Is it possible to compile with a static link in nvcc(cuda) using a method like the -static option in gcc?
I'd like to run cuda-stream benchmark on the simulator, but the simulator does not support dynamic links, so I need a method that can be statically linked with nvcc.
I don't care what version of CUDA you use for compilation.
Thank you for your cooperation.
As a minimum of manners when you ask questions, write more information about your environment (OS, hardware, ...).
Pattern language when asking questions in technical mailing lists
Although it does not specify exactly what you want to statically link to in CUDA, CUDA 5.5 and later supports static links in the Runtime API library.
NVCC::CUDA Toolkit Documentation
--cudart { none | shared | static }
Please read the official manual before asking any questions.
Also, it seems that Driver API, cuBLAS, cuFFT, etc. do not support static links.
What's New in CUDA 5.5 (2): CUDA Runtime Static Links - Total Disclosure Site (Facts as Is)
© 2024 OneMinuteCode. All rights reserved.