Mobile CUDA

Mobile CUDA is a set of deamon and CUDA's runtime library witch does "Managing Multiple CUDA Processes Dynamicaly for Hyper-Q".

Support CUDA-5.5.

*Utilize

Fix a path to NVIDIA's library written in /path to Mobile CUDA/include/mocu.h at line 298

#define CUDA_LIB_PATH path_to_NVIDIA's_runtime_library

Since the library is runtime one, you have to recompile your application with the flag -lcudart(or cudart=shared) like as follows

nvcc app.cu -lcudart

Run a deamon witch is located in /path to Mobile CUDA/daemon/mocu_daemon

./mocu_daemon

Open a library path like this

export LD_LIBRARY_PATH=/path for Mobile CUDA/lib64/

Try to run some CUDA application.

Hyper-Q is a hardware function on NVIDIA's GPU (supported Kepler and above gen) witch can run multiple applications on single device efficiently. If you have oppotunity for running multiple CUDA applications on single node, this project is very helpfull for you.

author Taichirou Suzuki.