PGI Compilers now available on the Cray CX Line of Supercomputers

The Portland Group, a wholly-owned subsidiary of STMicroelectronics and a leading supplier of compilers for high-performance computing (HPC), today announced an agreement under which Cray Inc will resell PGI® optimizing Fortran, C, and C++ compilers and development tools with the Cray CX1 line of deskside supercomputers and the recently announced Cray CX1000 line of rack mount supercomputers.

The PGI compilers and tools are used by scientists and engineers to create and run high-performance computing applications for complex modeling and simulations in weather forecasting, geophysical processing, aerodynamic simulation, structural analysis, automotive crash-testing, computational chemistry, and related fields. PGI compilers and tools support the latest 64-bit processors from AMD and Intel as well as CUDA™-enabled GPU accelerators from NVIDIA running under both Linux and Microsoft Windows®.

Cray has offered PGI compilers with its scalable supercomputers since 2004, and PGI compilers are installed on nearly all Cray XT systems worldwide. These same PGI compilers will now be available for purchase directly from Cray for use on its Intel® Xeon® processor-based Cray CX1 and CX1000 systems running either Linux or Windows HPC Server 2008. In addition, the PGI compilers for the Cray CX line will include the new PGI Accelerator features for NVIDIA GPUs, as well as the PGDBG® OpenMP/MPI graphical debugger and PGPROF® OpenMP/MPI graphical parallel performance profiling tools.

“Giving Cray’s customers the flexibility and simplicity of purchasing our high-performance compilers directly from Cray for the Cray CX1 and Cray CX1000 systems is a big win for everyone,” said Douglas Miles, director, The Portland Group. “Together we can offer a uniform PGI compiler environment on the Cray CX machines running Linux or Windows up through the high-end scalable Cray XT machines and provide significant ease of migration and application upscaling benefits for existing and future Cray users.”

First introduced in September 2008, the Cray CX1 line of deskside supercomputers are ideal for individuals and departments who want to harness the compute power of an integrated HPC cluster. Equipped with state-of-the-art visualization and storage capabilities and optionally configured with NVIDIA GPUs, the Cray CX1 system delivers performance leadership across a broad range of applications in a compact, deskside system. The Cray CX1000 system is the latest addition to Cray's CX line of supercomputing systems, scaling up to 128 nodes in a rack-mount configuration and offering HPC users the same integrated software stack as the Cray CX1system with a simplified, common operating environment for either Windows or Linux along with several system management packages.

“We are excited to offer our Cray CX1 and Cray CX1000 customers access to the same compiler technology enjoyed by our Cray XT customers,” said Ian Miller, senior vice president of the productivity solutions group and marketing at Cray. “In addition, PGI’s ability to deliver technologies that leverage the GPU capabilities in the Cray CX line make it extremely relevant as more customers explore the acceleration capabilities in NVIDIA Tesla-based systems.”

The PGI compiler suite includes the PGFORTRAN™, PGCC®, and PGC++™ compilers for the Fortran 95/03, C, and C++ programming languages, respectively. PGI compilers and tools feature full native support for OpenMP parallel programming extensions in Fortran, C, and C++, full support for 64-bit addressing, native integrated scalar and vector SSE and AVX code generation, directive-based x64+GPU programming, CUDA Fortran extensions for NVIDIA GPUs, and a bundled version of the ACML 4.4 library of highly optimized numeric functions for mathematical, engineering, scientific, and financial applications. PGI compilers are highly optimized for AMD and Intel CPUs, and NVIDIA CUDA-enabled GPUs running Linux, Mac OS X and Windows. The PGDBG debugger and PGPROF performance profiler are MPI- and OpenMP-enabled, and available for cluster configurations.