Crate oxiphysics_gpu 

GPU acceleration backends for the OxiPhysics engine.

This crate provides a GPU compute abstraction layer that can work with any backend, with a CPU fallback as the default implementation. No heavy GPU dependencies (such as wgpu) are required.

Re-exports

pub use compute::BufferHandle;

pub use compute::ComputeBackend;

pub use compute::ComputeKernel;

pub use compute::CpuBackend;

pub use neural_compute::*;

pub use particle_system::*;

pub use sparse_gpu::*;

Modules

bvh

CPU BVH (Bounding Volume Hierarchy) tree for broad-phase acceleration.

cell_list

Auto-generated module structure

collision_gpu

Auto-generated module structure

compute

Auto-generated module structure

compute_pipeline

CPU-side simulation of a wgpu compute pipeline abstraction.

deformable_gpu

Deformable GPU simulation module.

fluid_gpu

GPU fluid simulation: Navier-Stokes, LBM kernels, SPH kernels, vortex confinement, and shallow-water equations — CPU mock backend.

fluid_sim_gpu

Auto-generated module structure

flux_compute

CPU-side parallel flux and advection compute kernels.

gpu_cloth

GPU-accelerated cloth simulation (CPU mock implementation).

gpu_collision_detection

GPU-accelerated collision detection (CPU mock backend via Rayon).

gpu_collision_ext

GPU collision detection extensions (CPU mock backend).

gpu_fem_assembly

GPU-accelerated FEM matrix assembly (CPU mock implementation).

gpu_fluid

GPU-accelerated fluid simulation using a MAC (Marker-and-Cell) grid.

gpu_fluid_euler

GPU Eulerian fluid simulation on a staggered MAC grid (CPU mock backend).

gpu_lbm

GPU-accelerated Lattice Boltzmann Method (D2Q9) — CPU mock backend.

gpu_md_solver

GPU-side molecular dynamics (MD) solver — CPU mock backend.

gpu_mesh_processing

GPU triangle mesh processing (CPU mock implementation).

gpu_neural_solver

GPU neural network solver for physics (CPU mock backend).

gpu_nn

GPU-accelerated neural network compute (CPU mock backend).

gpu_particle_system

GPU-accelerated particle system (CPU mock implementation).

gpu_particles

GPU-style particle system simulated on the CPU.

gpu_ray_tracing

GPU ray tracing (CPU mock implementation).

gpu_reduction

GPU reduction and scan operations (CPU mock implementation).

gpu_rigid

GPU-accelerated rigid body batch simulation (CPU mock).

gpu_sdf

GPU-accelerated Signed Distance Field (SDF) computation (CPU mock).

gpu_sort

GPU-style parallel sorting algorithms (CPU simulation).

gpu_sparse_solver

GPU sparse linear system solver (CPU mock implementation).

gpu_sph_density

GPU-accelerated SPH density computation (CPU mock implementation).

gpu_sph_pressure

GPU SPH pressure solver (CPU mock backend).

gpu_sph_solver

GPU-side SPH solver — CPU mock backend.

gpu_thermal

GPU-accelerated thermal computation (CPU mock backend via Rayon).

gpu_voxel

GPU voxel grid operations (CPU mock implementation).

grid_reduce

GPU-style grid-based reduction and aggregation kernels.

kernels

GPU/CPU compute kernels for physics simulation.

memory

GPU memory management abstractions.

neural_compute

Auto-generated module structure

neural_physics

Neural network-based physics acceleration (CPU mock).

parallel

Auto-generated module structure

parallel_sort

Parallel sorting and reduction utilities (CPU-side, rayon-based).

particle_system

Auto-generated module structure

path_tracer

CPU path tracer using GPU-style algorithms.

pipeline

Auto-generated module structure

ray_marching

Ray marching and Signed Distance Field (SDF) rendering utilities.

ray_tracing_gpu

GPU ray tracing with f32 precision (CPU mock).

raytracing

Auto-generated module structure

scheduler

GPU workload scheduler.

sdf_compute

Auto-generated module structure

shader_registry

Auto-generated module structure

shaders

Auto-generated module structure

sparse_gpu

GPU-ready sparse matrix formats and iterative solvers with data-oriented layouts.

Structs

AnnotatedError

An error annotated with severity and an optional kernel name.

DispatchTimer

A simple timer utility for profiling GPU-like dispatches.

PipelineStageError

A pipeline-stage error: carries the stage name plus the underlying cause.

Enums

Error

Main error type for the gpu module.

ErrorSeverity

Severity level for a GPU error.

Functions

aligned_size

Compute the padded buffer size to meet alignment requirements.

alloc_err

Build a Error::BufferAllocationFailed error.

arg_mismatch_err

Build a Error::KernelArgCountMismatch error.

bandwidth_gb_s

Estimate memory bandwidth in GB/s.

buffer_size_2d

Compute the 2-D buffer size (rows × pitch) for a texture-like allocation.

check

Check a boolean condition; return Err(Error::General(msg)) if false.

clamp_f64

Clamp v to [lo, hi].

collect_errors

Collect multiple errors from a batch dispatch. Returns Ok(()) if the vec is empty, or Err containing the first error otherwise.

cross3

Compute the cross product of two 3-element arrays.

dispatch_count

Compute the optimal work group size for a given total work item count.

dispatch_limit_err

Build a Error::DispatchLimitExceeded error.

dot3

Compute the dot product of two 3-element arrays.

elements_in_budget

Compute the number of elements that fit in a given memory budget.

exclusive_scan

Parallel prefix sum (scan) on a slice of f64 values.

grid_oob_err

Build a Error::GridIndexOutOfBounds error.

inclusive_scan

Inclusive scan: result[i] = sum(data[0..=i]).

index_3d_from_linear

Convert a linear index back to 3D coordinates.

inv_lerp

Inverse lerp: returns t such that lerp(a, b, t) == v, or 0 if a==b.

is_power_of_two

True when value is a power of two (including 1).

length3

Length of a 3-D vector.

lerp

Linear interpolation: a + t*(b-a).

linear_index_3d

Flatten a 3D dispatch (x, y, z) into a linear index, given grid dimensions.

log2_pow2

Log2 of a power-of-two value. Panics in debug mode if v is not a power of two.

next_power_of_two

Round value up to the next power of two.

normalize3

Normalise a 3-D vector. Returns the zero vector if the length is < eps.

reduce_max

Parallel reduce: compute the maximum of all elements.

reduce_min

Parallel reduce: compute the minimum of all elements.

reduce_sum

Parallel reduce: compute the sum of all elements.

reflect3

Reflect vector d about normal n (both assumed normalised).

row_pitch

Stride (in bytes) of a row in a 2-D buffer, given the element count per row and the required alignment.

safe_recip

Safe reciprocal: returns 1/x when |x| > eps, else 0.

safe_sqrt

Safe square root: clamps negative values to 0 before taking sqrt.

shader_err

Build a Error::ShaderCompilationError.

smootherstep

Smoother-step (Ken Perlin’s quintic): 6t⁵ − 15t⁴ + 10t³.

smoothstep

Smooth-step function: 3t² - 2t³ with t = (v - lo) / (hi - lo).

tile_count_2d

Divides a 2-D problem of (width × height) into tiles of (tw × th) and returns (tiles_x, tiles_y).

tile_index_to_2d

Convert a flat tile index back to (tile_x, tile_y) for a grid with tiles_x columns.

total_tiles_2d

Total number of tiles for a 2-D problem.

unsupported_err

Build an Error::UnsupportedFeature.

wrap_angle

Wrap an angle in radians to (-π, π].

Type Aliases

Result

Result type alias