New PGI Community Edition supports NVIDIA V100 Tensor Cores in CUDA Fortean, the full C++17 language, PCAST CPU/GPU auto-compare directives, OpenACC 2.6…
Overview
The PGI Community Edition 19.10 introduces support for NVIDIA V100 Tensor Cores in CUDA Fortran, along with enhancements in C++17 language features and OpenACC 2.6. This release is aimed at scientists and engineers developing high-performance computing applications, providing tools for multicore CPU performance and GPU computing.
What You'll Learn
1
How to utilize CUDA Fortran for V100 tensor core operations
2
Why OpenACC directives facilitate GPU computing
3
When to apply PCAST CPU/GPU auto-compare directives
Key Questions Answered
What new features are included in PGI Community Edition 19.10?
The PGI Community Edition 19.10 includes support for NVIDIA V100 Tensor Cores in CUDA Fortran, the full C++17 language, PCAST CPU/GPU auto-compare directives, and OpenACC 2.6. Additionally, it introduces new Fortran 2008 features and updates the LLVM back-end to version 8.0.
How does PGI on AWS benefit developers?
Developers can use PGI on AWS to develop, test, benchmark, and deploy applications on V100s for as little as $3 per hour, providing a cost-effective solution for high-performance computing needs.
Technologies & Tools
Programming Language
Cuda Fortran
Used for high-performance computing applications with support for NVIDIA V100 Tensor Cores.
Parallel Computing Model
Openacc
Facilitates GPU computing through directives.
Compiler Infrastructure
Llvm
Updated to version 8.0 for enhanced performance on Linux x86-64 and OpenPOWER.
Key Actionable Insights
1Leverage the new CUDA Fortran features to optimize data movement for V100 tensor core operations.By using the 16-bit REAL(2) data type and optimized array assignments, developers can significantly enhance performance in their HPC applications.
2Utilize OpenACC auto-compare directives to ensure accuracy between CPU and GPU results.This feature allows developers to quickly identify and correct discrepancies in results, which is crucial for maintaining the integrity of scientific computations.
3Take advantage of the interoperability with GNU releases through GCC 9.1 for C++ development.This interoperability enhances the flexibility of using PGI compilers in diverse development environments, making it easier to integrate with existing codebases.