forums.silverfrost.com

[Anonymous]

I read some info that the plan for Salford Fortran 95 for .NET is to completely remove any dependency on salflibc.dll (or at least migrate it to managed code). It would obviously be great to use FTN95 to compile pure managed code apps that would run on MONO; particularly given most Fortran developers have a soft spot for Linux.

Could you please provide an update on when/if this is going to happen? (ie over the next 12 months? or are we talking a 2-3 year plan?).

My second question is have anyone tried running a .NET compiled FTN95 app on Linux and trying to run salflibc.dll under WINE?

Regards,

David Taylor

[Andrew]

We had done some work on getting this working some time ago and some code will now run on mono. There have been problems with different versions of mono breaking the support we had put in. You will basically just have to try it out and see how things work out for you.

You need to use the switch /NO_SALFLIBC on both the compiler and linker command line, you still need Windows to compile the assembly of course. Please note however that this is not documented or supported in any way, If you are using only standard Fortran you may find things work great for you, but just to unerline the fact this area is totally unsupported at present. With tests we did here (some time ago now) performance on mono was substantially poorer than with .NET.

[Andrew]

I have had a further look into this and at present the results are not promising. To begin with a special version of ftn95lib.dll (our .NET runtime assembly) is required, we could provide this in theory but there still seems to be problems. Tests that we have run using the switches I mentioned in my previous post and this special version of ftn95lib run fine under mono on Windows. Unfortunately it seems that the implementations are not consistent across platforms. When running the same assemblies under mono on Linux (Fedora Core 2 in this case) I have seen many errors. Some things simply segfault, others partially run, but there seems to be problems in general on Linux. Exactly where the problem lies, be it with FTN95 or mono is unclear.

[henric]

Hi,

I wonder if there are any progress in this area? Are you planning to support mono? I am very interesting in anything that could make FTN95-compiled Fortran code to run under mono in Linux.

We need to make some Fortran code that we previously only have executed in Windows, to be executable under mono in Linux. We have also tried to call a native, unmanaged version of the Fortran code, but there are still problems with parameters sent between Fortran and C#. Therefore I would be very grateful if you can provide me with anything that could make the Fortran code compile and run under Mono in Linux.

For example, you mentioned in an earlier post that a special version of ftn95lib.dll is required - could you provide a version of this one so I can try out? I understand that it would be without any type of support or guarantee. Or can I compile in a manner that makes the program more platform independent.

Best regards
Henric

[JohnHorspool]

Henric,

I have successfully run FTN95 compiled code in Linux running under "wine". If that is of any interest to you or anyone.

Regards,
John

[henric]

Hi John,

Thanks for your response. I would rather not use Wine, but if I run out of options it could be an alternative.

Regards
Henric

[acp693]

Henric,

I too have successfully run code compiled with FTN95 using "Wine". The only small problem I noticed were that some Simpleplot routines are a lot slower than under Windows, otherwise speed doesn't seem to have taken a hit.

Regards

Albert

[henric]

Since this is a program we probably wish to distribute we do not want to run it under Wine.

I would really like to hear from the Silverfrost team. You refer to the forum as your support channel but you haven't answered my question. What I wish to know is if there is any chance of getting a Linux version of the Windows specific DLL to try out. Maybe it works in our case since our code is fairly simple? Please get in touch.

[Martin]

Hi Henric,

I don't think its as easy as just compiling up a new mono-specific version of ftn95lib and sending it to you. Even when we were actively investigating it I don't think we were ever at a stage where non-trivial programs would work on linux. I suspect that more problems will have creeped in since we last looked at it.

There is also an undocumented /no_salflibc option that can be used when compiling with FTN95. I suspect that this doesn't do everything it needs to either. I'll try and have a look at this some time next week and get back to you - but of course this is an unsupported feature and we can't yet guarantee we will ever complete this work.

Martin

[henric]

Hi Martin,

Thanks for you reply. Great if you could look in to this. I am aware that it will be unsupported but I am willing to try it out.

Looking forward to hear from you again
Best regards
Henric

[Martin]

Hi Henric,

I've finally had chance to check out the current status of mono - sorry for the delay.

With a cross platform build of ftn95lib no programs can be linked using dbk_link at present, so it would be pointless to send it out to you. I'll look into this and report back here when progress is made but unfortunately we cannot say when or even if this work will be completed.

Martin

[DrTip]

How complete is this switch if we wanted to compile libraries for use on 64 bit windows machines? ie just letting the .NET framework on a 64 bit machine do the final compilation step.

would we be able to address more memory this way?

a C# program can I think if set up correctly

Carl