Simulation / Modeling / Design

Announcing the NVIDIA HPC SDK

AI-Generated Summary

  • The NVIDIA HPC SDK provides a comprehensive suite of compilers and libraries for programming accelerated computing systems from the GPU foundation through the CPU and interconnect.
  • NVIDIA Fortran, C, and C++ compilers enable GPU acceleration of standard parallel language constructs including C++17 parallel algorithms and Fortran intrinsics.
  • NVIDIA math libraries deliver drop-in GPU acceleration for linear algebra and signal processing algorithms with optimized performance on architectures such as A100 GPUs.
  • The new NVSHMEM library implements the OpenSHMEM standard for GPU memory with extensions for multi-GPU and multi-node communication from host, device, CUDA kernels, and OpenACC programs.
  • NVIDIA Nsight Compute and Nsight Systems provide performance profiling for multicore CPU applications in OpenMP and OpenACC and GPU-accelerated applications in OpenACC and CUDA.
  • The HPC Container Maker simplifies building and maintaining Docker and Singularity containers for HPC workloads.

Next Steps

  • Sign up for early access to receive notification when the SDK is ready.
  • Register for the webinar to learn more about the SDK capabilities.
  • Learn more about the NVIDIA HPC SDK features and documentation.
Powered by NVIDIA Nemotron. AI-generated content may summarize information incompletely. Verify important information. Learn more

The NVIDIA HPC SDK is a comprehensive suite of compilers and libraries enabling HPC developers to program the entire HPC platform from the GPU foundation to the CPU and through the interconnect. It is the only comprehensive, integrated SDK for programming accelerated computing systems.

The NVIDIA HPC SDK C++ and Fortran compilers  are the first and only compilers to support automatic GPU acceleration of standard language constructs including C++17 parallel algorithms and Fortran intrinsics.

The NVIDIA HPC SDK introduces new capabilities and performance optimizations for GPU-accelerated applications:

  • In addition to being the first compilers to enable GPU acceleration of standard parallel  language constructs, the NVIDIA Fortran, C, and C++ compilers enable the porting, writing, and tuning of parallel applications for heterogeneous CPU+GPU servers using GPU-accelerated math libraries, de facto standard OpenACC and OpenMP parallel directives, and CUDA. 
  • The NVIDIA math libraries provide drop-in, highly optimized GPU-acceleration for linear algebra and signal processing algorithms fundamental to HPC. In addition to providing an easy on-ramp to GPU acceleration, math libraries provide speed-of-light performance for supported routines and enable users to automatically benefit from newer GPU architectures including the latest A100 GPUs.
  • The new NVSHMEM library provides multi-GPU and multi-node communication primitives that can be initiated from the host or device and called from within a CUDA kernel and OpenACC program. NVSHMEM implements the OpenSHMEM standard for GPU memory, with extensions for improved performance on GPUs.
  • NVIDIA Nsight™ Compute and Nsight System are performance profiling tools for multicore-CPU applications in OpenMP and OpenACC and GPU-accelerated applications in OpenACC and CUDA. (Not all features are currently supported on all platforms.)
  • The NVIDIA debugger can be used to debug CUDA applications and is the foundation of higher-level third-party debuggers for OpenACC, OpenMP, and MPI.
  • The HPC Container Maker simplifies the process of building and maintaining Docker and Singularity containers.

Early access will be available soon. Sign up today here and we’ll notify you when it’s ready. The first general release is planned for later this summer.

Register for the Webinar>>

Learn more > >

Discuss (0)

Tags