,
}
.set_signed(is_signed);
let value = if is_signed {
integer.number.parse::<i128>().unwrap() as u128
} else {
integer.number.parse::<u128>().unwrap()
};
Ok(Self::int(value, dtype))
}
ast::Constant::Float(float) => {
let (dtype, value) = match float.suffix.format {
ast::FloatFormat::Float => {
// Casting from an f32 to an f64 is perfect and lossless (f32 -> f64)
// https://doc.rust-lang.org/stable/reference/expressions/operator-expr.html#type-cast-expressions
(Dtype::FLOAT, float.number.parse::<f32>().unwrap() as f64)
}
ast::FloatFormat::Double => {
(Dtype::DOUBLE, float.number.parse::<f64>().unwrap())
}
ast::FloatFormat::LongDouble => {
panic!("`FloatFormat::LongDouble` is_unsupported")
}
ast::FloatFormat::TS18661Format(_) => {
panic!("`FloatFormat::TS18661Format` is_unsupported")
}
};
Ok(Self::float(value, dtype))
}
ast::Constant::Character(character) => {
let dtype = Dtype::CHAR;
let value = character.parse::<char>().unwrap() as u128;
Ok(Self::int(value, dtype))
}
}
}
}
impl TryFrom<&ast::Expression> for Constant {
type Error = ();
fn try_from(expr: &ast::Expression) -> Result<Self, Self::Error> {
if let ast::Expression::Constant(constant) = expr {
Self::try_from(&constant.node)
} else {
Err(())
}
}
}
impl TryFrom<&ast::Initializer> for Constant {
type Error = ();
fn try_from(initializer: &ast::Initializer) -> Result<Self, Self::Error> {
if let ast::Initializer::Expression(expr) = &initializer {
Self::try_from(&expr.node)
} else {
Err(())
}
}
}
impl Constant {
#[inline]
pub fn is_integer_constant(&self) -> bool {
if let Self::Int { .. } = self {
true
} else {
false
}
}
pub fn unit() -> Self {
Constant::Unit
}
#[inline]
pub fn int(value: u128, dtype: Dtype) -> Self {
let width = dtype.get_int_width().expect("`dtype` must be `Dtype::Int`");
let is_signed = dtype.is_int_signed();
Constant::Int {
value,
width,
is_signed,
}
}
#[inline]
pub fn float(value: f64, dtype: Dtype) -> Self {
let width = dtype
.get_float_width()
.expect("`dtype` must be `Dtype::Float`");
Constant::Float { value, width }
}
#[inline]
pub fn global_variable(name: String, dtype: Dtype) -> Self {
Self::GlobalVariable { name, dtype }
}
}
impl fmt::Display for Constant {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
Self::Unit => write!(f, "unit"),
Self::Int { value, .. } => write!(f, "{}", value),
Self::Float { value, .. } => write!(f, "{}", value),
Self::GlobalVariable { name, .. } => write!(f, "%{}", name),
}
}
}
impl HasDtype for Constant {
fn dtype(&self) -> Dtype {
match self {
Self::Unit => Dtype::unit(),
Self::Int {
width, is_signed, ..
} => Dtype::int(*width).set_signed(*is_signed),
Self::Float { width, .. } => Dtype::float(*width),
Self::GlobalVariable { dtype, .. } => Dtype::pointer(dtype.clone()),
}
}
}
#[derive(Debug, PartialEq, Clone)]
pub enum Operand {
Constant(Constant),
Register { rid: RegisterId, dtype: Dtype },
}
impl Operand {
pub fn constant(value: Constant) -> Self {
Self::Constant(value)
}
pub fn register(rid: RegisterId, dtype: Dtype) -> Self {
Self::Register { rid, dtype }
}
pub fn get_constant(&self) -> Option<&Constant> {
if let Self::Constant(constant) = self {
Some(constant)
} else {
None
}
}
pub fn get_register(&self) -> Option<(&RegisterId, &Dtype)> {
if let Self::Register { rid, dtype } = self {
Some((rid, dtype))
} else {
None
}
}
}
impl fmt::Display for Operand {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
Self::Constant(value) => write!(f, "{}", value),
Self::Register { rid, .. } => write!(f, "{}", rid),
}
}
}
impl HasDtype for Operand {
fn dtype(&self) -> Dtype {
match self {
Self::Constant(value) => value.dtype(),
Self::Register { dtype, .. } => dtype.clone(),
}
}
}
#[derive(Debug, PartialEq, Clone)]
pub enum Instruction {
// TODO: the variants of Instruction will be added in the future
BinOp {
op: ast::BinaryOperator,
lhs: Operand,
rhs: Operand,
dtype: Dtype,
},
UnaryOp {
op: ast::UnaryOperator,
operand: Operand,
dtype: Dtype,
},
Store {
ptr: Operand,
value: Operand,
},
Load {
ptr: Operand,
},
Call {
callee: Operand,
args: Vec<Operand>,
return_type: Dtype,
},
TypeCast {
value: Operand,
target_dtype: Dtype,
},
}
impl HasDtype for Instruction {
fn dtype(&self) -> Dtype {
match self {
Self::BinOp { dtype, .. } => dtype.clone(),
Self::UnaryOp { dtype, .. } => dtype.clone(),
Self::Store { .. } => Dtype::unit(),
Self::Load { ptr } => ptr
.dtype()
.get_pointer_inner()
.expect("Load instruction must have pointer value as operand")
.deref()
.clone()
.set_const(false),
Self::Call { return_type, .. } => return_type.clone(),
Self::TypeCast { target_dtype, .. } => target_dtype.clone(),
}
}
}
#[derive(Debug, PartialEq, Eq, Hash, Clone, Copy)]
pub struct BlockId(pub usize);
impl fmt::Display for BlockId {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "b{}", self.0)
}
}
#[derive(Debug, PartialEq, Clone)]
pub struct JumpArg {
pub bid: BlockId,
pub args: Vec<Operand>,
}
impl JumpArg {
pub fn new(bid: BlockId, args: Vec<Operand>) -> Self {
Self { bid, args }
}
}
impl fmt::Display for JumpArg {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "{}({:?})", self.bid, self.args)
}
}
// TODO
#[derive(Debug, PartialEq, Clone)]
pub enum BlockExit {
Jump {
arg: JumpArg,
},
ConditionalJump {
condition: Operand,
arg_then: JumpArg,
arg_else: JumpArg,
},
Switch {
value: Operand,
default: JumpArg,
cases: Vec<(Constant, JumpArg)>,
},
Return {
value: Operand,
},
Unreachable,
}
#[derive(Debug, PartialEq, Clone)]
pub struct Block {
pub phinodes: Vec<Dtype>,
pub instructions: Vec<Instruction>,
pub exit: BlockExit,
}

View File

@@ -188,16 +188,16 @@ impl WriteOp for ast::UnaryOperator {
impl WriteString for BlockExit {
fn write_string(&self) -> String {
match self {
BlockExit::Jump { bid } => format!("j {}", bid),
BlockExit::Jump { arg } => format!("j {}", arg),
BlockExit::ConditionalJump {
condition,
bid_then,
bid_else,
arg_then,
arg_else,
} => format!(
"br {}, {}, {}",
condition.write_string(),
bid_then,
bid_else
arg_then,
arg_else
),
BlockExit::Switch {
value,
@@ -218,7 +218,3 @@ impl WriteString for BlockExit {
}
}
}
pub fn write_ir(ir: &TranslationUnit, write: &mut dyn Write) -> Result<()> {
ir.write_line(0, write)
}

View File

@@ -1,4 +1,4 @@
use std::fmt;
use core::fmt;
use lang_c::ast::*;

View File

@@ -1,37 +1,24 @@
#![deny(warnings)]
#![allow(unreachable_code)]
mod tests;
mod utils;
pub mod asm;
pub mod ir;
mod codegen;
mod irgen;
mod optimize;
mod parse;
pub mod run_ir;
mod write_asm;
mod write_base;
mod write_c;
mod write_ir;
pub mod assert_ast_equiv;
pub mod write_c_test;
mod asm;
mod c;
mod ir;
mod asmgen;
mod irgen;
mod opt;
pub use tests::*;
pub use utils::*;
pub use write_base::write;
pub use asm::Asm;
pub use c::Parse;
pub use codegen::Codegen;
pub use asmgen::Asmgen;
pub use irgen::Irgen;
pub use optimize::{O0, O1};
pub use parse::Parse;
pub use utils::{Optimize, Repeat, Translate};
pub use write_asm::write_asm;
pub use write_c::write_c;
pub use write_ir::write_ir;
pub use assert_ast_equiv::assert_ast_equiv;
pub use write_c_test::write_c_test;
pub use opt::{Gvn, Mem2reg, Optimize, Repeat, SimplifyCfg, Translate, O0, O1};

11
src/opt/gvn.rs Normal file
View File

@@ -0,0 +1,11 @@
use crate::ir;
use crate::*;
#[derive(Default)]
pub struct Gvn {}
impl Optimize<ir::TranslationUnit> for Gvn {
fn optimize(&mut self, _code: &mut ir::TranslationUnit) -> bool {
todo!()
}
}

11
src/opt/mem2reg.rs Normal file
View File

@@ -0,0 +1,11 @@
use crate::ir;
use crate::*;
#[derive(Default)]
pub struct Mem2reg {}
impl Optimize<ir::TranslationUnit> for Mem2reg {
fn optimize(&mut self, _code: &mut ir::TranslationUnit) -> bool {
todo!()
}
}

55
src/opt/mod.rs Normal file
View File

@@ -0,0 +1,55 @@
mod gvn;
mod mem2reg;
mod simplify_cfg;
pub use gvn::Gvn;
pub use mem2reg::Mem2reg;
pub use simplify_cfg::SimplifyCfg;
use crate::ir;
pub trait Translate<S> {
type Target;
type Error;
fn translate(&mut self, source: &S) -> Result<Self::Target, Self::Error>;
}
pub trait Optimize<T> {
fn optimize(&mut self, code: &mut T) -> bool;
}
#[derive(Default)]
pub struct Repeat<O> {
inner: O,
}
#[derive(Default)]
pub struct O0 {}
pub type O1 = Repeat<(SimplifyCfg, (Mem2reg, Gvn))>;
impl Optimize<ir::TranslationUnit> for O0 {
fn optimize(&mut self, _code: &mut ir::TranslationUnit) -> bool {
false
}
}
impl<T, O1: Optimize<T>, O2: Optimize<T>> Optimize<T> for (O1, O2) {
fn optimize(&mut self, code: &mut T) -> bool {
let changed1 = self.0.optimize(code);
let changed2 = self.1.optimize(code);
changed1 || changed2
}
}
impl<T, O: Optimize<T>> Optimize<T> for Repeat<O> {
fn optimize(&mut self, code: &mut T) -> bool {
if !self.inner.optimize(code) {
return false;
}
while self.inner.optimize(code) {}
true
}
}

48
src/opt/simplify_cfg.rs Normal file
View File

@@ -0,0 +1,48 @@
use crate::ir::*;
use crate::*;
pub type SimplifyCfg = Repeat<(SimplifyCfgConstProp, (SimplifyCfgReach, SimplifyCfgMerge))>;
impl Optimize<TranslationUnit> for SimplifyCfg {
fn optimize(&mut self, code: &mut TranslationUnit) -> bool {
code.decls.iter_mut().any(|(_, decl)| self.optimize(decl))
}
}
impl Optimize<Declaration> for SimplifyCfg {
fn optimize(&mut self, code: &mut Declaration) -> bool {
let (_fsig, fdef) = some_or!(code.get_function_mut(), return false);
let fdef = some_or!(fdef, return false);
self.optimize(fdef)
}
}
/// Simplifies block exits by propagating constants.
#[derive(Default)]
pub struct SimplifyCfgConstProp {}
/// Retains only those blocks that are reachable from the init.
#[derive(Default)]
pub struct SimplifyCfgReach {}
/// Merges two blocks if a block is pointed to only by another
#[derive(Default)]
pub struct SimplifyCfgMerge {}
impl Optimize<FunctionDefinition> for SimplifyCfgConstProp {
fn optimize(&mut self, _code: &mut FunctionDefinition) -> bool {
todo!("homework 3")
}
}
impl Optimize<FunctionDefinition> for SimplifyCfgReach {
fn optimize(&mut self, _code: &mut FunctionDefinition) -> bool {
todo!("homework 3")
}
}
impl Optimize<FunctionDefinition> for SimplifyCfgMerge {
fn optimize(&mut self, _code: &mut FunctionDefinition) -> bool {
todo!("homework 3")
}
}

View File

@@ -1,31 +0,0 @@
use crate::ir;
use crate::{Optimize, Repeat};
#[derive(Default)]
pub struct O0 {}
#[derive(Default)]
pub struct Mem2reg {}
#[derive(Default)]
pub struct Gvn {}
pub type O1 = Repeat<(Mem2reg, Gvn)>;
impl Optimize<ir::TranslationUnit> for O0 {
fn optimize(&mut self, _code: &mut ir::TranslationUnit) -> bool {
false
}
}
impl Optimize<ir::TranslationUnit> for Mem2reg {
fn optimize(&mut self, _code: &mut ir::TranslationUnit) -> bool {
unimplemented!()
}
}
impl Optimize<ir::TranslationUnit> for Gvn {
fn optimize(&mut self, _code: &mut ir::TranslationUnit) -> bool {
unimplemented!()
}
}

56
src/tests.rs Normal file
View File

@@ -0,0 +1,56 @@
use lang_c::ast::*;
use std::fs::File;
use std::path::Path;
use std::process::Command;
use tempfile::tempdir;
use crate::*;
pub fn test_write_c(unit: &TranslationUnit, _path: &Path) {
let temp_dir = tempdir().expect("temp dir creation failed");
let temp_file_path = temp_dir.path().join("temp.c");
let mut temp_file = File::create(&temp_file_path).unwrap();
crate::write(unit, &mut temp_file).unwrap();
let new_unit = c::Parse::default()
.translate(&temp_file_path.as_path())
.expect("parse failed while parsing the output from implemented printer");
drop(temp_file);
c::assert_ast_equiv(&unit, &new_unit);
temp_dir.close().expect("temp dir deletion failed");
}
pub fn test_irgen(unit: &TranslationUnit, path: &Path) {
// Check if the file has .c extension
assert_eq!(path.extension(), Some(std::ffi::OsStr::new("c")));
let file_path = path.display().to_string();
let bin_path = path.with_extension("exe").as_path().display().to_string();
// Compile c file
Command::new("gcc")
.args(&["-O1", &file_path, "-o", &bin_path])
.output()
.expect("failed to compile the given program");
// Execute compiled executable
let status = Command::new(bin_path.clone())
.status()
.expect("failed to execute the compiled executable")
.code()
.expect("failed to return an exit code");
// Remove compiled executable
Command::new("rm")
.arg(bin_path)
.status()
.expect("failed to remove compiled executable");
let ir = Irgen::default()
.translate(unit)
.expect("failed to generate ir");
let args = Vec::new();
assert_eq!(ir::interp(&ir, args), Ok(ir::Value::Int(status)));
}

View File

@@ -44,38 +44,3 @@ macro_rules! some_or_exit {
}
}};
}
pub trait Translate<S> {
type Target;
type Error;
fn translate(&mut self, source: &S) -> Result<Self::Target, Self::Error>;
}
pub trait Optimize<T> {
fn optimize(&mut self, code: &mut T) -> bool;
}
#[derive(Default)]
pub struct Repeat<O> {
inner: O,
}
impl<T, O1: Optimize<T>, O2: Optimize<T>> Optimize<T> for (O1, O2) {
fn optimize(&mut self, code: &mut T) -> bool {
let changed1 = self.0.optimize(code);
let changed2 = self.1.optimize(code);
changed1 || changed2
}
}
impl<T, O: Optimize<T>> Optimize<T> for Repeat<O> {
fn optimize(&mut self, code: &mut T) -> bool {
if !self.inner.optimize(code) {
return false;
}
while self.inner.optimize(code) {}
true
}
}

View File

@@ -1,5 +0,0 @@
use crate::asm::Asm;
pub fn write_asm(_asm: &Asm, _write: &mut dyn ::std::io::Write) {
unimplemented!();
}

View File

@@ -16,3 +16,7 @@ pub trait WriteString {
pub trait WriteOp {
fn write_operation(&self) -> String;
}
pub fn write<T: WriteLine>(t: &T, write: &mut dyn Write) -> Result<()> {
t.write_line(0, write)
}

View File

@@ -1,20 +0,0 @@
use lang_c::ast::*;
use std::fs::File;
use tempfile::tempdir;
use crate::*;
pub fn write_c_test(unit: &TranslationUnit) {
let temp_dir = tempdir().expect("temp dir creation failed");
let temp_file_path = temp_dir.path().join("temp.c");
let mut temp_file = File::create(&temp_file_path).unwrap();
write_c(&unit, &mut temp_file).unwrap();
let new_unit = Parse::default()
.translate(&temp_file_path.as_path())
.expect("parse failed while parsing file from implemented printer");
drop(temp_file);
assert_ast_equiv(&unit, &new_unit);
temp_dir.close().expect("temp dir deletion failed");
}

2
tests/.gitignore vendored
View File

@@ -1,2 +1,4 @@
/platform.info
/csmith-*
/test*.c*
/reduce-criteria.sh

View File

@@ -13,30 +13,71 @@ import itertools
import argparse
import sys
import re
from pathlib import Path
REPLACE_DICT = {
"#include \"csmith.h\"": "",
"volatile ": "",
"uint16_t": "unsigned int",
"uint32_t": "unsigned int",
"int16_t": "int",
"int32_t": "int",
"uint": "unsigned int",
"static ": "",
"extern ": "",
"__restrict": "",
"long __undefined;": "",
"return 0;": "return crc32_context % 128;",
r"__attribute__ \(\(.*\)\)": "",
"_Float128": "long double",
"union": "struct",
r"enum\s*\{[^\}]*\};": "",
"const char \*const sys_errlist\[\];": "",
r"[^\n]*printf[^;]*;": "",
r"[^\n]*scanf[^;]*;": "",
}
CSMITH_DIR = "csmith-2.3.0"
def install_csmith(tests_dir, bin_file):
def install_csmith(tests_dir):
global CSMITH_DIR
csmith_root_dir = os.path.join(tests_dir, CSMITH_DIR)
if not os.path.exists(bin_file):
subprocess.Popen(["curl", "https://embed.cs.utah.edu/csmith/" + CSMITH_DIR + ".tar.gz", "-o", CSMITH_DIR + ".tar.gz"], cwd=tests_dir).communicate()
subprocess.Popen(["tar", "xzvf", CSMITH_DIR + ".tar.gz"], cwd=tests_dir).communicate()
subprocess.Popen("cmake .; make -j", shell=True, cwd=csmith_root_dir).communicate()
else:
print("Using the existing csmith...")
def generate(tests_dir, bin_file, runtime, file_name):
usr_bin_path = "/usr/bin/csmith"
usr_inc_path = "/usr/include/csmith"
if os.path.exists(usr_bin_path):
assert os.path.exists(usr_inc_path)
return usr_bin_path, usr_inc_path
bin_path = os.path.abspath(os.path.join(tests_dir, CSMITH_DIR, "src/csmith"))
inc_path = os.path.abspath(os.path.join(tests_dir, CSMITH_DIR, "runtime"))
if not os.path.exists(bin_path):
csmith_filename = "{}.tar.gz".format(CSMITH_DIR)
try:
args = ["curl", "https://embed.cs.utah.edu/csmith/{}".format(csmith_filename), "-o", csmith_filename]
proc = subprocess.Popen(args, cwd=tests_dir)
proc.communicate()
if proc.returncode != 0:
raise Exception("Failed to download Csmith (exit code: {}): `{}`".format(proc.returncode, " ".join(args)))
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
try:
args = ["tar", "xzvf", csmith_filename]
proc = subprocess.Popen(args, cwd=tests_dir)
proc.communicate()
if proc.returncode != 0:
raise Exception("Failed to extract Csmith (exit code: {}): `{}`".format(proc.returncode, " ".join(args)))
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
csmith_root_dir = os.path.join(tests_dir, CSMITH_DIR)
try:
proc = subprocess.Popen("cmake . && make -j", shell=True, cwd=csmith_root_dir)
proc.communicate()
if proc.returncode != 0:
raise Exception("Failed to build Csmith (exit code: {})".format(proc.returncode))
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
return bin_path, inc_path
def generate(tests_dir, bin_path):
"""Feeding options to built Csmith to randomly generate test case.
For generality, I disabled most of the features that are enabled by default.
@@ -46,54 +87,103 @@ def generate(tests_dir, bin_file, runtime, file_name):
"""
global CSMITH_DIR
options = [
"--no-argc", "--no-arrays", "--no-checksum",
"--no-argc", "--no-arrays",
"--no-jumps", "--no-longlong", "--no-int8",
"--no-uint8", "--no-safe-math", "--no-pointers",
"--no-structs", "--no-unions", "--no-builtins"
"--no-structs", "--no-unions", "--no-builtins",
]
args = [bin_file] + options
dst_path = os.path.join(runtime, file_name)
args = [bin_path] + options
with open(dst_path, 'w') as f_dst:
subprocess.Popen(args, cwd=tests_dir, stdout=f_dst).wait()
f_dst.flush()
try:
proc = subprocess.Popen(args, cwd=tests_dir, stdout=subprocess.PIPE)
(src, err) = proc.communicate()
return src.decode()
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
return dst_path
def preprocess(src_path, file_name):
"""Preprocessing test case to fit in kecc parser specification.
It resolves an issue that arbitrarily included header file to hinder parsing.
def polish(src, inc_path):
"""Polishing test case to fit in kecc parser specification.
"""
global REPLACE_DICT, CSMITH_DIR
with open(src_path, 'r') as src:
src = str(src.read())
for _from, _to in REPLACE_DICT.items():
src = src.replace(_from, _to)
try:
args = ["gcc",
"-I", inc_path,
"-E",
"-",
]
proc = subprocess.Popen(args, stdin=subprocess.PIPE, stdout=subprocess.PIPE)
(src_preprocessed, err) = proc.communicate(src.encode())
src_preprocessed = src_preprocessed.decode()
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
with open(os.path.join(os.path.dirname(src_path), file_name), 'w') as dst:
dst.write(str(src))
src_replaced = src_preprocessed
for _from, _to in REPLACE_DICT.items():
src_replaced = re.sub(_from, _to, src_replaced)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Fuzzing KECC.')
parser.add_argument('-n', '--num', type=int, help='The number of tests')
parser.add_argument('-p', '--print', action='store_true', help='Fuzzing C AST printer')
args = parser.parse_args()
return src_replaced
if args.print:
cargo_arg = "-p"
else:
raise "Specify fuzzing argument"
def make_reduce_criteria(tests_dir, fuzz_arg):
"""Make executable reduce_criteria.sh
"""
# Make shell script i.e. dependent to KECC path
arg_dict = {
"$PROJECT_DIR": str(Path(tests_dir).parent),
"$FUZZ_ARG": fuzz_arg,
}
with open(os.path.join(tests_dir, "reduce-criteria-template.sh"), "r") as t:
temp = t.read()
for _from, _to in arg_dict.items():
temp = temp.replace(_from, _to)
with open(os.path.join(tests_dir, "reduce-criteria.sh"), "w") as f:
f.write(temp)
tests_dir = os.path.abspath(os.path.dirname(__file__))
csmith_bin = os.path.abspath(os.path.join(tests_dir, CSMITH_DIR, "src/csmith"))
csmith_runtime = os.path.abspath(os.path.join(tests_dir, CSMITH_DIR, "runtime/"))
install_csmith(tests_dir, csmith_bin)
# chmod the script executable
try:
args = ["chmod", "u+x", "reduce-criteria.sh"]
proc = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, cwd=tests_dir)
proc.communicate()
if proc.returncode != 0:
raise Exception("`{}` failed with exit code {}.".format(" ".join(args), proc.returncode))
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
# Run cargo test infinitely
raw_test_file = "raw_test.c"
test_file = "test.c"
def creduce(tests_dir, fuzz_arg):
"""Reduce `tests/test_polished.c` to `tests/test_reduced.c`
First, we copy test_polished.c to test_reduced.c.
Then, when Creduce reduces test_reduced.c, it overwrites partially reduced program to itself.
Original file is moved to test_reduced.c.orig which is then identical to test_polished.c.
"""
make_reduce_criteria(tests_dir, fuzz_arg)
try:
args = ["cp", "test_polished.c", "test_reduced.c"]
proc = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, cwd=tests_dir)
proc.communicate()
if proc.returncode != 0:
raise Exception("`{}` failed with exit code {}.".format(" ".join(args), proc.returncode))
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
try:
args = ["creduce", "./reduce-criteria.sh", "test_reduced.c"]
proc = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, cwd=tests_dir)
(out, err) = proc.communicate()
if proc.returncode != 0:
print(out.decode())
raise Exception("Reducing test_reduced.c by `{}` failed with exit code {}.".format(" ".join(args), proc.returncode))
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
def fuzz(tests_dir, fuzz_arg, num_iter):
csmith_bin, csmith_inc = install_csmith(tests_dir)
try:
print("Building KECC..")
try:
@@ -103,27 +193,55 @@ if __name__ == "__main__":
proc.kill()
raise e
if args.num is None:
if num_iter is None:
print("Fuzzing with infinitely many test cases. Please press [ctrl+C] to break.")
iterator = itertools.count(0)
else:
print("Fuzzing with {} test cases.".format(args.num))
iterator = range(args.num)
print("Fuzzing with {} test cases.".format(num_iter))
iterator = range(num_iter)
for i in iterator:
print("Test case #{}".format(i))
preprocess(generate(tests_dir, csmith_bin, csmith_runtime, raw_test_file), test_file)
args = ["cargo", "run", "--release", "--bin", "fuzz", "--", cargo_arg, os.path.join(csmith_runtime, test_file)]
src = generate(tests_dir, csmith_bin)
with open(os.path.join(tests_dir, "test.c"), 'w') as dst:
dst.write(src)
src_polished = polish(src, csmith_inc)
with open(os.path.join(tests_dir, "test_polished.c"), 'w') as dst:
dst.write(src_polished)
try:
args = ["cargo", "run", "--release", "--bin", "fuzz", "--", fuzz_arg, os.path.join(tests_dir, "test_polished.c")]
proc = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, cwd=tests_dir)
(out, err) = proc.communicate(timeout=10)
proc.communicate(timeout=10)
if proc.returncode != 0:
raise Exception("Test `{}` failed with exit code {}.".format(" ".join(args), proc.returncode))
except subprocess.TimeoutExpired as e:
proc.kill()
raise e
except KeyboardInterrupt:
proc.terminate()
print("\n[Ctrl+C] interrupted")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Fuzzing KECC.')
parser.add_argument('-n', '--num', type=int, help='The number of tests')
parser.add_argument('-p', '--print', action='store_true', help='Fuzzing C AST printer')
parser.add_argument('-i', '--irgen', action='store_true', help='Fuzzing irgen')
parser.add_argument('-r', '--reduce', action='store_true', help="Reducing input file")
args = parser.parse_args()
if args.print and args.irgen:
raise Exception("Choose an option used for fuzzing: '--print' or '--irgen', NOT both")
if args.print:
fuzz_arg = "-p"
elif args.irgen:
fuzz_arg = "-i"
else:
raise Exception("Specify fuzzing argument")
tests_dir = os.path.abspath(os.path.dirname(__file__))
if args.reduce:
creduce(tests_dir, fuzz_arg)
else:
fuzz(tests_dir, fuzz_arg, args.num)

View File

@@ -0,0 +1,3 @@
#!/usr/bin/env bash
! cargo run --manifest-path $PROJECT_DIR/Cargo.toml --release --bin fuzz -- $FUZZ_ARG test_reduced.c

View File

@@ -1,47 +1,43 @@
use std::ffi::OsStr;
use std::path::Path;
use lang_c::ast::*;
use kecc::run_ir::*;
use kecc::*;
fn test_dir<F>(path: &Path, f: F)
where
F: Fn(&TranslationUnit),
F: Fn(&TranslationUnit, &Path),
{
let mut parse = Parse::default();
let dir = path.read_dir().expect("read_dir call failed");
for entry in dir {
let entry = ok_or!(entry, continue);
let path = entry.path();
if path.is_dir() {
if !(path.is_file() && path.extension() == Some(&OsStr::new("c"))) {
continue;
}
println!("[testing {:?}]", path);
let test_unit = parse.translate(&path.as_path()).expect(
&format!("parse failed {:?}", path.into_os_string().to_str().unwrap()).to_owned(),
&format!(
"parse failed {:?}",
path.clone().into_os_string().to_str().unwrap()
)
.to_owned(),
);
f(&test_unit);
f(&test_unit, &path);
}
}
#[test]
fn test_examples_write_c() {
test_dir(Path::new("examples/"), write_c_test);
test_dir(Path::new("examples/hw1"), write_c_test);
test_dir(Path::new("examples/"), test_write_c);
test_dir(Path::new("examples/hw1"), test_write_c);
}
#[test]
fn test_examples_irgen() {
test_dir(Path::new("examples/"), |test_unit| {
let ir = Irgen::default()
.translate(test_unit)
.expect("failed to generate ir");
// TODO: insert randomly generated command line arguments
let args = Vec::new();
assert_eq!(run_ir(&ir, args), Ok(Value::Int(1)));
});
test_dir(Path::new("examples/"), test_irgen);
}