nvhpc

Version:

24.1, 23.1, 22.11, 22.7

Category:

compiler

Cluster:

Vali

Author / Distributor

https://developer.nvidia.com/hpc-sdk

Description

NVIDIA HPC SDK (NVHPC) provides a comprehensive suite of compilers, libraries, and tools for developing high-performance computing (HPC) applications on NVIDIA GPUs and CPUs.

Version 22.11 supports CUDA 11.7.0 and includes:

  • Fortran, C, and C++ compilers (nvc, nvc++, nvfortran)

  • Full support for GPU programming via: - CUDA Fortran and CUDA C++ - OpenACC 2.7 - OpenMP 4.5/5.x GPU offloading

  • Optimized math libraries (cuBLAS, cuFFT, LAPACK, OpenMPI)

  • Integrated profiling and debugging tools

  • Multi-architecture support: x86_64, Ampere, Volta, and more

Documentation

$ nvc --help
$ nvc++ --help
$ nvfortran --help

Full documentation available at:
https://docs.nvidia.com/hpc-sdk/

Programming model support:
- CUDA
- OpenACC
- OpenMP (target offload)
- MPI and multi-GPU with NVIDIA-aware libraries

Examples/Usage

  • Load the module:

$ module load NVHPC/22.11-CUDA-11.7.0

  • Compile a CUDA C++ program:

$ nvc++ -cuda matrixMul.cu -o matrixMul

  • Compile with OpenACC:

$ nvc -acc -Minfo=acc vector_add.c -o vector_add

  • Compile with OpenMP GPU offloading:

$ nvc++ -mp=gpu -o omp_gpu omp_target.cpp

  • Unload the module:

$ module unload NVHPC/22.11-CUDA-11.7.0

Installation

Source: https://developer.nvidia.com/nvidia-hpc-sdk-downloads