# CUDA

RightNow AI is an innovative AI-powered CUDA code optimization platform designed to help developers quickly improve GPU performance. Through powerful AI technology, it automatically analyzes CUDA kernels, identifies performance bottlenecks, and generates optimized code, saving significant time and effort compared to manual optimization. The platform supports various optimization strategies such as shared memory utilization, thread collaboration, and loop unrolling, achieving up to a 4x performance improvement. It primarily targets developers and enterprises requiring high-performance GPU computing, especially those lacking expertise in GPU optimization. RightNow AI offers various paid plans, including pay-as-you-go, developer, professional, and enterprise packages, to meet the needs of users of different scales.

DeepGEMM is a CUDA library focused on high-performance FP8 matrix multiplication. Through fine-grained scaling and various optimization techniques such as Hopper TMA features, persistent thread specialization, and a fully JIT design, it significantly improves matrix computation performance. Primarily aimed at deep learning and high-performance computing, it's suitable for scenarios requiring efficient matrix operations. It supports NVIDIA Hopper architecture Tensor Cores and demonstrates superior performance across various matrix shapes. DeepGEMM boasts a concise design with a core codebase of approximately 300 lines, making it easy to learn and use while achieving performance comparable to or exceeding expert-optimized libraries. Its open-source and free nature makes it an ideal choice for researchers and developers engaged in deep learning optimization and development.

VisoMaster is a desktop client software focused on video replacement and editing. It leverages advanced AI technology to achieve high-quality replacements in images and videos, creating natural and realistic effects. The software is easy to operate, supports various input and output formats, and enhances processing efficiency through GPU acceleration. VisoMaster's main advantages are its user-friendliness, efficient processing, and high customizability, making it suitable for video creators, post-production professionals, and everyday users with video editing needs. The software is currently provided free of charge to help users quickly generate high-quality video content.

karpathy/llm.c is a project that implements LLM training using a simple C/CUDA implementation. It aims to provide a clean and straightforward reference implementation while also including optimized versions that can approach the performance of PyTorch with drastically reduced code and dependencies. It is currently under development for a direct CUDA implementation, CPU version optimization using SIMD instructions, and support for more modern architectures such as Llama2 and Gemma.