回首頁 > 產品介紹 > 中介軟體 > 平行程式編譯套件 > PGI

平行程式編譯套件 -PGI

What's New in 18.10

All Compilers
Updated PGI compilers on Windows to use Microsoft's Visual Studio 2017 toolchain.
PGI compilers on Wndows now require Microsoft Visual Studio 2017 as an installation

Combined the PGI compiler packages on Windows into a single installation executable;
there are no longer separate install packages for PGI Community Edition and PGI Professional/Enterprise.

Enhanced output format and readability of the autocompare feature of PGI Compiler Assisted Software Testing (PCAST).

Added support for patch-and-continue to autocompare. When an incorrect value is detected, autocompare will emit a warning, replace the bad value with the known good value,
and continue execution. More details can be found in the PCAST Overview.

Our C and C++ compilers now have GCC-compatible support for visibility. Visibility in
this context describes how symbols should be exported from shared libraries. By default,
functions have global scope and can be called from other objects. Functions with hidden
visibility have local scope and cannot be called from other objects.

The compiler now implements -fvisibility and -fvisibility-inlines-hidden switches.

This switch tells the compiler how to set the default ELF image symbol visibility.
All symbols are marked with global visibility unless overridden with this switch or with
the visibility attribute.

This switch tells the compiler that the code does not compare pointers to inline functions
where the addresses of the two functions are from different objects. This option has the
same effect as marking inline methods with __attribute__((visibility(“hidden”))).

Added support for Fortran 2018 DO CONCURRENT. The body of the DO CONCURRENT
block is executed serially.
We now generate debug metadata for Fortran COMMON blocks when compiling with the
LLVM-based code generator.
Improved Fortran MATMUL performance for small matrices.

OpenACC and CUDA Fortran
Added support for the CUDA Toolkit version 10.0. Read more about support for CUDA
Toolkit versions in CUDA Toolkit Versions.
Added support for the use of relocatable device code within shared objects on Linux
Added definition of _CUDA and __CUDA_API_VERSION to C and C++ compilers when
compiling with -Mcuda.
Disabled the nollvm sub-option for -ta=tesla and -Mcuda switches on Windows.
The Windows compilers will ignore nollvm and emit a warning to that effect.

Changed the default value of OMP_MAX_ACTIVE_LEVELS to INT_MAX.

Some Key Information:

CUDA 10.0 support –

PGI 18.10 includes components of CUDA Tookit versions 9.1, 9.2 and 10.0.

PGI command-line compilers for Windows now support Visual Studio 2017, which is required as a prerequisite.

PGI Visual Fortran (PVF) is no longer available for sale to new users. Support and updates for existing licenses will be provided through 2019.

PGI 18.7 was the last release to include bundled Microsoft toolchain components. 18.10 and future releases require users to have the Microsoft toolchain components pre-installed on their system. IMPORTANT: Only users who have purchased a PGI Professional for Windows license on or before August 31, 2018 will have access to PGI 18.7 and earlier PGI Professional for Windows releases. Since 18.10 is also a Community Edition release, we encourage you to defer new PGI Professional for Windows license sales until the release of 19.1. Sales of maintenance renewals to existing users can continue during this time.

Today we released PGI 17.5 [1], a quality-improvement release that also expands operating system support to include Ubuntu 17.04 and Redhat Enterprise Linux 6.9.

Only users with a PGI account linked to a PGI for-fee license can download 17.5 [2] without assistance from PGI. Licensed end-users in large organizations who are _ not_ linked to a PGI for-fee license should _ first contact their license administrator_ or IT department for
assistance. PGI Sales [3] can also assist with granting download access to such users.

For purchase evaluations, we first encourage prospective users to try PGI Community Edition [4], which is currently 17.4. If prospective customers wish to try 17.5, they should contact PGI Sales for a temporary evaluation [5] download link.

Thank you for your continued support of PGI Compilers & Tools!

[1] http://www.pgroup.com/support/release.htm?utm_source=175_reseller
[2] http://www.pgroup.com/support/downloads.php?utm_source=175_reseller
[3] mailto: [email protected]
[4] http://www.pgroup.com/community?utm_source=175_reseller
[5] mailto: [email protected]?subject=PGI%20Professional%20Evaluation%20For%20Purchase%20Request


PGI 2017 Features and Performance

PGI Accelerator OpenACC Compilers
- New OpenACC 2.5 features:
 . Added if_present clause to the update directive
 . Added finalize clause to the exit data directive
 . Added new init, shutdown and set directives
 . Added new API routines to get and set the default async queue value
 . Added support for the redefined routine bind clause

- Behavior changes:
 . Changed exit data directive behavior to decrement dynamic reference count

- .Improved performance of the cache directive
- Added support for CUDA 8.0; CUDA 7.5 is now default
- Support for compute capability 2.0 (Fermi) devices no longer included by default

PGI C++ Compiler
- Comprehensive support of C++14, requires GCC version 5.1 or greater; full interoperability with GCC 5.1 through 6.2
- Added C++11 support when used as the nvcc host compiler; requires updated nvcc
- Improved exception handling

PGI Fortran Compiler for OpenPOWER CPUs
- Preliminary support for OpenMP 4.5 on OpenPOWER CPUs; GPU offload not yet supported
- Average performance improvements of 5–10% on floating-point intensive codes

All PGI Compilers
- Floating point divides are now guaranteed to be uniform for both scalar and vector operations
- Additional tuned numerical math intrinsic functions for Intel Haswell and Broadwell x86 CPUs
- Improved inlining

PGI Tools
- PGDBG debugger supports variable rollover
- Updated Eclipse plug-in for use with Neon and Mars (x86) and Mars and Luna (OpenPOWER)

- Linux products now ship with pre-built Open MPI 1.10.2; MVAPICH 2.2 and MPICH 3.2 available to current support subscribers
- Pre-built versions of NetCDF and Parallel NetCDF 1.7.0 available online
- Pre-built versions of the Earth System Modeling Framework (ESMF) 7.0.0 available online; one - per PGI-built MPI distribution

Other Features, Changes and Additions
- New operating systems supported include Fedora 24, RHEL 7.2, Ubuntu 16.10, macOS 10.12 Sierra and Windows Server 2016

Deprecations and Eliminations
- PGI 2017 supports 64-bit operating systems only. Compiling 32-bit applications for execution on either 32-bit or 64-bit operating systems is ot longer supported on any platform
- PGI 2017 for macOS no longer supports CUDA Fortran or OpenaCC running on GPUS, or CUDA-x86 running on CPUs. OpenACC targeting multi-core CPUs is still supported on all platforms including macOS.
- Dropped CUDA 7.0 toolkit support
- Dropped support for macOS 10.7 (Lion) and 10.8 (Mountain Lion)

PGI 2016 Features and Performance

PGI C++ Compiler
Comprehensive support of C++14, requires GCC version 5.0 or greater; full interoperability with GCC 5.1

PGI Fortran Compiler
Further optimizations of scalar/vector single precision pow, exp, log, and atan intrinsic functions for Intel Haswell processors

PGI Accelerator OpenACC Compilers
- New OpenACC 2.5 features:
  •default(present) clause
  •acc_memcpy_device API routine
  •asynchronous versions of data API routines
  •profiling and trace tools interface
- Behavior changes:
  •copy, copyin, copyout and create data clauses now act like present_or_copy, etc.
  •acc_copyin, acc_create, acc_copyout and acc_delete API routines now act like
    acc_present_or_copyin, etc.
  •declare create directive with a Fortran allocatable
  •reductions on orphaned gang loops explicitly disallowed
  •reductions on array and struct elements, and C++ class members explicitly disallowed
  •acc routine without gang/worker/vector/seq now an error
- More stable option -ta=tesla:pinned replaces -ta=tesla:pin
- Kernels launched using -ta=tesla:managed now synchronous by default
- New ACC_BIND environment variable set by default with -ta=multicore

PGI Tools
- All new PGPROF can profile CPU code or CPU+GPU code; both graphical and
  command-line modes
- PGDBG debugger supports disassembly of AVX3 instructions
- Enhanced debugger display of Fortran character types and named commons
- Improved support for debugging OS X shared objects
- Significantly reduced load time when debugging large applications

- New pre-built BLAS and LAPACK library based on OpenBLAS included with all PGI products
- Linux products now ship with Open MPI 1.10.1 instead of MPICH
- Pre-built versions of NetCDF and Parallel NetCDF 1.6.1 available online
- Pre-built versions of the Earth System Modeling Framework (ESMF) 6.3.0rp1 available
  online; one per PGI-built MPI distribution

Other Features, Changes and Additions
- New license key format allows combining multiple floating licenses of varying types
  and seat counts into a single license file
- License key format change requires users to retrieve new license keys
- New operating systems supported include CentOS 7, Fedora23, RHEL 7.2, SLES 12, Ubuntu 15.10 and OS X El Capitan

Deprecations and Eliminations
- Deprecated 32-bit support; expected to be dropped in PGI 2017
- Deprecated PGI Accelerator on OS X support; expected to be dropped in PGI 2017
- Dropped 32-bit PGI Accelerator support on all platforms; 32-bit support retained for
  the -ta=host and -ta=multicore sub-options
- Dropped legacy PGI C++ compiler compatible with the STLPort
- Dropped CUDA 6.5 toolkit support
- Dropped bundled AMD Core Math Library (ACML)
- Dropped RHEL 4 and SLES 10 support


The Portland Group

   PGI 產品介紹
   PGI Accelerator
   PGI 2018 Release Notes