servo: Merge #15535 - geckolib: use a global thread pool for styling (from froydnj:global-style-thread-pool); r=bholley

By having a single thread pool, rather than one per document, we use less memory.  This addresses https://bugzilla.mozilla.org/show_bug.cgi?id=1324250.

This may be obvious to an experienced Rust programmer, but I went with raw pointers because trying to use `Option` global variables resulted in complaints about turning on feature flags in nightly Rust.  Since this is for stylo, nightly features are not appropriate here.

---
<!-- Thank you for contributing to Servo! Please replace each `[ ]` by `[X]` when the step is complete, and replace `__` with appropriate data: -->
- [X] `./mach build -d` does not report any errors
- [X] `./mach test-tidy` does not report any errors
- [ ] These changes fix #__ (github issue number if applicable).

<!-- Either: -->
- [ ] There are tests for these changes OR
- [ ] These changes do not require tests because _____

<!-- Pull requests that do not address these steps are welcome, but they will require additional verification as part of the review process. -->

Source-Repo: https://github.com/servo/servo
Source-Revision: 56a99577b31a942e340624f97377980b0e612088

--HG--
extra : subtree_source : https%3A//hg.mozilla.org/projects/converted-servo-linear
extra : subtree_revision : 72e6b32e7662f3801e79c3894d89a0f9d25a861f

servo/Cargo.lock

@@ -906,6 +906,7 @@ dependencies = [
  "log 0.3.6 (registry+https://github.com/rust-lang/crates.io-index)",
  "num_cpus 1.2.1 (registry+https://github.com/rust-lang/crates.io-index)",
  "parking_lot 0.3.8 (registry+https://github.com/rust-lang/crates.io-index)",
+ "rayon 0.6.0 (registry+https://github.com/rust-lang/crates.io-index)",
  "selectors 0.18.0",
  "servo_url 0.0.1",
  "style 0.0.1",
@@ -2721,7 +2722,6 @@ dependencies = [
  "nsstring_vendor 0.1.0",
  "num-integer 0.1.32 (registry+https://github.com/rust-lang/crates.io-index)",
  "num-traits 0.1.36 (registry+https://github.com/rust-lang/crates.io-index)",
- "num_cpus 1.2.1 (registry+https://github.com/rust-lang/crates.io-index)",
  "ordered-float 0.4.0 (registry+https://github.com/rust-lang/crates.io-index)",
  "owning_ref 0.2.4 (registry+https://github.com/rust-lang/crates.io-index)",
  "parking_lot 0.3.8 (registry+https://github.com/rust-lang/crates.io-index)",
@@ -2794,6 +2794,7 @@ dependencies = [
  "log 0.3.6 (registry+https://github.com/rust-lang/crates.io-index)",
  "num_cpus 1.2.1 (registry+https://github.com/rust-lang/crates.io-index)",
  "parking_lot 0.3.8 (registry+https://github.com/rust-lang/crates.io-index)",
+ "rayon 0.6.0 (registry+https://github.com/rust-lang/crates.io-index)",
  "selectors 0.18.0",
  "servo_url 0.0.1",
  "style 0.0.1",

servo/components/style/Cargo.toml

@@ -13,7 +13,7 @@ path = "lib.rs"
 doctest = false
 
 [features]
-gecko = ["nsstring_vendor", "num_cpus", "rayon/unstable"]
+gecko = ["nsstring_vendor", "rayon/unstable"]
 use_bindgen = ["bindgen", "regex"]
 servo = ["serde/unstable", "serde", "serde_derive", "heapsize_derive",
          "style_traits/servo", "servo_atoms", "html5ever-atoms",
@@ -57,10 +57,6 @@ servo_url = {path = "../url"}
 time = "0.1"
 unicode-segmentation = "1.0"
 
-[dependencies.num_cpus]
-optional = true
-version = "1.0"
-
 [target.'cfg(windows)'.dependencies]
 kernel32-sys = "0.2"
 

servo/components/style/gecko/data.rs

@@ -11,13 +11,9 @@ use gecko_bindings::bindings::RawServoStyleSet;
 use gecko_bindings::structs::RawGeckoPresContextOwned;
 use gecko_bindings::sugar::ownership::{HasBoxFFI, HasFFI, HasSimpleFFI};
 use media_queries::Device;
-use num_cpus;
 use parking_lot::RwLock;
 use properties::ComputedValues;
-use rayon;
-use std::cmp;
 use std::collections::HashMap;
-use std::env;
 use std::sync::Arc;
 use std::sync::mpsc::{Receiver, Sender, channel};
 use stylesheets::Stylesheet;
@@ -48,29 +44,12 @@ pub struct PerDocumentStyleDataImpl {
     /// Unused. Will go away when we actually implement transitions and
     /// animations properly.
     pub expired_animations: Arc<RwLock<HashMap<OpaqueNode, Vec<Animation>>>>,
-
-    /// The worker thread pool.
-    /// FIXME(bholley): This shouldn't be per-document.
-    pub work_queue: Option<rayon::ThreadPool>,
-
-    /// The number of threads of the work queue.
-    pub num_threads: usize,
 }
 
 /// The data itself is an `AtomicRefCell`, which guarantees the proper semantics
 /// and unexpected races while trying to mutate it.
 pub struct PerDocumentStyleData(AtomicRefCell<PerDocumentStyleDataImpl>);
 
-lazy_static! {
-    /// The number of layout threads, computed statically.
-    pub static ref NUM_THREADS: usize = {
-        match env::var("STYLO_THREADS").map(|s| s.parse::<usize>().expect("invalid STYLO_THREADS")) {
-            Ok(num) => num,
-            _ => cmp::max(num_cpus::get() * 3 / 4, 1),
-        }
-    };
-}
-
 impl PerDocumentStyleData {
     /// Create a dummy `PerDocumentStyleData`.
     pub fn new(pres_context: RawGeckoPresContextOwned) -> Self {
@@ -86,14 +65,6 @@ impl PerDocumentStyleData {
             new_animations_receiver: new_anims_receiver,
             running_animations: Arc::new(RwLock::new(HashMap::new())),
             expired_animations: Arc::new(RwLock::new(HashMap::new())),
-            work_queue: if *NUM_THREADS <= 1 {
-                None
-            } else {
-                let configuration =
-                    rayon::Configuration::new().set_num_threads(*NUM_THREADS);
-                rayon::ThreadPool::new(configuration).ok()
-            },
-            num_threads: *NUM_THREADS,
         }))
     }
 
@@ -141,9 +112,3 @@ unsafe impl HasFFI for PerDocumentStyleData {
 }
 unsafe impl HasSimpleFFI for PerDocumentStyleData {}
 unsafe impl HasBoxFFI for PerDocumentStyleData {}
-
-impl Drop for PerDocumentStyleDataImpl {
-    fn drop(&mut self) {
-        let _ = self.work_queue.take();
-    }
-}

servo/components/style/lib.rs

@@ -61,7 +61,6 @@ extern crate matches;
 #[cfg(feature = "gecko")] extern crate nsstring_vendor as nsstring;
 extern crate num_integer;
 extern crate num_traits;
-#[cfg(feature = "gecko")] extern crate num_cpus;
 extern crate ordered_float;
 extern crate owning_ref;
 extern crate parking_lot;

servo/ports/geckolib/Cargo.toml

@@ -23,6 +23,7 @@ libc = "0.2"
 log = {version = "0.3.5", features = ["release_max_level_info"]}
 num_cpus = "1.1.0"
 parking_lot = "0.3"
+rayon = "0.6"
 selectors = {path = "../../components/selectors"}
 servo_url = {path = "../../components/url"}
 style = {path = "../../components/style", features = ["gecko"]}

servo/ports/geckolib/glue.rs

@@ -8,10 +8,13 @@ use cssparser::Parser;
 use cssparser::ToCss as ParserToCss;
 use env_logger::LogBuilder;
 use euclid::Size2D;
+use num_cpus;
 use parking_lot::RwLock;
+use rayon;
 use selectors::Element;
 use servo_url::ServoUrl;
 use std::borrow::Cow;
+use std::cmp;
 use std::env;
 use std::fmt::Write;
 use std::mem;
@@ -23,7 +26,7 @@ use style::context::{ThreadLocalStyleContext, ThreadLocalStyleContextCreationInf
 use style::data::{ElementData, ElementStyles, RestyleData};
 use style::dom::{ShowSubtreeData, TElement, TNode};
 use style::error_reporting::StdoutErrorReporter;
-use style::gecko::data::{NUM_THREADS, PerDocumentStyleData, PerDocumentStyleDataImpl};
+use style::gecko::data::{PerDocumentStyleData, PerDocumentStyleDataImpl};
 use style::gecko::restyle_damage::GeckoRestyleDamage;
 use style::gecko::selector_parser::{SelectorImpl, PseudoElement};
 use style::gecko::traversal::RecalcStyleOnly;
@@ -90,6 +93,45 @@ use stylesheet_loader::StylesheetLoader;
  * depend on but good enough for our purposes.
  */
 
+struct GlobalStyleData {
+    // How many threads parallel styling can use.
+    pub num_threads: usize,
+
+    // The parallel styling thread pool.
+    pub style_thread_pool: Option<rayon::ThreadPool>,
+}
+
+impl GlobalStyleData {
+    pub fn new() -> Self {
+        let stylo_threads = env::var("STYLO_THREADS")
+            .map(|s| s.parse::<usize>().expect("invalid STYLO_THREADS value"));
+        let num_threads = match stylo_threads {
+            Ok(num) => num,
+            _ => cmp::max(num_cpus::get() * 3 / 4, 1),
+        };
+
+        let pool = if num_threads <= 1 {
+            None
+        } else {
+            let configuration =
+                rayon::Configuration::new().set_num_threads(num_threads);
+            let pool = rayon::ThreadPool::new(configuration).ok();
+            pool
+        };
+
+        GlobalStyleData {
+            num_threads: num_threads,
+            style_thread_pool: pool,
+        }
+    }
+}
+
+lazy_static! {
+    static ref GLOBAL_STYLE_DATA: GlobalStyleData = {
+        GlobalStyleData::new()
+    };
+}
+
 #[no_mangle]
 pub extern "C" fn Servo_Initialize() -> () {
     // Initialize logging.
@@ -148,7 +190,7 @@ fn traverse_subtree(element: GeckoElement, raw_data: RawServoStyleSetBorrowed,
         }
     }
 
-    let mut per_doc_data = PerDocumentStyleData::from_ffi(raw_data).borrow_mut();
+    let per_doc_data = PerDocumentStyleData::from_ffi(raw_data).borrow();
 
     let token = RecalcStyleOnly::pre_traverse(element, &per_doc_data.stylist, unstyled_children_only);
     if !token.should_traverse() {
@@ -159,7 +201,9 @@ fn traverse_subtree(element: GeckoElement, raw_data: RawServoStyleSetBorrowed,
     debug!("{:?}", ShowSubtreeData(element.as_node()));
 
     let shared_style_context = create_shared_context(&per_doc_data);
-    let traversal_driver = if per_doc_data.num_threads == 1 || per_doc_data.work_queue.is_none() {
+    let ref global_style_data = *GLOBAL_STYLE_DATA;
+
+    let traversal_driver = if global_style_data.style_thread_pool.is_none() {
         TraversalDriver::Sequential
     } else {
         TraversalDriver::Parallel
@@ -169,7 +213,7 @@ fn traverse_subtree(element: GeckoElement, raw_data: RawServoStyleSetBorrowed,
     let known_depth = None;
     if traversal_driver.is_parallel() {
         parallel::traverse_dom(&traversal, element, known_depth, token,
-                               per_doc_data.work_queue.as_mut().unwrap());
+                               global_style_data.style_thread_pool.as_ref().unwrap());
     } else {
         sequential::traverse_dom(&traversal, element, token);
     }
@@ -334,7 +378,7 @@ pub extern "C" fn Servo_AnimationValues_Populate(anim: RawGeckoAnimationValueLis
 
 #[no_mangle]
 pub extern "C" fn Servo_StyleWorkerThreadCount() -> u32 {
-    *NUM_THREADS as u32
+    GLOBAL_STYLE_DATA.num_threads as u32
 }
 
 #[no_mangle]

servo/ports/geckolib/lib.rs

@@ -9,9 +9,12 @@ extern crate atomic_refcell;
 extern crate cssparser;
 extern crate env_logger;
 extern crate euclid;
+#[macro_use] extern crate lazy_static;
 extern crate libc;
 #[macro_use] extern crate log;
+extern crate num_cpus;
 extern crate parking_lot;
+extern crate rayon;
 extern crate selectors;
 extern crate servo_url;
 extern crate style;

servo/tests/unit/stylo/Cargo.toml

@@ -22,6 +22,7 @@ libc = "0.2"
 log = {version = "0.3.5", features = ["release_max_level_info"]}
 num_cpus = "1.1.0"
 parking_lot = "0.3"
+rayon = "0.6"
 selectors = {path = "../../../components/selectors"}
 servo_url = {path = "../../../components/url"}
 style_traits = {path = "../../../components/style_traits"}

servo/tests/unit/stylo/lib.rs

@@ -8,8 +8,11 @@ extern crate cssparser;
 extern crate env_logger;
 extern crate euclid;
 extern crate geckoservo;
+#[macro_use] extern crate lazy_static;
 #[macro_use] extern crate log;
+extern crate num_cpus;
 extern crate parking_lot;
+extern crate rayon;
 extern crate selectors;
 extern crate servo_url;
 extern crate style;