Struct ParStrided

pub struct ParStrided<T> { /* private fields */ }

A parallel strided iterator over tensor elements.

This struct provides mutable access to tensor elements with strided access patterns optimized for parallel processing.

Implementations

impl<T: CommonBounds> ParStrided<T>

pub fn shape(&self) -> &Shape

Retrieves the shape of the tensor.

Returns

A reference to the Shape struct representing the tensor’s dimensions.

pub fn strides(&self) -> &Strides

Retrieves the strides of the tensor.

Returns

A reference to the Strides struct representing the tensor’s stride information.

pub fn new<U: TensorInfo<T>>(tensor: U) -> Self

Creates a new ParStrided instance from a given tensor.

This constructor initializes the ParStrided iterator by determining the appropriate number of threads based on the tensor’s outer loop size and Rayon’s current number of threads. It then divides the iteration workload into intervals for parallel execution.

Arguments

• tensor - The tensor implementing the TensorInfo<T> trait to iterate over mutably.

Returns

A new instance of ParStrided initialized with the provided tensor.

pub fn par_strided_fold<ID, F>( self, identity: ID, fold_op: F, ) -> ParStridedFold<Self, ID, F>

where F: Fn(ID, T) -> ID + Sync + Send + Copy, ID: Sync + Send + Copy,

Performs a parallel fold (reduce) operation over the tensor elements.

This method applies a folding function fold_op to accumulate tensor elements into an initial identity value identity. It leverages parallel iteration to perform the fold operation efficiently.

Type Parameters

• ID - The type of the accumulator.
• F - The folding function.

Arguments

• identity - The initial value for the accumulator.
• fold_op - A function that takes the current accumulator and an element, returning the updated accumulator.

Returns

A ParStridedFold instance that represents the fold operation.

pub fn strided_map<'a, F, U>( self, f: F, ) -> ParStridedMap<'a, ParStrided<T>, T, F>

where F: Fn((&mut U, T)) + Sync + Send + 'a, U: CommonBounds,

Transforms the zipped iterators by applying a provided function to their items.

This method allows for element-wise operations on the zipped iterators by applying func to each item.

Type Parameters

• 'a - The lifetime associated with the iterators.
• F - The function to apply to each item.
• U - The output type after applying the function.

Arguments

• f - A function that takes an item from the zipped iterator and returns a transformed value.

Returns

A ParStridedMap instance that applies the provided function during iteration.

Trait Implementations

impl<T: Clone> Clone for ParStrided<T>

fn clone(&self) -> ParStrided<T>

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<T: CommonBounds> IterGetSet for ParStrided<T>

type Item = T

The type of the iterator’s elements.

fn set_end_index(&mut self, end_index: usize)

set the end index of the iterator, this is used when rayon perform data splitting

fn set_intervals(&mut self, intervals: Arc<Vec<(usize, usize)>>)

set the chunk intervals of the iterator, we chunk the outer loop

fn set_strides(&mut self, strides: Strides)

set the strides for the iterator, we call this method normally when we do broadcasting

fn set_shape(&mut self, shape: Shape)

set the shape for the iterator, we call this method normally when we do broadcasting

fn set_prg(&mut self, prg: Vec<i64>)

set the loop progress for the iterator

fn intervals(&self) -> &Arc<Vec<(usize, usize)>>

get the intervals of the iterator

fn strides(&self) -> &Strides

get the strides of the iterator

fn shape(&self) -> &Shape

get the shape of the iterator

fn layout(&self) -> &Layout

get the layout of the iterator

fn broadcast_set_strides(&mut self, shape: &Shape)

set the strides for all the iterators

fn outer_loop_size(&self) -> usize

get the outer loop size

fn inner_loop_size(&self) -> usize

get the inner loop size

fn next(&mut self)

update the loop progress

fn inner_loop_next(&mut self, index: usize) -> Self::Item

get the next element of the inner loop

impl<T: CommonBounds> ParStridedIteratorZip for ParStrided<T>

fn zip<'a, C>(self, other: C) -> ParStridedZip<'a, Self, C>

where C: UnindexedProducer + 'a + IterGetSet + ParallelIterator + ShapeManipulator, <C as IterGetSet>::Item: Send, Self: UnindexedProducer + ParallelIterator + ShapeManipulator, <Self as IterGetSet>::Item: Send,

Combines this iterator with another iterator, enabling simultaneous parallel iteration.

impl<T> ParallelIterator for ParStrided<T>

where T: CommonBounds, T::Vec: Send,

type Item = T

The type of item that this parallel iterator produces. For example, if you use the for_each method, this is the type of item that your closure will be invoked with.

fn drive_unindexed<C>(self, consumer: C) -> C::Result

where C: UnindexedConsumer<Self::Item>,

Internal method used to define the behavior of this parallel iterator. You should not need to call this directly.

fn for_each<OP>(self, op: OP)

where OP: Fn(Self::Item) + Sync + Send,

Executes OP on each item produced by the iterator, in parallel.

fn for_each_with<OP, T>(self, init: T, op: OP)

where OP: Fn(&mut T, Self::Item) + Sync + Send, T: Send + Clone,

Executes OP on the given init value with each item produced by the iterator, in parallel.

fn for_each_init<OP, INIT, T>(self, init: INIT, op: OP)

where OP: Fn(&mut T, Self::Item) + Sync + Send, INIT: Fn() -> T + Sync + Send,

Executes OP on a value returned by init with each item produced by the iterator, in parallel.

fn try_for_each<OP, R>(self, op: OP) -> R

where OP: Fn(Self::Item) -> R + Sync + Send, R: Try<Output = ()> + Send,

Executes a fallible OP on each item produced by the iterator, in parallel.

fn try_for_each_with<OP, T, R>(self, init: T, op: OP) -> R

where OP: Fn(&mut T, Self::Item) -> R + Sync + Send, T: Send + Clone, R: Try<Output = ()> + Send,

Executes a fallible OP on the given init value with each item produced by the iterator, in parallel.

fn try_for_each_init<OP, INIT, T, R>(self, init: INIT, op: OP) -> R

where OP: Fn(&mut T, Self::Item) -> R + Sync + Send, INIT: Fn() -> T + Sync + Send, R: Try<Output = ()> + Send,

Executes a fallible OP on a value returned by init with each item produced by the iterator, in parallel.

fn count(self) -> usize

Counts the number of items in this parallel iterator.

fn map<F, R>(self, map_op: F) -> Map<Self, F>

where F: Fn(Self::Item) -> R + Sync + Send, R: Send,

Applies map_op to each item of this iterator, producing a new iterator with the results.

fn map_with<F, T, R>(self, init: T, map_op: F) -> MapWith<Self, T, F>

where F: Fn(&mut T, Self::Item) -> R + Sync + Send, T: Send + Clone, R: Send,

Applies map_op to the given init value with each item of this iterator, producing a new iterator with the results.

fn map_init<F, INIT, T, R>( self, init: INIT, map_op: F, ) -> MapInit<Self, INIT, F>

where F: Fn(&mut T, Self::Item) -> R + Sync + Send, INIT: Fn() -> T + Sync + Send, R: Send,

Applies map_op to a value returned by init with each item of this iterator, producing a new iterator with the results.

fn cloned<'a, T>(self) -> Cloned<Self>

where T: 'a + Clone + Send, Self: ParallelIterator<Item = &'a T>,

Creates an iterator which clones all of its elements. This may be useful when you have an iterator over &T, but you need T, and that type implements Clone. See also copied().

fn copied<'a, T>(self) -> Copied<Self>

where T: 'a + Copy + Send, Self: ParallelIterator<Item = &'a T>,

Creates an iterator which copies all of its elements. This may be useful when you have an iterator over &T, but you need T, and that type implements Copy. See also cloned().

fn inspect<OP>(self, inspect_op: OP) -> Inspect<Self, OP>

where OP: Fn(&Self::Item) + Sync + Send,

Applies inspect_op to a reference to each item of this iterator, producing a new iterator passing through the original items. This is often useful for debugging to see what’s happening in iterator stages.

fn update<F>(self, update_op: F) -> Update<Self, F>

where F: Fn(&mut Self::Item) + Sync + Send,

Mutates each item of this iterator before yielding it.

fn filter<P>(self, filter_op: P) -> Filter<Self, P>

where P: Fn(&Self::Item) -> bool + Sync + Send,

Applies filter_op to each item of this iterator, producing a new iterator with only the items that gave true results.

fn filter_map<P, R>(self, filter_op: P) -> FilterMap<Self, P>

where P: Fn(Self::Item) -> Option<R> + Sync + Send, R: Send,

Applies filter_op to each item of this iterator to get an Option, producing a new iterator with only the items from Some results.

fn flat_map<F, PI>(self, map_op: F) -> FlatMap<Self, F>

where F: Fn(Self::Item) -> PI + Sync + Send, PI: IntoParallelIterator,

Applies map_op to each item of this iterator to get nested parallel iterators, producing a new parallel iterator that flattens these back into one.

fn flat_map_iter<F, SI>(self, map_op: F) -> FlatMapIter<Self, F>

where F: Fn(Self::Item) -> SI + Sync + Send, SI: IntoIterator, <SI as IntoIterator>::Item: Send,

Applies map_op to each item of this iterator to get nested serial iterators, producing a new parallel iterator that flattens these back into one.

fn flatten(self) -> Flatten<Self>

where Self::Item: IntoParallelIterator,

An adaptor that flattens parallel-iterable Items into one large iterator.

fn flatten_iter(self) -> FlattenIter<Self>

where Self::Item: IntoIterator, <Self::Item as IntoIterator>::Item: Send,

An adaptor that flattens serial-iterable Items into one large iterator.

fn reduce<OP, ID>(self, identity: ID, op: OP) -> Self::Item

where OP: Fn(Self::Item, Self::Item) -> Self::Item + Sync + Send, ID: Fn() -> Self::Item + Sync + Send,

Reduces the items in the iterator into one item using op. The argument identity should be a closure that can produce “identity” value which may be inserted into the sequence as needed to create opportunities for parallel execution. So, for example, if you are doing a summation, then identity() ought to produce something that represents the zero for your type (but consider just calling sum() in that case).

fn reduce_with<OP>(self, op: OP) -> Option<Self::Item>

where OP: Fn(Self::Item, Self::Item) -> Self::Item + Sync + Send,

Reduces the items in the iterator into one item using op. If the iterator is empty, None is returned; otherwise, Some is returned.

fn try_reduce<T, OP, ID>(self, identity: ID, op: OP) -> Self::Item

where OP: Fn(T, T) -> Self::Item + Sync + Send, ID: Fn() -> T + Sync + Send, Self::Item: Try<Output = T>,

Reduces the items in the iterator into one item using a fallible op. The identity argument is used the same way as in reduce().

fn try_reduce_with<T, OP>(self, op: OP) -> Option<Self::Item>

where OP: Fn(T, T) -> Self::Item + Sync + Send, Self::Item: Try<Output = T>,

Reduces the items in the iterator into one item using a fallible op.

fn fold<T, ID, F>(self, identity: ID, fold_op: F) -> Fold<Self, ID, F>

where F: Fn(T, Self::Item) -> T + Sync + Send, ID: Fn() -> T + Sync + Send, T: Send,

Parallel fold is similar to sequential fold except that the sequence of items may be subdivided before it is folded. Consider a list of numbers like 22 3 77 89 46. If you used sequential fold to add them (fold(0, |a,b| a+b), you would wind up first adding 0 + 22, then 22 + 3, then 25 + 77, and so forth. The parallel fold works similarly except that it first breaks up your list into sublists, and hence instead of yielding up a single sum at the end, it yields up multiple sums. The number of results is nondeterministic, as is the point where the breaks occur.

fn fold_with<F, T>(self, init: T, fold_op: F) -> FoldWith<Self, T, F>

where F: Fn(T, Self::Item) -> T + Sync + Send, T: Send + Clone,

Applies fold_op to the given init value with each item of this iterator, finally producing the value for further use.

fn try_fold<T, R, ID, F>( self, identity: ID, fold_op: F, ) -> TryFold<Self, R, ID, F>

where F: Fn(T, Self::Item) -> R + Sync + Send, ID: Fn() -> T + Sync + Send, R: Try<Output = T> + Send,

Performs a fallible parallel fold.

fn try_fold_with<F, T, R>(self, init: T, fold_op: F) -> TryFoldWith<Self, R, F>

where F: Fn(T, Self::Item) -> R + Sync + Send, R: Try<Output = T> + Send, T: Clone + Send,

Performs a fallible parallel fold with a cloneable init value.

fn sum<S>(self) -> S

where S: Send + Sum<Self::Item> + Sum,

Sums up the items in the iterator.

fn product<P>(self) -> P

where P: Send + Product<Self::Item> + Product,

Multiplies all the items in the iterator.

fn min(self) -> Option<Self::Item>

where Self::Item: Ord,

Computes the minimum of all the items in the iterator. If the iterator is empty, None is returned; otherwise, Some(min) is returned.

fn min_by<F>(self, f: F) -> Option<Self::Item>

where F: Sync + Send + Fn(&Self::Item, &Self::Item) -> Ordering,

Computes the minimum of all the items in the iterator with respect to the given comparison function. If the iterator is empty, None is returned; otherwise, Some(min) is returned.

fn min_by_key<K, F>(self, f: F) -> Option<Self::Item>

where K: Ord + Send, F: Sync + Send + Fn(&Self::Item) -> K,

Computes the item that yields the minimum value for the given function. If the iterator is empty, None is returned; otherwise, Some(item) is returned.

fn max(self) -> Option<Self::Item>

where Self::Item: Ord,

Computes the maximum of all the items in the iterator. If the iterator is empty, None is returned; otherwise, Some(max) is returned.

fn max_by<F>(self, f: F) -> Option<Self::Item>

where F: Sync + Send + Fn(&Self::Item, &Self::Item) -> Ordering,

Computes the maximum of all the items in the iterator with respect to the given comparison function. If the iterator is empty, None is returned; otherwise, Some(max) is returned.

fn max_by_key<K, F>(self, f: F) -> Option<Self::Item>

where K: Ord + Send, F: Sync + Send + Fn(&Self::Item) -> K,

Computes the item that yields the maximum value for the given function. If the iterator is empty, None is returned; otherwise, Some(item) is returned.

fn chain<C>(self, chain: C) -> Chain<Self, <C as IntoParallelIterator>::Iter>

where C: IntoParallelIterator<Item = Self::Item>,

Takes two iterators and creates a new iterator over both.

fn find_any<P>(self, predicate: P) -> Option<Self::Item>

where P: Fn(&Self::Item) -> bool + Sync + Send,

Searches for some item in the parallel iterator that matches the given predicate and returns it. This operation is similar to find on sequential iterators but the item returned may not be the first one in the parallel sequence which matches, since we search the entire sequence in parallel.

fn find_first<P>(self, predicate: P) -> Option<Self::Item>

where P: Fn(&Self::Item) -> bool + Sync + Send,

Searches for the sequentially first item in the parallel iterator that matches the given predicate and returns it.

fn find_last<P>(self, predicate: P) -> Option<Self::Item>

where P: Fn(&Self::Item) -> bool + Sync + Send,

Searches for the sequentially last item in the parallel iterator that matches the given predicate and returns it.

fn find_map_any<P, R>(self, predicate: P) -> Option<R>

where P: Fn(Self::Item) -> Option<R> + Sync + Send, R: Send,

Applies the given predicate to the items in the parallel iterator and returns any non-None result of the map operation.

fn find_map_first<P, R>(self, predicate: P) -> Option<R>

where P: Fn(Self::Item) -> Option<R> + Sync + Send, R: Send,

Applies the given predicate to the items in the parallel iterator and returns the sequentially first non-None result of the map operation.

fn find_map_last<P, R>(self, predicate: P) -> Option<R>

where P: Fn(Self::Item) -> Option<R> + Sync + Send, R: Send,

Applies the given predicate to the items in the parallel iterator and returns the sequentially last non-None result of the map operation.

fn any<P>(self, predicate: P) -> bool

where P: Fn(Self::Item) -> bool + Sync + Send,

Searches for some item in the parallel iterator that matches the given predicate, and if so returns true. Once a match is found, we’ll attempt to stop process the rest of the items. Proving that there’s no match, returning false, does require visiting every item.

fn all<P>(self, predicate: P) -> bool

where P: Fn(Self::Item) -> bool + Sync + Send,

Tests that every item in the parallel iterator matches the given predicate, and if so returns true. If a counter-example is found, we’ll attempt to stop processing more items, then return false.

fn while_some<T>(self) -> WhileSome<Self>

where Self: ParallelIterator<Item = Option<T>>, T: Send,

Creates an iterator over the Some items of this iterator, halting as soon as any None is found.

fn panic_fuse(self) -> PanicFuse<Self>

Wraps an iterator with a fuse in case of panics, to halt all threads as soon as possible.

fn collect<C>(self) -> C

where C: FromParallelIterator<Self::Item>,

Creates a fresh collection containing all the elements produced by this parallel iterator.

fn unzip<A, B, FromA, FromB>(self) -> (FromA, FromB)

where Self: ParallelIterator<Item = (A, B)>, FromA: Default + Send + ParallelExtend<A>, FromB: Default + Send + ParallelExtend<B>, A: Send, B: Send,

Unzips the items of a parallel iterator into a pair of arbitrary ParallelExtend containers.

fn partition<A, B, P>(self, predicate: P) -> (A, B)

where A: Default + Send + ParallelExtend<Self::Item>, B: Default + Send + ParallelExtend<Self::Item>, P: Fn(&Self::Item) -> bool + Sync + Send,

Partitions the items of a parallel iterator into a pair of arbitrary ParallelExtend containers. Items for which the predicate returns true go into the first container, and the rest go into the second.

fn partition_map<A, B, P, L, R>(self, predicate: P) -> (A, B)

where A: Default + Send + ParallelExtend<L>, B: Default + Send + ParallelExtend<R>, P: Fn(Self::Item) -> Either<L, R> + Sync + Send, L: Send, R: Send,

Partitions and maps the items of a parallel iterator into a pair of arbitrary ParallelExtend containers. Either::Left items go into the first container, and Either::Right items go into the second.

fn intersperse(self, element: Self::Item) -> Intersperse<Self>

where Self::Item: Clone,

Intersperses clones of an element between items of this iterator.

fn take_any(self, n: usize) -> TakeAny<Self>

Creates an iterator that yields n elements from anywhere in the original iterator.

fn skip_any(self, n: usize) -> SkipAny<Self>

Creates an iterator that skips n elements from anywhere in the original iterator.

fn take_any_while<P>(self, predicate: P) -> TakeAnyWhile<Self, P>

where P: Fn(&Self::Item) -> bool + Sync + Send,

Creates an iterator that takes elements from anywhere in the original iterator until the given predicate returns false.

fn skip_any_while<P>(self, predicate: P) -> SkipAnyWhile<Self, P>

where P: Fn(&Self::Item) -> bool + Sync + Send,

Creates an iterator that skips elements from anywhere in the original iterator until the given predicate returns false.

fn collect_vec_list(self) -> LinkedList<Vec<Self::Item>>

Collects this iterator into a linked list of vectors.

fn opt_len(&self) -> Option<usize>

Internal method used to define the behavior of this parallel iterator. You should not need to call this directly.

impl<T: CommonBounds> ShapeManipulator for ParStrided<T>

fn reshape<S: Into<Shape>>(self, shape: S) -> Self

reshape the iterator, we can change the iteration behavior by changing the shape

fn transpose<AXIS: Into<Axis>>(self, axes: AXIS) -> Self

transpose the iterator, we can change the iteration behavior by changing the axes

fn expand<S: Into<Shape>>(self, shape: S) -> Self

expand the iterator, we can change the iteration behavior by changing the shape

impl<T> UnindexedProducer for ParStrided<T>

where T: CommonBounds, T::Vec: Send,

type Item = T

The type of item returned by this producer.

fn split(self) -> (Self, Option<Self>)

Split midway into a new producer if possible, otherwise return None.

fn fold_with<F>(self, folder: F) -> F

where F: Folder<Self::Item>,

Iterate the producer, feeding each element to folder, and stop when the folder is full (or all elements have been consumed).