rust/src/librustc/middle/trans/common.rs

1115 lines
34 KiB
Rust

// Copyright 2012-2013 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
//! Code that is useful in various trans modules.
use driver::session;
use driver::session::Session;
use lib::llvm::{ValueRef, BasicBlockRef, BuilderRef};
use lib::llvm::{True, False, Bool};
use lib::llvm::{llvm};
use lib;
use middle::trans::base;
use middle::trans::build;
use middle::trans::datum;
use middle::trans::glue;
use middle::trans::write_guard;
use middle::ty::substs;
use middle::ty;
use middle::typeck;
use middle::borrowck::root_map_key;
use util::ppaux::{Repr};
use middle::trans::type_::Type;
use std::cast::transmute;
use std::cast;
use std::hashmap::{HashMap};
use std::libc::{c_uint, c_longlong, c_ulonglong};
use std::vec;
use syntax::ast::ident;
use syntax::ast_map::{path, path_elt};
use syntax::codemap::span;
use syntax::parse::token;
use syntax::{ast, ast_map};
pub use middle::trans::context::CrateContext;
pub fn gensym_name(name: &str) -> ident {
token::str_to_ident(fmt!("%s_%u", name, token::gensym(name)))
}
pub struct tydesc_info {
ty: ty::t,
tydesc: ValueRef,
size: ValueRef,
align: ValueRef,
take_glue: Option<ValueRef>,
drop_glue: Option<ValueRef>,
free_glue: Option<ValueRef>,
visit_glue: Option<ValueRef>
}
/*
* A note on nomenclature of linking: "extern", "foreign", and "upcall".
*
* An "extern" is an LLVM symbol we wind up emitting an undefined external
* reference to. This means "we don't have the thing in this compilation unit,
* please make sure you link it in at runtime". This could be a reference to
* C code found in a C library, or rust code found in a rust crate.
*
* Most "externs" are implicitly declared (automatically) as a result of a
* user declaring an extern _module_ dependency; this causes the rust driver
* to locate an extern crate, scan its compilation metadata, and emit extern
* declarations for any symbols used by the declaring crate.
*
* A "foreign" is an extern that references C (or other non-rust ABI) code.
* There is no metadata to scan for extern references so in these cases either
* a header-digester like bindgen, or manual function prototypes, have to
* serve as declarators. So these are usually given explicitly as prototype
* declarations, in rust code, with ABI attributes on them noting which ABI to
* link via.
*
* An "upcall" is a foreign call generated by the compiler (not corresponding
* to any user-written call in the code) into the runtime library, to perform
* some helper task such as bringing a task to life, allocating memory, etc.
*
*/
pub struct Stats {
n_static_tydescs: uint,
n_glues_created: uint,
n_null_glues: uint,
n_real_glues: uint,
n_fns: uint,
n_monos: uint,
n_inlines: uint,
n_closures: uint,
n_llvm_insns: uint,
llvm_insn_ctxt: ~[~str],
llvm_insns: HashMap<~str, uint>,
fn_stats: ~[(~str, uint, uint)] // (ident, time-in-ms, llvm-instructions)
}
pub struct BuilderRef_res {
B: BuilderRef,
}
impl Drop for BuilderRef_res {
fn drop(&self) {
unsafe {
llvm::LLVMDisposeBuilder(self.B);
}
}
}
pub fn BuilderRef_res(B: BuilderRef) -> BuilderRef_res {
BuilderRef_res {
B: B
}
}
pub type ExternMap = HashMap<@str, ValueRef>;
// Types used for llself.
pub struct ValSelfData {
v: ValueRef,
t: ty::t,
is_copy: bool,
}
// Here `self_ty` is the real type of the self parameter to this method. It
// will only be set in the case of default methods.
pub struct param_substs {
tys: ~[ty::t],
self_ty: Option<ty::t>,
vtables: Option<typeck::vtable_res>,
self_vtable: Option<typeck::vtable_origin>
}
impl param_substs {
pub fn validate(&self) {
for self.tys.iter().advance |t| { assert!(!ty::type_needs_infer(*t)); }
for self.self_ty.iter().advance |t| { assert!(!ty::type_needs_infer(*t)); }
}
}
fn param_substs_to_str(this: &param_substs, tcx: ty::ctxt) -> ~str {
fmt!("param_substs {tys:%s, vtables:%s}",
this.tys.repr(tcx),
this.vtables.repr(tcx))
}
impl Repr for param_substs {
fn repr(&self, tcx: ty::ctxt) -> ~str {
param_substs_to_str(self, tcx)
}
}
// Function context. Every LLVM function we create will have one of
// these.
pub struct fn_ctxt_ {
// The ValueRef returned from a call to llvm::LLVMAddFunction; the
// address of the first instruction in the sequence of
// instructions for this function that will go in the .text
// section of the executable we're generating.
llfn: ValueRef,
// The implicit environment argument that arrives in the function we're
// creating.
llenv: ValueRef,
// The place to store the return value. If the return type is immediate,
// this is an alloca in the function. Otherwise, it's the hidden first
// parameter to the function. After function construction, this should
// always be Some.
llretptr: Option<ValueRef>,
// These elements: "hoisted basic blocks" containing
// administrative activities that have to happen in only one place in
// the function, due to LLVM's quirks.
// A block for all the function's static allocas, so that LLVM
// will coalesce them into a single alloca call.
llstaticallocas: BasicBlockRef,
// A block containing code that copies incoming arguments to space
// already allocated by code in one of the llallocas blocks.
// (LLVM requires that arguments be copied to local allocas before
// allowing most any operation to be performed on them.)
llloadenv: Option<BasicBlockRef>,
llreturn: BasicBlockRef,
// The 'self' value currently in use in this function, if there
// is one.
//
// NB: This is the type of the self *variable*, not the self *type*. The
// self type is set only for default methods, while the self variable is
// set for all methods.
llself: Option<ValSelfData>,
// The a value alloca'd for calls to upcalls.rust_personality. Used when
// outputting the resume instruction.
personality: Option<ValueRef>,
// If this is a for-loop body that returns, this holds the pointers needed
// for that (flagptr, retptr)
loop_ret: Option<(ValueRef, ValueRef)>,
// True if this function has an immediate return value, false otherwise.
// If this is false, the llretptr will alias the first argument of the
// function.
has_immediate_return_value: bool,
// Maps arguments to allocas created for them in llallocas.
llargs: @mut HashMap<ast::node_id, ValueRef>,
// Maps the def_ids for local variables to the allocas created for
// them in llallocas.
lllocals: @mut HashMap<ast::node_id, ValueRef>,
// Same as above, but for closure upvars
llupvars: @mut HashMap<ast::node_id, ValueRef>,
// The node_id of the function, or -1 if it doesn't correspond to
// a user-defined function.
id: ast::node_id,
// If this function is being monomorphized, this contains the type
// substitutions used.
param_substs: Option<@param_substs>,
// The source span and nesting context where this function comes from, for
// error reporting and symbol generation.
span: Option<span>,
path: path,
// This function's enclosing crate context.
ccx: @mut CrateContext
}
impl fn_ctxt_ {
pub fn arg_pos(&self, arg: uint) -> uint {
if self.has_immediate_return_value {
arg + 1u
} else {
arg + 2u
}
}
pub fn out_arg_pos(&self) -> uint {
assert!(self.has_immediate_return_value);
0u
}
pub fn env_arg_pos(&self) -> uint {
if !self.has_immediate_return_value {
1u
} else {
0u
}
}
}
pub type fn_ctxt = @mut fn_ctxt_;
pub fn warn_not_to_commit(ccx: &mut CrateContext, msg: &str) {
if !ccx.do_not_commit_warning_issued {
ccx.do_not_commit_warning_issued = true;
ccx.sess.warn(msg.to_str() + " -- do not commit like this!");
}
}
// Heap selectors. Indicate which heap something should go on.
#[deriving(Eq)]
pub enum heap {
heap_managed,
heap_managed_unique,
heap_exchange,
heap_exchange_vector,
heap_exchange_closure
}
#[deriving(Eq)]
pub enum cleantype {
normal_exit_only,
normal_exit_and_unwind
}
pub enum cleanup {
clean(@fn(block) -> block, cleantype),
clean_temp(ValueRef, @fn(block) -> block, cleantype),
}
// Used to remember and reuse existing cleanup paths
// target: none means the path ends in an resume instruction
pub struct cleanup_path {
target: Option<BasicBlockRef>,
size: uint,
dest: BasicBlockRef
}
pub fn shrink_scope_clean(scope_info: &mut scope_info, size: uint) {
scope_info.landing_pad = None;
scope_info.cleanup_paths = scope_info.cleanup_paths.iter()
.take_while(|&cu| cu.size <= size).transform(|&x|x).collect();
}
pub fn grow_scope_clean(scope_info: &mut scope_info) {
scope_info.landing_pad = None;
}
pub fn cleanup_type(cx: ty::ctxt, ty: ty::t) -> cleantype {
if ty::type_needs_unwind_cleanup(cx, ty) {
normal_exit_and_unwind
} else {
normal_exit_only
}
}
pub fn add_clean(bcx: block, val: ValueRef, t: ty::t) {
if !ty::type_needs_drop(bcx.tcx(), t) { return; }
debug!("add_clean(%s, %s, %s)", bcx.to_str(), bcx.val_to_str(val), t.repr(bcx.tcx()));
let cleanup_type = cleanup_type(bcx.tcx(), t);
do in_scope_cx(bcx, None) |scope_info| {
scope_info.cleanups.push(clean(|a| glue::drop_ty(a, val, t), cleanup_type));
grow_scope_clean(scope_info);
}
}
pub fn add_clean_temp_immediate(cx: block, val: ValueRef, ty: ty::t) {
if !ty::type_needs_drop(cx.tcx(), ty) { return; }
debug!("add_clean_temp_immediate(%s, %s, %s)",
cx.to_str(), cx.val_to_str(val),
ty.repr(cx.tcx()));
let cleanup_type = cleanup_type(cx.tcx(), ty);
do in_scope_cx(cx, None) |scope_info| {
scope_info.cleanups.push(
clean_temp(val, |a| glue::drop_ty_immediate(a, val, ty),
cleanup_type));
grow_scope_clean(scope_info);
}
}
pub fn add_clean_temp_mem(bcx: block, val: ValueRef, t: ty::t) {
add_clean_temp_mem_in_scope_(bcx, None, val, t);
}
pub fn add_clean_temp_mem_in_scope(bcx: block, scope_id: ast::node_id, val: ValueRef, t: ty::t) {
add_clean_temp_mem_in_scope_(bcx, Some(scope_id), val, t);
}
pub fn add_clean_temp_mem_in_scope_(bcx: block, scope_id: Option<ast::node_id>,
val: ValueRef, t: ty::t) {
if !ty::type_needs_drop(bcx.tcx(), t) { return; }
debug!("add_clean_temp_mem(%s, %s, %s)",
bcx.to_str(), bcx.val_to_str(val),
t.repr(bcx.tcx()));
let cleanup_type = cleanup_type(bcx.tcx(), t);
do in_scope_cx(bcx, scope_id) |scope_info| {
scope_info.cleanups.push(clean_temp(val, |a| glue::drop_ty(a, val, t), cleanup_type));
grow_scope_clean(scope_info);
}
}
pub fn add_clean_return_to_mut(bcx: block,
scope_id: ast::node_id,
root_key: root_map_key,
frozen_val_ref: ValueRef,
bits_val_ref: ValueRef,
filename_val: ValueRef,
line_val: ValueRef) {
//! When an `@mut` has been frozen, we have to
//! call the lang-item `return_to_mut` when the
//! freeze goes out of scope. We need to pass
//! in both the value which was frozen (`frozen_val`) and
//! the value (`bits_val_ref`) which was returned when the
//! box was frozen initially. Here, both `frozen_val_ref` and
//! `bits_val_ref` are in fact pointers to stack slots.
debug!("add_clean_return_to_mut(%s, %s, %s)",
bcx.to_str(),
bcx.val_to_str(frozen_val_ref),
bcx.val_to_str(bits_val_ref));
do in_scope_cx(bcx, Some(scope_id)) |scope_info| {
scope_info.cleanups.push(
clean_temp(
frozen_val_ref,
|bcx| write_guard::return_to_mut(bcx, root_key, frozen_val_ref, bits_val_ref,
filename_val, line_val),
normal_exit_only));
grow_scope_clean(scope_info);
}
}
pub fn add_clean_free(cx: block, ptr: ValueRef, heap: heap) {
let free_fn = match heap {
heap_managed | heap_managed_unique => {
let f: @fn(block) -> block = |a| glue::trans_free(a, ptr);
f
}
heap_exchange | heap_exchange_vector | heap_exchange_closure => {
let f: @fn(block) -> block = |a| glue::trans_exchange_free(a, ptr);
f
}
};
do in_scope_cx(cx, None) |scope_info| {
scope_info.cleanups.push(clean_temp(ptr, free_fn,
normal_exit_and_unwind));
grow_scope_clean(scope_info);
}
}
// Note that this only works for temporaries. We should, at some point, move
// to a system where we can also cancel the cleanup on local variables, but
// this will be more involved. For now, we simply zero out the local, and the
// drop glue checks whether it is zero.
pub fn revoke_clean(cx: block, val: ValueRef) {
do in_scope_cx(cx, None) |scope_info| {
let cleanup_pos = scope_info.cleanups.iter().position(
|cu| match *cu {
clean_temp(v, _, _) if v == val => true,
_ => false
});
for cleanup_pos.iter().advance |i| {
scope_info.cleanups =
vec::append(scope_info.cleanups.slice(0u, *i).to_owned(),
scope_info.cleanups.slice(*i + 1u,
scope_info.cleanups.len()));
shrink_scope_clean(scope_info, *i);
}
}
}
pub fn block_cleanups(bcx: block) -> ~[cleanup] {
match bcx.scope {
None => ~[],
Some(inf) => /*bad*/copy inf.cleanups
}
}
pub struct scope_info {
parent: Option<@mut scope_info>,
loop_break: Option<block>,
loop_label: Option<ident>,
// A list of functions that must be run at when leaving this
// block, cleaning up any variables that were introduced in the
// block.
cleanups: ~[cleanup],
// Existing cleanup paths that may be reused, indexed by destination and
// cleared when the set of cleanups changes.
cleanup_paths: ~[cleanup_path],
// Unwinding landing pad. Also cleared when cleanups change.
landing_pad: Option<BasicBlockRef>,
// info about the AST node this scope originated from, if any
node_info: Option<NodeInfo>,
}
impl scope_info {
pub fn empty_cleanups(&mut self) -> bool {
self.cleanups.is_empty()
}
}
pub trait get_node_info {
fn info(&self) -> Option<NodeInfo>;
}
impl get_node_info for ast::expr {
fn info(&self) -> Option<NodeInfo> {
Some(NodeInfo {id: self.id,
callee_id: self.get_callee_id(),
span: self.span})
}
}
impl get_node_info for ast::blk {
fn info(&self) -> Option<NodeInfo> {
Some(NodeInfo {id: self.node.id,
callee_id: None,
span: self.span})
}
}
impl get_node_info for Option<@ast::expr> {
fn info(&self) -> Option<NodeInfo> {
self.chain_ref(|s| s.info())
}
}
pub struct NodeInfo {
id: ast::node_id,
callee_id: Option<ast::node_id>,
span: span
}
// Basic block context. We create a block context for each basic block
// (single-entry, single-exit sequence of instructions) we generate from Rust
// code. Each basic block we generate is attached to a function, typically
// with many basic blocks per function. All the basic blocks attached to a
// function are organized as a directed graph.
pub struct block_ {
// The BasicBlockRef returned from a call to
// llvm::LLVMAppendBasicBlock(llfn, name), which adds a basic
// block to the function pointed to by llfn. We insert
// instructions into that block by way of this block context.
// The block pointing to this one in the function's digraph.
llbb: BasicBlockRef,
terminated: bool,
unreachable: bool,
parent: Option<block>,
// The current scope within this basic block
scope: Option<@mut scope_info>,
// Is this block part of a landing pad?
is_lpad: bool,
// info about the AST node this block originated from, if any
node_info: Option<NodeInfo>,
// The function context for the function to which this block is
// attached.
fcx: fn_ctxt
}
pub fn block_(llbb: BasicBlockRef, parent: Option<block>,
is_lpad: bool, node_info: Option<NodeInfo>, fcx: fn_ctxt)
-> block_ {
block_ {
llbb: llbb,
terminated: false,
unreachable: false,
parent: parent,
scope: None,
is_lpad: is_lpad,
node_info: node_info,
fcx: fcx
}
}
pub type block = @mut block_;
pub fn mk_block(llbb: BasicBlockRef, parent: Option<block>,
is_lpad: bool, node_info: Option<NodeInfo>, fcx: fn_ctxt)
-> block {
@mut block_(llbb, parent, is_lpad, node_info, fcx)
}
pub struct Result {
bcx: block,
val: ValueRef
}
pub fn rslt(bcx: block, val: ValueRef) -> Result {
Result {bcx: bcx, val: val}
}
impl Result {
pub fn unpack(&self, bcx: &mut block) -> ValueRef {
*bcx = self.bcx;
return self.val;
}
}
pub fn val_ty(v: ValueRef) -> Type {
unsafe {
Type::from_ref(llvm::LLVMTypeOf(v))
}
}
pub fn in_scope_cx(cx: block, scope_id: Option<ast::node_id>, f: &fn(si: &mut scope_info)) {
let mut cur = cx;
let mut cur_scope = cur.scope;
loop {
cur_scope = match cur_scope {
Some(inf) => match scope_id {
Some(wanted) => match inf.node_info {
Some(NodeInfo { id: actual, _ }) if wanted == actual => {
debug!("in_scope_cx: selected cur=%s (cx=%s)",
cur.to_str(), cx.to_str());
f(inf);
return;
},
_ => inf.parent,
},
None => {
debug!("in_scope_cx: selected cur=%s (cx=%s)",
cur.to_str(), cx.to_str());
f(inf);
return;
}
},
None => {
cur = block_parent(cur);
cur.scope
}
}
}
}
pub fn block_parent(cx: block) -> block {
match cx.parent {
Some(b) => b,
None => cx.sess().bug(fmt!("block_parent called on root block %?",
cx))
}
}
// Accessors
impl block_ {
pub fn ccx(&self) -> @mut CrateContext { self.fcx.ccx }
pub fn tcx(&self) -> ty::ctxt { self.fcx.ccx.tcx }
pub fn sess(&self) -> Session { self.fcx.ccx.sess }
pub fn ident(&self, ident: ident) -> @str {
token::ident_to_str(&ident)
}
pub fn node_id_to_str(&self, id: ast::node_id) -> ~str {
ast_map::node_id_to_str(self.tcx().items, id, self.sess().intr())
}
pub fn expr_to_str(&self, e: @ast::expr) -> ~str {
e.repr(self.tcx())
}
pub fn expr_is_lval(&self, e: &ast::expr) -> bool {
ty::expr_is_lval(self.tcx(), self.ccx().maps.method_map, e)
}
pub fn expr_kind(&self, e: &ast::expr) -> ty::ExprKind {
ty::expr_kind(self.tcx(), self.ccx().maps.method_map, e)
}
pub fn def(&self, nid: ast::node_id) -> ast::def {
match self.tcx().def_map.find(&nid) {
Some(&v) => v,
None => {
self.tcx().sess.bug(fmt!(
"No def associated with node id %?", nid));
}
}
}
pub fn val_to_str(&self, val: ValueRef) -> ~str {
self.ccx().tn.val_to_str(val)
}
pub fn llty_str(&self, ty: Type) -> ~str {
self.ccx().tn.type_to_str(ty)
}
pub fn ty_to_str(&self, t: ty::t) -> ~str {
t.repr(self.tcx())
}
pub fn to_str(&self) -> ~str {
unsafe {
match self.node_info {
Some(node_info) => fmt!("[block %d]", node_info.id),
None => fmt!("[block %x]", transmute(&*self)),
}
}
}
}
// Let T be the content of a box @T. tuplify_box_ty(t) returns the
// representation of @T as a tuple (i.e., the ty::t version of what T_box()
// returns).
pub fn tuplify_box_ty(tcx: ty::ctxt, t: ty::t) -> ty::t {
let ptr = ty::mk_ptr(
tcx,
ty::mt {ty: ty::mk_i8(), mutbl: ast::m_imm}
);
return ty::mk_tup(tcx, ~[ty::mk_uint(), ty::mk_type(tcx),
ptr, ptr,
t]);
}
// LLVM constant constructors.
pub fn C_null(t: Type) -> ValueRef {
unsafe {
llvm::LLVMConstNull(t.to_ref())
}
}
pub fn C_undef(t: Type) -> ValueRef {
unsafe {
llvm::LLVMGetUndef(t.to_ref())
}
}
pub fn C_integral(t: Type, u: u64, sign_extend: bool) -> ValueRef {
unsafe {
llvm::LLVMConstInt(t.to_ref(), u, sign_extend as Bool)
}
}
pub fn C_floating(s: &str, t: Type) -> ValueRef {
unsafe {
do s.as_c_str |buf| {
llvm::LLVMConstRealOfString(t.to_ref(), buf)
}
}
}
pub fn C_nil() -> ValueRef {
return C_struct([]);
}
pub fn C_bool(val: bool) -> ValueRef {
C_integral(Type::bool(), val as u64, false)
}
pub fn C_i1(val: bool) -> ValueRef {
C_integral(Type::i1(), val as u64, false)
}
pub fn C_i32(i: i32) -> ValueRef {
return C_integral(Type::i32(), i as u64, true);
}
pub fn C_i64(i: i64) -> ValueRef {
return C_integral(Type::i64(), i as u64, true);
}
pub fn C_int(cx: &CrateContext, i: int) -> ValueRef {
return C_integral(cx.int_type, i as u64, true);
}
pub fn C_uint(cx: &CrateContext, i: uint) -> ValueRef {
return C_integral(cx.int_type, i as u64, false);
}
pub fn C_u8(i: uint) -> ValueRef {
return C_integral(Type::i8(), i as u64, false);
}
// This is a 'c-like' raw string, which differs from
// our boxed-and-length-annotated strings.
pub fn C_cstr(cx: &mut CrateContext, s: @str) -> ValueRef {
unsafe {
match cx.const_cstr_cache.find_equiv(&s) {
Some(&llval) => return llval,
None => ()
}
let sc = do s.as_c_str |buf| {
llvm::LLVMConstStringInContext(cx.llcx, buf, s.len() as c_uint, False)
};
let gsym = token::gensym("str");
let g = do fmt!("str%u", gsym).as_c_str |buf| {
llvm::LLVMAddGlobal(cx.llmod, val_ty(sc).to_ref(), buf)
};
llvm::LLVMSetInitializer(g, sc);
llvm::LLVMSetGlobalConstant(g, True);
lib::llvm::SetLinkage(g, lib::llvm::InternalLinkage);
cx.const_cstr_cache.insert(s, g);
return g;
}
}
// NB: Do not use `do_spill_noroot` to make this into a constant string, or
// you will be kicked off fast isel. See issue #4352 for an example of this.
pub fn C_estr_slice(cx: &mut CrateContext, s: @str) -> ValueRef {
unsafe {
let len = s.len();
let cs = llvm::LLVMConstPointerCast(C_cstr(cx, s), Type::i8p().to_ref());
C_struct([cs, C_uint(cx, len + 1u /* +1 for null */)])
}
}
// Returns a Plain Old LLVM String:
pub fn C_postr(s: &str) -> ValueRef {
unsafe {
do s.as_c_str |buf| {
llvm::LLVMConstStringInContext(base::task_llcx(), buf, s.len() as c_uint, False)
}
}
}
pub fn C_zero_byte_arr(size: uint) -> ValueRef {
unsafe {
let mut i = 0u;
let mut elts: ~[ValueRef] = ~[];
while i < size { elts.push(C_u8(0u)); i += 1u; }
return llvm::LLVMConstArray(Type::i8().to_ref(),
vec::raw::to_ptr(elts), elts.len() as c_uint);
}
}
pub fn C_struct(elts: &[ValueRef]) -> ValueRef {
unsafe {
do elts.as_imm_buf |ptr, len| {
llvm::LLVMConstStructInContext(base::task_llcx(), ptr, len as c_uint, False)
}
}
}
pub fn C_packed_struct(elts: &[ValueRef]) -> ValueRef {
unsafe {
do elts.as_imm_buf |ptr, len| {
llvm::LLVMConstStructInContext(base::task_llcx(), ptr, len as c_uint, True)
}
}
}
pub fn C_named_struct(T: Type, elts: &[ValueRef]) -> ValueRef {
unsafe {
do elts.as_imm_buf |ptr, len| {
llvm::LLVMConstNamedStruct(T.to_ref(), ptr, len as c_uint)
}
}
}
pub fn C_array(ty: Type, elts: &[ValueRef]) -> ValueRef {
unsafe {
return llvm::LLVMConstArray(ty.to_ref(), vec::raw::to_ptr(elts), elts.len() as c_uint);
}
}
pub fn C_bytes(bytes: &[u8]) -> ValueRef {
unsafe {
let ptr = cast::transmute(vec::raw::to_ptr(bytes));
return llvm::LLVMConstStringInContext(base::task_llcx(), ptr, bytes.len() as c_uint, True);
}
}
pub fn C_bytes_plus_null(bytes: &[u8]) -> ValueRef {
unsafe {
return llvm::LLVMConstStringInContext(base::task_llcx(),
cast::transmute(vec::raw::to_ptr(bytes)),
bytes.len() as c_uint, False);
}
}
pub fn get_param(fndecl: ValueRef, param: uint) -> ValueRef {
unsafe {
llvm::LLVMGetParam(fndecl, param as c_uint)
}
}
pub fn const_get_elt(cx: &CrateContext, v: ValueRef, us: &[c_uint])
-> ValueRef {
unsafe {
let r = do us.as_imm_buf |p, len| {
llvm::LLVMConstExtractValue(v, p, len as c_uint)
};
debug!("const_get_elt(v=%s, us=%?, r=%s)",
cx.tn.val_to_str(v), us, cx.tn.val_to_str(r));
return r;
}
}
pub fn const_to_int(v: ValueRef) -> c_longlong {
unsafe {
llvm::LLVMConstIntGetSExtValue(v)
}
}
pub fn const_to_uint(v: ValueRef) -> c_ulonglong {
unsafe {
llvm::LLVMConstIntGetZExtValue(v)
}
}
pub fn is_undef(val: ValueRef) -> bool {
unsafe {
llvm::LLVMIsUndef(val) != False
}
}
pub fn is_null(val: ValueRef) -> bool {
unsafe {
llvm::LLVMIsNull(val) != False
}
}
// Used to identify cached monomorphized functions and vtables
#[deriving(Eq,IterBytes)]
pub enum mono_param_id {
mono_precise(ty::t, Option<@~[mono_id]>),
mono_any,
mono_repr(uint /* size */,
uint /* align */,
MonoDataClass,
datum::DatumMode),
}
#[deriving(Eq,IterBytes)]
pub enum MonoDataClass {
MonoBits, // Anything not treated differently from arbitrary integer data
MonoNonNull, // Non-null pointers (used for optional-pointer optimization)
// FIXME(#3547)---scalars and floats are
// treated differently in most ABIs. But we
// should be doing something more detailed
// here.
MonoFloat
}
pub fn mono_data_classify(t: ty::t) -> MonoDataClass {
match ty::get(t).sty {
ty::ty_float(_) => MonoFloat,
ty::ty_rptr(*) | ty::ty_uniq(*) |
ty::ty_box(*) | ty::ty_opaque_box(*) |
ty::ty_estr(ty::vstore_uniq) | ty::ty_evec(_, ty::vstore_uniq) |
ty::ty_estr(ty::vstore_box) | ty::ty_evec(_, ty::vstore_box) |
ty::ty_bare_fn(*) => MonoNonNull,
// Is that everything? Would closures or slices qualify?
_ => MonoBits
}
}
#[deriving(Eq,IterBytes)]
pub struct mono_id_ {
def: ast::def_id,
params: ~[mono_param_id],
impl_did_opt: Option<ast::def_id>
}
pub type mono_id = @mono_id_;
pub fn umax(cx: block, a: ValueRef, b: ValueRef) -> ValueRef {
let cond = build::ICmp(cx, lib::llvm::IntULT, a, b);
return build::Select(cx, cond, b, a);
}
pub fn umin(cx: block, a: ValueRef, b: ValueRef) -> ValueRef {
let cond = build::ICmp(cx, lib::llvm::IntULT, a, b);
return build::Select(cx, cond, a, b);
}
pub fn align_to(cx: block, off: ValueRef, align: ValueRef) -> ValueRef {
let mask = build::Sub(cx, align, C_int(cx.ccx(), 1));
let bumped = build::Add(cx, off, mask);
return build::And(cx, bumped, build::Not(cx, mask));
}
pub fn path_str(sess: session::Session, p: &[path_elt]) -> ~str {
let mut r = ~"";
let mut first = true;
for p.iter().advance |e| {
match *e {
ast_map::path_name(s) | ast_map::path_mod(s) => {
if first {
first = false
} else {
r.push_str("::")
}
r.push_str(sess.str_of(s));
}
}
}
r
}
pub fn monomorphize_type(bcx: block, t: ty::t) -> ty::t {
match bcx.fcx.param_substs {
Some(substs) => {
ty::subst_tps(bcx.tcx(), substs.tys, substs.self_ty, t)
}
_ => {
assert!(!ty::type_has_params(t));
assert!(!ty::type_has_self(t));
t
}
}
}
pub fn node_id_type(bcx: block, id: ast::node_id) -> ty::t {
let tcx = bcx.tcx();
let t = ty::node_id_to_type(tcx, id);
monomorphize_type(bcx, t)
}
pub fn expr_ty(bcx: block, ex: &ast::expr) -> ty::t {
node_id_type(bcx, ex.id)
}
pub fn expr_ty_adjusted(bcx: block, ex: &ast::expr) -> ty::t {
let tcx = bcx.tcx();
let t = ty::expr_ty_adjusted(tcx, ex);
monomorphize_type(bcx, t)
}
pub fn node_id_type_params(bcx: block, id: ast::node_id) -> ~[ty::t] {
let tcx = bcx.tcx();
let params = ty::node_id_to_type_params(tcx, id);
if !params.iter().all(|t| !ty::type_needs_infer(*t)) {
bcx.sess().bug(
fmt!("Type parameters for node %d include inference types: %s",
id, params.map(|t| bcx.ty_to_str(*t)).connect(",")));
}
match bcx.fcx.param_substs {
Some(substs) => {
do params.iter().transform |t| {
ty::subst_tps(tcx, substs.tys, substs.self_ty, *t)
}.collect()
}
_ => params
}
}
pub fn node_vtables(bcx: block, id: ast::node_id)
-> Option<typeck::vtable_res> {
let raw_vtables = bcx.ccx().maps.vtable_map.find(&id);
raw_vtables.map(
|&vts| resolve_vtables_in_fn_ctxt(bcx.fcx, *vts))
}
pub fn resolve_vtables_in_fn_ctxt(fcx: fn_ctxt, vts: typeck::vtable_res)
-> typeck::vtable_res {
resolve_vtables_under_param_substs(fcx.ccx.tcx,
fcx.param_substs,
vts)
}
pub fn resolve_vtables_under_param_substs(tcx: ty::ctxt,
param_substs: Option<@param_substs>,
vts: typeck::vtable_res)
-> typeck::vtable_res {
@vts.iter().transform(|ds|
@ds.iter().transform(
|d| resolve_vtable_under_param_substs(tcx, param_substs, copy *d))
.collect::<~[typeck::vtable_origin]>())
.collect::<~[typeck::vtable_param_res]>()
}
// Apply the typaram substitutions in the fn_ctxt to a vtable. This should
// eliminate any vtable_params.
pub fn resolve_vtable_in_fn_ctxt(fcx: fn_ctxt, vt: typeck::vtable_origin)
-> typeck::vtable_origin {
resolve_vtable_under_param_substs(fcx.ccx.tcx,
fcx.param_substs,
vt)
}
pub fn resolve_vtable_under_param_substs(tcx: ty::ctxt,
param_substs: Option<@param_substs>,
vt: typeck::vtable_origin)
-> typeck::vtable_origin {
match vt {
typeck::vtable_static(trait_id, tys, sub) => {
let tys = match param_substs {
Some(substs) => {
do tys.iter().transform |t| {
ty::subst_tps(tcx, substs.tys, substs.self_ty, *t)
}.collect()
}
_ => tys
};
typeck::vtable_static(
trait_id, tys,
resolve_vtables_under_param_substs(tcx, param_substs, sub))
}
typeck::vtable_param(n_param, n_bound) => {
match param_substs {
Some(substs) => {
find_vtable(tcx, substs, n_param, n_bound)
}
_ => {
tcx.sess.bug(fmt!(
"resolve_vtable_in_fn_ctxt: asked to lookup but \
no vtables in the fn_ctxt!"))
}
}
}
typeck::vtable_self(_trait_id) => {
match param_substs {
Some(@param_substs
{self_vtable: Some(ref self_vtable), _}) => {
copy *self_vtable
}
_ => {
tcx.sess.bug(fmt!(
"resolve_vtable_in_fn_ctxt: asked to lookup but \
no self_vtable in the fn_ctxt!"))
}
}
}
}
}
pub fn find_vtable(tcx: ty::ctxt, ps: &param_substs,
n_param: uint, n_bound: uint)
-> typeck::vtable_origin {
debug!("find_vtable(n_param=%u, n_bound=%u, ps=%s)",
n_param, n_bound, ps.repr(tcx));
/*bad*/ copy ps.vtables.get()[n_param][n_bound]
}
pub fn dummy_substs(tps: ~[ty::t]) -> ty::substs {
substs {
self_r: Some(ty::re_bound(ty::br_self)),
self_ty: None,
tps: tps
}
}
pub fn filename_and_line_num_from_span(bcx: block,
span: span) -> (ValueRef, ValueRef) {
let loc = bcx.sess().parse_sess.cm.lookup_char_pos(span.lo);
let filename_cstr = C_cstr(bcx.ccx(), loc.file.name);
let filename = build::PointerCast(bcx, filename_cstr, Type::i8p());
let line = C_int(bcx.ccx(), loc.line as int);
(filename, line)
}
// Casts a Rust bool value to an i1.
pub fn bool_to_i1(bcx: block, llval: ValueRef) -> ValueRef {
build::ICmp(bcx, lib::llvm::IntNE, llval, C_bool(false))
}