cuda11.0

Version:

11.0.3

Category:

tools

Cluster:

Loki

Author / Distributor

https://developer.nvidia.com/cuda-toolkit-archive

Description

The NVIDIA CUDA® Toolkit 11.0 provides a powerful development environment for building GPU-accelerated applications using C, C++, and Fortran.

CUDA 11.0 introduces support for:

• Ampere GPUs (e.g., A100)

• Improved compatibility and performance for Volta and Turing

• Updated GPU-accelerated libraries including cuBLAS, cuFFT, cuSOLVER, and cuSPARSE

• Enhanced compiler toolchain with CUDA Graph and cooperative groups support

• Latest Nsight profiling tools (bundled within the toolkit module)

Available modules:

• cuda11.0/toolkit/11.0.3 Core CUDA development environment (includes nvcc, runtime, driver APIs, and tools)

• cuda11.0/blas/11.0.3 GPU-accelerated linear algebra library (cuBLAS)

• cuda11.0/fft/11.0.3 GPU-accelerated FFT routines for signal processing and scientific applications (cuFFT)

Documentation

• Toolkit: https://docs.nvidia.com/cuda/archive/11.0/

• cuBLAS: https://docs.nvidia.com/cuda/archive/11.0/cublas

• cuFFT: https://docs.nvidia.com/cuda/archive/11.0/cufft

• Programming Guide: https://docs.nvidia.com/cuda/archive/11.0/cuda-c-programming-guide/

Examples/Usage

• Load the main toolkit module:

$ module load cuda11.0/toolkit/11.0.3

• Compile a CUDA program (targeting Ampere):

$ nvcc -arch=sm_80 my_code.cu -o my_code

• Run your executable:

$ ./my_code

• Use cuBLAS and cuFFT separately:

$ module load cuda11.0/blas/11.0.3
$ module load cuda11.0/fft/11.0.3

• Check your environment:

$ echo $CUDA_HOME
$ echo $PATH
$ echo $LD_LIBRARY_PATH

• Unload modules:

$ module unload cuda11.0/toolkit/11.0.3
$ module unload cuda11.0/blas/11.0.3
$ module unload cuda11.0/fft/11.0.3

Installation

The CUDA 11.0 Toolkit was installed from the NVIDIA archive: https://developer.nvidia.com/cuda-11.0-download-archive