- The new methods
ParallelSlice::par_chunk_by
andParallelSliceMut::par_chunk_by_mut
work like the slice methodschunk_by
andchunk_by_mut
added in Rust 1.77.
- The new methods
IndexedParallelIterator::by_exponential_blocks
andby_uniform_blocks
allow processing items in smaller groups at a time. - The new
iter::walk_tree
,walk_tree_prefix
, andwalk_tree_postfix
functions enable custom parallel iteration over tree-like structures. - The new method
ParallelIterator::collect_vec_list
returns items as a linked list of vectors, which is an efficient mode of parallel collection used by many of the internal implementations ofcollect
. - The new methods
ParallelSliceMut::par_split_inclusive_mut
,ParallelSlice::par_split_inclusive
, andParallelString::par_split_inclusive
all work like a normal split but keeping the separator as part of the left slice. - The new
ParallelString::par_split_ascii_whitespace
splits only on ASCII whitespace, which is faster than including Unicode multi-byte whitespace. OsString
now implementsFromParallelIterator<_>
andParallelExtend<_>
for a few item types similar to the standardFromIterator
andExtend
.- The internal
Pattern
trait for string methods is now implemented for[char; N]
and&[char; N]
, matching any of the given characters.
- The new
"web_spin_lock"
crate feature makes mutexes spin on the main browser thread in WebAssembly, rather than suffer an error about forbiddenatomics.wait
if they were to block in that context. Thanks @RReverser!
- The minimum supported
rustc
is now 1.63. - Added
ThreadPoolBuilder::use_current_thread
to use the builder thread as part of the new thread pool. That thread does not run the pool's main loop, but it may participate in work-stealing if it yields to rayon in some way. - Implemented
FromParallelIterator<T>
forBox<[T]>
,Rc<[T]>
, andArc<[T]>
, as well asFromParallelIterator<Box<str>>
andParallelExtend<Box<str>>
forString
. ThreadPoolBuilder::build_scoped
now usesstd::thread::scope
.- The default number of threads is now determined using
std::thread::available_parallelism
instead of thenum_cpus
crate. - The internal logging facility has been removed, reducing bloat for all users.
- Many smaller performance tweaks and documentation updates.
- The minimum supported
rustc
is now 1.59. - Added a fallback when threading is unsupported.
- The new
ParallelIterator::take_any
andskip_any
methods work like unorderedIndexedParallelIterator::take
andskip
, counting items in whatever order they are visited in parallel. - The new
ParallelIterator::take_any_while
andskip_any_while
methods work like unorderedIterator::take_while
andskip_while
, which previously had no parallel equivalent. The "while" condition may be satisfied from anywhere in the parallel iterator, affecting all future items regardless of position. - The new
yield_now
andyield_local
functions will cooperatively yield execution to Rayon, either trying to execute pending work from the entire pool or from just the local deques of the current thread, respectively.
- Fixed miri-reported UB for SharedReadOnly tags protected by a call.
- Simplified
par_bridge
to only pull one item at a time from the iterator, without batching. Threads that are waiting for iterator items will now block appropriately rather than spinning CPU. (Thanks @njaard!) - Added protection against recursion in
par_bridge
, so iterators that also invoke rayon will not cause mutex recursion deadlocks.
- Fixed a race condition with threads going to sleep while a broadcast starts.
- The minimum supported
rustc
is now 1.56. - The new
IndexedParallelIterator::fold_chunks
andfold_chunks_with
methods work likeParallelIterator::fold
andfold_with
with fixed-size chunks of items. This may be useful for predictable batching performance, without the allocation overhead ofIndexedParallelIterator::chunks
. - New "broadcast" methods run a given function on all threads in the pool.
These run at a sort of reduced priority after each thread has exhausted their
local work queue, but before they attempt work-stealing from other threads.
- The global
broadcast
function andThreadPool::broadcast
method will block until completion, returning aVec
of all return values. - The global
spawn_broadcast
function and methods onThreadPool
,Scope
, andScopeFifo
will run detached, without blocking the current thread.
- The global
- Panicking methods now use
#[track_caller]
to report the caller's location. - Fixed a truncated length in
vec::Drain
when given an empty range.
Thanks to all of the contributors for this release!
- @cuviper
- @idanmuze
- @JoeyBF
- @JustForFun88
- @kianmeng
- @kornelski
- @ritchie46
- @ryanrussell
- @steffahn
- @TheIronBorn
- @willcrozi
- The new
ParallelSliceMut::par_sort_by_cached_key
is a stable sort that caches the keys for each item -- a parallel version ofslice::sort_by_cached_key
.
- Fixed a use-after-free race in job notification.
- The new
ParallelSlice::par_rchunks()
andpar_rchunks_exact()
iterate slice chunks in reverse, aligned the against the end of the slice if the length is not a perfect multiple of the chunk size. The newParallelSliceMut::par_rchunks_mut()
andpar_rchunks_exact_mut()
are the same for mutable slices. - The
ParallelIterator::try_*
methods now supportstd::ops::ControlFlow
andstd::task::Poll
items, mirroring the unstableTry
implementations in the standard library. - The
ParallelString
pattern-based methods now support&[char]
patterns, which match when any character in that slice is found in the string. - A soft limit is now enforced on the number of threads allowed in a single
thread pool, respecting internal bit limits that already existed. The current
maximum is publicly available from the new function
max_num_threads()
. - Fixed several Stacked Borrow and provenance issues found by
cargo miri
.
Thanks to all of the contributors for this release!
- @atouchet
- @bluss
- @cuviper
- @fzyzcjy
- @nyanzebra
- @paolobarbolini
- @RReverser
- @saethlin
- The new
in_place_scope
andin_place_scope_fifo
are variations ofscope
andscope_fifo
, running the initial non-Send
callback directly on the current thread, rather than moving execution to the thread pool. - With Rust 1.51 or later, arrays now implement
IntoParallelIterator
. - New implementations of
FromParallelIterator
make it possible tocollect
complicated nestings of items.FromParallelIterator<(A, B)> for (FromA, FromB)
works likeunzip
.FromParallelIterator<Either<L, R>> for (A, B)
works likepartition_map
.
- Type inference now works better with parallel
Range
andRangeInclusive
. - The implementation of
FromParallelIterator
andParallelExtend
forVec<T>
now usesMaybeUninit<T>
internally to avoid creating any references to uninitialized data. ParallelBridge
fixed a bug with threads missing available work.
Thanks to all of the contributors for this release!
- @atouchet
- @cuviper
- @Hywan
- @iRaiko
- @Qwaz
- @rocallahan
- Update crossbeam dependencies.
- The minimum supported
rustc
is now 1.36.
Thanks to all of the contributors for this release!
- @cuviper
- @mbrubeck
- @mrksu
- The new
flat_map_iter
andflatten_iter
methods can be used to flatten sequential iterators, which may perform better in cases that don't need the nested parallelism offlat_map
andflatten
. - The new
par_drain
method is a parallel version of the standarddrain
for collections, removing items while keeping the original capacity. Collections that implement this throughParallelDrainRange
support draining items from arbitrary index ranges, whileParallelDrainFull
always drains everything. - The new
positions
method finds all items that match the given predicate and returns their indices in a new iterator.
- Fixed an overflow panic on high-contention workloads, for a counter that was meant to simply wrap. This panic only occurred with debug assertions enabled, and was much more likely on 32-bit targets.
- Implemented a new thread scheduler, RFC 5, which uses targeted wakeups for new work and for notifications of completed stolen work, reducing wasteful CPU usage in idle threads.
- Implemented
IntoParallelIterator for Range<char>
andRangeInclusive<char>
with the same iteration semantics as Rust 1.45. - Relaxed the lifetime requirements of the initial
scope
closure.
Thanks to all of the contributors for this release!
- @CAD97
- @cuviper
- @kmaork
- @nikomatsakis
- @SuperFluffy
- Fixed a use-after-free race in calls blocked between two rayon thread pools.
- Collecting to an indexed
Vec
now drops any partial writes while unwinding, rather than just leaking them. If dropping also panics, Rust will abort.- Note: the old leaking behavior is considered safe, just not ideal.
- The new
IndexedParallelIterator::step_by()
adapts an iterator to step through items by the given count, likeIterator::step_by()
. - The new
ParallelSlice::par_chunks_exact()
and mutable equivalentParallelSliceMut::par_chunks_exact_mut()
ensure that the chunks always have the exact length requested, leaving any remainder separate, like the slice methodschunks_exact()
andchunks_exact_mut()
.
Thanks to all of the contributors for this release!
- @adrian5
- @bluss
- @cuviper
- @FlyingCanoe
- @GuillaumeGomez
- @matthiasbeyer
- @picoHz
- @zesterer
- Tuples up to length 12 now implement
IntoParallelIterator
, creating aMultiZip
iterator that produces items as similarly-shaped tuples. - The
--cfg=rayon_unstable
supporting code forrayon-futures
is removed. - The minimum supported
rustc
is now 1.31.
Thanks to all of the contributors for this release!
- @cuviper
- @c410-f3r
- @silwol
Send
bounds have been added for theItem
andError
associated types on all genericF: Future
interfaces. While technically a breaking change, this is a soundness fix, so we are not increasing the semantic version for this.- This crate is now deprecated, and the
--cfg=rayon_unstable
supporting code will be removed inrayon-core 1.7.0
. This only supported the now-obsoleteFuture
fromfutures 0.1
, while support forstd::future::Future
is expected to come directly inrayon-core
-- although that is not ready yet.
Thanks to all of the contributors for this release!
- @cuviper
- @kornelski
- @jClaireCodesStuff
- @jwass
- @seanchen1991
- Update crossbeam dependencies.
- Add top-level doc links for the iterator traits.
- Document that the iterator traits are not object safe.
Thanks to all of the contributors for this release!
- @cuviper
- @dnaka91
- @matklad
- @nikomatsakis
- @Qqwy
- @vorner
- The new
ParallelIterator::copied()
converts an iterator of references into copied values, likeIterator::copied()
. ParallelExtend
is now implemented for the unit()
.- Internal updates were made to improve test determinism, reduce closure type sizes, reduce task allocations, and update dependencies.
- The minimum supported
rustc
is now 1.28.
Thanks to all of the contributors for this release!
- @Aaron1011
- @cuviper
- @ralfbiedert
- FIFO spawns are now supported using the new
spawn_fifo()
andscope_fifo()
global functions, and their correspondingThreadPool
methods.- Normally when tasks are queued on a thread, the most recent is processed first (LIFO) while other threads will steal the oldest (FIFO). With FIFO spawns, those tasks are processed locally in FIFO order too.
- Regular spawns and other tasks like
join
are not affected. - The
breadth_first
configuration flag, which globally approximated this effect, is now deprecated. - For more design details, please see RFC 1.
ThreadPoolBuilder
can now take a customspawn_handler
to control how threads will be created in the pool.ThreadPoolBuilder::build_scoped()
uses this to create a scoped thread pool, where the threads are able to use non-static data.- This may also be used to support threading in exotic environments, like
WebAssembly, which don't support the normal
std::thread
.
ParallelIterator
has 3 new methods:find_map_any()
,find_map_first()
, andfind_map_last()
, likeIterator::find_map()
with ordering constraints.- The new
ParallelIterator::panic_fuse()
makes a parallel iterator halt as soon as possible if any of its threads panic. Otherwise, the panic state is not usually noticed until the iterator joins its parallel tasks back together. IntoParallelIterator
is now implemented for integralRangeInclusive
.- Several internal
Folder
s now have optimizedconsume_iter
implementations. rayon_core::current_thread_index()
is now re-exported inrayon
.- The minimum
rustc
is now 1.26, following the update policy defined in RFC 3.
Thanks to all of the contributors for this release!
- @cuviper
- @didroe
- @GuillaumeGomez
- @huonw
- @janriemer
- @kornelski
- @nikomatsakis
- @seanchen1991
- @yegeun542
ParallelExtend
is now implemented for tuple pairs, enabling nestedunzip()
andpartition_map()
operations. For instance,(A, (B, C))
items can be unzipped into(Vec<A>, (Vec<B>, Vec<C>))
.ParallelExtend<(A, B)>
works likeunzip()
.ParallelExtend<Either<A, B>>
works likepartition_map()
.
ParallelIterator
now has a methodmap_init()
which calls aninit
function for a value to pair with items, likemap_with()
but dynamically constructed. That value type has no constraints, not evenSend
orSync
.- The new
for_each_init()
is a variant of this for simple iteration. - The new
try_for_each_init()
is a variant for fallible iteration.
- The new
Thanks to all of the contributors for this release!
- @cuviper
- @dan-zheng
- @dholbert
- @ignatenkobrain
- @mdonoughe
- The
ParallelBridge
trait with methodpar_bridge()
makes it possible to use anySend
ableIterator
in parallel!- This trait has been added to
rayon::prelude
. - It automatically implements internal synchronization and queueing to
spread the
Item
s across the thread pool. Iteration order is not preserved by this adaptor. - "Native" Rayon iterators like
par_iter()
should still be preferred when possible for better efficiency.
- This trait has been added to
ParallelString
now has additional methods for parity withstd
string iterators:par_char_indices()
,par_bytes()
,par_encode_utf16()
,par_matches()
, andpar_match_indices()
.ParallelIterator
now has fallible methodstry_fold()
,try_reduce()
, andtry_for_each
, plus*_with()
variants of each, for automatically short-circuiting iterators onNone
orErr
values. These are inspired byIterator::try_fold()
andtry_for_each()
that were stabilized in Rust 1.27.Range<i128>
andRange<u128>
are now supported with Rust 1.26 and later.- Small improvements have been made to the documentation.
rayon-core
now only depends onrand
for testing.- Rayon tests now work on stable Rust.
Thanks to all of the contributors for this release!
- @AndyGauge
- @cuviper
- @ignatenkobrain
- @LukasKalbertodt
- @MajorBreakfast
- @nikomatsakis
- @paulkernfeld
- @QuietMisdreavus
- Added more documentation for
rayon::iter::split()
. - Corrected links and typos in documentation.
Thanks to all of the contributors for this release!
- @cuviper
- @HadrienG2
- @matthiasbeyer
- @nikomatsakis
ParallelIterator
added theupdate
method which applies a function to mutable references, inspired byitertools
.IndexedParallelIterator
added thechunks
method which yields vectors of consecutive items from the base iterator, inspired byitertools
.String
now implementsFromParallelIterator<Cow<str>>
andParallelExtend<Cow<str>>
, inspired bystd
.()
now implementsFromParallelIterator<()>
, inspired bystd
.- The new
ThreadPoolBuilder
replaces and deprecatesConfiguration
.- Errors from initialization now have the concrete
ThreadPoolBuildError
type, rather thanBox<Error>
, and this type implementsSend
andSync
. ThreadPool::new
is deprecated in favor ofThreadPoolBuilder::build
.initialize
is deprecated in favor ofThreadPoolBuilder::build_global
.
- Errors from initialization now have the concrete
- Examples have been added to most of the parallel iterator methods.
- A lot of the documentation has been reorganized and extended.
- Rayon now requires rustc 1.13 or greater.
IndexedParallelIterator::len
andParallelIterator::opt_len
now operate on&self
instead of&mut self
.IndexedParallelIterator::collect_into
is nowcollect_into_vec
.IndexedParallelIterator::unzip_into
is nowunzip_into_vecs
.- Rayon no longer exports the deprecated
Configuration
andinitialize
from rayon-core.
Thanks to all of the contributors for this release!
- @Bilkow
- @cuviper
- @Enet4
- @ignatenkobrain
- @iwillspeak
- @jeehoonkang
- @jwass
- @Kerollmops
- @KodrAus
- @kornelski
- @MaloJaffre
- @nikomatsakis
- @obv-mikhail
- @oddg
- @phimuemue
- @stjepang
- @tmccombs
- bors[bot]
Configuration
now has abuild
method.ParallelIterator
addedflatten
andintersperse
, both inspired by itertools.IndexedParallelIterator
addedinterleave
,interleave_shortest
, andzip_eq
, all inspired by itertools.- The new functions
iter::empty
andonce
create parallel iterators of exactly zero or one item, like theirstd
counterparts. - The new functions
iter::repeat
andrepeatn
create parallel iterators repeating an item indefinitely orn
times, respectively. - The new function
join_context
works likejoin
, with an addedFnContext
parameter that indicates whether the job was stolen. Either
(used byParallelIterator::partition_map
) is now re-exported from theeither
crate, instead of defining our own type.Either
also now implementsParallelIterator
,IndexedParallelIterator
, andParallelExtend
when both of itsLeft
andRight
types do.
- All public types now implement
Debug
. - Many of the parallel iterators now implement
Clone
where possible. - Much of the documentation has been extended. (but still could use more help!)
- All rayon crates have improved metadata.
- Rayon was evaluated in the Libz Blitz, leading to many of these improvements.
- Rayon pull requests are now guarded by bors-ng.
The spawn_future()
method has been refactored into its own rayon-futures
crate, now through a ScopeFutureExt
trait for ThreadPool
and Scope
. The
supporting rayon-core
APIs are still gated by --cfg rayon_unstable
.
- Two breaking changes have been made to
rayon-core
, but since they're fixing soundness bugs, we are considering these minor changes for semver.Scope::spawn
now requiresSend
for the closure.ThreadPool::install
now requiresSend
for the return value.
- The
iter::internal
module has been renamed toiter::plumbing
, to hopefully indicate that while these are low-level details, they're not really internal or private to rayon. The contents of that module are needed for third-parties to implement new parallel iterators, and we'll treat them with normal semver stability guarantees. - The function
rayon::iter::split
is no longer re-exported asrayon::split
.
Thanks to all of the contributors for this release!
- @AndyGauge
- @ChristopherDavenport
- @chrisvittal
- @cuviper
- @dns2utf8
- @dtolnay
- @frewsxcv
- @gsquire
- @Hittherhod
- @jdr023
- @laumann
- @leodasvacas
- @lvillani
- @MajorBreakfast
- @mamuleanu
- @marmistrz
- @mbrubeck
- @mgattozzi
- @nikomatsakis
- @smt923
- @stjepang
- @tmccombs
- @vishalsodani
- bors[bot]
ParallelSliceMut
now has six parallel sorting methods with the same variations as the standard library.par_sort
,par_sort_by
, andpar_sort_by_key
perform stable sorts in parallel, using the default order, a custom comparator, or a key extraction function, respectively.par_sort_unstable
,par_sort_unstable_by
, andpar_sort_unstable_by_key
perform unstable sorts with the same comparison options.- Thanks to @stjepang!
- The following core APIs are being stabilized:
rayon::spawn()
-- spawns a task into the Rayon threadpool; as it is contained in the global scope (rather than a user-created scope), the task cannot capture anything from the current stack frame.ThreadPool::join()
,ThreadPool::spawn()
,ThreadPool::scope()
-- convenience APIs for launching new work within a thread-pool.
- The various iterator adapters are now tagged with
#[must_use]
- Parallel iterators now offer a
for_each_with
adapter, similar tomap_with
. - We are adopting a new approach to handling the remaining unstable
APIs (which primarily pertain to futures integration). As awlays,
unstable APIs are intended for experimentation, but do not come with
any promise of compatibility (in other words, we might change them
in arbitrary ways in any release). Previously, we designated such
APIs using a Cargo feature "unstable". Now, we are using a regular
#[cfg]
flag. This means that to see the unstable APIs, you must doRUSTFLAGS='--cfg rayon_unstable' cargo build
. This is intentionally inconvenient; in particular, if you are a library, then your clients must also modify their environment, signaling their agreement to instability.
- Added the
map_with
andfold_with
combinators, which help for passing along state (like channels) that cannot be shared between threads but which can be cloned on each thread split. - Added the
while_some
combinator, which helps for writing short-circuiting iterators. - Added support for "short-circuiting" collection: e.g., collecting
from an iterator producing
Option<T>
orResult<T, E>
into aOption<Collection<T>>
orResult<Collection<T>, E>
. - Support
FromParallelIterator
forCow
. - Removed the deprecated weight APIs.
- Simplified the parallel iterator trait hierarchy by removing the
BoundedParallelIterator
andExactParallelIterator
traits, which were not serving much purpose. - Improved documentation.
- Added some missing
Send
impls. - Fixed some small bugs.
- We now have more documentation.
- Renamed the (unstable) methods
spawn_async
andspawn_future_async
-- which spawn tasks that cannot hold references -- to simplyspawn
andspawn_future
, respectively. - We are now using the coco library for our deque.
- Individual threadpools can now be configured in "breadth-first" mode, which causes them to execute spawned tasks in the reverse order that they used to. In some specific scenarios, this can be a win (though it is not generally the right choice).
- Added top-level functions:
current_thread_index
, for querying the index of the current worker thread within its thread-pool (previously available asthread_pool.current_thread_index()
);current_thread_has_pending_tasks
, for querying whether the current worker that has an empty task deque or not. This can be useful when deciding whether to spawn a task.
- The environment variables for controlling Rayon are now
RAYON_NUM_THREADS
andRAYON_LOG
. The older variables (e.g.,RAYON_RS_NUM_CPUS
are still supported but deprecated).
- Added a new game-of-life benchmark.
Thanks to the following contributors:
- @ChristopherDavenport
- @SuperFluffy
- @antoinewdg
- @crazymykl
- @cuviper
- @glandium
- @julian-seward1
- @leodasvacas
- @leshow
- @lilianmoraru
- @mschmo
- @nikomatsakis
- @stjepang
This release is a targeted performance fix for #343, an issue where rayon threads could sometimes enter into a spin loop where they would be unable to make progress until they are pre-empted.
This release marks the first step towards Rayon 1.0. For best
performance, it is important that all Rayon users update to at least
Rayon 0.7. This is because, as of Rayon 0.7, we have taken steps to
ensure that, no matter how many versions of rayon are actively in use,
there will only be a single global scheduler. This is achieved via the
rayon-core
crate, which is being released at version 1.0, and which
encapsulates the core schedule APIs like join()
. (Note: the
rayon-core
crate is, to some degree, an implementation detail, and
not intended to be imported directly; it's entire API surface is
mirrored through the rayon crate.)
We have also done a lot of work reorganizing the API for Rayon 0.7 in preparation for 1.0. The names of iterator types have been changed and reorganized (but few users are expected to be naming those types explicitly anyhow). In addition, a number of parallel iterator methods have been adjusted to match those in the standard iterator traits more closely. See the "Breaking Changes" section below for details.
Finally, Rayon 0.7 includes a number of new features and new parallel iterator methods. As of this release, Rayon's parallel iterators have officially reached parity with sequential iterators -- that is, every sequential iterator method that makes any sense in parallel is supported in some capacity.
- The internal
Producer
trait now featuresfold_with
, which enables better performance for some parallel iterators. - Strings now support
par_split()
andpar_split_whitespace()
. - The
Configuration
API is expanded and simplified:num_threads(0)
no longer triggers an error- you can now supply a closure to name the Rayon threads that get created
by using
Configuration::thread_name
. - you can now inject code when Rayon threads start up and finish
- you can now set a custom panic handler to handle panics in various odd situations
- Threadpools are now able to more gracefully put threads to sleep when not needed.
- Parallel iterators now support
find_first()
,find_last()
,position_first()
, andposition_last()
. - Parallel iterators now support
rev()
, which primarily affects subsequent calls toenumerate()
. - The
scope()
API is now considered stable (and part ofrayon-core
). - There is now a useful
rayon::split
function for creating custom Rayon parallel iterators. - Parallel iterators now allow you to customize the min/max number of
items to be processed in a given thread. This mechanism replaces the
older
weight
mechanism, which is deprecated. sum()
and friends now use the standardSum
traits
In the move towards 1.0, there have been a number of minor breaking changes:
- Configuration setters like
Configuration::set_num_threads()
lost theset_
prefix, and hence become something likeConfiguration::num_threads()
. Configuration
getters are removed- Iterator types have been shuffled around and exposed more consistently:
- combinator types live in
rayon::iter
, e.g.rayon::iter::Filter
- iterators over various types live in a module named after their type,
e.g.
rayon::slice::Windows
- combinator types live in
- When doing a
sum()
orproduct()
, type annotations are needed for the result since it is now possible to have the resulting sum be of a type other than the value you are iterating over (this mirrors sequential iterators).
Experimental features require the use of the unstable
feature. Their
APIs may change or disappear entirely in future releases (even minor
releases) and hence they should be avoided for production code.
- We now have (unstable) support for futures integration. You can use
Scope::spawn_future
orrayon::spawn_future_async()
. - There is now a
rayon::spawn_async()
function for using the Rayon threadpool to run tasks that do not have references to the stack.
Thanks to the following people for their contributions to this release:
- @Aaronepower
- @ChristopherDavenport
- @bluss
- @cuviper
- @froydnj
- @gaurikholkar
- @hniksic
- @leodasvacas
- @leshow
- @martinhath
- @mbrubeck
- @nikomatsakis
- @pegomes
- @schuster
- @torkleyy
This release includes a lot of progress towards the goal of parity with the sequential iterator API, though there are still a few methods that are not yet complete. If you'd like to help with that effort, check out the milestone to see the remaining issues.
Announcement: @cuviper has been added as a collaborator to the Rayon repository for all of his outstanding work on Rayon, which includes both internal refactoring and helping to shape the public API. Thanks @cuviper! Keep it up.
- We now support
collect()
and not justcollect_with()
. You can usecollect()
to build a number of collections, including vectors, maps, and sets. Moreover, when building a vector withcollect()
, you are no longer limited to exact parallel iterators. Thanks @nikomatsakis, @cuviper! - We now support
skip()
andtake()
on parallel iterators. Thanks @martinhath! - Breaking change: We now match the sequential APIs for
min()
andmax()
. We also supportmin_by_key()
andmax_by_key()
. Thanks @tapeinosyne! - Breaking change: The
mul()
method is now renamed toproduct()
, to match sequential iterators. Thanks @jonathandturner! - We now support parallel iterator over ranges on
u64
values. Thanks @cuviper! - We now offer a
par_chars()
method on strings for iterating over characters in parallel. Thanks @cuviper! - We now have new demos: a traveling salesman problem solver as well as matrix multiplication. Thanks @nikomatsakis, @edre!
- We are now documenting our minimum rustc requirement (currently v1.12.0). We will attempt to maintain compatibility with rustc stable v1.12.0 as long as it remains convenient, but if new features are stabilized or added that would be helpful to Rayon, or there are bug fixes that we need, we will bump to the most recent rustc. Thanks @cuviper!
- The
reduce()
functionality now has better inlining. Thanks @bluss! - The
join()
function now has some documentation. Thanks @gsquire! - The project source has now been fully run through rustfmt. Thanks @ChristopherDavenport!
- Exposed helper methods for accessing the current thread index. Thanks @bholley!
- Breaking change: The
reduce
method has been vastly simplified, andreduce_with_identity
has been deprecated. - Breaking change: The
fold
method has been changed. It used to always reduce the values, but now instead it is a combinator that returns a parallel iterator which can itself be reduced. See the docs for more information. - The following parallel iterator combinators are now available (thanks @cuviper!):
find_any()
: similar tofind
on a sequential iterator, but doesn't necessarily return the first matching itemposition_any()
: similar toposition
on a sequential iterator, but doesn't necessarily return the index of first matching itemany()
,all()
: just like their sequential counterparts
- The
count()
combinator is now available for parallel iterators. - We now build with older versions of rustc again (thanks @durango!),
as we removed a stray semicolon from
thread_local!
. - Various improvements to the (unstable)
scope()
API implementation.
- Parallel iterators now offer an adaptive weight scheme,
which means that explicit weights should no longer
be necessary in most cases! Thanks @cuviper!
- We are considering removing weights or changing the weight mechanism before 1.0. Examples of scenarios where you still need weights even with this adaptive mechanism would be great. Join the discussion at #111.
- New (unstable) scoped threads API, see
rayon::scope
for details.- You will need to supply the cargo feature
unstable
.
- You will need to supply the cargo feature
- The various demos and benchmarks have been consolidated into one
program,
rayon-demo
. - Optimizations in Rayon's inner workings. Thanks @emilio!
- Update
num_cpus
to 1.0. Thanks @jamwt! - Various internal cleanup in the implementation and typo fixes. Thanks @cuviper, @Eh2406, and @spacejam!
- Updated crates.io metadata.
- New
chain
combinator for parallel iterators. Option
,Result
, as well as many more collection types now have parallel iterators.- New mergesort demo.
- Misc fixes.
Thanks to @cuviper, @edre, @jdanford, @frewsxcv for their contributions!
- Make use of latest versions of catch-panic and various fixes to panic propagation.
- Add new prime sieve demo.
- Add
cloned()
andinspect()
combinators. - Misc fixes for Rust RFC 1214.
Thanks to @areilb1, @Amanieu, @SharplEr, and @cuviper for their contributions!
- Expanded
par_iter
APIs now available:into_par_iter
is now supported on vectors (taking ownership of the elements)
- Panic handling is much improved:
- if you use the Nightly feature, experimental panic recovery is available
- otherwise, panics propagate out and poision the workpool
- New
Configuration
object to control number of threads and other details - New demos and benchmarks
- try
cargo run --release -- visualize
indemo/nbody
:)- Note: a nightly compiler is required for this demo due to the
use of the
+=
syntax
- Note: a nightly compiler is required for this demo due to the
use of the
- try
Thanks to @bjz, @cuviper, @Amanieu, and @willi-kappler for their contributions!
No release notes were being kept at this time.