bjoernager/rust
bjoernager/rust
1
Fork 0
Code Activity

Rollup merge of #37229 - nnethercote:FxHasher, r=nikomatsakis

Browse source 
Replace FNV with a faster hash function.

Hash table lookups are very hot in rustc profiles and the time taken within `FnvHash` itself is a big part of that. Although FNV is a simple hash, it processes its input one byte at a time. In contrast, Firefox has a homespun hash function that is also simple but works on multiple bytes at a time. So I tried it out and the results are compelling:

```
futures-rs-test  4.326s vs  4.212s --> 1.027x faster (variance: 1.001x, 1.007x)
helloworld       0.233s vs  0.232s --> 1.004x faster (variance: 1.037x, 1.016x)
html5ever-2016-  5.397s vs  5.210s --> 1.036x faster (variance: 1.009x, 1.006x)
hyper.0.5.0      5.018s vs  4.905s --> 1.023x faster (variance: 1.007x, 1.006x)
inflate-0.1.0    4.889s vs  4.872s --> 1.004x faster (variance: 1.012x, 1.007x)
issue-32062-equ  0.347s vs  0.335s --> 1.035x faster (variance: 1.033x, 1.019x)
issue-32278-big  1.717s vs  1.622s --> 1.059x faster (variance: 1.027x, 1.028x)
jld-day15-parse  1.537s vs  1.459s --> 1.054x faster (variance: 1.005x, 1.003x)
piston-image-0. 11.863s vs 11.482s --> 1.033x faster (variance: 1.060x, 1.002x)
regex.0.1.30     2.517s vs  2.453s --> 1.026x faster (variance: 1.011x, 1.013x)
rust-encoding-0  2.080s vs  2.047s --> 1.016x faster (variance: 1.005x, 1.005x)
syntex-0.42.2   32.268s vs 31.275s --> 1.032x faster (variance: 1.014x, 1.022x)
syntex-0.42.2-i 17.629s vs 16.559s --> 1.065x faster (variance: 1.013x, 1.021x)
```

(That's a stage1 compiler doing debug builds. Results for a stage2 compiler are similar.)

The attached commit is not in a state suitable for landing because I changed the implementation of FnvHasher without changing its name (because that would have required touching many lines in the compiler). Nonetheless, it is a good place to start discussions.

Profiles show very clearly that this new hash function is a lot faster to compute than FNV. The quality of the new hash function is less clear -- it seems to do better in some cases and worse in others (judging by the number of instructions executed in `Hash{Map,Set}::get`).

CC @brson, @arthurprs
This commit is contained in:
Eduard-Mihai Burtescu 2016-11-09 20:51:15 +02:00 • committed by GitHub
commit dc8ac2679a
92 changed files with 703 additions and 588 deletions
Download patch file Download diff file Expand all files Collapse all files

56
src/librustdoc/html/render.rs
View file

@ -59,7 +59,7 @@ use rustc::hir::def_id::{CrateNum, CRATE_DEF_INDEX, DefId, LOCAL_CRATE};
use rustc::middle::privacy::AccessLevels;
use rustc::middle::stability;
use rustc::hir;
use rustc::util::nodemap::{FnvHashMap, FnvHashSet};
use rustc::util::nodemap::{FxHashMap, FxHashSet};
use rustc_data_structures::flock;
use clean::{self, Attributes, GetDefId, SelfTy, Mutability};
@ -111,9 +111,9 @@ pub struct SharedContext {
/// `true`.
pub include_sources: bool,
/// The local file sources we've emitted and their respective url-paths.
pub local_sources: FnvHashMap<PathBuf, String>,
pub local_sources: FxHashMap<PathBuf, String>,
/// All the passes that were run on this crate.
pub passes: FnvHashSet<String>,
pub passes: FxHashSet<String>,
/// The base-URL of the issue tracker for when an item has been tagged with
/// an issue number.
pub issue_tracker_base_url: Option<String>,
@ -208,7 +208,7 @@ pub struct Cache {
/// Mapping of typaram ids to the name of the type parameter. This is used
/// when pretty-printing a type (so pretty printing doesn't have to
/// painfully maintain a context like this)
pub typarams: FnvHashMap<DefId, String>,
pub typarams: FxHashMap<DefId, String>,
/// Maps a type id to all known implementations for that type. This is only
/// recognized for intra-crate `ResolvedPath` types, and is used to print
@ -216,35 +216,35 @@ pub struct Cache {
///
/// The values of the map are a list of implementations and documentation
/// found on that implementation.
pub impls: FnvHashMap<DefId, Vec<Impl>>,
pub impls: FxHashMap<DefId, Vec<Impl>>,
/// Maintains a mapping of local crate node ids to the fully qualified name
/// and "short type description" of that node. This is used when generating
/// URLs when a type is being linked to. External paths are not located in
/// this map because the `External` type itself has all the information
/// necessary.
pub paths: FnvHashMap<DefId, (Vec<String>, ItemType)>,
pub paths: FxHashMap<DefId, (Vec<String>, ItemType)>,
/// Similar to `paths`, but only holds external paths. This is only used for
/// generating explicit hyperlinks to other crates.
pub external_paths: FnvHashMap<DefId, (Vec<String>, ItemType)>,
pub external_paths: FxHashMap<DefId, (Vec<String>, ItemType)>,
/// This map contains information about all known traits of this crate.
/// Implementations of a crate should inherit the documentation of the
/// parent trait if no extra documentation is specified, and default methods
/// should show up in documentation about trait implementations.
pub traits: FnvHashMap<DefId, clean::Trait>,
pub traits: FxHashMap<DefId, clean::Trait>,
/// When rendering traits, it's often useful to be able to list all
/// implementors of the trait, and this mapping is exactly, that: a mapping
/// of trait ids to the list of known implementors of the trait
pub implementors: FnvHashMap<DefId, Vec<Implementor>>,
pub implementors: FxHashMap<DefId, Vec<Implementor>>,
/// Cache of where external crate documentation can be found.
pub extern_locations: FnvHashMap<CrateNum, (String, ExternalLocation)>,
pub extern_locations: FxHashMap<CrateNum, (String, ExternalLocation)>,
/// Cache of where documentation for primitives can be found.
pub primitive_locations: FnvHashMap<clean::PrimitiveType, CrateNum>,
pub primitive_locations: FxHashMap<clean::PrimitiveType, CrateNum>,
// Note that external items for which `doc(hidden)` applies to are shown as
// non-reachable while local items aren't. This is because we're reusing
@ -257,7 +257,7 @@ pub struct Cache {
parent_stack: Vec<DefId>,
parent_is_trait_impl: bool,
search_index: Vec<IndexItem>,
seen_modules: FnvHashSet<DefId>,
seen_modules: FxHashSet<DefId>,
seen_mod: bool,
stripped_mod: bool,
deref_trait_did: Option<DefId>,
@ -275,9 +275,9 @@ pub struct Cache {
/// Later on moved into `CACHE_KEY`.
#[derive(Default)]
pub struct RenderInfo {
pub inlined: FnvHashSet<DefId>,
pub inlined: FxHashSet<DefId>,
pub external_paths: ::core::ExternalPaths,
pub external_typarams: FnvHashMap<DefId, String>,
pub external_typarams: FxHashMap<DefId, String>,
pub deref_trait_did: Option<DefId>,
pub deref_mut_trait_did: Option<DefId>,
}
@ -376,10 +376,10 @@ impl ToJson for IndexItemFunctionType {
thread_local!(static CACHE_KEY: RefCell<Arc<Cache>> = Default::default());
thread_local!(pub static CURRENT_LOCATION_KEY: RefCell<Vec<String>> =
RefCell::new(Vec::new()));
thread_local!(static USED_ID_MAP: RefCell<FnvHashMap<String, usize>> =
thread_local!(static USED_ID_MAP: RefCell<FxHashMap<String, usize>> =
RefCell::new(init_ids()));
fn init_ids() -> FnvHashMap<String, usize> {
fn init_ids() -> FxHashMap<String, usize> {
[
"main",
"search",
@ -406,7 +406,7 @@ pub fn reset_ids(embedded: bool) {
*s.borrow_mut() = if embedded {
init_ids()
} else {
FnvHashMap()
FxHashMap()
};
});
}
@ -431,7 +431,7 @@ pub fn derive_id(candidate: String) -> String {
pub fn run(mut krate: clean::Crate,
external_html: &ExternalHtml,
dst: PathBuf,
passes: FnvHashSet<String>,
passes: FxHashSet<String>,
css_file_extension: Option<PathBuf>,
renderinfo: RenderInfo) -> Result<(), Error> {
let src_root = match krate.src.parent() {
@ -442,7 +442,7 @@ pub fn run(mut krate: clean::Crate,
src_root: src_root,
passes: passes,
include_sources: true,
local_sources: FnvHashMap(),
local_sources: FxHashMap(),
issue_tracker_base_url: None,
layout: layout::Layout {
logo: "".to_string(),
@ -510,22 +510,22 @@ pub fn run(mut krate: clean::Crate,
.collect();
let mut cache = Cache {
impls: FnvHashMap(),
impls: FxHashMap(),
external_paths: external_paths,
paths: FnvHashMap(),
implementors: FnvHashMap(),
paths: FxHashMap(),
implementors: FxHashMap(),
stack: Vec::new(),
parent_stack: Vec::new(),
search_index: Vec::new(),
parent_is_trait_impl: false,
extern_locations: FnvHashMap(),
primitive_locations: FnvHashMap(),
seen_modules: FnvHashSet(),
extern_locations: FxHashMap(),
primitive_locations: FxHashMap(),
seen_modules: FxHashSet(),
seen_mod: false,
stripped_mod: false,
access_levels: krate.access_levels.clone(),
orphan_impl_items: Vec::new(),
traits: mem::replace(&mut krate.external_traits, FnvHashMap()),
traits: mem::replace(&mut krate.external_traits, FxHashMap()),
deref_trait_did: deref_trait_did,
deref_mut_trait_did: deref_mut_trait_did,
typarams: external_typarams,
@ -572,7 +572,7 @@ pub fn run(mut krate: clean::Crate,
/// Build the search index from the collected metadata
fn build_index(krate: &clean::Crate, cache: &mut Cache) -> String {
let mut nodeid_to_pathid = FnvHashMap();
let mut nodeid_to_pathid = FxHashMap();
let mut crate_items = Vec::with_capacity(cache.search_index.len());
let mut crate_paths = Vec::<Json>::new();
@ -2655,7 +2655,7 @@ fn render_union(w: &mut fmt::Formatter, it: &clean::Item,
#[derive(Copy, Clone)]
enum AssocItemLink<'a> {
Anchor(Option<&'a str>),
GotoSource(DefId, &'a FnvHashSet<String>),
GotoSource(DefId, &'a FxHashSet<String>),
}
impl<'a> AssocItemLink<'a> {