This is a series of GPU optimization topics. Here we will introduce how to optimize the CUDA kernel in detail. I will introduce several basic kernel optimizations, including: elementwise, reduce, sgemv, sgemm, etc. The performance of these kernels is basically at or near the theoretical limit.

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

MegBA: A GPU-Based Distributed Library for Large-Scale Bundle Adjustment

G3: A Programmable GNN Training System on GPU

Code supporting the WAFR paper "A Performance Analysis of Differential Dynamic Programming on a GPU," and the ICRA workshop follow on work deploying the algorithm onto robot hardware.

GPU Framework for Radio Astronomical Image Synthesis

zkDL, an open source toolkit for zero-knowledge proofs of deep learning powered by CUDA

Lattice-Botlzmann fluid simulator using CUDA and OpenGL

Code supporting the RAL paper "Accelerating Robot Dynamics Gradients on a CPU, GPU, and FPGA"

NAS Parallel Benchmarks for GPU

CUDA Finite Difference Library

GPU accelerated spiking neural network simulator

LBM (Lattice Boltzmann method) software for flow simulation, using GPU acceleration

UltraFast GPU Grammar eXtractor for Machine Translation (He et al., TACL 2015 & NAACL 2013)

Get started with CUDA programming

Case studies constitute a modern interdisciplinary and valuable teaching practice which plays a critical and fundamental role in the development of new skills and the formation of new knowledge. This research studies the behavior and performance of two interdisciplinary and widely adopted scientific kernels, a Fast Fourier Transform and Matrix M…

Playing with CUDA and GPUs in Google Colab

Speeding up blind deconvolution of a blurred image by using GPUs