AMD ROCm-Simulation#
2025-12-15
3 min read time
ROCm-Simulation is an open-source software toolkit built on the ROCm™ platform, enabling high-performance physical simulations and advanced computational graphics on AMD GPUs. This toolkit enables workloads such as scientific computing, computer graphics, robotics, and AI-driven simulation to run on AMD Instinct™ GPUs and benefit from the optimizations available on those GPUs. ROCm-Simulation builds on the core ROCm libraries to combine frameworks and specialized libraries that accelerate physics-based and numerical simulations.
These tools use ROCm’s HIP runtime, optimized math libraries, and PyTorch integration to deliver high throughput for compute-intensive tasks. This provides you with efficient, scalable solutions for real-time and offline simulation workloads. Physical simulation workloads, such as fluid mechanics, rigid-body dynamics, and volumetric rendering, require significant computational resources. By leveraging ROCm’s open-source GPU stack together with the AMD Instinct product line, you gain performance from optimized kernels, flexibility from integration with Python and machine learning frameworks, and scalability with multi-GPU clusters and high- performance computing (HPC) support.
ROCm-Simulation provides a cohesive set of libraries and frameworks that support the simulation workflow, from physics kernels and numerical solvers to rendering and multi-GPU scaling. Each component is optimized for GPU performance on AMD Instinct GPUs while offering Python-friendly APIs and integrations with popular tools such as PyTorch, so that you can plug simulation workloads into existing research and production pipelines with minimal friction.
The ROCm Simulation Domain includes the following components:
Taichi Lang is an open-source, imperative, and parallel programming language embedded in Python, designed for high-performance numerical computation and real-time physical simulation. It uses just-in-time (JIT) compilation frameworks such as LLVM to accelerate compute-intensive Python code by compiling it into optimized GPU or CPU instructions. Taichi Lang is widely used in domains such as fluid dynamics, particle-based simulations, robotics, computer vision, augmented reality, artificial intelligence, and visual effects for gaming and film.
GSplat (Gaussian splatting) is an open-source library for GPU-accelerated differentiable rasterization of 3D Gaussians with Python bindings. It is inspired by the SIGGRAPH paper “3D Gaussian Splatting for Real-Time Rendering of Radiance Fields”. The ROCm-enabled release of GSplat is built on top of PyTorch, enabling innovators working at the intersection of computer graphics, machine learning, and 3D vision to leverage GPU acceleration for building, research, and innovation with Gaussian Splatting.
To contribute to the documentation, see Contributing to ROCm.
You can find licensing information on the Licensing page.