FLW6I25DCL Supplementary Information

NAG Fortran Library, Mark 25

FLW6I25DCL - License Managed

Microsoft Windows XP/Vista/7/8 x64, 64-bit, Intel Fortran, 32-bit integers

Supplementary Information

1. Introduction
2. Applicability
3. Test Platforms
4. Calling an Example Program from NAG Fortran Builder
5. Floating-point Exception Behaviour
6. Limitations with Visual Studio 2015

1. Introduction

This page is for supplementary information relating to the implementation described in Section 2.1 of the Installer's Note. It may be augmented over the life-time of the product if further pertinent information becomes available.

2. Applicability

This implementation is a compiled, tested, ready-to-use version of the NAG Fortran Library that is considered suitable for operation on the computer systems detailed below:

hardware:          Intel x64 systems and compatible
operating system:  Microsoft Windows 64-bit editions (XP / Vista / 7 / 8 and compatible)
Fortran compiler:  Intel 64-bit Fortran Compiler Version 15.0 and compatible

3. Test Platforms

Note: This list is not exhaustive; other combinations of hardware, operating system and compiler may also work.

This implementation has also been successfully tested using the following:

hardware:         Intel Core i7-3770 3.4 GHz
operating system: Microsoft Windows 7 x64 Edition
Fortran compiler: Intel(R) Fortran Compiler Version Version 15.0.1.148 Build 20141023

hardware:         AMD Opteron 2.6 GHz
operating system: Microsoft Windows Server 2003 Enterprise x64 Edition
Fortran compiler: Intel(R) Fortran Compiler
                  Version 11.1    Build 20100414

hardware:         AMD Opteron 2.6 GHz
operating system: Microsoft Windows Server 2003 Enterprise x64 Edition
Fortran compiler: Intel(R) Fortran Compiler
                  Version 12.1    Build 20110811

hardware:         Intel Core i7-3770 3.4 GHz
operating system: Microsoft Windows 7 x64 Edition
Fortran compiler: Intel(R) Fortran Compiler
                  Version 13.1    Build 20130313

hardware:         Intel Core i7-3770 3.4 GHz
operating system: Microsoft Windows 7 x64 Edition
Fortran compiler: Intel(R) Fortran Compiler
                  Version 14.0    Build 20131008

hardware:         Intel Core i7-2600 3.4 GHz
operating system: Microsoft Windows Server 2012 R2 Standard
Fortran compiler: Intel(R) Fortran Compiler
                  Version 15.0.2.179    Build 20150121

4. Calling an Example Program from NAG Fortran Builder

NAG Fortran Builder has in-built knowledge of some versions of the NAG Fortran Library. However, as new versions of the NAG library are produced, Fortran Builder will not know about them until it too has a new release. Nevertheless, it is perfectly possible to call a newer version of the NAG Library by following the instructions here.

Note: we assume that when you installed the NAG Library, you allowed the installer to modify your Windows environment variables - this enables the NAG library to be found at program run time.

For this example we will use one of the Mark 25 example programs, d01ahfe.f90, which comes with your installation of FLW6I25DCL.

Start Fortran Builder, and click on File, then New, and Project.

Choose project type Console Application (note that you should not choose NAG Library Application - that project type will try to use older versions of the NAG Fortran Library rather than the one we want to use)
On the Create Project page, select the check box marked Initial Hello World source
Leave the other settings as they are, and click the OK button

Fortran Builder will create a file named main.f90 for you, containing a small program which prints "Hello World". Delete the contents of the file (but not the file itself). Locate one of the NAG Mark 25 example programs - by default they are installed under c:\Program Files\NAG\FL25\flw6i25dcl\examples\source. For this example, we choose a simple one, d01ahfe.f90. You can use the Fortran Builder File/Open buttons to find this file, or you can locate it yourself using any editor. However you do it, copy and paste the contents of d01ahfe.f90 into your file main.f90.

From the Fortran Builder menu bar, click on Project and then Project Settings.

Then:

On the Basic Settings tab, make sure that 64-bit is selected.
On the Directories tab:
- Click Include
  - Click the Add button
    - Navigate to the NAG Mark 25 library installation directory which contains interface blocks for the nagfor compiler; by default this will be
      c:\Program Files\NAG\FL25\flw6i25dcl\nag_interface_blocks_nagfor
On the Link tab:
- In the Link Libraries section
  - Click the Add button
  - Navigate to the DLL library FLW6I25DC_nag.dll which by default will be c:\Program Files\NAG\FL25\flw6i25dcl\bin\FLW6I25DC_nag.dll
Click the OK button to exit from Project Settings
Click Project then Build - the program should compile and link.

This example program needs to get some input from the user when it runs. You can get this input by redirecting from the data file that comes with d01ahfe.f90.

By default, the date file is located in C:\Program Files\NAG\FL25\flw6i25dcl\examples\data\d01ahfe.d, and so in Fortran Builder, click Run, then Command Line ... and in the text box type

  < "C:\Program Files\NAG\FL25\flw6i25dcl\examples\data\d01ahfe.d"

(note that the quotation marks are necessary if the folder name contains a space as here).

Again click Run, then Execute - the program should run and produce output like this:

  D01AHF Example Program Results

  Integral =  3.14159

  Estimated relative error =   0.58E-08

  Number of function evaluations =    15

5. Floating-point Exception Behaviour

Some of the routines in this implementation of the NAG library can set the IEEE arithmetic "invalid operation" flag during their internal calculations. This is as a consequence of certain optimizations performed by the compiler used to build the library. It is only known to affect a handful of routines, and it does not affect the accuracy of results.

However, this may be a concern if your computing environment raises an exception as a result of the invalid operation. By default, the Intel Fortran compiler does not raise an exception, and so the invalid operation is not likely to be noticed. With the NAG Fortran compiler, however, the default behaviour is to raise an exception when an invalid operation is encountered.

Therefore, if using the DLLs from this product with the NAG compiler, you may find it useful to ensure that the compiler's floating-point exception handling is set to allow full IEEE 754 floating-point exception behaviour (non-stop arithmetic). This can be done by using the option -ieee=full rather than the default option -ieee=stop.

Routines known to be affected by this issue include:

  F01FQF  F01KEF  F01KFF  F01KJF  F01KFF

6. Limitations with Visual Studio 2015

Due to Microsoft refactoring the run-time libraries in Visual Studio 2015, the NAG static libraries in this product will only work with older versions of Visual Studio. Unfortunately, the legacy_stdio_definitions.lib file provided by Microsoft does not supply all the missing symbols required.

A possible workaround would be to link to the NAG dynamic link libraries (DLLs) provided in this product instead of the static libraries, as these are unaffected by the change. See Section 3.1 of the Users' Note for information on linking with the DLLs.

NAG has produced new implementations to work with Visual Studio 2015; the equivalent VS2015-compatible Library to this one is FLW6I25DEL.