-
Notifications
You must be signed in to change notification settings - Fork 159
Overview of the compiler
---
title: GCCRS Graph
---
stateDiagram-v2
Rust --> Tokens: Lexer
Tokens --> AST: Parser
state AST {
[*] --> attr_checking
attr_checking: Attribute Checker
cfg_stripping: Cfg stripping
toplevel_nr: Top level name resolution
early_nr: Early name resolution
validation: Validation
late: Late name resolution
feature_gating: Feature gating
attr_checking --> cfg_stripping
cfg_stripping --> toplevel_nr
toplevel_nr --> early_nr
early_nr --> expansion
expansion --> cfg_stripping
expansion --> validation
validation --> feature_gating
feature_gating --> late
late --> [*]
}
state HIR {
type_resolve: Type resolver
privacy: Privacy checker
safety: Safety checker
const: Const checker
dead: Dead code scanner
[*] --> type_resolve
type_resolve --> privacy
privacy --> safety
safety --> const
const --> dead
dead --> [*]
}
AST --> HIR: Lowering
metadata: Crate metadata
HIR --> metadata
HIR --> Generic
state Generic {
unused: Unused variables checker
readonly: Read only checker
[*] --> unused
unused --> readonly
readonly --> [*]
}
state BIR {
wrapper: Polonius wrapper
polonius: Polonius
wrapper --> polonius
}
HIR --> BIR
Generic --> Gimple
state RTL {
CFGC: Control flow graph cleanup
fwd_prop: Forward propagation
loop: Loop optimisation
if: If conversion
combination: Instruction combination
register: Register allocation
bb_reordering: Basic block reordering
[*] --> CFGC
CFGC --> fwd_prop
fwd_prop --> CSE
CSE --> GCSE
GCSE --> loop
loop --> if
if --> combination
combination --> scheduling
scheduling --> register
register --> bb_reordering
}
Gimple --> RTL
RTL --> Assembly
Lexer gcc/rust/lex
Take rust code and recognize different token patterns. Separate those patterns in a buffered queue (rust-lex
).
Tokens are defined in a macro.
Parser gcc/rust/parse
Parser implementation can be found here.
Accepted syntax may be relaxed, this is fine because it might get modified later by macros or removed using conditional compilation attributes.
For example we might accept at this stage functions without body. As long as a macro expand it to a proper function later or strip it out this is fine, we'll error out later.
Attribute checker gcc/rust/util/rust-attribute
Check the validity of attributes. Check if attributes are applied on correct items, within allowed context.
Expansion stage gcc/rust/expand
Fixed point expansion loop. If the AST has changed during one iteration we loop again until it stays the same or the maximum allowed number of iteration has been reached.
Cfg stripping gcc/rust/expand/rust-cfg-strip
Remove disabled items (eg. #[cfg(false)]
). See conditional compilation.
Top level name resolution gcc/rust/resolve/rust-ast-resolve-toplevel
Collect all definitions (Struct, Functions, Modules, Traits)
Early name resolution gcc/rust/resolve/rust-early-name-resolver-2.0
Resolve all macro invocations (mbe, proc-macros...)
Expansion gcc/rust/expand
Expand all macros invocation found. Do not error out if something is missing as it might be expanded during next iterations.
AST Validation gcc/rust/checks/errors/rust-ast-validation
Ensure the AST is valid and properly constructed. We error out on some syntax previously accepted during parsing.
Feature gating gcc/rust/checks/rust-feature-gate
Some features should only be accepted if they are enabled (nightly features)
Late name resolution gcc/rust/resolve/rust-late-name-resolver-2.0
Resolve everything else
-
gcc/rust/util/rust-hir-map
- Retrieve an item/ast node from it's node id. -
gcc/rust/lex/token.h
- Retrieve a macro to build a list of tokens. -
gcc/rust/util/lang-items.h
- Retrieve existing lang items. -
gcc/rust/check/errors/rust-feature-gate.cc
- Retrieve the list of existing nightly features. -
gcc/rust/util/rust-keyword-values.h
- Retrieve the list of existing keywords and weak keywords.