OpenVMS 8.4 rx3600 -> OpenVMS x86-64 Port Considerations

[vmskostoff]

Currently working in a OpenVMS 8.4 environment on an rx3600 Integrity.

Trying to wrap my mind around the considerations and concerns for porting to OpenVMS x86-64.

Any white papers to review items of consideration?

Application code written in: FORTRAN, C.

Any thoughts would be greatly appreciated.

Thanks.
Vince

[arne_v]

VMS is VMS.

VMS x86-64 9.2-2 and VMS Itanium 8.4-something are very very compatible.

So my suggestion is:
* check that all the software you need are available on VMS x86-64
* get a license
* create a VM and install VMS on it
* move the code over and try build as is

If you are lucky then it just works with zero changes.

Obviously no guarantee for luck.

Any code that check on HW platform will need to do the right thing for the new platform.

The C++ compiler is rather different (more standard clang than old VMS C++).

Any Itanium assembler need to be rewritten.

An probably a few things that I am not aware of it.

But still I would suggest the Nike way "Just do it".

[sms]

Code: Select all

> VMS is VMS.

   Generally true.  And where it might break down in your case is tough
to guess without very much more information.

> * get a license

   An evaluation license should be available to a real customer.

> [...] "Just do it".

   That'd be my suggestion, too.

[joukj]

It depens on what is in your code. I had not much problem with C. But if I see FORTRAN I smell scientific computing. hope in your code no complex variables are used, since they will give problems; i.e. all complex valued intrinsic functions give wrong answers. In this respect the current FORTRAN compiler does not even comply with the F77 standard.

Jouk

[arne_v]

It depens on what is in your code. I had not much problem with C. But if I see FORTRAN I smell scientific computing. hope in your code no complex variables are used, since they will give problems; i.e. all complex valued intrinsic functions give wrong answers. In this respect the current FORTRAN compiler does not even comply with the F77 standard.

True. But it should be noted that it is not a problem that will require people to change their code - it is a problem that require VSI to release a bug fix. Based on what we have been told then I it requires VSI to build the math library with the native compilers instead of the cross compilers (I actually think that it would make sense for VSI to release such a new build soon and not wait for the rest of VMS to be build by native compilers).

[vmskostoff]

Thanks for all responses and food for thought.


arne_v:

Is the version you noted, VMS x86-64 9.2-2 , stable and able to be used in a production environment, 24x7x365?

The only "software" I would need would be the compilers for FORTRAN and C.

[sms]

Code: Select all

> Is the version you noted, VMS x86-64 9.2-2 , stable and able to be used
> in a production environment, [...]

   As usual, it depends.  If you browse these forums, you can probably
find cases where someone found a way to crash the OS, or where some
compiler version misbehaved, but V8.4 and its friends were not perfect,
either.

   Evaluation licenses exist to allow people to evaluate the stuff in
their own environments.

> [...] 24x7x365?

   Eh?  24 hours/day * 7 days/week * 365 weeks/what?

   I doubt that you'll get an always-works warranty (that's worth
anything) here.  Whether the current versions of VMS, C, and Fortran are
adequate for your needs is beyond my ken.  On my x86_64 systems, I do
mostly freeware software development (in C), which is not a very
comprehensive or strenuous test, but these days they seem to work as
expected.

[arne_v]

Thanks for all responses and food for thought.

arne_v:

Is the version you noted, VMS x86-64 9.2-2 , stable and able to be used in a production environment, 24x7x365?

Is VMS 9.2-2 ready to start porting/testing effort on x86-64?

Yes!

Is VMS 9.2-2 ready for production?

In my experience the quality of VMS 9.2-2 is no worse than many other widely used OS. There are no reasons to consider it generally unfit for production.

So you do what you do on any other OS: you test your application carefully. If you encounter problems in test, then you report them to VSI and VSI fix them. If you do not encounter problems in test, then it is as production ready as it can be.

++++

So start the test tomorrow.

Expect it to be production ready "soon".

"soon" may be when your test is done and it did not find any problems.

"soon" may be time of test finding problems + time for VSI to fix problems + time to retest.

If you go that route then I am 95% sure that at Christmas you will have been running it in production for months.