cuda11.8

Version:: 11.8.0
Category:: tools
Cluster:: Loki, Vali

Author / Distributor

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

Description

The NVIDIA CUDA® Toolkit 11.8 is the final release in the 11.x series, delivering a stable, high-performance development environment for GPU-accelerated applications.

Available modules:

cuda11.8/toolkit/11.8.0 Main CUDA toolkit, including nvcc, runtime libraries, Nsight tools, headers, and device drivers
cuda11.8/blas/11.8.0 cuBLAS 11.10 — GPU-accelerated dense linear algebra (BLAS Level 1, 2, and 3)
cuda11.8/fft/11.8.0 cuFFT 10.7 — Highly optimized FFT library for 1D, 2D, and 3D transforms on NVIDIA GPUs

This version includes:

Full support for Ampere (A100), Turing, Volta, and older NVIDIA GPUs
Enhanced compatibility for CUDA-enabled libraries and frameworks
Compiler and linker performance improvements
Updated libraries: - cuBLAS 11.10 - cuSPARSE 11.7 - cuFFT 10.7 - cuSOLVER 11.4
Updated Nsight Compute 2022.2 and Nsight Systems 2022.3
Compatibility with newer Linux distributions and GCC versions

Documentation

Toolkit: https://docs.nvidia.com/cuda/archive/11.8/
cuBLAS: https://docs.nvidia.com/cuda/archive/11.8/cublas
cuFFT: https://docs.nvidia.com/cuda/archive/11.8/cufft
CUDA Programming Guide: https://docs.nvidia.com/cuda/archive/11.8/cuda-c-programming-guide/

Examples/Usage

Load the CUDA toolkit module:

$ module load cuda11.8/toolkit/11.8.0

Compile your program for Ampere GPUs:

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

Run the application:

$ ./my_kernel

Load specific libraries if needed:

$ module load cuda11.8/blas/11.8.0
$ module load cuda11.8/fft/11.8.0

Check environment variables:

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

Unload the modules:

$ module unload cuda11.8/toolkit/11.8.0
$ module unload cuda11.8/blas/11.8.0
$ module unload cuda11.8/fft/11.8.0

Installation

The CUDA 11.8 Toolkit was installed from: https://developer.nvidia.com/cuda-11-8-0-download-archive