This repository provides a collection of CUDA-accelerated modules for large-scale vector processing and clustering.
It includes both the K-means Clustering module, optimized for building Inverted File (IVF) indices, and the Load Manager, which efficiently handles massive vector data on GPUs.
These components can be used independently or together as part of a GPU-based vector indexing and search pipeline.
src/
├── kmeans-vector-clustering/ # CUDA-based K-means clustering for IVF index build
│ ├── kmeans_openai.cpp
│ ├── kmeans.cu
│ ├── README.md
│ └── ...
│
├── load-manager/ # GPU Load Manager with tile-based memory control
│ ├── load-manager.cpp
│ ├── load-manager.cu
│ ├── README.md
│ └── ...
│
├── gpu-topk-sort/ # GPU Top-k Sort for vector candidate lists
├── gpu_sort.cpp
├── gpu_sort.cu
├── README.md
└── ...