gpu-simulation

Rex is a JAX-powered framework for sim-to-real robotics.

Implementation of curl noise for particles simulated on GPU with OpenGL

A high performance framework for radiation therapy simulation and numerical solutions for kinetic equations.

Learn and understand compute shader operations and control flow.

基于 JAX 的全张量化 GPU 炸弹人模拟器与纯自博弈强化学习框架。实测对战人类不败率 96%，对战各类高阶规则敌人不败率 93.8%+；环境、训练脚本与权重全开源。

CASLab-GPU simulator in SystemC

Module for GPU-Accelerated Kinetic Wealth Exchange Models on Complex Networks

AeroManip-VLA: Scalable Vision-Language-Action Learning for Aerial Manipulation with RL-Generated Demonstrations

Hardware-Accelerated Cryptographic RAG System — GPU simulation of SHA-256 hash engine with AES-256-GCM encryption and Merkle tree integrity. NOT real ASIC hardware.

Underwater Corals Scene 1:1 镜像、中文实现分析与效果实验室

A GPU-simulated field of 100,000+ glowing particles morphing between nine shapes, controllable by mouse, keyboard, or your webcam-tracked hand. React 19 + Three.js + custom GLSL.

JSBSim 1.3.0's F-16 as a hand-written CUDA kernel: 2.95 billion aircraft-frames/s on one GPU, verified frame by frame against JSBSim

A Go-centric systems engineering lab simulating HPC batch scheduling (Slurm), bare-metal BMCs (Redfish), collective GPU topologies (NCCL), and custom Kubernetes scaling operators for local macOS development.

Java based GPU simulator using Java 21 virtual threads and the Strategy pattern. Team project, Computer Architecture course, OTH Regensburg.

Keyframe jump optimization for the Unitree H1-2 with batched CMA-ES in Genesis

Bringing a quadruped Isaac Lab does not ship (Barkour vB) through URDF conversion, a manager-based RL environment, PPO training, evaluation and a simulator-free deploy runtime. 1388x parallel scaling measured on a 6 GB laptop GPU.

Make a beefy NVIDIA GPU pretend to be a smaller consumer card. MIG + clock-lock + memory-fraction cosplay.

Základy počítačové grafiky - Projekt - Vykreslovací řetězec a programování shaderů

A lightweight cli utlitiy tool acting as architectural validator

Cycle-level performance and power of real Hopper LLM kernels with Accel-Sim 2.0 + AccelWattch, run on a Mac; traced Qwen2.5-0.5B layer on H100; split-K deadlock report (accel-sim#561)