Expand description
GPU Compute Pipeline
This module provides a complete GPU compute subsystem built on top of OpenGL 4.3+
compute shaders via the glow crate. It includes:
- buffer: SSBO management, typed buffers, double-buffered particle buffers, atomic counters, buffer pools, memory tracking, mapped ranges, copy engines.
- dispatch: Compute shader compilation with
#defineinjection, workgroup sizing, 1D/2D/3D dispatch, indirect dispatch, pipeline caching, profiling. - kernels: 10 built-in compute kernels as embedded GLSL source strings.
- sync: Fence synchronization, memory barriers, async compute queue, frame timeline, CPU fallback path, resource transition state machine.
Re-exports§
pub use buffer::TypedBuffer;pub use buffer::ParticleBuffer;pub use buffer::AtomicCounter;pub use buffer::BufferPool;pub use buffer::BufferBarrierType;pub use buffer::MemoryTracker;pub use buffer::MappedRange;pub use buffer::BufferCopyEngine;pub use buffer::BufferHandle;pub use buffer::BufferUsage;pub use dispatch::ShaderSource;pub use dispatch::ComputeProgram;pub use dispatch::WorkgroupSize;pub use dispatch::ComputeDispatch;pub use dispatch::DispatchDimension;pub use dispatch::IndirectDispatchArgs;pub use dispatch::PipelineCache;pub use dispatch::SpecializationConstant;pub use dispatch::ComputeProfiler;pub use dispatch::TimingQuery;pub use kernels::KernelLibrary;pub use kernels::KernelId;pub use kernels::ParticleIntegrateParams;pub use kernels::ParticleEmitParams;pub use kernels::ForceFieldDesc;pub use kernels::MathFunctionType;pub use kernels::FluidDiffuseParams;pub use kernels::HistogramParams;pub use kernels::PrefixSumPlan;pub use kernels::RadixSortPlan;pub use kernels::FrustumCullParams;pub use kernels::SkinningParams;pub use sync::FenceSync;pub use sync::FenceStatus;pub use sync::MemoryBarrierFlags;pub use sync::PipelineBarrier;pub use sync::AsyncComputeQueue;pub use sync::FrameTimeline;pub use sync::CpuFallback;pub use sync::ResourceTransition;pub use sync::ResourceState;
Modules§
- buffer
- SSBO management: typed buffers, double-buffering, atomic counters, buffer pools, barrier types, memory tracking, mapped ranges, copy engine.
- dispatch
- Compute shader compilation, dispatch, pipeline caching, and profiling.
- kernels
- Built-in compute kernels as embedded GLSL source strings.
- sync
- Compute-to-render synchronization: fences, memory barriers, async compute queue, frame timeline, CPU fallback, resource state machine.