Module kernels 

GPU/CPU compute kernels for physics simulation.

This module groups all low-level compute kernels. Each sub-module exposes a CPU-mock implementation that mirrors a GPU kernel in its data layout and dispatch model, but executes on the CPU using Rayon for parallelism.

Modules

broadphase

Broadphase AABB kernels for parallel overlap detection.

group_sizes

Typical work-group sizes used by NVIDIA/AMD GPUs.

md_force

Molecular dynamics force kernels.

rigid

Auto-generated module structure

sph

SPH (Smoothed Particle Hydrodynamics) compute kernels.

Structs

DispatchDims

Describes the 3-D work-group dispatch dimensions for a kernel launch.

KernelPerfCounters

Lightweight performance counters attached to a single kernel invocation.

Enums

KernelFamily

Identifier for a built-in kernel family.

Functions

dispatch_size_1d

Compute the 1-D dispatch size needed to cover n items with group_size threads per work-group.

smem_bytes_matmul

Calculate the shared-memory footprint (bytes) for a tiled matrix-multiply kernel with tiles of size tile × tile of T-sized elements.

workgroup_barrier

Simulated GPU barrier: in CPU mock this is a no-op but documents synchronisation points for future GPU backend porting.