Jump to content

The Portland Group

From Alfawiki
PGI
Company typeSubsidiary
IndustrySoftware, High-performance computing
PredecessorFloating Point Systems Inc. (compiler technology)
Founded1989; 37 years ago (1989)
FateAcquired by Nvidia
SuccessorNvidia HPC SDK
Headquarters
United States
Area served
Worldwide
ProductsCompilers for Fortran, C, C++, CUDA Fortran, OpenCL, OpenACC; Programming tools (debugger, profiler)
OwnerNvidia
ParentNvidia
Websitedeveloper.nvidia.com/hpc-sdk

PGI—originally known as The Portland Group, Inc.—specialized in producing commercially available compilers for Fortran, C, and C++ designed specifically for high-performance computing (HPC) environments. On July 29, 2013, Nvidia completed the acquisition of PGI. Since August 5, 2020, the compiler technologies developed by PGI have been incorporated into Nvidia's HPC SDK, which is distributed as a free download.

Company history

Established in 1989 as a privately held firm, PGI built its foundation upon compiler technology obtained from Floating Point Systems Inc.. Its first offerings, released in 1991, included pipelined Fortran and C compilers aimed at the Intel i860 processor and gained widespread adoption on Intel supercomputers like the Intel Paragon.

During the early 1990s, PGI was instrumental in advancing High Performance Fortran (HPF), producing the PGHPF compiler until its final update in 2015. By 1996, the company had developed x86 compilers for the ASCI Red Supercomputer at Sandia National Laboratories, marking the first system to achieve sustained teraflop speeds. The subsequent year saw PGI introduce x86 compilers tailored for Linux workstations.

On December 19, 2000, PGI became a wholly owned subsidiary of STMicroelectronics, continuing its focus on HPC compilers and development tools compatible with Linux, Windows, Mac OS, and STMicroelectronics’ embedded DSP cores.

PGI significantly contributed to the growth of general-purpose GPU (GPGPU) computing by collaborating with Nvidia to create CUDA Fortran compilers and developing PGI Accelerator Fortran and C compilers that utilized programming directives. Both companies played a pivotal role in establishing the OpenACC standard for GPU programming since its introduction in 2011. In 2013, PGI launched an OpenCL compiler targeting multi-core ARM processors.

Following Nvidia’s acquisition in 2013, PGI’s compiler technology continued under the "PGI Compilers and Tools" brand. By 2020, this suite had evolved into the Nvidia HPC SDK, which integrated rebranded PGI compilers enhanced for high-performance application development.

Compilers

PGI’s compiler suite provides advanced features such as global optimization, vectorization, software pipelining, and shared-memory parallelism optimized for Intel and AMD CPUs. The supported programming languages include:

These compilers have been rebranded within the Nvidia HPC SDK as follows:

  • Fortran: nvfortran (previously pgfortran)
  • C: nvc (previously pgcc)
  • C++: nvc++ (previously pgc++)

Programming tools

PGI offered parallel debugging and profiling utilities—PGDBG and PGPROF—that supported OpenMP and MPI across Linux, Windows, and Mac OS platforms. The PGI Fortran compiler and debugger were also integrated into Microsoft Visual Studio under the name PGI Visual Fortran (PVF) on Windows systems. Support for Mac OS and PVF was discontinued after the release of PGI version 19.10 in November 2019.

These legacy tools have since been phased out and replaced within the Nvidia HPC SDK by:

  • Debugger: PGDBG replaced by cuda-gdb
  • Profiler: PGPROF replaced by Nsight

PGI milestones

  • 1989 – Establishment of PGI
  • 1991 – Introduction of i860 Compilers with Pipelining
  • 1994 – Launch of Parallel i860 Compilers
  • 1996 – Development of ASCI Red TFLOPS Compilers
  • 1997 – Release of Linux/x86 Compilers
  • 1998 – Implementation of OpenMP for Linux/x86
  • 1999 – SSE/SIMD Vectorization and debut of PGI CDK Cluster Development Kit
  • 2000 – Acquisition by STMicroelectronics
  • 2001 – VLIW ST100 Compiler release
  • 2003 – 64-bit Linux/x86 Compiler availability
  • 2004 – Introduction of ASCI Red Storm Compilers
  • 2005 – Launch of PGI Unified Binary Technology
  • 2006 – Debut of PGI Visual Fortran
  • 2007 – Release of 64-bit Mac OS Compilers
  • 2008 – Introduction of PGI Accelerator Compilers
  • 2009 – Launch of CUDA Fortran Compiler
  • 2010 – Release of CUDA X86 Compiler
  • 2011 – Support for AVX/FMA Vectorization added
  • 2012 – Adoption of OpenACC directives to enhance GPU computing
  • 2013 – PGI OpenCL compiler for Multi-core ARM CPUs introduced (later discontinued after acquisition); Nvidia acquires PGI from STMicroelectronics, continuing development under "PGI Compilers and Tools"
  • 2015 – Launch of Flang, an open source Fortran front-end integrated with LLVM
  • 2018 – Initiation of new Flang Fortran front-end development based on the Fortran 2018 standard
  • 2020 – Nvidia incorporates PGI technology into the NVIDIA HPC SDK, discontinuing the "PGI Compilers and Tools" brand

See also

External links

Retrieved from "https://www.alfawiki.org/index.php?title=The_Portland_Group&oldid=23871"