T0b1/wasmtime
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

moved crates in lib/ to src/, renamed crates, modified some files' text (#660)

Browse Source 
moved crates in lib/ to src/, renamed crates, modified some files' text (#660)
This commit is contained in:
lazypassion
2019-01-28 18:56:54 -05:00
committed by Dan Gohman
parent 54959cf5bb
commit 747ad3c4c5
508 changed files with 94 additions and 92 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
cranelift/reader/src/error.rs Normal file
View File

@@ -0,0 +1,52 @@
//! Define the `Location`, `ParseError`, and `ParseResult` types.
#![macro_use]
use std::fmt;
/// The location of a `Token` or `Error`.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
pub struct Location {
/// Line number. Command-line arguments are line 0 and source file
/// lines start from 1.
pub line_number: usize,
}
/// A parse error is returned when the parse failed.
#[derive(Debug)]
pub struct ParseError {
/// Location of the error.
pub location: Location,
/// Error message.
pub message: String,
}
impl fmt::Display for ParseError {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
if self.location.line_number == 0 {
write!(f, "command-line arguments: {}", self.message)
} else {
write!(f, "{}: {}", self.location.line_number, self.message)
}
}
}
/// Result of a parser operation. The `ParseError` variant includes a location.
pub type ParseResult<T> = Result<T, ParseError>;
// Create an `Err` variant of `ParseResult<X>` from a location and `format!` args.
macro_rules! err {
( $loc:expr, $msg:expr ) => {
Err($crate::ParseError {
location: $loc.clone(),
message: $msg.to_string(),
})
};
( $loc:expr, $fmt:expr, $( $arg:expr ),+ ) => {
Err($crate::ParseError {
location: $loc.clone(),
message: format!( $fmt, $( $arg ),+ ),
})
};
}

65
cranelift/reader/src/isaspec.rs Normal file
View File

@@ -0,0 +1,65 @@
//! Parsed representation of `set` and `isa` commands.
//!
//! A test case file can contain `set` commands that set ISA-independent settings, and it can
//! contain `isa` commands that select an ISA and applies ISA-specific settings.
//!
//! If a test case file contains `isa` commands, the tests will only be run against the specified
//! ISAs. If the file contains no `isa` commands, the tests will be run against all supported ISAs.
use crate::error::{Location, ParseResult};
use crate::testcommand::TestOption;
use cranelift_codegen::isa::TargetIsa;
use cranelift_codegen::settings::{Configurable, Flags, SetError};
/// The ISA specifications in a `.clif` file.
pub enum IsaSpec {
/// The parsed file does not contain any `isa` commands, but it may contain `set` commands
/// which are reflected in the finished `Flags` object.
None(Flags),
/// The parsed file does contains `isa` commands.
/// Each `isa` command is used to configure a `TargetIsa` trait object.
Some(Vec<Box<TargetIsa>>),
}
impl IsaSpec {
/// If the `IsaSpec` contains exactly 1 `TargetIsa` we return a reference to it
pub fn unique_isa(&self) -> Option<&TargetIsa> {
if let IsaSpec::Some(ref isa_vec) = *self {
if isa_vec.len() == 1 {
return Some(&*isa_vec[0]);
}
}
None
}
}
/// Parse an iterator of command line options and apply them to `config`.
pub fn parse_options<'a, I>(iter: I, config: &mut Configurable, loc: Location) -> ParseResult<()>
where
I: Iterator<Item = &'a str>,
{
for opt in iter.map(TestOption::new) {
match opt {
TestOption::Flag(name) => match config.enable(name) {
Ok(_) => {}
Err(SetError::BadName(name)) => return err!(loc, "unknown flag '{}'", name),
Err(_) => return err!(loc, "not a boolean flag: '{}'", opt),
},
TestOption::Value(name, value) => match config.set(name, value) {
Ok(_) => {}
Err(SetError::BadName(name)) => return err!(loc, "unknown setting '{}'", name),
Err(SetError::BadType) => return err!(loc, "invalid setting type: '{}'", opt),
Err(SetError::BadValue(expected)) => {
return err!(
loc,
"invalid setting value for '{}', expected {}",
opt,
expected
);
}
},
}
}
Ok(())
}

646
cranelift/reader/src/lexer.rs Normal file
View File

@@ -0,0 +1,646 @@
//! Lexical analysis for .clif files.
use crate::error::Location;
use cranelift_codegen::ir::types;
use cranelift_codegen::ir::{Ebb, Value};
#[allow(unused_imports, deprecated)]
use std::ascii::AsciiExt;
use std::str::CharIndices;
use std::u16;
/// A Token returned from the `Lexer`.
///
/// Some variants may contains references to the original source text, so the `Token` has the same
/// lifetime as the source.
#[derive(Debug, PartialEq, Eq, Clone, Copy)]
pub enum Token<'a> {
Comment(&'a str),
LPar, // '('
RPar, // ')'
LBrace, // '{'
RBrace, // '}'
LBracket, // '['
RBracket, // ']'
Minus, // '-'
Plus, // '+'
Comma, // ','
Dot, // '.'
Colon, // ':'
Equal, // '='
Arrow, // '->'
Float(&'a str), // Floating point immediate
Integer(&'a str), // Integer immediate
Type(types::Type), // i32, f32, b32x4, ...
Value(Value), // v12, v7
Ebb(Ebb), // ebb3
StackSlot(u32), // ss3
GlobalValue(u32), // gv3
Heap(u32), // heap2
Table(u32), // table2
JumpTable(u32), // jt2
FuncRef(u32), // fn2
SigRef(u32), // sig2
UserRef(u32), // u345
Name(&'a str), // %9arbitrary_alphanum, %x3, %0, %function ...
HexSequence(&'a str), // #89AF
Identifier(&'a str), // Unrecognized identifier (opcode, enumerator, ...)
SourceLoc(&'a str), // @00c7
}
/// A `Token` with an associated location.
#[derive(Debug, PartialEq, Eq)]
pub struct LocatedToken<'a> {
pub token: Token<'a>,
pub location: Location,
}
/// Wrap up a `Token` with the given location.
fn token(token: Token, loc: Location) -> Result<LocatedToken, LocatedError> {
Ok(LocatedToken {
token,
location: loc,
})
}
/// An error from the lexical analysis.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum LexError {
InvalidChar,
}
/// A `LexError` with an associated Location.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct LocatedError {
pub error: LexError,
pub location: Location,
}
/// Wrap up a `LexError` with the given location.
fn error<'a>(error: LexError, loc: Location) -> Result<LocatedToken<'a>, LocatedError> {
Err(LocatedError {
error,
location: loc,
})
}
/// Get the number of decimal digits at the end of `s`.
fn trailing_digits(s: &str) -> usize {
// It's faster to iterate backwards over bytes, and we're only counting ASCII digits.
s.as_bytes()
.iter()
.rev()
.take_while(|&&b| b'0' <= b && b <= b'9')
.count()
}
/// Pre-parse a supposed entity name by splitting it into two parts: A head of lowercase ASCII
/// letters and numeric tail.
pub fn split_entity_name(name: &str) -> Option<(&str, u32)> {
let (head, tail) = name.split_at(name.len() - trailing_digits(name));
if tail.len() > 1 && tail.starts_with('0') {
None
} else {
tail.parse().ok().map(|n| (head, n))
}
}
/// Lexical analysis.
///
/// A `Lexer` reads text from a `&str` and provides a sequence of tokens.
///
/// Also keep track of a line number for error reporting.
///
pub struct Lexer<'a> {
// Complete source being processed.
source: &'a str,
// Iterator into `source`.
chars: CharIndices<'a>,
// Next character to be processed, or `None` at the end.
lookahead: Option<char>,
// Index into `source` of lookahead character.
pos: usize,
// Current line number.
line_number: usize,
}
impl<'a> Lexer<'a> {
pub fn new(s: &'a str) -> Self {
let mut lex = Self {
source: s,
chars: s.char_indices(),
lookahead: None,
pos: 0,
line_number: 1,
};
// Advance to the first char.
lex.next_ch();
lex
}
// Advance to the next character.
// Return the next lookahead character, or None when the end is encountered.
// Always update cur_ch to reflect
fn next_ch(&mut self) -> Option<char> {
if self.lookahead == Some('\n') {
self.line_number += 1;
}
match self.chars.next() {
Some((idx, ch)) => {
self.pos = idx;
self.lookahead = Some(ch);
}
None => {
self.pos = self.source.len();
self.lookahead = None;
}
}
self.lookahead
}
// Get the location corresponding to `lookahead`.
fn loc(&self) -> Location {
Location {
line_number: self.line_number,
}
}
// Starting from `lookahead`, are we looking at `prefix`?
fn looking_at(&self, prefix: &str) -> bool {
self.source[self.pos..].starts_with(prefix)
}
// Starting from `lookahead`, are we looking at a number?
fn looking_at_numeric(&self) -> bool {
if let Some(c) = self.lookahead {
if c.is_digit(10) {
return true;
}
match c {
'-' => return true,
'+' => return true,
'.' => return true,
_ => {}
}
if self.looking_at("NaN") || self.looking_at("Inf") || self.looking_at("sNaN") {
return true;
}
}
false
}
// Scan a single-char token.
fn scan_char(&mut self, tok: Token<'a>) -> Result<LocatedToken<'a>, LocatedError> {
assert_ne!(self.lookahead, None);
let loc = self.loc();
self.next_ch();
token(tok, loc)
}
// Scan a multi-char token.
fn scan_chars(
&mut self,
count: usize,
tok: Token<'a>,
) -> Result<LocatedToken<'a>, LocatedError> {
let loc = self.loc();
for _ in 0..count {
assert_ne!(self.lookahead, None);
self.next_ch();
}
token(tok, loc)
}
/// Get the rest of the current line.
/// The next token returned by `next()` will be from the following lines.
pub fn rest_of_line(&mut self) -> &'a str {
let begin = self.pos;
loop {
match self.next_ch() {
None | Some('\n') => return &self.source[begin..self.pos],
_ => {}
}
}
}
// Scan a comment extending to the end of the current line.
fn scan_comment(&mut self) -> Result<LocatedToken<'a>, LocatedError> {
let loc = self.loc();
let text = self.rest_of_line();
token(Token::Comment(text), loc)
}
// Scan a number token which can represent either an integer or floating point number.
//
// Accept the following forms:
//
// - `10`: Integer
// - `-10`: Integer
// - `0xff_00`: Integer
// - `0.0`: Float
// - `0x1.f`: Float
// - `-0x2.4`: Float
// - `0x0.4p-34`: Float
//
// This function does not filter out all invalid numbers. It depends in the context-sensitive
// decoding of the text for that. For example, the number of allowed digits in an `Ieee32` and
// an `Ieee64` constant are different.
fn scan_number(&mut self) -> Result<LocatedToken<'a>, LocatedError> {
let begin = self.pos;
let loc = self.loc();
let mut is_float = false;
// Skip a leading sign.
match self.lookahead {
Some('-') => {
self.next_ch();
if !self.looking_at_numeric() {
// If the next characters won't parse as a number, we return Token::Minus
return token(Token::Minus, loc);
}
}
Some('+') => {
self.next_ch();
if !self.looking_at_numeric() {
// If the next characters won't parse as a number, we return Token::Minus
return token(Token::Plus, loc);
}
}
_ => {}
}
// Check for NaNs with payloads.
if self.looking_at("NaN:") || self.looking_at("sNaN:") {
// Skip the `NaN:` prefix, the loop below won't accept it.
// We expect a hexadecimal number to follow the colon.
while self.next_ch() != Some(':') {}
is_float = true;
} else if self.looking_at("NaN") || self.looking_at("Inf") {
// This is Inf or a default quiet NaN.
is_float = true;
}
// Look for the end of this number. Detect the radix point if there is one.
loop {
match self.next_ch() {
Some('-') | Some('_') => {}
Some('.') => is_float = true,
Some(ch) if ch.is_alphanumeric() => {}
_ => break,
}
}
let text = &self.source[begin..self.pos];
if is_float {
token(Token::Float(text), loc)
} else {
token(Token::Integer(text), loc)
}
}
// Scan a 'word', which is an identifier-like sequence of characters beginning with '_' or an
// alphabetic char, followed by zero or more alphanumeric or '_' characters.
fn scan_word(&mut self) -> Result<LocatedToken<'a>, LocatedError> {
let begin = self.pos;
let loc = self.loc();
assert!(self.lookahead == Some('_') || self.lookahead.unwrap().is_alphabetic());
loop {
match self.next_ch() {
Some('_') => {}
Some(ch) if ch.is_alphanumeric() => {}
_ => break,
}
}
let text = &self.source[begin..self.pos];
// Look for numbered well-known entities like ebb15, v45, ...
token(
split_entity_name(text)
.and_then(|(prefix, number)| {
Self::numbered_entity(prefix, number)
.or_else(|| Self::value_type(text, prefix, number))
})
.unwrap_or_else(|| match text {
"iflags" => Token::Type(types::IFLAGS),
"fflags" => Token::Type(types::FFLAGS),
_ => Token::Identifier(text),
}),
loc,
)
}
// If prefix is a well-known entity prefix and suffix is a valid entity number, return the
// decoded token.
fn numbered_entity(prefix: &str, number: u32) -> Option<Token<'a>> {
match prefix {
"v" => Value::with_number(number).map(Token::Value),
"ebb" => Ebb::with_number(number).map(Token::Ebb),
"ss" => Some(Token::StackSlot(number)),
"gv" => Some(Token::GlobalValue(number)),
"heap" => Some(Token::Heap(number)),
"table" => Some(Token::Table(number)),
"jt" => Some(Token::JumpTable(number)),
"fn" => Some(Token::FuncRef(number)),
"sig" => Some(Token::SigRef(number)),
"u" => Some(Token::UserRef(number)),
_ => None,
}
}
// Recognize a scalar or vector type.
fn value_type(text: &str, prefix: &str, number: u32) -> Option<Token<'a>> {
let is_vector = prefix.ends_with('x');
let scalar = if is_vector {
&prefix[0..prefix.len() - 1]
} else {
text
};
let base_type = match scalar {
"i8" => types::I8,
"i16" => types::I16,
"i32" => types::I32,
"i64" => types::I64,
"f32" => types::F32,
"f64" => types::F64,
"b1" => types::B1,
"b8" => types::B8,
"b16" => types::B16,
"b32" => types::B32,
"b64" => types::B64,
_ => return None,
};
if is_vector {
if number <= u32::from(u16::MAX) {
base_type.by(number as u16).map(Token::Type)
} else {
None
}
} else {
Some(Token::Type(base_type))
}
}
fn scan_name(&mut self) -> Result<LocatedToken<'a>, LocatedError> {
let loc = self.loc();
let begin = self.pos + 1;
assert_eq!(self.lookahead, Some('%'));
while let Some(c) = self.next_ch() {
if !(c.is_ascii() && c.is_alphanumeric() || c == '_') {
break;
}
}
let end = self.pos;
token(Token::Name(&self.source[begin..end]), loc)
}
fn scan_hex_sequence(&mut self) -> Result<LocatedToken<'a>, LocatedError> {
let loc = self.loc();
let begin = self.pos + 1;
assert_eq!(self.lookahead, Some('#'));
while let Some(c) = self.next_ch() {
if !char::is_digit(c, 16) {
break;
}
}
let end = self.pos;
token(Token::HexSequence(&self.source[begin..end]), loc)
}
fn scan_srcloc(&mut self) -> Result<LocatedToken<'a>, LocatedError> {
let loc = self.loc();
let begin = self.pos + 1;
assert_eq!(self.lookahead, Some('@'));
while let Some(c) = self.next_ch() {
if !char::is_digit(c, 16) {
break;
}
}
let end = self.pos;
token(Token::SourceLoc(&self.source[begin..end]), loc)
}
/// Get the next token or a lexical error.
///
/// Return None when the end of the source is encountered.
pub fn next(&mut self) -> Option<Result<LocatedToken<'a>, LocatedError>> {
loop {
let loc = self.loc();
return match self.lookahead {
None => None,
Some(';') => Some(self.scan_comment()),
Some('(') => Some(self.scan_char(Token::LPar)),
Some(')') => Some(self.scan_char(Token::RPar)),
Some('{') => Some(self.scan_char(Token::LBrace)),
Some('}') => Some(self.scan_char(Token::RBrace)),
Some('[') => Some(self.scan_char(Token::LBracket)),
Some(']') => Some(self.scan_char(Token::RBracket)),
Some(',') => Some(self.scan_char(Token::Comma)),
Some('.') => Some(self.scan_char(Token::Dot)),
Some(':') => Some(self.scan_char(Token::Colon)),
Some('=') => Some(self.scan_char(Token::Equal)),
Some('+') => Some(self.scan_number()),
Some('-') => {
if self.looking_at("->") {
Some(self.scan_chars(2, Token::Arrow))
} else {
Some(self.scan_number())
}
}
Some(ch) if ch.is_digit(10) => Some(self.scan_number()),
Some(ch) if ch.is_alphabetic() => Some(self.scan_word()),
Some('%') => Some(self.scan_name()),
Some('#') => Some(self.scan_hex_sequence()),
Some('@') => Some(self.scan_srcloc()),
Some(ch) if ch.is_whitespace() => {
self.next_ch();
continue;
}
_ => {
// Skip invalid char, return error.
self.next_ch();
Some(error(LexError::InvalidChar, loc))
}
};
}
}
}
#[cfg(test)]
mod tests {
use super::trailing_digits;
use super::*;
use crate::error::Location;
use cranelift_codegen::ir::types;
use cranelift_codegen::ir::{Ebb, Value};
#[test]
fn digits() {
assert_eq!(trailing_digits(""), 0);
assert_eq!(trailing_digits("x"), 0);
assert_eq!(trailing_digits("0x"), 0);
assert_eq!(trailing_digits("x1"), 1);
assert_eq!(trailing_digits("1x1"), 1);
assert_eq!(trailing_digits("1x01"), 2);
}
#[test]
fn entity_name() {
assert_eq!(split_entity_name(""), None);
assert_eq!(split_entity_name("x"), None);
assert_eq!(split_entity_name("x+"), None);
assert_eq!(split_entity_name("x+1"), Some(("x+", 1)));
assert_eq!(split_entity_name("x-1"), Some(("x-", 1)));
assert_eq!(split_entity_name("1"), Some(("", 1)));
assert_eq!(split_entity_name("x1"), Some(("x", 1)));
assert_eq!(split_entity_name("xy0"), Some(("xy", 0)));
// Reject this non-canonical form.
assert_eq!(split_entity_name("inst01"), None);
}
fn token<'a>(token: Token<'a>, line: usize) -> Option<Result<LocatedToken<'a>, LocatedError>> {
Some(super::token(token, Location { line_number: line }))
}
fn error<'a>(error: LexError, line: usize) -> Option<Result<LocatedToken<'a>, LocatedError>> {
Some(super::error(error, Location { line_number: line }))
}
#[test]
fn make_lexer() {
let mut l1 = Lexer::new("");
let mut l2 = Lexer::new(" ");
let mut l3 = Lexer::new("\n ");
assert_eq!(l1.next(), None);
assert_eq!(l2.next(), None);
assert_eq!(l3.next(), None);
}
#[test]
fn lex_comment() {
let mut lex = Lexer::new("; hello");
assert_eq!(lex.next(), token(Token::Comment("; hello"), 1));
assert_eq!(lex.next(), None);
lex = Lexer::new("\n ;hello\n;foo");
assert_eq!(lex.next(), token(Token::Comment(";hello"), 2));
assert_eq!(lex.next(), token(Token::Comment(";foo"), 3));
assert_eq!(lex.next(), None);
// Scan a comment after an invalid char.
let mut lex = Lexer::new("$; hello");
assert_eq!(lex.next(), error(LexError::InvalidChar, 1));
assert_eq!(lex.next(), token(Token::Comment("; hello"), 1));
assert_eq!(lex.next(), None);
}
#[test]
fn lex_chars() {
let mut lex = Lexer::new("(); hello\n = :{, }.");
assert_eq!(lex.next(), token(Token::LPar, 1));
assert_eq!(lex.next(), token(Token::RPar, 1));
assert_eq!(lex.next(), token(Token::Comment("; hello"), 1));
assert_eq!(lex.next(), token(Token::Equal, 2));
assert_eq!(lex.next(), token(Token::Colon, 2));
assert_eq!(lex.next(), token(Token::LBrace, 2));
assert_eq!(lex.next(), token(Token::Comma, 2));
assert_eq!(lex.next(), token(Token::RBrace, 2));
assert_eq!(lex.next(), token(Token::Dot, 2));
assert_eq!(lex.next(), None);
}
#[test]
fn lex_numbers() {
let mut lex = Lexer::new(" 0 2_000 -1,0xf -0x0 0.0 0x0.4p-34 +5");
assert_eq!(lex.next(), token(Token::Integer("0"), 1));
assert_eq!(lex.next(), token(Token::Integer("2_000"), 1));
assert_eq!(lex.next(), token(Token::Integer("-1"), 1));
assert_eq!(lex.next(), token(Token::Comma, 1));
assert_eq!(lex.next(), token(Token::Integer("0xf"), 1));
assert_eq!(lex.next(), token(Token::Integer("-0x0"), 1));
assert_eq!(lex.next(), token(Token::Float("0.0"), 1));
assert_eq!(lex.next(), token(Token::Float("0x0.4p-34"), 1));
assert_eq!(lex.next(), token(Token::Integer("+5"), 1));
assert_eq!(lex.next(), None);
}
#[test]
fn lex_identifiers() {
let mut lex = Lexer::new(
"v0 v00 vx01 ebb1234567890 ebb5234567890 v1x vx1 vxvx4 \
function0 function b1 i32x4 f32x5 \
iflags fflags iflagss",
);
assert_eq!(
lex.next(),
token(Token::Value(Value::with_number(0).unwrap()), 1)
);
assert_eq!(lex.next(), token(Token::Identifier("v00"), 1));
assert_eq!(lex.next(), token(Token::Identifier("vx01"), 1));
assert_eq!(
lex.next(),
token(Token::Ebb(Ebb::with_number(1234567890).unwrap()), 1)
);
assert_eq!(lex.next(), token(Token::Identifier("ebb5234567890"), 1));
assert_eq!(lex.next(), token(Token::Identifier("v1x"), 1));
assert_eq!(lex.next(), token(Token::Identifier("vx1"), 1));
assert_eq!(lex.next(), token(Token::Identifier("vxvx4"), 1));
assert_eq!(lex.next(), token(Token::Identifier("function0"), 1));
assert_eq!(lex.next(), token(Token::Identifier("function"), 1));
assert_eq!(lex.next(), token(Token::Type(types::B1), 1));
assert_eq!(lex.next(), token(Token::Type(types::I32X4), 1));
assert_eq!(lex.next(), token(Token::Identifier("f32x5"), 1));
assert_eq!(lex.next(), token(Token::Type(types::IFLAGS), 1));
assert_eq!(lex.next(), token(Token::Type(types::FFLAGS), 1));
assert_eq!(lex.next(), token(Token::Identifier("iflagss"), 1));
assert_eq!(lex.next(), None);
}
#[test]
fn lex_hex_sequences() {
let mut lex = Lexer::new("#0 #DEADbeef123 #789");
assert_eq!(lex.next(), token(Token::HexSequence("0"), 1));
assert_eq!(lex.next(), token(Token::HexSequence("DEADbeef123"), 1));
assert_eq!(lex.next(), token(Token::HexSequence("789"), 1));
}
#[test]
fn lex_names() {
let mut lex = Lexer::new("%0 %x3 %function %123_abc %ss0 %v3 %ebb11 %_");
assert_eq!(lex.next(), token(Token::Name("0"), 1));
assert_eq!(lex.next(), token(Token::Name("x3"), 1));
assert_eq!(lex.next(), token(Token::Name("function"), 1));
assert_eq!(lex.next(), token(Token::Name("123_abc"), 1));
assert_eq!(lex.next(), token(Token::Name("ss0"), 1));
assert_eq!(lex.next(), token(Token::Name("v3"), 1));
assert_eq!(lex.next(), token(Token::Name("ebb11"), 1));
assert_eq!(lex.next(), token(Token::Name("_"), 1));
}
#[test]
fn lex_userrefs() {
let mut lex = Lexer::new("u0 u1 u234567890 u9:8765");
assert_eq!(lex.next(), token(Token::UserRef(0), 1));
assert_eq!(lex.next(), token(Token::UserRef(1), 1));
assert_eq!(lex.next(), token(Token::UserRef(234567890), 1));
assert_eq!(lex.next(), token(Token::UserRef(9), 1));
assert_eq!(lex.next(), token(Token::Colon, 1));
assert_eq!(lex.next(), token(Token::Integer("8765"), 1));
assert_eq!(lex.next(), None);
}
}

42
cranelift/reader/src/lib.rs Normal file
View File

@@ -0,0 +1,42 @@
//! Cranelift file reader library.
//!
//! The `cranelift_reader` library supports reading .clif files. This functionality is needed for
//! testing Cranelift, but is not essential for a JIT compiler.
#![deny(
missing_docs,
trivial_numeric_casts,
unused_extern_crates,
unstable_features
)]
#![warn(unused_import_braces)]
#![cfg_attr(feature = "clippy", plugin(clippy(conf_file = "../../clippy.toml")))]
#![cfg_attr(feature = "cargo-clippy", allow(clippy::new_without_default))]
#![cfg_attr(
feature = "cargo-clippy",
warn(
clippy::float_arithmetic,
clippy::mut_mut,
clippy::nonminimal_bool,
clippy::option_map_unwrap_or,
clippy::option_map_unwrap_or_else,
clippy::print_stdout,
clippy::unicode_not_nfc,
clippy::use_self
)
)]
pub use crate::error::{Location, ParseError, ParseResult};
pub use crate::isaspec::{parse_options, IsaSpec};
pub use crate::parser::{parse_functions, parse_test};
pub use crate::sourcemap::SourceMap;
pub use crate::testcommand::{TestCommand, TestOption};
pub use crate::testfile::{Comment, Details, TestFile};
mod error;
mod isaspec;
mod lexer;
mod parser;
mod sourcemap;
mod testcommand;
mod testfile;

2987
cranelift/reader/src/parser.rs Normal file
View File

File diff suppressed because it is too large Load Diff

240
cranelift/reader/src/sourcemap.rs Normal file
View File

@@ -0,0 +1,240 @@
//! Source map associating entities with their source locations.
//!
//! When the parser reads in a source file, it records the locations of the
//! definitions of entities like instructions, EBBs, and values.
//!
//! The `SourceMap` struct defined in this module makes this mapping available
//! to parser clients.
use crate::error::{Location, ParseResult};
use crate::lexer::split_entity_name;
use cranelift_codegen::ir::entities::AnyEntity;
use cranelift_codegen::ir::{
Ebb, FuncRef, GlobalValue, Heap, JumpTable, SigRef, StackSlot, Table, Value,
};
use std::collections::HashMap;
/// Mapping from entity names to source locations.
#[derive(Debug, Default)]
pub struct SourceMap {
// Store locations for entities, including instructions.
locations: HashMap<AnyEntity, Location>,
}
/// Read-only interface which is exposed outside the parser crate.
impl SourceMap {
/// Look up a value entity.
pub fn contains_value(&self, v: Value) -> bool {
self.locations.contains_key(&v.into())
}
/// Look up a EBB entity.
pub fn contains_ebb(&self, ebb: Ebb) -> bool {
self.locations.contains_key(&ebb.into())
}
/// Look up a stack slot entity.
pub fn contains_ss(&self, ss: StackSlot) -> bool {
self.locations.contains_key(&ss.into())
}
/// Look up a global value entity.
pub fn contains_gv(&self, gv: GlobalValue) -> bool {
self.locations.contains_key(&gv.into())
}
/// Look up a heap entity.
pub fn contains_heap(&self, heap: Heap) -> bool {
self.locations.contains_key(&heap.into())
}
/// Look up a table entity.
pub fn contains_table(&self, table: Table) -> bool {
self.locations.contains_key(&table.into())
}
/// Look up a signature entity.
pub fn contains_sig(&self, sig: SigRef) -> bool {
self.locations.contains_key(&sig.into())
}
/// Look up a function entity.
pub fn contains_fn(&self, fn_: FuncRef) -> bool {
self.locations.contains_key(&fn_.into())
}
/// Look up a jump table entity.
pub fn contains_jt(&self, jt: JumpTable) -> bool {
self.locations.contains_key(&jt.into())
}
/// Look up an entity by source name.
/// Returns the entity reference corresponding to `name`, if it exists.
pub fn lookup_str(&self, name: &str) -> Option<AnyEntity> {
split_entity_name(name).and_then(|(ent, num)| match ent {
"v" => Value::with_number(num).and_then(|v| {
if !self.contains_value(v) {
None
} else {
Some(v.into())
}
}),
"ebb" => Ebb::with_number(num).and_then(|ebb| {
if !self.contains_ebb(ebb) {
None
} else {
Some(ebb.into())
}
}),
"ss" => StackSlot::with_number(num).and_then(|ss| {
if !self.contains_ss(ss) {
None
} else {
Some(ss.into())
}
}),
"gv" => GlobalValue::with_number(num).and_then(|gv| {
if !self.contains_gv(gv) {
None
} else {
Some(gv.into())
}
}),
"heap" => Heap::with_number(num).and_then(|heap| {
if !self.contains_heap(heap) {
None
} else {
Some(heap.into())
}
}),
"table" => Table::with_number(num).and_then(|table| {
if !self.contains_table(table) {
None
} else {
Some(table.into())
}
}),
"sig" => SigRef::with_number(num).and_then(|sig| {
if !self.contains_sig(sig) {
None
} else {
Some(sig.into())
}
}),
"fn" => FuncRef::with_number(num).and_then(|fn_| {
if !self.contains_fn(fn_) {
None
} else {
Some(fn_.into())
}
}),
"jt" => JumpTable::with_number(num).and_then(|jt| {
if !self.contains_jt(jt) {
None
} else {
Some(jt.into())
}
}),
_ => None,
})
}
/// Get the source location where an entity was defined.
pub fn location(&self, entity: AnyEntity) -> Option<Location> {
self.locations.get(&entity).cloned()
}
}
impl SourceMap {
/// Create a new empty `SourceMap`.
pub fn new() -> Self {
Self {
locations: HashMap::new(),
}
}
/// Define the value `entity`.
pub fn def_value(&mut self, entity: Value, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define the ebb `entity`.
pub fn def_ebb(&mut self, entity: Ebb, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define the stack slot `entity`.
pub fn def_ss(&mut self, entity: StackSlot, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define the global value `entity`.
pub fn def_gv(&mut self, entity: GlobalValue, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define the heap `entity`.
pub fn def_heap(&mut self, entity: Heap, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define the table `entity`.
pub fn def_table(&mut self, entity: Table, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define the signature `entity`.
pub fn def_sig(&mut self, entity: SigRef, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define the external function `entity`.
pub fn def_fn(&mut self, entity: FuncRef, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define the jump table `entity`.
pub fn def_jt(&mut self, entity: JumpTable, loc: Location) -> ParseResult<()> {
self.def_entity(entity.into(), loc)
}
/// Define an entity. This can be used for instructions whose numbers never
/// appear in source, or implicitly defined signatures.
pub fn def_entity(&mut self, entity: AnyEntity, loc: Location) -> ParseResult<()> {
if self.locations.insert(entity, loc).is_some() {
err!(loc, "duplicate entity: {}", entity)
} else {
Ok(())
}
}
}
#[cfg(test)]
mod tests {
use crate::parse_test;
#[test]
fn details() {
let tf = parse_test(
"function %detail() {
ss10 = incoming_arg 13
jt10 = jump_table [ebb0]
ebb0(v4: i32, v7: i32):
v10 = iadd v4, v7
}",
None,
None,
)
.unwrap();
let map = &tf.functions[0].1.map;
assert_eq!(map.lookup_str("v0"), None);
assert_eq!(map.lookup_str("ss1"), None);
assert_eq!(map.lookup_str("ss10").unwrap().to_string(), "ss10");
assert_eq!(map.lookup_str("jt10").unwrap().to_string(), "jt10");
assert_eq!(map.lookup_str("ebb0").unwrap().to_string(), "ebb0");
assert_eq!(map.lookup_str("v4").unwrap().to_string(), "v4");
assert_eq!(map.lookup_str("v7").unwrap().to_string(), "v7");
assert_eq!(map.lookup_str("v10").unwrap().to_string(), "v10");
}
}

103
cranelift/reader/src/testcommand.rs Normal file
View File

@@ -0,0 +1,103 @@
//! Test commands.
//!
//! A `.clif` file can begin with one or more *test commands* which specify what is to be tested.
//! The general syntax is:
//!
//! <pre>
//! test <i>&lt;command&gt;</i> </i>[options]</i>...
//! </pre>
//!
//! The options are either a single identifier flag, or setting values like `identifier=value`.
//!
//! The parser does not understand the test commands or which options are valid. It simply parses
//! the general format into a `TestCommand` data structure.
use std::fmt::{self, Display, Formatter};
/// A command appearing in a test file.
#[derive(Clone, PartialEq, Eq, Debug)]
pub struct TestCommand<'a> {
/// The command name as a string.
pub command: &'a str,
/// The options following the command name.
pub options: Vec<TestOption<'a>>,
}
/// An option on a test command.
#[derive(Clone, PartialEq, Eq, Debug)]
pub enum TestOption<'a> {
/// Single identifier flag: `foo`.
Flag(&'a str),
/// A value assigned to an identifier: `foo=bar`.
Value(&'a str, &'a str),
}
impl<'a> TestCommand<'a> {
/// Create a new TestCommand by parsing `s`.
/// The returned command contains references into `s`.
pub fn new(s: &'a str) -> Self {
let mut parts = s.split_whitespace();
let cmd = parts.next().unwrap_or("");
Self {
command: cmd,
options: parts
.filter(|s| !s.is_empty())
.map(TestOption::new)
.collect(),
}
}
}
impl<'a> Display for TestCommand<'a> {
fn fmt(&self, f: &mut Formatter) -> fmt::Result {
write!(f, "{}", self.command)?;
for opt in &self.options {
write!(f, " {}", opt)?;
}
writeln!(f)
}
}
impl<'a> TestOption<'a> {
/// Create a new TestOption by parsing `s`.
/// The returned option contains references into `s`.
pub fn new(s: &'a str) -> Self {
match s.find('=') {
None => TestOption::Flag(s),
Some(p) => TestOption::Value(&s[0..p], &s[p + 1..]),
}
}
}
impl<'a> Display for TestOption<'a> {
fn fmt(&self, f: &mut Formatter) -> fmt::Result {
match *self {
TestOption::Flag(s) => write!(f, "{}", s),
TestOption::Value(s, v) => write!(f, "{}={}", s, v),
}
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn parse_option() {
assert_eq!(TestOption::new(""), TestOption::Flag(""));
assert_eq!(TestOption::new("foo"), TestOption::Flag("foo"));
assert_eq!(TestOption::new("foo=bar"), TestOption::Value("foo", "bar"));
}
#[test]
fn parse_command() {
assert_eq!(&TestCommand::new("").to_string(), "\n");
assert_eq!(&TestCommand::new("cat").to_string(), "cat\n");
assert_eq!(&TestCommand::new("cat ").to_string(), "cat\n");
assert_eq!(&TestCommand::new("cat 1 ").to_string(), "cat 1\n");
assert_eq!(
&TestCommand::new("cat one=4 two t").to_string(),
"cat one=4 two t\n"
);
}
}

57
cranelift/reader/src/testfile.rs Normal file
View File

@@ -0,0 +1,57 @@
//! Data structures representing a parsed test file.
//!
//! A test file is a `.clif` file which contains test commands and settings for running a
//! file-based test case.
//!
use crate::error::Location;
use crate::isaspec::IsaSpec;
use crate::sourcemap::SourceMap;
use crate::testcommand::TestCommand;
use cranelift_codegen::ir::entities::AnyEntity;
use cranelift_codegen::ir::Function;
/// A parsed test case.
///
/// This is the result of parsing a `.clif` file which contains a number of test commands and ISA
/// specs followed by the functions that should be tested.
pub struct TestFile<'a> {
/// `test foo ...` lines.
pub commands: Vec<TestCommand<'a>>,
/// `isa bar ...` lines.
pub isa_spec: IsaSpec,
/// Comments appearing before the first function.
/// These are all tagged as 'Function' scope for lack of a better entity.
pub preamble_comments: Vec<Comment<'a>>,
/// Parsed functions and additional details about each function.
pub functions: Vec<(Function, Details<'a>)>,
}
/// Additional details about a function parsed from a text string.
/// These are useful for detecting test commands embedded in comments etc.
/// The details to not affect the semantics of the function.
#[derive(Debug)]
pub struct Details<'a> {
/// Location of the `function` keyword that begins this function.
pub location: Location,
/// Annotation comments that appeared inside or after the function.
pub comments: Vec<Comment<'a>>,
/// Mapping of entity numbers to source locations.
pub map: SourceMap,
}
/// A comment in a parsed function.
///
/// The comment belongs to the immediately preceding entity, whether that is an EBB header, and
/// instruction, or one of the preamble declarations.
///
/// Comments appearing inside the function but before the preamble, as well as comments appearing
/// after the function are tagged as `AnyEntity::Function`.
#[derive(Clone, PartialEq, Eq, Debug)]
pub struct Comment<'a> {
/// The entity this comment is attached to.
/// Comments always follow their entity.
pub entity: AnyEntity,
/// Text of the comment, including the leading `;`.
pub text: &'a str,
}