Nvidia has announced news of version 14.4 of the PGI 2014 Compilers and Tools suite. This release is positioned as a means of bringing new levels of performance and "programming simplicity" to High Performance Computing (HPC) software application development.
Initially announced as a first product release in February of this year, PGI 2014 is a suite of high-performance parallelizing compilers and development tools. Version 14.4 provides a number of performance and programmability enhancements, all of which are (according to Nvidia) positioned to allow HPC-developers to maximize the power of accelerators and multi-core CPU processors.
Key features of version 14.4 include expanded OpenACC features — including loop directive collapse clause on deeply nested loops, parallel directive firstprivate clause, and C structs/Fortran derived type member arrays in data clauses.
There is also support for OpenACC in C++ Classes — including static array class members, member pointers to dynamic arrays, and member functions and operators. OpenACC device-side debugging also features here and this is supported using Allinea DTT.
Multi-core CPU features and enhancements include improved SPEC OMP 2012 performance compared to PGI 14.1 on Intel Sandy Bridge processors, support for new AVX2 instructions on the latest Intel Haswell CPUs, and a new EDG C++ front-end with C++11 support.
Other new features and enhancements of PGI 14.4 include comprehensive support for environment modules, prebuilt versions of the popular NetCDF and HDF5 libraries, and new tutorials and code examples.