Direct Parallelization of VecDeque, SplitVec and FixedVec

Upgrades the concurrent iterator dependency which will enable parallelization of the following additional collections:

• alloc::collections::VecDeque
• orx_split_vec::SplitVec
• orx_fixed_vec::FixedVec

v2: Redesign parallelization traits

Parallel computation through iterators is re-designed.

• Updates in the downstream crates are integrated, specifically the improvements in ConcurrentIter crate, v2.0, are benefited.
• Functional compositions are re-designed. This led to improvements in performance, as well as, simplification of the code and implementations.
  • In order to achieve this, required abstractions over values and arrays (in the mathematical sense) is defined. This allows us to perform on a value or a collection of values using the same generic code, while the actual implementation is monomorphized and optimized under the hood.
  • With this design, we are required to have only four generic computational variants of the iterators without loss of performance:
    • empty -> this is equivalent to the input concurrent iterator source, no computation is defined over it yet.
    • map -> this is a map operation on the input in the classical sense
    • xap -> this on the other hand is a map with superpowers, it is a generalization of (i) one-to-one map, (ii) filter-map and (iii) flat-map operations.
    • xap-filter-xap -> finally, this is a sequence of xap, followed by a filter, followed by another xap.
  • These four variants are sufficient to represent all possible iterator compositions without loss of performance due to abstraction.
• Significant performance improvements achieved and reported.
• The concepts defined in orx_iterable are brought to this crate defining traits such as IntoParIter, Parallelizable and ParallelizableCollection.
• Extensibility of input collections through concurrent iterators is defined.
• ParallelRunner is converted into a trait. This allows for customization of the parallel executor. Therefore, it enables improvements in performance in parallel workstreams. Meanwhile, a performant default parallel executor is provided.
• The complexity of collecting ordered or in-arbitrary-order is resolved using the CollectOrder parameter.
• The api is extended by enabling new methods such as inspect.
• Contribution guideline is revised.
• GenericIterator defined mainly for experimentation purposes. GenericIterator is a generalization of sequential iterators, rayon's parallel iterators and orx-parallel's parallel iterators. The computation can be defined as a composition of iterator transformations in the same way and can be executed with either of the variants.
• ci.yml is added.
• Fixes #16
• Fixes #26

Upgrade PseudoDefault dependency

Merge pull request #49 from orxfun/Upgrade-PseudoDefault-dependency

Upgrade PseudoDefault dependency

Dual license

Merge pull request #48 from orxfun/dual-license

dual license

Use of Collection constrained PinnedVecs

Upgraded pinned vectors to v3.12

Upgraded dependencies with no_std support

1.11.0

Merge pull request #44 from orxfun/upgraded-dependencies-with-no_std-…

Unordered iteration is revisited

Unordered iteration is revisited

• Upgraded the dependency to concurrent iterator to 1.27.0; see the release notes.
• Syntax for creating a parallel iterator from a sequential iterator is fixed.
  • prior: inputs.into_con_iter().into_par().
  • new: inputs.iter().par()
• Parallel reductions use transformation into ConcurrentIterX to have the performance benefits whenever possible.
• collect_x executions use transformation into ConcurrentIterX to have the performance benefits whenever possible.

ConIterOfIter simplification and performance

Additionally iter-map-collect benchmarks are defined.

Upgrade queue and iter deps

With these upgrades, all failing miri tests pass.

• orx-priority-queue = "1.3"
• orx-concurrent-iter = "1.25"

ConVecIter early exit memory leak issue is fixed

The fix is applied in one of the dependencies, see here: https://github.com/orxfun/orx-concurrent-iter/releases/tag/1.24.0