CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased - 2024-10-08

Added

• Between struct.
• SimdInteraction, ReduceSimdInteraction, BarnesHutInteraction and TreeInteraction traits.
• SequentialCompute, ParallelCompute and GpuCompute traits.
• sequential::Interactions and sequential::Interactions iterator structs.
• Acceleration and AccelerationSoftened structs.
• Position and Mass traits.
• ToSimd, TreeData, AccelerationAt, AccelerationPaired and AccelerationGPU helper traits for the implementation of xInteration traits for Acceleration and AccelerationSoftened structs.
• glam, ultraviolet and nalgebra features, required to enable implementations using these libraries.

Changed

• Renamed ComputeMethod trait to Interaction.
• Built-in algorithms (BruteForce, BarnesHut, ...) now generic over the interaction to compute which allows them to be more unified. BruteForceScalar is now named simply BruteForce and can compute over scalar or simd types. BruteForceSimd converts scalar types to simd types before computing interactions.
• Renamed ParticleTree to RootedOrthtree, ParticleOrdered to Ordered, and ParticleReordered to Reordered. These storages are also generic over their particles.
• Renamed PointMass to GravitationalField and renamed field mass to m. Removed most methods which are now implemented with the helper traits.
• Moved sequential, parallel and gpu modules to the root of the crate.

Removed

• Particle trait, replaced by the Position and Mass traits.
• math module and its traits.
• accelerations method on iterators. This method implicitly collected the particles in a Reordered storage, which is now required to be explicit.
• ParticleSystem struct.
• BruteForceSoftened variants. Softening information is now carried by the interaction.

0.7.0 - 2024-03-21

Added

• ParticleSystem, ParticleTree, ParticleOrdered and ParticleReordered structs used for storages in built-in compute methods.
• Array, Zero, One, Infinty, FloatOps, Float, IntoArray, FloatVectorOps, FloatVector, SIMD, SIMDElement, Reduce, FromPrimitive and AsPrimitive traits for math operation abstractions.
• force_scalar, force_simd methods and various constructors for PointMass.
• ScalarArray trait to bind a FloatVector and an array for Particle to PointMass conversion.
• Marker ReorderedCompute trait.
• Softened variants for all compute methods.
• GpuResources struct used for gpu::BruteForce.
• compute_shared.wgsl compute shader.
• MemoryStrategy to allow user selection between shared and global memory for massive particles within the compute shader used for gpu::BruteForce.

Changed

• Built-in compute methods implemented with the different provided storage structs.
• ComputeMethod no longer generic over V.
• Renamed WgpuData to WgpuResources and wgpu_data module to gpu_compute.
• gpu_compute (internals of gpu::BruteForce) is public.
• gpu::BruteForce holds a reference to a wgpu::Device, wgpu::Queue and a mutable reference to a GpuResources.
• write_particle_data and read_accelerations not generic and use ultraviolet::Vec3.
• read_accelerations is async.
• Update wgpu to 0.17
• Renamed BruteForce compute methods to BruteForceScalar for better differentiation with simd variants.
• BruteForceSIMD generic over the lane count.
• build_node takes position and compute functions instead of using traits and computes a square BoundingBox. Use build_node_with to provide a specific BoundingBox.
• Particle trait only has the Array associated type and position expects this array type.
• Deriving Particle requires #[dim] attribute.
• accelerations returns an iterator of arrays.
• Merge algorithms and compute_method in one module with the name of the latter.
• The size of SizedOrthant is no longer generic and is instead a BoundingBox
• 3D f32 vectors now fully scalar (unlike glam::Vec3A). Use BruteForceSIMD compute methods for simd benefits.

Removed

• glam dependency.
• BruteForcePairsAlt and BruteForcePairsCore compute methods.
• MassiveAffected, MassiveAffectedSIMD and ParticleSet structs.
• Storage, Compute, TreeData, Positionable and BarnesHutTree traits.
• acceleration_* and total_acceleration_* methods for PointMass.
• vector module and its members.

0.6.1 - 2023-07-19

Added

• acceleration_* and total_acceleration_* methods for PointMass to calculate the acceleration between particles of vectors implementing simd and internal.

Changed

• Built-in compute methods use total_acceleration_* methods when applicable.
• sequential::BruteForcePairs is more efficient with less allocations.

0.6.0 - 2023-07-14

Added

• sequential::BruteForcePairsAlt compute method.
• parallel::BruteForceSIMD and sequential::BruteForceSIMD compute methods making use of explicit SIMD instructions for major performance benefits on compatible platforms using ultraviolet.
• Tree, Node, SizedOrthant, BoundingBox structs and TreeData, Subdivide, Positionable, BarnesHutTree traits backing BarnesHut compute methods.
• PointMass struct representing a particle for built-in storages.
• internal::Scalar and internal::Vector traits to help genericity of built-in non-SIMD compute methods.
• simd::IntoVectorElement, simd::SIMD, simd::Scalar, simd::Vector and simd::ReduceAdd traits to help genericity of built-in SIMD compute methods.
• MassiveAffected and ParticleSet structs implementing Storage backing non-SIMD compute methods.
• MassiveAffectedSIMD struct implementing Storage backing SIMD compute methods.
• Storage trait for inputs of ComputeMethod::compute.
• Compute trait extending Iterator with a compute method used by the Accelerations trait.
• algorithms module.

Changed

• renamed sequential::BruteForce to sequential::BruteForcePairs. sequential::BruteForce is changed to a naive implementation iterating over all pairs.
• Compute trait renamed to Accelerations, no longer generic and only returns the computed accelerations. Zip its output with your collection instead.
• ComputeMethod generic over a storage and output type.
• ComputeMethod::compute expects a storage type.
• ComputeMethod has Output associated type returned by ComputeMethod::compute.
• Vector trait renamed to IntoVectorArray and its associated type to Vector and moved to internal submodule.
• InternalVector trait renamed to Vector and moved with Scalar trait to internal submodule.
• compute_method no longer glob imported in prelude.
• tree module and submodules made public.
• vector module made public and part of algorithms.
• Built-in compute methods moved to algorithms module.

Removed

• Accelerations iterator.
• MapCompute trait.

0.5.2 - 2023-05-16

Changed

• Use clone instead of copy for tuple particles.

Fixed

• NaN acceleration output when particles are in a certain mass order for gpu::BruteForce.

0.5.1 - 2023-03-22

Changed

• iterators module renamed to iterator.

Fixed

• Panic when BarnesHut tree is empty.

0.5.0 - 2023-03-22

Added

• BarnesHut compute method available for 2D and 3D float vectors. Available in sequential and parallel versions (only force computation is parallelized).
• Tuples of a position and a gravitational parameter implement Particle.
• accelerations method for iterators of particles that returns the particle and its computed acceleration.
• map_accelerations method for all iterators that returns the item and its computed acceleration using the mapped value (that implements Particle).
• Accelerations iterator obtained through the accelerations and map_accelerations methods available on iterators.

Changed

• ComputeMethod::compute takes slice of all the particles instead of two vectors, massive and massless.
• Generic brute-force compute methods implemented directly using Normed trait for internal vectors.
• Vector trait generic over full array type instead of dimension and scalar.
• 2D f32 vectors use glam::Vec2 instead of glam::Vec3A.
• gpu::BruteForce only available for 3D f32 vectors.

Removed

• ParticleSet. Use accelerations and map_accelerations methods on iterators instead.

0.4.0 - 2023-02-22

Added

• result_mut method of ParticleSet that returns mutable references and acceleration.
• ComputeMethod trait defining how the acceleration of particles is computed.
• Sequential, parallel and GPU BruteForce ComputeMethod implementations (structs in their respective modules).
• Scalar trait to represent components of a vector and the type of the gravitational parameter.

Changed

• result method of ParticleSet returns immutable references. Use result_mut instead for mutable references.
• Vector trait is now generic with a scalar to represent vector types of any float.
• A dimension is no longer required to be bound to a given vector by the user.
• result and accelerations methods of ParticleSet take a ComputeMethod as a parameter.
• Associated type Vector of Particle does not need to be wrapped in VectorDescriptor.
• Particle implementations require associated type Scalar.

Removed

• #[dim] attribute and particle attribute macro.
• Descriptor, IntoVector and FromVector traits.
• VectorDescriptor wrapper.

0.3.1 - 2022-11-26

Changed

• Improve documentation.

0.3.0 - 2022-10-01

Added

• New useful methods for iterating over the ParticleSet.
• particle attribute macro to derive Particle. Its argument is the dimension of the vector type used for the position.
• Descriptor trait for the type of an arbitrary vector.
• VectorDescriptor struct for an arbitrary vector and its dimension.
• FromVector and IntoVector traits to convert between an arbitrary vector and a SIMD vector for computations.

Changed

• Return tuple of result in ParticleSet is reversed.
• Associated type Vector of Particle now requires vector type to be wrapped with a VectorDescriptor, binding a dimension to it.
• Deriving Particle requires #[dim] attribute to associate a dimension to the vector type used for the position.
• Vector trait is now generic with a constant and only cares about From & Into arrays implementations.

Removed

• Vector and NormedVector traits.
• Features glam, nalgebra and cgmath.
• normed_vector! macro.

0.2.0 - 2022-09-17

Added

• Vector and NormedVector traits to describe generic vectors.
• Features glam (default feature), nalgebra and cgmath implements Vector and NormedVector for their vector types.
• normed_vector! macro to easily implement (recommended) Vector and NormedVector for a user-defined vector.

Changed

• Particle now generic regarding its position type to allow for any vector to be used (any dimension, SIMD, ...). Uses new Vector and NormedVector traits.

0.1.0 - 2022-08-18

Added

• Initial release.