Instant neural graphics primitives: lightning fast NeRF and more

how to optimize some algorithm in cuda.

Deformable ConvNets V2 (DCNv2) in PyTorch

[MICRO'23, MLSys'22] TorchSparse: Efficient Training and Inference Framework for Sparse Convolution on GPUs.

Quantized Attention that achieves speedups of 2.1-3.1x and 2.7-5.1x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.

Python library for Room Impulse Response (RIR) simulation with GPU acceleration

Neighborhood Attention Extension. Bringing attention to a neighborhood near you!

Deforming kernels to adapt towards object deformation. In ICLR 2020.

A high performance CUDA implementation of Scan Matching via the Iterative Closest Point Algorithm

高性能编程 笔记

Code for Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inference (CVPR2020)

Introduction to CUDA programming

An GPU/CUDA implementation of the Hungarian algorithm

CUDA implementation of parallel radix sort using Blelloch scan

Efficient CUDA Stream Compaction Library

Development a customized op in TensorFlow for convolution with sparse kernel

CUDA implementation of exclusive prefix sum via Blelloch's algorithm

MIT-licensed stand-alone CUDA utility functions.

Parallel Prefix Sum (Scan) with CUDA.

CUDA implementation of "A Fast Hybrid Approach for Stream Compaction on GPUs" by Rego, Sang and Yu