It is quite possible, but it may also be a consequence of the Cygwin environment. It is very hard to predict performance accurately, especially if you do not know anything about the program ;). Therefore: try it out, I'd say. My general experience is that Intel Fortran produces faster executables on Windows, but that is not a guarantee.
Thank you very much for a quick reply! Is there a possibility to get a trial version of intel fortran, please? I do not know much about Intel, so I apologize for the stupid question. I do not have much money on my project and to buy the intel and find out that it does not help...
And maybe might it be a solution to rewrite gnu makefile into nmake that should be the equivalent of gnu make under windows? Then I could if i understood well compile under windows. without cygwin. Or maybe both - nmake and intel or does intel have its own makefile environment, please?
Thank you very much.
Probably the code is not optimized for some reason. Should not be that much slower. Once you build in cygwin, it is running "native" as far as I know.
oneAPI should be the answer though, since it plugs into Visual Studio (if that works) and also works on Linux and Mac. I am having trouble integrating Visual Studio and oneAPI, but that may be something to do with the legacy "vfproj" files not loading, so I guess I can go back to building my application from scratch. Visual Studio 2022 17.2 apparently breaks the integration with oneAPI Fortran big time...
I suggest that you download the OneAPI Basic and HPC Toolkits – they are now free unless you want premium support.
Also download the Microsoft Visual Studio 2019 – there is a free version which is all you need.
The Fortran is based on the excellent Visual Fortran developed by Compaq (and DEC previously).
I have compiled NASTRAN and it runs fine on Win 10 64 bit.
EL Zorro Oro
Hi, I have a project that I compiled utilizing gfortran on Ubuntu. But I need to use windows. Hence, I compiled it using cygwin under windows. But I found out that the program is really slow (run for the same parameters 50minutes instead of 5 minutes).
In my extensive experience, I have found Gfortran + Cygwin to be usually within +/- 30 percent of Intel in regard to run time of IFort generated code, depending on options used, nature of the program, etc. Sometimes, the Gfortran geneated code that ran twice as long compared to Ifort. I have never seen a factor of 10, as you report.
If you can make your source code available, I should like to check if the factor is something closer to 1.