forked from M-Labs/nac3
1
0
Fork 0
nac3/nac3core/src/codegen/mod.rs

1101 lines
39 KiB
Rust

use crate::{
codegen::classes::{ListType, ProxyType, RangeType},
symbol_resolver::{StaticValue, SymbolResolver},
toplevel::{helper::PrimDef, TopLevelContext, TopLevelDef},
typecheck::{
type_inferencer::{CodeLocation, PrimitiveStore},
typedef::{CallId, FuncArg, Type, TypeEnum, Unifier},
},
};
use crossbeam::channel::{unbounded, Receiver, Sender};
use inkwell::{
attributes::{Attribute, AttributeLoc},
basic_block::BasicBlock,
builder::Builder,
context::Context,
debug_info::{
AsDIScope, DICompileUnit, DIFlagsConstants, DIScope, DISubprogram, DebugInfoBuilder,
},
module::Module,
passes::PassBuilderOptions,
targets::{CodeModel, RelocMode, Target, TargetMachine, TargetTriple},
types::{AnyType, BasicType, BasicTypeEnum},
values::{BasicValueEnum, FunctionValue, IntValue, PhiValue, PointerValue},
AddressSpace, IntPredicate, OptimizationLevel,
};
use itertools::Itertools;
use model::*;
use nac3parser::ast::{Location, Stmt, StrRef};
use object::ndarray::NDArray;
use parking_lot::{Condvar, Mutex};
use std::collections::{HashMap, HashSet};
use std::sync::{
atomic::{AtomicBool, Ordering},
Arc,
};
use std::thread;
pub mod builtin_fns;
pub mod classes;
pub mod concrete_type;
pub mod expr;
pub mod extern_fns;
mod generator;
pub mod irrt;
pub mod llvm_intrinsics;
pub mod model;
pub mod numpy;
pub mod object;
pub mod stmt;
#[cfg(test)]
mod test;
use concrete_type::{ConcreteType, ConcreteTypeEnum, ConcreteTypeStore};
pub use generator::{CodeGenerator, DefaultCodeGenerator};
#[derive(Default)]
pub struct StaticValueStore {
pub lookup: HashMap<Vec<(usize, u64)>, usize>,
pub store: Vec<HashMap<usize, Arc<dyn StaticValue + Send + Sync>>>,
}
pub type VarValue<'ctx> = (PointerValue<'ctx>, Option<Arc<dyn StaticValue + Send + Sync>>, i64);
/// Additional options for LLVM during codegen.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct CodeGenLLVMOptions {
/// The optimization level to apply on the generated LLVM IR.
pub opt_level: OptimizationLevel,
/// Options related to the target machine.
pub target: CodeGenTargetMachineOptions,
}
impl CodeGenLLVMOptions {
/// Creates a [`TargetMachine`] using the target options specified by this struct.
///
/// See [`Target::create_target_machine`].
#[must_use]
pub fn create_target_machine(&self) -> Option<TargetMachine> {
self.target.create_target_machine(self.opt_level)
}
}
/// Additional options for code generation for the target machine.
#[derive(Clone, Debug, Eq, PartialEq)]
pub struct CodeGenTargetMachineOptions {
/// The target machine triple.
pub triple: String,
/// The target machine CPU.
pub cpu: String,
/// Additional target machine features.
pub features: String,
/// Relocation mode for code generation.
pub reloc_mode: RelocMode,
/// Code model for code generation.
pub code_model: CodeModel,
}
impl CodeGenTargetMachineOptions {
/// Creates an instance of [`CodeGenTargetMachineOptions`] using the triple of the host machine.
/// Other options are set to defaults.
#[must_use]
pub fn from_host_triple() -> CodeGenTargetMachineOptions {
CodeGenTargetMachineOptions {
triple: TargetMachine::get_default_triple().as_str().to_string_lossy().into_owned(),
cpu: String::default(),
features: String::default(),
reloc_mode: RelocMode::Default,
code_model: CodeModel::Default,
}
}
/// Creates an instance of [`CodeGenTargetMachineOptions`] using the properties of the host
/// machine. Other options are set to defaults.
#[must_use]
pub fn from_host() -> CodeGenTargetMachineOptions {
CodeGenTargetMachineOptions {
cpu: TargetMachine::get_host_cpu_name().to_string(),
features: TargetMachine::get_host_cpu_features().to_string(),
..CodeGenTargetMachineOptions::from_host_triple()
}
}
/// Creates a [`TargetMachine`] using the target options specified by this struct.
///
/// See [`Target::create_target_machine`].
#[must_use]
pub fn create_target_machine(&self, level: OptimizationLevel) -> Option<TargetMachine> {
let triple = TargetTriple::create(self.triple.as_str());
let target = Target::from_triple(&triple).unwrap_or_else(|_| {
panic!("could not create target from target triple {}", self.triple)
});
target.create_target_machine(
&triple,
self.cpu.as_str(),
self.features.as_str(),
level,
self.reloc_mode,
self.code_model,
)
}
}
pub struct CodeGenContext<'ctx, 'a> {
/// The LLVM context associated with [this context][CodeGenContext].
pub ctx: &'ctx Context,
/// The [Builder] instance for creating LLVM IR statements.
pub builder: Builder<'ctx>,
/// The [`DebugInfoBuilder`], [compilation unit information][DICompileUnit], and
/// [scope information][DIScope] of this context.
pub debug_info: (DebugInfoBuilder<'ctx>, DICompileUnit<'ctx>, DIScope<'ctx>),
/// The module for which [this context][CodeGenContext] is generating into.
pub module: Module<'ctx>,
/// The [`TopLevelContext`] associated with [this context][CodeGenContext].
pub top_level: &'a TopLevelContext,
pub unifier: Unifier,
pub resolver: Arc<dyn SymbolResolver + Send + Sync>,
pub static_value_store: Arc<Mutex<StaticValueStore>>,
/// A [`HashMap`] containing the mapping between the names of variables currently in-scope and
/// its value information.
pub var_assignment: HashMap<StrRef, VarValue<'ctx>>,
pub type_cache: HashMap<Type, BasicTypeEnum<'ctx>>,
pub primitives: PrimitiveStore,
pub calls: Arc<HashMap<CodeLocation, CallId>>,
pub registry: &'a WorkerRegistry,
/// Cache for constant strings.
pub const_strings: HashMap<String, BasicValueEnum<'ctx>>,
/// [`BasicBlock`] containing all `alloca` statements for the current function.
pub init_bb: BasicBlock<'ctx>,
pub exception_val: Option<PointerValue<'ctx>>,
/// The header and exit basic blocks of a loop in this context. See
/// <https://llvm.org/docs/LoopTerminology.html> for explanation of these terminology.
pub loop_target: Option<(BasicBlock<'ctx>, BasicBlock<'ctx>)>,
/// The target [`BasicBlock`] to jump to when performing stack unwind.
pub unwind_target: Option<BasicBlock<'ctx>>,
/// The target [`BasicBlock`] to jump to before returning from the function.
///
/// If this field is [None] when generating a return from a function, `ret` with no argument can
/// be emitted.
pub return_target: Option<BasicBlock<'ctx>>,
/// The [`PointerValue`] containing the return value of the function.
pub return_buffer: Option<PointerValue<'ctx>>,
// outer catch clauses
pub outer_catch_clauses:
Option<(Vec<Option<BasicValueEnum<'ctx>>>, BasicBlock<'ctx>, PhiValue<'ctx>)>,
/// Whether `sret` is needed for the first parameter of the function.
///
/// See [`need_sret`].
pub need_sret: bool,
/// The current source location.
pub current_loc: Location,
}
impl<'ctx, 'a> CodeGenContext<'ctx, 'a> {
/// Whether the [current basic block][Builder::get_insert_block] referenced by `builder`
/// contains a [terminator statement][BasicBlock::get_terminator].
pub fn is_terminated(&self) -> bool {
self.builder.get_insert_block().and_then(BasicBlock::get_terminator).is_some()
}
}
type Fp = Box<dyn Fn(&Module) + Send + Sync>;
pub struct WithCall {
fp: Fp,
}
impl WithCall {
#[must_use]
pub fn new(fp: Fp) -> WithCall {
WithCall { fp }
}
pub fn run(&self, m: &Module) {
(self.fp)(m);
}
}
pub struct WorkerRegistry {
sender: Arc<Sender<Option<CodeGenTask>>>,
receiver: Arc<Receiver<Option<CodeGenTask>>>,
/// Whether any thread in this registry has panicked.
panicked: AtomicBool,
/// The total number of tasks queued or completed in the registry.
task_count: Mutex<usize>,
/// The number of threads available for this registry.
thread_count: usize,
wait_condvar: Condvar,
top_level_ctx: Arc<TopLevelContext>,
static_value_store: Arc<Mutex<StaticValueStore>>,
/// LLVM-related options for code generation.
pub llvm_options: CodeGenLLVMOptions,
}
impl WorkerRegistry {
/// Creates workers for this registry.
#[must_use]
pub fn create_workers<G: CodeGenerator + Send + 'static>(
generators: Vec<Box<G>>,
top_level_ctx: Arc<TopLevelContext>,
llvm_options: &CodeGenLLVMOptions,
f: &Arc<WithCall>,
) -> (Arc<WorkerRegistry>, Vec<thread::JoinHandle<()>>) {
let (sender, receiver) = unbounded();
let task_count = Mutex::new(0);
let wait_condvar = Condvar::new();
// init: 0 to be empty
let mut static_value_store = StaticValueStore::default();
static_value_store.lookup.insert(Vec::default(), 0);
static_value_store.store.push(HashMap::default());
let registry = Arc::new(WorkerRegistry {
sender: Arc::new(sender),
receiver: Arc::new(receiver),
thread_count: generators.len(),
panicked: AtomicBool::new(false),
static_value_store: Arc::new(Mutex::new(static_value_store)),
task_count,
wait_condvar,
top_level_ctx,
llvm_options: llvm_options.clone(),
});
let mut handles = Vec::new();
for mut generator in generators {
let registry = registry.clone();
let registry2 = registry.clone();
let f = f.clone();
let worker_thread_name =
format!("codegen-worker-{worker_id}", worker_id = generator.get_name());
let handle = thread::Builder::new()
.name(worker_thread_name)
.spawn(move || {
registry.worker_thread(generator.as_mut(), &f);
})
.unwrap();
let handle = thread::spawn(move || {
if let Err(e) = handle.join() {
if let Some(e) = e.downcast_ref::<&'static str>() {
eprintln!("Got an error: {e}");
} else {
eprintln!("Got an unknown error: {e:?}");
}
registry2.panicked.store(true, Ordering::SeqCst);
registry2.wait_condvar.notify_all();
}
});
handles.push(handle);
}
(registry, handles)
}
pub fn wait_tasks_complete(&self, handles: Vec<thread::JoinHandle<()>>) {
{
let mut count = self.task_count.lock();
while *count != 0 {
if self.panicked.load(Ordering::SeqCst) {
break;
}
self.wait_condvar.wait(&mut count);
}
}
for _ in 0..self.thread_count {
self.sender.send(None).unwrap();
}
{
let mut count = self.task_count.lock();
while *count != self.thread_count {
if self.panicked.load(Ordering::SeqCst) {
break;
}
self.wait_condvar.wait(&mut count);
}
}
for handle in handles {
handle.join().unwrap();
}
assert!(!self.panicked.load(Ordering::SeqCst), "tasks panicked");
}
/// Adds a task to this [`WorkerRegistry`].
pub fn add_task(&self, task: CodeGenTask) {
*self.task_count.lock() += 1;
self.sender.send(Some(task)).unwrap();
}
/// Function executed by worker thread for generating IR for each function.
fn worker_thread<G: CodeGenerator>(&self, generator: &mut G, f: &Arc<WithCall>) {
let context = Context::create();
let mut builder = context.create_builder();
let mut module = context.create_module(generator.get_name());
let target_machine = self.llvm_options.create_target_machine().unwrap();
module.set_data_layout(&target_machine.get_target_data().get_data_layout());
module.set_triple(&target_machine.get_triple());
module.add_basic_value_flag(
"Debug Info Version",
inkwell::module::FlagBehavior::Warning,
context.i32_type().const_int(3, false),
);
module.add_basic_value_flag(
"Dwarf Version",
inkwell::module::FlagBehavior::Warning,
context.i32_type().const_int(4, false),
);
let mut errors = HashSet::new();
while let Some(task) = self.receiver.recv().unwrap() {
match gen_func(&context, generator, self, builder, module, task) {
Ok(result) => {
builder = result.0;
module = result.1;
}
Err((old_builder, e)) => {
builder = old_builder;
errors.insert(e);
// create a new empty module just to continue codegen and collect errors
module = context.create_module(&format!("{}_recover", generator.get_name()));
let target_machine = self.llvm_options.create_target_machine().unwrap();
module.set_data_layout(&target_machine.get_target_data().get_data_layout());
module.set_triple(&target_machine.get_triple());
}
}
*self.task_count.lock() -= 1;
self.wait_condvar.notify_all();
}
assert!(
errors.is_empty(),
"Codegen error: {}",
errors.into_iter().sorted().join("\n----------\n")
);
let result = module.verify();
if let Err(err) = result {
println!("{}", module.print_to_string().to_str().unwrap());
panic!("{}", err.to_string())
}
let pass_options = PassBuilderOptions::create();
let target_machine = self
.llvm_options
.target
.create_target_machine(self.llvm_options.opt_level)
.unwrap_or_else(|| {
panic!(
"could not create target machine from properties {:?}",
self.llvm_options.target
)
});
let passes = format!("default<O{}>", self.llvm_options.opt_level as u32);
let result = module.run_passes(passes.as_str(), &target_machine, pass_options);
if let Err(err) = result {
panic!(
"Failed to run optimization for module `{}`: {}",
module.get_name().to_str().unwrap(),
err.to_string()
);
}
f.run(&module);
let mut lock = self.task_count.lock();
*lock += 1;
self.wait_condvar.notify_all();
}
}
pub struct CodeGenTask {
pub subst: Vec<(Type, ConcreteType)>,
pub store: ConcreteTypeStore,
pub symbol_name: String,
pub signature: ConcreteType,
pub body: Arc<Vec<Stmt<Option<Type>>>>,
pub calls: Arc<HashMap<CodeLocation, CallId>>,
pub unifier_index: usize,
pub resolver: Arc<dyn SymbolResolver + Send + Sync>,
pub id: usize,
}
/// Retrieves the [LLVM type][BasicTypeEnum] corresponding to the [Type].
///
/// This function is used to obtain the in-memory representation of `ty`, e.g. a `bool` variable
/// would be represented by an `i8`.
#[allow(clippy::too_many_arguments)]
fn get_llvm_type<'ctx, G: CodeGenerator + ?Sized>(
ctx: &'ctx Context,
module: &Module<'ctx>,
generator: &G,
unifier: &mut Unifier,
top_level: &TopLevelContext,
type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
ty: Type,
) -> BasicTypeEnum<'ctx> {
use TypeEnum::*;
// we assume the type cache should already contain primitive types,
// and they should be passed by value instead of passing as pointer.
type_cache.get(&unifier.get_representative(ty)).copied().unwrap_or_else(|| {
let ty_enum = unifier.get_ty(ty);
let result = match &*ty_enum {
TObj { obj_id, fields, .. } => {
// check to avoid treating non-class primitives as classes
if PrimDef::contains_id(*obj_id) {
return match &*unifier.get_ty_immutable(ty) {
TObj { obj_id, params, .. } if *obj_id == PrimDef::Option.id() => {
get_llvm_type(
ctx,
module,
generator,
unifier,
top_level,
type_cache,
*params.iter().next().unwrap().1,
)
.ptr_type(AddressSpace::default())
.into()
}
TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
let element_type = get_llvm_type(
ctx,
module,
generator,
unifier,
top_level,
type_cache,
*params.iter().next().unwrap().1,
);
ListType::new(generator, ctx, element_type).as_base_type().into()
}
TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
Ptr(Struct(NDArray)).get_type(generator, ctx).as_basic_type_enum()
}
_ => unreachable!(
"LLVM type for primitive {} is missing",
unifier.stringify(ty)
),
};
}
// a struct with fields in the order of declaration
let top_level_defs = top_level.definitions.read();
let definition = top_level_defs.get(obj_id.0).unwrap();
let TopLevelDef::Class { fields: fields_list, .. } = &*definition.read() else {
unreachable!()
};
let name = unifier.stringify(ty);
let ty = if let Some(t) = module.get_struct_type(&name) {
t.ptr_type(AddressSpace::default()).into()
} else {
let struct_type = ctx.opaque_struct_type(&name);
type_cache.insert(
unifier.get_representative(ty),
struct_type.ptr_type(AddressSpace::default()).into(),
);
let fields = fields_list
.iter()
.map(|f| {
get_llvm_type(
ctx,
module,
generator,
unifier,
top_level,
type_cache,
fields[&f.0].0,
)
})
.collect_vec();
struct_type.set_body(&fields, false);
struct_type.ptr_type(AddressSpace::default()).into()
};
return ty;
}
TTuple { ty, is_vararg_ctx } => {
// a struct with fields in the order present in the tuple
assert!(!is_vararg_ctx, "Tuples in vararg context must be instantiated with the correct number of arguments before calling get_llvm_type");
let fields = ty
.iter()
.map(|ty| {
get_llvm_type(ctx, module, generator, unifier, top_level, type_cache, *ty)
})
.collect_vec();
ctx.struct_type(&fields, false).into()
}
TVirtual { .. } => unimplemented!(),
_ => unreachable!("{}", ty_enum.get_type_name()),
};
type_cache.insert(unifier.get_representative(ty), result);
result
})
}
/// Retrieves the [LLVM type][`BasicTypeEnum`] corresponding to the [`Type`].
///
/// This function is used mainly to obtain the ABI representation of `ty`, e.g. a `bool` is
/// would be represented by an `i1`.
///
/// The difference between the in-memory representation (as returned by [`get_llvm_type`]) and the
/// ABI representation is that the in-memory representation must be at least byte-sized and must
/// be byte-aligned for the variable to be addressable in memory, whereas there is no such
/// restriction for ABI representations.
#[allow(clippy::too_many_arguments)]
fn get_llvm_abi_type<'ctx, G: CodeGenerator + ?Sized>(
ctx: &'ctx Context,
module: &Module<'ctx>,
generator: &G,
unifier: &mut Unifier,
top_level: &TopLevelContext,
type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
primitives: &PrimitiveStore,
ty: Type,
) -> BasicTypeEnum<'ctx> {
// If the type is used in the definition of a function, return `i1` instead of `i8` for ABI
// consistency.
if unifier.unioned(ty, primitives.bool) {
ctx.bool_type().into()
} else {
get_llvm_type(ctx, module, generator, unifier, top_level, type_cache, ty)
}
}
/// Whether `sret` is needed for a return value with type `ty`.
///
/// When returning a large data structure (e.g. structures that do not fit in 1-2 native words of
/// the target processor) by value, a synthetic parameter with a pointer type will be passed in the
/// slot of the first parameter to act as the location of which the return value is passed into.
///
/// See <https://releases.llvm.org/14.0.0/docs/LangRef.html#parameter-attributes> for more
/// information.
fn need_sret(ty: BasicTypeEnum) -> bool {
fn need_sret_impl(ty: BasicTypeEnum, maybe_large: bool) -> bool {
match ty {
BasicTypeEnum::IntType(_) | BasicTypeEnum::PointerType(_) => false,
BasicTypeEnum::FloatType(_) if maybe_large => false,
BasicTypeEnum::StructType(ty) if maybe_large && ty.count_fields() <= 2 => {
ty.get_field_types().iter().any(|ty| need_sret_impl(*ty, false))
}
_ => true,
}
}
need_sret_impl(ty, true)
}
/// Returns the [`BasicTypeEnum`] representing a `va_list` struct for variadic arguments.
fn get_llvm_valist_type<'ctx>(ctx: &'ctx Context, triple: &TargetTriple) -> BasicTypeEnum<'ctx> {
let triple = TargetMachine::normalize_triple(triple);
let triple = triple.as_str().to_str().unwrap();
let arch = triple.split('-').next().unwrap();
let llvm_pi8 = ctx.i8_type().ptr_type(AddressSpace::default());
// Referenced from parseArch() in llvm/lib/Support/Triple.cpp
match arch {
"i386" | "i486" | "i586" | "i686" | "riscv32" => {
ctx.i8_type().ptr_type(AddressSpace::default()).into()
}
"amd64" | "x86_64" | "x86_64h" => {
let llvm_i32 = ctx.i32_type();
let va_list_tag = ctx.opaque_struct_type("struct.__va_list_tag");
va_list_tag.set_body(
&[llvm_i32.into(), llvm_i32.into(), llvm_pi8.into(), llvm_pi8.into()],
false,
);
va_list_tag.into()
}
"armv7" => {
let va_list = ctx.opaque_struct_type("struct.__va_list");
va_list.set_body(&[llvm_pi8.into()], false);
va_list.into()
}
triple => {
todo!("Unsupported platform for varargs: {triple}")
}
}
}
/// Implementation for generating LLVM IR for a function.
pub fn gen_func_impl<
'ctx,
G: CodeGenerator,
F: FnOnce(&mut G, &mut CodeGenContext) -> Result<(), String>,
>(
context: &'ctx Context,
generator: &mut G,
registry: &WorkerRegistry,
builder: Builder<'ctx>,
module: Module<'ctx>,
task: CodeGenTask,
codegen_function: F,
) -> Result<(Builder<'ctx>, Module<'ctx>, FunctionValue<'ctx>), (Builder<'ctx>, String)> {
let top_level_ctx = registry.top_level_ctx.clone();
let static_value_store = registry.static_value_store.clone();
let (mut unifier, primitives) = {
let (unifier, primitives) = &top_level_ctx.unifiers.read()[task.unifier_index];
(Unifier::from_shared_unifier(unifier), *primitives)
};
unifier.put_primitive_store(&primitives);
unifier.top_level = Some(top_level_ctx.clone());
let mut cache = HashMap::new();
for (a, b) in &task.subst {
// this should be unification between variables and concrete types
// and should not cause any problem...
let b = task.store.to_unifier_type(&mut unifier, &primitives, *b, &mut cache);
unifier
.unify(*a, b)
.or_else(|err| {
if matches!(&*unifier.get_ty(*a), TypeEnum::TRigidVar { .. }) {
unifier.replace_rigid_var(*a, b);
Ok(())
} else {
Err(err)
}
})
.unwrap();
}
// rebuild primitive store with unique representatives
let primitives = PrimitiveStore {
int32: unifier.get_representative(primitives.int32),
int64: unifier.get_representative(primitives.int64),
uint32: unifier.get_representative(primitives.uint32),
uint64: unifier.get_representative(primitives.uint64),
float: unifier.get_representative(primitives.float),
bool: unifier.get_representative(primitives.bool),
none: unifier.get_representative(primitives.none),
range: unifier.get_representative(primitives.range),
str: unifier.get_representative(primitives.str),
exception: unifier.get_representative(primitives.exception),
option: unifier.get_representative(primitives.option),
..primitives
};
let mut type_cache: HashMap<_, _> = [
(primitives.int32, context.i32_type().into()),
(primitives.int64, context.i64_type().into()),
(primitives.uint32, context.i32_type().into()),
(primitives.uint64, context.i64_type().into()),
(primitives.float, context.f64_type().into()),
(primitives.bool, context.i8_type().into()),
(primitives.str, {
let name = "str";
match module.get_struct_type(name) {
None => {
let str_type = context.opaque_struct_type("str");
let fields = [
context.i8_type().ptr_type(AddressSpace::default()).into(),
generator.get_size_type(context).into(),
];
str_type.set_body(&fields, false);
str_type.into()
}
Some(t) => t.as_basic_type_enum(),
}
}),
(primitives.range, RangeType::new(context).as_base_type().into()),
(primitives.exception, {
let name = "Exception";
if let Some(t) = module.get_struct_type(name) {
t.ptr_type(AddressSpace::default()).as_basic_type_enum()
} else {
let exception = context.opaque_struct_type("Exception");
let int32 = context.i32_type().into();
let int64 = context.i64_type().into();
let str_ty = module.get_struct_type("str").unwrap().as_basic_type_enum();
let fields = [int32, str_ty, int32, int32, str_ty, str_ty, int64, int64, int64];
exception.set_body(&fields, false);
exception.ptr_type(AddressSpace::default()).as_basic_type_enum()
}
}),
]
.iter()
.copied()
.collect();
// NOTE: special handling of option cannot use this type cache since it contains type var,
// handled inside get_llvm_type instead
let ConcreteTypeEnum::TFunc { args, ret, .. } = task.store.get(task.signature) else {
unreachable!()
};
let (args, ret) = (
args.iter()
.map(|arg| FuncArg {
name: arg.name,
ty: task.store.to_unifier_type(&mut unifier, &primitives, arg.ty, &mut cache),
default_value: arg.default_value.clone(),
is_vararg: arg.is_vararg,
})
.collect_vec(),
task.store.to_unifier_type(&mut unifier, &primitives, *ret, &mut cache),
);
let ret_type = if unifier.unioned(ret, primitives.none) {
None
} else {
Some(get_llvm_abi_type(
context,
&module,
generator,
&mut unifier,
top_level_ctx.as_ref(),
&mut type_cache,
&primitives,
ret,
))
};
let has_sret = ret_type.map_or(false, |ty| need_sret(ty));
let mut params = args
.iter()
.filter(|arg| !arg.is_vararg)
.map(|arg| {
debug_assert!(!arg.is_vararg);
get_llvm_abi_type(
context,
&module,
generator,
&mut unifier,
top_level_ctx.as_ref(),
&mut type_cache,
&primitives,
arg.ty,
)
.into()
})
.collect_vec();
if has_sret {
params.insert(0, ret_type.unwrap().ptr_type(AddressSpace::default()).into());
}
debug_assert!(matches!(args.iter().filter(|arg| arg.is_vararg).count(), 0..=1));
let vararg_arg = args.iter().find(|arg| arg.is_vararg);
let fn_type = match ret_type {
Some(ret_type) if !has_sret => ret_type.fn_type(&params, vararg_arg.is_some()),
_ => context.void_type().fn_type(&params, vararg_arg.is_some()),
};
let symbol = &task.symbol_name;
let fn_val =
module.get_function(symbol).unwrap_or_else(|| module.add_function(symbol, fn_type, None));
if let Some(personality) = &top_level_ctx.personality_symbol {
let personality = module.get_function(personality).unwrap_or_else(|| {
let ty = context.i32_type().fn_type(&[], true);
module.add_function(personality, ty, None)
});
fn_val.set_personality_function(personality);
}
if has_sret {
fn_val.add_attribute(
AttributeLoc::Param(0),
context.create_type_attribute(
Attribute::get_named_enum_kind_id("sret"),
ret_type.unwrap().as_any_type_enum(),
),
);
}
let init_bb = context.append_basic_block(fn_val, "init");
builder.position_at_end(init_bb);
let body_bb = context.append_basic_block(fn_val, "body");
let mut var_assignment = HashMap::new();
let offset = u32::from(has_sret);
// Store non-vararg argument values into local variables
for (n, arg) in args.iter().enumerate().filter(|(_, arg)| !arg.is_vararg) {
let param = fn_val.get_nth_param((n as u32) + offset).unwrap();
let local_type = get_llvm_type(
context,
&module,
generator,
&mut unifier,
top_level_ctx.as_ref(),
&mut type_cache,
arg.ty,
);
let alloca =
builder.build_alloca(local_type, &format!("{}.addr", &arg.name.to_string())).unwrap();
// Remap boolean parameters into i8
let param = if local_type.is_int_type() && param.is_int_value() {
let expected_ty = local_type.into_int_type();
let param_val = param.into_int_value();
if expected_ty.get_bit_width() == 8 && param_val.get_type().get_bit_width() == 1 {
bool_to_i8(&builder, context, param_val)
} else {
param_val
}
.into()
} else {
param
};
builder.build_store(alloca, param).unwrap();
var_assignment.insert(arg.name, (alloca, None, 0));
}
// TODO: Save vararg parameters as list
let return_buffer = if has_sret {
Some(fn_val.get_nth_param(0).unwrap().into_pointer_value())
} else {
fn_type.get_return_type().map(|v| builder.build_alloca(v, "$ret").unwrap())
};
let static_values = {
let store = registry.static_value_store.lock();
store.store[task.id].clone()
};
for (k, v) in static_values {
let (_, static_val, _) = var_assignment.get_mut(&args[k].name).unwrap();
*static_val = Some(v);
}
builder.build_unconditional_branch(body_bb).unwrap();
builder.position_at_end(body_bb);
let (dibuilder, compile_unit) = module.create_debug_info_builder(
/* allow_unresolved */ true,
/* language */ inkwell::debug_info::DWARFSourceLanguage::Python,
/* filename */
&task
.body
.first()
.map_or_else(|| "<nac3_internal>".to_string(), |f| f.location.file.0.to_string()),
/* directory */ "",
/* producer */ "NAC3",
/* is_optimized */ registry.llvm_options.opt_level != OptimizationLevel::None,
/* compiler command line flags */ "",
/* runtime_ver */ 0,
/* split_name */ "",
/* kind */ inkwell::debug_info::DWARFEmissionKind::Full,
/* dwo_id */ 0,
/* split_debug_inling */ true,
/* debug_info_for_profiling */ false,
/* sysroot */ "",
/* sdk */ "",
);
let subroutine_type = dibuilder.create_subroutine_type(
compile_unit.get_file(),
Some(
dibuilder
.create_basic_type("_", 0_u64, 0x00, inkwell::debug_info::DIFlags::PUBLIC)
.unwrap()
.as_type(),
),
&[],
inkwell::debug_info::DIFlags::PUBLIC,
);
let (row, col) =
task.body.first().map_or_else(|| (0, 0), |b| (b.location.row, b.location.column));
let func_scope: DISubprogram<'_> = dibuilder.create_function(
/* scope */ compile_unit.as_debug_info_scope(),
/* func name */ symbol,
/* linkage_name */ None,
/* file */ compile_unit.get_file(),
/* line_no */ row as u32,
/* DIType */ subroutine_type,
/* is_local_to_unit */ false,
/* is_definition */ true,
/* scope_line */ row as u32,
/* flags */ inkwell::debug_info::DIFlags::PUBLIC,
/* is_optimized */ registry.llvm_options.opt_level != OptimizationLevel::None,
);
fn_val.set_subprogram(func_scope);
let mut code_gen_context = CodeGenContext {
ctx: context,
resolver: task.resolver,
top_level: top_level_ctx.as_ref(),
calls: task.calls,
loop_target: None,
return_target: None,
return_buffer,
unwind_target: None,
outer_catch_clauses: None,
const_strings: HashMap::default(),
registry,
var_assignment,
type_cache,
primitives,
init_bb,
exception_val: Option::default(),
builder,
module,
unifier,
static_value_store,
need_sret: has_sret,
current_loc: Location::default(),
debug_info: (dibuilder, compile_unit, func_scope.as_debug_info_scope()),
};
let loc = code_gen_context.debug_info.0.create_debug_location(
context,
row as u32,
col as u32,
func_scope.as_debug_info_scope(),
None,
);
code_gen_context.builder.set_current_debug_location(loc);
let result = codegen_function(generator, &mut code_gen_context);
// after static analysis, only void functions can have no return at the end.
if !code_gen_context.is_terminated() {
code_gen_context.builder.build_return(None).unwrap();
}
code_gen_context.builder.unset_current_debug_location();
code_gen_context.debug_info.0.finalize();
let CodeGenContext { builder, module, .. } = code_gen_context;
if let Err(e) = result {
return Err((builder, e));
}
Ok((builder, module, fn_val))
}
/// Generates LLVM IR for a function.
///
/// * `context` - The [LLVM Context][`Context`] used in generating the function body.
/// * `generator` - The [`CodeGenerator`] for generating various program constructs.
/// * `registry` - The [`WorkerRegistry`] responsible for monitoring this function generation task.
/// * `builder` - The [`Builder`] used for generating LLVM IR.
/// * `module` - The [`Module`] of which the generated LLVM function will be inserted into.
/// * `task` - The [`CodeGenTask`] associated with this function generation task.
///
pub fn gen_func<'ctx, G: CodeGenerator>(
context: &'ctx Context,
generator: &mut G,
registry: &WorkerRegistry,
builder: Builder<'ctx>,
module: Module<'ctx>,
task: CodeGenTask,
) -> Result<(Builder<'ctx>, Module<'ctx>, FunctionValue<'ctx>), (Builder<'ctx>, String)> {
let body = task.body.clone();
gen_func_impl(context, generator, registry, builder, module, task, |generator, ctx| {
generator.gen_block(ctx, body.iter())
})
}
/// Converts the value of a boolean-like value `bool_value` into an `i1`.
fn bool_to_i1<'ctx>(builder: &Builder<'ctx>, bool_value: IntValue<'ctx>) -> IntValue<'ctx> {
if bool_value.get_type().get_bit_width() == 1 {
bool_value
} else {
builder
.build_int_compare(
IntPredicate::NE,
bool_value,
bool_value.get_type().const_zero(),
"tobool",
)
.unwrap()
}
}
/// Converts the value of a boolean-like value `bool_value` into an `i8`.
fn bool_to_i8<'ctx>(
builder: &Builder<'ctx>,
ctx: &'ctx Context,
bool_value: IntValue<'ctx>,
) -> IntValue<'ctx> {
let value_bits = bool_value.get_type().get_bit_width();
match value_bits {
8 => bool_value,
1 => builder.build_int_z_extend(bool_value, ctx.i8_type(), "frombool").unwrap(),
_ => bool_to_i8(
builder,
ctx,
builder
.build_int_compare(
IntPredicate::NE,
bool_value,
bool_value.get_type().const_zero(),
"",
)
.unwrap(),
),
}
}
/// Generates a sequence of IR which checks whether `value` does not exceed the upper bound of the
/// range as defined by `stop` and `step`.
///
/// Note that the generated IR will **not** check whether value is part of the range or whether
/// value exceeds the lower bound of the range (as evident by the missing `start` argument).
///
/// The generated IR is equivalent to the following Rust code:
///
/// ```rust,ignore
/// let sign = step > 0;
/// let (lo, hi) = if sign { (value, stop) } else { (stop, value) };
/// let cmp = lo < hi;
/// ```
///
/// Returns an `i1` [`IntValue`] representing the result of whether the `value` is in the range.
fn gen_in_range_check<'ctx>(
ctx: &CodeGenContext<'ctx, '_>,
value: IntValue<'ctx>,
stop: IntValue<'ctx>,
step: IntValue<'ctx>,
) -> IntValue<'ctx> {
let sign = ctx
.builder
.build_int_compare(IntPredicate::SGT, step, ctx.ctx.i32_type().const_zero(), "")
.unwrap();
let lo = ctx
.builder
.build_select(sign, value, stop, "")
.map(BasicValueEnum::into_int_value)
.unwrap();
let hi = ctx
.builder
.build_select(sign, stop, value, "")
.map(BasicValueEnum::into_int_value)
.unwrap();
ctx.builder.build_int_compare(IntPredicate::SLT, lo, hi, "cmp").unwrap()
}
/// Returns the internal name for the `va_count` argument, used to indicate the number of arguments
/// passed to the variadic function.
fn get_va_count_arg_name(arg_name: StrRef) -> StrRef {
format!("__{}_va_count", &arg_name).into()
}