Rust-GPU/rust-cuda
Rust Cuda
rust-cuda is a GPU programming framework and device compiler that allows for the development and execution of high-performance kernels on NVIDIA hardware using Rust. It provides a driver wrapper to manage device memory allocation and kernel launching, effectively serving as a system for writing GPU compute logic without relying on C++.
The project includes a compute library with hardware-optimized primitives for neural network acceleration and hardware-accelerated raytracing. It utilizes a compilation toolchain that translates source code into a low-level intermediate representation for execution on graphics processors.
The framework covers device resource management, kernel development, and the simulation of high-precision integer operations. It also supports device-side random number generation and target-specific compute optimizations.
Pre-configured container images are available to simplify the provisioning of the compiler toolchain and development environment across different hardware architectures.
Features
- GPU Compilers - Provides a complete toolchain that translates Rust source code into optimized NVIDIA GPU hardware instructions.
- GPU Kernel Programming - Provides a framework for writing and executing high-performance GPU kernels using Rust.
- Compute Libraries - Ships a compute library with hardware-optimized primitives for neural networks and raytracing.
- GPU Resource Management - Implements a high-level driver interface to manage the lifecycle and allocation of GPU buffers and kernels.
- CUDA Driver Wrappers - Provides a programmatic abstraction for controlling NVIDIA GPU device memory and kernel execution.
- GPU Kernel Development - Provides a framework for managing thread indices and memory allocation to create device-side logic.
- CUDA Kernel Compilers - Compiles GPU source code into target-specific CUDA kernels with a Rust interface.
- GPU Compute Toolchains - Provides an integrated development environment and compiler for writing high-performance Rust compute kernels.
- Intermediate Representations - Compiles Rust source code into a PTX-based intermediate representation for GPU execution.
- CUDA Accelerated Neural Networks - Implements hardware-optimized primitives to accelerate deep learning workloads on NVIDIA GPUs.
- GPU Kernel Primitives - Provides highly tuned GPU kernels for standard deep learning operations tuned for tensor cores.
- Raytracing Primitives - Interfaces directly with specialized GPU circuits to compute light paths and perform image denoising.
- Raytracing Simulations - Uses specialized GPU circuits to implement physical light transport and path tracing simulations.
- Real-Time Raytracing - Implements real-time raytracing techniques using specialized hardware circuits for physical light transport.
- Hardware Driver API Mappings - Provides mappings that link high-level programmatic calls to native NVIDIA hardware driver and runtime APIs.
- Binary Instruction Execution - Executes low-level binary instructions on graphics hardware using a specialized intermediate representation.
- Compilation Target Specifications - Defines specific hardware compute capabilities during compilation to optimize machine code for target versions.
- Hardware-Targeted Compilation - Configures compilation rules to optimize binaries for specific GPU hardware revisions and architectures.
