diff options
Diffstat (limited to 'compiler/rustc_span/src')
| -rw-r--r-- | compiler/rustc_span/src/analyze_source_file.rs | 268 | ||||
| -rw-r--r-- | compiler/rustc_span/src/def_id.rs | 5 | ||||
| -rw-r--r-- | compiler/rustc_span/src/hygiene.rs | 9 | ||||
| -rw-r--r-- | compiler/rustc_span/src/lib.rs | 8 | ||||
| -rw-r--r-- | compiler/rustc_span/src/symbol.rs | 43 |
5 files changed, 74 insertions, 259 deletions
diff --git a/compiler/rustc_span/src/analyze_source_file.rs b/compiler/rustc_span/src/analyze_source_file.rs index fba20566580..141d261b5f0 100644 --- a/compiler/rustc_span/src/analyze_source_file.rs +++ b/compiler/rustc_span/src/analyze_source_file.rs @@ -29,165 +29,6 @@ pub(crate) fn analyze_source_file(src: &str) -> (Vec<RelativeBytePos>, Vec<Multi (lines, multi_byte_chars) } -#[cfg(bootstrap)] -cfg_match! { - cfg(any(target_arch = "x86", target_arch = "x86_64")) => { - fn analyze_source_file_dispatch( - src: &str, - lines: &mut Vec<RelativeBytePos>, - multi_byte_chars: &mut Vec<MultiByteChar>, - ) { - if is_x86_feature_detected!("sse2") { - unsafe { - analyze_source_file_sse2(src, lines, multi_byte_chars); - } - } else { - analyze_source_file_generic( - src, - src.len(), - RelativeBytePos::from_u32(0), - lines, - multi_byte_chars, - ); - } - } - - /// Checks 16 byte chunks of text at a time. If the chunk contains - /// something other than printable ASCII characters and newlines, the - /// function falls back to the generic implementation. Otherwise it uses - /// SSE2 intrinsics to quickly find all newlines. - #[target_feature(enable = "sse2")] - unsafe fn analyze_source_file_sse2( - src: &str, - lines: &mut Vec<RelativeBytePos>, - multi_byte_chars: &mut Vec<MultiByteChar>, - ) { - #[cfg(target_arch = "x86")] - use std::arch::x86::*; - #[cfg(target_arch = "x86_64")] - use std::arch::x86_64::*; - - const CHUNK_SIZE: usize = 16; - - let src_bytes = src.as_bytes(); - - let chunk_count = src.len() / CHUNK_SIZE; - - // This variable keeps track of where we should start decoding a - // chunk. If a multi-byte character spans across chunk boundaries, - // we need to skip that part in the next chunk because we already - // handled it. - let mut intra_chunk_offset = 0; - - for chunk_index in 0..chunk_count { - let ptr = src_bytes.as_ptr() as *const __m128i; - // We don't know if the pointer is aligned to 16 bytes, so we - // use `loadu`, which supports unaligned loading. - let chunk = unsafe { _mm_loadu_si128(ptr.add(chunk_index)) }; - - // For character in the chunk, see if its byte value is < 0, which - // indicates that it's part of a UTF-8 char. - let multibyte_test = unsafe { _mm_cmplt_epi8(chunk, _mm_set1_epi8(0)) }; - // Create a bit mask from the comparison results. - let multibyte_mask = unsafe { _mm_movemask_epi8(multibyte_test) }; - - // If the bit mask is all zero, we only have ASCII chars here: - if multibyte_mask == 0 { - assert!(intra_chunk_offset == 0); - - // Check if there are any control characters in the chunk. All - // control characters that we can encounter at this point have a - // byte value less than 32 or ... - let control_char_test0 = unsafe { _mm_cmplt_epi8(chunk, _mm_set1_epi8(32)) }; - let control_char_mask0 = unsafe { _mm_movemask_epi8(control_char_test0) }; - - // ... it's the ASCII 'DEL' character with a value of 127. - let control_char_test1 = unsafe { _mm_cmpeq_epi8(chunk, _mm_set1_epi8(127)) }; - let control_char_mask1 = unsafe { _mm_movemask_epi8(control_char_test1) }; - - let control_char_mask = control_char_mask0 | control_char_mask1; - - if control_char_mask != 0 { - // Check for newlines in the chunk - let newlines_test = unsafe { _mm_cmpeq_epi8(chunk, _mm_set1_epi8(b'\n' as i8)) }; - let newlines_mask = unsafe { _mm_movemask_epi8(newlines_test) }; - - if control_char_mask == newlines_mask { - // All control characters are newlines, record them - let mut newlines_mask = 0xFFFF0000 | newlines_mask as u32; - let output_offset = RelativeBytePos::from_usize(chunk_index * CHUNK_SIZE + 1); - - loop { - let index = newlines_mask.trailing_zeros(); - - if index >= CHUNK_SIZE as u32 { - // We have arrived at the end of the chunk. - break; - } - - lines.push(RelativeBytePos(index) + output_offset); - - // Clear the bit, so we can find the next one. - newlines_mask &= (!1) << index; - } - - // We are done for this chunk. All control characters were - // newlines and we took care of those. - continue; - } else { - // Some of the control characters are not newlines, - // fall through to the slow path below. - } - } else { - // No control characters, nothing to record for this chunk - continue; - } - } - - // The slow path. - // There are control chars in here, fallback to generic decoding. - let scan_start = chunk_index * CHUNK_SIZE + intra_chunk_offset; - intra_chunk_offset = analyze_source_file_generic( - &src[scan_start..], - CHUNK_SIZE - intra_chunk_offset, - RelativeBytePos::from_usize(scan_start), - lines, - multi_byte_chars, - ); - } - - // There might still be a tail left to analyze - let tail_start = chunk_count * CHUNK_SIZE + intra_chunk_offset; - if tail_start < src.len() { - analyze_source_file_generic( - &src[tail_start..], - src.len() - tail_start, - RelativeBytePos::from_usize(tail_start), - lines, - multi_byte_chars, - ); - } - } - } - _ => { - // The target (or compiler version) does not support SSE2 ... - fn analyze_source_file_dispatch( - src: &str, - lines: &mut Vec<RelativeBytePos>, - multi_byte_chars: &mut Vec<MultiByteChar>, - ) { - analyze_source_file_generic( - src, - src.len(), - RelativeBytePos::from_u32(0), - lines, - multi_byte_chars, - ); - } - } -} - -#[cfg(not(bootstrap))] cfg_match! { any(target_arch = "x86", target_arch = "x86_64") => { fn analyze_source_file_dispatch( @@ -227,9 +68,7 @@ cfg_match! { const CHUNK_SIZE: usize = 16; - let src_bytes = src.as_bytes(); - - let chunk_count = src.len() / CHUNK_SIZE; + let (chunks, tail) = src.as_bytes().as_chunks::<CHUNK_SIZE>(); // This variable keeps track of where we should start decoding a // chunk. If a multi-byte character spans across chunk boundaries, @@ -237,11 +76,10 @@ cfg_match! { // handled it. let mut intra_chunk_offset = 0; - for chunk_index in 0..chunk_count { - let ptr = src_bytes.as_ptr() as *const __m128i; + for (chunk_index, chunk) in chunks.iter().enumerate() { // We don't know if the pointer is aligned to 16 bytes, so we // use `loadu`, which supports unaligned loading. - let chunk = unsafe { _mm_loadu_si128(ptr.add(chunk_index)) }; + let chunk = unsafe { _mm_loadu_si128(chunk.as_ptr() as *const __m128i) }; // For character in the chunk, see if its byte value is < 0, which // indicates that it's part of a UTF-8 char. @@ -253,69 +91,36 @@ cfg_match! { if multibyte_mask == 0 { assert!(intra_chunk_offset == 0); - // Check if there are any control characters in the chunk. All - // control characters that we can encounter at this point have a - // byte value less than 32 or ... - let control_char_test0 = unsafe { _mm_cmplt_epi8(chunk, _mm_set1_epi8(32)) }; - let control_char_mask0 = unsafe { _mm_movemask_epi8(control_char_test0) }; - - // ... it's the ASCII 'DEL' character with a value of 127. - let control_char_test1 = unsafe { _mm_cmpeq_epi8(chunk, _mm_set1_epi8(127)) }; - let control_char_mask1 = unsafe { _mm_movemask_epi8(control_char_test1) }; - - let control_char_mask = control_char_mask0 | control_char_mask1; - - if control_char_mask != 0 { - // Check for newlines in the chunk - let newlines_test = unsafe { _mm_cmpeq_epi8(chunk, _mm_set1_epi8(b'\n' as i8)) }; - let newlines_mask = unsafe { _mm_movemask_epi8(newlines_test) }; + // Check for newlines in the chunk + let newlines_test = unsafe { _mm_cmpeq_epi8(chunk, _mm_set1_epi8(b'\n' as i8)) }; + let mut newlines_mask = unsafe { _mm_movemask_epi8(newlines_test) }; - if control_char_mask == newlines_mask { - // All control characters are newlines, record them - let mut newlines_mask = 0xFFFF0000 | newlines_mask as u32; - let output_offset = RelativeBytePos::from_usize(chunk_index * CHUNK_SIZE + 1); + let output_offset = RelativeBytePos::from_usize(chunk_index * CHUNK_SIZE + 1); - loop { - let index = newlines_mask.trailing_zeros(); + while newlines_mask != 0 { + let index = newlines_mask.trailing_zeros(); - if index >= CHUNK_SIZE as u32 { - // We have arrived at the end of the chunk. - break; - } + lines.push(RelativeBytePos(index) + output_offset); - lines.push(RelativeBytePos(index) + output_offset); - - // Clear the bit, so we can find the next one. - newlines_mask &= (!1) << index; - } - - // We are done for this chunk. All control characters were - // newlines and we took care of those. - continue; - } else { - // Some of the control characters are not newlines, - // fall through to the slow path below. - } - } else { - // No control characters, nothing to record for this chunk - continue; + // Clear the bit, so we can find the next one. + newlines_mask &= newlines_mask - 1; } + } else { + // The slow path. + // There are multibyte chars in here, fallback to generic decoding. + let scan_start = chunk_index * CHUNK_SIZE + intra_chunk_offset; + intra_chunk_offset = analyze_source_file_generic( + &src[scan_start..], + CHUNK_SIZE - intra_chunk_offset, + RelativeBytePos::from_usize(scan_start), + lines, + multi_byte_chars, + ); } - - // The slow path. - // There are control chars in here, fallback to generic decoding. - let scan_start = chunk_index * CHUNK_SIZE + intra_chunk_offset; - intra_chunk_offset = analyze_source_file_generic( - &src[scan_start..], - CHUNK_SIZE - intra_chunk_offset, - RelativeBytePos::from_usize(scan_start), - lines, - multi_byte_chars, - ); } // There might still be a tail left to analyze - let tail_start = chunk_count * CHUNK_SIZE + intra_chunk_offset; + let tail_start = src.len() - tail.len() + intra_chunk_offset; if tail_start < src.len() { analyze_source_file_generic( &src[tail_start..], @@ -369,29 +174,18 @@ fn analyze_source_file_generic( // string. let mut char_len = 1; - if byte < 32 { - // This is an ASCII control character, it could be one of the cases - // that are interesting to us. - + if byte == b'\n' { let pos = RelativeBytePos::from_usize(i) + output_offset; - - if let b'\n' = byte { - lines.push(pos + RelativeBytePos(1)); - } - } else if byte >= 127 { - // The slow path: - // This is either ASCII control character "DEL" or the beginning of - // a multibyte char. Just decode to `char`. + lines.push(pos + RelativeBytePos(1)); + } else if byte >= 128 { + // This is the beginning of a multibyte char. Just decode to `char`. let c = src[i..].chars().next().unwrap(); char_len = c.len_utf8(); let pos = RelativeBytePos::from_usize(i) + output_offset; - - if char_len > 1 { - assert!((2..=4).contains(&char_len)); - let mbc = MultiByteChar { pos, bytes: char_len as u8 }; - multi_byte_chars.push(mbc); - } + assert!((2..=4).contains(&char_len)); + let mbc = MultiByteChar { pos, bytes: char_len as u8 }; + multi_byte_chars.push(mbc); } i += char_len; diff --git a/compiler/rustc_span/src/def_id.rs b/compiler/rustc_span/src/def_id.rs index f61ce37131e..641bac88ad0 100644 --- a/compiler/rustc_span/src/def_id.rs +++ b/compiler/rustc_span/src/def_id.rs @@ -3,10 +3,9 @@ use std::hash::{BuildHasherDefault, Hash, Hasher}; use rustc_data_structures::AtomicRef; use rustc_data_structures::fingerprint::Fingerprint; -use rustc_data_structures::stable_hasher::{ - Hash64, HashStable, StableHasher, StableOrd, ToStableHashKey, -}; +use rustc_data_structures::stable_hasher::{HashStable, StableHasher, StableOrd, ToStableHashKey}; use rustc_data_structures::unhash::Unhasher; +use rustc_hashes::Hash64; use rustc_index::Idx; use rustc_macros::{Decodable, Encodable, HashStable_Generic}; use rustc_serialize::{Decodable, Encodable}; diff --git a/compiler/rustc_span/src/hygiene.rs b/compiler/rustc_span/src/hygiene.rs index 2910bcdf51d..84e89ff4b7d 100644 --- a/compiler/rustc_span/src/hygiene.rs +++ b/compiler/rustc_span/src/hygiene.rs @@ -33,9 +33,10 @@ use std::sync::Arc; use rustc_data_structures::fingerprint::Fingerprint; use rustc_data_structures::fx::{FxHashMap, FxHashSet}; -use rustc_data_structures::stable_hasher::{Hash64, HashStable, HashingControls, StableHasher}; +use rustc_data_structures::stable_hasher::{HashStable, HashingControls, StableHasher}; use rustc_data_structures::sync::{Lock, WorkerLocal}; use rustc_data_structures::unhash::UnhashMap; +use rustc_hashes::Hash64; use rustc_index::IndexVec; use rustc_macros::{Decodable, Encodable, HashStable_Generic}; use rustc_serialize::{Decodable, Decoder, Encodable, Encoder}; @@ -174,6 +175,12 @@ pub enum Transparency { Opaque, } +impl Transparency { + pub fn fallback(macro_rules: bool) -> Self { + if macro_rules { Transparency::SemiTransparent } else { Transparency::Opaque } + } +} + impl LocalExpnId { /// The ID of the theoretical expansion that generates freshly parsed, unexpanded AST. pub const ROOT: LocalExpnId = LocalExpnId::ZERO; diff --git a/compiler/rustc_span/src/lib.rs b/compiler/rustc_span/src/lib.rs index 0e146baef37..bca9323a50d 100644 --- a/compiler/rustc_span/src/lib.rs +++ b/compiler/rustc_span/src/lib.rs @@ -31,6 +31,7 @@ #![feature(round_char_boundary)] #![feature(rustc_attrs)] #![feature(rustdoc_internals)] +#![feature(slice_as_chunks)] #![warn(unreachable_pub)] // tidy-alphabetical-end @@ -87,9 +88,10 @@ use std::sync::Arc; use std::{fmt, iter}; use md5::{Digest, Md5}; -use rustc_data_structures::stable_hasher::{Hash64, Hash128, HashStable, StableHasher}; +use rustc_data_structures::stable_hasher::{HashStable, StableHasher}; use rustc_data_structures::sync::{FreezeLock, FreezeWriteGuard, Lock}; use rustc_data_structures::unord::UnordMap; +use rustc_hashes::{Hash64, Hash128}; use sha1::Sha1; use sha2::Sha256; @@ -403,7 +405,7 @@ impl fmt::Display for FileNameDisplay<'_> { impl<'a> FileNameDisplay<'a> { pub fn to_string_lossy(&self) -> Cow<'a, str> { match self.inner { - FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref), + FileName::Real(inner) => inner.to_string_lossy(self.display_pref), _ => Cow::from(self.to_string()), } } @@ -1441,7 +1443,7 @@ pub enum ExternalSourceKind { impl ExternalSource { pub fn get_source(&self) -> Option<&str> { match self { - ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src), + ExternalSource::Foreign { kind: ExternalSourceKind::Present(src), .. } => Some(src), _ => None, } } diff --git a/compiler/rustc_span/src/symbol.rs b/compiler/rustc_span/src/symbol.rs index c819d433235..172c2faca96 100644 --- a/compiler/rustc_span/src/symbol.rs +++ b/compiler/rustc_span/src/symbol.rs @@ -104,6 +104,8 @@ symbols! { Gen: "gen", // >= 2024 Edition only Try: "try", // >= 2018 Edition only + // NOTE: When adding new keywords, consider adding them to the ui/parser/raw/raw-idents.rs test. + // "Lifetime keywords": regular keywords with a leading `'`. // Matching predicates: `is_any_keyword` UnderscoreLifetime: "'_", @@ -190,9 +192,11 @@ symbols! { Capture, Cell, Center, + Child, Cleanup, Clone, CoercePointee, + CoercePointeeValidated, CoerceUnsized, Command, ConstParamTy, @@ -249,6 +253,7 @@ symbols! { Into, IntoFuture, IntoIterator, + IoBufRead, IoLines, IoRead, IoSeek, @@ -284,6 +289,7 @@ symbols! { OsString, Output, Param, + ParamSet, PartialEq, PartialOrd, Path, @@ -332,6 +338,7 @@ symbols! { SliceIter, Some, SpanCtxt, + Stdin, String, StructuralPartialEq, SubdiagMessage, @@ -514,6 +521,8 @@ symbols! { bang, begin_panic, bench, + bevy_ecs, + bikeshed_guaranteed_no_drop, bin, binaryheap_iter, bind_by_move_pattern_guards, @@ -594,6 +603,9 @@ symbols! { cfi, cfi_encoding, char, + char_is_ascii, + child_id, + child_kill, client, clippy, clobber_abi, @@ -619,6 +631,7 @@ symbols! { cmp_partialord_lt, cmpxchg16b_target_feature, cmse_nonsecure_entry, + coerce_pointee_validated, coerce_unsized, cold, cold_path, @@ -877,6 +890,7 @@ symbols! { extern_crate_self, extern_in_paths, extern_prelude, + extern_system_varargs, extern_types, external, external_doc, @@ -1024,6 +1038,7 @@ symbols! { generic_const_exprs, generic_const_items, generic_param_attrs, + generic_pattern_types, get_context, global_alloc_ty, global_allocator, @@ -1352,10 +1367,6 @@ symbols! { native_link_modifiers_whole_archive, natvis_file, ne, - nearbyintf128, - nearbyintf16, - nearbyintf32, - nearbyintf64, needs_allocator, needs_drop, needs_panic_runtime, @@ -1460,6 +1471,7 @@ symbols! { panic_2015, panic_2021, panic_abort, + panic_any, panic_bounds_check, panic_cannot_unwind, panic_const_add_overflow, @@ -1506,7 +1518,7 @@ symbols! { path_main_separator, path_to_pathbuf, pathbuf_as_path, - pattern_complexity, + pattern_complexity_limit, pattern_parentheses, pattern_type, pattern_types, @@ -1565,6 +1577,7 @@ symbols! { proc_macro_mod, proc_macro_non_items, proc_macro_path_invoc, + process_abort, process_exit, profiler_builtins, profiler_runtime, @@ -1671,20 +1684,16 @@ symbols! { return_position_impl_trait_in_trait, return_type_notation, rhs, - rintf128, - rintf16, - rintf32, - rintf64, riscv_target_feature, rlib, ropi, ropi_rwpi: "ropi-rwpi", rotate_left, rotate_right, - roundevenf128, - roundevenf16, - roundevenf32, - roundevenf64, + round_ties_even_f128, + round_ties_even_f16, + round_ties_even_f32, + round_ties_even_f64, roundf128, roundf16, roundf32, @@ -1755,7 +1764,6 @@ symbols! { rustc_insignificant_dtor, rustc_intrinsic, rustc_intrinsic_const_stable_indirect, - rustc_intrinsic_must_be_overridden, rustc_layout, rustc_layout_scalar_valid_range_end, rustc_layout_scalar_valid_range_start, @@ -1907,7 +1915,7 @@ symbols! { simd_shl, simd_shr, simd_shuffle, - simd_shuffle_generic, + simd_shuffle_const_generic, simd_sub, simd_trunc, simd_with_exposed_provenance, @@ -1967,6 +1975,10 @@ symbols! { str_from_utf8_mut, str_from_utf8_unchecked, str_from_utf8_unchecked_mut, + str_inherent_from_utf8, + str_inherent_from_utf8_mut, + str_inherent_from_utf8_unchecked, + str_inherent_from_utf8_unchecked_mut, str_len, str_split_whitespace, str_starts_with, @@ -1991,6 +2003,7 @@ symbols! { sub_assign, sub_with_overflow, suggestion, + supertrait_item_shadowing, surface_async_drop_in_place, sym, sync, |