Use AtomicU64 for head/tail index in deque, channel, and queues

Cargo.toml

@@ -68,7 +68,7 @@ default-features = false
 optional = true
 
 [dependencies.crossbeam-utils]
-version = "0.8.5"
+version = "0.8.6"
 path = "./crossbeam-utils"
 default-features = false
 

ci/no_atomic.sh

@@ -41,6 +41,7 @@ cat >"${file}" <<EOF
 // This file is @generated by $(basename "$0").
 // It is not intended for manual editing.
 
+#[allow(dead_code)] // Only crossbeam-{epoch,queue,skiplist,utils} use this.
 const NO_ATOMIC_CAS: &[&str] = &[
 EOF
 for target in "${no_atomic_cas[@]}"; do
@@ -49,7 +50,7 @@ done
 cat >>"${file}" <<EOF
 ];
 
-#[allow(dead_code)] // Only crossbeam-utils uses this.
+#[allow(dead_code)] // Only crossbeam-{channel,deque,queue,utils} use this.
 const NO_ATOMIC_64: &[&str] = &[
 EOF
 for target in "${no_atomic_64[@]}"; do

crossbeam-channel/Cargo.toml

@@ -27,7 +27,7 @@ std = ["crossbeam-utils/std"]
 cfg-if = "1"
 
 [dependencies.crossbeam-utils]
-version = "0.8"
+version = "0.8.6"
 path = "../crossbeam-utils"
 default-features = false
 optional = true

crossbeam-channel/build.rs

@@ -0,0 +1,43 @@
+// The rustc-cfg listed below are considered public API, but it is *unstable*
+// and outside of the normal semver guarantees:
+//
+// - `crossbeam_no_atomic_64`
+//      Assume the target does *not* support AtomicU64/AtomicI64.
+//      This is usually detected automatically by the build script, but you may
+//      need to enable it manually when building for custom targets or using
+//      non-cargo build systems that don't run the build script.
+//
+// With the exceptions mentioned above, the rustc-cfg emitted by the build
+// script are *not* public API.
+
+#![warn(rust_2018_idioms)]
+
+use std::env;
+
+include!("no_atomic.rs");
+
+fn main() {
+    let target = match env::var("TARGET") {
+        Ok(target) => target,
+        Err(e) => {
+            println!(
+                "cargo:warning={}: unable to get TARGET environment variable: {}",
+                env!("CARGO_PKG_NAME"),
+                e
+            );
+            return;
+        }
+    };
+
+    // Note that this is `no_*`, not `has_*`. This allows treating
+    // "max-atomic-width" as 64 when the build script doesn't run. This is
+    // needed for compatibility with non-cargo build systems that don't run the
+    // build script.
+    if NO_ATOMIC_64.contains(&&*target) {
+        println!("cargo:rustc-cfg=crossbeam_no_atomic_64");
+    } else {
+        // Otherwise, assuming `"max-atomic-width" == 64` or `"max-atomic-width" == 128`.
+    }
+
+    println!("cargo:rerun-if-changed=no_atomic.rs");
+}

crossbeam-channel/no_atomic.rs

@@ -0,0 +1 @@
+../no_atomic.rs

crossbeam-channel/src/flavors/array.rs

@@ -11,20 +11,21 @@
 use std::cell::UnsafeCell;
 use std::mem::MaybeUninit;
 use std::ptr;
-use std::sync::atomic::{self, AtomicUsize, Ordering};
+use std::sync::atomic::{self, Ordering};
 use std::time::Instant;
 
 use crossbeam_utils::{Backoff, CachePadded};
 
 use crate::context::Context;
 use crate::err::{RecvTimeoutError, SendTimeoutError, TryRecvError, TrySendError};
 use crate::select::{Operation, SelectHandle, Selected, Token};
+use crate::utils::AtomicU64;
 use crate::waker::SyncWaker;
 
 /// A slot in a channel.
 struct Slot<T> {
     /// The current stamp.
-    stamp: AtomicUsize,
+    stamp: AtomicU64,
 
     /// The message in this slot.
     msg: UnsafeCell<MaybeUninit<T>>,
@@ -37,7 +38,7 @@ pub(crate) struct ArrayToken {
     slot: *const u8,
 
     /// Stamp to store into the slot after reading or writing.
-    stamp: usize,
+    stamp: u64,
 }
 
 impl Default for ArrayToken {
@@ -55,20 +56,20 @@ pub(crate) struct Channel<T> {
     /// The head of the channel.
     ///
     /// This value is a "stamp" consisting of an index into the buffer, a mark bit, and a lap, but
-    /// packed into a single `usize`. The lower bits represent the index, while the upper bits
+    /// packed into a single `u64`. The lower bits represent the index, while the upper bits
     /// represent the lap. The mark bit in the head is always zero.
     ///
     /// Messages are popped from the head of the channel.
-    head: CachePadded<AtomicUsize>,
+    head: CachePadded<AtomicU64>,
 
     /// The tail of the channel.
     ///
     /// This value is a "stamp" consisting of an index into the buffer, a mark bit, and a lap, but
-    /// packed into a single `usize`. The lower bits represent the index, while the upper bits
+    /// packed into a single `u64`. The lower bits represent the index, while the upper bits
     /// represent the lap. The mark bit indicates that the channel is disconnected.
     ///
     /// Messages are pushed into the tail of the channel.
-    tail: CachePadded<AtomicUsize>,
+    tail: CachePadded<AtomicU64>,
 
     /// The buffer holding slots.
     buffer: Box<[Slot<T>]>,
@@ -77,10 +78,10 @@ pub(crate) struct Channel<T> {
     cap: usize,
 
     /// A stamp with the value of `{ lap: 1, mark: 0, index: 0 }`.
-    one_lap: usize,
+    one_lap: u64,
 
     /// If this bit is set in the tail, that means the channel is disconnected.
-    mark_bit: usize,
+    mark_bit: u64,
 
     /// Senders waiting while the channel is full.
     senders: SyncWaker,
@@ -95,7 +96,7 @@ impl<T> Channel<T> {
         assert!(cap > 0, "capacity must be positive");
 
         // Compute constants `mark_bit` and `one_lap`.
-        let mark_bit = (cap + 1).next_power_of_two();
+        let mark_bit = (cap as u64 + 1).next_power_of_two();
         let one_lap = mark_bit * 2;
 
         // Head is initialized to `{ lap: 0, mark: 0, index: 0 }`.
@@ -105,11 +106,11 @@ impl<T> Channel<T> {
 
         // Allocate a buffer of `cap` slots initialized
         // with stamps.
-        let buffer: Box<[Slot<T>]> = (0..cap)
+        let buffer: Box<[Slot<T>]> = (0..cap as u64)
             .map(|i| {
                 // Set the stamp to `{ lap: 0, mark: 0, index: i }`.
                 Slot {
-                    stamp: AtomicUsize::new(i),
+                    stamp: AtomicU64::new(i),
                     msg: UnsafeCell::new(MaybeUninit::uninit()),
                 }
             })
@@ -120,8 +121,8 @@ impl<T> Channel<T> {
             cap,
             one_lap,
             mark_bit,
-            head: CachePadded::new(AtomicUsize::new(head)),
-            tail: CachePadded::new(AtomicUsize::new(tail)),
+            head: CachePadded::new(AtomicU64::new(head)),
+            tail: CachePadded::new(AtomicU64::new(tail)),
             senders: SyncWaker::new(),
             receivers: SyncWaker::new(),
         }
@@ -151,7 +152,7 @@ impl<T> Channel<T> {
             }
 
             // Deconstruct the tail.
-            let index = tail & (self.mark_bit - 1);
+            let index = (tail & (self.mark_bit - 1)) as usize;
             let lap = tail & !(self.one_lap - 1);
 
             // Inspect the corresponding slot.
@@ -234,7 +235,7 @@ impl<T> Channel<T> {
 
         loop {
             // Deconstruct the head.
-            let index = head & (self.mark_bit - 1);
+            let index = (head & (self.mark_bit - 1)) as usize;
             let lap = head & !(self.one_lap - 1);
 
             // Inspect the corresponding slot.
@@ -452,8 +453,8 @@ impl<T> Channel<T> {
 
             // If the tail didn't change, we've got consistent values to work with.
             if self.tail.load(Ordering::SeqCst) == tail {
-                let hix = head & (self.mark_bit - 1);
-                let tix = tail & (self.mark_bit - 1);
+                let hix = (head & (self.mark_bit - 1)) as usize;
+                let tix = (tail & (self.mark_bit - 1)) as usize;
 
                 return if hix < tix {
                     tix - hix
@@ -524,8 +525,8 @@ impl<T> Drop for Channel<T> {
         let head = *self.head.get_mut();
         let tail = *self.tail.get_mut();
 
-        let hix = head & (self.mark_bit - 1);
-        let tix = tail & (self.mark_bit - 1);
+        let hix = (head & (self.mark_bit - 1)) as usize;
+        let tix = (tail & (self.mark_bit - 1)) as usize;
 
         let len = if hix < tix {
             tix - hix

crossbeam-channel/src/flavors/list.rs

@@ -12,6 +12,7 @@ use crossbeam_utils::{Backoff, CachePadded};
 use crate::context::Context;
 use crate::err::{RecvTimeoutError, SendTimeoutError, TryRecvError, TrySendError};
 use crate::select::{Operation, SelectHandle, Selected, Token};
+use crate::utils::AtomicU64;
 use crate::waker::SyncWaker;
 
 // TODO(stjepang): Once we bump the minimum required Rust version to 1.28 or newer, re-apply the
@@ -29,15 +30,15 @@ const READ: usize = 2;
 const DESTROY: usize = 4;
 
 // Each block covers one "lap" of indices.
-const LAP: usize = 32;
+const LAP: u64 = 32;
 // The maximum number of messages a block can hold.
-const BLOCK_CAP: usize = LAP - 1;
+const BLOCK_CAP: usize = LAP as usize - 1;
 // How many lower bits are reserved for metadata.
-const SHIFT: usize = 1;
+const SHIFT: u64 = 1;
 // Has two different purposes:
 // * If set in head, indicates that the block is not the last one.
 // * If set in tail, indicates that the channel is disconnected.
-const MARK_BIT: usize = 1;
+const MARK_BIT: u64 = 1;
 
 /// A slot in a block.
 struct Slot<T> {
@@ -66,7 +67,7 @@ struct Block<T> {
     next: AtomicPtr<Block<T>>,
 
     /// Slots for messages.
-    slots: [Slot<T>; BLOCK_CAP],
+    slots: [Slot<T>; BLOCK_CAP as usize],
 }
 
 impl<T> Block<T> {
@@ -97,7 +98,7 @@ impl<T> Block<T> {
     unsafe fn destroy(this: *mut Block<T>, start: usize) {
         // It is not necessary to set the `DESTROY` bit in the last slot because that slot has
         // begun destruction of the block.
-        for i in start..BLOCK_CAP - 1 {
+        for i in start..BLOCK_CAP as usize - 1 {
             let slot = (*this).slots.get_unchecked(i);
 
             // Mark the `DESTROY` bit if a thread is still using the slot.
@@ -118,7 +119,7 @@ impl<T> Block<T> {
 #[derive(Debug)]
 struct Position<T> {
     /// The index in the channel.
-    index: AtomicUsize,
+    index: AtomicU64,
 
     /// The block in the linked list.
     block: AtomicPtr<Block<T>>,
@@ -171,11 +172,11 @@ impl<T> Channel<T> {
         Channel {
             head: CachePadded::new(Position {
                 block: AtomicPtr::new(ptr::null_mut()),
-                index: AtomicUsize::new(0),
+                index: AtomicU64::new(0),
             }),
             tail: CachePadded::new(Position {
                 block: AtomicPtr::new(ptr::null_mut()),
-                index: AtomicUsize::new(0),
+                index: AtomicU64::new(0),
             }),
             receivers: SyncWaker::new(),
             _marker: PhantomData,
@@ -207,7 +208,7 @@ impl<T> Channel<T> {
             }
 
             // Calculate the offset of the index into the block.
-            let offset = (tail >> SHIFT) % LAP;
+            let offset = ((tail >> SHIFT) % LAP) as usize;
 
             // If we reached the end of the block, wait until the next one is installed.
             if offset == BLOCK_CAP {
@@ -302,7 +303,7 @@ impl<T> Channel<T> {
 
         loop {
             // Calculate the offset of the index into the block.
-            let offset = (head >> SHIFT) % LAP;
+            let offset = ((head >> SHIFT) % LAP) as usize;
 
             // If we reached the end of the block, wait until the next one is installed.
             if offset == BLOCK_CAP {
@@ -520,7 +521,7 @@ impl<T> Channel<T> {
                 head >>= SHIFT;
 
                 // Return the difference minus the number of blocks between tail and head.
-                return tail - head - tail / LAP;
+                return (tail - head - tail / LAP) as usize;
             }
         }
     }
@@ -567,7 +568,7 @@ impl<T> Channel<T> {
         let backoff = Backoff::new();
         let mut tail = self.tail.index.load(Ordering::Acquire);
         loop {
-            let offset = (tail >> SHIFT) % LAP;
+            let offset = ((tail >> SHIFT) % LAP) as usize;
             if offset != BLOCK_CAP {
                 break;
             }
@@ -585,7 +586,7 @@ impl<T> Channel<T> {
         unsafe {
             // Drop all messages between head and tail and deallocate the heap-allocated blocks.
             while head >> SHIFT != tail >> SHIFT {
-                let offset = (head >> SHIFT) % LAP;
+                let offset = ((head >> SHIFT) % LAP) as usize;
 
                 if offset < BLOCK_CAP {
                     // Drop the message in the slot.
@@ -645,7 +646,7 @@ impl<T> Drop for Channel<T> {
         unsafe {
             // Drop all messages between head and tail and deallocate the heap-allocated blocks.
             while head != tail {
-                let offset = (head >> SHIFT) % LAP;
+                let offset = ((head >> SHIFT) % LAP) as usize;
 
                 if offset < BLOCK_CAP {
                     // Drop the message in the slot.

crossbeam-channel/src/utils.rs

@@ -69,6 +69,46 @@ pub(crate) fn convert_timeout_to_deadline(timeout: Duration) -> Instant {
     }
 }
 
+#[cfg(not(crossbeam_no_atomic_64))]
+pub(crate) use core::sync::atomic::AtomicU64;
+
+#[cfg(crossbeam_no_atomic_64)]
+#[derive(Debug)]
+#[repr(transparent)]
+pub(crate) struct AtomicU64 {
+    inner: crossbeam_utils::atomic::AtomicCell<u64>,
+}
+
+#[cfg(crossbeam_no_atomic_64)]
+impl AtomicU64 {
+    pub(crate) const fn new(v: u64) -> Self {
+        Self {
+            inner: crossbeam_utils::atomic::AtomicCell::new(v),
+        }
+    }
+    pub(crate) fn load(&self, _order: Ordering) -> u64 {
+        self.inner.load()
+    }
+    pub(crate) fn store(&self, val: u64, _order: Ordering) {
+        self.inner.store(val);
+    }
+    pub(crate) fn compare_exchange_weak(
+        &self,
+        current: u64,
+        new: u64,
+        _success: Ordering,
+        _failure: Ordering,
+    ) -> Result<u64, u64> {
+        self.inner.compare_exchange(current, new)
+    }
+    pub(crate) fn fetch_add(&self, val: u64, _order: Ordering) -> u64 {
+        self.inner.fetch_add(val)
+    }
+    pub(crate) fn fetch_or(&self, val: u64, _order: Ordering) -> u64 {
+        self.inner.fetch_or(val)
+    }
+}
+
 /// A simple spinlock.
 pub(crate) struct Spinlock<T> {
     flag: AtomicBool,