The Portland Group

The Portland Group, commonly known as PGI, was a privately held high-performance computing software company founded in 1989. It developed commercial compilers and development tools for scientific, engineering, academic, government, and enterprise users building applications for supercomputers, clusters, workstations, embedded processors, and accelerator-based systems. Its technology originated partly from compiler work developed at and acquired from Floating Point Systems. PGI's first products appeared in 1991 as pipelining Fortran and C compilers for Intel's i860 processor. These products were used with several early Intel-based supercomputing systems, including the iPSC/860, Touchstone Delta, and Paragon. The company's early market position was built around compiler performance, architecture-specific optimization, vectorization, software pipelining, and support for parallel computing workloads. During the 1990s, PGI also participated in the development and adoption of High Performance Fortran, or HPF, a data-parallel extension associated with Fortran 90. Its PGHPF compiler remained part of the product family until its final listed release in 2015. PGI expanded from specialized i860 systems into general-purpose x86 computing. In 1996 it developed x86 compilers for the ASCI Red supercomputer at Sandia National Laboratories, a system associated with the first sustained teraflop-level performance. In 1997, it introduced x86 compilers for general Linux workstations, followed by support for OpenMP, SIMD and SSE-oriented optimization, 64-bit Linux, and other increasingly important processor features. Its software portfolio broadened beyond compilers through tools such as PGDBG, a parallel debugger, PGPROF, a performance profiler, PGI Visual Fortran for Microsoft Visual Studio, and the PGI Cluster Development Kit. The company was acquired by STMicroelectronics on December 19, 2000. Under STMicroelectronics, PGI operated as a wholly owned subsidiary and continued developing high-performance computing compilers and tools for Linux, Windows, and Mac OS, while also supporting STMicroelectronics' ST100 series of embedded digital signal processor cores. Its products covered Fortran, C, C++, High Performance Fortran, OpenMP, OpenCL, and later accelerator-oriented programming models. A major phase of PGI's later development involved heterogeneous and GPU computing. The company worked with NVIDIA on CUDA Fortran and developed PGI Accelerator compilers for Fortran and C using programming directives. PGI and NVIDIA also participated in the development of the OpenACC directive standard, announced in 2011, which sought to make accelerator programming more portable and accessible without requiring developers to rewrite entire applications in low-level GPU languages. PGI later released an OpenCL compiler targeting multicore ARM processors in 2013. NVIDIA acquired PGI from STMicroelectronics on July 29, 2013. NVIDIA initially continued the technology under the PGI Compilers and Tools product line. The PGI brand was subsequently folded into NVIDIA's broader software portfolio. On August 5, 2020, NVIDIA announced that PGI compiler technology had evolved into the NVIDIA HPC SDK, distributed as a free download. The SDK incorporated rebranded PGI compiler technologies, including nvfortran, nvc, and nvc++, as well as additional tools and libraries for high-performance and accelerated computing. The former PGI Compilers and Tools identity was therefore retired as an independent product brand. PGDBG was superseded by cuda-gdb, while PGPROF was replaced by NVIDIA Nsight tools. Mac OS support and PGI Visual Fortran had already been discontinued after PGI version 19.10, released in November 2019.

The Portland Group was established in 1989 as a privately held compiler company using technology developed at and acquired from Floating Point Systems. Its initial commercial products, released in 1991, were pipelining Fortran and C compilers for Intel's i860 processor. These tools targeted early parallel and supercomputing installations, including the Intel iPSC/860, Touchstone Delta, and Paragon systems, and became widely used across i860-based platforms. During the early 1990s, PGI became involved in High Performance Fortran, a data-parallel language extension associated with Fortran 90. The company developed PGHPF, allowing users to express parallel computation in a more portable Fortran-based programming model. PGI subsequently expanded its compiler technology to other processor families and operating systems, while retaining an emphasis on optimization for scientific and technical workloads. In 1996, PGI developed x86 compilers for the ASCI Red supercomputer at Sandia National Laboratories. The following year it released x86 compilers for general Linux workstations, helping move its technology beyond specialized i860 systems. Later milestones included OpenMP support, SSE and SIMD vectorization, 64-bit Linux compilers, cluster development tools, and support for the ST100 embedded DSP architecture. STMicroelectronics acquired The Portland Group on December 19, 2000. PGI operated within STMicroelectronics as a wholly owned subsidiary, producing compilers and tools for Linux, Windows, Mac OS, and selected embedded targets. Its product family included Fortran, C, C++, HPF, parallel debugging, profiling, and Visual Studio integration through PGI Visual Fortran. PGI was also an early commercial participant in accelerator-oriented programming. It worked with NVIDIA on CUDA Fortran and created PGI Accelerator compilers for Fortran and C. These products used directives to enable offloading and parallel execution while preserving much of the original source structure. PGI and NVIDIA participated in the OpenACC initiative, announced in 2011, which defined directives for programming accelerators and GPUs. In 2013, PGI released an OpenCL compiler for multicore ARM processors. NVIDIA purchased PGI from STMicroelectronics on July 29, 2013. The acquired technology initially remained visible through the PGI Compilers and Tools product line. NVIDIA later incorporated the compiler technology into the NVIDIA HPC SDK. In 2020, NVIDIA described the SDK as the successor to the PGI product line and made it available as a free download. The former PGI compiler names were mapped to NVIDIA names: pgfortran became nvfortran, pgcc became nvc, and pgc++ became nvc++. PGDBG was replaced by cuda-gdb and PGPROF by NVIDIA Nsight tools. The standalone PGI brand consequently ceased to operate as an independent product identity.

1. 2020PGI technology becomes part of NVIDIA HPC SDK

   NVIDIA retires the standalone PGI Compilers and Tools branding and incorporates the technology into NVIDIA HPC SDK.

   Source
2. 2019Mac OS and PGI Visual Fortran support ends

   Support for Mac OS and the PGI Visual Fortran product is discontinued after PGI 19.10.

   Source
3. 2015Final listed PGHPF release

   PGHPF reaches its final listed release, version 15.10.

   Source
4. 2013Acquisition by NVIDIA

   NVIDIA acquires PGI from STMicroelectronics and continues its technology under the PGI Compilers and Tools name.

   Source
5. 2011Participation in OpenACC

   PGI and NVIDIA participate in the OpenACC directive standard for accelerator programming.

   Source
6. 2009CUDA Fortran compiler introduced

   PGI develops CUDA Fortran technology with NVIDIA for GPU-accelerated computing.

   Source
7. 2000Acquisition by STMicroelectronics

   STMicroelectronics acquires PGI and operates it as a wholly owned subsidiary.

   Source
8. 1997Linux x86 compilers introduced

   PGI releases x86 compilers for general Linux workstation use.

   Source
9. 1996Compilers developed for ASCI Red

   PGI develops x86 compilers for Sandia National Laboratories' ASCI Red supercomputer.

   Source
10. 1991First i860 pipelining compilers released

    PGI releases Fortran and C compilers optimized for Intel's i860 processor.

    Source
11. 1989The Portland Group is founded

    The company is established as a privately held compiler and high-performance computing software business.

    Source

A performance-focused compiler and development-tools provider for scientific, engineering, and technical computing. PGI differentiated itself through architecture-specific optimization, parallel programming support, and early commercial support for accelerator and GPU-oriented programming models.

• PGI Fortran CompilerFortran compiler1991

  PGI's Fortran compiler family supported traditional Fortran dialects and progressively added features for parallel, vector, accelerator, and high-performance computing. The later pgfortran compiler was rebranded as nvfortran within the NVIDIA HPC SDK.

• PGI C and C++ CompilersC and C++ compilers

  PGI supplied optimizing C and C++ compilers designed for scientific and technical workloads, including vectorization, shared-memory parallelism, and processor-specific optimization. The former pgcc and pgc++ products became nvc and nvc++ in NVIDIA's HPC SDK.

• PGHPFHigh Performance Fortran compiler

  PGHPF implemented High Performance Fortran, a data-parallel extension associated with Fortran 90. It provided a Fortran-oriented way to express parallel computation across supported architectures, with version 15.10 identified as its final release.

• PGI Accelerator CompilersAccelerator compiler2008

  The PGI Accelerator compiler family supported directive-based acceleration for Fortran and C. It was intended to let developers target GPUs and other accelerators while retaining a comparatively high-level programming model.

• CUDA FortranGPU programming compiler2009

  Developed with NVIDIA, CUDA Fortran extended Fortran development into NVIDIA GPU computing. It served scientific and engineering programmers who wanted to use CUDA-oriented acceleration while working in a Fortran environment.

• PGDBGParallel debugger

  PGDBG was a debugger for parallel applications and supported programming environments involving OpenMP and MPI. It was offered across selected Linux, Windows, and Mac OS configurations and was later superseded by cuda-gdb.

• PGPROFPerformance profiler

  PGPROF provided performance-analysis capabilities for parallel programs, including applications using OpenMP and MPI. NVIDIA later directed users toward Nsight tools for profiling and analysis.

• PGI Visual FortranIntegrated development tool2006

  PGI Visual Fortran integrated the PGI Fortran compiler into Microsoft Visual Studio, giving Windows developers an environment for building Fortran applications with PGI's optimization and parallel-programming capabilities.

• 2020Transition from PGI Compilers and Tools to NVIDIA HPC SDKGeneration change

  NVIDIA consolidated its HPC compiler, library, debugger, profiler, and accelerator-development capabilities into a broader software development kit.

  What changed. NVIDIA retired the standalone PGI Compilers and Tools branding and made the successor NVIDIA HPC SDK available as a free download.

  Aftermath. PGI compiler technologies continued under NVIDIA names, including nvfortran, nvc, and nvc++, while PGDBG and PGPROF were replaced by NVIDIA debugging and profiling tools.

  Source
• 2013NVIDIA acquires PGI technologyM&A

  PGI had developed strong capabilities in GPU, CUDA Fortran, directive-based acceleration, and OpenACC-related compiler technology.

  What changed. NVIDIA acquired The Portland Group from STMicroelectronics and placed its technology in the PGI Compilers and Tools product line.

  Aftermath. PGI technology became part of NVIDIA's high-performance computing software strategy and was later incorporated into NVIDIA HPC SDK.

  Source
• 2000Join STMicroelectronicsM&A

  PGI was acquired by STMicroelectronics as the high-performance computing compiler market developed alongside broader processor and embedded-system opportunities.

  What changed. STMicroelectronics acquired The Portland Group and operated it as a wholly owned subsidiary.

  Aftermath. PGI continued developing compilers and tools for multiple operating systems, HPC platforms, and STMicroelectronics embedded DSP targets.

  Source

• 2020NVIDIA transitions PGI technology into the NVIDIA HPC SDK

  NVIDIA announced that PGI Compilers and Tools had evolved into the NVIDIA HPC SDK, which incorporated rebranded PGI compilers and additional high-performance computing features.

  Product generation
  Source
• 2019PGI Visual Fortran and Mac OS support are discontinued

  PGI discontinued Mac OS support and the PGI Visual Fortran product after release 19.10.

  Other
  Source
• 2013NVIDIA acquires The Portland Group from STMicroelectronics

  NVIDIA acquired PGI and continued offering its compiler and development technology under the PGI Compilers and Tools product line.

  M&A
  Source
• 2013PGI releases an OpenCL compiler for multicore ARM processors

  PGI announced a compiler for OpenCL targeting multicore ARM processors; the offering was later removed after NVIDIA acquired PGI.

  Product launch
  Source
• 2000STMicroelectronics acquires The Portland Group

  The Portland Group became a wholly owned subsidiary of STMicroelectronics and continued producing high-performance computing compilers and tools.

  M&A
  Source

1. The Portland Group
2. NVIDIA HPC SDK