Index

AArch64 (see ARM64)

ABA problem, #

aborting the process, #

AcqRel, #

• (see also release and acquire memory ordering)

acquire memory ordering (see release and acquire memory ordering)

add instruction (ARM), #

add instruction (x86), #

address-based waiting (Windows), #

• (see also futex)

air, out of thin, #

alignment, #

allocating IDs (see ID allocation)

AMD processors, #

and instruction (x86), #

Arc, #

• building our own, #
• cyclic structures, #
• get_mut, #
• memory ordering, #, #, #, #
• naming clones, #
• using for channel state, #
• weak pointers, #
  • performance cost, #

arguments, consuming, #

ARM64 (processor architecture), #

• aarch64-unknown-linux-musl target, #
• other-multi-copy atomic, #
• weakly ordered, #

ARM64 instructions

• add, #
• ARMv8.1 atomic instructions, #, #
• b.ne (branch if not equal), #
• cbnz (compare and branch on nonzero), #
• clrex (clear exclusive), #
• cmp (compare), #
• dmb (data memory barrier), #
• ldar (load-acquire register), #
• ldaxr (load-acquire exclusive register), #
• ldr (load register), #
• ldxr (load exclusive register), #
• load-linked and store-conditional instructions, #
• mov (move), #
• overview, #
• ret (return), #
• stlr (store-release register), #
• stlxr (store-release exclusive register), #
• str (store register), #
• stxr (store exclusive register), #

ARMv8 (see ARM64)

ARMv8.1 atomic instructions, #, #

• overview, #

array::from_fn, #

assembler, #

assembly, #

• inspecting compiler output, #

atomic, #, #

• compare-and-exchange operations, #
  • weak, #, #, #, #, #, #, #, #, #, #, #, #
• fetch-and-modify operations, #
  • wrapping behavior (add and sub), #, #, #, #
• load and store operations, #
  • example, stop flag, #, #, #, #, #
• memory ordering (see memory ordering)
• reference counting (see Arc)

atomic barriers (see fences)

atomic fences (see fences)

atomic types, #, #

• compare_exchange, #
• compare_exchange_weak, #
• fetch_add, #
  • wrapping behavior, #
    • (see also overflows)
• fetch_and, #
• fetch_max, #
• fetch_min, #
• fetch_nand, #
• fetch_or, #
• fetch_store (see swap)
• fetch_sub, #
  • wrapping behavior, #
    • (see also overflows)
• fetch_update, #
• fetch_xor, #
• get_mut, #
• load, #
• store, #
• swap, #

atomic-wait crate, #

AtomicBool, #

• (see also atomic types)
• locking using, #, #

AtomicI8 (see atomic types)

AtomicI16 (see atomic types)

AtomicI32 (see atomic types)

AtomicI64 (see atomic types)

AtomicIsize (see atomic types)

AtomicPtr, #

• (see also atomic types)
• compare-and-exchange, #
• lazy initialization, #

AtomicU8 (see atomic types)

AtomicU16 (see atomic types)

AtomicU32 (see atomic types)

AtomicU64 (see atomic types)

AtomicUsize (see atomic types)

auto traits, #

b.ne (branch if not equal) instruction (ARM), #

barriers (see fences)

basics, #

benchmarking, #, #

• black_box, avoiding optimizations with, #, #

binary semaphore, #

black_box, #, #

blocking, #

• channel, #
• condition variables, #
• futex wait operation, #
  • (see also futex)
• mutexes, #
• Once and OnceLock, #, #
• semaphores, #
• spin loop, #
• thread parking (see thread parking)

boolean (atomic) (see AtomicBool)

borrowing, #

• bending the rules, #
• error, #
• exclusive, #
• from multiple threads (Sync), #
• immutable, #
  • (see also shared)
• local variables in a thread, #
• mutable, #
  • (see also exclusive)
• shared, #
• splitting, #
• undefined behavior, #

Box

• from_raw, #, #
• into_raw, #
• leak, #, #
• unmovable type, wrapping in, #

btc (bit test and complement) instruction (x86), #

btr (bit test and reset) instruction (x86), #

bts (bit test and set) instruction (x86), #

building our own

• Arc, #
• channels, #
• condition variables, #
• mutexes, #
• reader-writer locks, #
• spin locks, #

busy-looping, #

• (see also spinning)

C standard library, #

• (see also libc)

cache coherence, #

• protocol, #
  • write-through, #, #, #, #

cache lines, #

• performance experiment, #

cache miss, #

caching (processors), #

• (see also cache coherence)
• compare-and-exchange operations, effect of, #
• per core, #
• performance experiment, #

cargo-show-asm, #

cas (compare and swap) instruction (ARM), #

casa (compare and swap, acquire) instruction (ARM), #

casal (compare and swap, acquire and release) instruction (ARM), #

casl (compare and swap, release) instruction (ARM), #

cbnz (compare and branch on nonzero) instruction (ARM), #

Cell, #

• unsafe (see UnsafeCell)

channels

• blocking, #
• borrowing, #
• building our own, #
• dropping, #
• one-shot, #
• safe interface, #
• Sender and Receiver types, #, #
• storing in Arc, #
  • avoiding, #
• unsafe interface, #

Clone trait, #, #, #, #, #

closures

• captured values
  • moving, #, #
• spawning scoped threads using, #
• spawning threads using, #

clrex (clear exclusive) instruction (ARM), #

cmp (compare) instruction (ARM), #

cmpxchg (compare and exchange) instruction (x86), #

#[cold], #

compare-and-exchange operations (atomic), #

• on ARM64, #
• caching, effect on, #
• compiler optimization, #
• example, ID allocation, #
• example, lazy initialization, #, #
• memory ordering, #
• using for channel state, #
• using for mutex state, #
• using for reader-writer lock state, #
• using on AtomicPtr, #
• using to lock reference counter, #
• weak, #
  • on ARM64, #
• on x86-64, #

Compiler Explorer, #

compiler fence, #, #

compiler optimization

• black_box, avoiding with, #, #
• #[cold], #
• of compare-and-exchange loops, #
• enabling, #, #
• #[inline], #
• reordering, #

complex instruction set computer (CISC), #

concurrency, basics, #

condition variables, #

• building our own, #
• example, #
• memory ordering, #
• pthread_cond_t, #
• thundering herd problem, #
• timeout, #
• using to build a channel, #
• Windows, #

Condvar, #

• (see also condition variables)

consume memory ordering, #

consuming arguments by value, #

contention (mutexes), #, #

• benchmarking, #

Copy trait, #, #

• atomic types, not implementing, #
• moving, #

critical section (Windows), #

current thread, #, #, #

cyclic structures (Arc), #

data races, #

• avoiding using atomics, #, #

Deref trait, #, #, #

DerefMut trait, #, #, #, #

disassembler, #, #

dmb (data memory barrier) instruction (ARM), #

drop function, #, #, #

Drop trait, #, #, #, #, #, #, #, #, #, #, #, #

dword, #

--emit=asm (rustc), #

exclusive references, #

fair locks, #

false sharing, #

fences, #, #, #

• on ARM64, #
• compiler fence, #, #
• instructions, #
• process-wide memory barriers, #
• on x86-64, #

fetch-and-modify operations (atomic), #

• on ARM64, #
• example, ID allocation, #
• example, progress reporting, #
• example, statistics, #
• wrapping behavior (add and sub), #
  • (see also overflows)
• on x86-64, #

fetch_store operation (atomic) (see swap operation)

fetch_update (atomic), #

FlushProcessWriteBuffers (Windows), #

forgetting (see leaking)

FreeBSD, umtx_op syscall, #

• (see also futex)

from_fn (array), #

futex, #

• cross-platform futex-like functionality, #
• example, #
• memory safety, #
• on other platforms, #
• operations (Linux), #
  • FUTEX_WAIT, #, #, #, #, #, #, #, #, #, #, #
• requeuing, #, #
• spurious wake-ups, #
• timeout, #, #
• wait operation, #
• wake operation, #

globally consistent order, #

• (see also sequentially consistent memory ordering)

Godbolt, #

good luck, #

guards

• dropping, #
• join guard, #
• mutex guard, #, #
• read guard, #, #
• spin lock guard, #
• write guard, #, #

hand, things getting out of, #

happens-before relationships, #

• in Arc, #, #
• between threads, #
• locking and unlocking, #, #
• spawning and joining threads, #
• through a release-acquire pair, #, #
  • chaining, #
• within the same thread, #

hint::black_box, #, #

hint::spin_loop, #

ID allocation

• using compare_exchange_weak, #
• using fetch_add, #
• using fetch_update, #

ideas and inspiration, #

if let statement

• lifetime of temporaries, #

ignorance, blissful, #

immutable references, #

• (see also shared references)

indivisible, #

#[inline], #

inspiration, #

Instant, #

instruction reordering (see reordering)

instructions, #

• (see also ARM64 instructions; x86-64 instructions)
• compare-and-exchange operations, #, #
• fences, #
• load and store operations, #
• load-linked/store-conditional (LL/SC) instructions, #
• memory ordering, #
• overview, #
• read-modify-write operations, #

Intel processors, #

interior mutability, #, #, #

invalidation queues, #

jne (jump if not equal) instruction (x86), #

join method, #

JoinGuard, #

JoinHandle, #

joining threads, #

• happens-before relationship, #

kernel, #, #

• interfacing with, #
• kernel-managed objects (Windows), #

L1/L2/L3/L4 cache, #

label (assembly), #

lazy initialization

• using compare_exchange, #
• using compare_exchange and allocation, #
• using load and store, #

ldadd (load and add) instruction (ARM), #

ldadda (load and add, acquire) instruction (ARM), #

ldaddal (load and add, acquire and release) instruction (ARM), #

ldaddl (load and add, release) instruction (ARM), #

ldar (load-acquire register) instruction (ARM), #

ldaxr (load-acquire exclusive register) instruction (ARM), #

ldr (load register) instruction (ARM), #

ldxr (load exclusive register) instruction (ARM), #

leaking, #, #

• by mistake, #
• a MutexGuard, #

“Leakpocalypse”, #

libc, #

• pthreads functionality in, #

libpthread, #

• (see also pthreads)

lifetime

• elision, #, #
• in a struct, #
• of mutex guard, #
• specifying using plain English, #
• static, #

linked list, #

Linux

• futex syscall, #
  • (see also futex)
  • arguments, #
• futex_waitv syscall, #
• interfacing with the kernel, #
• libc, role of, #
• membarrier syscall, #
• process-wide memory barrier, #
• RCU, #

load and store operations (atomic), #

• on ARM64 and x86-64, #
• compared to non-atomic operations, #, #
• example, lazy initialization, #
• example, progress reporting, #
• example, stop flag, #

load-linked/store-conditional (LL/SC) loop, #

• on ARM64, #
• compiler optimization, #

lock poisoning, #

lock prefix (x86), #

lock_api crate, #

luck, good, #

machine code, #

machine instructions (see instructions)

macOS

• futex-like functionality on, #
• interfacing with the kernel, #, #
• os_unfair_lock, #

main thread, #

ManuallyDrop, #

MaybeUninit, #, #, #

mem::forget, #

membarrier syscall, #

memory barriers (see fences)

memory fences (see fences)

memory model, #

memory ordering, #, #

• on ARM64, #
• compiler fence, #, #
• consume, #
• experiment, using relaxed instead of release and acquire, #
• fences, #, #, #
• happens-before relationship, #, #
• Miri, detecting problems with, #
• misconceptons about, #
• out-of-thin-air values, #
• at processor level, #
• reference counting, #, #, #, #, #
• relaxed, #, #
• release and acquire, #
  • (see also release and acquire memory ordering)
  • locking and unlocking, #
• sequentially consistent, #
  • (see also sequentially consistent memory ordering)
• specifying using plain English, #
• total modification order, #, #, #, #
• on x86-64, #

MESI cache coherence protocol, #

MESIF cache coherence protocol, #

mfence (memory fence) instruction (x86), #

microinstructions, #

Miri, #

MOESI cache coherence protocol, #

mov (move) instruction (ARM), #

mov (move) instruction (x86), #

movable, not

• critical section (Windows), #
• Pin, #
• pthread types, #
• wrapping in Box, #

move closure, #

multi-copy atomicity, #

mutability, interior (see interior mutability)

mutable references, #

• (see also exclusive references)

Mutex, #, #

• (see also mutexes)

mutexes, #

• building our own, #
• as container, #
• contention, #, #
• example, #
• fair, #
• happens-before relationship, #
• into_inner, #
• lifetime of mutex guard, #
• memory ordering, #
• Mutex type in Rust, #
• os_unfair_lock (macOS), #
• in other languages, #
• poisoning, #
• pthread
  • wrapping in Rust, #
• pthread_mutex_t, #
• recursive, #
• robust, #
• Send requirement, #
• spin locks, #
  • (see also spin locks)
• spinning, #, #
• using to build a channel, #

MutexGuard, #

• dropping, #
• lifetime of, #

mutual exclusion (see mutexes)

name of a thread, #

NetBSD, futex support, #

• (see also futex)

NonNull, #

-O flag (rustc), #, #

Once and OnceLock, #, #

one-shot channels, #

OpenBSD, limited futex support, #

• (see also futex)

operating systems, #

• (see also Linux; macOS; Windows)
• libraries shipped with, #
• synchronization primitives, #

optimization (see compiler optimization)

or instruction (x86), #, #

Ordering, #, #

• AcqRel, #
  • (see also release and acquire memory ordering)
• Acquire, #
  • (see also release and acquire memory ordering)
• Consume, #
• Relaxed, #, #
• Release, #
  • (see also release and acquire memory ordering)
• SeqCst, #
  • (see also sequentially consistent memory ordering)

os_unfair_lock (macOS), #

other-multi-copy atomicity, #

out of order execution (see reordering)

out-of-thin-air values, #

output locking, #

overflows (atomic), #

• (see also wrapping behavior)
• aborting on, #
• notification counter, #
• panicking on, #
• preventing (compare-and-exchange), #
• reference counter, #
• usize, big enough, #

overview of atomic instructions, #

ownership

• moving, #, #
• sharing, #
• transferring to another thread (Send), #

panicking

• poisoned mutexes, #
• RefCell, borrowing, #
• thread name in panic messages, #
• using a Condvar with multiple mutexes, #
• when joining a thread, #, #
• when spawning a thread, #

parking (see thread parking)

parking lot-based locks, #

parking_lot crate, #

PhantomData, #, #

Pin, #

pipelining, #

pointers

• atomic (see AtomicPtr)
• neither Send nor Sync, #
• NonNull, #

poisoning, lock, #

POSIX, #

• pthreads, #

println, use of output locking, #

priority inheritance, #

priority inversion, #

privacy (modules), #

process-wide memory barriers, #

processes, #

processor architecture, #

• (see also ARM64; x86-64)
• strongly ordered, #
• weakly ordered, #

processor caching (see caching)

processor instructions (see instructions)

processor registers, #

• return value, #

pthreads, #

• pthread_cond_t, #, #
• pthread_mutex_t, #
  • dropping while locked, #
• pthread_rwlock_t, #
• wrapping in Rust, #

queue-based locks, #

racing, #

Rc, #

RCU (read, copy, update), #, #

reader-writer locks, #

• avoiding accidental spinning, #
• building our own, #
• pthread_rwlock_t, #
• Send requirement, #
• SRW locks (Windows), #
• Sync requirement, #
• writer starvation, #, #

recursive locking, #, #

reduced instruction set computer (RISC), #

RefCell, #

• RwLock compared to, #

reference counting, #

• (see also Arc)

references

• exclusive, #
• immutable, #
  • (see also shared)
• mutable, #
  • (see also exclusive)
• shared, #

registers, #

• return value, #

relaxed memory ordering, #, #, #

• counter-intuitive results, #
• misconceptions about, #, #
• out-of-thin-air values, #
• reference counting, #
• total modification order, #, #, #, #

release and acquire memory ordering, #

• acquire fence, #, #, #, #, #
• on ARM64, #
• example, lazy initialization, #
• experiment, using relaxed instead, #
• happens-before relationship, #, #
  • chaining, #
• locking and unlocking, #, #
• reference counting, #, #, #, #
• release fence, #
• on x86-64, #

--release flag (cargo), #, #

reordering (instructions), #, #

• memory ordering, #

#[repr(align)], #

requeuing waiting threads, #, #

ret (return) instruction (ARM), #

ret (return) instruction (x86), #

robust mutexes, #

rustup, #

RwLock, #, #

• (see also reader-writer locks)

RwLockReadGuard, #

RwLockWriteGuard, #

safe interface, #, #, #

safety requirements of unsafe functions, #

scheduler, #

scoped threads, #

semaphores, #

Send trait, #, #, #

• error, #
• implementing for Arc, #
• implementing for spin lock guard, #
• requirement by Mutex and RwLock, #

SeqCst (see sequentially consistent memory ordering)

sequence locks, #

sequentially consistent memory ordering, #

• on ARM64, #
• fence, #
• misconceptions about, #, #
• on x86-64, #

shadowing, #

shared ownership, #

• leaking, #
• reference counting, #
• statics, #

shared references, #

• mutating atomics through, #

slim reader-writer locks (Windows), #, #

spawning threads, #

• failing to, #
• happens-before relationship, #
• scoped, #

spin locks

• building our own, #
• cache lines, effect of, #
• compare-and-exchange, (not) using, #
• experiment, using wrong memory ordering, #
• guard, #
• memory ordering, #

spin loop hint, #, #

spinning, #, #, #

• avoiding accidental (reader-writer lock), #

splitting (borrowing), #

spurious wake-ups, #, #, #

SRW locks (Windows), #, #

stack size, #

starvation, #, #

static lifetime, #

statics, #

stlr (store-release register) instruction (ARM), #

stlxr (store-release exclusive register) instruction (ARM), #

stop flag, #

store buffers, #

store operations (atomic) (see load and store operations)

store-conditional (see load-linked/store-conditional)

str (store register) instruction (ARM), #

stress, reducing, #

strongly ordered architecture, #

stxr (store exclusive register) instruction (ARM), #

sub (subtract) instruction (x86), #

swap operation (atomic), #

• locking using, #

Sync trait, #, #

• implementing for Arc, #
• implementing for channel, #
• implementing for mutex, #
• implementing for reader-writer lock, #
• implementing for spin lock, #
• implementing for spin lock guard, #
• requirement by RwLock, #

SYS_futex (Linux), #

• (see also futex)
• arguments, #

syscalls, #

• avoiding, #, #

--target (rustc), #

teaching, #

thin air, out of, #

thread builder, #

thread name, #

Thread object, #

• id, #
• unpark, #, #

thread parking, #, #, #, #

• spurious wake-ups, #
• timeout, #
  • example, #

thread safety, #, #

• keeping objects on one thread, #

ThreadId, #

threads, #

• joining, #
• panicking, #, #
• returning a value, #
• scoped, #
• spawning, #

thundering herd problem, #

time travel, #

timeout

• condition variables, #
• futex, #, #
• thread parking, #
  • example, #

total modification order, #, #, #, #

uncontended (mutexes), #, #

• benchmarking, #

undefined behavior, #

• borrowing, #
• data races, #
• Miri, detecting with, #
• time travel, #

uninitialized memory, #

Unix systems

• interfacing with the kernel, #
• libc, role of, #

unmovable

• critical section (Windows), #
• Pin, #
• pthread types, #
• wrapping in Box, #

unpark (Thread), #

unparking (see thread parking)

unsafe code, #

unsafe functions, #

unsafe trait implementation, #

UnsafeCell, #, #, #

• get_mut, #

unsound, #

VecDeque, #

waiting (see blocking)

WaitOnAddress (Windows), #

WakeByAddressAll (Windows), #

WakeByAddressSingle (Windows), #

Weak (see Arc; weak pointers)

weakly ordered architecture, #

• experiment, using relaxed instead of release and acquire, #

Windows, #

• condition variables, #
• critical section, #
• FlushProcessWriteBuffers, #
• interfacing with the kernel, #
• kernel-managed objects, #
• Native API, #
• process-wide memory barrier, #
• SRW locks, #, #
• WaitOnAddress, #
• WakeByAddressAll, #
• WakeByAddressSingle, #
• Win32 API, #

windows crate, #

windows-sys crate, #

wrapping behavior (fetch_add and fetch_sub), #

• (see also overflows (atomic))

wrapping unmovable object in Box, #

write-through cache coherence protocol, #

writer starvation, #, #

x86-64 (processor architecture), #

• other-multi-copy atomic, #
• strongly ordered, #
• x86_64-unknown-linux-musl target, #

x86-64 instructions

• add, #
• and, #
• btc (bit test and complement), #
• btr (bit test and reset), #
• bts (bit test and set), #
• cmpxchg (compare and exchange), #
• jne (jump if not equal), #
• lock prefix, #
• mfence (memory fence), #
• mov (move), #
• or, #, #
• overview, #
• ret (return), #
• sub (subtract), #
• xadd (exchange and add), #
• xchg (exchange), #, #
• xor, #

xadd (exchange and add) instruction (x86), #

xchg (exchange) instruction (x86), #, #

xor instruction (x86), #