mirror of
https://github.com/carbon-language/carbon-lang.git
synced 2026-09-24 22:02:23 +01:00
These are based on looking at our compilation benchmark and looking at function bodies that seem surprising to not get inlined. Note that this will have a bit more impact on x86 where function call overhead (especially due to pushing and popping registers) is a bit higher than Arm. For a recent AMD server, this makes parsing around 15% faster, and full "check" phase 5% faster. Benchmark results: ``` name old cpu/op new cpu/op delta BM_CompileAPIFileDenseDecls<Phase::Lex>/256 40.2µs ± 2% 37.8µs ± 1% -5.89% (p=0.000 n=19+17) BM_CompileAPIFileDenseDecls<Phase::Lex>/1024 190µs ± 2% 181µs ± 2% -4.93% (p=0.000 n=19+18) BM_CompileAPIFileDenseDecls<Phase::Lex>/4096 779µs ± 1% 745µs ± 2% -4.29% (p=0.000 n=19+19) BM_CompileAPIFileDenseDecls<Phase::Lex>/16384 3.44ms ± 1% 3.32ms ± 3% -3.32% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Lex>/65536 14.6ms ± 2% 14.3ms ± 3% -2.46% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Lex>/262144 66.7ms ± 2% 65.0ms ± 4% -2.52% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/256 85.7µs ± 2% 71.3µs ± 2% -16.77% (p=0.000 n=20+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/1024 421µs ± 2% 352µs ± 2% -16.38% (p=0.000 n=20+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/4096 1.71ms ± 2% 1.44ms ± 2% -15.89% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/16384 7.19ms ± 2% 6.10ms ± 2% -15.24% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/65536 29.8ms ± 2% 25.3ms ± 2% -14.91% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/262144 127ms ± 2% 109ms ± 2% -14.28% (p=0.000 n=20+20) BM_CompileAPIFileDenseDecls<Phase::Check>/256 785µs ± 1% 752µs ± 1% -4.13% (p=0.000 n=20+18) BM_CompileAPIFileDenseDecls<Phase::Check>/1024 1.71ms ± 1% 1.62ms ± 1% -5.17% (p=0.000 n=20+18) BM_CompileAPIFileDenseDecls<Phase::Check>/4096 5.28ms ± 1% 4.97ms ± 1% -6.04% (p=0.000 n=20+19) BM_CompileAPIFileDenseDecls<Phase::Check>/16384 20.2ms ± 1% 19.0ms ± 2% -5.98% (p=0.000 n=20+20) BM_CompileAPIFileDenseDecls<Phase::Check>/65536 83.8ms ± 1% 78.9ms ± 2% -5.84% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Check>/262144 354ms ± 1% 335ms ± 1% -5.41% (p=0.000 n=19+20) ```
410 lines
16 KiB
C++
410 lines
16 KiB
C++
// Part of the Carbon Language project, under the Apache License v2.0 with LLVM
|
|
// Exceptions. See /LICENSE for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
|
|
#ifndef CARBON_TOOLCHAIN_PARSE_CONTEXT_H_
|
|
#define CARBON_TOOLCHAIN_PARSE_CONTEXT_H_
|
|
|
|
#include <optional>
|
|
|
|
#include "common/check.h"
|
|
#include "common/vlog.h"
|
|
#include "toolchain/lex/token_kind.h"
|
|
#include "toolchain/lex/tokenized_buffer.h"
|
|
#include "toolchain/parse/node_kind.h"
|
|
#include "toolchain/parse/precedence.h"
|
|
#include "toolchain/parse/state.h"
|
|
#include "toolchain/parse/tree.h"
|
|
|
|
namespace Carbon::Parse {
|
|
|
|
// An amount by which to look ahead of the current token. Lookahead should be
|
|
// used sparingly, and unbounded lookahead should be avoided.
|
|
//
|
|
// TODO: Decide whether we want to avoid lookahead altogether.
|
|
//
|
|
// NOLINTNEXTLINE(performance-enum-size): Deliberately matches index size.
|
|
enum class Lookahead : int32_t {
|
|
CurrentToken = 0,
|
|
NextToken = 1,
|
|
};
|
|
|
|
// Context and shared functionality for parser handlers. See state.def for state
|
|
// documentation.
|
|
class Context {
|
|
public:
|
|
// Possible operator fixities for errors.
|
|
enum class OperatorFixity : int8_t { Prefix, Infix, Postfix };
|
|
|
|
// Possible return values for FindListToken.
|
|
enum class ListTokenKind : int8_t { Comma, Close, CommaClose };
|
|
|
|
// Used for restricting ordering of `package` and `import` declarations.
|
|
enum class PackagingState : int8_t {
|
|
FileStart,
|
|
InImports,
|
|
AfterNonPackagingDecl,
|
|
// A warning about `import` placement has been issued so we don't keep
|
|
// issuing more (when `import` is repeated) until more non-`import`
|
|
// declarations come up.
|
|
InImportsAfterNonPackagingDecl,
|
|
};
|
|
|
|
// Used to track state on state_stack_.
|
|
struct StateStackEntry : public Printable<StateStackEntry> {
|
|
// Prints state information for verbose output.
|
|
auto Print(llvm::raw_ostream& output) const -> void {
|
|
output << state << " @" << token << " subtree_start=" << subtree_start
|
|
<< " has_error=" << has_error;
|
|
}
|
|
|
|
// The state.
|
|
State state;
|
|
// Set to true to indicate that an error was found, and that contextual
|
|
// error recovery may be needed.
|
|
bool has_error = false;
|
|
|
|
// Precedence information used by expression states in order to determine
|
|
// operator precedence. The ambient_precedence deals with how the expression
|
|
// should interact with outside context, while the lhs_precedence is
|
|
// specific to the lhs of an operator expression.
|
|
PrecedenceGroup ambient_precedence = PrecedenceGroup::ForTopLevelExpr();
|
|
PrecedenceGroup lhs_precedence = PrecedenceGroup::ForTopLevelExpr();
|
|
|
|
// A token providing context based on the subtree. This will typically be
|
|
// the first token in the subtree, but may sometimes be a token within. It
|
|
// will typically be used for the subtree's root node.
|
|
Lex::TokenIndex token;
|
|
// The offset within the Tree of the subtree start.
|
|
int32_t subtree_start;
|
|
};
|
|
|
|
// We expect StateStackEntry to fit into 12 bytes:
|
|
// state = 1 byte
|
|
// has_error = 1 byte
|
|
// ambient_precedence = 1 byte
|
|
// lhs_precedence = 1 byte
|
|
// token = 4 bytes
|
|
// subtree_start = 4 bytes
|
|
// If it becomes bigger, it'd be worth examining better packing; it should be
|
|
// feasible to pack the 1-byte entries more tightly.
|
|
static_assert(sizeof(StateStackEntry) == 12,
|
|
"StateStackEntry has unexpected size!");
|
|
|
|
explicit Context(Tree& tree, Lex::TokenizedBuffer& tokens,
|
|
Lex::TokenDiagnosticEmitter& emitter,
|
|
llvm::raw_ostream* vlog_stream);
|
|
|
|
// Adds a node to the parse tree that has no children (a leaf).
|
|
auto AddLeafNode(NodeKind kind, Lex::TokenIndex token, bool has_error = false)
|
|
-> void {
|
|
tree_->node_impls_.push_back(Tree::NodeImpl(kind, has_error, token));
|
|
}
|
|
|
|
// Adds a node to the parse tree that has children.
|
|
auto AddNode(NodeKind kind, Lex::TokenIndex token, bool has_error) -> void {
|
|
tree_->node_impls_.push_back(Tree::NodeImpl(kind, has_error, token));
|
|
}
|
|
|
|
// Replaces the placeholder node at the indicated position with a leaf node.
|
|
//
|
|
// To reserve a position in the parse tree, you may add a placeholder parse
|
|
// node using code like:
|
|
// ```
|
|
// context.PushState(State::WillFillInPlaceholder);
|
|
// context.AddLeafNode(NodeKind::Placeholder, *context.position());
|
|
// ```
|
|
// It may be replaced with the intended leaf parse node with code like:
|
|
// ```
|
|
// auto HandleWillFillInPlaceholder(Context& context) -> void {
|
|
// auto state = context.PopState();
|
|
// context.ReplacePlaceholderNode(state.subtree_start, /* replacement */);
|
|
// }
|
|
// ```
|
|
auto ReplacePlaceholderNode(int32_t position, NodeKind kind,
|
|
Lex::TokenIndex token, bool has_error = false)
|
|
-> void;
|
|
|
|
// Returns the current position and moves past it.
|
|
auto Consume() -> Lex::TokenIndex { return *(position_++); }
|
|
|
|
// Consumes the current token. Does not return it.
|
|
auto ConsumeAndDiscard() -> void { ++position_; }
|
|
|
|
// Parses an open paren token, possibly diagnosing if necessary. Creates a
|
|
// leaf parse node of the specified start kind. The default_token is used when
|
|
// there's no open paren. Returns the open paren token if it was found.
|
|
auto ConsumeAndAddOpenParen(Lex::TokenIndex default_token,
|
|
NodeKind start_kind)
|
|
-> std::optional<Lex::TokenIndex>;
|
|
|
|
// Parses a closing symbol corresponding to the opening symbol
|
|
// `expected_open`, possibly skipping forward and diagnosing if necessary.
|
|
// Creates a parse node of the specified close kind. If `expected_open` is not
|
|
// an opening symbol, the parse node will be associated with `state.token`,
|
|
// no input will be consumed, and no diagnostic will be emitted.
|
|
auto ConsumeAndAddCloseSymbol(Lex::TokenIndex expected_open,
|
|
StateStackEntry state, NodeKind close_kind)
|
|
-> void;
|
|
|
|
// Composes `ConsumeIf` and `AddLeafNode`, returning false when ConsumeIf
|
|
// fails.
|
|
auto ConsumeAndAddLeafNodeIf(Lex::TokenKind token_kind, NodeKind node_kind)
|
|
-> bool;
|
|
|
|
// Returns the current position and moves past it. Requires the token is the
|
|
// expected kind.
|
|
auto ConsumeChecked(Lex::TokenKind kind) -> Lex::TokenIndex;
|
|
|
|
// If the current position's token matches this `Kind`, returns it and
|
|
// advances to the next position. Otherwise returns an empty optional.
|
|
auto ConsumeIf(Lex::TokenKind kind) -> std::optional<Lex::TokenIndex> {
|
|
if (!PositionIs(kind)) {
|
|
return std::nullopt;
|
|
}
|
|
return Consume();
|
|
}
|
|
|
|
// Find the next token of any of the given kinds at the current bracketing
|
|
// level.
|
|
auto FindNextOf(std::initializer_list<Lex::TokenKind> desired_kinds)
|
|
-> std::optional<Lex::TokenIndex>;
|
|
|
|
// If the token is an opening symbol for a matched group, skips to the matched
|
|
// closing symbol and returns true. Otherwise, returns false.
|
|
auto SkipMatchingGroup() -> bool;
|
|
|
|
// Skips forward to move past the likely end of a declaration or statement.
|
|
//
|
|
// Looks forward, skipping over any matched symbol groups, to find the next
|
|
// position that is likely past the end of a declaration or statement. This
|
|
// is a heuristic and should only be called when skipping past parse errors.
|
|
//
|
|
// The strategy for recognizing when we have likely passed the end of a
|
|
// declaration or statement:
|
|
// - If we get to a close curly brace, we likely ended the entire context.
|
|
// - If we get to a semicolon, that should have ended the declaration or
|
|
// statement.
|
|
// - If we get to a new line from the `SkipRoot` token, but with the same or
|
|
// less indentation, there is likely a missing semicolon. Continued
|
|
// declarations or statements across multiple lines should be indented.
|
|
//
|
|
// Returns the last token consumed.
|
|
auto SkipPastLikelyEnd(Lex::TokenIndex skip_root) -> Lex::TokenIndex;
|
|
|
|
// Skip forward to the given token. Verifies that it is actually forward.
|
|
auto SkipTo(Lex::TokenIndex t) -> void;
|
|
|
|
// Returns true if the current token satisfies the lexical validity rules
|
|
// for an infix operator.
|
|
auto IsLexicallyValidInfixOperator() -> bool;
|
|
|
|
// Determines whether the current trailing operator should be treated as
|
|
// infix.
|
|
auto IsTrailingOperatorInfix() -> bool;
|
|
|
|
// Diagnoses whether the current token is not written properly for the given
|
|
// fixity. For example, because mandatory whitespace is missing. Regardless of
|
|
// whether there's an error, it's expected that parsing continues.
|
|
auto DiagnoseOperatorFixity(OperatorFixity fixity) -> void;
|
|
|
|
// If the current position is a `,`, consumes it, adds the provided token, and
|
|
// returns `Comma`. Returns `Close` if the current position is close_token
|
|
// (for example, `)`). `CommaClose` indicates it found both (for example,
|
|
// `,)`). Handles cases where invalid tokens are present by advancing the
|
|
// position, and may emit errors. Pass already_has_error in order to suppress
|
|
// duplicate errors.
|
|
auto ConsumeListToken(NodeKind comma_kind, Lex::TokenKind close_kind,
|
|
bool already_has_error) -> ListTokenKind;
|
|
|
|
// Gets the kind of the next token to be consumed. If `lookahead` is
|
|
// provided, it specifies which token to inspect.
|
|
auto PositionKind(Lookahead lookahead = Lookahead::CurrentToken) const
|
|
-> Lex::TokenKind {
|
|
return tokens_->GetKind(position_[static_cast<int32_t>(lookahead)]);
|
|
}
|
|
|
|
// Tests whether the next token to be consumed is of the specified kind. If
|
|
// `lookahead` is provided, it specifies which token to inspect.
|
|
auto PositionIs(Lex::TokenKind kind,
|
|
Lookahead lookahead = Lookahead::CurrentToken) const -> bool {
|
|
return PositionKind(lookahead) == kind;
|
|
}
|
|
|
|
// Pops the state and keeps the value for inspection.
|
|
auto PopState() -> StateStackEntry {
|
|
auto back = state_stack_.pop_back_val();
|
|
CARBON_VLOG("Pop {0}: {1}\n", state_stack_.size(), back);
|
|
return back;
|
|
}
|
|
|
|
// Pops the state and discards it.
|
|
auto PopAndDiscardState() -> void {
|
|
CARBON_VLOG("PopAndDiscard {0}: {1}\n", state_stack_.size() - 1,
|
|
state_stack_.back());
|
|
state_stack_.pop_back();
|
|
}
|
|
|
|
// Pushes a new state with the current position for context.
|
|
auto PushState(State state) -> void { PushState(state, *position_); }
|
|
|
|
// Pushes a new state with a specific token for context. Used when forming a
|
|
// new subtree when the current position isn't the start of the subtree.
|
|
auto PushState(State state, Lex::TokenIndex token) -> void {
|
|
PushState({.state = state, .token = token, .subtree_start = tree_->size()});
|
|
}
|
|
|
|
// Pushes a new expression state with specific precedence.
|
|
auto PushStateForExpr(PrecedenceGroup ambient_precedence) -> void {
|
|
PushState({.state = State::Expr,
|
|
.ambient_precedence = ambient_precedence,
|
|
.token = *position_,
|
|
.subtree_start = tree_->size()});
|
|
}
|
|
|
|
// Pushes a new state with detailed precedence for expression resume states.
|
|
auto PushStateForExprLoop(State state, PrecedenceGroup ambient_precedence,
|
|
PrecedenceGroup lhs_precedence) -> void {
|
|
PushState({.state = state,
|
|
.ambient_precedence = ambient_precedence,
|
|
.lhs_precedence = lhs_precedence,
|
|
.token = *position_,
|
|
.subtree_start = tree_->size()});
|
|
}
|
|
|
|
// Pushes a constructed state onto the stack.
|
|
auto PushState(StateStackEntry state) -> void {
|
|
CARBON_VLOG("Push {0}: {1}\n", state_stack_.size(), state);
|
|
state_stack_.push_back(state);
|
|
CARBON_CHECK(state_stack_.size() < (1 << 20),
|
|
"Excessive stack size: likely infinite loop");
|
|
}
|
|
|
|
// Pushes a constructed state onto the stack, with a different parse state.
|
|
auto PushState(StateStackEntry state_entry, State parse_state) -> void {
|
|
state_entry.state = parse_state;
|
|
PushState(state_entry);
|
|
}
|
|
|
|
// Propagates an error up the state stack, to the parent state.
|
|
auto ReturnErrorOnState() -> void { state_stack_.back().has_error = true; }
|
|
|
|
// Adds a node for a declaration's semicolon. Includes error recovery when the
|
|
// token is not a semicolon, using `decl_kind` and `is_def_allowed` to inform
|
|
// diagnostics.
|
|
auto AddNodeExpectingDeclSemi(StateStackEntry state, NodeKind node_kind,
|
|
Lex::TokenKind decl_kind, bool is_def_allowed)
|
|
-> void;
|
|
|
|
// Emits a diagnostic for a declaration missing a semi.
|
|
auto DiagnoseExpectedDeclSemi(Lex::TokenKind expected_kind) -> void;
|
|
|
|
// Emits a diagnostic for a declaration missing a semi or definition.
|
|
auto DiagnoseExpectedDeclSemiOrDefinition(Lex::TokenKind expected_kind)
|
|
-> void;
|
|
|
|
// Handles error recovery in a declaration, particularly before any possible
|
|
// definition has started (although one could be present). Recover to a
|
|
// semicolon when it makes sense as a possible end, otherwise use the
|
|
// introducer token for the error.
|
|
auto RecoverFromDeclError(StateStackEntry state, NodeKind node_kind,
|
|
bool skip_past_likely_end) -> void;
|
|
|
|
// Handles parsing of the library name. Returns the name's ID on success,
|
|
// which may be invalid for `default`.
|
|
// TODO: Add an invalid node on error, fix callers to adapt.
|
|
auto ParseLibraryName(bool accept_default)
|
|
-> std::optional<StringLiteralValueId>;
|
|
|
|
// Handles parsing `library <name>`. Requires that the position is a `library`
|
|
// token. Returns the name's ID on success, which may be invalid for
|
|
// `default`.
|
|
auto ParseLibrarySpecifier(bool accept_default)
|
|
-> std::optional<StringLiteralValueId>;
|
|
|
|
// Sets the package declaration information. Called at most once.
|
|
auto set_packaging_decl(Tree::PackagingNames packaging_names, bool is_impl)
|
|
-> void {
|
|
CARBON_CHECK(!tree_->packaging_decl_);
|
|
tree_->packaging_decl_ = {.names = packaging_names, .is_impl = is_impl};
|
|
}
|
|
|
|
// Adds an import.
|
|
auto AddImport(Tree::PackagingNames package) -> void {
|
|
tree_->imports_.push_back(package);
|
|
}
|
|
|
|
// Adds a function definition start node, and begins tracking a deferred
|
|
// definition if necessary.
|
|
auto AddFunctionDefinitionStart(Lex::TokenIndex token, bool has_error)
|
|
-> void;
|
|
// Adds a function definition node, and ends tracking a deferred definition if
|
|
// necessary.
|
|
auto AddFunctionDefinition(Lex::TokenIndex token, bool has_error) -> void;
|
|
|
|
// Prints information for a stack dump.
|
|
auto PrintForStackDump(llvm::raw_ostream& output) const -> void;
|
|
|
|
auto tree() const -> const Tree& { return *tree_; }
|
|
|
|
auto tokens() const -> const Lex::TokenizedBuffer& { return *tokens_; }
|
|
|
|
auto emitter() -> Lex::TokenDiagnosticEmitter& { return *emitter_; }
|
|
|
|
auto position() -> Lex::TokenIterator& { return position_; }
|
|
auto position() const -> Lex::TokenIterator { return position_; }
|
|
|
|
auto state_stack() -> llvm::SmallVector<StateStackEntry>& {
|
|
return state_stack_;
|
|
}
|
|
|
|
auto state_stack() const -> const llvm::SmallVector<StateStackEntry>& {
|
|
return state_stack_;
|
|
}
|
|
|
|
auto packaging_state() const -> PackagingState { return packaging_state_; }
|
|
auto set_packaging_state(PackagingState packaging_state) -> void {
|
|
packaging_state_ = packaging_state;
|
|
}
|
|
auto first_non_packaging_token() const -> Lex::TokenIndex {
|
|
return first_non_packaging_token_;
|
|
}
|
|
auto set_first_non_packaging_token(Lex::TokenIndex token) -> void {
|
|
CARBON_CHECK(!first_non_packaging_token_.is_valid());
|
|
first_non_packaging_token_ = token;
|
|
}
|
|
|
|
private:
|
|
// Prints a single token for a stack dump. Used by PrintForStackDump.
|
|
auto PrintTokenForStackDump(llvm::raw_ostream& output,
|
|
Lex::TokenIndex token) const -> void;
|
|
|
|
Tree* tree_;
|
|
Lex::TokenizedBuffer* tokens_;
|
|
Lex::TokenDiagnosticEmitter* emitter_;
|
|
|
|
// Whether to print verbose output.
|
|
llvm::raw_ostream* vlog_stream_;
|
|
|
|
// The current position within the token buffer.
|
|
Lex::TokenIterator position_;
|
|
// The FileEnd token.
|
|
Lex::TokenIterator end_;
|
|
|
|
llvm::SmallVector<StateStackEntry> state_stack_;
|
|
|
|
// The deferred definition indexes of functions whose definitions have begun
|
|
// but not yet finished.
|
|
llvm::SmallVector<DeferredDefinitionIndex> deferred_definition_stack_;
|
|
|
|
// The current packaging state, whether `import`/`package` are allowed.
|
|
PackagingState packaging_state_ = PackagingState::FileStart;
|
|
// The first non-packaging token, starting as invalid. Used for packaging
|
|
// state warnings.
|
|
Lex::TokenIndex first_non_packaging_token_ = Lex::TokenIndex::Invalid;
|
|
};
|
|
|
|
} // namespace Carbon::Parse
|
|
|
|
#endif // CARBON_TOOLCHAIN_PARSE_CONTEXT_H_
|