CUDA Toolkit Container: The CUDA (Compute Unified Device Architecture) Toolkit provides a development environment for creating high performance GPU-accelerated applications. The toolkit includes GPU-accelerated libraries, debugging and optimization tools, a C/C++ compiler and a runtime library to deploy your application. All NGC Container images are based on the CUDA platform layer
( ).

CUDA Libraries
cuBLAS 9.2.174 is an update to CUDA Toolkit 9.2 and is available separately as a patch update to CUDA 9.2. It includes the following changes:
* Improved GEMM performance for large matrices of certain tile sizes.
* Improved performance of half precision GEMM APIs with optimized algorithm selection heuristics for FP16 compute type.
* Improved performance of complex GEMM APIs for small n matrices.
* Improved performance of half precision GEMM APIs – cublashgemm(), cublashgemmbatched() and cublashgemmstridedbatched()
The sparse matrix-vector multiplication routine in based on the “mergepath” approach, cusparse<t>csrmv_mp() randomly crashes in cuSPARSE 9.2.

Environment variable setup:
To add this path to the PATH variable:
Add the following lines in your .bashrc file
$vi ~/.bashrc
$ export PATH=/usr/local/cuda-9.2/bin:$PATH
In addition, when using the runfile installation method, the LD_LIBRARY_PATH variable needs to contain /usr/local/cuda-9.2/lib64 on a 64-bit system.
To change the environment variables for 64-bit operating systems:
$ export LD_LIBRARY_PATH=/usr/local/cuda-9.2/lib64
Run the command source .bashrc


For any queries, raise a ticket in the helpdesk or please contact System Administrator, #103,SERC.