took 6 time more to compile in Intel fortran 7.2 than Visual Fortran 6.0

[hiskiyas]

I have a multi language environment Visua C++ 6.0, Object Ada 7.2 and Intel Fortran Compiler 7.1.
I have Microsoft Visual Studio 6.0 environment

All the code compiling and linking fine before. The problem started when I installed Intel Fortran Compiler 7.1, I used to have Visual Fortran 6.0.
I was getting LINK: fatal error LINK1104: can not open file libIEPCF90.lib, but thanks for your help, I put the library path in tools, option, directory tab and I do not have that error.

Now I am getting the following error: libIEPCF90.lib:error LNK2005: _qw32used already defined in dfconsol.lib.
I will try to compile all and see if this problem go away

The other problem I have is It is taking to long to compile so far I compile three times it took 3.5 hrs average. But with Visual Fortran 6.0. it only took 30 minutes. What could be the problem?

[Steven_L_Intel1]

It's hard to say what might cause the longer compilation time. Perhaps your code exercises an algorithm in the compiler that is not time-efficient. Perhaps the compilation requires more virtual memory and your PC starts swapping because you need more RAM. Only with an example can this be determined. You can submit a sample source with compile instructions to Intel Support.

Steve

[TimP]

If your build involves much vectorization, it could easily take longer with IFL 7.1 than with CVF, even if you used the CVF /optimize:5 switch. As Steve mentioned, the amount of RAM could be a factor. A possible work-around is to reduce optimization to /O1 for the parts of your code which aren't performance sensitive.

[hiskiyas]

Thanks, for your help. I have two different machine the same code running on the faster machine and larger RAM the only different is Intel Fortran compiler 7.2 on the faster machine and Visual Fortran 6.0 on the slower machine. But the faster machine takes 6 times longer to compile with Intel Fortran 7.2. Very minimal code change I did to be compatible with Intel Fortran 7.2 compiler.

[larkov]

Please tell me one thing.

Compilation of a C++ project in MSVCPP takes about 10 seconds (Pentium 4, 1000 Mb RAM).

Compilation of a Fortran project (approx same summary size) in CVF6.5 takes about 60 seconds.

Both projects are Windows Applications (multiple "use dfwina" are used if CVF).

What is the reason for this difference in time required for compilation between VCPP and CVF?

It was tested on several machines, compile time ratio is the same.

[Steven_L_Intel1]

I'd have to see the two programs and also know what compilation options are used. Size of the code doesn't matter - it's what the code does. For example, using nested modules in Fortran can increase compilation time. Also, Fortran has language features such as array operations that can mean increased compilation time.

You may want to try CVF 6.6B - we've made a lot of improvements in the last three years.

Steve

[gfthomas8]

Here's a small project to see whether the same code compiles and runs at disparate times with MSVC++ and CVF 6.6b or any other choice of compilers. Go here

http://www.netlib.org

and download the FORTRAN 77 version of Hindmarsh's LSODA code, a classic in any language. Go here

http://server.ccl.net/cca/software/SOURCES/C/kinetics1/lsoda.c.shtml

to get it in C. Hindmarsh has issued a C version but I forget where it's located.

It comes with samples so pick one that exercises the code for a very stiff problem. Compile, link, and run the identical sample under Fortran and C on a level playing field. Report back on your findings.

HTH,
Gerry T.

[larkov]

Steve, Thomas, Thanks for your answers!

Thomas,
Sorry, but I have not enough time now to perform a test You suggested.

Steve,
Compile options are defaults both in CVF6.5A and MSVC++6.0.

I will take the things You said into consideration in future.