About PGI Compilers & Tools
PGI is a premier supplier of software compilers and tools for parallel computing, known as PGI® products. PGI offers optimizing parallel Fortran 2003, C11 and C++11 compilers and tools for workstations, servers and clusters running Linux, macOS or Windows operating systems based on the following microprocessors:
- 64-bit Intel 64 technology CPU (Intel Core/Xeon/i3,i5,i7 families and related processors)
- 64-bit AMD64 technology CPU (AMD Opteron family and related processors)
- 64-bit OpenPOWER 8
- NVIDIA Tesla GPUs (supporting compute capability 2.0 and newer)
HPC is the segment of the computing industry that uses high-end parallel computing systems to accelerate computationally intensive tasks, commonly referred to as "Supercomputing". With the emergence of workstation, server and cluster architectures based on a variety of microprocessor technologies, and the emergence of GPGPU and accelerated computing technologies, users of these powerful systems rely on portable parallel programming models and hardware-independent optimizing compilers so that their applications can execute and scale efficiently across all heterogeneous systems. PGI delivers high-performance, portable and cost-effective production-quality compilers and software development tools that play an essential role in the use of HPC systems. PGI engages with vendors to support and optimize for their latest processor designs as they become available. PGI fully supports industry and de facto language standards such as OpenACC, OpenMP and MPI, has partnered with NVIDIA to support CUDA Fortran and CUDA C/C++ extensions natively in the PGI compilers, and is leading the HPC industry in defining and implementing higher-level scalable programming with the PGI Accelerator heterogeneous CPU+GPU compilers.
Product Offering Summary
Native FORTRAN, C and C++ Compilers for Linux, macOS and Windows
PGI is an industry leader in developing advanced compiler technologies. PGI compilers incorporate global optimization, interprocedural optimization, profile-feedback optimization, SIMD vectorization, multi-core auto-parallelization, and OpenMP 3.x capabilities to maximize performance on the latest multi-core x86 and OpenPOWER CPUs. PGI compilers support explicit GPGPU programming using NVIDIA CUDA, or they can automatically offload compute-intensive loops and regions to NVIDIA GPUs using the OpenACC directive-based programming model. The PGI CUDA compilers support either NVIDIA GPUs or multi-core x86 CPUs as a device target, allowing PGI-compiled CUDA programs to run on any x86 processor based system, with or without NVIDIA GPUs. PGI compilers do not rely on external pre-processors to provide these capabilities. PGI is the only supplier of compilers and tools with the capabilities to perform optimized native compilation for all multi-core CPU+NVIDIA platforms.
PGI offers optimizing compiler products for the following high-level languages:
- Fortran 2003 with CUDA Fortran extensions
- ANSI C11 with K&R extensions
- ANSI/ISO C++11 with CUDA C/C++ extensions
- PGI Accelerator Fortran, C and C++ with OpenACC
Compilers for these languages are offered in these packages:
Both PGI Community Edition and PGI Professional Edition are offered as single-seat node-locked licenses, meaning one user at a time can use any of the components in the suite on the system on which the PGI software is installed. PGI Professional Edition is also offered as a multi-user network-floating license, meaning that the licensed number of users can use any of the components in the suite on any system networked to the system on which the PGI software is installed. Executables produced by the PGI compilers are unconstrained, and can be run on any compatible system regardless of whether the PGI compilers are installed on that system. Not all products are available on all platforms.
Debuggers, Profilers and Other Tools
In addition to the full suite of parallel language compilers, most PGI products contain the PGI Debugger, an OpenMP/MPI graphical parallel debugger and the PGI Profiler, an OpenMP/OpenACC performance profiler. Both tools have an identical user interface and uniform feature set across all supported platforms. The versions included with PGI single-user node-locked licenses support debugging and profiling of up to 16 local OpenMP threads or MPI processes. The versions included in the PGI multi-user network floating licenses support debugging and profiling of distributed-memory MPI programs up to 256 processes on Linux clusters using a wide variety of MPI implementations. The PGI tools user interfaces are designed to be intuitive and easy-to-use, allowing developers to perform the most common parallel debugging and profiling tasks quickly and easily.
PGI Visual Fortran (PVF) integrates the PGI Fortran 2003 compiler, including CUDA and OpenACC directives for 64-bit x86 (x64) CPUs or NVIDIA GPUs, into Microsoft Visual Studio. It includes Fortran language editor extensions, a PGI-custom project system and parallel debug engine. PVF is a state-of-the-art visual programming environment for Windows Fortran, extended to support parallel debugging of OpenMP programs on multi-core x64, and of MSMPI programs running on Windows HPC Server clusters all from within Microsoft Visual Studio. PVF is included in all PGI products for Windows.
* All trademark items within this text are the property of their respective owners.