Merge pull request #313 from Amanieu/hle_feature

.github/workflows/rust.yml

@@ -25,6 +25,9 @@ jobs:
           - channel: nightly
             feature: nightly
             os: ubuntu
+          - channel: nightly
+            feature: hardware-lock-elision
+            os: ubuntu
 
     steps:
     - uses: actions/checkout@v2

Cargo.toml

@@ -28,6 +28,7 @@ nightly = ["parking_lot_core/nightly", "lock_api/nightly"]
 deadlock_detection = ["parking_lot_core/deadlock_detection"]
 serde = ["lock_api/serde"]
 send_guard = []
+hardware-lock-elision = []
 
 [workspace]
 exclude = ["benchmark"]

README.md

@@ -50,6 +50,7 @@ in the Rust standard library:
    library versions of those types.
 7. `RwLock` takes advantage of hardware lock elision on processors that
    support it, which can lead to huge performance wins with many readers.
+   This must be enabled with the `hardware-lock-elision` feature.
 8. `RwLock` uses a task-fair locking policy, which avoids reader and writer
    starvation, whereas the standard library version makes no guarantees.
 9. `Condvar` is guaranteed not to produce spurious wakeups. A thread will
@@ -93,8 +94,6 @@ There are a few restrictions when using this library on stable Rust:
 - You will have to use the `const_*` functions (e.g. `const_mutex(val)`) to
   statically initialize the locking primitives. Using e.g. `Mutex::new(val)`
   does not work on stable Rust yet.
-- `RwLock` will not be able to take advantage of hardware lock elision for
-  readers, which improves performance when there are multiple readers.
 - The `wasm32-unknown-unknown` target is only supported on nightly and requires
   `-C target-feature=+atomics` in `RUSTFLAGS`.
 
@@ -126,6 +125,10 @@ To allow sending `MutexGuard`s and `RwLock*Guard`s to other threads, enable the
 Note that the `deadlock_detection` and `send_guard` features are incompatible
 and cannot be used together.
 
+Hardware lock elision support for x86 can be enabled with the
+`hardware-lock-elision` feature. This requires Rust 1.59 due to the use of
+inline assembly.
+
 The core parking lot API is provided by the `parking_lot_core` crate. It is
 separate from the synchronization primitives in the `parking_lot` crate so that
 changes to the core API do not cause breaking changes for users of `parking_lot`.

src/elision.rs

@@ -28,14 +28,14 @@ pub trait AtomicElisionExt {
 #[inline]
 pub fn have_elision() -> bool {
     cfg!(all(
-        feature = "nightly",
+        feature = "hardware-lock-elision",
         any(target_arch = "x86", target_arch = "x86_64"),
     ))
 }
 
 // This implementation is never actually called because it is guarded by
 // have_elision().
-#[cfg(not(all(feature = "nightly", any(target_arch = "x86", target_arch = "x86_64"))))]
+#[cfg(not(all(feature = "hardware-lock-elision", any(target_arch = "x86", target_arch = "x86_64"))))]
 impl AtomicElisionExt for AtomicUsize {
     type IntType = usize;
 
@@ -50,13 +50,14 @@ impl AtomicElisionExt for AtomicUsize {
     }
 }
 
-#[cfg(all(feature = "nightly", any(target_arch = "x86", target_arch = "x86_64")))]
+#[cfg(all(feature = "hardware-lock-elision", any(target_arch = "x86", target_arch = "x86_64")))]
 impl AtomicElisionExt for AtomicUsize {
     type IntType = usize;
 
     #[inline]
     fn elision_compare_exchange_acquire(&self, current: usize, new: usize) -> Result<usize, usize> {
         unsafe {
+            use core::arch::asm;
             let prev: usize;
             #[cfg(target_pointer_width = "32")]
             asm!(
@@ -87,6 +88,7 @@ impl AtomicElisionExt for AtomicUsize {
     #[inline]
     fn elision_fetch_sub_release(&self, val: usize) -> usize {
         unsafe {
+            use core::arch::asm;
             let prev: usize;
             #[cfg(target_pointer_width = "32")]
             asm!(