Product lifecycle / long-term support

[John560]

Given that Intel Fortran Compiler used to be a paid product in the past, our developer team is concerned that at some point in the future Intel will decide to discontinue updates to the product altogether.  If this were to happen, we could continue using the final release versions for a period of time, but eventually we will be unable to continue using the development toolchain.

What we've seen is Microsoft releasing new versions of visual studio frequently (every 2 to 4 years). With each release, older versions are no longer supported by Microsoft.  Eventually, the final version of Fortran compiler would only be able to work with an expired version of visual studio. The expired VS has a high probability of being unsafe to use due to unpatched exploits, and potentially future versions of windows may be incompatible with the expired visual studio. There is also a chance of future windows versions being incompatible with the older fortran runtimes.

What we are looking for, is some assurance on Intel's long-term commitment to providing Intel Fortran Compiler.

With these fears of depreciation, my team has decided to rewrite our software in C# and discarding the current version which is written in ~650,000 lines of Fortran77 code. Rewriting will take several years given the volume and complexity and probably will have inferior performance compared to Fortran.

If Intel would comment on these concerns and explain reasons why intel will continue to use their resources to maintain the compiler over the coming decades, that would help me make a case to stick with Fortran.

[caplanr]

Hi,

Just chiming in as a fellow Fortran developer:

I do not see any need to rewrite the code into C# as the compilers one would use for C# also support Fortran.

GCC-gfortran, LLVM-Flang, nvfortran, and eventually LFortran are all free Fortran compilers that should work for you even if the Intel compiler were to disappear (which I highly doubt).

A lot of vendors are moving to using LLVM under the hood too, making updates easier to do on their front ends.

That said, it may be a good idea to at least update the Fortran77 to Fortran 2023 and take advantage of the many improvements to the language.

Check out this recent talk on some of these features here:

https://youtu.be/mwmYiKE1unE?si=yyJNF1wLfIyvGh_z

You can even offload to GPUs with Fortran - check out: https://tinyurl.com/fortrangpu

 - Ron

[Steve_Lionel]

Not speaking for Intel: Manufacturers of CPUs recognize that providing their own compilers makes their processors look better. I once heard it said that Intel compilers made a "one bin level" difference in performance, that was worth a million dollars or more. No hardware vendor treats compilers as a profit-making product - this was true back in my DEC days as well, though we did charge for compilers then, even though software groups were not treated as profit-and-loss centers. 

Microsoft is not the same - they are largely a software company. Note that AMD, NVidia and ARM are all building their own compilers (the industry has largely coalesced around LLVM as a base for compilers.) 

I do understand the concern regarding the Windows product being dependent on Microsoft Visual Studio, even for command-line building (linker and libraries). MS has changed their strategy regarding Visual Studio several times over my career, and they might change it again.  What I don't understand is your choosing to move to C#, which is even more closely tied to Microsoft. This seems to me to be making your position even worse.

[Mentzer__Stuart]

Fortran certainly has some limitations but converting to another language just to assure you can develop with Visual Studio seems extreme. GFortran, Flang, and other Fortran compilers are unlikely to be abandonded any time soon and allow for efficient development without (the benefits and friction of) VS. If your Fortran has a lot of Intel/oneAPI dependencies then, of course, your risk exposure is higher. If so, the community here may be able to help with suggestions on how to port your code to other Fortran compilers.

If there are other reasons your team needs to move away from Fortran I suggest that C++ is a better match to Fortran than C#. Automated Fortran (with Intel extensions) to C++ conversion can be done with no performance loss. But first it makes sense to explore if you can address your Fortran risk exposures while staying with Fortran.