The cuosqp from zhenshenglee

zhenshenglee / cuosqp Goto Github PK

View Code? Open in Web Editor NEW

CUDA based osqp compatible with apollo openspace planner.

License: Apache License 2.0

CMake 6.43% C 76.78% Cuda 11.27% Batchfile 0.59% Shell 0.92% C++ 0.06% Python 3.95%

黎振胜 (Zhensheng(Victor) Lee)

Software engineer in Robotics Middleware

wechat: zhensheng_li

Contribution

Streaming based automotive middleware specifically in Compute Graph Framework (CGF) and System Task Manager (STM) Scheduler from Nvidia DriveWorks SDK

Pubsub-messaging based Robotics Middleware specifically in rclcpp, rcl and rmw from OSRF ROS2

GPU based Hardware Acceleration specifically in CUDA

Rust programming specifically with cpp

iceoryx-rust

Education

Master's degree in Robotics Engineering, Central South University, China [2014-2017]

Experience

Software Engineer in automotive middleware for ADAS system, GACRND, China [2022-]
Software Engineer in robotics middleware for LSAV system, BZL Corporation, China [2019-2022]
Software Engineer in application server middleware for 5G fronthaul network switch, ZTE Corporation, China [2017-2019]

Certification

Project Management Professional (PMI-PMP), Project Management Institute, [2019-]
System Architect, Qualification Certificate of Computer and Software Technology Proficiency, MIIT, China[2019-]

Publication

Design and Implementation of a software platform for interaction between human and remote mobile robots

cuosqp's People

Contributors

Stargazers

Watchers

cuosqp's Issues

Cmake build wrong CUDA architecture

Hello,

In cuosqp/CMakeLists.txt, I've added the following lines:

set(CMAKE_CUDA_ARCHITECTURES 75)
set(CUDA_ARCHITECTURES 75)

However, when I cmake --build ., it runs into error: no instance of overloaded function "atomicAdd" matches the argument list argument types are: (double *, double).

This is because the compiler keeps setting arch=compute52_52,code=sm_52 (I found this by running grep -rn . -e "sm_" in cuosqp/build folder). I have no idea where the number 52 comes from, because it runs fine on Google Colab but not on my PC.

So, what's the correct way to set CUDA_ARCHITECTURES ? My machine has cuda-11.4 and 2080ti

Thanks!

Recommend Projects