Auto merge of rust-lang#139241 - bvanjoi:less-decoding, r=<try>

don't store opaque info during encoding

Now `remapped_ctxts` reserved and let's check the performance.

r? `@petrochenkov`

Author: bors

compiler/rustc_metadata/src/rmeta/mod.rs

@@ -36,7 +36,7 @@ use rustc_serialize::opaque::FileEncoder;
 use rustc_session::config::{SymbolManglingVersion, TargetModifier};
 use rustc_session::cstore::{CrateDepKind, ForeignModule, LinkagePreference, NativeLib};
 use rustc_span::edition::Edition;
-use rustc_span::hygiene::{ExpnIndex, MacroKind, SyntaxContextData};
+use rustc_span::hygiene::{ExpnIndex, MacroKind, SyntaxContextKey};
 use rustc_span::{self, ExpnData, ExpnHash, ExpnId, Ident, Span, Symbol};
 use rustc_target::spec::{PanicStrategy, TargetTuple};
 use table::TableBuilder;
@@ -193,7 +193,7 @@ enum LazyState {
     Previous(NonZero<usize>),
 }
 
-type SyntaxContextTable = LazyTable<u32, Option<LazyValue<SyntaxContextData>>>;
+type SyntaxContextTable = LazyTable<u32, Option<LazyValue<SyntaxContextKey>>>;
 type ExpnDataTable = LazyTable<ExpnIndex, Option<LazyValue<ExpnData>>>;
 type ExpnHashTable = LazyTable<ExpnIndex, Option<LazyValue<ExpnHash>>>;
 

compiler/rustc_middle/src/query/on_disk_cache.rs

@@ -16,7 +16,7 @@ use rustc_serialize::opaque::{FileEncodeResult, FileEncoder, IntEncodedWithFixed
 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
 use rustc_session::Session;
 use rustc_span::hygiene::{
-    ExpnId, HygieneDecodeContext, HygieneEncodeContext, SyntaxContext, SyntaxContextData,
+    ExpnId, HygieneDecodeContext, HygieneEncodeContext, SyntaxContext, SyntaxContextKey,
 };
 use rustc_span::source_map::Spanned;
 use rustc_span::{
@@ -75,9 +75,9 @@ pub struct OnDiskCache {
     alloc_decoding_state: AllocDecodingState,
 
     // A map from syntax context ids to the position of their associated
-    // `SyntaxContextData`. We use a `u32` instead of a `SyntaxContext`
+    // `SyntaxContextKey`. We use a `u32` instead of a `SyntaxContext`
     // to represent the fact that we are storing *encoded* ids. When we decode
-    // a `SyntaxContext`, a new id will be allocated from the global `HygieneData`,
+    // a `SyntaxContextKey`, a new id will be allocated from the global `HygieneData`,
     // which will almost certainly be different than the serialized id.
     syntax_contexts: FxHashMap<u32, AbsoluteBytePos>,
     // A map from the `DefPathHash` of an `ExpnId` to the position
@@ -305,7 +305,7 @@ impl OnDiskCache {
             let mut expn_data = UnhashMap::default();
             let mut foreign_expn_data = UnhashMap::default();
 
-            // Encode all hygiene data (`SyntaxContextData` and `ExpnData`) from the current
+            // Encode all hygiene data (`SyntaxContextKey` and `ExpnData`) from the current
             // session.
 
             hygiene_encode_context.encode(
@@ -566,7 +566,7 @@ impl<'a, 'tcx> SpanDecoder for CacheDecoder<'a, 'tcx> {
             // We look up the position of the associated `SyntaxData` and decode it.
             let pos = syntax_contexts.get(&id).unwrap();
             this.with_position(pos.to_usize(), |decoder| {
-                let data: SyntaxContextData = decode_tagged(decoder, TAG_SYNTAX_CONTEXT);
+                let data: SyntaxContextKey = decode_tagged(decoder, TAG_SYNTAX_CONTEXT);
                 data
             })
         })

compiler/rustc_middle/src/ty/parameterized.rs

@@ -111,7 +111,7 @@ trivially_parameterized_over_tcx! {
     rustc_span::Span,
     rustc_span::Symbol,
     rustc_span::def_id::DefPathHash,
-    rustc_span::hygiene::SyntaxContextData,
+    rustc_span::hygiene::SyntaxContextKey,
     rustc_span::Ident,
     rustc_type_ir::Variance,
     rustc_hir::Attribute,

compiler/rustc_span/src/hygiene.rs

@@ -24,17 +24,14 @@
 // because getting it wrong can lead to nested `HygieneData::with` calls that
 // trigger runtime aborts. (Fortunately these are obvious and easy to fix.)
 
-use std::cell::RefCell;
-use std::collections::hash_map::Entry;
-use std::collections::hash_set::Entry as SetEntry;
 use std::hash::Hash;
 use std::sync::Arc;
 use std::{fmt, iter, mem};
 
 use rustc_data_structures::fingerprint::Fingerprint;
 use rustc_data_structures::fx::{FxHashMap, FxHashSet};
 use rustc_data_structures::stable_hasher::{HashStable, HashingControls, StableHasher};
-use rustc_data_structures::sync::{Lock, WorkerLocal};
+use rustc_data_structures::sync::Lock;
 use rustc_data_structures::unhash::UnhashMap;
 use rustc_hashes::Hash64;
 use rustc_index::IndexVec;
@@ -59,10 +56,10 @@ impl !PartialOrd for SyntaxContext {}
 
 /// If this part of two syntax contexts is equal, then the whole syntax contexts should be equal.
 /// The other fields are only for caching.
-type SyntaxContextKey = (SyntaxContext, ExpnId, Transparency);
+pub type SyntaxContextKey = (SyntaxContext, ExpnId, Transparency);
 
 #[derive(Clone, Copy, PartialEq, Debug, Encodable, Decodable)]
-pub struct SyntaxContextData {
+struct SyntaxContextData {
     outer_expn: ExpnId,
     outer_transparency: Transparency,
     parent: SyntaxContext,
@@ -1266,7 +1263,7 @@ impl HygieneEncodeContext {
     pub fn encode<T>(
         &self,
         encoder: &mut T,
-        mut encode_ctxt: impl FnMut(&mut T, u32, &SyntaxContextData),
+        mut encode_ctxt: impl FnMut(&mut T, u32, &SyntaxContextKey),
         mut encode_expn: impl FnMut(&mut T, ExpnId, &ExpnData, ExpnHash),
     ) {
         // When we serialize a `SyntaxContextData`, we may end up serializing
@@ -1315,18 +1312,12 @@ struct HygieneDecodeContextInner {
     // so that multiple occurrences of the same serialized id are decoded to the same
     // `SyntaxContext`. This only stores `SyntaxContext`s which are completely decoded.
     remapped_ctxts: Vec<Option<SyntaxContext>>,
-
-    /// Maps serialized `SyntaxContext` ids that are currently being decoded to a `SyntaxContext`.
-    decoding: FxHashMap<u32, SyntaxContext>,
 }
 
 #[derive(Default)]
 /// Additional information used to assist in decoding hygiene data
 pub struct HygieneDecodeContext {
     inner: Lock<HygieneDecodeContextInner>,
-
-    /// A set of serialized `SyntaxContext` ids that are currently being decoded on each thread.
-    local_in_progress: WorkerLocal<RefCell<FxHashSet<u32>>>,
 }
 
 /// Register an expansion which has been decoded from the on-disk-cache for the local crate.
@@ -1397,10 +1388,10 @@ pub fn decode_expn_id(
 // to track which `SyntaxContext`s we have already decoded.
 // The provided closure will be invoked to deserialize a `SyntaxContextData`
 // if we haven't already seen the id of the `SyntaxContext` we are deserializing.
-pub fn decode_syntax_context<D: Decoder, F: FnOnce(&mut D, u32) -> SyntaxContextData>(
+pub fn decode_syntax_context<D: Decoder, F: FnOnce(&mut D, u32) -> SyntaxContextKey>(
     d: &mut D,
     context: &HygieneDecodeContext,
-    decode_data: F,
+    decode_ctxt_key: F,
 ) -> SyntaxContext {
     let raw_id: u32 = Decodable::decode(d);
     if raw_id == 0 {
@@ -1409,115 +1400,39 @@ pub fn decode_syntax_context<D: Decoder, F: FnOnce(&mut D, u32) -> SyntaxContext
         return SyntaxContext::root();
     }
 
-    let pending_ctxt = {
-        let mut inner = context.inner.lock();
-
-        // Reminder: `HygieneDecodeContext` is per-crate, so there are no collisions between
-        // raw ids from different crate metadatas.
-        if let Some(ctxt) = inner.remapped_ctxts.get(raw_id as usize).copied().flatten() {
-            // This has already been decoded.
-            return ctxt;
-        }
-
-        match inner.decoding.entry(raw_id) {
-            Entry::Occupied(ctxt_entry) => {
-                let pending_ctxt = *ctxt_entry.get();
-                match context.local_in_progress.borrow_mut().entry(raw_id) {
-                    // We're decoding this already on the current thread. Return here and let the
-                    // function higher up the stack finish decoding to handle recursive cases.
-                    // Hopefully having a `SyntaxContext` that refers to an incorrect data is ok
-                    // during reminder of the decoding process, it's certainly not ok after the
-                    // top level decoding function returns.
-                    SetEntry::Occupied(..) => return pending_ctxt,
-                    // Some other thread is currently decoding this.
-                    // Race with it (alternatively we could wait here).
-                    // We cannot return this value, unlike in the recursive case above, because it
-                    // may expose a `SyntaxContext` pointing to incorrect data to arbitrary code.
-                    SetEntry::Vacant(entry) => {
-                        entry.insert();
-                        pending_ctxt
-                    }
-                }
-            }
-            Entry::Vacant(entry) => {
-                // We are the first thread to start decoding. Mark the current thread as being
-                // progress.
-                context.local_in_progress.borrow_mut().insert(raw_id);
-
-                // Allocate and store SyntaxContext id *before* calling the decoder function,
-                // as the SyntaxContextData may reference itself.
-                let new_ctxt = HygieneData::with(|hygiene_data| {
-                    // Push a dummy SyntaxContextData to ensure that nobody else can get the
-                    // same ID as us. This will be overwritten after call `decode_data`.
-                    hygiene_data.syntax_context_data.push(SyntaxContextData::decode_placeholder());
-                    SyntaxContext::from_usize(hygiene_data.syntax_context_data.len() - 1)
-                });
-                entry.insert(new_ctxt);
-                new_ctxt
-            }
-        }
-    };
+    // Reminder: `HygieneDecodeContext` is per-crate, so there are no collisions between
+    // raw ids from different crate metadatas.
+    if let Some(ctxt) = context.inner.lock().remapped_ctxts.get(raw_id as usize).copied().flatten()
+    {
+        // This has already been decoded.
+        return ctxt;
+    }
 
     // Don't try to decode data while holding the lock, since we need to
     // be able to recursively decode a SyntaxContext
-    let ctxt_data = decode_data(d, raw_id);
-    let ctxt_key = ctxt_data.key();
-
-    let ctxt = HygieneData::with(|hygiene_data| {
-        match hygiene_data.syntax_context_map.get(&ctxt_key) {
-            // Ensure that syntax contexts are unique.
-            // If syntax contexts with the given key already exists, reuse it instead of
-            // using `pending_ctxt`.
-            // `pending_ctxt` will leave an unused hole in the vector of syntax contexts.
-            // Hopefully its value isn't stored anywhere during decoding and its dummy data
-            // is never accessed later. The `is_decode_placeholder` asserts on all
-            // accesses to syntax context data attempt to ensure it.
-            Some(&ctxt) => ctxt,
-            // This is a completely new context.
-            // Overwrite its placeholder data with our decoded data.
-            None => {
-                let ctxt_data_ref =
-                    &mut hygiene_data.syntax_context_data[pending_ctxt.as_u32() as usize];
-                let prev_ctxt_data = mem::replace(ctxt_data_ref, ctxt_data);
-                // Reset `dollar_crate_name` so that it will be updated by `update_dollar_crate_names`.
-                // We don't care what the encoding crate set this to - we want to resolve it
-                // from the perspective of the current compilation session.
-                ctxt_data_ref.dollar_crate_name = kw::DollarCrate;
-                // Make sure nothing weird happened while `decode_data` was running.
-                if !prev_ctxt_data.is_decode_placeholder() {
-                    // Another thread may have already inserted the decoded data,
-                    // but the decoded data should match.
-                    assert_eq!(prev_ctxt_data, *ctxt_data_ref);
-                }
-                hygiene_data.syntax_context_map.insert(ctxt_key, pending_ctxt);
-                pending_ctxt
-            }
-        }
-    });
-
-    // Mark the context as completed
-    context.local_in_progress.borrow_mut().remove(&raw_id);
+    let (parent, expn_id, transparency) = decode_ctxt_key(d, raw_id);
+    let ctxt =
+        HygieneData::with(|hygiene_data| hygiene_data.alloc_ctxt(parent, expn_id, transparency));
 
     let mut inner = context.inner.lock();
     let new_len = raw_id as usize + 1;
     if inner.remapped_ctxts.len() < new_len {
         inner.remapped_ctxts.resize(new_len, None);
     }
     inner.remapped_ctxts[raw_id as usize] = Some(ctxt);
-    inner.decoding.remove(&raw_id);
 
     ctxt
 }
 
-fn for_all_ctxts_in<F: FnMut(u32, SyntaxContext, &SyntaxContextData)>(
+fn for_all_ctxts_in<F: FnMut(u32, SyntaxContext, &SyntaxContextKey)>(
     ctxts: impl Iterator<Item = SyntaxContext>,
     mut f: F,
 ) {
     let all_data: Vec<_> = HygieneData::with(|data| {
         ctxts.map(|ctxt| (ctxt, data.syntax_context_data[ctxt.0 as usize].clone())).collect()
     });
     for (ctxt, data) in all_data.into_iter() {
-        f(ctxt.0, ctxt, &data);
+        f(ctxt.0, ctxt, &data.key());
     }
 }